Studies in Sound Symbolism

GOTHENBURG MONOGRAPHS IN LINGUISTICS 17 

Studies in 

Sound Symbolism 

Åsa Abelin 

Doctoral Dissertation 

publicly defended in Stora Hörsalen, 

Humanisten, Göteborg University, 

on May 14, 1999, at 10.00 

for the degree of Doctor of Philosophy 

Department of Linguistics, Göteborg University, Sweden, 1999

ABSTRACT 

This thesis investigates how the Swedish lexicon is structured with 

respect to sound symbolism, the productivity of phonesthemes and cross 

language similarities in certain areas of sound symbolism. 

The Swedish lexicon has been analyzed with emphasis on the sound 

symbolic properties of initial and final consonant clusters, and to a 

certain extent of vowels. Approximately 1, 000 lexemes were judged to 

be sound symbolic and the outcome of the analysis are tentative 

phonesthemes, i.e. motivated connections between meanings and 

consonant clusters. 

Almost all Swedish initial consonant clusters and many of the final 

consonant clusters can carry sound symbolic meanings. Lexically 

infrequent clusters are utilized to a larger extent than lexically frequent 

clusters. No two consonant clusters have exactly the same semantic 

profile. Phonesthemes have different sound symbolic strength, i.e. some 

are clearly sound symbolic (i.e. a high percentage of the words 

beginning with a certain cluster are sound symbolic), and carry either 

one meaning or several meanings. Other (candidates for) phonesthemes 

are weaker and not so clearly sound symbolic. 

The meanings of most phonesthemes are relatable to the senses: 

hearing, vision or tactile sensation, or they are metaphorically or 

metonymically connected to the senses. The most common semantic 

features occurring are often relatable to synaesthesia. 

The productivity of phonesthemes was tested in experiments of 

production and understanding. The experiments show, for 

interpretation, that no constructed word is interpreted as expected by 

all subjects, but that all of the constructed words are interpreted 

correctly by some subjects. The most common semantic features found 

in the lexical analysis are often also the most successfully interpreted by 

subjects. For production, the experiments indicate that subjects tend to 

encode the semantic features in initial clusters rather than in final 

clusters. Final consonant clusters seem to be of less importance than the 

initial clusters in new sound symbolic words in Swedish. 

For the contrastive studies, the general results are that there are both 

similarities and dissimilarities between the expressions in the different 

languages. The variation is greater for some semantic fields than for 

others. 

KEY WORDS: sound symbolism, lexical structure, synaesthesia, 

productivity, universals

Acknowledgements 

The first idea to write this thesis emerged when I was working in the 

Lexical Database and Svensk Ordbok project at Språkdata in the 1980's, 

where I alphabetically plowed through large parts of the Swedish 

vocabulary. 

My greatest thanks goes to my supervisor Jens Allwood, for always 

supporting me and for being so consistent in always misunderstanding 

what is not perfectly clear. I am also grateful to Elisabeth Ahlsén who has 

given me many valuable comments and especially encouraged my 

experimenting. 

I also want to thank all my colleagues and former colleagues at the 

linguistics department, especially Sally Boyd, Beatrice Dorriots, Johan 

Hagman, Jerker Järborg, Per Lindblad, Lars Malmsten, Kerstin Nelfelt, 

Shirley Nicholson, Joakim Nivre, Sören Sjöström, Sven Strömqvist, 

Nicole Takolander, Hans Vappula, Åsa Wengelin, Ulla Veres, and all 

others who have helped me in various ways over the years. A special 

thanks goes to Catherine Paterson who did the cover design. 

I also want to express my gratitude to my informants from many parts of 

the world. 

The persons who I want to thank especially are my children Tove and 

Ellen who are for sure very tired of the word "avhandling". They have 

shown great patience with my intermittent absentmindedness, but they 

have also shown great interest in the subject of this thesis and given me 

interesting comments from their point of view. Finally I want to thank my 

family, my mother and father, my sisters Ulla and Susanne and all my 

other friends. 

Göteborg in April, 2000 

Åsa Abelin

Contents 

Chapter 0 Introduction 1 

Chapter 1 Background 3 

Chapter 2 Theoretical framework 49 

Chapter 3 Method 71 

Chapter 4 Analysis of initial consonant clusters 77 

Chapter 5 Analysis of final consonant clusters, vowels and 

combinations. 143 

Chapter 6 Some contrastive studies in sound symbolism 189 

Chapter 7 Experiments with words constructed from phonesthemes 219 

Chapter 8 Summary and discussion 245 

Bibliography 273 

Appendix 1: Sound symbolic roots of initial clusters 

Appendix 2: Interjections of cross language study 

Appendix 3: Test sheets of chapter 7

Studies in Sound symbolism 

0 Introduction 1 

1 Background 3 

1.1 Purpose of the chapter 3 

1.2 Terminology 3 

1.2.1 Onomatopoeia 3 

1.2.2 Sound symbolism 3 

1.2.3 Phonestheme 4 

1.2.4 Ideophone 4 

1.2.5 Morpheme 5 

1.2.6 Conclusion 8 

1.3 Is sound symbolism the rule or the exception in 

language? 9 

1.3.1 Sound symbolism is an exception 9 

1.3.2 Symbolism is fundamental to language 9 

1.3.3 Sound symbolism is both inside and outside of 

language 11 

1.3.4 Evaluation of discussion of sound symbolism 12 

1.4 Is sound symbolism productive or not? 12 

1.4.1 Sound symbolism is not productive 12 

1.4.2 Sound symbolism is productive 13 

1.4.3 Greater or lesser degree of productivity 13 

1.4.4 Evaluation of the discussion of productivity 13 

1.5 The question of etymology 14 

1.6 The phylogenesis of language 18 

1.7 Universality versus language specificity 20 

1.7.1 Sound symbolism is universal 20 

1.7.2 Symbolism is not universal 21 

1.7.3 Evaluation of the discussion of universality 

versus language specificity in sound symbolism 22 

1.8 Context 23 

1.9 A framework for models of sound symbolism 26 

1.10 Expression and content of sound symbolism 28 

1.10.1 Expression 29

1.10.2 Content 29 

1.10.3 Expression and content in different analyses 29 

1.10.4 Results - data from different authors 31 

1.10.5 Experimental results 37 

1.11 Possible explanations of sound symbolism 41 

1.11.1 Miscellaneous explanations and proprioception 41 

1.11.2 Synaesthesia 42 

1.11.3 Other neurological and biological explanations 44 

1.11.4 Non- biological explanations 46 

2 Theoretical framework 49 

2.1 General considerations 49 

2.2 Static-dynamic, conventionality and 

arbitrariness 49 

2.3 Semantic analysis 51 

2.3.1 Conceptions of meaning 51 

2.3.2 Semantic features and semantic fields 52 

2.4 Basic relations between expression and content 54 

2.5 The nature of phonesthemes 57 

2.6 Considerations for a model 59 

2.6.1 Relations between the categories 63 

2.6.2 An explanatory model for sound symbolism 66 

3 Method 71 

3.1 Stage 1: Collection of lexical material 71 

3.2 Stage 2: Cross-language comparisons 74 

3.2.1 Cross-language thesaurus studies 74 

3.2.2 Cross-language informant studies 74 

3.3 Stage 3: Experiments 76 

3.3.1 Experiments with neologisms 76 

3.4 Further method 76 

4 Analysis of the initial consonant clusters 77 

4.1 Data analysis, a short overview 77 

4.2 Results 80 

4.2.1 More and less sound symbolic clusters 80 

4.2.2 Proportions of motivated root morphemes. Summary 86 

4.2.3 Types of meaning 88

4.3 Frequent semantic features 93 

4.3.1 Pejorative 93 

4.3.2 Sound 100 

4.3.3 Long thin form 105 

4.3.4 Quick or strong movement 107 

4.3.5 Wetness 111 

4.3.6 The most sound symbolic clusters 115 

4.4 Frequent clusters 116 

4.4.1 The cluster sl- 118 

4.4.2 The cluster sn- 119 

4.4.3 The cluster kn- 121 

4.4.4 The cluster kr- 124 

4.4.5 The cluster fn- 126 

4.4.6 The cluster kn- 127 

4.4.7 The cluster gn- 127 

4.4.8 The cluster spr- 129 

4.4.9 The cluster pj- 130 

4.5 Typical or unique meanings 131 

4.6 Patterns of semantic features 134 

4.7 Discussion and conclusions 140 

5 Analysis of final consonant clusters, vowels and 

combinations 143 

5.1 Final clusters 144 

5.2 Summary of the analysis of semantic features for 

final clusters 150 

5.3 Properties of consonant clusters of Nusvensk Frekvensordbok 152 

5.3.1 Summary of the analysis of properties of different 

final clusters 164 

5.4 Vowels 164 

5.4.1 Vowel pairs and triplets 164 

5.4.2 Vowels in light/gaze-words 166 

5.4.3 The vowel [P] 167 

5.4.4 Summary of vowels 167 

5.5 Comparisons of final clusters of roots from different sources 

167 

5.5.1 NFO4 and Sigurd (1965) 167

5.5.2 The most frequent final clusters in NFO4 171 

5.5.3 Discussion of final clusters in Svensk 

Baklängesordbok and of Nusvensk Frekvensordbok 172 

5.6 Combinations of initial and final consonant clusters 174 

5.6.1 Initial cluster + no final cluster 175 

5.6.2 No initial cluster + final cluster 179 

5.6.3 Initial cluster + final cluster 182 

5.6.4 Summary of combinations 185 

5.7 Summary and discussion of initial and final clusters, 

and vowels 186 

6 Some contrastive studies in sound symbolism 189 

6.1 Introduction 189 

6.2 The Thesaurus study 189 

6.2.1 Method 189 

6.2.2 Results 190 

6.2.2.1Words for 'stupidity' in English 190 

6.2.2.2 Words for 'stupidity' in Swedish 190 

6.2.2.3 Words for 'surface structure' in English 191 

6.2.2.4 Words for 'surface structure' in Swedish 192 

6.2.3 Conclusions of the Thesaurus study 193 

6.3 Some interjections in different languages 194 

6.3.1 Swedish expressive interjections 195 

6.3.2 Discussion of Swedish expressive interjections, 

commands and greetings 196 

6.3.3 Phonological and phonetic similarities and 

dissimilarities between interjections of different languages 197 

6.4 Imitations of animal calls 202 

6.4.1 Expressions for animal calls in Swedish and other 

languages 202 

6.4.2 A test of expressions for animal calls of different 

languages 205 

6.4.3 Results from a test of expressions for animal calls 

in different languages 206 

6.4.4 Discussion of the test on identifying animal calls 208 

6.5 Conclusions of studies of expressive interjections 

and expressions of animal calls 208

6.6 Test of cross language interpretation of Swedish 

onomatopoeic and other sound symbolic words 210 

6.6.1 Method 210 

6.6.2 Results of interpretation of cross language Swedish 

onomatopoeic and other sound symbolic words 211 

6.6.3 Conclusions from the test on cross language 

interpretation of Swedish onomatopoeic and other 

sound symbolic words 215 

6. 7 General conclusions and discussion of the cross 

language studies 217 

7 Experiments with words constructed from 

phonesthemes 219 

7.1 Production and understanding 215 

7.1.1 Forced choice for production - from meaning to 

phonological (graphic) form 223 

7.1.2 Forced choice for understanding - from phonological 

form to meaning 226 

7.1.3 Free production test from constructed words to 

meanings 229 

7.1.4 Free production from meaning to constructed word 232 

7.1.5 Matching test of nonsense words and meanings 238 

7.1.6 Summary of results of tests on interpreting meanings 

and sounds 239 

8 Summary and discussion 245 

8.1 The research questions were as follows: 245 

8.1.1 Question 1 246 

8.1.2 Question 2 252 

8.1.3 Question 3 255 

8.1.4 Question 4 257 

8.2 Comparison of the studies 258 

8.3 Possible explanations of onomatopoeia and sound symbolism 263 

8.3.1 Pejoratives 265 

8.3.2 Summary 266 

8.4 Predictions for sound symbolism in Swedish 268 

8.5 Main objectives and further research 271

Introduction 

The purpose of this thesis is to study some aspects of the wide field of 

onomatopoeia and other sound symbolism with special reference to 

Swedish. A large part of the study is devoted to a description, semantic and 

phonological, of onomatopoeia and other sound symbolism 1 of initial and 

final consonant clusters, and also of vowels. 

One main interest is the issue of productivity, which is studied mainly with 

the aid of experiments. 

Furthermore, the issue of universals is addressed through partial 

comparisons with other languages and also through tests of perception of 

Swedish onomatopoeia and sound symbolism. 

The purpose is finally to construct a model for sound symbolism as a 

central part of language, and to construct an explanatory model for the 

semantic aspect of sound symbolism in Swedish. 

The research questions of the thesis are the following: 

1. What are the properties of sound symbolic sequences in Swedish? More 

specifically the questions are: 

Which initial and final consonant clusters are used in sound 

symbolism? 

Which meanings are used in sound symbolism? 

How do these combine in phonesthemes? 

What are the sound symbolic characteristics of some vowels? 

How do initial and final clusters and vowels combine in words? 

2. Are phonesthemes productive in Swedish? And, if so, are some 

phonesthemes more productive than others? Are neologisms created or 

interpreted in accordance with the semantic model of chapter 2 and the 

analysis of chapters 4 and 5? 

3. Are there similarities or dissimilarities, in some specific aspects of sound 

symbolism, between different languages? 

4. Do non-Swedish speakers interpret Swedish phonesthemes in accordance 

with the semantic model of chapter 2 and the analysis of chapters 4 and 5? 

1 cf. p. 4 

1

1 Background 

1.1 Purpose of the chapter 

The area of onomatopoeia and other sound symbolism has not been 

central in linguistics. Nevertheless, most linguists have had something to 

say on the subject, in many cases using different terminologies. In this 

chapter there will be a thematic overview of some of the most interesting 

contributions. The themes that are especially important are the questions 

of integration into language (grammar and lexicon), of productivity, of 

universality and of explanation. The chapter will end with a summary of 

forms and meanings of onomatopoeic and sound symbolic expressions. 

1.2 Terminology 

There are two problems with the terminology in this area. First, there are 

a lot of terms. Second, the terms are not always used in the same way, and 

they are seldom defined. The greatest confusion, if one doesn't know the 

author's purpose, is perhaps that the term symbolic is often used to mean 

sound symbolic (i. e. words or other expressions having a sound structure 

that is not independent of their meaning), and not as a contrast to 

indexical and iconic (cf. Peirce (1955) and Allwood and Andersson 

(1976)). In this thesis, the term symbolic is therefore not used as a 

synonym to sound symbolic. If the term is used, it is used in the 

traditional way, for conventional non-motivated signs. 

1.2.1 Onomatopoeia 

Nordberg (1986) gives a definition. ‘‘Onomatopoeia in a restricted sense 

refers to imitation of natural sounds, e.g. of animals’’. I do not believe it 

is necessary to restrict the term onomatopoeia to natural sounds even 

though, of course, this is a special kind of onomatopoeia. I will use the 

term onomatopoeia for all kinds of sound imitation. 

1.2.2 Sound symbolism 

Nordberg (1986) writes ‘‘Sound symbol or phonestheme ... is the 

synesthetic combination of a certain sound or sound sequence with a 

particular notion or a particular connotative content.’’ Jakobson and 

Waugh (1979) define sound symbolism as ‘‘an inmost, natural association 

between sound and meaning’’. Malkiel (1994) uses the term 

phonosymbolism to mean the same, he writes, as sound symbolism. 

3

I will use the term sound symbolism as a general term for an iconic or 

indexical relationship between sound and meaning, and also between 

sound and sound (which is onomatopoeia). Onomatopoeia is then a special 

case of sound symbolism. Onomatopoeia and other sound symbolism will 

also be termed motivated expressions. 

1.2.3 Phonestheme 

Phonesthemes are sometimes called word affinities or verbal affinities and 

are described as associated with ‘‘a marginal set of vocables which are 

semantically fluid, more expressive than cognitive’’ (Jakobson and 

Waugh, 1979) or as ‘‘the grouping of similar meanings about similar 

sound’’ (Bolinger, 1965). (The relation between phonesthemes and 

morphemes is discussed in 1.2.5). Householder's (1946) definition of a 

phonestheme is: ‘‘a phoneme or cluster of phonemes shared by a group of 

words which also have in common some element of meaning or function, 

though the words may be etymologically unrelated’’ The definition above 

could be improved by changing ‘‘words’’ into ‘‘morphemes’’ (for a 

discussion of the morpheme, see 1.2.5). 

Another term is psychomorphs (Markell and Hemp, 1960), which is used 

in approximately the same sense as phonesthemes. Bolinger (1950) also 

uses the term submorphemic differentials or submorphs which are 

described as ‘collocations of phonemes common to a set of words and 

suggestive of a stronger or vaguer semantic interconnection. He uses the 

term affective morphemes as synonymous to phonesthemes. Nordberg 

(1986) coins the term sound words (onomatopoeic elements) as the joint 

designation for ideophones, sound symbols, phonesthemes and 

onomatopoeia since he claims that the boundaries are fuzzy. Nordberg 

also writes that onomatopoeic elements are not such vaguely imitative 

words with normal phonotactics as e.g. Swedish susa 'sigh', klucka 

'cluck', fladdra 'flutter', etc. but purely sound-illustrative sequences, 

reminiscent of the sound balloons of comic strips. 

1.2.4. Ideophone 

Childs (1994) discusses the problem of defining ideophones and claims 

that often one or more of the following criteria are met: ideophones often 

have unusual phonological characteristics, and they often display very 

little morphology. Syntactically they are often set apart from the rest of 

4

an utterance. In some languages they constitute a separate syntactic 

category, in others not. Semantically they often underscore the meaning 

of a verb, and often, but not always, they are sound symbolic. In many 

cases ideophones show a close connection to gestures. 

This kind of phenomenon seems to correspond best to the ‘‘sound words’’ 

of adolescent language, described by Nordberg (1986), (e.g. krch, ppff, 

dåing, ‘‘har hon håret så här, så här: tsscchht ’’ (‘‘she wears her hair like 

this, like this: tsscchh’’, etc.)) but not to phonesthemes. 

The term ideophone will not be used in my own analysis of onomatopoeia 

and sound symbolism in Swedish, partly because different authors use the 

term in a non-uniform way. Even if I do not treat ideophones in my own 

analysis I will use the term as it is described above, in discussing some of 

the authors below. 

1.2.5 Morpheme 

What exactly is the relation between phonesthemes and morphemes? Are 

phonesthemes really morphemes or are they something different? The 

most common definition of a morpheme is ‘‘a minimal meaning carrying 

unit’’, but that definition in this context seems to apply to phonesthemes 

rather than to what traditionally has been labeled morphemes: The lexeme 

bjäfs (gewgaws) can be analyzed as one morpheme. But it can, at the same 

time, be analyzed as consisting of the minimal meaning units bj 1- 

'pejorative', -E- 'pejorative, and -fs 'pejorative'. It seems that these are 

really the minimal meaning carrying units. 

However, these minimal meaning carrying units do not build up the whole 

meaning of the word bjäfs. The meaning of a sound symbolic word is 

more than the meaning of its parts, but this also goes for many lexicalized 

compounds, e.g. blackbird means more than 'black bird'. In the case of 

other sound symbolic words, however, it seems that the meaning of a 

word is almost always more than the meaning of its parts. As a contrast, 

the meaning of onomatopoeic words, like plask, is often not more than the 

meaning of its parts: pl- 'wetness' and -sk 'wetness'. 

1Phonemes will be marked with bold type and are not put between slashes. The symbol P 

will be used for the short rounded close-mid vowel, e.g. the first vowel in muttra 

(mutter). 

5

Also, in a productive perspective, neologisms are sometimes created or 

understood out of (one or more) minimal meaning units (phonesthemes) 

and these can thus be seen as building blocks for word meaning. 

Distributionally, phonesthemes are not whole words (lexemes) but parts 

of words and could thus not qualify as free morphemes. They could 

perhaps be described as bound morphemes, i. e. as affixes. However, 

there is a fact that contradicts this interpretation, on the expression side: 

bound morphemes in Swedish can include a vowel or consist of a single 

consonant or of a consonant cluster. The domain of the phonestheme is 

often a consonant cluster. This shows that the expression side of 

phonesthemes is only partially different from other bound morphemes, in 

Swedish. But there are also other more problematic facts on the 

expression side in traditional morphology (for a discussion see e.g. 

Källström (1988), e.g. suppletion (be, am are, is, was, were) and 

inflecting and fusioned declination (e.g. finna, fann, funnen), so it is not 

clear that this would distinguish phonesthemes from morphemes. 

Another problem can be illustrated by the word flämta (pant), where flis 

a phonestheme imitating movement. The rest of the word -ämta would 

have to be called a restmorph. Restmorphs comprise a problematic, but 

not unusual, category in morphology. A way to avoid too many 

restmorphs is to say that in sound symbolic words there are meaning units 

in independent dimensions to morphemes, e. g. in the morphemes bjäfs 

and flämta; bjäfs and flämta are meaning units on one level which at the 

same time contain bj- (in bjäfs ) and fl- (in flämta), with pejorative 

meanings on another level. Examples such as these show that it is a 

problem to determine how far the morphological analysis should go (see 

Källström, 1988). 

Phonesthemes can be placed in a hierarchy between phonemes and 

morphemes, where morphemes and phonesthemes both are meaning 

bearing units while phonemes are not (figure 1.1). 

6

morphemes 

phonesthemes 

phonemes 

Figure 1.1. The hierarchy between phonemes and morphemes. 

Phonesthemes are built up from phonemes, morphemes can (partly) be 

built up from phonesthemes, but phonesthemes are never built up of 

morphemes. 

The morpheme is a connection between expression and meaning. When 

the expression varies, e.g. dog-s [z] - cat-s [s] they are called allomorphs. 

A parallel distinction could be made between e.g. the pejoratives pj-, bj-, 

fj- which can be called ‘‘allophonests 2’’ to a phonestheme (with pejorative 

meaning), in an item-and-arrangement analysis. (Alternatively, pj-, bjand 

fj- can be analyzed as labial obstruent + j.) These examples show that 

these types of analyses can be done with phonesthemes, as well as with 

regular morphemes. 

A certain consonant cluster can sometimes be the expression of a 

phonestheme, but sometimes not, (e.g. kl- kladdig (sound symbolic), 

klöver (not sound symbolic). The phonestheme appears in interaction with 

the word meaning, when it fits the expression 3. That means that we have 

homonymy/polysemy in phonesthemes, e.g. kl- can mean 'adhesion' as in 

kladdig (sticky), klibbig (sticky), kletig (smeary), klick (dollop), 'wetness' 

as in klafsa (squelch), it can be 'onomatopoeic' as in klang (clang), 

klappra (clatter), klatsch (slap), klicka (click), kling (tinkle), klirra 

(jingle), klämta (toll), klucka (lap) and it can mean a 'certain form' as in 

klimp (lump), kloss (block), klot (ball), klubba (club), klump (lump) and 

it is 'pejorative' as in kludda (daub), klotter (doodle), etc, cf. chapter 4. 

2This term will, however, not be used further. 

3There can also be interaction between different sounds in the word, cf. analysis in 5.5. 

7

We also have a variation between some phonesthemic meaning and none, 

e.g. we have no phonesthemic meaning in klöver (clover). Certain 

consonant clusters are almost always phonesthemic, e.g. pj-, while others 

are less phonesthemic, e.g. tr-. This means that there is a variation in the 

strength of the connection between content and expression in 

phonesthemes, a quantitative dimension of the morphology. 

Consequently, there are arguments for and against whether phonesthemes 

are morphemes or not. The meaning of the concept morpheme is vague, 

but I suggest that phonesthemes belong to morphology. It is one type of 

morpheme, a type which is special on the semiotic and semantic side: the 

relation to what is denoted is often iconic or indexical and what is denoted 

are often sounds, experiences of the senses, emotions, etc. 

Phonesthemes are also special in that they have a low degree of autonomy. 

They are bound morphemes which often can be analytically distinguished 

in a fashion similar to free morphemes. They will be referred to as 

phonesthemes and regarded as a type of morpheme which, because of 

their dependent and motivated nature, perhaps could rather be called 

submorphemes. 

1.2.6 Conclusion 

Above, there has been a survey of different terms and usages. 

Subsequently, the term onomatopoeia will be used to mean all kinds of 

sound imitation, phonestheme will be used to mean bound submorphemic 

(cf. above) strings (e.g. consonant clusters) which have in common a 

certain element of meaning or function. The relation between sound and 

meaning is often iconic or indexical, as well as symbolic. The term sound 

symbolism will be used for the general phenomenon of motivated 

relations between sound and meaning, including onomatopoeia. Thus 

sound symbolism, in a sense, is used almost oppositely to the sense of 

symbolism that was suggested by Peirce, i. e. it focuses on what he called 

icons and indexes but not symbols. Most of the words discussed also 

contain a conventional arbitrary element which means we are dealing with 

what could be called iconic and indexical symbols. (For further 

discussion, see chapter 2.) 

There also seems to be a further terminological issue here, in that sound 

symbolism tends to be reserved for universal phenomena and 

8

phonesthemes for language specific phenomena. I would, however, prefer 

to use sound symbolism as a more general term and use the term universal 

sound symbolism when this is the issue. 

The relationship between the most important terms discussed here can be 

illustrated in the following way (figure 1.2). 

sound symbolism or 

motivated expressions 

onomatopoeia other sound symbolism 

phonesthemes 

free morphemes 

phonesthemes 

9 

free morphemes 

Figure 1.2. The relationship between some of the most important terms 

discussed in 1.2. 

1.3. Is sound symbolism the rule or the exception 

in language? 

1.3.1 Sound symbolism is an exception 

Saussure (1916) who held the view that linguistic signs are arbitrary 

wrote that ‘‘onomatopoeic words are never organic elements of a 

linguistic system’’. Bühler (1933/1969) said that onomatopoeia is a 

reversion, since language has evolved beyond primitive needs and means 

of self-expression. The genesis of language is a measure of its success at 

arbitrary symbolic (i. e. not what I call sound symbolic) representation. 

1.3.2 Sound symbolism is fundamental to language 

Already von der Gabelentz (1891) wrote about the ‘‘sound symbolic 

feeling’’, the experience that sound and meaning are inalienably 

interconnected for the naive members of the speech community. For

example, naive Germans would say ‘‘that Frenchmen are silly when they 

name ein Pferd Schewall’’ (Jakobson and Waugh, p. 182). 

An opposite view was held by Saussure's contemporary Jespersen (1922 

a), who went as far as to claim that ‘‘languages in the course of time grow 

richer and richer in symbolic’’ (that is sound symbolic) ‘‘words’’ and 

‘‘develop towards a greater number of easy and adequate expressions – 

expressions in which sound and sense are united in a marriage-union 

closer than was ever known to our remote ancestors’’. 

To the extent that African so called ideophones are discussed (by e.g. 

Samarin, 1978 4 ) these are usually considered part of language proper. 

One might question if this is because ideophones play a much more central 

role in certain African languages than in European languages, or whether 

there is another, freer tradition in describing non-European and primarily 

spoken languages. Diffloth (1976) writes that in describing languages with 

a structure remote from ones own language, it is often self-evident to 

segment roots into smaller, significant units with their own sound 

symbolic value, thereby touching on the subject of how grammatical 

tradition may obscure interesting phenomena in a language. 

Lakoff and Johnson (1980), showing many examples, claim that in all of 

language, including syntax, there are many cases of non-arbitrary 

relations between form and meaning. This seems to point to a view of 

non-arbitrariness being integrated in language and not in a separate 

category. 

A phenomenon in Swedish which is similar to ideophones (cf. 1.2.4) are 

the ‘‘sound words’’ of adolescent language, described by Nordberg 

(1986), e.g. krch, ppff, dåing. These are usually not considered to belong 

to the core of Swedish, if described at all. However, they show interesting 

similarities to e.g. the ideophones in Gbeya (a dialect of Gbaya, spoken in 

Central Africa), (Samarin, 1978); they show reduplication, lengthening, 

phonotactic ‘‘divergencies’’ and sentence final position. The greatest 

4 

Samarin defines ideophony as ‘‘the foregrounding of phonological elements in word (or 

lexeme) composition in both spontaneous creations and fully institutionalized lexicons, 

usually associated with semantic categories of an attributive nature, commonly affective, 

and sometimes also in true onomatopoeia’’ 

10

difference to Gbeyan seems to be on the semantic side; Gbeyan words 

seem to have more specific meanings, that can be assigned dictionary-type 

definitions, e.g. ndadak ndadak ‘‘(wood that) doesn’t split well in 

chopping’’. 

1.3.3 Sound symbolism is both inside and outside 

of language 

Householder (1946), writes that the vocabulary of English falls into three 

parts with regard to arbitrariness of structure: 1) those clearly and 

completely arbitrary, their meaning unaffected by the sound, 2) those 

made up, in whole or part, of phonesthemes, 3) those belonging primarily 

to the first group but with their meaning colored or altered in varying 

degree by secondary association with phonesthemes. 

In the Oxford Dictionary of English Etymology (1967) the consonant wr - 

is a separate entry meaning ‘‘consonant combination occurring initially in 

many words implying twisting or distortion’’. The dictionary further 

states that the meaning of twisting has many correspondences in other 

Germanic languages5. Wescott (1975) uses the term ‘‘microlanguage’’ for the core of language, 

which is subject to well-known grammatical rules – i.e. conventional 

language. For other domains of speech, e.g. baby talk, exclamations, 

verbal art – ‘‘language that is alienated from conventionally structured 

speech’’, he coins the cover term ‘‘allolanguage’’. He says, ‘‘one of the 

characteristics of allolanguage is a closer relation between sound and sense 

than obtains in microlanguage’’, ‘‘a retention ... of an older and simpler 

manner of self expression alongside one that is more recent and 

complex.’’ Sound symbolism, being one aspect of allolanguage, is thus 

part of language, but on the periphery, according to Wescott. 

Sounds can enter a language by means of sound symbolic words (e.g. 

Hinton, 1986). Also, sound changes often do not affect sound symbolic 

words, so that phonemes that should have disappeared in a language or 

have become restricted to certain environments are still to be found in 

sound symbolic vocabulary (as pointed out by Hinton, Nichols, Ohala 

5Cf. the Swedish clusters kr- and sn-, in chapter 4. 

11

1994). E. g. initial r- is rare in the non sound-symbolic vocabulary of 

Finnish (Austerlitz, 1994). 

There are also suggestions of tendencies to use a more reduced phonemic 

inventory in sound symbolism (e.g. Oswalt 1994). 

1.3.4 Evaluation of discussion of sound symbolism 

The view of the difference between sound symbolic words and other 

words assumed in this thesis is the following: For most words, the 

ordinary speaker will, on reflection, agree that there is no motivation for 

them (e.g. for horse) but he/she will say that there is a motivation for 

sound symbolic words. Of course there will be a border area where 

different speakers will disagree or find it hard to judge. 

Bolinger (1950) takes a radical methodological approach to morphology 

in general, with phonesthemes (which are also called affective morphemes 

in contrast to neutral morphemes, in spite of the impossibility of stating 

where the neutral ends and the affective begins) as one type of possible 

output from a strictly synchronic analysis. According to Bolinger, the 

alternative approach of taking etymology into account is not available 

since ordinary, naive speakers' judgements are important for discovering 

morphemes. Bolinger suggests the existence of numerous phonesthemes 

but also claims that they are too fluid to be penned with limits. One 

possible way to evaluate morphemes, according to Bolinger, would be 

through lexical and textual frequencies (cf. chapter 4). Bolinger claims 

that roughly half of the words, in English, beginning with gl- have the 

implication ‘‘visual’’. Bolinger claims that as percentages go this is better 

than some of the paradigmatic suffixes (with regard to a constant 

association of meaning and form), though of course gl- is never more 

than sporadically productive. 

1.4 Is sound symbolism productive or not? 

1.4.1 Sound symbolism is not productive 

Samarin (1978), in his studies of Gbeyan, has not found support for the 

creation of new ideophones (cf. 1.2.4). In traditional etymology the 

explanation of new coinages is often just by ‘‘analogy’’ with one other 

word (which implies non-productivity). 

12

1.4.2 Sound symbolism is productive 

Most of the above mentioned linguists who are specifically interested in 

the phenomenon and who view it as an integral part of language, also 

regard it as productive. For example, Nordberg (1986) who studied sound 

words, deals almost exclusively with new coinages. 

1.4.3 Greater or lesser degree of productivity 

Rhodes (1994) discusses onomatopoeia, aural images (mapping sound onto 

sound) and forms based on aural images. He distinguishes between ‘‘wild’’ 

and ‘‘tame’’ words, these being the ends of a scale. ‘‘At the extreme wild 

end the possibilities of the human vocal tract are utilized to their fullest to 

imitate sounds of other than human origin. At the tame end the imitated 

sound is simply approximated by an acoustically close phoneme or 

phoneme combination.’’ 

Bolinger does an assonance-rime analysis of English monosyllables (cf. 

Bolinger 1950) where the initial consonant(s) constitute the assonance and 

the remainder of the syllable is the rime. He argues that assonance-rime 

analysis (of tame words) is morphology because assonances and rimes do 

not combine productively. That, however, does not mean that a 

construction is frozen. He introduces the term ‘‘active’’ for constructions 

that produce monosyllables continuously, at a slow rate. 

1.4.4 Evaluation of the discussion of productivity 

The hypothesis that will be tested in the experiments is that phonesthemes 

are productive to a greater or lesser degree, i.e. that some phonesthemes 

are more productive than others. The intermittent occurrence of new 

forms in speech, prose and fiction, (which fit into a pattern), especially in 

child literature, constitutes an argument for this claim. The opposite view 

would mean that new coinages would be phonetically and semantically 

haphazard. However, with that view, the fairly wide-spread and easy 

comprehension of new forms would be difficult to account for. The 

concept of phonestheme involves stronger or weaker productivity. 

The distinction between understanding and production might also be 

fruitful. When being presented with deliberately constructed nonsense 

words in the experiments of this thesis, listeners have no objections to or 

difficulties in assigning some interpretation to them (cf. 7.1.3). 

13

Another problem, in this area of research is to decide where the 

borderline goes between lexicalized and more temporary, newly created, 

forms. In other words, what I experience as a neologism can be an 

established word in a subgroup or an (extinct) dialectal word. 6 

1.5 The question of etymology 

In an etymological perspective this part of the vocabulary is less static in 

one aspect, more static in another; onomatopoeic words are constantly 

recreated, but this also makes them keep much the same form throughout 

the ages, as they don't always undergo general changes of sound and 

meaning Jespersen (1922 a) gives the example of cuckoo which has not 

changed its vowel from [u] to [U], as in cut, but is pronounced [kuku]. An 

onomatopoeic word is constrained by the sound it imitates. 

In accounting for onomatopoeic and other sound symbolic expressions, a 

synchronic explanation will be given, irrespective of whether the 

explanation is historically ‘‘true’’ or ‘‘false’’. Some examples will be given 

of what some critics of classical etymology have stated. First, however, it 

should be pointed out that there seems to be a general agreement that 

onomatopoeic (and also sound symbolic) words usually do not undergo 

the same phonological changes as other words, e.g. according to Grimm's 

law, as long as they still have a sound imitative meaning. They constantly 

get renewed and sometimes reshaped by fresh imitation. Wescott (1975) 

writes: ‘‘Thus, the reconstructed proto-indoeuropean forms *pap- ‘‘teat’’, 

*tut- ‘‘to hoot’’ and *kuku- ‘‘cuckoo’’, appear in English as pap, toot, and 

cuckoo rather than, as comparative philologists would normally predict, 

*faf, *thuth and *houhg.’’ 

6 

Existing forms, like place-names, can also be given a (new) meaning, befitting their 

form, as in the book ‘‘The meaning of Liff’’ (Adams & LLoyd, 1983), which states that 

‘‘In life, there are many hundreds of common experiences, feelings, situations and even 

objects which we all know and recognize, but for which no words exist. On the other 

hand, the world is littered with thousands of spare words, which spend their time doing 

nothing but loafing on signposts pointing at places...’’ These names and meanings are 

then paired like e.g., Cranleigh: a mood of irrational irritation with everyone and 

everything, Burbage: The sound made by a lift full of people all trying to breathe politely 

through their noses, or Plymouth: to relate an amusing story to someone without 

remembering that it was they who told it to you in the first place. 

14

On the other hand, Jespersen (1922 a) also remarks that ‘‘words that have 

been symbolically expressive may cease to be so in consequence of 

historical development, either phonetic, semantic or both.’’ An example, 

according to Jespersen, is the word crow which is not now so good an 

imitation of the sound made by the bird as OE crawe was. And whine, 

pipe were better imitations when the vowel was still i as in Danish hvine, 

pibe. But the sound made by a small bird is still pronounced with an i in 

peep. 

Furthermore, writes Jespersen, some words have in the course of time 

become more expressive than they were at first. This phenomenon he calls 

secondary echoism or secondary symbolism. Patter ‘‘to talk rapidly or 

glibly’’ is to all intents a truly symbolical word, even though it comes 

from pater (paternoster) and at first meant to repeat that prayer. 

An interesting example in Swedish is (atjo) prosit as a polite reaction to a 

sneeze, of which the latter word is probably apprehended as 

onomatopoeic, at least by children. (The word atjo is of course motivated. 

It is a conventionalization of a sound from a bodily reaction and the 

relation between the sound and the meaning is indexical (cf. the crosslinguistic 

comparison in chapter 5). The Latin original meaning of prosit 

(which is ‘‘måtte det gagna’’ (may it be of use)) is probably quite dead in 

the mental lexicons of most speakers. On the other hand, according to 

Jespersen (1922 a), there are words which we feel instinctively to be 

adequate to express the ideas they stand for and others, the sound of which 

are felt to be more or less incongruous with their signification. These 

feelings of adequacy or incongruity are both examples of etymological 

creativity. A Swedish example of incongruity is perhaps munter (joyful) 7. 

Jespersen (1922 a, b) talks about ‘‘symbolism at work’’ where both sound 

and sense fit. Also, he writes, through changes in meaning, too, some 

words have become more expressive than they were formerly. His 

example is miniature, which, because of the i, has come to mean ‘‘a small 

picture’’ instead of ‘‘image painted with a minimum of vermilion’’. 

Jespersen claims that ‘‘sound symbolism makes some words more fit to 

survive’’. The word roll, in French rouler, etc. derived from Latin rota + 

7Because of the [P], cf. 5.4.3. 

15

diminutive ending -ul- gained its popularity in English, Dutch, German, 

and Scandinavian languages, because the sound is suggestive of the sense. 

He also talks of sound symbolism being in action when borrowing words 

from other languages and when coining nouns, verbs, etc. from (place) 

names. To sum up then, according to Jespersen, onomatopoeia and other 

sound symbolism do not always date back to the earliest times, and it is 

mostly uninteresting to lay any aspects of ‘‘correct etymology’’ on words 

affected by onomatopoeia and sound symbolism. 

Wescott (1978) would make the same claim for most words, but especially 

for slang and proper names. He takes a polygenetic view of word origins 

and assumes that ‘‘lexical ancestry is relative rather than absolute in 

nature’’ and that because of this derivational relativity ‘‘there is a gradual 

‘fade-out’ in the etymological antecedence of any lexeme and that this 

fade-out effect, in turn, leads inescapably to subjectivity in the assessment 

of degrees of lexical ancestry’’. Of the various processes that contribute to 

the development of polygenetic lexemes are, e.g. sound repetition, 

indifferent varieties, elision or ‘‘chopping’’, the conversion of spoken 

language into written language and the consequent feedback effect of 

writing on speech. 8 

Von der Gabelentz (1891) observes historically ‘‘false’’ but synchronically 

‘‘true’’ etymologies based on collective agreement within a given speech 

community. He writes that words linked together by both sound and 

meaning manifest ‘‘elective affinities’’ (wahlverwandtschaften) able to 

modify both the shape and the content of the words involved. Here, it is 

natural to refer to Householder's (1946) definition of a phonestheme: ‘‘a 

phoneme or cluster of phonemes shared by a group of words which also 

have in common some element of meaning or function, though the words 

may be etymologically unrelated.’’ (Traditionally, words are said to be 

etymologically related if they can be traced back to the same word, but 

not if they contain the same phonestheme, as a part of words. It would 

probably be fruitful for etymological study to explore the concept of 

8 

There is another kind of etymology, namely folk etymology, which I believe is an 

expression for the same mental process as in productive sound symbolism. There seems 

to exist a human inclination for motivated signs. Instead of using sounds that fit the 

meaning, existing words are reshaped through using other but similar-sounding 

morphemes to make the resulting word more morphologically transparent. 

16

phonestheme, i.e. sound symbolic links could be just as valid as other 

morphological links.) 

Bolinger (1968) also holds the view that traditional etymology is not 

relevant for explaining phonesthemes (for methodological reasons, cf. 

above). He also describes phonesthemes like this: ‘‘Given a particular 

word for a particular thing, if other words for similar things come to 

resemble that word in sound, then, no matter how arbitrary the 

relationship between sound and sense was to begin with, the sense is now 

obviously tied to the sound. The relationship between sound and sense is 

still arbitrary, as far as the outside world is concerned (and would appear 

that way absolutely to a foreigner), but within the system it is no longer 

so’’ (p. 242). Bolinger (1950) also argues that when two expressions are 

encountered in the same area of greater or lesser specificity of meaning, 

and are also similar in form, they are likely to exercise a kind of magnetic 

attraction one upon the other. The attraction may be extremely remote. 

Most speakers of English, according to Bolinger (1950), when they hear 

ambush, are likely to think of someone hiding in the bushes. Likewise 

with hierarchy: one tends to hear the element higher. (This phenomenon 

could be called 'interpretive folk etymology'.) Bolinger (1950) is very 

consistent in his synchronic (and spoken language) approach, which 

sometimes leads to absurdities. His main arguments for the irrelevance of 

etymology is that it doesn't match speakers' judgements and that it isn't 

compatible with morpheme convergence and divergence. 

Malkiel (1994) discusses the role of phonosymbolic (i.e. sound symbolic) 

interference in the sound development of words. He especially discusses 

the example of the older French word for close developed from Latin 

claudere being replaced (in Late Antiquity and the Middle Ages) by the 

word fermer under the influence of the word for iron (ferrum), a then 

newly introduced, highly prestigious metal. 

It is sometimes pointed out that every word has its own history, which is 

an anomalistic point of view (in the sense of antique Greek-Roman 

distinction of ‘‘anomaly vs. analogy’’). But one could also talk about 

different rule systems conflicting. Malkiel, for example, refers to 

Meillet's (1931) discussion on phonetic mini-systems explaining 

exceptional lexical units handed down from one generation to another. 

17

These mini-systems could well have existed side by side with the basic 

phonological structure. 

Rhodes (1994) writes that the abundant etymological dictionary comments 

like ‘‘origin uncertain’’ or ‘‘prob. akin to’’ for English monosyllables 

should convince one that these are innovated continuously, but at a slow 

rate (active combination, in contrast to productive and frozen). There is 

also, he writes, an ongoing reinterpretation of forms of various historical 

sources as new instances of assonances (initial consonants). An example is 

the largely Germanic sl- liquid (slop, slush) classifier (a type of 

phonestheme) which also, semantically, includes the sl- in Latinate sluice 

(< Vulgar Latin *exclusa), (Swe: sluss), in spite of different etymological 

origins. 

1.6 The phylogenesis of language 

From the question of etymology it is natural to go on to a related question 

in which onomatopoeia and sound symbolism have figured, namely the 

question of the phylogenesis of language. 

Plato, in the dialogue Cratylos, treats the contest between the two basic 

linguistic forces, convention and nature. The argument concerned the 

nature of names; Cratylos meant that names were given by nature, 

Socrates that they were conventional. 

Most theories of the origin of language have built on onomatopoeia, the 

‘‘bow-wow theory’’ (e.g. Herder, 1772) (cf. Müller, 1861) for the names 

‘‘bow-wow’’, etc) or on gestures, (e.g. Herman Paul, according to Fano, 

1962). Other variants are ‘‘divine origin’’, (e.g. Süssmilch, 1767) or ‘‘the 

pooh-pooh theory’’ (e.g. Rousseau, 1822), which says that the origin of 

language lies in interjectional, emotional cries triggered by strong feelings 

of fear, pain, lust, etc. The ‘‘yo-he-ho theory’’ says that in collective 

labor, rhythmically produced spoken sounds help coordinate the actions of 

many individuals. ‘‘The sing-song theory’’ (e.g. Jespersen, 1922 a) means 

that the origin of language lies in dance, song and related expressive 

vocalizing. 

Max Müller (1861), (who has coined humorous expressions) is a 

representative for the ‘‘ding-dong theory’’, which might appear to give a 

sound symbolic explanation. It is, however, based on his reconstruction of 

18

400-500 Indo-European roots. He writes that man had an innate 

inclination to associate certain types of sound with certain types of objects 

and actions which echoed in him in a way which is analogous to an 

object's resonance when struck. The reconstructed Indo-European roots 

are an outgrowth of the ding-dong effect. Müller would not, as would e.g. 

Jespersen, say that sound symbolism is productive in language, but that 

this instinct to give articulate expression for rational concepts in the 

human mind has disappeared since there is no longer a need for it, once 

language is established. 

One theory of how spoken language has developed from gestural language 

is ‘‘the mouth-gesture theory’’ (Paget, 1930 and Jóhannesson, 1949). It 

says, briefly, that the organs of speech tend to move in unison with hand 

and arm movements when these are used in sign language or when using 

tools. If such movements of the speech organs are accompanied by 

vocalizations, then the resulting sounds (that are similar to sounds in 

articulated speech) eventually get the same meaning as the gestures. The 

gestures of the organs of articulation are recognized by the hearer because 

the hearer unconsciously reproduced in his mind the actual gesture which 

had produced the sound. 

Jóhannesson (1949) presents material from both Indo-European and 

Semitic sources. In Indo-European languages, he claims that about 5% of 

the words could be attributed to interjection, probably of emotion, and 

10% to onomatopoeia, leaving 85% explainable as direct or indirect 

derivations from mouth-gesture. 

Other theories of the origin of language have focused more on why and 

less on how language appeared, e.g. because of social need (Révész, 1946) 

or cerebral development or as a consequence of early child language 

development. An example of the later is ‘‘the babble-luck theory’’ 

(Thorndike, 1943), which is to be understood as a theory of the selective 

reinforcement of initial, random babbling, which later became words. 

Thorndike has been criticized for not explaining how the parents of the 

earliest babblers were able to shape infant vocalizations. 

Some linguists (explicitly) believe in monogenesis, others in polygenesis 

of human language. For a much more extensive review on the subject, see 

Hewes (1977). I will not take a stand on these theories of the origin of 

19

language, but suspect that the solution probably lies in a combination of 

different theories, one of them being the bow-wow variant. 

The foregoing section leads up to the question of the role of onomatopoeia 

and sound symbolism in the ontogenesis of language. One of the more 

recent investigations, Williams (1991), concerns the phenomenon of [da] 

universally having a deictic function in early language acquisition. She 

finds support for the mechanism of cross-modal transfer existing as a 

bridge to the acquisition of language and she holds that this can shed some 

light on sound symbolism in adult language. The cross-modal transfer 

doesn't disappear entirely once it has fulfilled its function, even though 

this aspect may later be regarded as superfluous. Without any other 

comparison this is similar to Müller's (1861) views on the genesis of 

language. 

1.7 Universality versus language specificity 

One important question, which is not always clearly accounted for, is that 

of the universality or language specificity of sound symbolism. 

Universality can be a feature of the semantic side, the expressive side or 

more abstractly. 

1.7.1 Sound symbolism is universal 

Samarin (e.g. 1978) writes that it is a universal fact that all languages use 

onomatopoeia and sound symbolism, but that the expressions differ. He 

does not make any claims for the semantic side, other than that all 

languages have an expressive function. 

Bolinger (1965) writes that the tendency of forms to mold themselves on 

other forms with like meanings and of meanings to mold themselves on 

other meanings, conveyed by like words, is universal. Allot (1973), using 

data from many languages, claims that there is a tendency for words that 

sound similar to have similar meanings in languages not known to be 

historically related. 

The meaning of [i] has been well studied, and claims have been made for 

its universality in sound symbolism (Ultan, 1978; Jespersen, 1933). 

There seems to be a general agreement that phonesthemes are language 

specific, a conclusion which is not necessary; besides, phonesthemes are 

20

often similar in related languages. A reasonable claim as concerns 

universality is that ‘‘the intralinguistic variation recapitulates the 

interlinguistic variation’’ (Allwood, 1985). Wescott (1975) handles the 

same phenomena by claiming that what is not in the microlanguage 

(conventionally structured speech) is to be found in the allolanguage 

(speech which is alienated from conventionally structured speech e.g. 

infantile babbling, non-grammatical interjection, poetry and song). He 

claims universality in processes like reduplication and lengthening. Some 

experimental linguists have explicitly tackled the question of universality 

by testing people speaking different languages, e.g. Osgood (1962) (see 

1.10.5). 

Allwood (1983) discusses language in general, about the relationship 

between language and thought and its connection with universal and 

relativistic standpoints. The conclusions relevant to onomatopoeia and 

other sound symbolism are, in brief, that universal traits in language 

mostly concern the semantic, not the expression side and that semantic 

universals are most likely in areas that are biologically grounded, e.g. 

motoric and perceptual activities. Examples are spatial relations and color 

terms. 

1.7.2 Sound symbolism is not universal 

Some linguists seem to have taken the fact that there are differences 

between languages concerning (presumptive) onomatopoeia and other 

sound symbolism as a proof of the non-existence of sound symbolism, on 

the grounds that if sound symbolism did exist, it should be universal in 

form and content. This conclusion rests on some unarticulated 

assumptions concerning the way in which sound symbolism ought to be 

accounted for, e. g. as innate, cf. Bolinger (1968), ‘‘If there were a real 

connection between the sound of a word and its meaning, a person who 

did not know the language would be able to guess the word if he knew the 

meaning and guess the meaning if he heard the word. This almost never 

happens, even with words that imitate sounds’’. 

Austerlitz (1994) has studied vowels in e.g. Finnish and suggests that there 

is a language-specific correlation between recentness and exploitability in 

sound symbolism. In Finnish the vowels (and the consonants) are 

unequally rooted in the system. The vowel ö /ø/ is the most recent arrival 

21

(and is secondary, and marked, as all front rounded vowels are) in the 

sound system and it is the vowel which is the most sound symbolic. 

1.7.3 Evaluation of the discussion of universality 

versus language specificity in sound symbolism 

An important question is whether phonesthemes are language specific, as 

Bolinger (1950) claims. If so, then the origin of each phonestheme could 

be attributed solely to chance, a ‘‘clustering effect’’ (which would be 

interesting in itself). Another possibility is that phonesthemes show 

universal tendencies, even if ‘‘only’’ in cases concerning semantic 

features. If they show universal tendencies, the explanation could partly 

have to do with innateness, and partly involve a basic relationship between 

environment and individual. 

But it may be that, instead of the dichotomy between certain sounds which 

are universal and certain phonesthemes which are language specific, one 

should distinguish between (a) universal sound symbolism pertaining to 

certain sounds and sound clusters and (b) the phenomenon of 

‘‘phonesthemicity’’, i. e. the tendency of certain sound-meaning 

combinations to mold themselves on other combinations, pertaining to all 

sounds and clusters, etc. In the latter case the result could well be 

language specific while the thrust for analogy itself is universal. The next 

question is if the clustering effect is enhanced in certain types of meaning 

or sound. There might also be a frequency effect emanating from 

syntagmatic context. 

If the semantic-phonetic relationships of motivated words could be 

analytically treated one by one, my assumption is that the existence of 

universality in phonesthemes on the phonetic side (i. e. that e.g. imitation 

of ‘‘wet sounds’’ is done with the same speech sounds in different 

languages) is most likely at a level of (combinations of) distinctive 

features, e.g. voiceless, fricative, etc. 

On the semantic/functional side phonesthemes are probably partly 

universal and partly language specific. The result here probably depends, 

to a great extent, on how abstract the semantic categorization is. Also, 

some semantic fields are more likely to contain sound symbolism, e.g. 

[DIMINUTIVE]. 

22

According to Allwood, as referred to above (1983), the universals in 

general to be expected, viz. universals of content are likely to be related 

to perception and perhaps to motoric behavior. This proposal does not, 

however, explain all semantic categories involved in sound symbolism. 

Whorf (1956) has an argument for universals of content: ‘‘in the 

psychological experiments, human subjects seem to associate the 

experiences of bright, cold, sharp, hard, high, light (in weight), quick, 

high-pitched, narrow, and so on in a long series, with each other; and 

conversely, the experiences of dark, warm, yielding, soft, blunt, low, 

heavy, slow, low-pitched, wide, etc. in another long series. This occurs 

whether the WORDS for such associated experiences resemble them or 

not, but the ordinary person is likely to NOTICE a relation to words only 

when it is a relation of likenesses to such a series in the vowels and 

consonants of words’’ (p. 267 f). 

It is not always clear if the authors mentioned above have discussed 

understanding, production or both. It is most clear in the experiments 

conducted, cf. 1.10.5). 

1.8 Context 

Words (and parts of them) are always perceived in a context, which 

influences their interpretation. The types of context that are interesting 

for sound symbolism are: 

i) phonetic/phonological and semantic 

ii) only semantic 

iii) situational 

The interpretation of the meaning connected with e.g. a consonant cluster 

can thus be influenced by phonetic/phonological and semantic 

(phonesthemic) context e.g. other consonants, vowels, intonation, etc. 

connected with a certain meaning. 

The interpretation can also be influenced (e.g. disambiguated) by semantic 

context only. For example, a consonant cluster can have as a weak 

phonestheme meaning 'wetness'; the cluster pl-, which is otherwise 

mainly 'pejorative', can also mean 'wetness'. In the context of other words 

23

having to do with wetness (and e.g. a word like rinna (flow)) a neologism 

beginning with pl- could adopt a meaning of wetness. 

The disambiguation of a consonant cluster can also depend on the speech 

situation in which e.g. a neologism or an ambiguous word is uttered, as 

for all words. An example of an ambiguous word is klabb (‘‘wet snow’’ 

or ‘‘chunk of wood’’). The cluster kl- can mean 'wetness' (wet snow) or 

'short wide form' (chunk of wood). 

It is probable that at least some motivated words, e.g. the sound words of 

teenagers (like tssccht in ‘‘she wears her hair like this : tssccht’’), are 

more dependent on situational context than more arbitrary lexical 

morphemes; a person, who was not present in the situation were the word 

‘‘tssccht’’ was uttered, cannot understand exactly how the girl's hair 

looked. 

Grammont (1933), who was mainly interested in the evocative value of 

vowels, especially in word forms reduplicated with a vowel change, in 

different languages (e.g. ritsch-ratsch, piff-paff-puff) claimed that the 

meaning of a vowel manifests itself when it is prompted by the meaning 

of the text or when it at least does not stand in contradiction to it. The 

degree depends on the subjectivity of speakers and listeners as well as on 

situations, e.g. affective speech and poetry being favorable. 

As will be seen in chapter 7, the experiments in the present work are 

mainly done without variation of context. The reason for this is the desire 

to know how much (if any) of the sound symbolic meaning is conveyed by 

certain consonant clusters themselves, without context consisting of 

vowels, other consonants, other words, intonation, gestures, etc. 

However, in these experiments situational context is present from another 

aspect. As is shown in chapter 4, some consonant clusters have a higher 

frequency of motivated words with a certain meaning while other 

consonant clusters are less dominated by motivated words. This is often 

reflected in the results of the tests described in chapter 7. It is reasonable 

to believe that neologisms beginning with not so clearly profiled clusters 

are more dependent on the linguistic or extralinguistic context for their 

interpretation. 

24

An example of phonetic/phonological and semantic (phonesthemic) 

context could be an invented name like Pjäfser which isn't very attractive 

because of the pejorative pj- and the pejorative -fs (like in hafs, slafs, 

tjafs, krafs, rufs, bjäfs). These two clusters make the weakly pejorative - 

E- come to life and add to the pejorative impression. (In addition to this 

the suffix -er also has a pejorative nuance.) 

The cluster fl-, which mainly means 'quick movement' also has the 

phonestheme meaning 'pejorative' (and few others). In a neologism like 

flafs it is likely that the whole word will be interpreted as a pejorative 

because of the pejorative ending -fs. The less common pejorative meaning 

of flis activated because of the phonological/semantic context of -fs. 

From this we can see that words with sound symbolism are neither more 

context dependent nor less context dependent than other words. Certain 

clusters are more context dependent, while other are less context 

dependent, depending on the lexical sound symbolic strength of the cluster 

(cf. chapter 4). 

While the cluster pj- is mostly pejorative, another clearly sound symbolic 

cluster kl- carries several meanings, i.e. 'sound', 'talking', 'pejorative', 

'wetness', 'adhesion' and 'short wide form'. What determines the meaning 

of a neologism beginning with kl- must be either phonesthemic, other 

semantic or situational context. Looking at the examples in Appendix 1 

there seems to be a tendency that kl- words meaning 'short wide form' 

are monosyllabic and ending with a geminate consonant. They do not end 

with fricative clusters; these appear here to be reserved for 'sound' and 

'wetness' meanings. Phonesthemic context seems to be important but it is 

not possible, at this stage, to give rules for this. (Cf. the discussion in 

Allwood (1982) about meaning potential and contextual conditions for 

different meanings. Cf. also 5.6.) 

The influence of context explains why the same sound (sequence) mostly 

unambiguously can appear in both sound symbolic words (as 

phonesthemes) and in non-sound symbolic words (as just a phoneme 

sequence). 

Nerman (1954) makes an interesting analysis of the sound symbolism of 

vowels and consonants in Swedish poetry. He writes that they always stand 

25

in a context, and really the whole poem, or at least the stanza, ought to be 

quoted for every example and that all means conspire: content, rhythm 

and ring. However, his analysis is not preceded by a lexical 

(paradigmatic) analysis. 

1.9 A framework for models of sound symbolism 

Models of onomatopoeia and sound symbolism must take such factors into 

account as discussed in 1.3-1.8. One factor is the position of onomatopoeia 

and other sound symbolism in language, (i. e. in grammar and lexicon). 

Are they primarily central or somewhere on the periphery of language? I 

claim that onomatopoeia and sound symbolism are central and inside 

language and a part of morphology. 

However, some newly created motivated expressions (like the sound 

words of teenagers, e.g. krch, ppff, dåing) seem to be more context 

dependent than most established adjectives, nouns and verbs and therefore 

on the periphery of language - words are created in subgroups and are 

used for a limited time (cf. Kotsinas, 1994). The study of sound 

symbolism naturally leads to the study of processes of language 

development: How do words and sounds enter language or disappear from 

language? 

Phonesthemes vary in regard to productivity. The question of 

productivity also involves the issue of production of new expressions vs. 

understanding of such expressions. Sound symbolic neologisms are not 

created all the time, but are created now and then. But when they are 

created, they are easily understood by the listener; nonsense words can be 

interpreted almost immediately. 

Onomatopoeia and other sound symbolism are probably driving forces in 

etymological development (and some claim they were so in the genesis of 

language). 

The general framework has to include a view of language as a dynamic 

phenomenon with components of different dignity for different functions. 

Sound symbolism is always possibly present, but new expressions (words 

or phonesthemes) can be seen as floating in and out of language. Whether 

or not they are actually ‘‘in language’’ depends on situation, stylistic 

context, time span under study and geographical area studied. 

26

The description and explanation of sound symbolic phenomena is related 

to the question of universality. If there are universals in (different aspects 

of) sound symbolism, this is compatible with innateness of sound 

symbolism. 

Many of the above mentioned points are relevant for most aspects of 

language, but perhaps more typically so for onomatopoeia and other 

sound symbolism, which are often more productive and universal. This 

leads back to the first question of their status in language. Other sound 

symbolism and onomatopoeia are different from other expressions, 

considering some of the points mentioned above (productivity and 

universality). But, above all, they are different because of the nonarbitrary 

dimension. 

The dimensions of the framework are: 

1) universality - language specificity 

2) innateness 

3) degree of conventionalization 

4) productivity 

5) centrality in language (historically, genetically, frequentially) 

6) types of context determination 

and they can be related 9 in the following way: 

innate 

universal 

not innate language specific 

conventional 

conventionalization 

centrality productivity type of context 

determination 

Figure 1.3 A framework model for sound symbolism. 

27 

actual sound 

symbolic meaning 

9The arrows in the model in figure 1.3 stand for different types of relationships, which 

are spelled out in the text below.

The model in figure 1.3 shows the following: Assuming that sound 

symbolism is central in language, this fact is compatible with both 

innateness and non-innateness of sound symbolism. If it is innate it must 

be universal and if it is not innate it is language specific (or universal due 

to pure chance). In both cases new expressions can be produced, based on 

innate capacity and phonesthemes which are pre-existing because of 

innateness or because of convention. Naturally there are no innate forms 

that are unaffected by convention, and therefore universals, like i 

connected to smallness, are not absolute. Smallness does not implicate i 

and i does not implicate smallness. Language specific expressions can be 

said to be created by convention while universal expressions are affected 

by convention. 

Context affects the meaning of all expressions. The meaning potentials of 

both the language specific conventional clusters like kl- ('sound', 

'talking', 'wetness', 'adhesion', 'shortwide form' and 'pejorative') and 

more universal phonesthemes like i (smallness, high pitch, light) are 

disambiguated by context (phonesthemic, situational, etc.). 

A lexical description of sound symbolism in Swedish is presented in 

chapters 4 and 5. The lexical description, which shows the meaning 

potentials for consonant clusters, treats sound symbolism as a part of 

language and not as an exception. The description is a basis for 

predictions of sound symbolism (chapter 8). It does not claim 

universality. The explanatory model of chapter 2, which draws on the 

above discussion and describes in detail the nature of the motivated 

relation between expression and meaning of sound symbolic words, is, 

however, easily relatable to many languages. 

1.10 Expression and content of sound symbolism 

Before I continue with the presentation of more concrete results obtained 

by various linguists, I will make some comments about the expressions 

and contents of onomatopoeia and other sound symbolism, i.e. phonemes, 

sounds, phonological features, semantic fields, and level of semantic 

categories. 

28

1.10.1 Expression 

The expressive side of onomatopoeia and other sound symbolism is 

usually described in terms of phonemes, i.e. as belonging to the language 

in question, or as sounds (phones) not belonging to the phonemic 

inventory of the language in question. More seldom it is described in 

terms of some kind of phonological or phonetic features. 

In this presentation the phonemes are written as in the literature, but one 

has to keep in mind that phonemes are relative to the phoneme system of 

the language under consideration. 

1.10.2 Content 

In terms of content, especially if one is interested in similarities between 

languages, it is convenient to work with semantic fields (Trier, 1934, 

Lehrer, 1974) and semantic features. When dealing with certain kinds of 

onomatopoeia, this is not so important; dogs sound about the same in 

different countries, so different expressions can be fairly 

straightforwardly compared. Interjections, on the other hand, often 

express different aspects of human emotions and are therefore often more 

difficult to compare. 

While an expression such as aj (ouch) can probably be translated fairly 

easily into other languages, many other exclamations are probably better 

described with semantic components in relation to a semantic field, rather 

than just translated to the nearest synonym. E.g. ohoj (ahoy) has an 

expression with a similar content (EwE) in Ososo, but this is not used to 

adults. Schas (shoo) in Icelandic has a special expression when directed to 

sheep (hau hau instead of sch). Hoppsan (whoops) in Slovenian is 

translated differently depending on whether the surprise is experienced in 

a negative or positive way (Oho: and Oi). 

1.10.3 Expression and content in different analyses 

It is often difficult to compare the results obtained by different linguists 

because of different levels of analytic semantic categories. The categories 

can be very broad, e.g. things and appearances (Jespersen, 1922 a) or 

more narrow, e.g. color (Samarin, 1978). Also, the categories differ 

from one linguist to another. In the context of onomatopoeia and other 

sound symbolism it is difficult but perhaps desirable to strive towards a 

29

ase level à la Brown (1958). Bolinger (1950) is of the opinion that the 

level of specificity determines which morpheme analysis is made. ‘‘The 

lower the specificity of meaning, the larger is the number of forms that 

may be subsumed under one morpheme.’’ As an example of this, he gives 

Nida's characterization of the 'suffix' /-´r/ i hammer, ladder, spider, 

otter, badger, and water as having some sort of 'grammatical meaning'. 

In spite of these difficulties I will try to compare the semantic 

categorizations of different linguists, especially those who have done a 

more extensive analysis. 

The most elaborated taxonomy for onomatopoeia and other sound 

symbolism is Jespersen's (1922 a). He describes it as ‘‘a preliminary 

enumeration of the most obvious classes, with a small fraction of the 

examples ... collected.’’ The classes are: 

1. direct imitation, e.g. splash, klonk (onomatopoeia) 

2. originator of the sound, e.g. cuckoo, or nicknames of nations from 

recurring words. 

3. movement (inseparable from sound), e.g. flicker, snatch, slide 

4. things and appearances (this seems to be form and light), e. g. ‘‘ a 

thick stick, a knot of wood, a bit of food, a protuberance, a small 

hill’’; gloom, light, dunkel 

5. states of mind, e.g. grunt, sulky, also pejoratives, e.g. bosh 

6. size and distance, e.g. teeny 

7. length and strength of words and sounds; this gives an emotional 

effect, e.g. Danish langsommelig or Hungarian short words for commands 

and long words for entreaty; the category also mirrors perfective, plural, 

distance in time and space. 

A few points of criticism of Jespersen's taxonomy are the following: 

1) The basis of division is not homogeneous. The basic principle is 

semantic, but how this is applied is also open to discussion. The main 

critique is, however: a) category 4 is a very broad semantic category, it 

covers almost everything. I believe Jespersen wants to emphasize the 

interpretation of visual perception. If so, then auditory perception ought 

to be treated in the same way, as things and appearances; b) class 7 is the 

result of a categorization based on form and not on content. 

30

2) The categories are not mutually exclusive, either empirically (many 

words can be placed under both 1 and 3) or, analytically. Category 7 has 

several types of content which characterize other categories, e.g. 

'distance' and 'emotional effect'). 

3) The categories are not exhaustive. It is, for example, evident that there 

is sound symbolism in the semantic spheres of 'wetness' and 'dryness'. 

The categories suggested in this thesis (see chapter 2) agree on some 

points with those of Jespersen though they are a bit more detailed. The 

category 'things and appearances' takes up examples which would be 

classified as 'form' and 'light' by me. The class of 'pejoratives' of the 

present thesis falls under Jespersen's 'states of mind'. The category 

'originator of the sound' is just a normal extension of meaning from Act 

to Actor, (e. g. sökande (pres. part). – sökande (noun), (see Malmgren, 

1988), which also applies to words which are not sound symbolic in any 

way. Jespersen's last category does not have a correspondence in this 

thesis but is interesting and deals with, e.g. ‘‘the emotional value of some 

'mouth-filling' words’’, e.g. evig – evinderlig (eternal), vex – aggravate 

and slang words like splendidious for splendid. 

None of the authors below has tried to relate the categories which they 

have found to each other, in order to explain, other than partially, the 

phenomena of sound symbolism. In contrast, one important aim of this 

thesis is to give an explanatory model for the semantic categories that 

reoccur in sound symbolism in Swedish. 

1.10.4 Results - data from different authors 

The results presented in this section are primarily second-hand data, a 

large part of them from the authors discussed in 1.1-1.8. These data 

arrived at by earlier linguists will be summarized, without any evaluation. 

Therefore, this section is a guide to what sounds and meanings different 

linguists have studied. Comparisons between the authors are difficult for 

several reasons; e.g. Bolinger claims there is an unclear boundary 

between neutral and affective morphemes, Wescott includes the whole of 

allolanguage, others have studied contrast, etc. 

None of the authors has claimed to give a complete inventory of sound 

symbolism, usually they are e.g. simply illustrating an argument. 

31

Sometimes they just mention sound symbolic contents without giving 

examples of the relevant sounds or sound combinations. 

Table 1 Phonesthemic sounds and meanings of different authors. 

Examples are given in regular spelling except for some IPA-symbols. The 

language described is English if nothing else is written. 

Bolinger (1950): 

sound content 

gl- phenomena 

light 

of 

fl- phenomena 

movement 

of 

itr intermittent 

ow steady 

Er intense 

kr- bent 

-amble locomotion 

-ust surface formation 

-usty old 

-lessness indifference 

i diminutive 

-utter discontinuity 

-ash hit, fragments 

tw- twisting motion 

st- arrest 

sp-t rush of liquid 

str-p line having 

st-nt 

breadth 

piece of 

performance 

s k swift movement 

-ump awkward, heavy 

Bloomfield (1933): 

sound content 

32 

fl- moving light 

fl- movement in air 

gl- unmoving light 

sl- smoothly wet 

k r - noisy impact 

skr- grating impact or 

sound 

sn- breath-noise 

s n - quick separation 

snor 

movement 

creep 

dJ- up-and-downmovement 

b - dull impact 

-E violent movement 

-E´ big light or noise 

-awns quick movement 

-im (´ ) small light or 

noise 

-Um clumsy 

-Et(´ ) particled 

movement 

Humboldt (1836/1907, German) 

sound content 

st- firmness 

n - sharp cutting 

w - random movement 

u hollow and dark

Rhodes (1994): 

Classifiers: 

sound content 

st- 1 dimensional (stick, staff, stem) 

str- 1 dimensional 

flexible 

33 

(string, strand, strip) 

fl- 2 dimensional (flap, flat, floor) 

S/sk - 2 dimensional, flexible (sheet, scarf) 

n - 3 dimensional (knob, knot, node, nut) 

sp- cylindrical (spool, spine, spike) 

dr-/tr- liquid (drink, drain, trickle, trough) 

Paths: 

sound content 

tr-/dr- simple (track, trip, drive, drag) 

p-/b- ‘‘anchored’’ (push, pop, bump, bounce) 

j-/tS - short (jerk, jiggle, jagged, chop) 

w - back and forth (wag, wiggle, wobble) 

p - abrupt onset (pop, ping, peep) 

b - abrupt, loud onset (boom, bang, beep) 

b l - loud, air-induced sound (blat, blast, blab) 

kl- abrupt onset (clank, click, clip, clop) 

r - irregular onset (rip, roar, roll) 

y loud, vocal tract noise (yell, yap, yak) 

Q- low pitch, slow onset (thump, thud) 

p l - abrupt onset (plink, plop, 

plunk) 

k r - abrupt onset (creak, crack, crunch) 

tS- irregular onset (chirp, cheep, chatter) 

w- poorly resolvable onset (whiz, whack, wham) 

z- poorly resolvable onset (zip, zing, zap, zoom) 

dr- liquid (drip, drain, drop, drizzle) 

sl- liquid (slop, slush) 

fl- liquid (flow, flush, flood) 

m - liquid (mud, mush, mire, marsh)

Plato: 

(sounds and meanings discussed 

in Cratylos, on Greek)) 

form content 

r movement 

i all fineness (it 

can penetrate 

ph, ps, s, z 

everything) 

everything 

similar to 

airstreams 

d, t binding and 

standing still 

l gliding 

movement 

gl something 

sticky 

n - inside (it is 

pronounced 

inside the 

mouth) 

a size (largeness?) 

e (eta) length (long?) 

o roundness 

Sapir (1921): 

sound process content 

reduplication distribution, 

plurality, 

repetition, 

customary 

activity, 

increase of size, 

added intensity, 

continuance 

34 

Malkiel (1978): 

sound process content 

reduplication 

+vowel change 

Jespersen (1918): 

disorder, 

confusion, 

rubbish, thrash 

sound content 

m roundness 

Jespersen (1922 a): 

no sounds content 

direct imitation 

originator of 

the sound 

movement 

things and 

appearances 

states of mind 

size and 

distance 

length and 

strength of 

words and 

sounds 

Wescott (1975): (content is often 

not mentioned) 

sound content 

u w emotive 

z ‘‘an unusual 

semantic 

function’’ 

‘‘sound 

repetition’’

‘‘sound 

alternation’’ 

‘‘allolinguistic 

prefix’’ 

‘‘pentestheme’’ 

Householder (1946) 

no sounds content 

dislike 

destruction 

projection 

protuberance 

(short and 

roundish) 

collectives: heap 

or pile, 

cluster or knot, 

large 

shapeless piece, 

indefinite 

number or 

amount 

35 

thick, coarse, 

soft substance 

dull, loud, 

indistinct noise 

coward, failure 

or deceiving 

past tense, past 

participle 

deficient in 

some desirable 

quality 

Sigurd, B. (1965) (Swedish): 

sound meaning 

fj- pejorative 

fn- pejorative 

pj- pejorative 

-ms pejorative 

-mp pejorative 

-sk pejorative 

-b(e)l pejorative 

-m(e)l pejorative 

Some generalizations 

Some phonesthemes are the same in English and Swedish, e.g. fl- 

‘‘phenomena of movement’’, e.g. flicker, flutter; fladdra, flaxa, while 

others are different in English and Swedish, e.g. English: fl-‘‘moving 

light’’ e.g. flicker, gl- ‘‘unmoving light’’, e.g. gleam, and Swedish bl-, 

gn-, e.g. blänka, gnistra ‘‘light’’ (not to mention the discrepancies 

between different analysts of English). Therefore, there is not complete 

universality of expression on the phoneme level, since for example 

English does not use the cluster gn- for light phenomena (it is not even a 

consonant cluster in English).

There are sound symbolic contents in some languages which do not occur 

in Swedish, e.g. in the field of color. This indicates that all sound 

symbolic contents are not the same in all languages. On the other hand, 

size (diminutive) stands out for itself since it seems to occur in almost all 

languages – and in a similar phonetic form (Ultan, 1978). 

The categories on the content side of the lists above can be summarized as 

belonging to the following semantic fields. This is one possible 

classification, mainly with a departure in the senses: 

Hearing 

sound, noise (dull, loud, indistinct, big, small, rushing liquid) 

Vision 

light (small, big, moving, dark) 

Touch 

surface structure or substance (thick, coarse, soft) 

Movements 

(random, twisting, swift, locomotion, up and down, quick, violent; or 

stop) 

Forms 

(bent, line having breadth, projection or protuberance, round, hollow) 

Mind 

attitude, emotive (indifference, dislike) 

Pejorative 

(old, awkward, heavy, coward, failure, deceiving, deficient, clumsy, 

disorder, confusion, rubbish, thrash) 

Size 

diminutive 

augmentative (increase in size, added intensity) 

Number 

collectives (heap or pile, cluster or knot, indefinite number) 

intermittent (distribution in space, plurality, repetition, discontinuity) 

Various 

liquid 

steady, firmness 

destruction (hit, fragments, sharp cutting) 

inside 

tense (past tense, past participle) 

piece of performance 

unusual semantic function 

36

Three of the features are perceptive: 'hearing', 'vision' and 'touch'. Taste 

and smell, however, do not occur. Perception of 'movement' is often cooccurring 

with perception of sound, a contiguity relationship (cf. 

discussion of Hinton, Nichols and Ohala (1994) in 1.11). 'Form' is 

likewise, perceptually, closely connected with 'vision' and secondarily also 

with 'touch' (cf. Brown, 1958). 

The semantic category 'mind' (attitudes and emotions) is a category that 

probably comes naturally and can be seen as indexically related to 

expressions for dislike, etc. This may also be the connection to at least 

some pejorative expressions. The size categories can be seen as iconically 

related to speech sound (in accordance with Ohala, 1994). 

It seems, from this overview of the literature, that the most common 

semantic categories are related to the three senses hearing, vision, touch 

(situated in the cortex of the human brain, in contrast with smell and 

taste) or they are metaphorically, metonymically, indexically or iconically 

related to the senses. 

1.10.5 Experimental results 

Sapir (1929), (who communicated with Jespersen on sound symbolism and 

who wrote his master's thesis on J.G. Herder's (1772) "Essay on the 

origin of speech") wrote about ‘‘latent expressive symbolism’’, a type of 

relationship, e.g. in words like teeny and tiny, which is ‘‘directly 

expressive of the difference in meaning’’. He conducted some 

experiments. In one of them more than 500 subjects were asked to decide 

which of the nonsense words mil and mal meant a large table and which 

meant a small table. 80% agreed that mal is better suited to the large 

table. An investigation by Bentley and Varon (1933) indicated that [a] 

sounds are felt to be larger than [i] sounds in the proportion 4:1, and also 

that [a] is rounder and [i] is more angular in the proportion 3:1 and [a] is 

softer and [i] is harder in the proportion 2:1. 

Sapir's experiments were further developed by Newman (1933). He tested 

both vowels and consonants with respect to the small-large and the brightdark 

dimensions. His results were the following: vowels agree with 

articulatory position (front-back:small-large); also consonants agree with 

37

articulatory position (labial-dental: small-large); accentuation of vowels is 

heavily in favor of largeness and darkness. 

Newman (1933) also made a study of a Thesaurus with respect to the 

categories of 'greatness', 'smallness', 'size' and 'littleness'. He did not, 

however, find great support for sound symbolism here. 

The subject of correspondence between meaning of speech sound 

sequences and abstract graphic figures was investigated by Usnadze (1924) 

and Köhler (1930). Two nonsense line drawings and two nonsense words 

maluma and takete were presented to subjects who were asked to decide 

which sound matched which drawing. The overwhelming majority 

assigned maluma to the rounded figure and takete to the angular one. This 

result has been shown for several other languages (Holland and 

Wertheimer, 1964; Davis, 1961). 

The longest series of experiments concerns the question of whether, and 

to what degree, lexical oppositions in meaning bear any consistent lawful 

relationship to the symbolic properties of sounds. Tsuru and Fries (1933) 

initiated matching experiments, where verbal data in the manner of 

Köhler's maluma study were used. Lists of pairs of opposites in two 

different languages were prepared, e.g. big/small - gross/klein, and 

presented orally. Subjects who knew only one of these languages were to 

match the corresponding words. They succeeded with a certainty 

exceeding chance. Other linguists have followed trying to eliminate the 

semantic cues, e. g. by matching two languages that are both unknown to 

the speakers. The results were less successful in these cases as no semantic 

dimension could be available to the subject. 

A recent continuation of these experiments was done by Lapolla (1994). 

Since many of the earlier results have had methodological weaknesses he 

started a new set of experiments, on tonal morphology in Mandarin 

Chinese. The results show that there is a cross-linguistic tendency toward 

associating acoustically acute 10 segments with 'small' category words, and 

acoustically grave 11 segments with 'big' category words. These results are 

explained with reference to the ‘‘frequency code’’ theory developed by 

10high frequency energy 

11low frequency energy 

38

Ohala (1994). This theory implies that sound symbolism is a manifestation 

of a larger ethological phenomenon that is also seen in the vocal 

communication and certain facial expressions 12 of other species. In 

intonational communication of affect and in sound symbolic vocabulary 

there are sound-meaning correlations where high F0 signifies smallness, 

non-threatening attitude, desire for goodwill of the receiver, etc, whereas 

low F0 conveys largeness, threat, self-confidence, and self-sufficiency. 

The common connections between segments (consonants and vowels) with 

low frequency energy 13 and largeness and segments with high frequency 

energy and smallness are also in accordance with the frequency code. 

Wilde (1958), Müller (1960) and others have studied another area of 

motivated expressions, namely the expressions of emotions. The present 

thesis will not investigate this area, see however Abelin and Allwood 

(1984). Wisseman (1954) studied the creation of onomatopoeic words 

from noises. Various noises were presented to the subjects, who were not 

able to observe how they were produced. They were then instructed to 

invent or select names for the noises. The research design has been 

criticized, e.g. Hörmann (1979), but it seems that vowels represented the 

pitch and qualitative color feature of the noise: the i sound imitates a 

light, spiky noise, the u sound a low dark noise. The number of syllables 

in the invented words was not proportional to the length of the noise; it 

reflected sections of the noise sequence. A noise with an abrupt beginning 

was represented by a word beginning with a voiceless plosive p, t, k. A 

gradually starting noise was described by a word beginning with s or ts. 

Also Osgood (1962) studied (universal) phonetic symbolism. Osgood's 

semantic differential (Osgood et al, 1957) testing the basic dimensions of 

‘‘value’’, ‘‘strength’’ and ‘‘potency’’ has been widely used in testing 

emotional content in words, also for phonetic symbolism. Nonsense 

syllables were rated on the semantic differential. The results for 

American and Japanese speakers were almost identical. For both groups 

12The smile is connected with high F2 (and higher formant frequencies) since retracting 

the corners of the mouth shortens the vocal tract and thus raises those frequencies. 

13In consonants, voiceless obstruents have higher frequency than voiced, and dental, 

alveolar, palatal and front velars have higher frequencies (of bursts and frication noise) 

than labials and back velars. Dentals have higher frequencies of formant transitions than 

labials. High front vowels have higher F2 and high back vowels the lowest F2. 

39

front consonants (e. g. p) were more pleasant than back consonants (e.g. 

g); high frequency sounds were associated with smallness and impotence. 

However, it seems that his semantic categories are too restricted and the 

task of placing words/concepts on a 1–7 scale is not really feasible. 

Symbolism of French vowels was studied by Chastaing (1958). i is treated 

in association with acuteness, smallness, lightness, rapidity, and closeness. 

Also consonantal oppositions were studied by Chastaing (1965, 1966). 

Stops are hard, continuants soft, r is rough, strong, hard, etc. in contrast 

to l which is smooth, weak, light-weighted, etc. 

Fonagy (1963) compared i and u in Hungarian. In his investigations of 

children and adults i was quicker than u for 94%, smaller for 88%, 

prettier for 83%, friendlier for 82%, harder for 71 %, whereas u was 

thicker for 98%, hollower and darker for 97%, sadder and blunter for 

92%, more bitter for 86% and stronger for 80% . The responses to r 

were that r was wild, pugnacious, manly, rolling, harder for the 

overwhelming majority. 

A recent study by Sereno (1994) concerned lexical organization. 

Departing from the results of a lexical study concerning which vowels are 

the most common in English verbs and nouns, a reaction time experiment 

was performed. In this experiment the subjects categorized nouns and 

verbs. The results showed that there was a connection between syntactic 

class and phonological form in English. Verbs, with front vowels (which 

are lexically most frequent) were recognized faster than verbs with back 

vowels, while nouns with back vowels (which are lexically more frequent) 

were recognized more quickly than nouns with front vowels. This 

connection is independent of the frequency of the words. It is interesting 

to note this distinction between front and back vowels. The distinction 

between front and back vowels is perceptually salient and often also 

occurring in size-sound symbolism. 

Sereno (1994) thus proposes that noun/verb-categories and front/back 

classification of vowels (i. e. acoustical-perceptual classification) are 

organization principles of the lexicon and explicitly coded. 

As we have seen, there has been quite a large amount of different kinds of 

experimentation in this area. The tests, which have been commented upon 

40

above, have mostly been concerned with isolated consonants (i. e. not with 

consonant clusters) and vowels and, semantically, with contrasting ends on 

a scale (semantic oppositions). The experiments of this thesis treat 

Swedish (mostly) and more specifically the semantic value (not necessarily 

in opposition to some other category) of initial consonant clusters. 

1.11 Possible explanations of sound symbolism 

The different explanations of sound symbolism are of varying types but 

have often focussed on synaesthesia and proprioception and, generally, on 

the question of a biological or non-biological base. 

1.11.1 Miscellaneous explanations and 

proprioception 

Let us now further investigate some of the ideas concerning the nature of 

the relations between meaning and expression in sound symbolism (cf. fig. 

2.3). This discussion will then serve as a basis for a suggestion of an 

explanatory model of sound symbolism to be presented in chapter 2. 

Different explanations have been put forward by different linguists. Some 

linguists like Publius Nigidius Figulus, 98–45 BC (according to Jespersen, 

1922 a, p. 396), Fónagy (1963) or Peterfalvi (1965) have speculated on 

the connection between articulatory movements and meaning. Nigidius 

Figulus claimed that in pronouncing vos one puts forward one's lips and 

sends out breath in the direction of the other person, while this is not the 

case with nos 14 (i.e. an indexical relation). 

Peterfalvi (1965) claims that vowels articulated towards the exterior of 

the body are judged to be ‘‘light’’ whereas those articulated towards the 

interior of the body are judged to be ‘‘dark’’ because ‘‘the further you 

penetrate the body, the darker it is there.’’ It seems that he is influenced 

by sense analogy, i.e. the word dark can be used for different pe r ceptual 

phenomena. Fonagy (1963) claims that various movements of the tongue 

and the jaw bear likeness to different emotions. 

These and similar theories connect articulation with meaning directly and 

disregard the acoustic (or visual) link in a communicative situation. As 

for the area of non-verbal communication, Fonagy's and Peterfalvi's 

14This is of course not correct. 

41

‘‘gestures’’ cannot even have an observer, if one excepts that front 

articulations can be seen. However, there might be proprioception 

involved. 

Bolinger (1968) suggests the following metaphor: ‘‘the digital island floats 

on an analog sea’’. Digital stands for consonants and vowels, which are 

arbitrary, analogue stands for phenomena like prosody and gestures, 

which are not completely arbitrary. ‘‘... now and then a bit of the 

analogue sea washed over the digital island’’, e.g. when i stands for 

smallness and o for largeness. ‘‘The size of the mouth cavity ... is matched 

with the meaning’’. If there is something in this explanation, a number of 

factors are left out, e.g. the acoustic link. The explanation is a bit better if 

one compares the frequency of acoustic energy in i with that in o, or a, 

which is connected with the size of the mouth cavity and other vocal tract 

cavities. (see e.g. Jakobson, Fant, Halle, 1957). 

Darwin (1872) proposes a possible explanation for sound symbolic words 

related to emotions, e.g. disgust (related to 'pejorative'), and for 

interjections. The explanation is based on the instinctive contractions of 

facial muscles connected with a certain emotion. This type of explanation 

would mean that pejoratives can be classified as indexical (cf. discussion 

in chapter 2). 

A recent investigation on the topic of universal sound symbolism in 

deictic words was done by Traunmüller (1996). He found that pairs of 

demonstratives in which there is a vocalic opposition have an advantage in 

the struggle for existence in languages when F2' is higher in the proximal 

than in the distal form; he also found that nasals are preferred in first 

person pronouns while stops and other obstruents are preferred in second 

person pronouns. He also offers explanations that involve affinities with 

the association of pitch with size, the proprioceptive qualities of speech 

sounds and oral pointing gestures. 

1.11.2 Synaesthesia 

Sometimes it is stated in passing that sound symbolism can be explained by 

synesthesia, i.e. neurological connections between the sound side and the 

semantic side of a word, morpheme or phonestheme, when the meaning is 

in some way connected to one of the other senses, e.g. 'sight', 'touch' or 

to categories perceived with several of the senses, e.g. 'form', 'surface 

42

structure', 'movement'. (This resembles Müller's (1861) ‘‘ding dong 

theory’’.) 

Luria's (1977) patient, the mnemonist S. is a well known synesthete. 

When he heard a tone vibrating with 50 cycles/second at 100 dB he saw a 

brown stripe against a darker background with red tongues at the edges. 

At the same time he experienced the taste of sweet-sour borsjtj. When he 

heard [r] he always saw a ragged line. Persons with synaesthesia are 

unusual, and the experiences vary from person to person (cf. e.g. 

Cytowic, 1989). 

Aristotle claimed that the senses were clearly separated from each other. 

Newton, on the other hand, tried to find a numerical correlation between 

e.g. the wave length of green light and a certain frequency of sound. 

However, he could not find this. 

The problems with an explanation in terms of synaesthesia are several. 

According to Cytowic (1989) synaesthesia is an idiosyncratic 

phenomenon, which varies between the persons who experience it. 

Cytowic argues that ‘‘rather than being merely a more intense form of 

metaphoric speech, one can look at cross-modal metaphor as an abstract, 

linguistic derivative of the stuff of synaesthesia’’. Jakobson and Waugh 

(1979), however, mention different findings of tendencies of the 

correlation of speech sounds with colors: redness of a, yellowness and 

whitishness of e and i and darkness of o and u. 

According to Marks (1990) the synaesthetic connections do not vary too 

much between the experiences of different persons. For example, the 

frequency of the second formant of vowels can be related to black (dark) 

and white (light) so that [u] has a low frequency and dark color while [i] 

has a high frequency and light color. 

In order to give a more convincing explanation, one would need to do it 

in terms of neurophysiology. For example Freud (1891) suggested that 

language could be represented in a ‘‘field’’ in the border area between the 

temporal, parietal and occipital lobes, where all properties of the 

perception of a thing were connected in a network (smell, taste, visual 

appearance, sound, etc.) 

43

Considering the apparent similarities between the semantic categories in 

s y n a e s t h e s i a – color, form, motion, texture, luminescence, 

dynamics (but also numbers, days of the week and months of the year, 

moral judgements) – s e n s e a n a l o g i e s (sense analogies are 

linguistic metaphors, e.g. the expressions dark tones, warm colors, cf. 

Abelin, 1988) and s o u n d s y m b o l i s m , i.e. sense related categories, 

one might hypothesize a common ground for synaesthesia on the one hand 

and language phenomena like sense analogies and sound symbolism on the 

other. 

This is in agreement with Geschwind (1965, according to Cytowic, 1989): 

language depends on stable intermodal associations, especially visualauditory 

and tactile-auditory. These are the most common modes of 

synaesthesia. The relation between sound and meaning in sound 

symbolism should then stand for a neurological connection; this can be 

interpreted as an indexical relation since it is not iconic or conventional. 

According to Cytowic (1989), there is behind synaesthesia a purely 

neurological process which is connected to certain cell clusters. He studied 

the blood flow to cortical association centra for vision-hearing-touch 

during subjects synaesthetic experiences. He expected an increase in these 

association centra but did not find such an increase. Instead, the 

connections seemed to occur in the limbic system, which was studied by 

introducing electric stimulation during operations. Cytowic argues that, in 

some individuals, the developmentally earlier, ‘‘suppressed’’ limbic 

system sometimes overrides the cortex and that the boundaries between 

the senses then disappear. 

Related to the issue of synaesthesia and language universals is the study of 

verbs of perception in 50 languages by Viberg (1984). This study showed 

that there are implicational universals in the order: 

SEE>HEAR>FEEL>TASTE>SMELL. That is, if a language has only one 

verb of perception, the basic meaning is 'see'. If it has two, the basic 

meanings are 'see' and 'hear' etc. 

1.11.3 Other neurological and biological 

explanations 

A specific, neurological explanation for why languages often represent 

movement with the same sort of sound symbolic forms that they use for 

44

the representation of non-linguistic sounds is found in Hinton, Nichols and 

Ohala (1994): ‘‘ Certain rhythmic movements often directly produce 

sound. But beyond that, the rhythms of sound and the rhythms of 

movement are so closely linked in the human neural system that they are 

virtually inseparable. This is illustrated in the very natural human 

physical response to rhythmic music, in the forms of hand clapping, foot 

tapping, dancing, rhythmical physical labor, etc. ... humans ... are also 

capable of the reverse : translating rhythmic movements into sounds, 

including sound-symbolic language forms.’’ (pp. 3-4) In other words, at 

least part of ‘‘the sound symbolic feeling’’ is not something that one has 

learned. 

Related to the area of synaesthesia and sense analogy is the research area 

of bimodal perception e.g. in the form of the McGurk effect (e.g. 

Massaro et al 1993). Various experiments show how e.g. auditory 

perception is influenced by e.g. visual or tactile perception, and points to 

an interconnection of the senses. Which phonemes (or sound types) that 

are perceived do not, then, depend purely on acoustic information but also 

on information from other senses. 

Ohala (e.g.1994) discusses the frequency code and offers a solution for 

sound symbolism of diminutives and augmentatives. In contrast with 

Brown (1958) his theory points to innateness through an ethologically 

based, phonetically plausible theory for why sound symbolism exists in 

language. He identifies a link between sound symbolism in vowels, 

consonants, tones and intonation. The common factor is high-low F0 (of 

tones and intonation), noise frequency (of obstruents) or F2 (of vowels 

and sonorant consonants). High frequency is connected to smallness, low 

frequency to largeness. However, in Swedish there are several examples 

of initial [p] usually denoting smallness in spite of being grave ([p] has 

noise at low frequencies), e.g. pipa, pil, pilla, pilt, pingla, pinne, pippi, 

(however followed by [i], which has F2 at a high frequency). Another 

possible explanation of why [p] can denote smallness is the fact that [p] has 

weak noise. 

Clark and Clark (1977) discuss different kinds of categories, perceptual, 

e.g. color, shape, spatiality, cognitive categories, e.g. number, negation, 

evaluation, cause and effect, time, and social categories, e.g. kinship 

terms, personal pronouns. 

45

1.11.4 Non-biological explanations 

A non-biological explanation of size-sound symbolism is offered by 

Brown (1958), who claims that associations between, e.g., sound and size 

are simply learned through experience. Large objects usually produce 

dark (low frequency) sounds when pushed or moved in certain ways, 

whereas small objects produce bright (high frequency) sounds, (i. e. the 

relation is indexical). Thus, universality does not have to imply 

innateness. Brown (1958) also anticipates the studies on multimodal 

perception and writes that perception does not have to be connected to a 

particular receptor but is a matter of the whole body. 

Lakoff and Johnsson (1980), explain part of sound symbolism within the 

framework of their theory for metaphors. Underlying sound symbolism is 

the ‘‘conduit metaphor’’ defining a spatial relationship between form and 

content: ‘‘linguistic expressions are containers’’, and their meanings are 

the content of those containers. We expect small containers to have small 

contents, large containers to have large contents. Therefore: more form is 

more content. Examples of this is reduplication: He ran and ran and ran, 

He is very, very tall or lengthening: He is bi-i-i-i-i-ig. 

Diffloth (1994) goes against mainstream work on sound symbolism with 

unusual data and unusual explanations. He shows data from Bahnar, a 

language of Vietnam, in which i expresses largeness and a expresses 

smallness. He claims that there is nevertheless an iconic basis for this – in 

proprioceptive sensation. He concludes that iconicity can be both 

physiologically motivated and culturally relative at the same time. From 

this follows, he proposes, that iconic patterns, being language specific, 

must be described in the grammars of language. He criticizes current 

linguistic models for not being able to describe a direct relation between 

phonetics and semantics. Also, phonetics is unable to represent the 

phonetic parameters needed for Bahnar, e.g. perceived size of the tongue 

in the oral cavity. He proposes an aesthetic component of the grammar. 

His discussions on a proprioceptive explanation are very interesting but 

seem isolated in comparison with the elaborate Frequency Code theory of 

Ohala (see 1.10.5). Nevertheless, I think, it is preferable to give an 

explanation of Bahnar’s size-sound symbolism rather than discarding it as 

an exception. A multimodal model for sound symbolism should 

preferably incorporate the facts of Bahnar, i. e. proprioception. 

46

Hamano (1994) has studied palatalization in Japanese sound symbolism. 

Palatalization of alveolar stops and fricatives is associated with 

‘‘childishness’’ and ‘‘immaturity’’. He connects this to studies on language 

acquisition reporting palatalization as one of the universal characteristics 

of early stages of children's language acquisition. Palatalization is also 

reported as one of the commonest features of baby-talk, i.e. child directed 

speech (Snow and Ferguson, 1977). A possible explanation of the Swedish 

pejorative clusters pj-, bj-, fj- which do not fit into a sense related 

pattern (see chapters 2 and 4) could be that they have another origin, in 

child directed speech. Considering Hamanos findings these clusters can be 

interpreted as marked (they are unusual), rather than having their origin 

in interjections. Hamanos explanation is basically indexical (causal). 

A study of the frequency of Swedish consonant clusters in running text 

(Stenström, 1984) shows that the seven most infrequent initial consonant 

clusters are fj-, bj-, pj-, mj-, spl-, vr-, nj- (where nj- is the most 

unusual). In other words, bj-, pj- and fjare among the most infrequent 

clusters, also in running text. 

A pragmatic explanation for a part of sound symbolism, namely 

vocatives, is given by Jacobsen (1994). In Nootka languages there seem to 

be two main sound-symbolic tendencies that shape vocative forms: 

saliency and brevity. Saliency means that the word contains a prominent 

syllable that will attract the attention of the addressee. For example, nonhigh 

vowels are more salient because they are intrinsically more intense; 

falling pitch is argued to be more salient because it involves a rapid 

change in pitch. The notion of saliency can, however, be criticized. Even 

if non-high vowels are acoustically more intense, they are perceived as 

having the same strength as high vowels. Also, a raising F0 can be as 

quick as a falling. 

The discussion of different explanations of sound symbolism is continued 

in 8.3 where the arguments are related to the findings of the lexical and 

experimental studies of this thesis. 

47

2 Theoretical framework 

2.1 General considerations 

The preceding chapter was an overview of terminological issues and 

general questions concerning onomatopoeia and other sound symbolism. 

This chapter will penetrate further into the relation between meaning and 

expression in sound symbolism. There will be an attempt at an explanation 

of the relation between expression and meaning, with the aid of the 

concepts 'icon' and 'index'. 

In this thesis the standpoint is that language in important ways is a 

psychological phenomenon. The semantics of lexical analysis is therefore 

not concerned with referents, but with reference: correspondence between 

linguistic form and external data. For a discussion on this subject, see 

Allwood and Andersson (1976). 

The validity of the model presented, for example concerning the 

explanations or the number and types of semantic categories provided, will 

be supported in terms of coherence, e.g. recurrent relations between 

expression and content in the analyses, and by consensus, between subjects 

in the experiments. 

2.2 Static-dynamic, conventionality and arbitrariness 

The vocabulary of language is both static and dynamic (in the sense of 

undergoing change). Some parts are less static than other parts, e.g. sound 

symbolic words can be less static in both a diachronic and a microgenetic 

perspective (i.e., during the development of language in a situation) than 

more arbitrary words are, since sound symbolic words can easily be 

created. The phonesthemes, however, are stable and sound symbolic 

words are created with the assistance of phonesthemes. 

Thus, onomatopoeic or other sound symbolic neologisms come and go, but 

the phonesthemes, out of which neologisms can be created, are stable over 

49

longer periods of time. An example (concerning the pejorative fj-) from 

SAOB (Ordbok över svenska språket, utgiven av Svenska Akademien) is 

the word fjåka (now, if existing at all, only in some dialects) which meant 

'våp, narr, tok' (silly or crazy person). The first written instance is from 

1732; there are similar words with similar meanings in Norwegian dialects. 

The origin of the word is described as "unclear". In other words, fjåka is no 

longer existing in standard Swedish, but the phonestheme fj- 'pejorative' 

seems to have been the same in 1732 as today; the phonesthemes are more 

stable than the words. Another more recent example is the word slobb. It 

was created for a newspaper article (around 1985) and describes the sound 

of porridge landing in a bowl. The same expression may be created again in 

another context. One phonestheme used is the phonestheme sl- 'wetness'. 

In words, the connection between expression and content can be described 

as arbitrary or motivated. The motivated words can be + or - conventional; 

neologisms are -conventional. In reality a word cannot be completely nonconventional 

because the phonesthemes of sound symbolic neologisms are 

not. As discussed in connection with the model shown in figure 1.3, 

language specific expressions are created by convention while expressions 

emanating from innate (universal) processes are affected by convention. 

arbitrary door 

+conventional -conventional 

motivated shriek iiiiik 

The field of interest in this thesis consists of the motivated expressions, 

both + and - conventional. Arbitrariness is a precondition for 

conventionalization. Arbitrary words are thus always conventional while 

motivated words can be conventionalized or not. In language there can not 

be non-conventional arbitrary words. Somebody can invent words that fit 

50

into this square but nobody would understand them. The present claim is 

that the non-conventional can be understood if it is motivated. In some 

cases non-conventional words can be said to be constructed out of 

conventional phonesthemes, e.g. the expression fnölp is constructed for 

something that is 'silly' (i.e. 'pejorative'). 

The perspective on words in this thesis is synchronic, not diachronic. This 

is also the case as concerns semantic relations and extension of meaning. 

Language is not static, but in the microperspective static enough. 

2.3 Semantic analysis 

2.3.1 Conceptions of meaning 

Meaning is more or less context dependent, especially so in the reference of 

a lexeme. The meanings of concrete nouns like horse or spoon are less 

context dependent than pronouns like he or it. 

Lexemes can be claimed to have a central, intensional meaning. It is 

possible to describe this meaning as prototypes, meaning components or 

distinctive features. One can also make a distinction between core meaning 

and the vaguer emotive and associative meanings and sometimes between 

denotation and connotation. Another conception, proposed by Allwood 

(1989) is 'meaning potential'. This means, briefly, that e.g. a word form has 

a union of potential meanings, one of which is decided on by different 

contextual conditions. Homonymy only has to be resorted to in word 

classification where the different meanings are not relatable via semantic 

relations like metonymy or metaphor. 

The concept of meaning potential is useful also in the analysis of 

phonesthemes, and will be used in this thesis. A certain consonant cluster 

can have one or several meanings. Where there are several meanings many 

of these are usually relatable to each other, e.g. light – form letting through 

light – looking – gaze – reflecting smooth surface – movement on such a 

surface (gl-), which all can be characterized as metonymical (indexical). 

Other examples are sound – wetness – adhesion – pejorative (kl-) and 

51

wetness – smooth surface – quick movement – long thin form (sl-). (Some 

of these are described in more detail in chapter 4.6.) 

The meanings of onomatopoeic and other sound symbolic neologisms are 

very context sensitive. What does, for example, the word slasli (which is 

newly created) mean? Well, something having to do with wetness. If we 

get more context, like in the test matching columns of neologisms, where 

each column has a special meaning, (cf. table 7.8), subjects correctly pair 

word columns with meanings. The results depend on the possibility to 

compare the words in the different columns and compare these with the 

different suggested meanings. We get an effect from phonological and 

semantic context. 

Emotive and associative meanings are often not considered part of the core 

meaning of words. I will not adopt this standpoint but I claim that 

associative/emotive meanings belong to the meaning potential of words and 

phonesthemes. I claim that, apart from the lexicalized emotive/associative 

meanings which individual lexemes may have, there are also phonesthemes 

as parts of individual lexemes, e. g. sl- meaning wetness in e. g. slaska 

(splash), slafsa (squelch), slabba (splash), slem (slime). 

2.3.2 Semantic features and semantic fields 

The status of different types of semantic units, i.e. how concepts are best 

described, has been the subject of a long debate (see Allwood and 

Andersson (1976), Allwood (1989), Kukkonen (1989), Lehrer (1974), 

Miller and Johnsson Laird (1976) and Lyons (1977)). The most important 

units proposed are essences (defined by necessary and sufficient 

conditions) discussed e.g. by Aristotle and semantic features or 

components, discussed by e.g. Katz and Postal (1964), Leech (1969). The 

semantic features can be of different kinds, e.g. in the form of a restricted 

number of primitive (universal) distinctive features with +, – values (i. e. 

meaning components) or more generally: features found empirically in 

semantic analyses and which are possible to decompose further (i. e. 

semantic features or meaning postulates). Meaning components and 

semantic features have also proven to be a useful tool in systematic 

52

lexicology. In this thesis are used semantic features, i.e. features found 

empirically in the semantic analysis, but not meaning components, in the 

sense of a restricted number of apriori primitive features. 

Semantic fields can be characterized, using semantic features, as consisting 

of words (lexemes) with related meanings which have at least one common 

semantic feature in common and are analyzed through meaning 

components or semantic features. Fields can be used to show paradigmatic 

relations between the words (Lehrer, 1974) or syntagmatic relations 

(Porzig, 1934). 

Examples of semantic fields relevant for this study are 'pejoratives' and 

'form'. The aim was not to do a semantic field analysis. Instead fields and 

properties of fields will be used in the discussions, e.g. sometimes the field 

of 'pejoratives' is discussed, sometimes the feature [PEJORATIVE] and 

sometimes words like fjantig (fussy) or fjollig (foolish). In this way 

different authors and different informants who have given information 

related to semantic fields at different levels of abstraction could be included 

in the discussion . 

One aim of this thesis is to give a psychologically valid description and the 

principle assumed here is that the appropriate level is the one that works 

best in the analysis. During data collection the root morpheme level was 

preferred. The meanings of roots were later, in the analysis, decomposed 

into semantic features. 

Since some of the studies concern a number of words with similar (and 

vague) content, within a restricted number of semantic categories, which 

are translated into different languages, it is helpful to work with semantic 

fields, and semantic features. An example: when translating to an unknown 

language, and with informants of varying proficiency in Swedish (or 

English) it is difficult to ask for word to word translations of e.g. words like 

oh, ah, whoops (to take some interjections) but easier to ask for words with 

meaning of positive surprise or negative mood etc. 

53

It can also be the case that an expression, for example for pain, can be 

translated differently in different contexts. For example there is, in Ososo, a 

special expression of pain for when somebody has died. 

Other interjections, like atjo (for a sneeze) are easily translatable word for 

word, e.g. [haptsI] in Hungarian and [Itçími] in Nigerian Ososo. 

2.4 Basic relations between expression and content 

The field of interest which was analyzed here can be defined as 

"synchronously motivated verbal signs". By this definition non-verbal signs 

such as gestures, sign-posts etc are excluded. For practical reasons the 

greater part of prosody such as sentence intonation, phrase intonation etc 

was also excluded. 

Sound symbolism can be divided into three types: 

Free morphemes: 1) onomatopoeic 

2) expressive interjections 

Non-free morphemes: 3) sound symbolic phonesthemes 

(these can be part of 1) and 2)) 

The basic distinctions between these three groups are: 

1) onomatopoeia is mainly iconic, e.g. mjau, plask 

2) expressive interjections are mainly indexical, e.g. aj, atjo 

3) sound symbolic phonesthemes are motivated, but not so clearly iconic 

or indexical, e.g. glänsa, fjanta 

The three groups can be characterized in more detail: 

1) Onomatopoeia is realized in complete free morphemes e.g. mjau, voff. 

Onomatopoeia is a verbal imitation of any sound. The relation between 

expression and content has an element of arbitrariness. The meaning is a 

54

(sort of) sound. The expression side is more or less conventionalized, that 

is, adjusted to the phonology of the language in question. 

The expression has universal traits with a language specific superstructure 

which is affected by the phonology of the individual language. The relation 

between expression and content is basically iconic. (The sound of the 

rooster can serve as an illustration; it is (with informants' transcriptions) in 

Swedish kuckeliku, in German kikeriki, in Serbocroatian kukuricu, 

Macedonian kukurica, Italian chicchiricchi, Syrian kuckylyku, Urdu 

kokelongkong). Figure 2.1 illustrates onomatopoeia: 

sound 

step1: iconic relation 

linguistic form 

(step2: 

linguistic expression) 


Figure 2.1 Schematic illustration 1 of onomatopoeia. 

2) Expressive interjections are also realized in complete free morphemes 

e.g. atjo, aj, oj. The relation between expression and content is nonarbitrary. 

The expression can be described as standing in an indirect causal 

relation to the content which is a bodily or mental reaction. The expression 

is more or less conventionalized and there are universal traits in the 

expressions, naturally with a language specific superstructure. Its base is 

indexical. Figure 2.2 can illustrate: 

1 Linguistic form is a phonetic form associated with a meaning. 

55

emotion, bodily reaction 

step 1: indexical relation 

expression (cry, i.e. 

sound) 

step 2: iconic relation 

linguistic form 

step 3: conventionalization 

linguistic 

expression 

Figure 2.2 Schematic illustration of expressive interjections. 

That is, a bodily reaction produces a sound, e.g a sneeze or a scream. That 

scream etc. is then verbally imitated (like in onomatopoeia) and then 

conventionalized into a word of a language. 

The more vivid the connection between e.g. an emotion and its verbal 

expression is, the stronger the causal component is. But the form of the 

linguistic sign is not an icon of the emotion etc.! The causal relation is 

interesting insofar as the expression could not be just any kind of 

expression. 

Probably many signs are a combination of iconicity, indexicality and 

symbol. Atjo is a good example of a word were the indexical relation (a 

natural relation) is clear. The sound of the sneeze is imitated and then 

conventionalized - the symbol for a sneeze differs in different languages. 

The word aj is not so clearly indexical; the connection with pain is not as 

obvious as the connection between the sneeze and the sound of a sneeze, 

even though the sound for pain can be seen as dependent on the shape of 

the mouth cavity while, e.g. one is screaming from pain. The mouth is 

opened wide, causing an [a]-like sound in all languages (see Darwin, 

1872). 

56

3) Phonesthemes are realised systematically within complete (traditional) 

morphemes, e.g. fl- in flicker, flutter. In sound symbolic phonesthemes the 

relation between expression and content is of a type which is experienced 

as motivated by the typical speaker of a language. (In a more compelling 

degree than the feeling, e.g. by English speakers that horse is a better 

expression than häst, for that four-legged animal.) The expression side is 

language specific, and conventional, but the content side has universal 

traits. 

The meaning of a phonestheme is experienced by the speaker/listener as 

somehow corresponding to the expression. The main problem is to describe 

the type of relation that exists between, e.g. fl- and 'quick or strong 

movement'. What the relation might be, in detail and in relation to iconicity 

and indexicality, is discussed in sections 2.5 and 2.6. 

2.5 The nature of phonesthemes 

First there is the problem with onomatopoeic phonesthemes like fr- in e.g. 

frusta or gn- in e.g. gnägga, gnälla. In the kind of analysis undertaken in 

this chapter these words ought to belong to category 1, onomatopoeic 

words, since they are iconic (even though in a word like gnägga it is 

conventionalized almost beyond recognition). However, I have preferred to 

put the onomatopoeic phonesthemes in category 3, even though they are 

not problematic semiotically, because they are often metaphorically 

connected with other phonesthemes. Gn- which is 'sound', like in (gnissla, 

gnälla) is metaphorically connected with gn- : 'way of talking' (like in 

gnata, gnola, gnälla). Here is again the problem of distinguishing meanings 

(which also holds for other morphology) since 'way of talking' really 

belongs to the same meaning potential as 'sound' does. In this analysis, 

however, 'way of talking' is isolated since it is both frequent and 

interesting. 

The following can be stated and should be considered in a model for 

phonesthemes: 

57

1. Some sounds/sound combinations are (judged to be) better suited for 

some (types of) meanings, within a given language or for many languages. 

2. Some meanings are better suited for being expressed with some of these 

sounds/sound combinations. 

Why? It isn´t plausible that the only explanation is that a number of words 

happen to have similar phonetic structures linked to meanings and that this 

"tie" lately has become productive, which would suggest that 

diachronically the phonesthemes would be arbitrary. In that case the 

meaning dimensions (see chapter 4) would not be so few and they ought to 

be more difficult to relate to each other. The relatable meaning dimensions 

can be structure internal proofs (coherence). 

The expression has universal traits (especially if the meaning of the 

phonestheme can be related to onomatopoeia.) The content of the 

phonesthemes is partly universal and partly language specific because of an 

interaction between e.g. innateness on the one hand and environment on the 

other (see discussion on explanations of sound symbolism in 1.11) 

The phonesthemes that appear from the lexical analysis (chapter 4) are 

preliminary until either there have appeared neologisms that support them, 

the more the better, or until they have found support in tests. 

Figure 2.3 can illustrate the problem of explaining other sound symbolism, 

i.e., phonesthemes which are neither iconic or indexical at first sight. 

meaning 

linguistic expression 

?-relation 

Figure 2.3 Schematic illustration of other sound symbolism. 

58

There is a motivated relation between meaning and form, but the question 

is: How should the relation illustrated by the arrow best be described? 

The analogous representation for a conventional, arbitrary sign would be: 


linguistic expression 

purely conventional relation 

Figure 2.4 Schematic illustration of a conventional, arbitrary sign. 

without an arrow, showing that the form is not motivated by the meaning 

(see figure 2.4). What exactly does then the arrow represent in 

phonesthemes? It can be stated that it represents an ordinary speakers' 

intuition that the form is motivated by the content. In section 1.11 it was 

shown how this intuition can be explained in different ways, and it was 

suggested that the relation is in fact indexical or iconic. It can be concluded 

that most of the explanations discussed in 1.11 support the idea that the 

relation is indexical, iconic or a combination of both. 

2.6. Considerations for a model 

As concerns more specified explanations an eclectic approach to 

explanation of sound symbolism would be to say that most of the 

explanations discussed in 1.11 could be valid, either simultaneously or for 

different types of sound symbolism. 

However, some explanations seem more plausible than others. One basic 

distinction is that between innateness and learning. Innateness can be of at 

least three different kinds. It can mean: 1) that there are innate abilities by 

nature of biological endowment, 2) that there are innate specific 

59

predispositions for the ability, which develops in a certain way over time, 

in interaction with the environment, and 3) that there are innate nonspecific 

predispositions, the development of which heavily relies on the 

environment. Learning is compatible with the last two types of innateness. 

It seems that there are recurring semantic features, which also are easy to 

relate to each other. This suggests that a model of innate predispositions for 

learning certain connections between expression and content is the most 

plausible one. If there were no such predispositions, the semantic 

categories involved in sound symbolism (if there were sound symbolism at 

all) would most probably be haphazard. As we will see in chapter 4 the 

semantic categories in sound symbolism are not unlimited but restricted to 

a number of types. In the analysis of sound symbolic words (the method of 

excerption of words is described in chapter 3) the following categories, 

which are differentiated by their meaning, were found: 

Sound 

Movement 

Light 

Surface structure 

Consistency (plasticity) 

Wetness 

Dryness 

Attitude 

Slang 

Jocular 

Pejorative 

Mental feeling 

Bodily feeling 

Separation 

Putting together (convergence) 

Diminutive 

Augmentative 

Form 

Iterative 

60

It is obvious that some of the categories seem to be of similar types, and 

they can be structured in the following way (figure 2.5): 

61

cognition spatiality repetition 

sense attitude mental bodily structure size form iterative 

impr. feeling feeling 

sound attitude pejorative separation diminutive augmentative 

movement slang putting together 

light jocular (convergence) 

surface 

structure 

consistency 

(plasticity) 

wetness 

dryness 

Figure 2.5 Relations between the recurring semantic categories in Swedish sound symbolism.

Most of the categories can also be described as related to sense impressions 

in a more or less direct way. The categories are either related to one sense 

or to several senses. The simpler cases are 'sound', related to hearing, 'light' 

and 'movement' related to seeing and 'consistency' related to touch. The 

experience of 'surface structure', 'wetness' or 'dryness' is a combination of 

tactile, auditive and visual experience. 'Form' involves the senses seeing 

and touching. 'Separation' and 'putting together' have a more abstract 

relation to the expression. 'Diminutive' and 'augmentative' are also more 

abstract (but can be explained with reference to the frequency code, cf. 

discussion in 1.10.5.) as well as 'iterative' (but it is clearly iconic). Finally 

there are the expressive categories of mental and bodily emotion and 

attitude which are cognitive categories. 

The categories are not mutually exclusive. E.g. 'cognition' and 'spatiality' 

could be combined e.g. in prepositions like in and other basic relations. 

'Mental feeling' and 'bodily feeling' can combine. However, for the analysis 

of significant properties of phonesthemes the above categorisation is 

useful. 

The categories are identical to or comparable with the semantic categories 

listed in the end of 1.10.4, which were based on the studies of different 

linguists. The only really new classes are 'separation' (e.g. spl- splinta, 

split, splits, splitter, spr- spreta, sprida) and 'putting together' (e.g. knknipa, 

knyta, knyckla, tv- tving, tvinna). 

2.6.1 Relations between the categories 

The onomatopoeic free morphemes or phonesthemes are in a special class 

since the iconicity concerns language sounds imitating sounds. Almost all 

the initial consonant clusters correspond to an onomatopoeic phonestheme. 

This means that this category is co-occurring with many of the other 

categories, e.g. many consonant clusters can be both 'onomatopoeic' and 

'pejorative', however not necessarily in the same root morpheme (see 

further discussion in chapter 4). 

63

One semantic category, 'light', can be classified as a metaphorical extension 

from sound phenomena. The basis for extension is sense analogy - an 

iconic phenomenon based on some sort of similarity, e.g. the word warm 

can be used not only for temperature but also for colour ('light') and tones 

('sound') i. e. using phonesthemes of one sensory modality for words of 

another sensory modality. (cf Abelin 1988). (This is not the same as 

synesthesia - a neurological, indexical phenomenon, e.g. the experience of 

hearing a certain tone because of tasting something that is salty, cf. 1.11.2). 

Appealing to sense analogy we can explain the category of 'light'. Sense 

analogy concerns conventional relations between words. These relations 

are there because we see some sort of similarity between a particular type 

of sound and a particular type of light. Underlying sense analogy (as well 

as synaesthesia and sound symbolism) could be "stable intermodal 

connections". An example of sense analogy with sound as a point of 

departure is gn-, as in gnälla (squeak), which can be seen as extended to 

gn- in gnistra (glimmer) in a fashion similar to the meaning extension of 

e.g. the words skrikig (loud), gäll (shrill), dov (dull) which originally 

described sounds but later also colours. 

Many of the remaining categories can be explained satisfactorily with the 

metaphoric and metonymical model of Lakoff and Johnson (1980) and 

Lakoff (1987). Sound symbolism can be seen as depending on 

1) innate capacities for metaphor (similarity, icon), especially 

personification, e.g. gr- 'negative mood' and for metonymy (contiguity, 

index) i. e. closeness in time and space, e.g. gl- denotes both 'light' and 

'form letting through light'). 

and on 

2) the result of learning from one's senses through use of the innate 

capacities, (e. g. how something sounds when moving in a certain way) 

where the innate capacity for metaphor is put to use. 

This model seems to incorporate the ideas of both Cytowic and Brown, as 

described above. 

64

Lakoff (1987, p. 8) writes "human categorization is essentially a matter of 

both human experience and imagination - of perception, motor activity and 

culture on the one hand, and of metaphor, metonymy and mental imagery 

on the other." This would also explain why sound symbolism is partly 

universal, partly language specific. Universality from the semantic 

perspective depends among other things on human capacity for metaphor 

and metonymy; experience is shaped by metaphor and metonymy. 

Lakoff and Johnsson (1980) write: "the most obvious ontological 

metaphors are those where the physical object is further specified as being 

a person." How does this apply to sound symbolism? Sounds of language 

naturally cannot be seen as persons, but they can be seen (heard) as 

belonging to persons of a certain type, mood etc. Doing this is a kind of 

personification, which is a basic conceptual strategy. The material this 

personification is working on is based on our experience of how people 

sound when of a certain type, in a certain mood etc. This personification 

could explain the 'negative mood' applying to persons. 

The categories Movement, Form, Consistency, Surface structure, Wetness, 

Dryness and Diminutive are simply explained through metonymy 

(closeness in time and space) between sound and manipulation of objects 

during the language learning process. Already Brown (1958) said that what 

was mysteriously called physiognomic (by Werner, 1953) is a result and 

consequence of a simple act of learning. In everyday experience there is a 

correlation between certain physical attributes of objects and the noises 

produced by these objects. A large object is more likely to produce a deep 

sound than a small object when it is pushed or when it falls. The 

relationship between size and deep sounds can therefore be learned; equally 

a connection between sharpness and high frequency and between roundness 

or bluntness and a low frequency, cf. however Ohala (1994). 

Werner (1953) introduced the concept of physiognomy. According to him, 

in the primitive perception of animals, children etc. things are not treated 

objectively, but physiognomically, i. e. as if they expressed an inner life 

and had a mind; a landscape is cheerful, a cup on its side tired etc. Werner, 

in contrast to Lakoff and Johnsson (1980) treated this as an exception. 

65

Werner and Kaplan (1963) developed a genetic theory for symbol 

formation according to which objects are expressive, e. g. two headlights 

look like eyes, music can be threatening. This transcendence of expressive 

qualities does not only form the basis for analogies, metaphors and similes, 

it is also the basis for seeing similarities in such unrelated things as rounded 

shapes in a drawing and the word "maluma". 

Lakoff and Johnsson's theory is in agreement with both Brown, Werner and 

Kaplan, and Cytowic (i. e. sound symbolism is formed by experience of the 

world (learning) (cf. Brown, 1958) but it is also formed by predispositions 

of perception (Werner and Kaplan, 1963, Cytowic, 1989). This points to 

the conclusion that sound symbolism works in much the same way as other 

concepts except that much more is dependent on iconicity and indexicality. 

2.6.2 An explanatory model for sound symbolism 

From the considerations discussed above a number of dimensions for a 

model of sound symbolism can be extracted. The general background to the 

model is the framework presented in 1.9, e.g. sound symbolism is central to 

language, it can be innate or conventional, it is productive and context 

sensitive. The present model tries to explain the nature of the relation 

between meaning and expression, "the arrow with a question-mark". Some 

dimensions from the model in 1.9 then appear again, e.g. innateness vs. 

learning (the latter connected with conventionality). The dimensions are the 

following: 

1. innateness 

2. learning 

3. icon (metaphor) 

4. index (metonymy) 

5. conventionalization 

6. the sound symbolic categories: 

Sound 

Movement 

Light 

Surface structure 

66

Consistency (plasticity) 

Wetness 

Dryness 

Attitude 

Slang 

Jocular 

Pejorative 

Mental feeling 

Bodily feeling 

Separation 

Putting together (convergence) 

Diminutive 

Augmentative 

Form 

Iterative 

The sound symbolic categories concern Swedish since they build on the 

analysis of the Swedish lexical material. In a description of another 

language they might be slightly different but they would be explainable in 

terms of icon and index and mostly relatable to sense impressions or 

emotions. 

innateness 

(icon,index) 

learning 

sound 

symbolism 

Figure 2.6 An explanatory model for sound symbolism, which shows the 

cause of the motivated relation between expression and meaning. 

67 


phonesthemes 

for 

onomatopoeia 

pejoratives, 

surface 

structure, 

movement, 

form, etc.

The model in figure 2.6 tries to capture the relations between the different 

concepts that are fundamental to sound symbolism. It says that the relation 

between expression and content in sound symbolism can be caused by 

either innateness or learning. What is predisposed can be affected by 

learning, and what is learned is depending on innate categories for thinking, 

such as index (metonymy) or icon (metaphor). Sound symbolism is the 

result of both innate capacities and learning. Some types of sound 

symbolism depend more on an indexical relation (connected with learning) 

e.g. phonesthemes for 'movement', 'form' etc. while other types of sound 

symbolism depend more on an iconic relation, e.g. phonesthemes for 

'onomatopoeia' and perhaps 'pejoratives', but these have nevertheless been 

put in the same box. The model also shows that sound symbolic expression, 

and maybe also conent, is always conventionalized, to a greater or lesser 

degree. This explains the fact that e.g. words for a certain animal call can 

differ between languages at the same times as it is motivated. The model 

can be seen as a static model where the processes to the left are necessary 

prerequisites for the existence of phonesthemes. It could also be seen as a 

basis for how sound symbolic expressions have emerged in individuals in 

interaction within a language system, with the result of sound symbolic 

expressions in that particular language. 

The concepts iconicity and indexicality can be used for describing the 

nature of the different explanations of the relation between expression and 

form in sound symbolism described above. The concepts iconicity and 

indexicality (superordinate to metaphor and metonymy) can also be used 

for describing the semantic relations between the semantic features 

reoccurring in sound symbolism (cf. 2.6) The relations between some of the 

semantic categories will be further specified in 4.6. Figure 2.6 An 

explanatory model for sound symbolism, which shows the cause of the 

motivated relation between expression and meaning. 

The model in figure 2.6 tries to capture the relations between the different 

concepts that are fundamental to sound symbolism. It says that the relation 

between expression and content in sound symbolism can be caused by 

either innateness or learning. What is predisposed can be affected by 

68

learning, and what is learned is depending on innate categories for thinking, 

such as index (metonymy) or icon (metaphor). Sound symbolism is the 

result of both innate capacities and learning. Some types of sound 

symbolism depend more on an indexical relation (connected with learning) 

e.g. phonesthemes for 'movement', 'form' etc. while other types of sound 

symbolism depend more on an iconic relation, e.g. phonesthemes for 

'onomatopoeia' and perhaps 'pejoratives', but these have nevertheless been 

put in the same box. The model also shows that sound symbolic expression, 

and maybe also conent, is always conventionalized, to a greater or lesser 

degree. This explains the fact that e.g. words for a certain animal call can 

differ between languages at the same times as it is motivated. The model 

can be seen as a static model where the processes to the left are necessary 

prerequisites for the existence of phonesthemes. It could also be seen as a 

basis for how sound symbolic expressions have emerged in individuals in 

interaction within a language system, with the result of sound symbolic 

expressions in that particular language. 

The concepts iconicity and indexicality can be used for describing the 

nature of the different explanations of the relation between expression and 

form in sound symbolism described above. The concepts iconicity and 

indexicality (superordinate to metaphor and metonymy) can also be used 

for describing the semantic relations between the semantic features 

reoccurring in sound symbolism (cf. 2.6) The relations between some of the 

semantic categories will be further specified in 4.6. 

69

3 Method 

The method used in this investigation is a combination of intuition and 

empirical studies (experiments). Using intuition about language is an 

indispensable part of all linguistic research, as the researcher cannot avoid 

having some knowledge of the subject. However, intuition is not 

sufficient, and will therefore be supported by empirical studies. 

This investigation of sound symbolic words in language is concentrated on 

initial and final consonant clusters in Swedish. It consists of three main 

stages. 

1. A number of Swedish dictionaries were excerpted with the purpose of 

establishing tentative phonesthemes for Swedish. The material collected 

was stored in spreadsheats, which could be manually searched for roots, 

key words and semantic features. 

2. Stage 2, which is a smaller part of the study, consists of partial 

excerptions from other languages. The excerptions were done using 

meanings, or, in some cases, phonesthemes as the point of departure. 

3. A number of tests were performed in order to study the tentative 

Swedish phonesthemes found in stage 1. 

Below follows an elaboration of the three stages. In each chapter there is a 

more detailed description of the method. 

3.1 Stage 1: Collection of lexical material 

Initial consonant clusters 

Since the study has its point of departure in the Swedish language it was 

initiated by an excerption of sound symbolic words in Swedish. The 

lexical material consists of excerptions from Svenska Akademiens Ordlista 

(SAOL 10, The Wordlist of the Swedish Academy), a word list with some 

semantic information, and Svensk ordbok (SOB, Dictionary of Swedish), 

a dictionary with definitions, exemplifications etc. 

71

The excerption of words from SOB was done in the following way. To 

qualify as a candidate for being a possibly interesting word, one or two of 

the following criteria should be met: 

(i) The word is clearly onomatopoeic. 

(ii) At least two 1 word roots with the same consonant sequence and 

similar meaning can be found in the lexical material, where the likeness 

does not come from trivial morphological relatedness like derivation of 

e.g. nouns or adjectives. As an example, fjant, fjanta, fjantig are counted 

as forms of the same root. Kladd, kladda och kladdig are another 

example. In other words: compounds, derivations or words belonging to 

different parts of speech are seen as different instances of one root 

morpheme (containing a phonestheme). 

Through the method described above there is no absolute guarantee that 

all possible phonesthemes for Swedish are found. Those that are found are 

chosen with approximately the same criteria. The emphasis has been on an 

analysis of all words, rather than a detailed analysis of a few words. The 

purpose of this is to get an overview of the main traits in the Swedish 

lexical material. 

The Swedish lexicon (represented by SOB) was manually excerpted for 

root morphemes with onomatopoeic or other sound symbolic meanings, 

(motivated words.) Only words beginning with consonant clusters were 

considered in this excerption. 

"Key words" from the SOB were also registered in the excerption. A key 

word is a formal representative of the sense connected to a root 

morpheme and usually denotes e.g. 'sound', 'shape', 'texture' (sense 

related categories), 'pejorative' etc. 

A key word is either a word in the definition of the actual word, or a 

word in the definition of a word which is in the definition of the actual 

word, or another word in the paraphrase of a more peripheral mening of 

the actual word, or a synonym. Ideally the key word should have been in 

the definition of the actual word, but unfortunately SOB is not completely 

1 For lexically infrequent clusters, with less than 13 roots, 1 root sufficed if it had one of 

the recurring semantic features. 

72

consistent in its way of giving definitions. Sometimes a word is only given 

a very short definition, e.g. a synonym. 

The key word is always a word which is written into the definition of 

the lexicon entry of the word under consideration and, in a most cases, a 

word in the definition of the actual word. 

This function of key words is to ensure a certain amount of objectivity, 

through consensus, in the selection procedure. It depends then on earlier 

semantic analyses by several persons. Key words are sound symbolic 

categories or hyponyms to these. They can be found in Appendix 1. 

Examples of key words from SOB are: ljud (sound), rörelse (movement), 

spetsig (pointed), ljus (light), äcklig (nasty), vatten (water), tjock (thick), 

klibbig, (sticky), ohyfsad (rude), löjlig (ridiculous). Naturally there were 

expectations as to what semantic categories would be interesting; the key 

words are related to the senses, to 'form', 'mental feeling' etc. Words 

with definitions containing appropriate key words were, however, not 

considered if a connection between expression and meaning was not felt, 

using linguistic intuition, as e.g. for skam. No part of the expression side 

of this word seems to mirror the words oförmåga or förkastlig of the 

definition. The opposite is also the case; a connection seemed to be present 

but an appropriate key word was not to be found in the definition. 2 

Some words which can be found in SAOL 10 are described as colloquial, 

while they in fact are out of date. Nevertheless they are not difficult to get 

a vague understanding of. Examples are words like pjask, pjasker and 

pjalt 3 that have a clearly pejorative ring. These and the words of the 

previous paragraph could probably easily be (mis)interpreted in an 

appropriate context! 

With an attitude of accepting doubtful words rather than discarding them, 

about 1,000 words – in the sense of root morphemes – were registered 

for further analysis. According to Nusvensk frekvensordbok 4 (Allén et 

2 There are also words that, from the point of view of sound, would seem to be 

appropriate for mirroring a certain content but do not. Examples are fläns (part of a 

machine), skralna (to begin to head, of wind; sailing term), and slubb (sort of spinning 

machine). 

3 In SAOL 12 these words are no longer included. 

73

al, 1980) the number of lexical morphemes in Swedish amount to around 

8,300 (suffixes add another 562 morphemes). 

It is important to keep in mind, both what concerns the data collection and 

the data analysis, that the borders between the different meanings are 

fuzzy and partly subjective, and the resulting diagrams and tables are to 

be seen as points of departure for different investigations, experiments 

and discussions. The primary aim of the data collection is to get an 

overview of the structure of the clusters and to show patterns of forms 

and contents. 

Final clusters 

Final consonant combinations with recurring meanings were also 

excerpted from Svensk baklängesordbok (1981) (Reverse Order 

Dictionary of Swedish) and Nusvensk frekvensordbok (Allén et al, 1980) 

(Frequency Dictionary of Present-Day Swedish). They were analyzed in a 

fashion similar to the one described above. 

3.2 Stage 2: Cross-language comparisons 

With a focus on both the expression and the content side of phonesthemes, 

some comparisons with other languages were done. The emphasis was on 

certain easily defined meanings. Also, some contrastive experiments of 

interpretation, testing Swedish phonesthemes on native speakers of other 

languages, were made. 

3.2.1 Cross-language thesaurus studies 

The lexical study consists of two studies, one in Swedish and one in 

English. It is concentrated on the semantic fields of 'stupidity' and 

'surface structure' – 'rough' and 'smooth'. These fields were chosen 

because they were different in type, they were quite common ('stupidity' 

is a subclass of 'pejorative') and also quite uncomplicated semantically. 

There were three informants from each language. 

3.2.2 Cross-language informant studies 

Self-imitative interjections 

The Swedish primary interjections (see Ideforss, 1928) have been 

translated to 8 different languages with the help of informants. The 

languages are Icelandic, English, Polish, Hungarian, Finnish, Ososo, 

74

Malagasi and Slovenian. The material is presented in Appendix 2 and it is 

grouped according to type of interjection. In some cases it was difficult to 

translate word for word. The semantic contents of an interjection instead 

had to be described (with semantic features) after which the informant 

gave the closest correspondence in her own language. 

Interjections imitating animal sounds 

Swedish interjections imitating animal calls were translated, with the help 

of informants, into 17 different languages: Icelandic, English, Polish, 

Hungarian, Finnish, Ososo, Malagasi, Slovenian, Korean, Japanese, 

Chinese, Estonian, Urdu, Persian, Kurdish, Arabic, and Spanish. 

Interpretation of animal sounds 

Sixteen of the interjections imitating animal calls from different languages 

were chosen to test how speakers of other languages would interpret 

them. The ones chosen were those whose expression sides were most 

different between the languages studied, in order to make the task as 

difficult as possible. The subjects in this test were 15 persons, with the 

following first languages: Swedish (8 subjects) French (2 subjects), 

English, Hungarian, Czech, Slovenian, Ososo. All were tested on the same 

occasion. They listened to pronunciations of the words for the animal calls 

and they saw them transcribed in IPA. They were told to guess which 

animal had gotten its call conventionalized in each way, and to write down 

their answers. 

Cross-language experiment 

Furthermore, one person representing each of the languages: Arabic, 

Spanish, German, Dutch, Ibo and English, took part in an experimental 

study of persons who did not have Swedish as a first language. The study 

was conducted in much the same way as one of the experiments described 

below and in 7.1, (test 2.a) – a free choice test from expression to 

meaning. The main difference was that the subjects were confronted 

mostly with real words. There were, however, also a few neologisms 

based on phonesthemes. 

75

3.3 Stage 3: Experiments 

3.3.1 Experiments with neologisms 

To further penetrate the status and productivity of the tentative Swedish 

phonesthemes of language users of today, a number of tests were 

performed (see Appendix 3). 

The tests investigated both production and understanding of written 

neologisms, which were, in three of the tests, modelled on the tentative 

phonesthemes. The tests consisted of free and forced choice tasks. There 

was was an additional matching test where two neologisms were to be 

matched with two meanings. The subjects were 14–15 native Swedish 

speakers. In one test they were instructed to freely interpret neologisms 

and in another they were supposed to produce neologisms out of different 

concepts written on the test sheet. In the next set of tests the subjects 

should, in interpreting neologisms, chose from three alternative meanings 

written on the test sheet, or, in producing neologisms, choose among three 

alternative neologisms, also written on the test sheet. The design of the 

experiments is described in detail in 7.1. 

3.4 Further method 

Explanatory models for sound symbolism are also constructed, see 1.9, 

2.6.2. These models aim at explaining the nature of the motivated 

connection between expression and meaning in sound symbolism. 

Comparisons between the results of the lexical studies, the experiments – 

also the cross-language experiments – some of the cross language 

comparisons 4 ,and the models are made in chapter 8. 

4 Interjections and animal sounds of the cross language comparisons belong to the classes 

expressive interjections and onomatopoeia which are not included in the models of 1.9 

and 2.6.2. 

76

4 Analysis of the initial consonant clusters 

4.1. Data analysis, a short overview 

All the entries of the lexicon Svensk Ordbok (1986) were examined 

(around 65, 000 lexemes, which corresponds to approximately 8, 300 1 

morphemes, which is the number of morphemes in Swedish according to 

Allén et al, 1980). Around 1,250 words (root morphemes, cf. 3.1) were 

judged to be possibly motivated. 

Of all the 37 initial consonant clusters of Swedish all but one are used for 

sound symbolism. The 37 clusters are: bj-, bl-, br-, dr-, dv-, fj-, fl-, 

fn-, fr-, gl-, gn-, gr-, kl-, kn-, kr-, kv-, mj-, nj-, pj-, pl-, pr-, 

sk-, skr-, skv-, sl-, sm-, sn-, sp-, spj-, spl-, spr-, st-, str-, sv-, 

tr-, tv-, vr-. The unused one (dv- ) has a lexical frequency of 3 

morphemes. Thus, almost all clusters seem to be used, but they are used 

to varying extents and for different purposes. 

(Now and then one hears authentic examples of e.g. onomatopoeic 

expressions which use non-standard clusters (cf. Garlén (1988) like 

sklofsa ("walk in mud"). It is also the case that a cluster like spl-, which 

normally is not onomatopoeic in Swedish, is often used for imitating 

sounds, for example in comic strips – perhaps influenced by English. In 

comic strips expressions like splofs and splafs are common.) 

All initial consonant clusters were analyzed closely, both those that appear 

to have a greater and those that seem to have a smaller amount of 

motivated root morphemes (cf. 3.1). 

For each cluster, each motivated root morpheme was classified according 

to its motivated semantic features. (cf. 2.6). A root morpheme can have 

one or more motivated semantic features, e.g. skvalpa (lap, ripple, splash, 

spill) has 'sound', 'wetness'. Even though part of the meaning might also 

be due to -alpa and not only to skv-, such possibilities were not 

1This number of morphemes is the result of an investigation of a one million word 

newspaper corpus (NFO 4). It is most probable that newspaper language contains fewer 

roots than spoken language, but these figures are not known at present. The estimation of 

65, 000 lexemes is excluding transparent compounds. 

77

considered at this stage. 2 The semantic features 'wetness' and 'sound' of 

skvalpa were, in this analysis, attributed to the cluster skv- because there 

are two or more root morphemes beginning with skv- that have the same 

semantic features. The SOB-definition of skvalpa is: "vara i rörelse och 

därvid ge ifrån sig ett kluckande och plaskande ljud – om vatten o.d."(my 

italics) ("to be moving and thereby emit a gurgling and splashing sound – 

of water, etc") 

The SOB-definitions of other skv- lexemes are: 

skvimpa (splash to and fro) "skvalpa med små rörelser – ofta så att vätska 

spills ut" (The semantic features are the same as for skvalpa (lap with 

small movements – often so that liquid is spilled out), with the addition of 

'diminutive' – which is probably due to the i, cf. Ultan, 1978.) 

skvätta (splash, squirt) "fara i väg i skvättar" (go off in squirts), skvätt: 

"liten mängd vätska" (small amount of liquid) The semantic features are 

'movement', 'diminutive', 'wetness'. 

skvala (pour, gush, rush) "rinna rikligt och ljudligt (flow abundantly and 

noisily). Rinna (flow) is defined as "förflytta sig nedåt (längs viss yta) i 

sammanhängande formlös mängd – om vätska e.d." (move downwards 

(along a certain surface) in a continuous shapeless quantity – about liquid, 

etc) and riklig (abundant) is defined as "förekommande i stor mängd eller 

omfattning" (occurring in large quantity or range) and so the semantic 

features of skvala are 'movement', 'wetness', 'sound' and 'augmentative'. 

As we see the features 'movement', 'wetness' and 'sound' are all present 

in three of these four words. There are around eight more words 

(considering the root morphemes) beginning with skv-. 

(The cluster -Vlpa ends the following words (considering the root 

morphemes): valpa (whelp), skvalpa (lap), stolpa (walk with long paces), 

pulpa (pulp), hjälpa (help), stjälpa (tip over), skölpa (hollow out with a 

special tool) – no other root morphemes with the features 'wetness' or 

'sound', but 4 root morphemes with the feature 'movement'.) 

2 For an analysis of combined effects of initial cluster, final clusters and, to some extent, 

of vowels, see 5.6. 

78

Diagram 4.1 shows the number of motivated root morphemes for every 

cluster. The number of all root morphemes and of the motivated root 

morphemes for each cluster can be studied in diagrams 4.2 and 4.3. Also 

the ratios between the number of sound symbolic root morphemes on the 

one hand (irrespective of exact meaning) and total number of root 

morphemes beginning with a certain cluster was calculated, see diagram 

4.4. 

In addition to this, the various semantic features (cf. 4.2.3) for every 

cluster and the ratios for the occurrence of different semantic features per 

total number of root morphemes (for every cluster) were put into the 

chart. These figures are used for the following diagrams. 3 

Diagram 4.5 shows the absolute numbers of all initial consonant clusters 

for all semantic features. 

In diagrams 4.6 to 4.15 the distributions of separate semantic features 

over different clusters can be seen. 

Diagrams 4.16 to 4.23 show the sound symbolic profiles of different 

clusters, i.e. how the semantic features are differently distributed. 

In addition to this, the relations between the different types of meaning 

within a cluster were analyzed synchronically. 

4.2 Results 

In this section some of the results from the lexical analysis are presented. 

The same material will be studied from different angles and in greater or 

lesser detail. The focus will be either on the different consonant clusters, 

or on the different meanings. There will be general patterns and patterns 

for certain clusters or meanings. 

3 If all dialectal, archaic and slang words had been counted the ratios of motivated root 

morphemes would have been higher for most consonant clusters. (This reflects the 

phenomenon of slang words floating in and out of language at another rate, cf. 1.3 - 1.4). 

As an example the following root morphemes, found in Svenska Akademiens ordlista 

över svenska språket (SAOL 10), are not included: skvattra "snattra"(quack, gabble, 

chatter), skrålla (coll.) "löjlig damhatt" (silly lady's hat), skryp (dial.) "slösaktig" 

(wasteful). The main reason for not counting these is that they are not covered in SOB; 

special studies would have to be made of dialectal and slang dictionaries. There seem to 

be more dialectal and archaic words in SAOL than in SOB. 

79

4.2.1 More and less sound symbolic clusters 

The 36 initial consonant clusters are very different with respect to how 

many root morphemes they contain. They also differ considerably in how 

many sound symbolic root morphemes there are for each cluster. 

Diagram 4.1 shows the number of motivated root morphemes for all 36 

clusters. 

Diagram 4.1 shows e.g. that sl- has the highest number of sound symbolic 

clusters, namely 83, while nj- has the fewest, namely only 1. In absolute 

numbers, the 6 clusters that have most motivated root morphemes are (in 

order of descending frequency): sl-, sn-, kn-, kr-, kl-, sp-. Of these 6 

clusters 3 begin with an s and 3 begin with a k. There is reason to wonder 

if nj- should be counted as a sound symbolic cluster at all, but nj-, as well 

as all other clusters, must be seen in the light of how many root 

morphemes there are in the total vocabulary for each cluster. This 

comparison is done in diagram 4.2. 

80

sl 

sn 

kn 

kr 

kl 

sp 

st 

gl 

tr 

fl 

skr 

pl 

bl 

sm 

str 

gr 

sk 

sv 

spr 

pr 

gn 

kv 

dr 

vr 

br 

fr 

fn 

skv 

mj 

fj 

pj 

spl 

bj 

spj 

tv 

nj 

0 20 40 60 80 100 

Diagram. 4.1. More and less sound symbolic initial consonant clusters. Number of 

motivated root morphemes per cluster. 

81

st 

pr 

tr 

kr 

sl 

sp 

br 

sn 

gr 

sk 

kl 

pl 

fl 

fr 

str 

kn 

dr 

gl 

sv 

bl 

skr 

kv 

sm 

spr 

fj 

gn 

tv 

vr 

bj 

mj 

skv 

fn 

spl 

pj 

spj 

nj 

0 50 100 150 200 250 300 

82 

no of motiv roots 

all roots 

Diagram. 4.2 More and less sound symbolic consonant clusters. 

Number of root morphemes and number of motivated root 

morphemes per cluster.

Diagram 4.2 shows the proportions between total number of root 

morphemes and the number of motivated root morphemes. The diagram 

shows for example that the cluster pj- contains both few root morphemes 

and few motivated root morphemes. So there is a high proportion of 

motivated ones. It also shows that tr- has many root morphemes, quite a 

few motivated root morphemes, but a lower proportion of motivated root 

morphemes. For percentages of the proportion motivated root 

morphemes/all root morphemes, see diagram 4.4 (It should be noted that 

one motivated root morpheme may contain one or more motivated 

semantic feature.) 

Diagram. 4.3 shows the same information as diagram 4.2. but here the 

clusters are ranked from the most sound symbolic to the least sound 

symbolic, as in 4.1. The diagram shows that lexically frequent clusters do 

not necessarily contain a large proportion of sound symbolic morphemes, 

e.g. pr-. (A large proportion of the non-motivated morphemes beginning 

with pr- are due to loan words 4 and it is also quite difficult to count the 

morphemes.) 

Table 4.1 (related to diagrams 4.1, 4.2 and 4.3) shows the number of root 

morphemes and the number of motivated root morphemes for the 36 

clusters. The table is ordered alphabetically. The clusters are ranked from 

the most sound symbolic (in absolute numbers) to the least sound 

symbolic. 

4 It would be interesting, in further research, to study the influence of loan words on the 

frequencies of sound symbolic root morphemes per cluster. Words older than a 100 

years, etc. could easily be found with the aid of the machine readable version of SAOB 

(Ordbok över svenska språket, utgiven av Svenska Akademien). It could be hypothesized 

that there would be a higher frequency of sound symbolic root morphemes among the 

older root morphemes. On the other hand, root morphemes that confirm with the native 

pattern could be preferred in loans, cf. discussion in 1.5.) 

83

sl 

sn 

kn 

kr 

kl 

sp 

gl 

st 

tr 

fl 

skr 

pl 

bl 

sm 

str 

gr 

sk 

spr 

sv 

pr 

gn 

kv 

dr 

vr 

br 

fn 

fr 

skv 

mj 

fj 

pj 

spl 

bj 

spj 

tv 

nj 

0 50 100 150 200 250 300 

84 

no of motiv roots 

all roots 

Diagram 4.3. Number of root morphemes and number of motivated root 

morphemes per cluster, which are ordered from highest number of sound 

symbolic root morphemes to lowest number of sound symbolic root 

morphemes.

Table 4.1 Number of root morphemes and motivated root morphemes. 

total number of motivated root Rank 

root morphemes morphemes 

bj- 13 5 32 

bl- 64 33 13 

br- 130 11 25 

dr- 74 16 23 

fj- 24 7 29 

fl- 96 36 10 

fn- 10 10 26 

fr- 87 10 26 

gl- 69 45 7 

gn- 22 20 21 

gr- 126 26 16 

kl- 115 51 5 

kn- 77 59 3 

kr- 169 53 4 

kv- 56 17 22 

mj- 11 7 29 

nj- 3 1 36 

pj- 7 6 31 

pl- 100 34 12 

pr- 200 21 20 

sk- 116 25 17 

skr- 63 35 11 

skv- 10 9 28 

sl- 150 83 1 

sm- 55 31 14 

sn- 126 62 2 

sp- 137 47 6 

spj- 6 4 34 

spl- 9 5 32 

spr- 34 23 18 

st- 279 45 7 

str- 77 30 15 

sv- 65 23 18 

tr- 169 44 9 

tv- 20 3 35 

vr- 19 12 23 

85

Table 4.1 shows, for example, that 5 out of 13 root morphemes beginning 

with bj- have a sound symbolic meaning. The remaining 8 (13-5) 

unmotivated root morphemes beginning with bj- are: 

bjuda (invite, offer, etc) 

bjugg (barley) 

bjussa (slang for bjuda) 

bjälke (balk) 

björk (birch) 

björkna (a fish) 

björn (bear) 

björna (demand repayment) 

and the 5 motivated root morphemes are: 

Key word 

bjäbba uppnosigt prat (cheeky talk) 

bjäfs överdriven (exaggerated) 

bjällra klingande (chiming bell) 

bjärt lysande (shining) 

bjässe mycket stor (very big) 

(Only key words, not the complete definitions, are given above.) 

4.2.2 Proportions of motivated root morphemes. 

Summary 

To sum up part of the foregoing comparisons, diagram 4.4 shows the 

proportions between number of sound symbolic morphemes and total 

number of morphemes in percentages. 

86

fn 

gn 

skv 

pj 

kn 

spr 

spj 

gl 

mj 

vr 

spl 

sm 

skr 

sl 

bl 

sn 

kl 

str 

fl 

bj 

sv 

sp 

pl 

nj 

kr 

kv 

fj 

tr 

sk 

dr 

gr 

st 

tv 

pr 

fr 

br 

0 20 40 60 80 100 

Diagram 4.4. Percentage of motivated root morphemes per cluster. 

87 

%

Diagram 4.4. shows the proportion of motivated root morphemes in 

percentage of the total number of root morphemes per individual cluster. 

The root morphemes are the same as in diagrams 4.1–4.3. We can see that 

the clusters fn-, gn-, skv-, pj-, kn-, spr-, spj-, gl- mj- and vr- have a 

greater ratio of motivated root morphemes, well over 60%. The cluster 

fnis the most sound symbolic, at 100 %. These clusters are all lexically 

infrequent clusters with the exception for kn- and gl- which are 

comparatively larger, (cf. diagrams 4.2 and 4.3). It is clear that lexically 

infrequent clusters are exploited sound symbolically to a higher degree 

than lexically frequent clusters are (cf. the discussion in 8.3). The clusters 

br-, pr-, and fr- have the lowest proportion of sound symbolic root 

morphemes. Of these clusters pr- is very frequent lexically and br- is 

quite frequent too (cf. diagrams 4.2, 4.3). This phenomenon is however 

not symmetrical; there are lexically very frequent clusters like sl- which 

are also sound symbolic to quite a high degree (55%). An interesting 

result is that the two most sound symbolic clusters, fn- and gn-, end with 

n while the three least sound symbolic clusters, pr-, fr- and br-, all end 

with an r. 

4.2.3 Types of meaning 

After calculating the ratios of motivated root morphemes per cluster, an 

analysis was made of the semantic features appearing in each cluster. The 

analysis can be done in detail or more abstractly. The analysis presented 

in this chapter is quite detailed, with more specific semantic features. For 

example there are the categories 'narrow form', round form', 'thin 

form', hollow form', 'short wide form', 'crooked form', 'long thin 

form', 'winding form' and 'small end form'. On a more abstract level 

these can of course be classified as 'form'. The same goes for e.g. 'surface 

structure' which is further analyzed into 'rough surface structure', 

'smooth surface structure', 'hardness'. Some categories are broader than 

their names imply. This should be noted for 'pejorative', which includes 

something more generally negative and 'talking' which includes all sounds 

made by humans. 

The categories of diagram. 4.5 are related to those presented in chapter 2 

in the manner listed below. They are not mutually exclusive, since it was 

judged to be interesting to count certain frequent special cases of 

meaning. Therefore e.g. 'mental feeling' is subdivided into 'bad mood' 

and 'other mental feeling'. However, in the following discussions and 

88

diagrams 'other mental feeling', etc is simply called 'mental feeling' 5. The 

feature 'slang' is a stylistic feature and in fact adds another dimension. 

Sound: Sound 

Talking 

Beat 

Movement: Movement 

Quick or strong movement 

Walking 

Falling 

Potential movement 

Quickness 

Light: Light 

Gaze 

Surface structure: Rough surface structure 

Smooth surface structure 

Consistency: Soft consistency 

(Plasticity) Hardness 

Slackness 

Stiffness 

Wetness: Wetness 

Adhesion 

Dryness: Dryness 

Attitude: Attitude 

Secrecy 

Slang: Slang 

Jocular: Jocular 

Pejorative: Pejorative 

Destruction 

Mental feeling: Mental feeling 

Bad mood 

Bodily feeling: Bodily feeling 

Suffocation 

Separation: Separation 

5 Other such cases are 'other movement' which is called 'movement', 'light emission' 

which is called 'light', 'other wetness' which is called 'wetness', 'other attitude' which is 

called 'attitude', 'other mental feeling' which is called 'mental feeling', 'other bodily 

feeling' which is called 'bodily feeling', 'other form' which is called 'form', and 'other 

iterative' which is called 'iterative'. 

89

Putting together Putting together 

Diminutive: Diminutive 

Augmentative: Augmentative 

Form: Form 

Round form 

Short-wide form 

Thin form 

Hollow form 

Winding form 

Long thin form 

Narrow form 

Small end form 

Bent form 

Iterative: Iterative 

Fine grain 

The most frequent meanings (shown in table 4.2) are, in descending 

order: 

Table 4.2 The most frequent semantic features, in descending order. 

semantic freq semantic freq semantic freq 

feature 

feature 

feature 

'pejorative' 163 'gaze' 15 'attitude' 5 

'sound' 107 'thin form' 14 'jocular' 5 

'long thin form' 97 'smooth surface' 13 'falling' 5 

'quick or strong 67 'slang' 12 'soft 

4 

movement' 

consistency' 

'wetness' 63 'beat' 12 'small end 

form' 

4 

'talking' 55 'slackness' 11 'secrecy' 4 

'light' 32 'rough surface 

structure' 

10 'iterative' 3 

'diminutive' 31 'separation' 9 'bent form' 3 

'round form' 23 'putting 

together' 

9 'narrow form' 3 

'way of walking' 22 'hardness' 8 'stiffness' 2 

90

'destruction' 21 'bodily feeling' 8 'potential 

movement' 

2 

'winding form' 20 'fine grain' 8 'light tactile 

sensation' 

2 

'short-wide 

form' 

18 'hollow form' 8 'augmentative' 1 

'bad mood' 18 'quickness' 6 'dryness' 1 

'form' 16 'adhesion' 6 

91

pejorative 

sound 

long thin form 

quick or strong movem 

wetness 

talking 

light 


round form 

way of walking 

destruction 

winding form 

bad mood 

shortwide form 

form 

gaze 

thin form 

smooth surface 

beat 

slang 

slackness 

rough surf. str 

separation 

putting together 

hardness 

bodily feeling 

fine grain 

hollow form 

quickness 

adhesion 

attitude 

jocular 

falling 

soft cons 

small end form 

secrecy 

bent form 

narrow form 

stiffness 

pot. movem. 

light tactile sens. 

augmentative 

dryness 

0 20 40 60 80 100 120 140 160 180 

Diagram, 4.5 The diagram shows the extent to which different semantic 

features are exploited by all initial consonant clusters, in absolute 

numbers. It shows e.g. that 'pejorative' is the most frequent semantic 

feature. 

92

Combinations of features 

The high frequency of some of the features is probably due to their 

ability to combine with other features. 'Pejorative' and 'sound' often 

combine with other features in a root morpheme, e.g. 'walking', 'talking' 

('way of walking' and 'talking' are very often combined with 

'pejorative'). A feature like 'winding form', on the other hand, is very 

low frequent (only found in two clusters, kr- and sn-) and this low 

frequency might be partly a consequence of lower ability to combine with 

other features. Likewise 'bad mood' is confined to the clusters gr-, vrand 

tr-. The frequencies of different semantic features for each cluster 

and how different features combine can be studied in Appendix 1. 

4.3 Frequent semantic features 

A detailed account for how semantic features exploit the five most 

frequent clusters (cf. diagram. 4.5) is shown below. These features are 

'pejorative', 'wetness', 'sound', 'long thin form' and 'quick or strong 

movement'. 

Diagram 4.6 shows the feature 'pejorative' and how it is distributed in 

terms of percentage over 28 clusters. (The percentage is calculated on 

number of features per total number of root morphemes, for every 

cluster.) 

4.3.1 Pejorative 

Diagram 4.6 shows the percentage of morphemes with a pejorative 

feature. For example, 71 % of all morphemes beginning with pj- and 

44% of all morphemes beginning with fn- have a pejorative meaning. 

Almost all clusters have morphemes with pejorative features. As can be 

seen, pj- is the cluster with the highest percentage of pejorative root 

morphemes. fn- comes second and then two more cluster ending in j: nj- 

(which is however to small to be considered interesting) and fj-. 

93

80 

70 

60 

50 

40 

30 

20 

10 

0 

sp 

sk 

pr 

str 

tr 

sv 

sm 

kl 

kr 

gr 

pl 

fl 

gl 

mj 

skv 

sn 

dr 

bl 

kn 

skr 

bj 

vr 

sl 

gn 

fj 

nj 

fn 

pj 

Diagram 4.6. Percent root morphemes with the feature 'pejorative' for different clusters.

It is also interesting to study the absolute numbers of morphemes with 

pejorative features, since some clusters are lexically very frequent and 

others are very infrequent. The morphemes with a pejorative feature for 

the clusters are shown in table 4.3 (the absolute number of root 

morphemes for each cluster is placed after every cluster). The clusters 

are presented in order of descending relative frequency (cf. diagram 4.6). 

Table 4.3 The morphemes with the feature pejorative, the initial 

consonant clusters involved, and their absolute frequencies. 

pj-: pjatt (squirt), pjoller (babble, 

twaddle), pjosk (coddling), pjunk (coddling; 

puling), pjåkig (mawkish, bad) (SAOL also 

lists pjalt (stackare), pjunk ("gnäll, pjosk"), 

pjåk ("pjunk")) 5 

fn-: fnask (prostitute), fnatta (run about), 

fnoskig (dotty), fnurra (cf. grumpy) 4 

nj-: njugg (niggardly) 1 

fj-: fjant (busybody), fjollig (foolish), fjompig 

(foolish), fjuttig (insignificant), fjäsa 

(show off), fjäska (fawn on) (SAOL also lists 

fjoskig (fnoskig) (dotty) 6 

gn-: gnat (nagging), gneta (stint), gnidare 

(miser) 3 

sl-: slabba (splash), slabbertacka 

(gossipmonger), sladdra (gossip), slafs 

(sloppiness), slampa (slut), slams 

(slovenliness), slarvig (slipshod), slas 

(slugish and slipshod person), slasa 

(walk lazily), slask (slush), slatt (heeltap), 

slattrig (gossiping), slidder (gossip), slinka (wench), 

slisk (sweet stuff), slok (bloke), 

sludder (slurred speech), slum (slum), 

slusk (shabby-looking fellow), slyna (bitch), 

slyngel (young rascal), slö (indolent), 

slödder (riff-raff), slösa (squander) 24 

vr-: vräkig (flashy), vränga 

(turn inside out), vrövel (silly talk) 3 

bj-: bjäbba (squabble, bicker), bjäfs 

95

(gewgaws) 2 

skr-: skral (poor, bad), skrodera (brag), 

skrubb (cubby-hole), skrutt (rubbish), 

skrymt (hypocrisy), skräppa (brag), 

skrävla (brag), skröna (tall story) 8 

kn-: knackig (quite bad), knal 

(economically weak), knasig (crazy), 

kneg (job), kneken (vara på kneken: 

be down one's luck), knodd (bounder), 

knorva (make creased 

and wrinkled), knutte (person who is 

one-sidedly focused on a certain activity), 

knöl (bastard), knös (rich fellow) 10 

bl-: bladdra (babble), blaffa (big ugly 

colour patch), blaj (silly talk), blarr (silly 

talk), blaska (splash), bliga (stare stupidly), 

bluddra (talk nonsense), blunder ( 

blunder) (SAOL also lists (blaha 

(nonsense talk)) 8 

dr-: drasut (lanky fellow), dratta (fall), 

dravel (twaddle), dregel (dribble), drulle 

(clumsy fool), drummel (lout), drägg 

(dregs), drälla (spill), dröse (unstructured 

amount) 9 

sn-: snafs (dirt), snaskig (smutty), snatta 

(pinch), snicksnack (chit-chat), sniken 

(greedy), snobb (dandy), snorkig (snooty), 

snusk (smuttiness), snylta (sponge), 

snål (stingy), snärta (young thing), 

snöd (sordid), snöplig (inglorious) 13 

skv-: skvallra (gossip) 1 

mj-: mjäkig (namby-pamby) 1 

gl-: glop (whipper-snapper), glufsa 

(scoff), glunkas (det glunkas: there is a 

rumour), glupande (ravenous), glupsk 

(greedy), glåpord (taunt) 6 

fl-: flacka (rove), flamsig (silly, giggly), 

96

flina (sneer, cackle), flitter (?), floskel 6 

(empty phrase), flabba (cackle), flopp 

(ignominious failure), flummig (intellectually 

unclear). SAOL also lists flepig ("mjäkig", 

namby-pamby) 8 

pl-: pladuska (irregular spot), plottra 

(potter about), plufsig (bloated), plump 

(rude), plump (blot), plundra (plunder), 

plussig (bloated) 7 

gr-: grumlig (muddy), grummel 

(dregs), grums (dregs), grumsa 

(grumble), gräll (glaring), gräma (grieve), 

gräslig (terrible) 7 

kr-: krafs (trash), krake (wretch), kram 

(trash), krams (trash), kratta (funk), kreta 

(badly whittle), krimskrams (knick-knacks), 

kruserlig (ceremonious), krångel (fuss), kräk 

(wretch) 10 

tr-: tradig (boring), traggla (plod through), 

trams (rubbish, drivel), traska (trot, 

jog), trassel (tangle, muddle), troll 

(troll), truls (unordered collection) 7 

kl-: klanka (grouse, grumble), klantig 

(clumsily stupid), klotter (scribble, doodle), 

kludda (daub, smudge), klyscha 7 (cliché; 

lump of spittle), klåpare (bungler, botcher) 6 

sm-: smicker (blarney), smolk 

(particle of dirt), smuts (filth) 3 

sv-: svamla (drivel), svassa (strut, 

swagger), svulstig (bombastic) 3 

str-: strul (fuss), strunt (rubbish), 

(stuck-up blighter) 2 

pr-: pracka (fob), pryl (awl), prångla (utter - 

counterfeit coin) 3 

6 The word floskel comes from Latin flosculus which means "little flower; showy 

decoration in speech" but it also fits into the Swedish pattern (cf. discussion in 1.5) 

7 Another interesting word, which can be said to come from cliché, but which fits into the 

pejorative group (and also with all the root morphemes denoting way of talking, etc.) 

especially in connection with the dialectal meaning of klyscha (lump of spittle). 

97

sk-: skavank (flaw), skorv (old tub) 2 

sp-: spill (waste), spoling (whipper-snapper) 2 

The following diagram (4.7) shows the absolute number of features for 

the clusters. 8 This diagram shows that sl- has 24 root morphemes with a 

pejorative feature, sn- has 13, etc. It is clear that a large number of all 

pejorative morphemes begin with an s. Sl-, sn- and skr- begin a large 

number of morphemes that have a pejorative component (but these 

clusters are not the most dominated by sound symbolic root morphemes, 

cf. diagram 4.4). 

Note again that one root morpheme can have more than one motivated 

meaning component. For example 'sound' and 'pejorative' are often 

combined as in bluddra, or 'sound', 'wetness' and 'movement' as in 

skvimpa. 'Walking' is often combined with an additional slightly 

pejorative meaning component (e. g. svassa has 'pretentious manner'). 

'Talking' is also often combined with an additional pejorative meaning 

component (e.g. pladdra implies 'nonsense'). 

8 The reason why there are 30 clusters here instead of 28 as in diagram 4.6 is that the 

frequency of 1 morpheme is below 1% in a large cluster like st-. 

98

25 

20 

15 

10 

5 

0 

fl 

spr 

nj 

mj 

st 

sp 

sk 

str 

bj 

sv 

sm 

vr 

pr 

gn 

fn 

pj 

gl 

kl 

fj 

pl 

gr 

tr 

bl 

s kr 

dr 

kr 

kn 

sn 

sl 

k 

Diagram 4.7 Number of root morphemes with the 'pejorative' feature.

Comments on the feature 'pejorative' 

As stated above, pj- is the cluster with the highest percentage of 

pejorative root morphemes. All two-consonant clusters containing a jphoneme 

are pejorative and to quite a high degree. 

Looking at absolute figures, clusters beginning with s – especially sl- snand 

skr- – are dominating among the pejorative morphemes. Sl- is by far 

connected with the greatest number of pejorative morphemes. 

4.3.2 Sound 

The semantic feature 'sound' ranks as number 2 in sound symbolic 

frequency (cf. diagram 4.5). Diagram 4.8 shows that fnis the cluster 

with the highest proportion of morphemes having the feature 'sound'. Fnis 

a lexically infrequent cluster and thus the tendency from 'pejorative' 

and is repeated: lexically infrequent clusters tend to be highly sound 

symbolic. Fn- is also the most sound symbolic cluster of all (100%) 

counting the different semantic features. 

100

35 

30 

25 

20 

15 

10 

5 

0 

st 

pr 

tr 

fl 

fr 

pl 

gl 

fj 

sk 

kr 

br 

skr 

sm 

sn 

mj 

bj 

kl 

kv 

gn 

kn 

skv 

fn 

Diagram 4.8 Percent root morphemes with the feature 'sound' for different clusters.

The morphemes with the feature 'sound' for the clusters are shown in 

table 4.4 (the absolute number of root morphemes for each cluster is 

placed after every cluster). The clusters are presented after descending 

relative frequency (cf. diagram 4.8): 

Table 4.4 The morphemes with the feature 'sound', the initial consonant 

clusters involved and their absolute frequencies. 

fn-: fnissa (giggle), fnittra (giggle), fnysa 

(snort) 3 

skv-: skval (pouring), skvalp (lapping) 2 

kn-: knacka (knock), knaka (creak), knall 

(bang), knapra (nibble), knarr (creaking), 

knastra (crackle), knatter (rattle), knirka 

(creak), knirra (creak), knistra (crackle), 

knittra (crackle), knorr (grumbling), knot 

(murmuring), knyst (the least sound), 

knäppa (click) 15 

gn-: gnissla (squeak), gnoding (a fish with a 

grumbling sound), gnägga (whinny), gnälla 

(squeak) 4 

kv-: kvacka (quack), kvarka (strangles), 

kvida (whimper), kvillra (twitter), kvirra 

(grumble), kvitter (chirp), kväda (sing), 

kväka (croak) 8 

kl-: klafs (squelch), klamp (tramp), klang 

(clang), klappa (knock), klappra (klatter), 

klatsch (crack), klick (lump), klifs (squelch), 

kling (ringing), klink (tinkling), klirra (jingle), 

klocka (bell), klucka (cluck), klämta (toll), 

kläpp (clapper) 15 

bj-: bjällra (bell) 1 

mj-: mjau (meow) 1 

sn-: snarka (snore), snarpa (creak), snarra 

(burr), snattra (quack), snusa (sniff), 

snyfta (sob), snyta (blow one's nose), 

snäppa (click), snärp (duck's sound), 

snörvla (snuffle) 10 

102

sm-: smack (smack), smaska (guzzle), 

smatter clatter), smätta (flick) 4 

skr-: skramla (rattle), skrap (scraping), 

skrälla (blare) 3 

br-: braka (crash), brassa (fire away), 

brum (grumble), brus (buzz), bräka (bleat), 

bröl (growl) 6 

kr-: krafsa (scratch), krakel (row), kras 

(crack), krasch (crash), kraxa (croak), 

krysta (bear down), kråka (crow), krämta 

(hawk) 8 

sk-: skall (barking), skalla (resound), skallra 

(rattle), skorr (burr), skott (shot), skångra 

(vibrate) 6 

fj-: fjärta (fart) 1 

gl-: glam (laughing and talking), glissando, 

gläfs (yelp) 3 

pl-: pladask (flop), plask (splash), pling 

(ting-a-ling), plums (plop) 4 

fr-: frasa (rustle), frusta (snort), fräsa (hiss) 3 

fl-: flabb (cackle), flöjt (flute) 2 

tr-: trumma (drum), trumpet 

(trumpet) 2 

pr-: prassel (rustle), prutt (fart) 2 

st-: stampa (stamp), stepp (tap-dance), 

stön (groan) 3 

The list (as well as diagrams 4.8 and 4.9) shows that kn- is frequent both 

percentally and absolutely. 

Comments on the feature 'sound' 

Diagram 4.9 shows how many root morphemes have the feature 'sound', 

in absolute figures. Four of the five largest ones begin with k: kl-, kn-, 

kv- and kr-. Kl-, kn- and kv- are also percentally (cf. diagram 4.8) 

among the clusters more dominating for the feature 'sound'. 

103

16 

14 

12 

10 

8 

6 

4 

2 

0 

mj 

fj 

bj 

skv 

fl 

pr 

tr 

st 

gl 

fr 

fn 

skr 

sm 

pl 

gn 

sk 

br 

kr 

kv 

sn 

kn 

kl 

Diagram 4.9 Number of root morphemes with the feature 'sound' for different clusters.

4.3.3 Long thin form 

The feature 'long thin form' ranks as number 3 in sound symbolic 

frequency according to diagram 4.5. This is one of the many 'form'features. 

35 

30 

25 

20 

15 

10 

5 

0 

spj 

s tr 

spr 

sp 

sl 

sn 

Diagram 4.10 Percent of the feature 'long thin form' for different 

clusters. 

Diagram 4.10 shows that the semantic feature 'long thin form' is 

dominated by three three-consonant clusters: spj-, str-, and spr-. All 

these begin with s (a fact of all Swedish initial three consonant clusters) 

just as all but two of the other clusters, namely sp-, sl-, sn-, st-, sk- and 

sv-. The only ones that do not begin with s are tr- and tv-. ( However, t 

is a dental like s.) Again the highest ranked cluster spj- is a lexically very 

infrequent cluster. 

The morphemes with the feature 'long thin form' for the clusters are 

shown in table 4.5 (the absolute number of root morphemes for each 

cluster are placed after every cluster). The clusters are presented after 

descending relative frequency (cf. diagram 4.10). 

tr 

105 

tv 

st 

sk 

sv

Table 4.5 The morphemes with the feature 'long thin form', the initial 

consonant clusters involved and their absolute frequencies. 

spj-: spjut (spear), spjäla (lath) 2 

str-: streamer (streamer), streck (stroke), 

stretch (stretch), strigel (strop), stril 

(nozzle), strimla (strip), strimma (streak), 

stripa (wisp), strut (cone), strå (straw), 

stråk (band), stråle (ray), stråt (way), 

sträcka (stretch), sträng (string), ström 

(stream), strössel (sprinkles) 17 

spr-: spricka (crack), springa (narrow 

opening), sprint (pin), sprits (squirt), 

sprund (slit), spröjs (window-bar), 

spröt (rib) 7 

sp-: spaljé (trellis), spalt (column), 

spant (rib), sparre (rafter), sparris 

(asparagus), spatel (spatula), spene (dug), 

spenslig (slender), spetig (skinny), spets 

(point), spett (spit), spigg (stickleback), spik 

(nail), spila (spike ling), spindel (spider), 

spinkig (thin), spinna (spin), spira (spire), 

spole (bobbin), spont (tongue), spång 

(foot-bridge), spänta (split wood), spö (twig) 23 

sl-: sladd (cord), slamsa (rag), slana (scaffold 

pole), slang (tube), slank (slender), slant 

(a fishing rod), slejf (strap), slimmad (fitted), 

slinga (coil), slips (tie), slits (slit), släde (sleigh) 12 

sn-: snabel (trunk), snigel (snail), snilja 

(a thread), snitsel (paperstrip), sno (twine), 

snodd (cord), snok (grass snake), snorkel 

(snorkel), snugga (cutty), snöre (string) 10 

tr-: tratt (funnel), tross (hawser), truta 

(pout), tryne (snout), tråd (thread), tråg 

(trough), trål (trawl), träns (braid) 8 

tv-: tvinna (twine) 1 

st-: stake (stake), stav (staff), sticka 

(splinter), stift (pin), stig (path), stilett 

(stiletto), stock (log), stolpe (pole), stylta 

106

(stilt), stång (pole), stör (stake), stötta 

(prop) 12 

sk-: skakel (shaft), skalm (shaft), skalpell 

(scalpel) 3 

sv-: svabba (swab), svans (tail) 2 

25 

20 

15 

10 

5 

0 

sp 

s tr 

sl 

st 

sn 

tr 

Diagram 4.11 Number of morphemes with the feature 'long thin form' 

for different clusters. 

Comments on the feature 'long thin form'. 

Diagram 4.11 shows that sp- is the largest cluster, in absolute figures, for 

the semantic feature 'long thin form'. 

4.3.4 Quick or strong movement 9 

Diagram 4.12 shows that fl- has the highest percent of morphemes having 

the feature 'quick or strong movement'. Fl- is a quite large cluster. In 

second place comes spr- which is a lexically infrequent cluster even if it 

is not extremely small. Vr- in third place is also lexically infrequent. 

The morphemes with the feature 'quick or strong movement' for the 

clusters are shown in table 4.6 (the absolute number of root morphemes 

9 The reason for collapsing quick movement and strong movement is that in some root 

morphemes both these meanings are present (although in others only one is present). 

However, it seems cumbersome to make three categories instead of one. 

spr 

107 

sk 

spj 

sv 

tv

for each cluster are placed after every cluster). The clusters are presented 

according to descending relative frequency (cf. diagram 4.12). 

20 

18 

16 

14 

12 

10 

8 

6 

4 

2 

0 

fl 

spr 

vr 

sv 

gn 

fn 

skv 

Diagram 4.12 Percentage 'quick or strong movement' for different 

clusters. 

Table 4.6 The morphemes with the feature 'quick or strong movement', 

the initial consonant clusters involved, and their absolute frequencies. 

fl-: flacka (rove), fladdra (flutter), flagga 

(flag) flamma (flame), flanera (be out for a 

stroll), flaxa (flutter), flimra (flicker), fluga 

(fly), fluktuation (fluctuation), flyga (fly), fly 

(flee), flyta (float), flytta (move), flåsa 

(puff) fläkta (fan), flämta (pant), flänga 

(be dashing about), flöda (flow) 18 

spr-: sprallig (frolicsome), sprattla 

(flounder), spritta (jump), sprudla (bubble), 

spruta (spurt), sprutt (speed) 6 

vr-: vricka (scull), vrida (twist), vräka 

(heave) 3 

sv-: svaja (swing), svalla (surge), svepa 

(sweep), sving (swing), svirvel (swivel), 

svämma (overflow), sväng (sweep), 

svärm (swarm), sväva (float) 9 

sl 

108 

kn 

sk 

tr 

fr 

kr 

sp 

st

gn-: gnida, (rub) gno (rub), gnugga (rub) 3 

fn-: fnatta (run about)) 1 

skv-: skvimpa (splash to and fro) 1 

sl-: sladda (skid), slinka (slip), slinta (slide), 

slipprig (slippery), slira (skid), slita (pull), 

slugga (beat), slunga (fling), slänga (throw) 9 

kn-: knixa (bob), knyck (jerk) 2 

sk-: skaka (shake), skalv (quake), skippa 

(skip) 3 

tr-: tromb (tornado), tromla (rotating 

cylindrical sieve), trumla (treat 

something by enclosing it in a rotating 

drum) 3 

fr-: frossa (the shivers), frusa (gush) 2 

kr-: kratsa (scrape), kränga (heave over) 2 

sp-: spasm (spasm), spurt (spurt) 2 

st-: studsa (bounce), stöppla (give painted 

surface a certain look by striking a brush 

against it), stöt (thrust) 3 

18 

16 

14 

12 

10 

8 

6 

4 

2 

0 

fl 

sv 

sl 

spr 

vr 

tr 

st 

gn 

Diagram 4.13 Number of morphemes with the feature 'quick and strong 

movement' for different clusters. 

109 

sk 

fr 

kn 

kr 

sp 

skv 

fn

40 

35 

30 

25 

20 

15 

10 

5 

0 

kr 

str 

dr 

sn 

pl 

sp 

sv 

bl 

sl 

kl 

sm 

spr 

skv 

Diagram 4.14 Percent root morphemes with the feature 'wetness' for different clusters.

Comments on the feature 'quick or strong 

movement'. 

Diagram 4.13 shows that fl- has by far the greatest number of 

morphemes with the feature 'quick and strong movement'. Fl- is then 

most frequent both absolutely and percentally. 

4.3.5 Wetness 

The last semantic feature to be examined is 'wetness', which ranks as 

number 5 in sound symbolic frequency according to diagram 4.5. 

Diagram 4.14 shows that skv- is the cluster with the highest proportion of 

morphemes having the feature wetness. (Almost half of the morphemes 

beginning with skv- have the feature 'wetness', cf. table 4.16) The next to 

largest cluster, sm-, has only 13% morphemes with this feature. The 

cluster skv- is small and contains a total of ten root morphemes. The 

morphemes with the 'wetness' feature for the clusters are the following 

(the absolute number of root morphemes for each cluster are placed after 

every cluster). The clusters are presented according to descending relative 

frequency (cf. diagram 4.14). 

Table 4.7 The morphemes with the feature 'wetness', the initial consonant 

clusters involved, and their absolute frequencies. 

skv-: skval (gush), skvalta (ripple), skvimpa 

(splash), skvätt (drop) 4 

spr-: sprej (spray), sprinkler (sprinkler), 

sprudla (bubbla), spruta (spurt) 4 

sm-: smegma (secretion), smet (sludge), 

smetana (a sour thick cream), smink 

(make-up), smälta (melt), smörja (grease) 6 

kl-: klabb (stick), kladd (daub), klafs 

(squelch), klena (daub), kleta (daub), klibba 

(stick), klick (daub), klifs (squelch), klucka 

(gurgle), klunk (gulp) 10 

sl-: slabba (splash), slafsa (squelch), 

slam (ooze), slask (slush), slatt (drop), slem 

(slime), slicka (lick), slipprig (slippery), slisk 

(sweet stuff), sluring (soup), slurk (swig), 

111

sluss (sluice) 12 

bl-: blaska (splash), blod (blood), bläck (ink), 

blöt (wet) 4 

sv-: svabba (swab), svett (sweat), svämma 

(overflow) 3 

sp-: spackel (putty), spad (liquid), sperma 

(sperm), spilla (spill), spola (flush), spott 

(spittle), sputum (phlegm) 7 

pl-: plask (splash), plums (plop), plurr 

(water) 3 

sn-: snaskig (smutty), snigel (snail), snor 

(snot), snuva (cold) 4 

dr-: dregel (dribble), droppa (drip) 2 

str-: strila (sprinkle), ström (stream) 2 

kr-: kram (cloggy), kräm (cream) 2 

Comments on the feature 'wetness' 

Diagram. 4.15. shows in absolute numbers how many morphemes have 

the feature 'wetness' for each cluster. The largest cluster is sl- (twelve 

instances) followed by kl- (ten instances). 8 of the 13 'wetness' clusters 

begin with s. Three begin with an unvoiced stop and two begin with a 

voiced stop. 

Summary of and discussion of the five most 

frequent features 

As stated above, skv- is the cluster with the highest percentage of 

'wetness' root morphemes. This cluster only contains ten morphemes so – 

as well as for the pejorative root morphemes – small clusters are 

proportionally more utilized for sound symbolism. The clusters sl-, kl-, 

sp-, sm-, skv-, spr-, sn-, bl-, pl-, and sv- are most utilized for the 

meaning 'wetness', i.e. s or initial unvoiced plosives are preferred. Sland 

kl- (both ending with l) have the highest number of root morphemes 

with the feature 'wetness'. 

112

12 

10 

8 

6 

4 

2 

0 

sl kl sp sm skv sn bl spr sv pl kr dr str 

Diagram 4.15 Number of root morphemes with the feature 'wetness' for different clusters.

The cluster sl- is the most frequent sound symbolic cluster for both 

'pejorative' and 'wetness'. 

Looking at absolute figures, clusters beginning with s, especially sl- sn-, 

and skr-, are dominating among the 'pejorative' morphemes. Sl- has by 

far the greatest number of pejorative morphemes. 

Similar results for cluster types were obtained for other clusters : 'light' 

is dominated by voiced plosives + l or n: bl-, gl-, gn- (dl-, bn-, dn- are 

not allowed in Swedish) 10, 'bad (negative) mood' is constrained to some 

plosives + r: gr-, tr-, vr- 11 and 'wetness' is dominated by: sl-, kl-, sp-, 

sm-, skv-, spr-, sn-, bl-, pl-, sv-, i.e. s and initial unvoiced plosives are 

preferred (the only exception is bl-). The phoneme l is also quite 

common. 

10 There are root morphemes beginning with other clusters that have been classified as 

‘light’ but these mainly concern quality of colour: bjärt (gaudy), gräll (glaring), 

prunkande (dazzling), prålig (garish). 

11 There is only one exception: knarrig (creaky). 

114

4.3.6 The most sound symbolic clusters 

Table 4.8 shows which the most frequent clusters are, for every meaning. 

Table 4.8. The most sound symbolic clusters in absolute numbers and 

proportionally, for the five most frequent meanings. 

freq % examples in % % freq examples 

meaning in 

absolutenumbers 

'pejorative' sl- 24 16 sladder 

(gossip), slok 

(bloke), 

slödder (riffraff) 

'sound' kl-, 

kn- 

'long thin 

form' 

'quick or 

strong 

movement' 

15 13, 

19 

klang (clang), 

klirra (jingle) 

knarra 

(creak), 

knittra 

(crackle) 

sp- 23 17 spant (rib), 

spett (spit) 

spö (twig) 

fl- 18 19 fladdra 

(flutter), 

flamma 

(flame), 

flimra 

(flicker). 

fläkta (fan) 

'wetness' sl- 12 8 slask (slush) 

slem (slime) 

slipprig 

(slippery) 

115 

pj- 71 5 pjoller 

(babble), 

pjosk 

(coddling), 

pjunk 

fn- 33 3 

(coddling) 

fnissa 

(giggle), 

fnysa (snort) 

spj- 33 2 spjut (spear), 

spjäla (lath) 

fl- 19 18 fladdra 

(flutter), 

flamma 

(flame), 

flimra 

(flicker) 

fläkta (fan) 

skv- 40 4 skval (gush), 

skvalta 

(ripple), 

skvimpa 

(splash)

The information given in table 4.8, for the five most frequent meanings, 

will now be explained. For the feature 'pejorative' (which is the most 

frequent meaning, see diagram 4.5) the most high frequent cluster (in the 

sense of the greatest number of roots) is sl-, and it has a frequency of 24 

root morphemes. Sl- can therefore be said to be a very pejorative cluster. 

However, many words begin with sl- and only 16% of these are in fact 

pejorative. It is also interesting to look at which cluster is dominated to 

the highest degree by the feature 'pejorative'. This turns out to be pjwhere 

5 root morphemes (71% of the root morphemes beginning with 

pj-) are pejorative. So, in the sense of being dominated by a certain 

meaning, pj- can be said to be the most pejorative cluster. The same 

comparisons are made for the remaining 4 semantic features. 

The clusters that are most frequent, in absolute numbers, show a strong 

tendency to end with l. The clusters that are most frequent proportionally 

show a weak tendency to end with j. In both groups the clusters begin 

with a voiceless obstruent. 

4.4 Frequent clusters 

So far we have studied how clusters are distributed over different 

meanings. We will now look at what meanings different clusters hold. 

The following section shows how different meanings are distributed over 

different clusters. The clusters examined are the most frequent in absolute 

numbers: sl-, sn-, kn-, kr- (cf. diagram 4.3) and some of the most 

frequent percentally: fn-, kn-, gn-, spr-, pj- (cf. diagram 4.4). 

Also those clusters that are almost unique for a certain meaning will be 

discussed, e.g. 'pejorative' has the unique cluster fj-, i.e., fj- is almost 

only 'pejorative'. Fl- is to a high degree 'quick or strong movement', fn- 

'pejorative', gl- 'light', bl- 'light' or 'gaze', kr- 'winding form', kn- 

'round form', skr- 'pejorative' and 'destruction', sl- 'pejorative', str- 

'long thin form', br- 'sound' and 'destruction', gr- 'bad mood' or 'hollow 

form', (vr- 'bad mood', tr- 'bad mood'), kl- is 'wetness', 'short-wide 

form' or 'adhesion', kv- is 'suffocation', mj- 'softness', sk- 'hardness', 

skv- 'wetness', sl- 'slackness' and sp-, st- and str- are 'long thin form' 

(cf. table 4.16). 

116

25 

20 

15 

% m.comp/all 

10 

5 

0 

walking 

pot. movem. 

smooth surf. 

beat 


strong movem 

quick or 

slackness 

long thin 

wetness 

pejorative 

Diagram 4.16 Semantic features of the cluster sl-. Percentages of semantic features for all root morphemes.

4.4.1 The cluster sl- 

The first cluster to be studied is sl-, the largest cluster in absolute 

numbers. Diagram 4.16 shows the semantic features of the consonant 

cluster sl-. The words (representing root morphemes) of the dominating 

features are shown in table 4.8. The absolute frequencies (not the 

percentages) of the semantic features are shown to the right. 

Table 4.8 The sound symbolic morphemes of the sl- cluster, the semantic 

features involved, and their absolute frequencies. 

'pejorative': slabba (splash), 

slabbertacka (gossipmonger), 

sladdra (chatter), slafs (sloppiness) 

slampa (slut), slams (slovenliness), 

slarvig (slipshod), slas (lazy and 

careless person), slasa (walk 

heavily and shuffling), slask (slush), 

slatt (drop), slattrig (slack), slidder 

(gossip), slinka (wench), slisk 

(sweet stuff), sloka (droop), sludder 

(slurred speech), slum (slum), slusk 

(shabby looking fellow), slyna 

(bitch), slyngel (young rascal), slö 

(sluggish), slödder (riff-raff), 

slösa (waste) 24 

'wetness': slabba (splash), slafs 

(sloppiness), slam (ooze), slask 

(slush), slatt (drop), slem (slime), 

slicka (lick), slipprig (slippery), slisk 

(sweet stuff), sluring (a soup), slurk 

(swig), sluss (sluice) 

(cf. discussion in 1.5) 12 

'long thin form': sladd (cord), 

slamsa (rag), slana (scaffold pole), 

slang (tube), slank (slender), slant 

(a fishing rod), slejf (strap), slimmad 

(fitted), slinga (coil), slips (tie), slits (slit), 

släde (sleigh). 12 

118

'slackness': slack (slack), sladdrig 

(flabby), slak (slack), slamsa (rag), 

slana (scaffold pole), slang (tube), 

slankig (flaccid), slapp (slack), sliddrig 

(flabby), slinkig (slabby), sloka (droop) 11 

'quick or strong movement': sladda 

(skid), slinka (slip), slinta (slip), 

slipprig (slippery), slira (skid), slita 

(tear), slugga (slug), slunga (fling), slänga 

(fling) 9 

'talking': slabbertacka (gossipmonger), 

sladdra (chatter), slattrig (gossiping), 

slidder (gossip), sludder (slurred 

speech) 5 

'beat': slag (beat), slå (beat), slägga 

(sledge hammer) 3 

'smooth surface': slipa (polish), slät 

(smooth), slätt (plain) 3 

'potential movement': slutta (slant), 

slänt (slope) 2 

'walking': slasa (walk heavily and 

shuffling), släntra (saunter) 2 

The semantic feature 'slackness' is unique for this cluster. 

The relations between some of the different meanings of sl- are discussed 

below in 4.6. 

Sl- has a different meaning profile than sn- (cf. diagram 4.17). However, 

they are both dominated by the feature 'pejorative'. 

4.4.2 The cluster sn- 

The second most sound symbolic cluster, in absolute numbers, is sn-. 

Diagram 4.17 shows the semantic features of this consonant cluster. 

119

10 

9 

8 

7 

6 

5 

4 

3 

2 

1 

0 

pejorative 

sound 


quickness 

wetness 

Diagram 4.17 Semantic features of the cluster sn-. Percentages of 

semantic features for all root morphemes. 

In table 4.9 the words (representing root morphemes) of the dominating 

features are shown. The absolute frequencies (not the percentages) of the 

semantic features are shown to the right. 

Table 4.9 The sound symbolic morphemes of the sn- cluster, the semantic 


'pejorative': snafs (dirt), snaskig (smutty), 

snatta (pinch), snicksnack (chit-chat), 

sniken (greedy), snobb (dandy), snorkig 

(snooty), snusk (smuttiness), snylta 

(sponge), snål (greedy), snärta 

(young thing), snöd (sordid), snöplig 

(inglorious) 13 

'sound': snarka (snore), snarpa (creak), 

snarra (burr), snattra (quack), snusa 

120 

small end form 





slang

(sniff), snyfta (sob), snyta (blow nose), 

snäppa (click) , snärp (duck´s sound), 

snörvla (snuffle) 10 

'long thin form': snabel (trunk), snigel 

(snail), snilja (a thread) , snitsel (paper 

strip), sno (twine), snodd (cord), snok (grass 

snake), snorkel (snorkel), snultra (a long 

thin fish), snöre (string) 10 

'quickness': snabb (quick), snappa (snatch), 

snar (who reacts without delay), sno 

(twine), snudig (quick), snärj (hurry) 6 

'talking': snacka (chat), snicksnack 

(chit-chat), snubba (rebuke), snäsa (snub) 4 

'wetness': snaskig (smutty), snigel (snail), 

snor (snot), snuva (head cold) 4 

'small end form': snagga (clip short), snibb 

(tip), snip (lip), snopp (tip) 4 

'winding form': snirkel (scroll), snurra 

(twist), snäcka (shell) 3 

'diminutive': snitt (cut), snugga (cutty), 

snutt (small piece) 3 

'slang': snofsa upp (make elegant), snubbe 

(person), snut (cop) 3 

'walking': snava (trip), snubbla (trip) 2 

4.4.3 The cluster kn- 

The following diagram (4.18) shows the semantic features of the cluster 

kn-, the third largest cluster in absolute numbers. 

The words (representing root morphemes) of the semantic features of knare 

shown in table 4.10. The absolute frequency (not the percentages) of 

the semantic features are shown to the right. 

121

20 

18 

16 

14 

12 

10 

8 

6 

4 

2 

0 

round form 

sound 

pejorative 

together 

Diagram 4.18 Semantic features of the cluster kn-. Percentages of 


Table 4.10 The sound symbolic morphemes of the kn- cluster, the 

semantic features involved and their absolute frequencies. 

'sound': knacka (knock), knaka (creak) 

, knall (bang), knapra (nibble), knarr 

(creaking) , knastra (crackle), knatter (rattle) 

, knirka (creak), knirra (creak), knistra 

(crackle), knittra (crackle), knorr 

(grumbling), knot (murmuring), knyst 

(the least sound), knäppa (click) 15 

'round form': knagg (wooden stick), knalle 

(hillock), knapp (button), knast (twig in 

122 

way of walk 


beat 

quick or 

strong movem

wood), knick (sharp curve on pipe), knoge 

(fist), knollrig (frizzy), knop (knot), knopp 

(bud), knorr (curl), knota (bone), knottra 

(goose-pimples), knubbig (chubby), knödel 

(cooked bun of patatoes), knöl (bump) 15 

'pejorative': knackig (quite bad), knal 

(economically weak), knasig (crazy), kneg 

(job), kneken (vara på kneken: be down 

one's luck), knodd (bounder), knorva (make 

creased and wrinkled), knutte (person who 

is onesidedly focused on a certain activity), 

knöl (bastard), knös (rich fellow) 10 

'putting together': knipa (pinch), knippa 

(bunch), knipsa (staple), knut (knot), 

knyckla (crumple up), knyppla (make lace), 

knyta (tie) 7 

'walking': knaggla (move forward slowly 

and with difficulty), knalla (trot), knata 

(run), knoga (walk laboriously) 4 

'diminutive': knatte (little fellow), knott 

(gnat), knåp (finicky job) 3 

'beat': knocka (knock out), knuff (push), 

knäck (crack) 3 

'quick or strong movement': knixa (bob), 

knyck (jerk) 2 

Kn- contains many words with the meaning 'round form', but the cluster 

is not unique for this meaning. 

Kn- also belongs to the clusters in the next section, since it is also one of 

the 4 most sound symbolic cluster percentally. 

123

10 

9 

8 

7 

6 

5 

4 

3 

2 

1 

0 


pejorative 

sound 


124 


rough surf 

destruction 

quick or 

strong movem 

Diagram 4.19 Semantic features of the cluster kr-. Percentages of 


4.4.4 The cluster kr- 

The last of the four largest clusters (in absolute numbers) is kr-, shown in 

diagram 4.19. In table 4.11 the words (representing root morphemes) of 

the semantic features of kr- are shown. The absolute frequency (not the 


Table 4.11 The sound symbolic morphemes of the kr- cluster, the 

semantic features involved, and their absolute frequencies. 

'winding form': krans (wreath), 

kringelikrokar (lots of bends in different 

wetness

directions), kringla (pretzel), krok (hook), 

krokan (croquembouche), krulla (curl), krum 

(crooked), krumbukt (windings), krumelur 

(flourish), krumpen (crooked and bent from 

age or disease), krusa (curl), krusiduller 

(ornaments), krycka (crutch), kråma (prance 

about), kräkla (crosier), kräla (crawl), 

krök (bend) 17 

'pejorative': krafs (trash), krake (wretch), 

kram (trash), krams (trash), kratta (funk), 

kreta (badly whittle), krimskrams (knick-knacks), 

kruserlig (ceremonious), krångel (fuss), kräk 

(wretch) 10 

'sound': krafsa (scratch), krakel (row), kras 

(crack), krasch (crash), kraxa (croak), krysta 

(bear down), kråka (crow), krämta (hawk) 8 

'thin form': krakmandel (soft-shell 

almond), krokett (croquette), krusta (crust), 

krustad (croustade) 4 

'walking': kravla (crawl), krylla (crawl with), 

krypa (crawl), kräla (crawl) 4 

'rough surface structure': kratta (rake), 

krås (ruffle), kräpp (crepe) 3 

'destruction': kracka (decompose molecules 

through heating), krackelera (crackled), krossa 

(crush) 3 

'wetness': kram (cloggy), kräm (cream) 2 

'quick or strong movement': kratsa 

(scrape), kränga (heave over) 2 

Kr- is dominated by the unusual semantic feature 'winding form' and the 

cluster and meaning are almost uniquely connected. (However 'winding 

form' also occurs in some root morpheme with sn-.) 

The clusters described above are the four largest ones. Taken together 

they are dominated by the semantic features 'pejorative' and 'sound'. 

125

The largest clusters, percentally 

The next group of clusters that are to be studied more closely are the ones 

that are the percentally most sound symbolic, cf. diagram 4.4. 

4.4.5 The cluster fn- 

Fn- is the initial consonant cluster with the largest percentage of sound 

symbolic root morphemes. The percentages of the semantic features are 

shown in diagram 4.20. 

45 

40 

35 

30 

25 

20 

15 

10 

5 

0 

pejorative 

sound 

quick or strong 

movem 

Diagram 4.20 Semantic features of the cluster fn-. Percentages of 


The words (representing root morphemes) of the semantic features of fnare 

presented in table 4.12. The absolute number of words (not the 

percentages) of the semantic features are shown to the right. For all 

lexically infrequent clusters, i.e. clusters containing less than 13 root 

126 


dryness

morphemes, also root morphemes with the frequency 1 are counted (cf. 

chapter 3). 

Table 4.12 The sound symbolic morphemes of the fn- cluster, the 


'pejorative': fnask (prostitute), fnatta (run 

about), fnoskig (dotty), fnurra (cf. grumpy) 4 

'sound': fnissa (giggle), fnittra (giggle), 

fnysa (snort) 3 

'quick or strong movement': fnatta (run about) 1 

'thin form': fnasig (chapped) 1 

'dryness': fnöske (tinder) 1 

The cluster fnis dominated by 'pejorative'. All fn- root morphemes are 

sound symbolic. (However, as shown in diagrams 4.6 and 4.7, the feature 

'pejorative' is spread over many clusters.) 

4.4.6 The cluster kn- 

Kn- is the initial consonant cluster with the fourth largest percentage of 

sound symbolic root morphemes. It is also the third largest cluster in 

absolute numbers. The diagrams and the list of root morphemes can be 

studied above in 4.4.3. 

4.4.7 The cluster gn- 

Gn- is the initial consonant cluster with the second largest percentage of 

sound symbolic root morphemes. The percentages of the semantic features 

are shown in diagram 4.21. 

The words (representing root morphemes) of the semantic features of gnare 



127

25 

20 

15 

10 

5 

0 


sound 

pejorative 

quick or 

strong movem 

Diagram 4.21 Semantic features of the cluster gn-. Percentages of 


Table 4.13 The sound symbolic morphemes of the gn- cluster, the 


'talking': gnabb (bickering), gnat (nagging), 

gnola (hum), gny (din), gnöla (klaga) 5 

'sound': gnissla (squeak), gnoding (a fish 

with a grumbling sound), gnägga (whinny), 

gnälla (squeak) 4 

'pejorative': gnat (nagging), gneta (stint), 

gnidare (miser) 3 

'smooth surface structure': gnida (rub), 

gno (rub), gnugga (rub) 3 

'quick or strong movement': gnida,(rub), 

gno (rub), gnugga (rub) 3 

'light': gnista (spark), gnistra (sparkle) 2 

128 

smooth surf 

light

The semantic features of gnare primarily 'talking' and 'sound', which 

semantically are very close. An analysis of relations between some of the 

semantic features of gnare presented in 4.6. 

4.4.8 The cluster spr- 

The initial consonant cluster with the fifth largest percentage of sound 

symbolic root morphemes is spr-. The percentages of the semantic 

features are shown in diagram 4.22: 

25 

20 

15 

10 

5 

0 



movement 

Diagram 4.22 Semantic features of the cluster spr-. Percentages of 


129 

separation 

wetness

Table 4.14 shows the words (representing root morphemes) of the 

semantic features of spr-. The absolute number of words (not the 


Table 4.14 The sound symbolic morphemes of the spr- cluster, the 


'long thin form': spricka (crack), springa 

(narrow opening), sprint (pin), sprits (squirt), 

sprund (slit), spröjs (window-bar), spröt (rib) 7 

'quick or strong movement': sprallig 

(frolicsome), sprattla (flounder), spritta 

(jump), sprudla (bubble), spruta (spurt), 

sprutt (speed) 6 

'separation': sprej (spray), spreta (sprawl), 

sprida (spread), sprinkler (sprinkler), 

spränga (burst), sprätta (flick) 6 

'wetness': sprej (spray), sprinkler (sprinkler), 

sprudla (bubbla), spruta (spurt) 4 

Spr- is quite unique for the meaning 'separation'. (The meaning does, 

however, occur in sp- and spl-). 

4.4.9 The cluster pj- 

Pj- is the initial consonant cluster with the third largest percentage of 

sound symbolic root morphemes. The percentages of the semantic 

features are shown in the following diagram (4.23): 

The words (representing root morphemes) of the semantic features of pjare 



130

80 

70 

60 

50 

40 

30 

20 

10 

0 

pejorative way of 


Diagram 4.23 Semantic features of the cluster pj-. Percentages of 

semantic features for both root morphemes. 

Table 4.15 The sound symbolic morphemes of the pj- cluster, the 


'pejorative': pjatt (squirt), pjoller (babble, 

twaddle), pjosk (coddling), pjunk (coddling; 

puling), pjåkig (mawkish, bad) 5 

'sound': pjoller (babble, twaddle) 1 

The cluster is clearly dominated by the feature 'pejorative'. 

Summary for percentages 

Also for the four percentally most sound symbolic clusters the semantic 

features 'pejorative' and 'sound' are dominating, together with 'talking'. 

Only the cluster spr- is deviating from this tendency. The cluster kn- is 

among the most sound symbolic clusters, both in absolute numbers and 

percentally. 

4.5 Typical or unique meanings 

In Table 4.16 all the clusters with typical or unique meanings are 

presented. Clusters that are dominated by a certain meaning (e.g. 'quick 

or strong movement' in fl-), or have a meaning which is almost unique 

for that cluster (e.g. 'winding form': kr-), have those meanings in italics. 

For small clusters, i.e. those with less than 13 root morphemes, also 

meanings that have the frequency 1 are counted. 

131

Table 4.16 Typical 

or unique meanings 

of initial clusters. 

bl- 33 

pejorative 8 

light 9 

gaze 5 

talking 4 

wetness 4 

round form 3 

br- 11 

sound 6 

destruction 5 

fj- 7 

pejorative 6 

sound 1 

fl- 36 


movement 

18 

pejorative 8 

thin form 6 

sound 2 

light 2 

fn- 10 

pejorative 4 

sound 3 



movement 

1 

dryness 1 

gl- 45 

light 14 


structure 

7 

pejorative 6 

gaze 6 

diminutive 5 

form 4 

sound 3 

gn- 20 


sound 4 

pejorative 3 


structure 

3 


movement 

3 

light 2 

gr- 21 

hollow form 8 

bad mood 6 

pejorative 7 


kl- 51 

sound 15 

short wide form 11 

wetness 10 

adhesion 6 

pejorative 6 


kn- 59 

round form 15 

sound 15 

pejorative 10 

putting together 7 

walking 4 


beat 3 


movement 

2 

kr- 53 

winding form 17 

132 

pejorative 10 

sound 8 



rough surface 

structure 

3 

destruction 3 

wetness 2 


movement 

2 

kv- 17 

sound 8 

suffocation 5 

destruction 2 


mj- 7 

fine grain 3 

soft consistency 2 

sound 1 

pejorative 1 

pj- 6 

pejorative 5 


sk- 25 

hardness 8 

sound 6 

long thin form 4 


movement 

3 

pejorative 2 

round form 2 

skr- 35 


pejorative 8 

destruction 7 

rough surface 

structure 

5

sound 3 

skv- 9 

wetness 4 

sound 2 


pejorative 1 

movement 1 

sl- 83 

pejorative 24 

wetness 12 


slackness 11 


movement 

9 


beat 3 

smooth surface str 3 


potential 

movement 

2 

sn- 62 

pejorative 13 


sound 10 

quickness 6 


small end form 4 

wetness 4 

winding form 3 


slang 3 


sp- 47 


wetness 7 

jocular 4 


gaze 2 


movement 

2 

attitude 2 

pejorative 2 


spr- 23 


separation 6 


movement 

6 

wetness 4 

str- 30 



stiffness 2 

wetness 2 

light 2 

pejorative 2 

fine grain 2 

tr- 44 

bad mood 8 


pejorative 7 



round form 3 


movement 

3 


short wide form 2 

sound 2 

133

4.6 Patterns of semantic features 

The different semantic features of a certain consonant cluster often do not 

only constitute a list of haphazard features, but instead they are related. 

These kinds of relations between phonesthemes (e.g. sl-:'smooth surface 

structure' (slät) -> 'quick or strong movement (slira, slingra)) can also be 

found within the same phonestheme (e.g. bl-: 'light' (blixt) –>'reflecting 

surface structure potentially giving light' (blank), at a more detailed level 

of analysis. In other words, there are many clusters which contain root 

morphemes that have similar, but not exactly the same, semantic features. 

These features are related indexically or iconically and can be seen as 

belonging to the same meaning potential and activated in different 

contexts. This is especially striking in the case of root morphemes 

connected to 'light', beginning with the clusters gl-, bl- and gn-. Most of 

these words have been classified as 'light' or 'gaze' in the analysis above, 

but as shown here they could be further analyzed. Some root morphemes 

can denote light, others perception of light, yet others potentiality for 

perception of light, etc. The different relations to 'light' are paraphrased 

below. Almost all relations between phonesthemic meanings can be 

described as some kind of potentiality. Potentiality is not a semantic 

feature but it describes the relation between different meanings. By 

potentiality I mean a relation between meanings where one meaning 

describes a possible prerequisite for another meaning, e.g. form can be 

seen from the perspective of being able to let through a bit of light, or a 

reflecting (glossy) surface can send out light, cf. the paraphrases below. 

The emergence of these relations probably come from simultaneity in 

perception, i.e. a kind of indexical relation. In some cases the relation is 

iconic. 

The structures are similar but not exactly parallel, so the clusters will be 

presented one at a time: 

gl- 

'form through which light 

potentially can be perceived' 

glipa (narrow space), glugg 

(aperture), glänt (slightly open), 

glänta (glade) 

134

'perception of light' 

glana (stare), glo (stare), glutta 

(take a glance), glimt (glimpse) 

'production or source of light' 

glans (lustre), glimma (gleam), glimra (gleam), glindra 

(gleam), glisa (shine), glittra (glitter), glänsa (shine), glöd 

(glow), gloria (halo), glåmig (pale) 

'smooth surface that potentially 

reflects light' 

glas (glass), glasyr (glazing), glatt 

(smooth), glimmer (gleaming), 

glinder (a shiny trolling-spoon), 

glaciär (glacier) 

'movement on such a surface' 

glida (glide) 

135

l- 

gn- 

'perception of light' 

blick (gaze), bliga (stare), blänga 

(glare), blända (blind), blind (blind), 

blinka (blink) 

'light' 

blixt (lightning), blinka (twinkle), blänka (shine), blossa 

(flare), blek (pale) 

'reflecting surface potentially giving light' 

blank (shiny) 

136 

(iconic) 

'light surface' 

bläs (blaze), black (drab) 

'momentary light' 

gnista (spark), gnistra (sparkle) 

'action which produces smooth surface' 

gnida (rub), gno (rub), gnugga (rub)

'sound arising from, and simultaneous 

with, such an action' 

gnida (rub), gno (rub), gnugga (rub) 

'similar sound arising from similar action' 

gnissla (squeak) 

'similar sounds arising from humans or 

creatures' 

gnabb (bickering), gnat (nagging), 

gnoding (a murmuring fish), gnola 

(hum), gny (din), gnägga (neigh), gnälla 

(whine), gnöl (grumble) 

In the last cluster we can see indexical and iconic relations between such 

basic categories as 'sound', 'movement' and 'light', e.g. the relation 

between the sound and the action is often simultaneous (indexical) or the 

relation between this sound and another sound (like in gnissla) is likeness 

(iconic). Gn- can, in other words, be used for other sounds than those 

arising simultaneously with a polishing movement. These are synchronic 

relations but one can imagine the historical changes through meaning 

extension. (The etymological relations, as described in e.g. Hellquist 

(1966), are not clear.) 

Common for the three clusters is the meaning 'light', i. e. the process 

which is the source for visual perception. It is of course central. Different 

extensions of meaning from 'light' are possible. As can be seen, there are 

"empty slots" in the meaning patterns for the clusters. For example, gndoes 

not have adjectives having to do with 'reflecting surface', only verbs 

meaning 'producing reflecting surface'. And it does not have root 

morphemes for 'perception of light'. These can be accidental gaps. A 

speculative explanation of why gndoes not have words for 'perception 

137

of light' is that these words often are verbs and that the verbs beginning 

with gnare used for the feature 'sound'. 

Similar relations can be found for other semantic fields and for other 

clusters, sometimes as different root morphemes beginning with a certain 

cluster, sometimes as semantic features of a root morpheme. Meanings 

often combined are e.g. 'sound' - 'movement' - 'destruction' (e.g. braka), 

'destruction' - 'pejorative' (e.g bråte), 'wetness' - 'pejorative' (e.g. 

slemmig), 'sound' - 'surface structure' (e.g skrovlig), etc, and the relation 

between the meanings (except for 'wetness' - 'pejorative') is nearness in 

space and time (indexical). 

The semantic pattern of sl- is similar: 

pejorative (slabba (splash), slafs 

(sloppiness), slampa (slut), slarvig 

(slipshod), slasa (walk lazily), 

sladdrig (flabby)) 

wetness (slafsa (slop), slipprig (slipprig), slem (phlegm)) 

smooth surface structure (slipa 

(polish), slät (smooth)) 

quick or strong movement (slinka 

(slide), slinta (glide), slira (skid), 

slingra (wind)) 

138 

potential 

movement

139 

(slutta (slope), 

slänt (slope)) 

long thin form (slang (tube), 

slimmad (slimmed), slana (scaffold 

pole), slejf (strap), slinga (coil), slips 

(tie)) 

slackness (slamsa (rag), slankig 

(limp), slana (scaffold pole), slejf 

(strap), slinga (wreath), slips (tie)) 

The relations are the following: 

Wetness causes smooth surface structure (index). 

Smooth surface causes quick or strong movement (index). 

That which moves quickly often has a long thin form (index). 

That which has same form as long thin form is often slack (index). 

Meaning hierarchies? 

One aim of the analysis was to find meaning hierarchies within clusters, 

e.g. that sound symbolism presupposes sound imitation, that movement 

presupposes sound imitation, that wetness always implies pejorative, but 

such strong claims can not be made. However, there are strong 

tendencies: 

For all 36 clusters 'pejorative' is the largest category, followed by 

'sound'. As mentioned earlier, clusters are often both 'sound' (or 

'talking') and 'pejorative'. Exceptions are br-, fr-, kv- which are 

'sound' but not 'pejorative'. Spj-, spl-, spr- are neither 'sound' nor 

'pejorative'. 

Clusters with 'quick or strong movement' also have the semantic feature 

'sound' (things that move quickly often sound) except for one cluster: 

spr-. 'Wetness' is almost always co-occurring with 'sound imitation', also 

probably due to the fact that wet things often sound.

Some clusters are used for a variety of meanings (sl-, kl-) while others 

are more specialized on one meaning (fj-, pj-). 

4.7 Discussion and conclusions 

This chapter has shown that certain semantic features are connected with 

certain initial consonant clusters, in different combinations and to 

different extents. From the analysis of sound symbolism of the initial 

consonant clusters of Swedish some general conclusions can be drawn: 

* There is a difference between clusters in the number of root 

morphemes that are motivated. This is partly due to how lexically 

frequent the cluster is. There is also a difference between clusters in how 

big a proportion of the root morphemes are motivated. Some clusters are 

simply more used for motivated root morphemes. The variation is 

between 8% (br-) and 100% (fn-). It seems clear that lexically infrequent 

clusters are exploited for sound symbolism to a higher degree. 

(One might ponder about the importance of total or percental number for 

the language user's feeling of degree of motivation of a cluster. 

Nevertheless this analysis has mostly dealt with percentages. Another 

related issue is the issue of textual frequency of motivated root 

morphemes, which has not been dealt with here.) 

* The meaning profiles for most clusters differ. Different clusters (and 

different cluster types) seem to be fitting for different (types of) 

meanings. 

* There are often, on a more detailed level of analysis, interesting 

relations between the different root morphemes classified as belonging to 

the same phonestheme, and sometimes there is a connection between the 

different meanings of a cluster profile, see the analysis of gl-, bl- gnand 

sl- above. The meanings are related indexically or iconically and can 

be seen as belonging to the same meaning potential and activated in 

different contexts. It is a question of detail of analysis how many 

meanings are described. 

140

* There is also a frequency difference between the meanings. Some 

meanings are quite frequent, e.g. 'sound' and 'pejorative', other meanings 

are less frequent, e.g. 'adhesion' and 'round form'. 

* Some meanings seem to combine more often with other meanings, e.g. 

pejorative + walking or talking. 

The next chapter will present an analysis of sound symbolism of final 

consonant clusters and of vowels where phonesthemes connected with the 

semantic features found in the analysis of initial consonant clusters will be 

searched for. The results of the analysis of initial consonant clusters (as 

well as that of final clusters) have been the base for the experiments with 

neologisms presented in chapter 7. It was also the base for part of the 

cross linguistic studies presented in chapter 6. 

141

5. Analysis of final consonant clusters and 

vowels and of combinations 

It seems clear that a greater or lesser part of the sound symbolic 

meaning of root morphemes can be attributed not only to an initial 

consonant sequence but also to the final consonants, sometimes in 

combination with vowels (in different positions). It could be possible 

that the semantic feature 'quick or strong movement' of e. g. fladdra 

(flutter) is attributable not only to fl- (cf. 4.3.4) but also to -dr-. For all 

words ending in -dr- the dominating semantic feature is 'talking' 

(bladdra (talk nonsense), pladdra (babble), sladdra (chatter), bluddra 

(talk nonsense), sluddra (slur one's words) but also 'quick or strong 

movement' (fladdra and bläddra (turn over the pages). So, in the case 

of fladdra, flis most strongly tied to 'quick and strong movement', but 

-dr- also adds to the meaning of the word. In addition it is possible that 

the vowels add to the sound symbolic flavor of a word, and that a 

neologism like fliddra would depict a yet quicker movement. 

To facilitate understanding of the study, the reader is reminded of the 

procedure of deciding which final clusters should be counted as tentative 

phonesthemes (cf. 3.1). When there are at least two 1 root morphemes 

ending with the same cluster (which is sometimes followed by an 

obligatory vowel, i. e. strictly speaking they are semifinal) and having 

similar meaning (ascertained by key words), one phonestheme is 

established. If there are roots (normally at least two of each) with 

different, sound symbolic, meanings, different phonesthemes are 

established, e.g. -mla 'talking' (svamla, mumla) and 'quick or strong 

movement' (famla, vimla, tromla, fumla, rumla, drumla, tumla). It can 

also be the case that words have clearly different meanings, e.g. blaska. 

There is one meaning 'splash' and one pejorative meaning for 

'newspaper'. This word is then counted as two roots, and the analysis 

proceeds as above. In many cases a word has more than one meaning, 

but only one of them is sound symbolic. 

1 Or, for lexically infrequent clusters, 1 root 

143

5.1 Final clusters 

It is not as easy to enumerate the final sequences as it is to list the initial 

sequences. In final sequences, morphological structure has to be taken 

into account since some sequences only occur in inflected or derived 

forms, e.g. -ndsk in bondsk. Many forms are difficult to evaluate, 

according to the discussion in Sigurd (1965, pp. 67–69), who adopts a 

set of mechanical rules which exclude certain forms (secondary forms) 

which can be assumed to break the natural phonotactic pattern. 

Thus, for practical reasons, final clusters were studied in the following 

ways. Primarily they were studied with the aid of 'Svensk 

Baklängesordbok' (1981), which is a list of most Swedish lemmas 2 

(ordered alphabetically after the endings of the words only). The final 

clusters that are followed by a have been analyzed, partly since the 

clusters were easy to find in that way, but mainly because some clusters 

cannot have word final position (e.g. C+r- and C+l-sequences). These 

are mainly verbs. (In addition, roots from Nusvensk Frekvensordbok 4 

(NFO4) were analyzed, see 5.3 and 5.5). The roots that were excerpted 

were those that have one or more of the semantic features discussed 

earlier. For this reason also single occurrences of final clusters with 

these semantic features were registered. The aim has been to use 

different materials in order to study as many roots as possible. Not all 

final clusters are sound symbolic 3, as almost all initial clusters are. 

The resulting 27 sound symbolic final clusters before a (semi final) are 

shown in table 5.1 below. There, geminate consonants, and not only 

consonant clusters, are included. The semantic features recurring more 

than once for each final cluster are within brackets. 

Table 5.1 The 27 sound symbolic final clusters (before a and preceded 

by a short vowel), from Svensk Baklängesordbok. 100% indicates that 

there is only one root. This root is in all indicated cases sound symbolic. 

All other clusters have at least two roots with similar meaning (or, for 

lexically infrequent clusters, 1 root) according to the method described 

in 3.1. 

2Svensk Baklängesordbok contains lemmas from SAOL and from newspaper articles, 

and does not exactly match Svensk Ordbok. 

3 Approximately 22% of the final clusters from NFO4 are sound symbolic, cf. table 

5.17. 

144

% 

/tSa/ (sound, quick or strong movement) 

/f:a/ (slang4, quick or strong movement) 

/ska/ (wetness, pejorative) 

/bla/ (talking) 

/fla/ (pejorative, talking) 

/Nla/ (quick or strong movement) 

/l:a/ (pejorative, round form, walking) 

/mla/ (quick or strong movement, talking, sound) 

/pla/ (quick or strong movement) 1/15 100% 

/rpla/ (sound) 1/1 100% 

/spla/ (quick or strong movement) 1/1 100% 

/rla/ (sound) 

/sla/ (sound, talking) 

/ampa/ (walking) 

/mpa/ (walking, pejorative, quick or strong 

movement, short-wide form) 

/p:a/ (quick or strong movement) 

/bra/ (talking) 1/1 100% 

/dra/ (talking, quick or strong movement) 

/indra/ (light) 

/lra/ (quick or strong movement, sound, talking) 

/imra/ (light) 

/r:a/ (quick or strong movement, sound, talking) 

/tra/ (quick or strong movement, sound, talking, 

pejorative) 

/fsa/ (pejorative, quick or strong movement, 

walking, sound) 

/msa/ (long thin form, pejorative, talking, quick or 

strong movement) 

/lta/ (walking, quick or strong movement) 

4 'Slang' is a stylistic feature and thus belongs to another dimension, cf. 4.2.3. 

5 1/1 means that there is one root morpheme which is sound symbolic. 

145

sta/ (talking, quick or strong movement) 

Examples of sound symbolic root morphemes ending with these final 

clusters from Svensk Baklängesordbok are found in table 5.2 below. 

Some final clusters are rare and occur in only one or a few words. These 

are (translated in the table below): klatscha, ratscha, rutscha (three out of 

four root morphemes), knaggla, raggla, traggla (three out of four root 

morphemes), porla, sorla (two out of four root morphemes) and stöppla, 

sörpla, haspla, slabbra (there is only one root morpheme for each of these 

last four final clusters). 

146

Table 5.2 Sound symbolic roots, 

translation and categorization. 

klatscha crack sound 

ratscha rip sound 

rutscha slide quick or strong 

movement 

haffa nab slang 

blaffa large (ugly) color slang 

patch 

klaffa tally slang 

fiffa smarten (up) slang 

sniffa sniff slang 

piffa smarten (up) slang 

skoffa shovel slang 

roffa rob slang 

buffa nudge slang 

skuffa push slang 

luffa tramp slang 

fluffa fluff (up) slang 

knuffa push quick or strong 

movement 

puffa push quick or strong 

movement 

ruffa foul slang 

gruffa squabble slang 

tuffa puff slang 

töffa puff slang 

blaska splash wetness 

plaska splash wetness 

sjaska soil pejorative 

slaska splash wetness 

smaska slurp wetness 

snaska munch wetness 

vaska wash wetness 

babbla babble talking 

rabbla rattle talking 

gaffla gabble talking 

taffla muck things up pejorative 

fiffla fiddle pejorative 

ruffla fiddle pejorative 

knaggla plod along movement 

traggla go on about talking 

dangla dangle quick or strong 

movem. 

rangla be lanky quick or strong 

movem. 

skrangla be rickety quick or strong 

movem. 

dingla dangle quick or strong 

movem. 

147 

pingla tinkle sound 

ringla coil quick or strong 

movem. 

kringla pretzel winding form 

vingla stagger quick or strong 

movem. 

fjolla foolish woman pejorative 

lolla scatter-brained 

woman 

pejorative 

rolla paint with a round form 

roller 

skrolla scroll round form 

stolla fool of a woman pejorative 

bulla bun round form 

lulla reel walking 

rulla roll round 

drulla blunder walking 

krulla curl round form 

famla grope quick or strong 

movem. 

ramla tumble falling 

skramla rattle sound 

svamla drivel talking 

strimla strip long thin form 

vimla swarm quick or strong 

movem. 

tromla rotating quick or strong 

cylindrical sieve movement 

fumla fumble quick or strong 

movem. 

humla bumble-bee sound 

mumla mumble talking 

rumla be on the spree quick or strong 

movement 

drumla blunder quick or strong 

movem. 

tumla tumble quick or strong 

movem. 

stöppla give painted quick or strong 

surface a certain 

look by striking a 

brush against it 

movement 

sörpla drink noisily sound 

haspla reel quick or strong 

movem. 

porla ripple sound 

sorla ripple sound 

rassla rattle sound 

prassla rustle sound 

tassla whisper talking 

gnissla squeak sound 

tissla whisper talking 

vissla whistle sound 

rossla wheeze talking 

klampa tramp walking

slampa slut pejorative 

trampa tramp walking 

stampa stamp walking 

dimpa tumble quick or strong 

movem. 

fimpa stub short-wide form 

limpa loaf short-wide form 

klimpa get lumpy short-wide form 

skvimpa splash to and fro quick or strong 

movement 

fjompa be silly pejorative 

dumpa dump slang 

gumpa jog quick or strong 

jumpa jump from one 

piece of floating 

ice to another 

movem. 


skumpa jog quick or strong 

movem. 

klumpa form lumps short-wide form 

plumpa make blots form 

pumpa pump quick or strong 

movem. 

rumpa rump slang 

sumpa blow a thing slang 

stumpa tiny tot diminutive 

klappa pat quick or strong 

movement 

snappa snatch quick or strong 

movem. 

rappa strike quick or strong 

movem. 

tappa drop quick or strong 

movem. 

greppa grip quick or strong 

movem. 

steppa tap-dance quick or strong 

movem. 

kippa flop about quick or strong 

movem. 

skippa skip quick or strong 

movem. 

trippa trip along walking 

tippa tip over quick or strong 

movem. 

vippa swing up and quick or strong 

down 

movem. 

hoppa jump quick or strong 

movem. 

loppa flea diminutive 

glopp sleet wetness 

moppa mop quick or strong 

movem. 

noppa pluck quick or strong 

movem. 

snoppa top and tail quick or strong 

movem. 

148 

poppa pop up quick or strong 

movem. 

droppa drip quick or strong 

movem. 

proppa cram quick or strong 

movem. 

stropp sling round form 

guppa jolt quick or strong 

movem. 

knäppa flick quick or strong 

movem. 

snäppa snap one's fingers quick or strong 

movem. 

slabbra chatter talking 

bladdra talk nonsense talking 

fladdra flutter quick or strong 

movem 

pladdra babble talking 

sladdra chatter talking 

bluddra talk nonsense talking 

sluddra slur one's words talking 

bläddra turn over the quick or strong 

pages 

movem. 

glindra gleam light 

tindra twinkle light 

dallra wobble quick or strong 

movement 

skallra rattle sound 

pillra potter at quick or strong 

movem. 

tillra trickle quick or strong 

movem. 

kvillra ripple, twitter sound 

jollra babble talking 

pjollra babble talking 

knollra curl round form 

bullra rumble sound 

mullra rumble sound 

myllra swarm quick or strong 

movem. 

bjällra bell sound 

skimra shimmer light 

flimra flicker quick or strong 

movem., light 

glimra gleam light 

darra tremble quick or strong 

movem. 

blarra talk nonsense talking 

knarra creak sound 

snarra burr talking 

irra wander about quick or strong 

movem. 

dirra tremble quick or strong 

movem. 

klirra jingle sound

knirra creak sound 

pirra tingle bodily feeling 

stirra stare gaze 

virra wander about quick or strong 

movem. 

kvirra make a fuss talking 

svirra whirl quick or strong 

movem. 

skorra burr sound 

morra growl talking 

knorra grouse talking 

burra ruffle up quick or strong 

movem. 

hurra hurrah talking 

kurra rumble sound 

plurra fall into the water quick or strong 

movement 

murra growl talking 

snurra spin quick or strong 

movem. 

surra hum sound 

tjattra jabber talking 

klattra be awkward pejorative 

smattra clatter sound 

knattra rattle sound 

snattra quack sound 

skvattra quack sound 

glittra glitter quick or strong 

movement 

splittra splinter quick or strong 

movement 

fnittra giggle talking 

knittra sound high 

pitched and 

sound 

iterated 

kvittra chirp sound 

klottra scrawl quick or strong 

movem. 

plottra potter about pejorative 

knottra get goose- rough surface 

pimples structure 

huttra shiver quick or strong 

movement 

kuttra coo sound 

muttra mutter talking 

puttra chug sound 

hafsa scamp a thing pej. quick or 

strong movement 

sjafsa shuffle quick or strong 

movement 

tjafsa talk drivel pej., talking 

lafsa shuffle pej., walking 

klafsa squelch sound, wetness, 


slafsa slop sound, pejorative 

149 

nafsa snap quick or strong 

movem. 

snafsa snap quick or strong 

movem. 

rafsa rummage quick or strong 

movem. 

krafsa scratch quick or strong 

movem. 

tafsa fiddle about with quick or strong 

a thing movem. 

lufsa lumber walking 

glufsa gobble down pejorative 

plufsa plop quick or strong 

movem. 

rufsa ruffle quick or strong 

movem. 

tufsa tousle quick or strong 

movem. 

gläfsa yelp sound 

räfsa rake quick or strong 

movem. 

jamsa talk nonsense talking, pej. 

flamsa fool about pejorative 

slamsa rag slackness, long 


ramsa string long (thin) form 

tramsa talk rubbish talking, pej. 

remsa strip long thin form 

slimsa rag long thin form 

plumsa plop quick or strong 

movem. 

mumsa munch quick or strong 

movem. 

grumsa grumble talking 

dalta coddle pejorative 

halta limp walking 

palta muffle up quick or strong 

movem. 

skralta be rickety quick or strong 

movem. 

skvalta ripple wetness, quick or 

strong movem. 

bulta beat beat 

rulta waddle walking 

tulta toddle about walking 

stulta toddle walking 

bylta muffle up quick or strong 

movem 

stylta walk stiff-legged walking 

gasta yell talking 

hasta hasten quick or strong 

movem. 

kasta throw quick or strong 

movem. 

rista cut quick or strong 

movem.

ista burst destruction 

hosta cough talking 

pusta pant talking 

frusta snort talking 

hysta throw quick or strong 

movem. 

150 

nysta wind quick or strong 

movem. 

knysta breathing a word talking 

krysta bear down talking 

5.2. Summary of the analysis of semantic features 

for final clusters 

The most common semantic feature of the final clusters is 'quick or 

strong movement', which is present in 15 final clusters. Table 5.3 

compresses all the features and final clusters of table 5.2. The table can 

also be compared with table 5.1, where instead the semantic features for 

each final cluster are shown. The semantic features are shown, in 

descending order, in table 5.3. 

Table 5.3 The semantic features of roots from Svensk Baklängesordbok. 

Listed are also the clusters corresponding to each feature and the total 

frequencies of clusters. 

Semantic feature Frequency of 

final clusters 

Final clusters6 quick or strong movement 85 ffa, Nla, pla, spla, 

mpa, mla, ppa, dra, 

lra, rra, tra, fsa, msa, 

lta, sta 

talking 37 bla, fla, mla, sla, bra, 

lra, rra, msa, dra, 

tra, sta, ska 

sound 34 tSa, mla, rpla, rla, 

sla, lra, rra, tra, fsa 

pejorative 20 fla, lla, mpa, tra, fsa, 

msa 

slang 19 ffa, mpa 

walking 15 lla, ampa, fsa, lta 

wetness 9 ska 

long thin form 5 msa 

round form 7 lla 

light 5 indra, imra 

short-wide form 4 mpa 

6 Double consonant grapheme stands for a phonologically long consonant.

In table 5.4 the ranking of table 5.3 is compared with diagram 4.5, which 

shows the most frequent meanings for i n i t i a l consonant clusters. 

Table 5.4 The most frequent semantic features for initial and final 

clusters. 

rank initial clusters final clusters 

1 pejorative quick or strong 

movement 

2 sound talking 

3 long thin form sound 

4 quick or strong pejorative 

movement 

5 wetness slang 

6 talking walking 

7 light wetness 

8 diminutive long thin form 

9 round form round form 

10 walking light 

11 destruction short-wide form 

The semantic features in bold type are those nine that are among the 

eleven most common features for both groups. 'Sound' is ranked second 

for initial clusters and third for final clusters. 'Quick or strong 

movement' is ranked first for final clusters and fourth for initial clusters. 

'Pejorative' is ranked fourth for final clusters but first for initial clusters. 

The semantic features of the six most frequent final clusters are among 

the semantic features of the ten most frequent initial clusters. In other 

words, many of the most common semantic features are the same for 

initial clusters and final clusters. However, 'pejorative' is not as common 

in final clusters as in initial. 

Some clusters can occur both initially and finally (before a), namely sk, 

bl, fl, spl, sl, br, dr, tr and st. Four of these can have the same 

semantic feature. These are bl (talking), fl (pejorative), sl (talking), tr 

(quick or strong movement, sound, talking, pejorative). However, there 

are no conventional words that both begin and end with these 

combinations. 

151

5.3 Properties of consonant clusters of Nusvensk 

Frekvensordbok 

Using the root analysis with paraphrases of Nusvensk Frekvensordbok 4 

(NFO4), an analysis of final clusters that resembles the analysis of initial 

clusters was made, i.e. sound symbolic roots were excerpted from a 

larger set of roots with the aid of key words (cf. Appendix 1). NFO4 is a 

frequency dictionary of Swedish morphemes, and the material consists of 

1 million words from newspaper articles. List 4.3 of NFO4 contains the 

roots ordered in reverse. The reason not to analyze only the NFO4 

material is that there are many more words in Svensk Baklängesordbok. 

The roots from the NFO4 material are chosen in the following way: they 

shall contain at least one root ending of a cluster, if this root belongs to 

one of the semantic categories listed in chapter 4. 

The most common clusters, absolutely and percentally, are shown in 

diagrams 5.1 and 5.2, and in the following diagrams (5.3 to 5.11) the 

semantic profiles of some of the most common clusters are shown. 

The first diagram (5.1) shows how the clusters rank when the total 

number of sound symbolic roots are counted. -Nk and -sk are the clusters 

with the largest number of sound symbolic roots. Thereafter follow - fs, - 

nd and -tr. -Nk, -sk, -fs and -tr are described in detail below. Diagram 

5.1 corresponds to diagram 4.1 of initial clusters. 

The next diagram (5.2) shows the clusters that have the highest degree of 

sound symbolic roots. The first five have a percentage of 100% because 

there is only one – sound symbolic – root, for every cluster. These are 

not studied in further detail. The following ones with quite a high 

percentage are -fs, -dr and -lr, followed by -ml and -Nl, These are 

described in detail below. Also -bl is described. Diagram 5.2 corresponds 

to diagram 4.4 of initial clusters. 

152

20 

18 

16 

14 

12 

10 

8 

6 

4 

2 

0 

ksn 

ltn 

spl 

mj 

rS 

jf 

lb 

mt 

lt 

ps 

rl 

Sk 

rj 

ld 

ls 

lm 

rv 

nt 

ns 

ms 

sl 

Nl 

bl 

lk 

lr 

dr 

st 

ml 

mp 

tr 

nd 

fs 

sk 

Nk 

Diagram 5.1 7 More and less sound symbolic final clusters. Number of motivated root morphemes per cluster. 

7 In this and the following diagrams "N" stands for [N] and "S" stands for [S].

100 

90 

80 

70 

60 

50 

40 

30 

20 

10 

0 

nt 

lt 

st 

ns 

rj 

ld 

nd 

rv 

ps 

rl 

mt 

ls 

lk 

lm 

mp 

sk 

bl 

ms 

sl 

tr 

Nk 

Sk 

Nl 

ml 

lr 

dr 

ksn 

fs 

ltn 

spl 

mj 

rS 

jf 

lb 

Diagram 5.2 Percent of motivated root morphemes per cluster, for final consonant clusters.

The following diagrams, 5.3 to 5.11 show the semantic profiles for 

different final clusters. These diagrams can be compared with the 

diagrams 4.16 – 4.23, which show the semantic profiles for different 

initial clusters. Diagram 5.3 shows the semantic profile of -fs. 

60 

50 

40 

30 

20 

10 

0 

pejorative sound quickness 

Diagram 5.3 Semantic features of the cluster -fs. Percentages concern the 

number of semantic features in proportion to all roots of the cluster. 

Diagram 5.3 shows the semantic features of the consonant cluster -fs. The 

words (representing root morphemes) of the dominating features are 

shown in table 5.5. The absolute frequencies (not the percentages) of the 


Table 5.5 The sound symbolic morphemes of the -fs cluster, the semantic 


'pejorative': hafs (slovenliness), tjafs 

(drivel), slafs (sloppiness), rafs (trash), 

krafs (knick-knacks), lufsa (shamble), 

kalufs (forelock), rufsa (ruffle), 

tufsa (tousle), bjäfs (gewgaws) 10 

'sound': klafs (squelch), gläfs (yelp) 2 

'quickness': nafs (snap), rafs (trash) 2 

155 

%

10 

9 

8 

7 

6 

5 

4 

3 

2 

1 

0 



movement 

sound 

Diagram 5.4 The semantic profile of -Nk. Percentages concern the 


Diagram 5.4 shows the semantic features of the consonant cluster -Nk. In 

table 5.6. the words (representing root morphemes) of the dominating 



Table 5.6 The sound symbolic morphemes of the -Nk cluster, the semantic 


'long thin form': bank (bar), dank 

(thin candle), hank (a band), planka 

(plank), rank (tall and slender), spinkig 

(slender), rynka (furrow) 7 

'quick or strong movement': flink 

(quick), slink (slip), vink (wave) 3 

beat 

156 

slang 

wetness 

%

'sound': banka (knock), dunk (thumping), 

stånka (puff and blow) 3 

'beat': banka (knock), dunka (thump) 2 

'slang': pank (broke), grunka (gadget) 2 

'wetness': klunk (gulp), stänk (splash) 2 

The following diagram (5.5) shows the semantic profile of -sk. 

8 

7 

6 

5 

4 

3 

2 

1 

0 

wetness pejorative beat sound 

Diagram 5.5 The semantic profile of -sk. Percentages concern the number 

of semantic features in proportion to all roots of the cluster. 

Diagram 5.5 shows the semantic features of the consonant cluster -sk. The 

words (representing root morphemes) of the dominating features are 



Table 5.7 The sound symbolic morphemes of the -sk cluster, the semantic 


'wetness': blask (slops), plask (splash), 

slask (slush), smaska (guzzle), vaska 

(wash), loska (spit), mäsk (mash), 

träsk (swamp) 8 

157 

%

'pejorative': sjask (cad), fnask 

(prostitute), snaska (be messy), fnoskig 

(dotty), fjäsk (fawning) 5 

'beat': daska (slap), plask (splash), piska (whip) 3 

'sound': pladask (flop), smaska (guzzle) 2 

The following diagram (5.6) shows the semantic profile of -tr. 

14 

12 

10 

8 

6 

4 

2 

0 

talking sound pejorative 

Diagram 5.6 The semantic profile of -tr. Percentages concern the number 

of semantic features in proportion to all roots of the cluster. 

The diagram 5.6 shows the semantic features of the consonant cluster -tr. 

In table 5.8 the words (representing root morphemes) of the dominating 

features can be studied. The absolute frequencies (not the percentages) of 

the semantic features are shown to the right. 

Table 5.8 The sound symbolic morphemes of the -tr cluster, the semantic 


'talking': tjattra (jabber), snattra (quack), 

fnittra (giggle), muttra (mutter), 

yttra (utter) 5 

158 

%

'sound': smattra (clatter), knattra (rattle), 

kvittra (chirp), puttra (simmer) 4 

'pejorative': tjattra (jabber), plottra (fritter) 2 

The following diagram (5.7) shows the semantic profile of -Nl. 

40 

35 

30 

25 

20 

15 

10 

5 

0 

quick or strong movement 

Diagram 5.7 The semantic profile of -Nl. The percentage is of the 

semantic feature, for all roots of the cluster. 

Diagram 5.7 shows the semantic feature of the consonant cluster -NNl. The 

words (representing root morphemes) of the only feature is shown in 

table 5.9. The absolute frequency (not the percentage) of the semantic 

feature is shown to the right. 

159 

%

Table 5.9 The sound symbolic morphemes of the -Nl cluster, the semantic 

feature involved, and its absolute frequency. 

'quick or strong movement': dingla 

(dangle), singla (dance), vingla (wobble), 

jonglera (juggle) 4 

The following diagram (5.8) shows the semantic profile of -bl. The 

percentage is of the semantic feature for all roots of the cluster. 

25 

20 

15 

10 

5 

0 


Diagram 5.8 The semantic profile of -bl. The percentage is of semantic 

features for all roots of the cluster. 

Diagram 5.8 shows the semantic feature of the consonant cluster -bl. The 

words (representing root morphemes) of the only feature is shown in 

table 5.10. The absolute frequency (not the percentage) of the semantic 

feature is shown to the right. 

Table 5.10 The sound symbolic morphemes of the -bl cluster, the 


'talking': babbla (babble), rabbla (rattle), 

käbbla (bicker), jubla (shout with joy) 4 

160 

%

The following diagram (5.9) shows the semantic profile of -dr. 

25 

20 

15 

10 

5 

0 

quick or 

strong 

movement 

talking pejorative 

Diagram 5.9 The semantic profile of -dr. Percentages concern the 


This diagram shows the semantic features of the consonant cluster -dr. In 

table 5.11 the words (representing root morphemes) of the dominating 



Table 5.11 The sound symbolic morphemes of the -dr cluster, the 


'quick or strong movement': fladdra (flutter), 2 

bläddra (turn over the pages) 

'talking': pladdra (babble), sluddra (slur) 2 

'pejorative': pladdra (babble), sluddra (slur) 2 

The next diagram (5.10) shows the semantic profile of -lr. Percentages 

are of semantic features for all roots of the cluster. 

161 

%

45 

40 

35 

30 

25 

20 

15 

10 

5 

0 

sound quick or 

strong 

movement 

Diagram 5.10 The semantic profile of -lr. Percentages are of semantic 

features for all roots of the cluster. 

Diagram 5.10 shows the semantic features of the consonant cluster -lr. 

The words (representing root morphemes) of the dominating features are 



Table 5.12 The sound symbolic morphemes of the -lr cluster, the 


'sound': skallra (rattle), bullra (make a noise), 

mullra (rumble), bjällra (jingle) 4 

'quick or strong movement': dallra (tremble), 

myllra (swarm) 2 

The last diagram (5.11) shows the semantic profile of -ml. 

162 

%

20 

18 

16 

14 

12 

10 

8 

6 

4 

2 

0 

quick or 

strong 

movement 

movement sound 

Diagram 5.11 The semantic profile of -ml. Percentages concern the 


Diagram 5.11 shows the semantic features of the consonant cluster -ml. 

The words (representing root morphemes) of the dominating features are 



Table 5.13 The sound symbolic morphemes of the -ml cluster, the 


'quick or strong movement': vimla (swarm), 3 

fumla (fumble), tumla (tumble) 

'movement': famla (grope), ramla (fall down) 2 

'sound': skramla (rattle), humla (bumble-bee) 2 

163 

%

5.3.1 Summary of the analysis of properties of 

different final clusters 

As well as for the initial clusters, the meaning profiles for the sound 

symbolic final clusters vary. Different clusters are connected with 

different meanings. Also the final clusters have different semantic 

profiles. Some final clusters have several semantic features, others have 

only one. 

Of these absolutely and percentally most common clusters (-Nk, -sk,-fs, - 

nd, -tr, -dr, -lr, -ml, -Nl), 5 out of 9 end with a liquid and are mainly 

connected with 'quick or strong movement', 'talking', 'sound', and 

'pejorative'. -sk is dominated by 'wetness' and -fs is clearly 'pejorative' 

5.4. Vowels 

As mentioned earlier, vowels may also have sound symbolic meaning. To 

delimit the analysis of vowels, the contrastive effect of different vowels 

was studied. 

5.4.1 Vowel pairs and triplets 

In order to investigate the semantic effect of different vowels, word pairs 

were searched for. However, there are not so many cases of word pairs 

with contrasting vowels. Those that were found contrast the vowels [i]-[a], 

[i]-[a]-[u], [i]-[o], [i]-[P]. Most of them imitate sound (8 out of 19). The 

examples (taken from Svensk Baklängesordbok) are: klibba-klabba 

(adhesion 8), slibba-slabba (wetness), slidder-sladder 9 (talking, pejorative), 

klifsa-klafsa (wetness, sound), klitsch-klatsch (sound), ritsch-ratsch 

(sound), snick-snack (talking), ringla-rangla (movement), tissla-tassla 

(talking), rispa-raspa (rough surface structure), dirr and darr 

(movement), knirra-knarra (sound), knistra-knastra (sound), knittraknattra 

(sound), slimsa-slamsa (form) snipp-snapp-snorum (in nonsense 

verse), dripp-dropp (sound), tick-tock (sound), klimp-klump (form). 

From the definitions in Svensk Ordbok it is not evident that there is a 

meaning distinction (size or pitch) between [i] and [a] except for some 

cases, but clearly [i] never stands for anything large or low-pitched. (For 

an investigation of English reduplicatives see Thun, 1963.) 

8Shown within parenthesis is the semantic feature for both words in the pair. Some 

words share two semantic features. 

9 In fact, many of these pairs occur as semi-lexicalized compounds, e.g. in poetry and 

verse. 

164

However, there are a number of words found in Svensk Baklängesordbok 

which show that root morphemes with endings containing [i] often have 

the meanings 'smallness', 'quickness' or 'high pitch', while [P] has the 

meanings of darker sounds or bigger movements – while [a] seems quite 

neutral. Comparing the endings [inka] and [Pnka] in table 5.14, there is, in 

blinka (twinkle) and lunka (jog along), a contrast in small, quick 

movement versus big, slow movement. In klinka (tinkle) and dunka 

(thump) there is a contrast between high pitched and low pitched sound. 

Table 5.14 The vowels i, P, a before the ending -NNka 

kinka whine 

linka limp 

blinka twinkle 

klinka tinkle 

slinka slip 

spinka split (wood) 

vinka wave (hand) 

dunka thump 

pjunka coddle 

lunka jog along 

klunka gulp down 

runka shake 

banka bang 

klanka grouse 

planka plank 

slanka make slim 

ranka creeper 

vanka saunter 

svanka be sway-backed 

The vowels with the ending -lra can also be compared. Most of the -lra 

words imitate sounds or movements where i clearly stands for smallness 

or high pitch and [P] for low pitch, e.g. tillra (trickle) and mullra 

(rumble). 

165

Table 5.15 The vowels a, i, O, P before the ending -lra. 

dallra quiver 

skallra rattle 

pillra potter at 

tillra trickle 

kvillra ripple, twitter 

jollra prattle 

pjollra twaddle 

knollra frizz 

bullra rumble 

mullra rumble 

The i and [P] vowels in words ending with -ttra show similar tendencies, 

cf. e.g. kvittra (twitter) and muttra (mutter). 

Table 5.16 The vowels a, i, O, P before the ending -tra. 

tjattra chatter 

klattra make a fuss 

smattra clatter 

knattra rattle 

snattra quack 

skvattra quack 

glittra glitter 

splittra splinter 

fnittra giggle 

knittra sound high 

pitched and 

iterated 

kvittra twitter 

knottra get goosepimples 

huttra shiver 

kuttra coo 

muttra mutter 

puttra chug 

5.4.2 Vowels in light/gaze-words 

The following list contains most light or gaze root morphemes: 

blek (pale), blick (gaze), bliga (stare), blind (blind), blinka (blink), blixt 

(lightning), blossa (flare), blända (blind), blänga (glare), blänka (shine), 

flimra (flicker), glana (stare), glans (luster), glatt (glossy), glimma 

(gleam), glimra (gleam), glimta (gleam), glindra (gleam), glisa (shine), 

glittra (glitter), glo (stare), gloria (halo), glutta (take a glance), glåmig 

166

(pale), glänsa (shine), glöd (glow), gnista (spark), gnistra (sparkle), tindra 

(twinkle), skimra (shimmer). 

The stressed vowels of these 30 root morphemes are in 16 cases i, in 1 

case e, in 4 cases E, in 1 case Ø, in 3 cases a (2 [a] and 1 [A:]), in 2 cases o, 

in 2 cases u, and in 1 case [P]. The dominating vowel is clearly i. 

Furthermore, 24 of the vowels are front vowels; only 5 are back vowels 

and 1 is central. It seems thus that the feature 'light' is connected with 

front vowels (high F2). 

5.4.3 The vowel [P] 

Since there seemed to be a tendency for words with [P] (a short, medial, 

half closed, rounded vowel) to have a pejorative meaning (see table 5.2), 

this vowel was studied more closely. In the whole material (Appendix 1) 

the number of pejorative features is 163. Of these features 33 belong to 

roots with [P], i.e. 20%. The number of non-pejorative features is 789 and 

70 of these belong to roots with a [P], i.e. 8%. A Chi square test showed 

that there is a significant correlation between 'pejorative' and [P] (p. < 

0.0001). 

5.4.4 Summary of vowels 

From the examples in this material it seems quite clear that stressed 

vowels are connected with the semantic dimensions 'size' and 'high-low 

pitch' (which are connected through the frequency code); i tends to have 

the meanings 'smallness', 'quickness', 'high pitch', while [P] seems to have 

the meanings of 'low pitch' and 'largeness'. a seems quite neutral. The 

special study of [P] showed that this sound is over-represented among 

roots with the pejorative feature. 

5.5. Comparisons of final clusters of roots from 

different sources 

5.5.1 NFO4 and Sigurd (1965) 

Since the question of which final combinations should be taken into 

account is somewhat open, a comparison is made between NFO4 and 

Sigurd (1965). Table 5.17 shows the final clusters which are found in the 

two different sources. The comparison with Sigurd (1965) is included to 

remind the reader that the final clusters of NFO4 are not exactly the same 

167

as those of other analyses and to show which clusters are found in both 

sources. 

168

Table 5.17 The final clusters of 

NFO4, those of NFO4 that are 

sound symbolic, and, as a 

comparison, the primary final 

clusters according to Sigurd 

(1965). 

Clusters 

for all 

reverse 

order 

sorted 

roots of 

NFO4 

Clusters 

for sound 

symbolic 

roots 

from 

NFO4 

lb + + 

mb + 

rb 

gd + 

jd + 

ld + + 

md + 

nd + + 

vd + 

jf + + 

lf + 

mf + 

rf + 

sf 

rg + 

dg 

rj + + 

ntg 

nS + 

rk 

nSS 

rS + + 

tS + 

dj 

lj + 

mj + + 

nj + 

rj + 

lC 

tj 

vj 

jk + 

lk + + 

mk 

Nk + + 

rk + 

sk + + 

lsk + 

msk + 

Primary 

clusters 

according 

to Sigurd 

169 

nsk + 

psk + 

Sk + + 

bl + 

mbl 

ndl 

fl 

gl 

Nl + 

rjl 

gl 

jl 

kl 

NNkl 

rkl 

ml + 

pl 

mpl 

spl + 

rl + 

Nsl 

jsl 

nsl 

psl 

sl + 

ntl 

stl 

vl 

rvl 

ksl 

gm 

hm 

km 

lm + + 

rm + 

sm 

stm 

tm 

Nn + 

jn 

kn 

ln + 

mn + 

pn 

rn 

sn 

tn 

ltn + 

ksn + + 

jp 

lp 

mp + + 

rp +

sp + 

br 

rbr 

dr + 

ldr 

ndr 

dr + 

rdr 

fr 

gr 

kr 

Nkr 

lr + 

mr 

nr 

pr 

tr + 

rtr 

str 

kstr 

vr 

lvr 

bs 

ds 

fs + + 

js + 

ks + 

ls + + 

ms + + 

ns + + 

Ns + 

ps + + 

mps 

rps 

ts + 

ft + 

rft + 

kt 

Nt 

jt + 

kt + 

Nkt 

rkt 

lt + + 

mt + 

nt + + 

pt + 

st + + 

fst + 

Nst + 

jst + 

kst + 

mst + 

170 

nst + 

St + 

kv 

lv + 

rv + + 

nts

Table 5.17 shows the following: There are, in the NFO-material, 151 

different endings of roots. Of these, 33 are used systematically for sound 

symbolism. 

The rightmost column shows which of these clusters are primary final 

combinations according to Sigurd (1965). Quite a few root final 

combinations in NFO4 are not the same as Sigurd’s word final 

combinations. Also some of Sigurd’s final combinations do not occur in 

the NFO-material: tsk, pst, lft, lst, NNd, NNt. 

5.5.2 The most frequent final clusters in NFO4 

The following tables (5.18 and 5.19) show which final clusters are the 

most frequent for different semantic features, both absolutely and 

percentally. They show that fs is the most sound symbolic final cluster 

(pejorative) both in absolute numbers and percentally. 

Table 5.18 The most frequent final clusters in absolute numbers. Their 


Cluster Cluster 

frequency 

Semantic feature 

fs 10 pejorative 

sk 8 wetness 

Nk 

sk 

tr 

st 

7 

5 

5 

5 


pejorative 


sound 

lk 

bl 

Nl 

sl 

4 

4 

4 

4 




movement 

sound 

lr 4 sound 

tr 

ms 

4 

4 

sound 

pejorative 

171

Table 5.19 The most frequent final clusters in percent 10 of the total 

number of roots for clusters. Their semantic features are shown to the 

right. 

Cluster % Semantic feature 

fs 59 pejorative 

lr 44 sound 

Nl 40 quick or strong 

movement 

Sk 29 attitude 

sl 27 sound 

ms 27 pejorative 

dr 25 quick or strong 

movement 

dr 25 talking 

dr 25 pejorative 

bl 24 talking 

lr 22 quick or strong 

movement 

The accumulated frequencies of table 5.18 show that the most common 

semantic feature is 'pejorative' with 19 roots (distributed over 3 clusters) 

followed by 'sound' with 17 roots (distributed over 4 clusters), 'long thin 

form' with 11 roots distributed over 2 clusters, and 'talking' with 9 roots 

distributed over 2 clusters. 

5.5.3 Discussion of final clusters in Svensk 

Baklängesordbok and in Nusvensk Frekvensordbok 

As pointed out in the beginning of this chapter, it is not self-evident which 

are the final clusters in Swedish. The semifinal clusters (before a) of 

Svensk Baklängesordbok (the Reverse Order Dictionary), the final 

clusters of Nusvensk Frekvensordbok 4, and of Sigurd (1965) are 

partially overlapping sets. For practical reasons, percentages are 

calculated only on Nusvensk Frekvensordbok 4. 

Likewise, the roots of the two dictionaries also constitute partially 

overlapping sets. The lemmas (from which roots were extracted) of 

Svensk Baklängesordbok come from the Wordlist of the Swedish 

Academy (Svenska Akademiens ordlista över svenska språket, 1974) and 

10 The following clusters have a 100%-frequency because they have only one root, which 

is sound symbolic: lb (round form), jf (long thin form), rS (walking), mj (bodily 

feeling), spl (quick or strong movement), ltn (destruction) and ksn (mental feeling). 

172

from the same newspaper material as the material of Nusvensk 

Frekvensordbok. Consequently Svensk Baklängesordbok contains more 

roots. The similarities and the differences in the analyses of the two 

dictionaries are as follows (cf. also tables 5.4 and 5.24). 

The most common semantic features 

The most common semantic features are, in order: 

Svensk Baklängesordbok: 'quick or strong movement', 'talking', 'sound', 

'pejorative', 'slang', 'walking', 'wetness', 'long thin form', 'round form', 

'light', 'short-wide form'. 

Nusvensk Frekvensordbok 4: 'pejorative', 'sound', 'long thin form', 

'talking', 'wetness', 'quick or strong movement'. 

For both dictionaries the following features are the same: 'quick or strong 

movement', 'talking', 'sound', 'pejorative', 'wetness' and 'long thin form'. 

(These six features are also the six most common features of initial 

clusters, however in a different order, cf. table 5.4). 

The most common final consonant clusters (including all sound symbolic 

semantic features) are, in descending order (tables 5.20 and 5.21). 

Table 5.20 The most common final consonant clusters (for all sound 

symbolic semantic features) in Svensk Baklängesordbok. Cf. table 5.2. 

mp 20 

fs 18 

tr 18 

ml 13 

st 12 

lr 12 

lt 11 

ms 10 

Nl 8 

sl 7 

dr 7 

sk 6 

fl 4 

tS 3 

173

In Nusvensk Frekvensordbok 4 the most common final consonant clusters, 

including all sound symbolic semantic features, are (cf. diagram 5.1): 

Table 5.21 The most common final consonant clusters (for all sound 

symbolic semantic features) in Nusvensk Frekvensordbok 4. 

Nk 19 

sk 18 

fs 14 

nd 12 

tr 11 

mp 8 

ml 7 

st 7 

dr 6 

lr 6 

As can be seen in the two lists above, many, but not all, of the most 

common clusters are the same, namely mp, fs, tr, ml, st, lr, dr, and sk, 

in slightly different order. 

The following clusters used sound symbolically are found in Svensk 

Baklängesordbok but not in Nusvensk Frekvensordbok 4: fl, tS, gl, pl, 

rpl and br. 

The differences between the two dictionaries seem to be greater for the 

consonant clusters than for the semantic features, but the most common 

consonant clusters are to a great extent the same. Studying not only the 

most common clusters, there are more sound symbolic clusters in Svensk 

Baklängesordbok. This might be due to the fact, described above, that 

Svensk Baklängesordbok is based on a larger selection. 

5.6 Combinations of initial and final consonant 

clusters 

Initial and final clusters can combine to create meaning, e.g. 'wetness' as 

in blaska, plaska, slaska, smaska, snaska. The final combinaton -ska is 

connected with 'wetness' and the initial clusters bl-, pl-, sl-, sm- and sncan 

also stand for 'wetness' (among other things). 

174

The final clusters are combined in different ways with the initial clusters, 

apart from with single consonants. There are four groups: 

1. initial cluster + no final cluster (e.g. skval 11) 

2. no initial cluster + final cluster (e.g. babbla) 

3 a. initial cluster + final cluster with the same meaning 

(e.g. blaska, wetness+wetness) 

3 b. initial cluster + final cluster with a different meaning 

(e.g. drumla, pejorative + quick or strong movement) 

The analysis of group 1 is based on the analysis of chapter 4 (of Svensk 

Ordbok). The following analysis of groups 2 and 3 is based on the final 

clusters of Svensk Baklängesordbok. As the analysis in 5.5. shows, the 

most common semantic features are the same for both initial clusters and 

final clusters. 

5.6.1 Initial cluster + no final cluster 

In order to study sound symbolism of the initial clusters isolated from 

final clusters (cf. discussion in 4.1), the roots with a sound symbolic 

initial cluster n o t ending in a final cluster were studied more closely. 

Table 5.22 shows the frequencies of semantic features when roots ending 

with clusters or geminates are not taken into account. The table shows the 

semantic features, the frequencies and root examples, and the clusters are 

ranked in frequency order. The words are extracted from Appendix 1. 

The roots of Table 5.22 are a subset of the roots in Appendix 1. Table 

5.22 thus only shows the roots which do not end with a final cluster or 

geminate, whereas Appendix 1 contains roots both with and without final 

clusters or geminates. 

11 It is possible that there is sound symbolism also in single consonants, initially as well 

as finally. For example, [S] could have the semantic feature 'separation' as in skiva, 

skilja, skinna. Also, singular sounds with special characteristics could attract sound 

symbolic meanings, e.g. [r] is a sound which is different from other sounds and [C] has 

low frequency in Swedish. This question is not investigated in this thesis, but is left for 

further research. 

175

Table 5.22 Combinations of initial cluster + no final cluster. The 

recurring semantic features of initial clusters in frequency order. Listed 

to the right are the roots. 

features freq. the words 

pejorative 45 bliga, drasut, dröse, fjäsa, 

flina, glop, glupa, glåpord, 

gnat, gneta, gnidare, gräma, 

gräslig, klåpa, knal, knasig, 

kneg, kneken, knöl, knös, 

krake, kreta, kruserlig, 

kräk, mjäkig, pjåkig, pryl, 

skral, slas, slasa, slok, slyna, 

slö, slösa, sniken, snyta, 

snål, snöd, snöpa, spoling, 

strul, tradig, trasa, vräkig, 

vrövel 

long thin form 36 skåra, slana, släde, snabel, 

sno, snok, snöre, spene, 

speta, spik, spila, spira, 

spole, spö, spjut, spjäla 

spröt, stake, stav, stig, 

stilett, stör, streamer, 

strigel, stril, stripa, strut, 

strå, stråk, stråle, stråt, 

truta, tryne, tråd, tråg, trål 

talking 23 gnat, gnola, gny, gnöla, 

gråta, gräla, klaga, knota, 

kverulans, prat, pruta, 

skri, skrodera, skryta, 

skrål, skräna, skröna, 

smäda, snäsa, svada, svära, 

vrål, vrövel 

sound 21 braka, brus, bräka, bröl, 

fnysa, frasa, fräsa, gnoding, 

knaka, knot, krakel, kras, 

kråka, kvida, kväda, kväka, 

skval, snusa, snyta, stön, 

stril 

176


movement 

18 flanera, fluga, flyga, flyta, 

flåsa, flöda, frusa, gnida, 

gno, skaka, slira, slita, 

spruta, stöt, svepa, sväva, 

vrida, vräka 

wetness 15 blod, blöt, klena, kleta, 

kram, kräm, skval, smet, 

snor, snuva, spad, spola, 

sputum, spruta, stril 

thin form 13 flaga, flak, flarn, flisa, fnasig, 

krakmandel, krokett, 

krusta, krustad, spade, 


structure 

spat, spatel, spån 

11 glací, glaciär, glas, glatt, 

glida, glissando, gnida, gno, 

slipa, slät, spegel 

destruction 11 bryta, bråte, bräck, 

bränning, frat, fräta, kvarn, 

skrot, skräp, trasa, vrak 

light 10 blek, bläs, glacé, glisa, 

gloria, glåmig, glöd, prål, 

spat, spraka 

gaze 10 bliga, glana, glaukom, glo, 

glosögd, glosa, plira, snegla, 

spana, speja 

diminutive 8 gli, glipa, knåp, krabat, 

smula, små, späd, vret 

walking 8 knata, krypa, kräla, snava, 

strosa, ströva, trava, träda 

round form 7 blåsa, knop, knota, knödel, 

knöl, kruka, skopa 

winding form 7 krok, krokan, krusa, kråma, 

kräla, krök, spiral 

bad mood 6 gräla, trulig, tråkig, träta, 

vrede, vresig 

hollow form 5 grav, grop, gryt, gräva, 

gröpa 

form 4 glipa, glob, plös, pryl 

bodily feeling 4 kval, kvav, kväva, svida 

soft consistency 4 mjuk, plym, plymå, plysa 

fine grain 4 mjäla, mjöl, stöv, strö 

rough surface structure 4 fryna, krås, skråma, sträv 

separation 4 spagat, split, spreta, sprida 

falling 3 dråsa, drälla, drösa 

putting together 3 knipa, knut, knyta 

177

hardness 3 skal, skare, skava 

slackness 3 slak, slana, sloka 

quickness 3 snar, sno, snudig 

stiffness 3 styv, strak, stram 

slang 2 plit, snut 

beat 2 slag, slå 

mental feeling 2 snopen, tråna 

jocular 2 spe, spjuver 

fine grain 2 stöv, strö 

In order to make a general comparison with the earlier results of chapter 

4 concerning the semantic features found in the analysis of initial 

consonant clusters, part of the results of diagram 4.5 are presented as a 

table below (table 5.24). 

The 11 most frequent semantic features for the combination initial cluster 

+ no final cluster (and no geminate consonant) are shown in table 5.23. 

This ranking can be compared with diagram 4.5, which shows the 

frequencies of all the semantic features, both with and without final 

clusters. 

Table 5.23 The 11 most frequent 

semantic features for the combination 

initial cluster + no final cluster. 

178 

Table 5.24 The 11 most frequent 

semantic features of diagram 4.5 

showing initial clusters irrespective of 

final cluster. 

pejorative 45 pejorative 163 

long thin form 36 sound 107 

talking 23 long thin form 97 

sound 21 quick or strong 

movement 

67 


movement 

18 wetness 63 

wetness 15 talking 55 

thin form 13 light 32 


11 diminutive 31 

structure 

destruction 11 round form 23 

light 10 way of walking 22 

gaze 10 destruction 21

Table 5.23 shows the 11 most frequent semantic features of table 5.22 

(initial clusters + no final clusters or geminates), whereas table 5.24 

shows the 11 most frequent features of diagram 4.5 (or Appendix 1), 

which contains roots with initial clusters either with or without final 

clusters or geminates. 

Naturally, there are fewer words in the table showing initial cluster + no 

final cluster. In both tables, i.e. for both initial cluster + no final cluster 

and initial cluster + final cluster, 'pejorative' is the most frequent feature. 

The features 'long thin form', 'sound', 'quick or strong movement', 

'wetness' and 'talking' are among the six most frequent features in both 

tables. In other words the same six features are among the six most 

frequent features in both cases, only in partially different order. This 

means that the semantic features are the same, and also approximately as 

frequent, irrespective of whether one analyzes roots with only initial 

consonant clusters or roots with both initial and final consonant clusters. 

This result strengthens one outcome of the analysis of chapter 4: the 

analysis of semantic features of initial clusters actually showed the 

semantic features of the initial consonant clusters and not of the final 

clusters. 

5.6.2 No initial cluster + final cluster 

Table 5.25 shows the combinations of no initial cluster + final cluster, in 

frequency order. 

179

Table 5.25 Combinations of no initial cluster + final cluster. The 

recurring semantic features of final clusters in frequency order. Listed to 

the right are the roots. 

semantic feature freq. rank the words 


movement 

in 5.22 

51 5 rutscha, puffa, dangla, 

rangla, dingla, ringla, 

vingla, famla, vimla, 

fumla, rumla, tumla, 

haspla, dimpa, gumpa, 

pumpa, rappa, tappa, 

kippa, tippa, vippa, hoppa, 

moppa, poppa, guppa, 

dallra, pillra, tillra, 

myllra, darra, irra, 

dirra, virra, burra, 

huttra, hafsa, sjafsa, 

nafsa, rafsa, tafsa, rufsa, 

tufsa, räfsa, mumsa, 

palta, bylta, hasta, kasta, 

rista, hysta, nysta 

talking 18 3 babbla, rabbla, gaffla, 

mumla, tassla, tissla, 

rossla, jollra, morra, 

hurra, murra, tjattra, 

muttra, jamsa, gasta, 

hosta, pusta, tjafsa 

sound 14 4 ratscha, pingla, humla, 

sörpla, porla, sorla, rassla, 

vissla, bullra, mullra, 

kurra, surra, kuttra, 

puttra 

slang 12 - haffa, fiffa, piffa, roffa, 

buffa, luffa, ruffa, tuffa, 

töffa, dumpa, rumpa, 

sumpa 

pejorative 9 1 taffla, fiffla, ruffla, lolla, 

hafsa, tjafsa, lafsa, jamsa, 

dalta 

walking 7 13 lulla, jumpa, lafsa, lufsa, 

halta, rulta, tulta 

round form 3 14 rolla, bulla, rulla 

short-wide form 2 - fimpa, limpa 

light 2 10 tindra, skimra 

long thin form 2 2 ramsa, remsa 

180

The ranking of the most frequent words can be compared with the 

ranking of table 5.22. This comparison shows that 'pejorative' is not as 

frequently coded in the final clusters as in the initial clusters. 

On the other hand, 'slang' (mainly coded in -ffa), 'walking' and 'quick or 

strong movement' are more frequent in words with only final clusters, 

compared with words with only initial clusters. This might be due to the 

verbal nature of the clusters ending in a vowel. In the same way as table 

5.23 is compared with table 5.24, table 5.25 can be compared with table 

5.26, which contains all the roots (and not just the ones without initial 

clusters) taken from table 5.2. 

Table 5.26 All roots from table 5.2, ordered by semantic features, in 

frequency order. 

semantic feature freq. the words 


movement 

85 rutscha, knuffa, puffa, dangla, rangla, skrangla, 

dingla, ringla, vingla, famla, vimla, tromla, 

fumla, rumla, drumla, tumla, stöppla, haspla, 

rumla, dimpa, skvimpa, gumpa, skumpa, pumpa, 

klappa, snappa, rappa, tappa, greppa, steppa, 

kippa, skippa, tippa, vippa, hoppa, moppa, 

noppa, snoppa, poppa, droppa, proppa, guppa, 

knäppa, snäppa, fladdra, bläddra, dallra, pillra, 

tillra, myllra, flimra, darra, irra, dirra, virra, 

svirra, burra, plurra, snurra, glittra, splittra, 

klottra, huttra, hafsa, sjafsa, nafsa, snafsa, rafsa, 

krafsa, tafsa, plufsa, rufsa, tufsa, räfsa, plumsa, 

mumsa, palta, skralta, skvalta, bylta, hasta, kasta, 

rista, hysta, nysta 

talking 37 babbla, rabbla, gaffla, traggla, svamla, mumla, 

tassla, tissla, rossla, slabbra, bladdra, pladdra, 

sladdra, bluddra, sluddra, jollra, pjollra, blarra, 

snarra, kvirra, morra, knorra, hurra, murra, 

tjattra, fnittra, muttra, tjafsa, jamsa, grumsa, 

tramsa, gasta, hosta, pusta, frusta, knysta, krysta 

sound 34 klatscha, ratscha, pingla, skramla, humla, sörpla, 

porla, sorla, rassla, prassla, gnissla, vissla, 

kvillra, skallra, bullra, mullra, bjällra, knarra, 

klirra, knirra, skorra, kurra, surra, smattra, 

knattra, snattra, skvattra, knittra, kvittra, kuttra, 

puttra, gläfsa, klafsa, slafsa 

181

slang 19 haffa, blaffa, klaffa, fiffa, sniffa, piffa, skoffa, 

roffa, buffa, skuffa, luffa, fluffa, ruffa, gruffa, 

tuffa, töffa, dumpa, rumpa, sumpa 

pejorative 19 taffla, fiffla, ruffla, fjolla, lolla, stolla, slampa, 

fjompa, klattra, plottra, lafsa, glufsa, flamsa, 

dalta, slafsa, hafsa, tjafsa, tramsa, jamsa 

walking 15 lulla, drulla, klampa, trampa, stampa, jumpa, 

trippa, lufsa, halta, rulta, tulta, stulta, stylta, 

klafsa, lafsa 

wetness 9 blaska, plaska, slaska, smaska, snaska, vaska, 

glopp, klafsa, skvalta 

round form 7 rolla, skrolla, bulla, krulla, stropp, rulla, knollra 

long thin form 5 strimla, slamsa, ramsa, remsa, slimsa 

light 5 glindra, tindra, skimra, glimra, flimra 

short-wide form 4 fimpa, limpa, klimpa, klumpa 

diminutive 2 stumpa, loppa 

falling 1 ramla 

bodily feeling 1 pirra 

gaze 1 stirra 

rough surface structure 1 knottra 

beat 1 bulta 

destruction 1 brista 

slackness 1 slimsa 

movement 1 knaggla 

winding form 1 kringla 

form 1 plumpa 

The analysis shows that the six most frequent features are the same and 

that they are so in exactly the same order. This result strengthens the 

analysis of final clusters in table 5.25. 

5.6.3 Initial cluster + final cluster 

The tables 5.27 and 5.28 show combinations of final clusters and initial 

clusters. The examples are extracted from table 5.2. Table 5.27 shows 

combinations where the semantic features are the same initially and 

finally, and table 5.28 shows combinations where the features are 

different initially and finally. The following conclusions can be drawn: 

1. Fifty words have the same semantic feature initially and finally, 

whereas 34 words have different features. However, of these 34 words, 8 

have initial and final features that are similar to each other: 'sound' + 

'talking' and 'winding form' + 'round form'. It is more common, then, 

182

for semantic features to be the same, initially and finally, than to be 

different. 

2. The semantic features of a root can be the same initially and finally for 

each root with double features 12. These are the following, listed in order 

of frequency. The meanings of initial clusters come from the analysis in 

Appendix 1, and the meanings of final clusters come from the analysis in 

table 5.2. 

Table 5.27 Semantic features (with absolute frequencies of the roots) that 

are the same initially and finally in one and the same root. 

doubling 

freq. the words representing 

features 

root morphemes 

sound 18 klatscha, skramla, 

prassla, gnissla, skallra, 

kvillra, bjällra, knarra, 

klirra, knirra, skorra, 

smattra, knattra, snattra, 

knittra, kvittra, klafsa, gläfsa 

talking 9 svamla, slabbra, bladdra, 

sladdra, bluddra, sluddra, 

blarra, tramsa, grumsa 

pejorative 9 fjolla, stolla, slampa, fjompa, 

plottra, slafsa, glufsa, flamsa, 

tramsa 

wetness 4 blaska, plaska, slaska, snaska 

quick or strong 3 tromla, skippa, fladdra 

movement 

light 3 glindra, flimra, glimra 

walking 2 trampa, stulta 

short-wide form 2 klimpa, klumpa 

long thin form 1 slamsa 

Table 5.28 shows which different semantic features are combined initially 

and finally. A word representing a root morpheme may occur in two 

places if it contains two features initially or two features finally, e.g. 

plumsa: pl- means both 'sound' and 'wetness' in this root morpheme. 

12 There are 50 words but 51 features, as can be seen in the table, because one word 

(tramsa) contains two features initially and two features finally. 

183

Table 5.28 Semantic features that are different initially and finally in one and 

the same root. 

initial: 

final: 

semantic 

feature of 

final 

cluster: 

slang quick or 

strong 

movement 

semantic 

feature of 

initial 

cluster: 

sound klappa 

steppa 

knäppa 

snäppa 

krafsa 

plumsa 

pejorative blaffa drumla 

klottra 

plufsa 

beat knuffa 

bad mood gruffa 

talking gruffa 

destruction skrangla 

splittra 

skralta 

184 

wetness walking round 

form 

smaska klampa 

stampa 


snarra 

kvirra 

knorra 

fnittra 

frusta 

knysta 

krysta 

drulla pjollra 

winding 

form 

snurra krulla 

iterative stampa 

quickness snappa 

falling droppa 

wetness droppa 

plurra 

plumsa 

skvalta 

light glittra 

diminutive splittra 

The following conclusions can be drawn from table 5.28. 

1. The semantic feature which is most often combined with other features 

is final 'quick or strong movement' – but this could partly be an effect of 

the choice to study the semifinal "a- endings". 

2. The most common feature combinations are:

'sound' + 'talking' (7 words) 

'sound' + 'quick or strong movement' (6 words) 

'wetness' + 'quick or strong movement' (4 words) 

In the combination 'sound' + 'talking' the two features are closely related 

and they strengthen each other. 

3. 'Sound' is the most common initial feature in these combinations. 

4. The otherwise common semantic features 'sound' and 'pejorative' do not 

have final position in these combinations. 'Talking' and 'walking' are much 

more common in final position than in initial. 

5.6.4 Summary of combinations 

The general results of the study of combinations of initial and final 

clusters are the following: Sometimes the initial cluster carries the sound 

symbolic meaning because the word ends in a single consonant (Group 1). 

The most common semantic feature of Group 1 is 'pejorative', which is 

also the case for all the sound symbolic roots in Appendix 1. The same six 

features are among the six most frequent ones in both groups and this 

indicates that the analysis of chapter 4 showed what it was intended to 

show: the meaning of initial clusters. 

When there are both initial and final consonant clusters in a root, these 

can combine by having the same meaning (Group 3a), e.g. bl- and -aska 

('wetness') – cf. table 5.27. Alternatively, different meanings can combine 

in a word (Group 3b), e.g. glittra – cf. table 5.28. The most common 

combinations are those in which the meaning is the same initially and 

finally. The most common meaning for group 3a is 'sound' + 'sound', 

followed by 'talking' + 'talking' and 'pejorative' + 'pejorative'. The most 

common meaning combinations for Group 3b is 'sound' + 'talking', 

followed by 'sound' + 'quick or strong movement'. 

Finally, sometimes only the final cluster (or geminate) carries the sound 

symbolic meaning, because the word begins with a single consonant 

(Group 2). The most common meaning is 'quick or strong movement' 

followed by 'talking' and 'sound'. 

185

5.7 Summary and discussion of initial and final 

clusters, and vowels 

This and the preceding chapter have shown that certain semantic features 

are connected with certain initial and final consonant clusters and with 

vowels – in different constellations and to different extents. A study of 

combinations of initial and final clusters shows that the sound symbolic 

meaning of a root morpheme can also be dependent on the combinations 

of phonemes of the whole word, e.g. sl- 'wetness' + a + -sk(a) 'wetness' 

(=slaska (splatter)). (However, both 'slisk' and 'slusk' have the feature 

'pejorative' and not 'wetness' because of the semantic profile of sl-, cf. 

4.4.1). 

Partly the same semantic features are used in both initial and final 

clusters, while the vowels sometimes add a dimension of 'size' (smalllittle, 

high pitch-low pitch). The 'light/gaze' words, where the vowel is 

almost always a front vowel, could be included in the 'size' group, since 

'light' can be seen as synaesthetically connected with high pitch. The 

vowels, when they are used sound symbolically, thus often seem to have 

other semantic features than the consonant clusters do, and they add a 

special meaning to roots – which otherwise get their sound symbolic 

meaning from initial and/or final clusters – e.g. by showing if a sound is 

high pitched or low pitched, or if a movement is quick or slow. 

For typically sound-symbolic (where all or almost all root morphemes are 

sound symbolic) or high frequent initial clusters, vowels, and final 

clusters, the meaning of a word, especially of a neologism, should be 

predictable. 

A nonsense expression like pjaffla : pj- ('pejorative') + a + -fla 

('pejorative') ought to be interpreted as a pejorative word. A neologism 

like brullra : br- ('sound') + P ('low pitch') + -lra ('sound' or 'quick or 

strong movement') ought to be interpreted as imitating sound (a low 

pitched one) while a nonsense expression like flillra: fl- ('quick or strong 

movement') + i ('smallness', 'quickness' or 'high pitch') + -lra ('sound' 

186

or 'quick or strong movement') ought to be interpreted as something 

having to do with quick movement 13. 

Looking at real words, there can be a combination effect like in skrangla 

(skranglig-rickety) "which has an unsteady construction, or has become 

unsteady because of long usage". This definition goes well with the 

phonesthemes skr- 'destruction' and -ngla 'quick or strong movement'. 

Another example is klatscha (crack): kl- 'sound' and -tscha [tSa] 'sound'. 

A more unconventional word klitscha (which is not in the dictionary) 

would stand for a more high pitched sound 14. For real words there can 

always be an effect of conventionalization which makes them less 

predictable. 

Conventionalization can always take over, at least partly, either in the 

arbitrary direction or in choosing phonesthemic meanings. The 

phonestheme level is an underlying level which can be activated by 

different types of context. 

The next chapter will present contrastive studies of some semantic fields 

of sound symbolism. In chapter 7 final clusters, as well as initial clusters, 

will be discussed again in connection with neologisms. 

13 In the experiments with nonsense words in chapter 6 only the effect of the initial 

clusters were tested. In further research, however, the intention is to also test the 

combined effects. 

14Two roommates were asked about the meaning of this expression. They answered that 

it was not a real word, and then one of them gave a possible meaning 'the sound of a 

bicycle spoke that breaks'. The other person said it meant 'the sound that comes when 

you drive on a thin layer of ice'. Both these definitions have to do with a high pitched 

sound. 

187

6 Some contrastive studies in sound symbolism 

6.1 Introduction 

The phenomena of onomatopoeia and other sound symbolism have been 

described in part for different languages. Some earlier contrastive 

findings are described in the overview of 1.10.4. 

The aim of this chapter is to study some of the semantic features and 

consonant clusters discussed above in order to analyze what differences 

and similarities there are between some languages. 

In the first section I will present a contrastive lexical study made on some 

selected meanings in a Thesaurus. The languages compared are Swedish 

and English. I will then present a contrastive study of some expressive 

interjections, commands and greetings and a study of words for animal 

sounds. Furthermore I will present a contrastive experiment of 

interpretation of Swedish sound symbolic words. 

6.2 The Thesaurus study 

6.2.1 Method 

In this experiment informants chose the best words for 'stupidity', 

'roughness' and 'smoothness' in English and Swedish out of a Thesaurus. 

From the Swedish Thesaurus (Bring, 1930) a couple of meanings were 

chosen, namely: 'dumhuvud' (501), 'oförstånd' (499), and 'glatthet' (255), 

'skrovlighet' (256). In the English Roget´s Thesaurus (1977) the 

following meanings were chosen: 'fool' (501), 'imbecility' (499), 

'insanity' (503), 'smoothness' (255), and 'roughness' (256) 

Lists of all words under these headings were then inspected by native 

informants, three for the Swedish material and three for the English 

material. The informants were given the instruction "to mark words, the 

expressive forms of which are felt to be especially adequate for their 

content". 1 

1 Since I am not an active speaker of English and since I have deepened my intuitions of 

Swedish phonesthemes from the analysis of chapter 4, I used subjects for this task. 

189

6.2.2 Results 

6.2.2.1 Words for 'stupidity' in English 

Since English does not have the clusters typically pejorative for Swedish: 

fj-, fn-, pj-, there could be no similarities between Swedish and English 

here. (Cf. below and diagram 4.6.) 

There were 4 words where all three English subjects agreed: 

nincompoop, blockhead, dunce and dull. 

Two of the three subjects agreed on 9 words: simpleton, dolt, booby, oaf, 

clod, silly, muddleheaded, addleheaded, fool. 

Isolated contributions from either of the subjects were: tomfool, ass, 

noodle, gawk, mooncalf, dotard, driveler, old fogy, Simple Simon, goose, 

lout, idiot, dotage, idiocy, fatuity, giddiness, drivel, dote, stultify, bovine, 

feebleminded, obtuse, stolid, fatuous, driveling, bewildered, maudlin, 

stupidity, foolishness, rashness, brainless, childlike, vacant. 

Many of these words are not what we would normally think of as sound 

symbolic. Nevertheless there seems to be a slight tendency for the 

preferred ones to contain a long [u] or an [a] and perhaps [d] initially or 

medially. The value of the initial consonant clusters is difficult to judge 

from this material. 

6.2.2.2 Words for 'stupidity' in Swedish 

All three Swedish subjects agreed on the following words: pjosker 2 , 

pjasker, fjoller, fnasker, fjanter, flep, flepig, pjåk, fjolleri, fjant, pjoller, 

pjosk, fjollig, fjoskig, fjantig, pjoskig, pjollrig, pjåkig, pjaskig. 

The frequencies of the initial consonant clusters are the following: 

pj- : 9 words 

fj- : 7 words 

fl- : 2 words 

fn-: 1 word 

2 The Swedish words listed under the different categories of the Thesaurus are not 

translated since the Thesaurus heading, e.g. "stupidity" shows the semantic category. 

What is interesting is not the exact meaning of all these words, but which sounds are the 

most frequent for each Thesaurus heading. 

190

Except for fl- the result above mirrors the results of the percentally most 

frequent clusters well (diagram 4.6). 

Two of the subjects agreed on the following words: 

mähä, tafser, tossa, tafsig, fjoskighet, fjanta, fjollas, pjollra, pjoska, pjåka, 

pjask. 

The most common clusters here are: 

pj-: 4 words 

fj-: 3 words 

There is a clear preference for the fj- and pj- clusters among the Swedish 

informants. 

Isolated contributions were: våp, tåp, jöns, fån, sjasker, drummel, tosing, 

stolle, drönig, tåpig, -snut, pund-, mes, schajas, flack, flärdfull, larvig, 

korkad, tafs, sjask, jolt, fåna, jolta, drumlig, slapp, slö, våpig, fånig, 

sjaskig, taskig, tölpig, showing an additional preference for the vowels 

[o:], [O] and [a]. The consonant cluster which is the most common is dr- 

(3 roots). The cluster dr- is pejorative according to the analysis in 

chapter 4. 

6.2.2.3 Words for 'surface structure' in English 

For the categories of 'roughness' and 'smoothness' there where two 

English subjects who made an assessment. 

'Roughness' 

Both subjects agreed on the following words: crest, ruffle, crumple, 

rumple, rugged, jagged, gnarly, scraggly, scraggy, craggy, cragged, 

prickly, bristly, bushy. 

The clusters cr- and scr- and the consonant r are favored. 

Isolated contributions were: asperity, corrugation, shag, cross-grained, 

hirsute, shaggy, nappy, thatch, whiskers, feather, rough, crinkle. Among 

these words there are two instances of cr- and five instances of single r. 

'Smoothness' 

Both subjects agreed on the words sleek and glossy. 

191

Isolated contributions were: plane, level, polish, velvety, glassy, gloss, 

roller, roll, oily, silken, silky. 

The preferred consonant is l. 

6.2.2.4 Words for 'surface structure' in Swedish 

'Roughness' 

All three subjects agreed on the following words: knotter and knottrig. 

Two out of three agreed on the following: knotighet, knut, knyla, knotig, 

knagglig, klippig, knollrig, krullig, krusig, stripig, skrapa, skrynka, 

skrynkla, knottra, skrovlig, skrynkig, skrynklig 3 

These words show a preference for r and for the clusters kn- and skrand 

kr-. The kr- and skr- words both have the semantic feature 'rough 

surface structure' (in chapter 4) but kn- words are analyzed as 'round 

form'. The cluster with the highest number of root morphemes with the 

feature 'rough surface' structure (in the lexical analysis) is skr-. 

Isolated contributions were: raggighet, knöl, kornighet, rufsighet, burr, 

stripa, ludenhet, lurv, tovig, purrighet, ludd, borst, ragg, lugg, rugg, 

stubb, test, ull, kvist, ris, tistel, tova, tuva, kvast, visp, buckla, kugge, 

krusning, rysch, frans, dun, plysch, schagg, sträv, fårad, uddig, tandad, 

taggig, törnig, risig, buskig, tuvig, lummig, sträv, tofsig, noppig. 

These contributions, from singular subjects, show a preference for the 

vowel [P]. 

'Smoothness' 

All three subjects agreed on glätta, glatta, glanska and glansa showing 

preference for gl-. 

Two out of three agreed on: glättning, glans, slipprighet, slirning, 

glidning, blanka, glasera, glida, slira, glatt, glansig, glansk and slipprig. 

3 All words excerpted by the subjects were counted, and as a consequence the list also 

contains words that represent the same root. 

192

The recurring clusters are gl-, sl-, bl-, which all contain the liquid l. 

These are, except for sl-, related to 'light', but as was shown in 4.6 there 

is an indexical relation between 'light', 'reflecting surface' and 'movement 

on reflecting surface'. Slipprig and slira were analyzed as 'quick or 

strong' movement in the lexical analysis, (while slipa, slät, slätt were 

analyzed as 'smooth surface structure'). Sl- is also to a high degree 

connected with 'wetness'. 

Isolated contributions were: glänsa, glänsande, blank, glas, smörjning, 

smörja, smärgel, kristall, siden, silke, sammet, lackera, polera, slipa, 

kana, stryka, mangla, valsa, oljig, smidig, mjäll. 

Also in these words the l is very frequent, however not part of an initial 

cluster. 

6.2.3 Conclusions of the Thesaurus study 

There was greater agreement between Swedish and English among the 

preferred sounds for 'surface structure', both 'roughness' (r, skr) and 

'smoothness' (sl, gl) than for 'pejorative'. An explanation to this might 

be that 'surface structure' is closer to a potentially common phenomenon, 

namely sound imitation, since stroking different surfaces gives different 

sound effects. 

Sometimes the isolated contributions perhaps mirror the fact that the task 

at hand for the test subjects can be difficult to keep in mind or that the 

language feeling sometimes runs amock among words with similar 

meanings. Suddenly all words can feel motivated. Nevertheless, this 

method is a possible way of comparing different languages, through the 

intuitions of informants. A problem with using the Thesaurus is that the 

word corpus listed is very heterogeneous. Examples of this are several 

words (however not chosen by the Swedish subjects) under the category 

'roughness': ojämnhet (unevenness), knävelborr (big and bushy 

moustache), brottyta (fracture), kartnagel (deformed nail), vårta (wart) 

and tuppkam (cock's crest). 

For cross linguistic comparisons the Thesaurus method used above is an 

alternative to free production within different semantic fields, (for those 

languages that have a Thesaurus). Naturally, native informants have to be 

used. 

193

6.3 Some interjections in different languages 

Expressive interjections is one of the sound symbolic types described in 

2.4. The relation between expression and meaning is mainly indexical; the 

expression is caused by a bodily or mental reaction. It is an interesting 

task to analyze how similar the expressive interjections are in different 

languages (the question of universal traits of the expression). It is also 

interesting to see, for the expressive interjections of Swedish, to what 

extent the expressive interjections are unsystematical, i. e. not relatable to 

phonesthemes. 

Expressive interjections, commands and greetings (cf. Ideforss, 1928, 

categorization) are exemplified mostly with a fragment of the examples in 

Ideforss. These were translated into 8 different languages with the help of 

informants. The languages are Icelandic, English, Polish, Hungarian, 

Finnish, Ososo, Malagasi and Slovenian. The results are presented in 

Appendix 2 and grouped according to type of interjection. The expressive 

interjections, commands, and greetings are listed in table 6.2 and in 

Appendix 2. They are written in normal spelling in table 6.2. 

In some cases it was difficult to ask subjects to translate word for word. 

The semantic contents of the intended interjection instead had to be 

described (within the semantic categories of table 6.2), after which the 

informant gave the closest correspondence in his own language. The way 

the semantic contents was described to the informant depended on his or 

her competence in Swedish or English, e.g. "Give me a word or 

expression for when you want to be depreciatory" or "Give me a word or 

expression for when you think somebody or something is bad or ugly, 

etc". 

The semantic features used in earlier chapters, which correspond to the 

features of these interjections, are shown in table 6.1. 

194

Table 6.1 The semantic features of interjections and their 

correspondences in earlier chapters. 

Features of 

Corresponding features 

interjections 

of earlier chapters 

pejorative pejorative 

positive attitude 

surprise mental feeling 

song talking 

other bodily or mental bodily feeling or mental 

feeling 

feeling 

commands talking 

greetings talking 

6.3.1 Swedish expressive interjections 

In table 6.2, a sample of the Swedish expressive interjections is presented. 

Table 6.2 Examples (in ordinary spelling, mostly from Ideforss, 1928) of 

Swedish expressive interjections, commands and greetings. 

EXPRESSIVE 

hihi(hi) 

mums 

pejorative 

mm 

namnam 

bu 

iih 

usch 

grr 

hu 

surprise 

COMMANDS 

"t" [|] 

oj 

to animals 

blä 

oh 

schas 

ha 

åh 

ptroo 

håhå(jaja) 

ä 

"t" [|] 

tvi 

öh 

åhå 

"p" [Ö] 

ä 

song 

to persons 

äh 

lala(la) 

jaja 

bah 

trala(la) 

aja(baja) 

asch 

sch 

äsch 

other bodily or 

vyss 

isch 

mental feeling 

lull lull 

uh 

urrk 

aha 

aj 

pst 

bu 

fy 

brr 

tss 

positive 

hm 

åhej(åhå) 

GREETINGS 

ohoj 

tjo(hej) 

ah 

puh 

pust 

hoho 

oh 

ojojoj 

åh 

vojvoj 

haha(ha) 

håhå(jaja) 

atjo 

195

6.3.2 Discussion of Swedish expressive 

interjections, commands and greetings 

Some short comments are made here on the phonetic/phonological 

structure of the interjections in table 6.2 in order to relate them to the 

analyses in chapters 2, 4 and 5. They often consist of only a vowel, 

sometimes extra long, or of only a consonant, often extra long. They can 

begin with a vowel or a consonant. The more open vowels are preferred, 

especially [A] and [a] but also [O]. The most common phonemes are j and 

h. 

Of the initial clusters, bl-, tv-, ts, tr-, br, gr-, ptr and pst, that occur 

among these interjections, bl-, tr- and gr- conform with the analysis in 

chapter 4, i.e. bl- can be 'pejorative', tr- can be 'talking' and gr- can be 

'bad mood'. The clusters ts, ptr and pst are unconventional. The final 

clusters are very few: -rk, -st and -ms. None of these conform to the 

meanings of final clusters of chapter 5. 

This result is in accordance with the analysis presented in 2.4, which said 

that the sound side of expressive interjections are realized more 

unsystematically, i.e. they are not (partly) built up by phonesthemes as 

sound symbolic roots are. But, as can be seen, the borders are not 

absolute between expressive interjections and sound symbolic 

phonesthemes. Three interjections – blä (bl- 'pejorative'), tralala (tr- 

'talking') and grr (gr- 'bad mood') – conform with the phonesthemic 

analysis. 

Typical for the interjections are the special sounds or sound combinations 

that occur, e.g. click sounds and non-standard phonotactic combinations, 

e.g. [ptro:], [ts:], [pst] and [hm], and isolated consonants like [S]. CVstructure 

and reduplicated CV-structure are also common, as well as the 

lengthening of vowels or consonants. Consonant frames with a shift of 

vowel, e.g. asch, äsch, usch, isch also occur. Here it is quite obvious that 

the vowel quality imparts different meanings (cf. 5.4). The same goes for 

ah, äh, eh, i(h), ö(h). Prosody is probably important but has not been 

studied here. None of the phonological or phonetic characteristics is 

clearly connected with a certain semantic feature. 

196

To summarize, few of the interjections have (conventional) initial 

consonant clusters. Of those 5 conventional clusters that occur, 3 conform 

with the analysis of chapter 4: bl- (pejorative), tr- (talking) and gr- (bad 

mood). The final clusters of interjections are very few, and none of them 

conforms with the analysis in chapter 5. 

6.3.3 Phonological and phonetic similarities and 

dissimilarities between interjections of different 

languages 

One interesting question now is: Are there phonological and phonetic 

similarities between the interjections of different languages? Expressive 

terms, which mirror bodily and mental states, could be similar in 

different languages, with a larger or smaller language specific 

superstructure. (cf. Abelin and Allwood, 1984). An analysis of my 

material (presented in Appendix 2) gives the following result: 

Each semantic category shows its sound pattern tendencies for the 

different languages. These tendencies are enhanced by big differences 

between the categories. The pejorative interjections, for example, often 

contain an [u:] or an [O], the positive interjections an [i] or an [a], surprise 

often an [O] or an [a]. 'Pain' (cf. Swedish aj) has an [a] and a [j] or a 

diphtongized open vowel. 'Sneeze' (cf. Swedish atjo 4 ) has a voiceless 

fricative or an affricate in all the languages being studied, and the 

interjection for go away (to an animal) (cf. Swedish schas) has a voiceless 

fricative. 'Good taste' has a nasal in these 8 languages, and 'scaring 

somebody' has a voiced stop (mostly b) in all these languages except for 

two. 

When counting all instances of vowels and consonants in the interjections 

in Appendix 2 (excluding the Swedish ones, since they are greater in 

number) the following tendencies are found, shown in table 6.3. (Which 

languages contributed to which sounds can be seen in Appendix 2 and 

table 6.3). Many of the categories of table 6.3 are more specific than 

those in table 6.2, depending on what answers the informants gave, cf. 

4 Different ways of conventionalizing this physical outburst (which of course can vary for 

different individuals and from time to time) is, in Swedish, mirrored in older spellings of 

this word: atschi, atsji, aptschäh, apschohoj, aaah-tschah-katsch katsch, kaa-kah, tschah, 

tjas, hlutt, schtschi, tjihitt, tjihihitt, tjitjisit (Ideforss, 1928, p. 25). 

197

Appendix 2. The categories 'song' and 'greetings' were not elicited from 

the informants. 

Table 6.3 The most common vowels and consonants for different 

categories of interjections. The languages are Icelandic, English, Polish, 

Hungarian, Finnish, Ososo, Malagasi and Slovenian. 

vowels consonants 

pejorative O (9), u (9), i (5) 

O: 

Icelandic: 2 

English: 1 

Polish: 2 

Hungarian: 1 

Ososo: 2 

Slovenian: 1 

u: 

Icelandic: 1 

English: 3 

Hungarian: 3 

Finnish: 2 

i: 

Hungarian: 1 


Slovenian: 2 

positive i (6), a (5) 

i: 

Icelandic: 2 

English: 2 


a: 

Icelandic: 1 

English: 2 


Ososo: 1 

198 

j (10), f (6), h (6) 

j: 

Icelandic: 3 

English: 1 

Polish: 1 

Hungarian: 4 

Ososo: 1 

f: 

English: 2 

Hungarian: 1 

Ososo: 1 

Slovenian: 2 

h: 


Ososo: 1 

j (3), p (3) 

j: 

Icelandic: 1 

English: 1 


p: 

Icelandic: 1 

English: 1 

Finnish: 1

surprise O (7), a (4) 

O: 

English: 1 

Polish: 1 

Ososo: 1 

Slovenian: 4 

a: 

Icelandic: 1 

English: 1 

Ososo: 1 

Malagasi: 1 

pain a (9), u (6) 

a: 

Icelandic: 2 

English: 2 

Hungarian: 2 


Malagasi: 1 

u: 

Icelandic: 2 

Hungarian: 2 

Ososo: 2 

freezing O (2) 

O: 

Polish: 2 

199 

j (3), h (3) 

j: 

Hungarian: 1 

Ososo: 1 

Slovenian: 1 

h: 

Icelandic: 1 

Hungarian: 1 

Slovenian: 1 

j (7) 

j: 

Icelandic: 1 

Hungarian: 4 


Malagasi: 1 

b (3), r (3) 

b: 

English: 1 

Polish: 1 

Hungarian: 1 

r: 

English: 1 

Polish: 1 

Hungarian: 1 

thoughtfulness - h (5), m (6) 

h: 

Icelandic: 1 

English: 2 

Polish: 1 


m: 

Icelandic: 2 

English: 2 

Polish: 1 

Finnish: 1

exhaustion u (4) 

u: 

English: 1 

Polish: 1 


sudden insight a (10) 

a: 

English: 4 

Hungarian: 2 

Ososo: 2 

Slovenian: 2 

sneeze i (8), a (6) 

i: 

English: 1 

Polish: 1 

Hungarian: 1 

Ososo: 3 

Slovenian: 2 

a: 

Icelandic: 1 

English: 1 

Polish: 1 

Hungarian: 1 

Slovenian: 2 

good taste a (6) 

a: 

Icelandic: 3 

English: 1 

Polish: 2 

commands to 

animals 

i (3), u (5) 

i: 

Icelandic: 2 

Hungarian: 1 

u: 

Icelandic: 2 

English: 1 

Ososo: 1 

Malagasi: 1 

200 

h (4) 

h: 

Hungarian: 1 


h (7) 

h: 

English: 4 

Hungarian: 1 

Ososo: 1 

Slovenian: 1 

t (6), C (5) 

t: 

Icelandic: 1 

English: 1 

Hungarian: 1 

Ososo: 1 

Slovenian: 2 

C : 

Icelandic: 1 

Polish: 1 

Ososo: 1 

Slovenian: 2 

m (11), n (7) 

m: 

Icelandic: 4 

English: 1 

Polish: 2 


Slovenian: 3 

n: 

Icelandic: 3 

Polish: 2 

Slovenian: 2 

S (7) 

S: 

Icelandic: 1 

English: 1 

Polish: 1 

Hungarian: 2 


Malagasi: 1

mild warning to 

children 

- s (3) 

s: 

Icelandic: 3 

be quiet - S (4), t (3) 

S: 

English: 1 

Hungarian: 1 


Malagasi: 1 

t : 

Polish: 1 

scaring somebody u (5) 

u: 

English: 1 

Polish: 2 

Hungarian: 2 

201 

Hungarian: 2 

b (6) 

b: 

Icelandic: 1 

English: 1 

Polish: 2 

Hungarian: 1 


The table shows that there are many vowels and consonants that are 

similar in the languages Icelandic, English, Polish, Hungarian, Finnish, 

Ososo, Malagasi and Slovenian for the categories 'pejorative', 'pain', 

'sneeze', 'good taste', 'be quiet', 'thoughtfulness' respectively. These 

vowels and consonants of all the functions of table 6.3 are (approximately) 

[i], [a], [O], [u] and [p], [b], [m], [f], [t], [s], [n], [r], [j], [C], [S], [h]. Most of 

these consonants are labial or dental. There are no consonants produced 

behind the hard palate, except for [h]. The vowels are few and mostly [i], 

[u], and [a]. 

It also seems that a certain sound (or sound combination) is preferred in a 

certain language and is used for a variety of meanings. The Polish 

informant, for example, prefers the sound combinations [ux], [Ox] and [O], 

while the Hungarian informant prefers [jOj], [jaj], [juj]. 

There are also different degrees of conventionalization in the language, 

both depending on speakers and on the situation (wild and tame forms in 

Rhodes' (1995) terminology). Thus, in Swedish there are both expiration

and the interjection uh for 'tiredness', and an imitation of the horse's 

neigh or the interjection gnägg. 5 

6.4 Imitations of animal calls 

A subgroup to the onomatopoeic interjections (cf. 2.4) are those that 

imitate animal sounds. 

sound imitative 

Animal call imitative Imitative of physical and bodily reactions Other 

Figure 6.1. Relations between different types of onomatopoeic 

interjections. 

Many animals have laryngeal and supraglottal organs and cavities similar 

to man, but with different resonance properties. Most of them produce 

sounds in the same way as humans (cf. Lindblad, 1992). These sounds 

seem to be sufficiently similar to allow for imitation in human languages. 

6.4.1 Expressions for animal calls in Swedish and 

other languages 

The human expressions that imitate animal calls are easy to categorize 

semantically. (I presume that e.g. a Swedish cat sounds the same as a cat 

from geographically distant countries.) Interjections imitating animal calls 

have been translated, with the help of informants, into 16 different 

languages: Icelandic, English, Polish, Hungarian, Finnish, Ososo, 

Malagasi, Korean, Japanese, Chinese, Estonian, Urdu, Persian, Kurdish, 

Arabic, and Spanish (see table 6.4). 

5There is reason to believe that comic strips have contributed to a conventionalization of 

both writing and speech (via writing). 

202

Table 6.4.a Interjections imitating 7 animal calls, from 16 different languages. 

mouse cat goat/ 

sheep 

Korean tSikTSik jaON 

njaON 

Japanese tSiju: nija: 

nijaN 

Chinese tSitSitSi mimio 

mijao 

dog pig horse cow 

mœœ: mUNmUN kulkul hihi:hiN Im:E: 

mœhœ: kENkEN 

mE:mE: waNwaN bu:bu: hihiN mo: 

mimi: 

mimie: 

w´w´w´ IhIhIh ahahaha Na: 

Finnish pyppyp miau mœ:œ: hauhau 

vuhvuh 

rØhrØh ihaha: am:u: 

Estonian pi:p miau mØk: 

mœ: 

auxaux ØhØh ihaha mu: 

Urdu tSitSi miaow mE: 

bE: 

bONbON – - m: 

Persian - miau bE:bE: howhow korokoro hihihi: 

daNtSo 

mE: 

Kurdish DikDikDik miau be: wOwwOw – - ma: 

me: 

bOrabOr 

Arabic - miau me: 

ba: 

habhab – hEehEe bu: 

Polish pipi miau me: hauhau – hihi mu: 

English skwi:k 

i:k 

miau ba: bauwau 

wOfwOf 

japjap 

Swedish pi:p pi:p miau bœ: 

mœ: 

vuv:uv 

vufvuf 

203 

OiNkOiNk neigh mu: 

nØf:nØf: gnEg:nEg m¨ 

Spanish - miau be:be. guau OiNkOiNk - mu: 

Icelandic pi:p pi:p miau mœ: vufvuf ¯ njihaha møù 

Hungarian - miau bœ:œœ 

mœkmœk 

vauvau rØf:rØf: - mu: 

Ososo - miau bœœ bOubOu imitation - mu: 

Table 6.4.b Interjections imitating another 8 animal calls, from 16 different languag 

KoreanJapaneseChineseFinnishEstonian 

crow cuckoo owl duck hen rooster frog chicken 

ka:kka:k PUkukPuUNkwakkokotœpUkukpuUNkwakkoko ka:ka: kak:o: ho:kho:k ga:ga: kokoko kokekoko - - 

kokjo: - - 

tSijatSija pukupuku ahaha ka:kak kukuku kIkIkI - - 

va:kva:k kuk:u:kuk:u: 

huhu: kvak-kvak kO:tkotkot kuk:ukie:ku 

kva:k - 

- kuku uhu: prœ:k kaka kukeleku: - - 

Urdu kajkaj kokokoko - kvakva kONkONkON 

kOk´lONkON 

- -

PersianKurdishArabicPolishEnglishSwedish 

SpanishIcelandic 

kr¨k- 

--kr¨k 

HungarianOsoso 

204 

- - 

VarVar hoho huo: kwak- VodVod kukulikwakkoko 

- - ba:bu:kwak- kOkOkokO qOqOqOqO - - 

ba:bu:kwak - - hu:hu: waq ququ qoqoqo - - 

krakra kuku uhu: kwakwa koko kukuriku - - 

kOwkOw kuku: - kwak- kl´kkl´k kOk´kwakdud´ldu:krakskraks 

kuku: huhu: kvakkvak kakaka 

- - - kwakkwak 

- - 

kvak:vak: pi:p pi:p 

kPkElik¨: 

klPkklPk 

- kikiriki: - - 

- ¨hu: bra:bra: kakaka kPkElikP 

- pi:ppi:p 

ka:rka:r - - haphap kOtkOt kukuriku - tCiptCip 

- - u:u: kuOk klOkklOk kukuru:kuOkOkOrOkO 

The following conclusions can be drawn from comparisons between 

expressions for animal sounds: 

1. There is imitation of animal calls in all the languages in the study. 

2. No animal call imitation is exactly the same in all languages. 

krukkruk tCiatCia 

3. Some animal call imitations are similar in the different languages, 

while others vary more. For example, the cat's sound is conventionalized 

as [miau] in all languages except in Korean where it is [njaON], Japanese 

where it is [nijaN] and Chinese where it can be [mimio] (but also [mijao]). 

The expression for the dog's sound, however, varies greatly between the 

languages. One possible explanation for this is that the calls of the 

different animals differ in complexity so that for the complicated animal 

calls, different languages attend to different acoustic characteristics with a 

point of departure in the phonology of the language. Another possibility 

is that some animals simply have a bigger repertoire of calls. 

4. There are mostly similarities between expressions for animal calls on 

the level of phonological features. For example, all imitations of the cats 

meow include a nasal, and the imitation of the rooster always contains a

voiceless plosive, velar or uvular. Sometimes the similarity is in the form 

of reduplication. The dog says [vuv:uv] or [vuf:vuf] in Swedish, the same 

in Icelandic, and [vau vau] in Hungarian, so in these languages the initial 

consonant is the same. In Ososo the dog says [bOubOu]. The likeness with 

the previous examples lies in the initial consonant which is voiced and 

labial, and that there is an [u] vowel. 

5. Prosody is probably important but not analyzed here. 

6. Within each language here, as well as for the interjections, there are 

likenesses which can be supposed to depend on, or reflect, the 

phonological structure of the language in question. For example, Arabic 

has [q] where e.g. Swedish has [k], Swedish has [¨] instead of [u], Korean 

and Japanese often have a final [N]. 

6.4.2 A test of expressions for animal calls of 

different languages 

Sixteen of the expressions for animal calls from different languages were 

chosen to test how speakers from other languages would interpret them. 

Those chosen were the ones where the sound shape of the expression was 

the most deviant in the corpus in tables 6.4.a and 6.4.b, in order to make 

the test as difficult as possible. The expressions of animal calls are listed 

in table 6.5. 

The subjects in this test were 15 people, with the following mother 

tongues: Swedish (8 subjects), French (2 subjects), English, Hungarian, 

Czech, Slovenian, Ososo. All were tested on the same occasion. They 

listened to pronunciations of the words for the animal calls and they saw 

them transcribed as in table 6.5. They were told to guess which animals 

had gotten their calls conventionalized in this way, and to write down 


205

Table 6.5 The 16 expressions of animal calls chosen for the interpretation 

experiment. 

CALL ANIMAL LANGUAGE 

1. vau vau dog Hungarian 

mπ´ 2. : cow Icelandic 

3. njihaha horse Hungarian 

4. hap hap duck Hungarian 

5. rØf rØf pig Hungarian 

6. bra: bra: duck Icelandic 

7. kOt kOt hen Hungarian, Finnish 

8. kOkEkO‘ko: rooster Japanese 

9. prœ:k duck Estonian 

10. pakUk: pakUk: cuckoo Korean 

11. njaON cat Korean 

12. bu: cow, pig Arabic, Czech, 

Japanese 

13. kOk´lONkON rooster Urdu 

14. tSi tSi mouse Urdu 

15. maN maN dog Korean 

16. hab: hab: dog Arabic 

6.4.3 Results from a test of expressions of animal 

calls in different languages 

The results were as listed in table 6.6. Sometimes a subject had written 

down more than one answer. All answers have been counted except for 

when subjects gave a judgement on a word in his/her first language. 

Table 6.6 Assignment of animal calls to animals, by 8 speakers of Swedish 

and 7 speakers of other languages: French (2 subjects), English, 

Hungarian, Czech, Slovenian, Ososo. 

8 Swedish 7 other languages 

expected answer other answer expected answer other answer 

1.dog 8 6 

2. cow 8 7 

3. horse 5 donkey 5 3 donkey 2 

jackal 1 cat 1 

4. duck 0 ? 5 0 ? 4 

goose 1 frog 2 

hen 1 

hippopotamus 1 

5. pig 7 dog 1 2 dog 3 

? 1 

206

6. duck 0 sheep 5 1 donkey 1 

crow 2 raven 1 

lion 1 crow 1 

wolf 1 sheep 1 

? 1 ? 2 

7. hen 2 goose 1 4 ? 3 

squirrel 1 

? 4 

8. rooster 8 5 hen 1 

pigeon 1 

9. duck 0 turkey 1 2 frog 1 

sheep 1 ? 4 

magpie 1 

? 5 

10. cuckoo 0 hen 4 1 turkey 2 

rooster 1 hen 1 

? 3 rooster 1 

? 2 

11. katt 7 ? 1 7 

12. cow, pig 

(cow) 

2 sheep 2 5 (cow) frog 1 

owl 2 ? 1 

? 2 

13. rooster 7 hen 1 5 ? 2 

14. mouse 0 squirrel 1 1 bird 2 

great titmouse 1 sparrow 1 

small bird 1 chicken 1 

snake 1 ? 2 

? 3 

15. dog (chin.) 1 chin. peacock 1 0 Chin. peacock 1 

pekingese dog ? 6 

rabbit 1 

? 5 

16. dog 1 duck 1 1 frog 1 

? 6 duck 1 

? 4 

207

6.4.4 Discussion of the test on identifying animal 

calls 

The animals that were correctly identified by all listeners (who gave an 

answer) were: dog (in Hungarian, [vau vau]), cow (in [mπ´ 

Icelandic, :]) 

and cat (in Korean, [njaON]). The Arabic word for the dog’s bark, [hab: 

hab:] and the Korean [maN maN] were not as accurately identified. The 

three instances of the duck's sound (from Hungarian, Icelandic and 

Estonian) all gave a variety of answers. Only one animal sound was not 

identified by anybody, [hap hap] (Hungarian duck). The animal sounds 

that were identified by only 1 person were [bra: bra:] (Icelandic duck), 

[pakUk: pakUk:] (Korean cuckoo, however, was interpreted as other 

birds), [tSi tSi] (mouse in Urdu, however, interpreted as various small 

animals), [maN maN] (Korean dog). 

In general it can be said that certain animal calls were more difficult to 

identify, others were easier. The words for the duck sounds were difficult 

while the word for the cat's meow was easy, in spite of the fact that the 

word chosen for a meow for the test was the one that was most deviant. 

There was no clear difference between the larger group of Swedish 

speakers and speakers of the other languages. 

Several times the informants guessed on the correct kind of animal, even 

though they did not give exactly the expected answer, e.g. 'bird' (see 6, 8, 

10) or 'small animal' (see 14). This means that they identified the size of 

the animal in question. Some subjects were more unsure than others. No 

subject guessed throughout on certain animals. A few subjects were more 

imaginative and more specific. 6 

6.5 Conclusions of studies of expressive 

interjections and expressions of animal calls 

Interjections and expressions of animal calls exist in all the languages in 

the study. There are similar sounds and meanings in all the languages, but 

there are also categories that seem to be specific to a certain language, 

e.g. Icelandic has a special command 'go away' directed to sheep. The 

expressions of the interjections are not the same in all the languages, but 

on the other hand they are not totally different; for some categories they 

6 In question 15 there might be an error; somebody must have whispered "Chinese" aloud. 

It is still interesting that several subjects judged "Chinese" to be fitting. 

208

are very similar, e.g. 'sneeze', 'good taste' cf. table 6.3. These expressions 

have an onomatopoeic basis which can be conventionalized in different 

ways in the different languages (the expressive interjections also have an 

indexical basis, i.e. a bodily reaction produces a sound, e.g. a sneeze or a 

scream, cf. 2.4). The phonological system influences the perception of the 

sounds and the choice of an adequate expression for imitation, e.g. [D] in 

Icelandic. There are, however, tendencies for the whole material: most of 

these consonants are labial or dental. The vowels are few and mostly [i], 

[a], [u]. 

The study of expressions of animal calls shows similar phonological 

tendencies (the semantic categories investigated were fixed, as specific 

animal calls were asked for): The imitations of the animal calls are not the 

same in all the languages, but on the other hand they are not totally 

different. Also, some expressions of animal calls are more alike in the 

different languages, e.g. Persian cat [miau] and Finnish cat [miau] while 

others vary more, e.g. Korean pig [kulkul] and Swedish pig [nØf:nØf:] cf. 

table 6.4. The imitations of animal calls can be conventionalized in 

different ways in the different languages, e.g. Swedish has [P] or [¨] 

where other languages have [u], for example Swedish [kPkelik¨:] and 

Hungarian [kukuriku]. The most common vowels are [i], [a], [u], i.e., 

closed or open vowels. 

In the interpretation test of expressions of animal calls (cf. 6.4.2–6.4.4), 

some animals were identified by all listeners (dog, cow and cat), some 

less correctly identified and one animal was not identified at all (duck, 

Hungarian). The interpretation test shows that, given the context of 

animal calls, it was quite easy for listeners to interpret animal calls from 

languages other than their own. 

Prosody, voice quality and gestures are most probably important for both 

interjections and imitation of animal calls. These forms for expression 

need to be studied further. 

209

6.6 Test of cross language interpretation of 

Swedish onomatopoeic and other sound symbolic 

words 

6.6.1 Method 

In order to see how non-Swedish speakers interpret Swedish 

onomatopoeic and other sound symbolic expressions, an experiment 

(similar to that in 7.1.3) was performed. 

The subjects were six persons who did not have Swedish as a first 

language, one each from of the languages Arabic, Spanish, German, 

Dutch, Ibo and English. 

A list of 15 written words was presented and read aloud to each subject. 

One of the words was repeated. Three of the words were not real ones 

but were instead constructed out of phonesthemes. The subjects were told 

to write down their answers in a language of their own choice. The words 

are: 

fjompig 

skvalpa 

skrälle 

pjaltig (constructed) 

vresig 

glansig 

slabbig 

bjaltig (constructed) 

blankig (constructed ending) 

pladdig (constructed) 

trumpen 

kladdig 

slabbig 

fladdrig 

grubbel 

stripig 

The subjects were instructed to try to interpret each word and suggest an 

appropriate meaning. What is here called the conventional interpretation 

of a constructed word is one where the answer has a semantic feature that 

210

can be connected with a certain consonant cluster in accordance with the 

analysis of chapter 4. These initial clusters: pj-, bj-, pl- and bl- were 

successful in different parts of the test in chapter 7 (especially pj) and 

could thus be assumed to represent a Swedish norm. 

Subjects who knew some Swedish were told to mark the words that they 

knew already. After the test the subjects were free to orally elaborate on 


6.6.2 Results of interpretation of cross language 

Swedish onomatopoeic and other sound symbolic 

words 

The subjects seldom interpreted the words in the conventional way. 

However, nearly all of the answers given belong to semantic features of 

the model in 2.6.2 and the semantic features of chapter 4. It thus seems 

that certain semantic categories are preferred to be expressed sound 

symbolically, but that the expression can vary. 

The subjects will first be accounted for one by one: 

Arabic speaker : This subject had very little knowledge of Swedish and 

knew none of the words. He answered in English. He guessed according to 

convention on one word. The meaning suggestions belonged to the 

following categories: The ones interpreted in an acceptable way, as 

concerns phonesthemes, were: 

fjompig - big mouth ('pejorative') 

pjaltig - speaks wrong ('pejorative', 'talking') 

slabbig - slap him/her on face ('beat') 

bjaltig - speaks wrong ('pejorative', 'talking') 

trumpen - drum’s sound ('sound') 

fladdrig - eruption ('quick or strong movement') 

For some of the misinterpreted words (misinterpreted in terms of 

phonesthemes), the semantic categories of the given answers were 

nevertheless in accordance with the sound symbolic categories of the 

model in 2.6.2 (These misinterpreted words are labeled "unconventional 

but possible category" in table 6.7). 

211

'wetness' 

'light' 

'quick or strong movement', 'wetness' 

'quick or strong movement' 

The responses that did not fit into any category of the semantic model (or 

were difficult to interpret) are: shut up, close the door, kind of penalty, 

intuitive, my country. 

Spanish speaker : This subject had a good knowledge of Swedish, both 

passively and actively. Her answers were given in Spanish, after which 

they were translated. The subject knew 4 of the words and guessed 

according to convention, in terms of phonesthemes, on 3. The ones 

interpreted in an acceptable way, as concerns phonesthemes, are: 

skrälle - chiming ('sound') 7 

blankig - opaque ('light') 8 

As for the Arabic speaker, for some of the misinterpreted words (in 

terms of phonesthemes), the semantic categories of the given answers 

were in accordance with sound symbolic categories of the model in 2.6.2. 

'adhesion' 

'beat' 

'long thin form' (2) 

'rough surface structure' 

'destruction' 

German speaker: This subject had a good passive knowledge of Swedish, 

but he did not know any of the words on the list. The answers were given 

in English. He guessed according to convention on 3 words. The words 

that were given an acceptable interpretation, in terms of phonesthemes, 

were: 

fjompig - condescending ('pejorative') 

vresig - angry ('mental feeling') 

glansig - shiny ('light') 

7 Skr- does not imitate a chiming sound at all, but in the previous analysis no distinction 

has been made between different kinds of sound. 

8 This is of course the opposite of 'light'. 

212

lankig - shiny ('light') 

pladdig - looks like a blot of e.g. yogurt ('wetness') 

fladdrig - something which is decomposed, thin ('thin form') 

The remaining meaning suggestions can be classified in the following 

semantic fields: 

'small size' 

'rough surface structure' (wrinkled skin of tomato) 

'form' (distorted) 

'pejorative' (distorted; heavy, fat, uncontrolled; unordered) 

'mood' (aggressive - of females) 

More difficult to classify in the previously discussed categories are the 

answers: "powerful"; "no more energy". 

Dutch speaker : This speaker had studied Swedish for a year and was very 

fluent. She had lived in Sweden for a very short period. She gave her 

answers in Swedish or English. She knew 1 of the words and guessed 

according to convention on 1 word. The words that were given a 

plausible interpretation, in terms of phonesthemes, were: 

skvalpa - call names at someone ('talking', 'pejorative') 

skrälle - say something with a lot of sound ('sound', 'talking') 

pjaltig - snobbish ('pejorative') 

slabbig - unorganized ('pejorative') 

trumpen - garbage ('pejorative') 

kladdig - to write in an ugly way ('pejorative') 

stripig - something with many stripes ('long thin form') 

For the misinterpreted words (in terms of phonesthemes), the semantic 

categories given were, however, in accordance with the model: 

'form' 

'mental feeling' 

'round form' 

'quick or strong movement' (2). 

One answer could not be classified according to the model: 'void'. 

213

Ibo (and English) speaker: This speaker knew none of the words and did 

not guess according to convention in any case. The words that were given 

a plausible interpretation (acceptable category), in terms of phonesthemes, 

were: 

slabbig - lazy ('pejorative') 

pladdig - dirty ('pejorative') 

trumpen - loud ('sound') 

grubbel - grumble ('talking') 

The semantic categories given were in accordance with the model 

(unconventional but possible category) for some of the misinterpreted 

words (in terms of phonesthemes): 

'putting together' 

'bodily feeling' 

Seven of the answers could not be classified according to the model 

(unconventional category): deep, slowness, positive, closed, tight, empty, 

wide. 

English speaker This subject had good knowledge of Swedish, Norwegian 

and German and answered in English. She knew 4 of the test words and 

guessed according to convention on 2 of the words. The words that were 

given an acceptable interpretation, in terms of phonesthemes, were: 

skvalpa - to gossip ('talking') 

skrälle - to complain, whine ('talking') 

vresig - twisting ('quick or strong movement') 

fladdrig - flimsy ('thin form') 

grubbel - annoying, irritating stuff ('bad mood') 

However, the semantic categories given were in accordance with the 

model (unconventional but possible category), for some of the 

misinterpreted words (in terms of phonesthemes): 

'round form' 

'adhesion' 

'sound' 

214

One of the answers could not be classified according to the model 

(unconventional category): tricked 

6.6.3 Conclusions from the test on cross language 

interpretation of Swedish onomatopoeic and sound 

symbolic words 

The following table, 6.7, shows the numbers of answers with different 

degrees of accuracy for each subject. 

Table 6.7 Summary of the numbers of answers with different degrees of 

accuracy in the cross linguistic word interpretation study. 

subject’s 

language 

known 

words 

guess in 

accordance 

with 

conven- 

tion 

accept. 

category 

215 

unconventional 

but 

possible 

category 

unconventional 

category 

Arabic 0 1 6 4 5 0 

Spanish 4 3 2 6 0 0 

German 0 3 6 5 2 0 

Dutch 1 1 7 5 1 0 

Ibo 0 0 4 2 7 2 

English 4 2 5 3 1 0 

sum 9 10 30 25 15 2 

Few of the words were known to the subjects, not even to those who had 

good knowledge of Swedish. There were also quite few correct guesses. 

Their performance is better when their answers are classified into 

semantic categories which are related to phonesthemes of the initial 

clusters (column "acceptable category"). 

The frequencies of words that were judged according to convention (in 

terms of phonesthemes) are shown in table 6.8. 

All 15 test words are represented in the results and there is no 

great preference for any word or words. 

no 

answer

Table 6.8 Frequencies of words that were most often successfully 

interpreted according to the Swedish norm, by the six speakers. 

Test words Number of speakers 

fladdrig 3 

skrälle 3 

trumpen 3 

blankig 2 

fjompig 2 

grubbel 2 

pjaltig 2 

pladdig 2 

skvalpa 2 

slabbig 2 

vresig 2 

bjaltig 1 

glansig 1 

kladdig 1 

stripig 1 

Table 6.9 shows which semantic categories were used most often, when 

the subjects guessed in an unconventional way (i.e. there is no possibility, 

according to the lexical analysis of chapter 4, that the consonant cluster in 

question can have this meaning). 

This result means that even when subjects guess in an unconventional way, 

they still guess within the semantic categories of the model. The answers 

that can not be classified within the semantic categories of the model (the 

column “unconventional category” of table 6.7) are fewer. Most of them 

were produced by the Arabic and Ibo speaker, which suggests an 

influence of cultural (or linguistic) differences, i.e. European vs. non- 

European. 9 

9 These results are very preliminary and clearly a larger investigation is needed. 

216

Table 6.9 Most commonly preferred meanings for unconventional but 

possible meanings. 

semantic features number of 

words 

quick or strong movement 4 

wetness 2 

adhesion 2 


rough surface structure 2 

round form 2 

form 2 

pejorative 1 

putting together 1 

bodily feeling 1 

mood 1 

mental feeling 1 

small size 1 

destruction 1 

beat 1 

sound 1 

light 1 

6.7 General conclusions and discussion of the 

cross language studies 

The following conclusions can be drawn from the different contrastive 

studies reported in this chapter: The Thesaurus study has shown that the 

expression of phonesthemes can differ between languages as closely 

related as English and Swedish. In some phonesthemes, however, the 

expression (and meaning) is the same. 

The study of interjections shows that there are similarities of expression 

as well as of meaning between unrelated languages. There is variation of 

expression within semantic categories, but certain sounds are still 

preferred. 

The study of expressions for animal calls in different languages shows 

that there are similarities as well as differences. The understanding test 

217

showed that it is quite easy for subjects to guess correctly on expressions 

for animal calls in other languages. Some animals were more difficult to 

identify. Often the type of animal was identified, e.g. bird or small 

animal. 

The last study, on interpretation of Swedish sound symbolic words and 

non-words (constructed out of phonesthemes), showed that it was quite 

difficult for subjects to interpret the meanings in accordance with the 

Swedish norm, but that the misinterpretations most often were within the 

semantic categories of the model in 2.6.2. 

The general conclusion from these studies is that there are greater 

differences between languages for expression than for meaning, in sound 

symbolism. The more onomatopoeic expressions are easier to interpret 

than other sound symbolism for speakers of other languages. 

Interpretation of other sound symbolism often goes wrong (because 

expressions differ in different languages), but the semantic categories 

guessed on are most often within the semantic model of this thesis. 

218

7 Experiments with constructed words containing 

phonesthemes 

7.1 Production and understanding 

In section 1.4 the hypothesis was formulated that phonesthemes are 

productive, to a greater or lesser degree. In order to test this hypothesis I 

have carried out several experiments. The purpose of the experiments 

that will be described in this chapter is to test this hypothesis, also in 

more detail for some of the phonesthemes. In chapter 4 the results of the 

lexical analysis made by one person were presented. The results of this 

analysis are to a certain extent dependent on the idiosyncrasies of the 

individual subject, the material and other circumstances and are therefore 

preliminary. Therefore the results of the lexical analysis were used as a 

basis for testing the sound symbolic value of certain consonant clusters on 

a great number of individuals. 

Method and materials 

The tests (see Appendix 3) are constructed as free choice, multiple-choice 

and matching tests. 

Test 1.a. is a free choice test, which goes from meaning to expression, to 

test the production of sound symbolism, e.g. "Invent a short word for 

somebody who is stupid". 

Test 1.b. is a forced choice test which also goes from meaning to 

expression. An example of a task is: "Which of the following words fits 

best for a person who is silly: a) pjotig b) brotig c) splotig?" Only one 

answer should be possible according to the previous analysis. For some of 

the questions, one of the answers is supposed to be clearly contradictory, 

i. e. the expected answering score is zero (where there are contrasting 

meanings, e. g. 'dry'-'wet'). In a few questions a word which sounds 

similar to the test word, but with a non-expected cluster is included to test 

which is more important, word analogy or sound symbolic clusters, e.g. 

"Which of the following words best describes a broken (trasig) object". a) 

bjatig b) skratig c) tratig?" Skr- is the expected cluster but tratig is very 

similar to trasig. 

219

Test 2.a is a free choice test which goes from expression to meaning in 

order to test the understanding of presumptive sound symbolic clusters, 

e.g. "What would be a good meaning for the word fnotig?" 

In test 2.b, a forced choice test (which also goes from expression to 

meaning), each nonsense word shall be matched with a certain number of 

recurring meanings of a certain abstractness, e.g. "something which is 

soft". More specifically, for each nonsense word tested there will be 3 

alternative meanings, only 1 of which is possible according to the 

previous analysis. An example of this is "What do you think slatig means: 

a) somebody who is silly b) something that is dry c) something which is 

unpleasant?" 

Test 3 is a matching test where the subject has to choose between two 

different meanings and two different expressions. An example of this is: 

Which word best describes a thing that is wet and which word best 

describes a thing that is dry: fnottig or skvottig?" 

For the tests 2a and b, nonsense words were constructed from initial 

consonant clusters that are commonly used for onomatopoeia and other 

sound symbolism, according to the lexical analysis in chapter 4. To avoid 

motivated meaning appearing in the non-initial part of the neologism the 

words constructed are short. Also the endings (-t or -t:) of the neologisms 

were checked in Svensk baklängesordbok (1981) in order to a v o i d 

motivated meanings in the endings. 

A complication with using monosyllables is that this syllable structure, 

when repeated in a list, gives an impression of 'beat' or 'sound imitation', 

especially when it ends in a long consonant. For this reason the test words 

do not have a long consonant (except in a few test cases). Furthermore, 

the test words have been added with the semantically neutral ending -ig. 

A problem with this might be, however, that in using longer words, more 

consonants and vowels will be involved, and it is more difficult to keep 

control over entire test words. 

220

Aside from constructing monosyllabic words (with an added ending -ig), 

another simplification is that in natural sound symbolic words, the 

semantic feature, e. g. 'pejorative', of an initial consonant cluster is often 

repeated in the final consonant cluster (cf. 5.6). 

As was shown in diagram 4.5, the 13 most common semantic features of 

Swedish initial clusters are, in descending order: 'pejorative', 'sound', 

'long thin form', 'quick or strong movement', 'wetness', 'talking', 'light', 

'diminutive', 'round form', 'walking', 'destruction', 'winding form', 'bad 

mood'. These semantic features were used in the tests except for 'sound', 

'diminutive', 'round form' and 'way of walking' 1. Also some more 

infrequent features, shown in the list below, were used. The formulations 

on the test sheets correspond to the semantic features above in the way 

shown in table 7.1. 

Table 7.1 The semantic features corresponding to the formulations on the 

test sheets. 

formulations semantic features 

löjlig (silly) pejorative 

fånig (silly) pejorative 

obehaglig (unpleasant) pejorative 

dum och klumpig (stupid and 

clumsy) 

pejorative 

pratar dumheter (talks nonsense) pejorative, talking 

rak (straight) long thin form 

smal form (narrow form) long thin form 

rörelse fram och tillbaka (movement 

back and forth) 

quick or strong movement 

plötslig rörelse (sudden movement) quick or strong movement 

(mycket) blöt ((very) wet) wetness 

1 'Sound' was not tested since it is usually easier for speakers to produce sound imitative 

words. 'Diminutive', 'round form', and 'walking' were not included because the final lexical 

analysis was not completed at the time the test was constructed, and therefore these categories 

were not included. The category of 'form' is, however, well represented in the study. 

221

hur en människa kan låta (how a 

person can sound) 


starkt lysande (shining brightly) light 

trasig (broken) destruction 

slingrig form (winding form) winding form 

arg (angry) bad mood 

på dåligt humör (bad mood) bad mood 

ihålig (hollow) hollow form 

pekar åt olika håll (pointing in 

different directions) 

separation 

går ihop (comes together) putting together 

(mycket) torr ((very) dry) dryness 

tydlig hård ytstruktur (pronounced 

hard surface structure) 

rough surface structure 

följsam ytstruktur (adaptable surface 

structure) 

soft consistency 

sitter fast (sticks) adhesion 

vidhäftande (sticking) adhesion 

hårt slag (hard beat) beat 

The most common consonant clusters, in absolute numbers and 

percentally, are shown in diagrams 4.1 and 4.4. In this test all of these 

clusters except nj- are used. The tests includes the clusters: kl-, gr-, vr- 

, spj-, str-, sp-, mj-, kv-, skv-, skr-, kr-, bl-, sm-, gr-, st-, tr- 

, bl-, br-, spl-, gn-, spr-, fr-, sk-, kn-, gl-, sv-, dr-, pl-, sl-, 

sn, fl-, bj-, pj-, fn- and fj-. 

The test words are grotig, gratig, vrotig, vratig, spjotig, spjatig, stratig, 

spatig, mjatig, kvatig, skvatig, skratig, bjetig, bjatig, kretig, kratig, bletig, 

blatig, smatig, statig, tratig, platig, fnatig, krotig, strotig, trotig, pjotig, 

brotig, splotig, snitig, gnitig, gritig, klatig, spratig, smitig, gritig, spitig, 

platig, snatig, smatig, dratig, blattig, mjattig, flattig, snattig, smattig, 

drattig, etc (see Appendix 3). 

222

Test procedure 

The test questions were written four to a page and the pages were stapled 

together. The test questions were mixed in order to keep the test words 

from influencing each other (e.g. no two questions about 'light' were put 

on the same page). The subjects were told not to go back and check and 

were not given time for that either. Half of the subjects were tested in the 

order: 1 (a)(b) -2(a)(b) -3 and the other half in the order 2 (a)(b) -1(a)(b) 

-3 (see Appendix 3). 

Test (a) which is free production, is the most difficult, test (b) which is 

multiple-choice, is less difficult, and test 3, which is matching, is the 

easiest. The largest number of presumed answers are expected for test 3 

and the least number of presumed (or consistent) answers are expected for 

1a and 2a. 

The subjects were linguistics students at the beginner's level, with 

Swedish as a first language, males and females, ages 21–57. There were 

15 informants. 14 of them took the tests from meaning to expression and 

15 informants took the tests from expression to meaning. Thirteen of 

them took the matching test. 

7.1.1 Forced choice for production – from 

meaning to phonological (graphic) form 

Of the 39 test questions where a meaning was presented along with three 

alternatives for phonetic (graphic) form, one of which was possible 

according to the model, 28 test questions showed a majority for expected 

answers, 4 had a shared majority for the expected and an unexpected 

answer and 7 showed a majority for an unexpected answer. Sometimes 

the majority was overwhelming, sometimes not, see table 7.2. 

The following 19 (out of 39) test questions received the best results: no. 

37, 38, 39, 35, 30, 31, 32, 29, 21, 22, 23, 24, 13, 16, 9, 10, 5, 7 and 8 

(cf. Appendix 3). These questions concerned the features and words 

shown in table 7.2. 

223

Table 7.2 The semantic features of the 19 test questions that received the best 

results in the forced choice test of choosing phonological form from 

different semantic features. Words are presented in order from best match 

with expected to worst. 14 informants. 

semantic feature word ratio of expected 

answers 

'pejorative' fjotig 14/14 

'dryness' fnotig 13/14 

'long thin form' stratig 13/14 

'wetness' spatig 13/14 

'bad mood' vratig 13/14 

'pejorative' pjotig 13/14 

'wetness' platig 12/14 

'pejorative' bjatig 12/14 

'quick or strong 

movement' 

flattig 12/14 

'adhesion' klatig 12/14 

'bad mood' tratig 12/14 

'separation' spratig 11/14 

'hollow form' gratig 11/14 

'wetness' svotig 11/14 

'quick or strong 

movement' 

skottig 11/14 

'pejorative' drotig 11/14 

'hollow form' grotig 10/14 

'winding form' krotig 9/14 

'talking' snatig 9/14 

Seen in relation to the number of instances of each feature there were in 

the test, the semantic features shown in table 7.3 are the most successfully 

interpreted: 

224

Table 7.3 The most successfully interpreted semantic features of table 7.2., 

14 informants. 

'wetness:' 3/3 2 (100%) 

'hollow form': 2/2 (100%) 

'separation': 1/1(100%) 

'dryness': 1/1(100%) 

'pejorative': 4/6 (67%) 

'bad mood': 2/3 (67%) 

'quick or strong movement': 2/4 (50%) 

'winding form': 1/2 (50%) 

'talking': 1/2 (50%) 

'adhesion': 1/2 (50%) 

'long thin form': 1/3 (33%) 

The questions that did not give the expected answers pertained to the 

features and words in table 7.4 (for a majority of subjects). 

Table 7.4 Semantic features for neologisms that did not give the expected 

answers. The substitutions of words (with frequencies) are shown in the 

right column. 14 informants. 

'winding form' stratig (6) instead of kratig (4) 

'light' kratig (7) instead of blatig (3) 

'light' snitig (8) instead of gnitig (2) 

'quick or strong movement' mjatig (6) instead of flatig (3) 

'light' snotig (9) instead of gnotig (2) 

'light' kretig (6) instead of gletig (4) 

It is clear that the 'light' feature gets least expected answers. The clusters 

tested are the well known bl-, gl- and gn- clusters, but obviously 

something more is required to get a light-associative effect. The 'light' 

words might also belong to a closed lexical class which would imply that 

gl- , gn-, blare not productive, unlike phonesthemes such as kl- or 

2 i.e. 3 of the 3 wetness words were identified by most subjects. 

225

pj-. Another possible explanation, in light of the results in 5.4, is that the 

vowels have to be taken into account. The lexical analysis showed that 

front vowels, and especially i, are much more common in root 

morphemes with the features 'light' or 'gaze'. Two of the constructed 

words in the test, gnitig and gletig, had front vowels (i and e) but were 

nevertheless not interpreted according to expectations. 

The feature most successfully coded in phonesthemes is 'wetness', for 

which 3 out of 3 words where identified by most subjects, 'hollow form' 

(2 out of 2), 'separation' (1 out of 1) and 'dryness' (1 out of 1), cf. table 

7.3. The feature that is the least successfully interpreted is 'light'; none of 

the 4 test words were interpreted according to expectations, cf. table 7.4. 

Since some of the vowels in the constructed words were in accordance 

with expectations, the non-expected results of the 'light' words can not be 

explained with reference to the importance of vowels. It seems rather that 

the 'light' phonesthemes are not productive. One explanation for why 

they are not productive could have to do with the fact that 'light' is the 

only category which is metaphorically connected with 'sound', i .e. by 

sense analogy. Many other categories are connected metonymically with 

sound, i.e. they occur simultaneously, e.g. 'movement', 'surface 

structure', 'wetness', 'form', or are more abstract (cf. 2.5, 2.6). 

7.1.2 Forced choice for understanding - from 

phonological form to meaning 

Out of the 38 test questions where a constructed word was presented 

along with three meaning alternatives, one of which was possible (or most 

possible 3) according to the model of 2.6.2 and the analysis of chapter 4, 

29 showed a majority for expected answers, 1 had a shared majority for 

expected and unexpected answers and 8 had a majority for an unexpected 

answer. Sometimes the majority was overwhelming, sometimes not. The 

test questions that had the best results were the following: 24, 32, 26, 29, 

21, 23, 25, 18, 10, 12, 5, 3, 4, 14. These questions concerned the words 

and features presented in table 7.5. 

3 The only exception is fnotig (question 23) where 'rough surface structure' is slightly more 

expectable than 'dryness'. However, as can be seen in table 7.5, fnotig is interpreted as 

'dryness' by 15 out of 15 subjects. 

226

Table 7.5 The words in the test questions that received the best results in the 

forced choice test of choosing meanings for different constructed words. 15 

informants. 

word semantic feature ratio of expected 

answers 

grotig 'hollow form' 15/15 

fnotig 'dryness' 15/15 

snattig 'talking' 15/15 

pjotig 'pejorative' 15/15 

skottig 'quick or strong 

movement' 

14/15 

spratig 'separation' 14/15 

grotig 'bad mood' 11/15 

fjotig 'pejorative' 13/15 

vratig 'bad mood' 13/15 

blatig 'pejorative' 12/15 

spitig 'long thin form' 12/15 

snatig 'talking' 11/15 

skratig 'rough surface 

structure' 

10/15 

flattig 'quick or strong 

movement' 

10/15 

The semantic features which were the most successful in the forced choice 

test of choosing meanings for different constructed words were 'hollow 

form' (for gr-), 'dryness' (for fn-), 'talking' (for sn-), and 'pejorative' 

(for pj-) 

The questions that did not give the expected results concerned the forms 

and meanings in table 7.6 (for a majority of speakers). 

227

Table 7.6 Words that did not give the expected answers. The substitutions of 

features. 15 informants. 

word substitutions of 

features 

gnotig 'talking' instead of 

'light' 

flatig 'long thin form' instead 

of 'quick or strong 

movement' 

gletig 'separation' instead of 

'light' 

gnitig 'sticking' instead of 

'light' 

glatig 'pejorative' instead of 

'light' 

smatig 'pejorative' instead of 

'beat' 

blatig 'hollow form' instead 

of 'light' 

228 

comment 

This is in fact more 

expectable according to 

the analysis in 4.5 

This is also expectable. 

There might be an 

influence from the 

word flat (flat). 

In this case a similar 

word gnatig has no 

influence. The third, 

but not chosen, 

alternative was actually 

'talking' which is also 

possible according to 

the model 

'Pejorative' is in fact 

also expectable 

according to the model 

'Pejorative' is in fact 

also expectable 

according to the model 

The semantic feature which is the least successful is 'light' (for gn-, gland 

bl-). Even where the vowel was front, as in gletig and gnitig, the 

result was not according to expectations. 

Summary of forced choice tests 

In the first test, 28 out of 39 test questions showed a majority for 

expected answers according to the model, and 19 of these received good 

results, i.e. at least 64% (9/14) expected answers (table 7.2). In the second 

test, 29 out of 38 test questions showed a majority for expected answers,

and 13 of these received good results, i.e. at least 67% (10/15) expected 

answers. 

In both tests the semantic feature which was the least successfully 

interpreted or produced was 'light'. Even when the vowel was front, 

according to expectations, the constructed words were not interpreted or 

produced according to expectations. It thus seems as if the phonesthemes 

connected with 'light' are not productive. In the test of meaning to 

phonological form, 28 of the 39 test questions showed a majority for 

expected answers. The six phonesthemes that were the most successfully 

coded were 'pejorative': fj-, 'dryness': fn-, long thin form': str-, 

'wetness': sp-, 'bad mood': vr-, 'pejorative': pj-. (The features most 

successfully coded totally were 'wetness', 'hollow form', 'separation' and 

'dryness'.) In the test from phonological form to meaning, 29 of the 38 

test questions showed a majority for expected answers. The six 

phonesthemes most successfully interpreted were 'hollow form': gr-, 

'dryness': fn-, 'talking': sn-, 'pejorative': pj-, 'quick or strong 

movement': sk- and 'separation': spr-. For both the interpretation and 

production tests, the most successful cases were 'dryness': fn- and 

'pejorative': pj-. Both of these are lexically low frequency clusters that 

are sound symbolic to a very high degree. 

7.1.3 Free production test from constructed words 

to meanings 

The free production test from constructed forms to meanings asked the 

question "What do you think would be a good meaning for ...?" This gave 

the results shown in table 7.7. 

229

Table 7.7 Free interpretations by 15 subjects of 6 constructed words. 

skvatig fnotig vratig krötig pjotig skratig 

härsken tråkig vrensk stöddig tokig skarp 

(rancid) (boring) (refractory) (stuck up) (mad) (sharp) 

blöt obeslutsam tjatig trög pjoskig glad 

(wet) (irresolute) (nagging) (sluggish) (mawkish) (happy) 

tröttsam 

(tiring) 

- sned 

(crooked) 

ihopklumpad 

(lumped 

together) 

barnslig 

(childish) 

bristfällig 

(defective) 

blöt 

(wet) 

torr och 

fladdrig 

(dry and 

flapping) 

vriden 

(twisted) 

trött och lite 

sjuk 

(tired and a 

little ill) 

fånig 

(silly) 

snurrig 

(giddy, 

crazy) 

blöt fnasig grov tillgjord petig dålig 

(wet) (chapped) (coarse) (affected) (finical) (bad) 

kaxig krokig tjatig knotig dum glad 

(cocky) (crooked) (nagging) (bony) (stupid) (happy) 

schlager- pustande skrytsam sluddrande gnetig trög 

musik 

(popular 

song) 

(panting) (boastful) (slurring) (crabbed) (sluggish) 

flamsig löjlig pratig rörig pjåskig kantig 

(silly) (ridiculous) (chatty) (messy) (mawkish) (angular) 

galen gammal arg mosig ojämn knyckig 

(crazy) (old) (angry) (fuddled) (uneven) (jerky) 

blöt knäpp vrängd krokig plaskig smal 

(wet) (stupid) (turned 

inside out) 

(crooked) (splashy) (narrow) 

något kantigt 

och blött 

(sth angular 

and wet) 

något svårt 

(sth difficult) 

pratsam 

tråkig 

(talkative 

boring) 

inkrökt 

(focused on 

one´s own 

problems) 

besvärlig 

(troublesome) 

vissen 

(withered) 

skvalande skrumpen tjatig urgröpt fånig full med 

(pouring) (shrunk) (nagging) (hollowed (ridiculous) revor 

out) 

(full of rips) 

kantig eller 

hård 

(angular or 

hard) 

knölig 

(knobbly) 

arg 

(angry) 

berusad 

(drunk) 

liten 

rar 

(small 

sweet) 

trasig 

(ragged) 

230

skratta 

(laugh) 

halvdålig* 4 

(half bad) 

frysa 

(freeze) 

knäpp* 

tokrolig* 

(crazy 

funny) 

vrida 

(wring) 

halvnerriven 

vägreklam* 

(half torn 

down 

advertizement 

by the 

road side) 

krånglig 

(trouble- 

231 

some) 

försupen* 

(sottish) 

passande 

(suitable) 

löjlig* 

liten* 

(ridiculous 

small) 

skrattande 

(laughing) 

en härjad 

smal kvinna* 

(a worn and 

haggard thin 

woman) 

1. skvatig gives 6/15 expected answers, i.e., words with a semantic 

feature 'wetness'. Of the other meanings only one gives an answer with a 

feature 'wetness' – pjotig 

2. fnotig gets the expected 'pejorative' in 5/14 (perhaps more depending 

on which answers should be classified as 'pejorative'). It also gets the 

expected 'dry' in 2/14. (cf. diagram 4.20). 

3. vratig gives 3/15 clear answers containing the feature 'bad mood' and 

some other answers which are less clear. 

4. krötig does not give words with the expected semantic feature 'winding 

form' but rather point to some sort of 'pejorative', which is the second 

expected feature for kr- (cf. diagram 4.19). 

5. pjotig, where the expected semantic feature 'pejorative' gives 10/15. 

6. skratig, where expected semantic feature 'destruction' gives 7/15. 

'Pejorative' has a higher percentage of skr- than 'destruction', but 

'destruction' is typical for skr-, cf. table 4.16. 

The nonsense word that is most successful is pjotig (pejorative). Pj- is 

percentally the fourth most sound symbolic cluster (cf. diagram 4.4) and 

is mainly 'pejorative' (cf. diagram 4.23). The second most successful 

word is skratig, which only belongs to the eleventh most sound symbolic 

cluster (percentally), skr-. However, it is a typical cluster for 

'destruction'. In third place comes skvatig (wetness). 

In other words, no constructed word is interpreted as expected by all 

subjects, but all of the constructed words, except krötig (winding form), 

are interpreted correctly by some subjects. Krötig is, however, given a 

4 The * marks the answers from the subject who deviated most in the "matching test of 

nonsense words and meanings" (see 7.1.5).

second best interpretation, 'pejorative'. The success ranking of these six 

phonesthemes are thus, in order from best match to worst: 'pejorative' 

(for pj-), 'destruction' (for skr-), 'wetness' (for skv-), 'pejorative' (for 

fn-), 'bad mood' (for vr-), 'winding form' (for kr-). 

There are examples both of where the associations have gone to a word 

with a similar phonetic form, and examples of where the phonestheme has 

an influence in spite of a similar sounding word 5. An example of the first 

case is krötig where the associations often seem to go to kröka (drink 

alcohol): slang word): sluddrande, berusad, försupen. Skratig sometimes 

gives associations to skratt (laughter): glad, skrattande. There are, 

however, many answers under skratig which contain the meaning 

component of 'destruction': bristfällig, lite sjuk, dålig, vissen, full med 

revor, trasigt, en härjad smal kvinna. 

The conclusion is that these nonsense words are interpreted in accordance 

with expectations by some subjects, but not all. The nonsense words have 

a phonesthemic meaning potential that is used by some subjects in the test 

situation. The phonestheme most easily interpreted was 'pejorative': pj-. 

Explanations for this might have to do with the fact that pj- is a lexically 

low frequent highly sound symbolic cluster. 'Pejorative' is also the most 

frequent semantic feature according to the lexical analysis. The cluster 

most difficult to interpret was 'winding form': kr-. Kr- is lexically high 

frequent but also sound symbolic to quite a high degree, however not 

only with the feature 'winding form'. It seems as lexically low frequent 

highly sound symbolic clusters are easy to access. 

7.1.4 Free production from meaning to constructed 

word. 

The question "Make up a short word for somebody or something which is 

(has) ...?", where subjects were supposed to invent an expression for one 

of the meanings 'silly', 'winding', 'angry', 'dry', 'wet', and 'rough 

surface', gave the results shown in table 7.8. 

5 There are, as is well known, many types of relations in the lexicon, and thus relations 

between words, as well as between phonesthemes, cf. e.g. Garman (1990). 

232

Table 7.8 Words produced by 14 subjects for 6 different meanings. The 

words are non-words. Most of them follow the Swedish phonotactic rules. 

'silly' 'winding' 'angry' 'dry' 'wet' 'rough 

surface' 

smurk slirv gurp lirv slish pritt 

spjal islig kral spri subl flarb 

flong plyr orn krasp svåsk donk 

fjän ril vrag fnat plat raster 

koos pis faaby kirl trippp hitt 

flutt sjling grol fnus svurp gilb 

floppig siloln börr kritto slasli teppig 

pjöl krel vram fnöl svomm knupp 

fjutt kril burr krasp mollo skrak 

krumpig vrinlig trossk prusskig schjaflig dank 

fnölp tirori rark fnuskig splass nin 

- slio furn kln splurr klik 

loup* 6 evans* hram* srrats* paupe* kovo* 

knork vrom dramm frok blu tlak 

When it comes to the production of forms, we are not restricted to initial 

clusters; the subjects have also used vowel qualities, reduplication and 

final clusters to express the different meanings. However, initial clusters 

are much more common in these neologisms. An analysis of the 

neologisms produced follows below. 

Semantic feature 'pejorative' 

For the meaning 'silly' (semantic feature: 'pejorative') the produced 

initial clusters in the test are, in frequency order: fl- (3), fj- (2), and then 

one instance each of pj-, fn-, kn-, kr-, sm- and spj-. These results can 

be compared with diagrams 4.6 and 4.7 where we see that pj- fn- and fjare 

very frequent percentally, while kn- and kr- are very frequent in 

absolute numbers. Thus both percental and absolute frequency of 

phonesthemes in the lexical analysis of chapter 4 correlate with the 

6 The * marks the answers from the subject who deviated most in the "matching test of 

nonsense words and meanings" (see 7.1.5). 

233

clusters that the subjects use in free production of sound symbolic words. 

The most frequently produced cluster flis quite frequent in absolute 

numbers but not so frequent percentally. Only one of the produced words 

contains a cluster which is not pejorative at all, namely spj-. 

The final consonant clusters produced are, in frequency order: -rk (2), 

and one instance each of -mp and -lp. These clusters can be compared 

with tables 5.2 and 5.3 and diagrams 5.1 and 5.2. The tables show that - 

mp can be pejorative (but is usually not), and that -rk and -lp are not 

sound symbolic at all. The conclusion here must be that the final 

consonant clusters are of less importance than the initial ones in 

producing new pejorative words. 

Semantic feature 'winding form' 

For the semantic feature 'winding form', the produced initial clusters in 

the test are, with two instances each: kr-, vr-, sl- pl- and sjl-. These can 

be compared with table 4.16. The clusters that have the feature 'winding 

form' are: kr- (typical) and sn- (possible). In other words, kr- is 

expected. The cluster sjl- breaks the phonotactic pattern. 

The final consonant clusters produced are, with one instance each: -rv, - 

ln and -ns. Of these three, -rv and -ns can be sound symbolic according 

to the analysis in chapter 5 (cf. tables 5.2 and 5.3). There are however, in 

this analysis, no final consonant clusters with the feature 'winding form' 

(only one word kringla). 

Instead, there seems to be an iconicity in any position in the word for the 

meaning 'winding form' reflected e.g. in a contrast between s and l, in 

the words slirv, islig, sjling, siloln, slio. As seen earlier (cf. for 

example diagram 4.1), sl- is also the most sound symbolic cluster in 

absolute numbers. This issue is left for further research. 

The vowels are mainly i, and there is a dominance for front, closed or 

half closed vowels (with high F2). 

234

Semantic feature 'bad mood' 

For the meaning 'angry' (semantic feature: 'bad mood') the initial clusters 

produced in the test are, in order of frequency: vr- (2) and one instance 

of each of gr-, kr-, tr-, dr- and hr-. The most frequent ones, vr-, as trand 

gr- (and only these clusters) also have the semantic feature 'bad 

mood' according to the earlier analysis. All the produced clusters except 

one have an r, which is what the conventional ones have in common. The 

last one, hr- breaks the phonotactic pattern. The phonestheme 'bad mood' 

seems to have the expression obstruent + r. 

The final consonant clusters produced are: -rn (2) (it is unclear if the 

subjects pronounced this as a cluster or as a retroflex) and one instance 

each of -rp, -rk, and -ssk. Obviously r dominates also in the final 

clusters, the only exception being -ssk. According to the analysis in 

chapter 5, there are no conventional final clusters with the semantic 

feature 'bad mood', so the initial cluster phonemes seem to have been 

transferred to the final position. 

There are no front, closed or half closed vowels in these words but there 

is a preference for vowels with lower F2. 

Semantic feature 'dryness' 

For the semantic feature 'dryness' the initial clusters produced in the test 

are, in order of frequency: fn- (4), kr- (3) and one instance each of fr-, 

pr-, spr-, srr-, kln (a whole word). The most frequent one fn- has the 

semantic feature 'dryness' according to the earlier analysis, and fnis the 

only cluster with the feature 'dryness'. The remaining ones consequently 

do not have the feature 'dryness'; srr- and kln break the phonotactic 

pattern. 

The final consonant clusters produced are: -sp, -sk, -rv, -ts and -dt. 

'Dryness' is not a feature of final clusters according to the analysis of 

chapter 5. The cluster -dt breaks the phonotactics. Again, the subjects 

seem to encode the semantic feature in the initial cluster. 

235

For 'dryness' there are more clusters that break the phonotactic pattern 

than for the other semantic features. Perhaps this has to do with the fact 

that 'dryness' is a very infrequent sound symbolic feature, e.g. in 

comparison with wetness. 

Semantic feature 'wetness' 

For the semantic feature 'wetness', the initial clusters produced in the test 

are, in order of frequency: sl- (2), sv- (2), spl- (2) and one instance each 

of pl-, tr-, schj- [Sj], bl- and bw-. The clusters with the feature 

'wetness' are two of the most common ones, sl- and sv- (but not spl-). 

The remaining ones, pl- and bl-, also have the feature 'wetness', but not 

tr-. Schj- [Sj] breaks the phonotactic pattern and bw- has the sound [w] 

which is not even a phoneme in Swedish. So, for this feature, four 

expected clusters are used, two unexpected ones and two unconventional. 

The final consonant clusters produced are: -bl, -sk, -rp, -sl(i) and -fl. 

Of these, the only final clusters for 'wetness', according to the analysis of 

chapter 5, is -sk. -bl is not phonotactically possible in the absolute final 

position. 

In both the initial and final clusters produced for 'wetness' the most 

common phonemes are the following: l (6 instances), s (5 instances) and 

p (4 instances), as could be expected, cf. diagrams 4.14 and 4.15. 

Semantic feature 'rough surface structure' 

For the semantic feature 'rough surface structure', the initial clusters 

produced in the test are: fl- (2) and one instance each of pr-, skr-, kn-, 

kl-, tl-. Of these only skr- has the feature 'rough surface structure'. Tlbreaks 

the phonotactic pattern. There are other cluster with this feature, 

fr- and kr-, according to the analysis in chapter 4, but they were not used 

by the subjects. Perhaps the formulation in Swedish : 'tydlig, hård 

ytstruktur' was not clear enough. 

The final consonant clusters produced are: -Nk (2), -rb and -lb. In the 

analysis of chapter 5, the only roots classified as 'rough surface structure' 

236

are knottra (ttr) and rispa-raspa (sp). Rough surface structure is hardly a 

feature of final consonant clusters. 

Discussion of free production tests from meaning 

to constructed word 

Subjects seem to encode semantic features in the initial clusters rather 

than in the final ones when producing new sound symbolic words. This is 

mirrored in the results of chapter 5: many of the semantic features are 

not conventionally encoded in specific final consonant clusters. However, 

subjects do produce final clusters in these words, but these final clusters 

seem to be either mirrored in the initial clusters or are haphazard. There 

is of course also the possibility that semantic features other than those in 

the study are relevant for final clusters. 

Subjects do produce clusters that break phonotactics, both initially and 

finally. For sound symbolically low frequent features (like 'dryness'), 

with few possible clusters, there are more unconventional forms 

produced. 

The semantic features that to the greatest extent were expressed according 

to the model were 'pejorative' (163 7), 'bad mood' (18) and 'wetness' 

(63). Less successfully expressed were 'winding form' (20), 'dryness' (1) 

and 'rough surface' (10). There is a tendency for frequent features to be 

more successfully coded. The initial consonant clusters of 'pejorative', 

with the exception of one, are according to expectations. The most 

commonly produced initial consonant clusters for 'bad mood' are in 

accordance with the phonesthemes vr-, tr-, gr-. Of the rest of the words 

produced all except one have an r. For the initial consonant clusters of 

'wetness' the most common phonemes are: l, s and p. The feature 

'winding form' stands out in that it uses the non-expected phonemes s and 

l in different positions of the words. This might be an effect of the 

trigger word 'slingrig form' ('winding form'). 

7 The absolute frequencies of these features, for the initial clusters, are shown in table 4.2. 

237

A matching test with context 

Yet another study was done in order to shed further light on the results 

presented in table 7.8. A matching test between the meanings and all 

columns of words of table 7.8 8 was performed on a group of 9 linguistics 

students with Swedish as their first language. The subjects were told to 

match the six columns of expressions with a random list of the six 

meanings. 

This test gave 100 % correct results. The results depend partly on the 

possibility of comparing the words in the different columns. We get an 

effect from context. An isolated reading of the columns might be 

somewhat more difficult, and might not yield the same degree of correct 

results. Even more difficult would be to read (or hear) one word at a 

time and suggest a meaning. Another interesting test would be to ask 

people to pick out the word most suitable for every meaning of each 

labeled column. 

7.1.5 Matching test of nonsense words and 

meanings 

In this test (Test 3), pairs of nonsense words, e.g. fnottig-skvottig, were 

to be matched with pairs of contrasting meanings, e.g. 'wet' - 'dry'. The 

subjects were 13 of those in tests 1 and 2. This test gave the following 

results, for each question: 

Question 1 (skv- 'wetness', fn- 'dryness'): 92 % expected answers 

Question 2 (str- 'long thin form', kr- 'winding form'): 84% expected 

answers 

Question 3 (spr- 'separation', kn- 'putting together': 84% expected 

answers 

Question 4 (skr- 'rough surface structure', mj- 'soft consistency'): 100% 

expected answers 

Question 5 (fj- 'silly' (pejorative), vr- 'arg' (bad mood): 100% expected 

answers. 

8 One column of words is e.g. slirv, islig, plyr, ril, pis, sjling, siloln, krel, kril, vrinlig, 

tirori, slio, evans, vrom. 

238

In question 1, one person (out of 13) deviated, on questions 2 and 3 two 

persons deviated. The same person deviated on questions 1, 2 and 3. In 

other words, one person stands for 3 out of 5 unexpected answers. (This 

is the person who's answers are marked with an * in tables 7.8. and 7.7.) 

7.1.6 Summary of results of tests on interpreting 

meanings and sounds 

The forced choice tests 

The forced choice tests (1b and 2b) had very good results for 19 out of 

39 test questions in the test from meaning to constructed words, and 28 

test questions showed a majority for expected answers. In the test from 

constructed words to meaning (1a and 2a) 14 out of 38 test questions 

showed very good results, and 29 test questions showed a majority for 

expected answers. This is an overall good result for the interpretation of 

the hypothesized phonesthemes. The phonesthemes that were most 

successfully interpreted in the test from meaning to constructed words 

were fj- 'pejorative', fn- 'dryness', str- 'long thin form', sp- 'wetness', 

vr- 'bad mood', and pj- 'pejorative'. The phonesthemes that were most 

successfully interpreted in the test from constructed words to meaning 

were gr- hollow form', fn- 'dryness', sn- 'talking' and pj- 'pejorative'. 

Combining the results of both tests, pj- 'pejorative' and fn- 'dryness' 

were the most successful. 

The question arises whether the interpretation of semantic features in 

general correlate with the most frequent features according to the lexical 

analysis. The features 'pejorative': pj- and 'dryness': fn- (both lexically 

small clusters with a high degree of sound symbolism) showed the highest 

expected correlations between meaning and sound among the six most 

successful features for both forced choice meaning to constructed word 

and constructed word to meaning tests (tables 7.2 and 7.5). It seems then 

that it is generally easier to interpret and code phonesthemes of lexically 

low frequent, highly sound symbolic clusters. The meaning that gave the 

smallest number of expected answers was 'light', (also for interpreting 

sounds). There were five test words with the expected feature 'light' and 

the clusters tested were bl-, gl- and gn-. Even when the vowel was in 

accordance with the expectations of the lexical analysis, i.e. having an i or 

239

e vowel like in gletig and gnitig, the words were not interpreted as 'light' 

by a majority of speakers, and the category 'light' was not coded with bl- 

, gl- or gn- by a majority of speakers. , To quite a great extent, gn- is a 

lexically low frequent, highly sound symbolic cluster, as is the case with 

bl- and gl. However, this does not seem to be enough for productivity. 

As mentioned above, the 'light' phonesthemes (which are not productive 9) 

have the only semantic feature which is metaphorically connected with 

sound. 

The free production tests 

The free production of meanings or words tests (1a and 2a) gave a 

variety of meanings and new sound symbolic words, many of which were 

predicted but some were not. In the free production test from form to 

meaning no constructed word was interpreted as expected by all subjects, 

but all of the constructed words (except one which is given a second best 

interpretation) are interpreted correctly by some of the subjects. The 

success ranking of the six phonesthemes are, in order from best match to 

worst: 'pejorative' (for pj-), 'destruction' (for skr-), 'wetness' (for skv- 

), 'pejorative' (for fn-), 'bad mood' (for vr-), 'winding form' (for kr-). 

The features (of the six tested) that were most successful were 

'pejorative', 'bad mood', and 'wetness'. Also the fact that all the test 

subjects willingly produced nonsense words, and meanings from nonsense 

words, is an interesting result. Thus, these nonsense words are interpreted 

in accordance with expectations by some subjects, but not all. The 

nonsense words have a phonesthemic meaning potential that is used by 

some subjects in the test situation. 

The free production test from meaning to form shows that phonesthemes 

are used. This test also shows that for all semantic features both percental 

and absolute frequencies of the lexical analysis correlate with the clusters 

that the subjects use in free production of sound symbolic words. Subjects 

seem to encode semantic features in the initial clusters rather than in the 

final ones when producing new sound symbolic words. They also produce 

clusters that break phonotactic rules, both initially and finally. The 

9 There may well be productive 'light' phonesthemes, but it might be the case that the 

expression side of these phonesthemes has to include a vowel, e.g. i. 

240

semantic features that to the greatest extent were expressed according to 

the model, were 'pejorative', 'bad mood' and 'wetness'. Less successfully 

identified were 'winding form', 'dryness' and 'rough surface'. There is a 

tendency for frequent features to be more successfully coded. 

For both free production tests, the most successful features (out of six 

possible) were 'pejorative', 'bad mood' and 'wetness'. 

Thus in both the forced choice tests and free production tests the most 

successful semantic feature is 'pejorative'. 

The matching test 

The matching test (c), where two words are to be matched with two 

semantic features, produced the highest percentages of expected answers. 

The results partly depend on the possibility to compare the word pairs. 

We get an effect from context. 

Summary and discussion 

Most of the phonesthemes are more or less successfully interpreted or 

coded, while some are more clearly not interpreted or coded. In table 7.9 

below, the phonesthemes that were the most successfully interpreted and 

coded in the experiments are presented, in order to show which are 

recurring in the different experiments. 

241

Table 7.9 The phonesthemes most successfully interpreted or coded in 

experiments with neologisms. 

Test: 

Phonestheme: 

most 

successfulphonesthemes 

leastsuccessfulphonesthemes 

forced 

choice 

from 

meaning to 

phonol. 

form 

pejorative: 

fjdryness:fnlong 

thin 

form: 

strwetness:spbad 

mood: 

vrpejorative:pj- 

light: 

gn- 

forced 

choice from 

phonol. 

form to 


hollow form: 

grdryness:fntalking:snpejorative:pjquick 

or 

strong 

movement: 

skseparation:sprlight:gnlight:gl- 

242 

free 

production 

from 


phonol. 

form 

pejorative 

(several 

clusters) 

bad mood 

(several 

clusters) 

wetness 

(several 

clusters) 


form, 

dryness, 

rough 

surface 

free 

production 

from 

phonol. 

form to 


pejorative: 

pj- 


form: 

kr- 

As can be seen from table 7.9 the most successful phonestheme is 

'pejorative': pj- and the least successful is 'light': gn- and 'winding form' 

(kr-). Pj- and gnare lexically low frequent, highly sound symbolic 

clusters. 

The clusters of most of the most successful phonesthemes in table 7.9 

above – fj-, fn-, vr-, pj-, spr-, skr- and mj- – are lexically low 

matching test 

rough 

surface 

structure vs. 

soft 

consistency: 

skr- vs mjpejorative 

vs. 

bad mood: 

fj- vs. vr- 

-

frequent, very sound symbolic clusters. Str-, sp-, gr-, sn- and skare 

not. The most successful semantic feature of table 7.8 is 'pejorative'. The 

general success of the feature 'pejorative' is discussed in chapter 8. Of the 

least successful phonesthemes, gn- is lexically low frequent, highly sound 

symbolic, but not gl- and kr-. The degree of success in these tests is not 

then exclusively restricted to the lexically low frequent, highly sound 

symbolic clusters. 

243

8 Summary and discussion 

The purpose of this thesis has been to study different aspects of sound 

symbolism – with special reference to Swedish. The largest part of the 

descriptive study has been devoted to Swedish phonesthemes. Initial and 

final consonant clusters were primarily studied, but vowels were also 

included. 

Other important issues have been: studying productivity of new sound 

symbolic words and similarities between languages (connected with the 

issue of universals), and finding a reasonable explanatory model for (a 

part of) sound symbolism in Swedish. 

The role of sound symbolism in language was discussed and a model for 

the position of sound symbolism in language was constructed, taking a 

number of factors into account, such as innateness, learning, productivity, 

context and conventionalization. (cf. 1.9). 

The sound symbolic properties of consonant clusters and vowels were 

described – the expressions and the meanings. The meanings of 

phonesthemes were found to be relatable to each other and different 

explanations for the relationship between sound and meaning in sound 

symbolism were also discussed, especially in relation to indexicality and 

iconicity. A semantic model for explaining the different semantic features 

of sound symbolism was constructed. 

With a point of departure in the description of Swedish phonesthemes a 

number of studies were made in order to investigate universality and, 

above all, the role of productivity. 

8.1 The research questions were as follows: 

1. What are the properties of sound symbolic sounds and sound sequences 

in Swedish? More specifically the questions are: 

Which initial and final consonant clusters are used in sound 

symbolism? 

245

Which meanings are used in sound symbolism? 

How do these combine in phonesthemes? 

What are the sound symbolic characteristics of some vowels? 

How do initial and final clusters and vowels combine in words? 

2. Are phonesthemes productive in Swedish? And, if so, are some 

phonesthemes more productive than others? Are neologisms created or 

interpreted in accordance with the semantic model of chapter 2 and the 

analysis of chapters 4 and 5? 

3. Are there similarities or dissimilarities between different languages in 

some aspects of sound symbolism? 

4. Do non-Swedish speakers interpret Swedish phonesthemes in 

accordance with the semantic model of chapter 2 and the analysis of 

chapters 4 and 5? 

The results pertaining to these questions are summarized and discussed 

below: 

8.1.1 Question 1 

What are the properties of sound symbolic sequences in Swedish? 

The lexical study of initial consonant clusters hints at about 1,000 roots 

with sound symbolic beginnings of the 65,000 lexemes (or 8,300 root 

morphemes 1) in the Swedish vocabulary. The initial consonant clusters 

vary in the degree to which they are sound symbolic and the same is true 

for the final clusters and the vowels. Almost all initial consonant clusters 

and about 22% of the final clusters are used for sound symbolism. (From 

the experiments in chapter 7 it seems that the final consonant clusters are 

of less importance than the initial ones in producing new words, see 

below.) 

1 This number of morphemes is the result of an investigation of a one million word 

newspaper corpus (NFO 4). It is most probable that newspaper language contains 

more roots than spoken language, but these figures are not known at present. The 

estimate of 65, 000 lexemes excludes transparent compounds. 

246

A restricted number of meanings that are semantically relatable to each 

other are used in Swedish phonesthemes. These are in most cases 

connected with perception or cognition (cf. the models in 1.9 and 2.6.2). 

A consonant cluster usually has more than one possible sound symbolic 

meaning, and the semantic profiles vary for different clusters. 

The most frequent semantic features for initial and final clusters are 

partly the same and partly different. For the initial clusters, the 10 most 

common features are, in descending order: 

Table 8.1 The 10 most common 

semantic features of initial clusters. 

'pejorative' 

'sound' 

'long thin form' 

'quick or strong movement' 

'wetness' 

'talking' 

'light' 

'diminutive' 

'round form' 

'walking' 

For the final clusters, the 10 most common semantic features of Svensk 

Baklängesordbok and the 6 most common features of Nusvensk 

Frekvensordbok 4 (NFO 4) are shown in table 8.2. 

When the most common semantic features of final clusters of Svensk 

Baklängesordbok and Nusvensk Frekvensordbok 4 are combined, the 

following features are most frequent, in descending order: 'sound' and 

'pejorative', 'talking', 'quick or strong movement', 'long thin 

form', 'wetness' and 'slang' ,'walking', 'round form', 'light'. 

247

The first six features are also the six most common features of initial 

clusters, cf. table 8.1 (cf. also table 5.4). 

Table 8.2 The 10 most common semantic features of final clusters of 

Svensk Baklängesordbok and the six most common features of the 

Nusvensk Frekvensordbok 4. 

Svensk Baklängesordbok Nusvensk Frekvensordbok 4 2 

'quick or strong movement' 'pejorative' 

'talking' 'sound' 

'sound' 'long thin form' 

'pejorative' 'talking' 

'slang' 'wetness' 

'walking' 'quick or strong movement' 

'wetness' 

'long thin form' 

'round form' 

'light' 

This can be compared with the features that were the most successfully 

coded and interpreted in the experiments of chapter 7. In the experiments 

of free production tests from meaning to constructed word, 'pejorative', 

'bad mood', and 'wetness' were the most successful. 

Of the six features listed above 'pejorative and 'wetness' were the most 

successful in the experiments of free production tests from meaning to 

constructed word. 

In the forced choice tests the most successful semantic features were 

'pejorative' and 'dryness' ('sound', 'diminutive', 'round form' and 

'walking' were not tested in the forced choice experiments, see 7.1). 

2 There are only six semantic features in this column since they are the result of the 

analysis presented in table 5.18, which showed the most common clusters and their 

semantic features; other semantic features are therefore not very frequent in Nusvensk 

Frekvensordbok 4. 

248

Of the six features listed above, 'pejorative' was the most successful in the 

forced choice tests. 

The phonological characteristics of the most common sound symbolic 

clusters, initially and finally and in particular for certain meanings, were 

as follows: 

Initial clusters 

For initial clusters, the most common ones (cf. table 4.8) are sl- 

(absolutely) and pj- (percentally). For the five most frequent meanings, 

in absolute numbers, there is a strong tendency for the consonant clusters 

to end with l. For the five most frequent meanings, percentally, there is a 

slight tendency that the consonant clusters end with j. The clusters ending 

with j are lexically low frequent ones and are thus marked. This can 

make them more useful for sound symbolic functions, cf. 8.3. 

Considering specific semantic features, the results for the initial clusters 

are: all two-consonant clusters containing a j-phoneme are 'pejorative' 

and they are so to quite a great extent. Looking at absolute figures, 

clusters beginning with s, especially sl-, sn-, skr- are dominant among 

the pejorative roots. The semantic feature 'bad mood' is mainly coded 

with the clusters gr-, vr- and tr- (but these clusters can have many other 

meanings). Furthermore 'light' is dominated by voiced plosives + l or n – 

bl-, gl-, gn- (dl-, bn-, dn- are not allowed in Swedish) – and 'wetness' is 

dominated by sl-, kl-, sp-, sm-, skv-, spr-, sn-, bl-, pl-, sv-, i.e. s or 

initial unvoiced plosives are preferred (the only exception is bl-). The 

feature 'long thin form' is dominated by three three-consonant clusters: 

spj-, str-, spr-. These all begin with s (a fact of all Swedish initial threeconsonant 

clusters). In addition, with the exception of two clusters 

connected with the feature 'long thin form', the above mentioned clusters 

plus sp-, sl-, sn-, st-, sk- and sv-, also begin with an s. The only ones 

that do not begin with s are tr- and tv-. These are however, voiceless 

dentals like s. 

249

Most clusters that are percentally most frequent are lexically very 

infrequent ones, cf. discussion in 8.3. 

The results of the lexical analysis of chapter 4 concerning the pj- 

'pejorative' cluster, the gr-, tr-, vr- 'bad mood', the skr- 'destruction', 

the skv- 'wetness' and the fn- 'dryness' clusters correlate with the results 

of the free production or free interpretation experiments of chapter 7 (cf. 

question 2 below). 

Final clusters 

As discussed in the beginning of chapter 5, it is not self-evident which are 

the final clusters in Swedish. Morphological structure has to be taken into 

account since some sequences only occur in inflected or derived forms 

(e.g. -ndsk in bondsk) and some sequences, which are obviously 

interesting, cannot occur in the absolutely final position (e. g. -dr- as in 

fladdra. The meaning profiles for the sound symbolic final clusters differ, 

as is also the case for initial clusters, cf. diagrams 5.3–5.11. 

For the final clusters, the main results are as follows: The most common 

clusters with sound symbolic meaning, in order of frequency – 

independent of meaning - are the following (from NFO4): -Nk, -sk, -fs, 

-nd, -tr (absolutely) or -fs, -dr, -lr, -ml, -Nl (percentally) 3. Of these, 5 

of 9 end with a liquid. Of the 11 percentally most sound symbolic final 

consonant clusters, 8 end with a liquid and 3 with a voiceless obstruent (s 

or k). Of the 13 most common final consonant clusters in absolute 

numbers, 6 end with a liquid and 7 with a voiceless obstruent (s, t or k). 

The high frequency of liquids and voiceless obstruents among the final 

clusters could be due to lexical dominance of these clusters. (Swedish final 

clusters can also end in voiced obstruents and nasals.) It is beyond the 

scope of this thesis to investigate this issue. The high sound symbolic 

frequency of liquids and voiceless obstruents could also have to do with 

the fact that they are especially suited for certain meanings. As seen in 

table 5.3, however, none of the two groups are especially tied to any 

3 The five most frequent final clusters have a percentage of 100% because there is only 

one – sound symbolic – root, for every cluster. These are not considered here, cf. 

however, diagram 5.2. 

250

semantic feature. In the NFO 4 material, -fs (pejorative) is the most 

common phonestheme both absolutely and percentally (cf. tables 5.18 and 

5.19). 

It is not the case, for final clusters, that the high frequent semantic 

features seem to prefer certain sounds or sound combinations, cf. table 

5.3. The more low frequent semantic features, however, use certain 

consonant clusters: 'wetness' -ska; 'long thin form' -msa; 'round form' - 

lla; 'light' -indra, -imra; 'shortwide form' -mpa. These clusters, except 

for -indra, -imra ('light') are, as can be seen in table 5.3, used by other 

semantic features as well. 

Vowels 

Vowels, a selection of which were studied in minimal pairs, seem to have 

other semantic characteristics than consonant clusters, often modifying the 

meaning of consonant clusters, e.g. being diminutive or augmentative. [i] 

tends to have the meanings 'smallness', 'quickness', 'high pitch', while [P] 

seems to have the meanings of 'low pitch' and 'largeness'. [a] seems 

neutral. The vowel [P] is also connected with 'pejorative'. The semantic 

feature 'light' is connected with front vowels, especially with [i]. 

Combinations 

The study of combinations of initial and final consonant clusters showed 

that it is more common for semantic features to be the same initially and 

finally, than to be different. If new semantic features had been searched 

for in the investigation of final clusters, there is a possibility that this 

result would have to be modified, because of the appearance of new roots. 

The semantic features under consideration are however, mostly the same 

initially and finally. 

It seems then, that initial and final clusters strengthen each other rather 

than contributing to different sound symbolic meanings in the same root. 

However, for those consonant clusters that can occur both initially and 

finally it is not the case that they are combined in the same word, e.g. bland 

-bl, but neologisms as blabbla are conceivable. 

251


Are phonesthemes productive in Swedish? And, if so, are some 

phonesthemes more productive than others? Are neologisms created and 

interpreted in accordance with the semantic model of chapters 1 and 2 and 

the analysis of chapters 4 and 5? The model of 1.9 says, among other 

things, that sound symbolism is productive. Furthermore, the model of 

2.6.2 claims that semantic features of sound symbolism are, due to the 

innateness of categories of thinking - such as predispositions for seeing 

contiguity and similarity, and due to learning - relatable to sense 

impressions and emotions, restricted to certain types. 

The characteristics of the consonant clusters (cf. above, question 1) are 

mirrored in the experiments of chapter 7 in the following way: 

Forced choice tests 

The phonesthemes, in the constructed words, that were the most 

successfully coded, i.e. given a phonological form based on a presented 

semantic feature, were 'pejorative': fj-, 'dryness': fn-, long thin form': 

str-, 'wetness': sp-, 'bad mood': vr-, 'pejorative': pj-. The features most 

successfully coded totally were 'wetness', 'hollow form', 'separation' and 

'dryness'. The six phonesthemes most successfully interpreted were 

'hollow form': gr-, 'dryness': fn-, 'talking': sn-, 'pejorative': pj-, 'quick 

or strong movement': sk- and 'separation': spr-. For both the 

interpretation and production tests, the most successful phonesthemes 

were 'dryness': fn- and 'pejorative': pj-. Both of these are (lexically low 

frequent) clusters that are sound symbolic to a very high degree. A 

possible explanation for the success of these phonesthemes is that they can 

be more easily accessed. Phonesthemes could be stored in a way different 

from other morphemes since there is a motivated connection between 

sound and meaning. This could make them more reliable and more 

frequent in production and interpretation. 

The 'light' words stand out because of many instances of bad results, both 

in the word-to-meaning and meaning-to-word tests. It is clear that the 

'light' category gets the least number of expected answers. The 'light' 

words might belong to a closed lexical class which would imply that gl-, 

252

gn-, bl-, are not productive, unlike phonesthemes such as kl- or pj-. 

Even when the 'light' words had a vowel with high F2, as many of the 

'light' words have according to 5.4, the results of the experiment were no 

better. The semantic feature 'light' is positioned quite high in the analysis 

of the most common features (cf. diagram 4.5). 

Free production tests 

Phonesthemes 

In the free production test from expression to meaning, the percentally 

most common phonestheme pj- (pejorative) was the most accurately 

identified. In the free production test from meaning to expression the jclusters, 

among others, for 'pejorative' were produced as expected. The 

r-clusters for 'bad mood' were also produced as expected, plus some 

additional r-clusters, both initial and final. For the feature 'wetness', 

there was a majority of s-, l- and p-clusters (as expected), both initially 

and finally. 

Explanations of these results, for the j-clusters, could have to do with the 

facts that these are (lexically low frequent) clusters that are sound 

symbolic to a very high degree; they could thus be more easily accessed 

and this could, in turn, make them more reliable – and frequent – in 

production and interpretation. As concerns the r-clusters for 'bad mood' 

they can have their base in expressions which are spontaneous 

vocalizations in connection with an angry feeling, and the s-, l- and pclusters 

for 'wetness' have their base in sound imitation. The question of 

whether one of these bases gives better results than the other in 

interpretation and production cannot be answered from these experiments 

alone; further experimentation is needed. 

The phonestheme most difficult to interpret was 'winding form': kr-. 

Semantic features 

The semantic features of the free production meaning to word 

experiments that were most successfully expressed (in accordance with the 

lexical analysis) were 'pejorative' (163), 'bad mood' (18) and 'wetness' 

(63). Less successfully expressed were 'winding form' (20), 'dryness' (1) 

253

and 'rough surface' (10) (The absolute frequencies of these semantic 

features, for the initial clusters, are shown within the parenthesis, cf. table 

4.2.) These results can be compared with the most frequent semantic 

features according to the lexical analysis summarized in tables 8.1 and 

8.2. Of the three most successful features – 'pejorative', 'bad mood' and 

'wetness' – of the free production experiments, 'pejorative' and 'wetness' 

are among the six most common features according to the lexical analysis. 

A probable explanation for these similarities is that phonesthemes of the 

most frequent semantic features are stored in such a way that they are 

more accurately accessed by the language user. The way they are stored is 

dependent on stable intermodal connections, cf. the discussion in 1.11.2. 

Of the three less successful features of the free production experiments - 

'winding form', 'dryness' and 'rough surface' - none belonged to the 10 

most common semantic features of the lexical analysis. A probable 

explanation for this is that the phonesthemes of the less frequent semantic 

features are stored in such a way that they are less accurately accessed. 

The conclusion is that phonesthemes are – more or less – productive, both 

in production of new forms and understanding of neologisms. There is a 

tendency for the most common semantic features to be more successfully 

coded, in accordance with the main results of the analysis of 

phonesthemes in chapter 4; these might be more accessible. It could also 

be the case that categories related to negative emotions (e.g. 'pejorative', 

'bad mood') are more important to humans (at least Swedes) than more 

abstract categories like 'form'. It also seems that subjects tend to encode 

semantic features in the initial clusters rather than in the final ones in free 

production, and it seems that for low frequent features (like 'dryness'), 

with few possible clusters, there are a larger number of unconventional 

forms produced. These tendencies need to be investigated further. 

An interesting result from the free choice test from expression to 

meaning is that the meanings produced all belong to the classes found in 

the lexical study. Even the informants who started the test by freely 

suggesting meanings based on constructed neologisms (and therefore had 

no expectancies as to what the answers ought to be) produced meanings 

254

within these classes (although not always within the expected class for a 

certain nonsense word - the reason for this might be individual contextual 

influences at the moment of the test). These classes are: 'pejorative' (often 

'destruction'), 'mental feeling' (often irritated), 'surface structure', 

'wetness', 'form', 'consistency', 'movement', 'diminutive', 'sound'. These 

results can also be compared with the most frequent semantic features 

according to the lexical analysis summarized in table 8.1. Six of the nine 

general semantic features resulting from the free choice test from 

expression to meaning ('pejorative', 'wetness', 'form', 'movement', 

'diminutive' and 'sound') are among (or superordinate to) the nine most 

common semantic features of table 8.1. This supports the model of 2.6.2 

where the phonesthemes for these categories are seen as a result of 

innateness, learning and conventionalization; the semantic categories of 

phonesthemes are predictable rather than haphazard. This model shows 

that in many cases the semantic features of phonesthemes are potentially 

relatable to neurological connections between the senses, cf. 8.3.2. It 

could also imply that phonesthemes that concern the most frequent 

semantic features are stored in such a way that they are most readily and 

accurately accessed 4. 


Are there similarities or dissimilarities between different languages in 

various aspects of sound symbolism,? 

In the Thesaurus study of the concepts 'stupidity', 'rough surface 

structure', and 'smooth surface structure', for English and Swedish, the 

following was found: The phonological agreement between words 

belonging to these semantic fields in English and Swedish was greater 

among the words for 'surface structure' than for the words for 

'pejoratives'. One obvious reason is that some of the clusters used in 

Swedish (fj-, fn-, pj-) are not allowed in English. Another reason could 

be that 'surface structure' is closer to a potentially common phenomenon, 

namely sound imitation, since stroking different surfaces give different 

4 This is a question for further experiments, e.g. lexical decision experiments, and 

development of on line models. 

255

sound effects. Of course, Swedish and English have many cognates but not 

in these results (cf. 6.2), except for some cases. 

In the tests with informants concerning different interjections in 8 

languages the following similarities and dissimilarities were found 

concerning expressions: The pejorative interjections often contain an u or 

an O, the positive interjections an a, 'surprise' often an a or an O. 'Pain' 

(e.g. Swedish aj) has a diphtongized open vowel, 'sneeze' (e.g. Swedish 

atjo) has an affricate and the interjection for go away (to an animal) (e.g. 

Swedish schas) has a fricative in all the 8 languages. 

In the tests with informants of 16 different languages concerning different 

expressions that are imitative of animal noises, the main results are the 

following: No animal call imitation is exactly the same in all the 16 

languages. Some animal call imitations are more alike, e.g. the cat's 

meow, while others vary more, e.g. the dog's barking. One reason for 

this could be that some animal cries are more complicated than others. 

There are, however, always similarities on the level of phonetic features 

between the expressions for the same animal call in the different 

languages. Within each language there are, as well as for the expressive 

interjections, similarities which can be assumed to depend on the 

phonological structure of the language in question. 

In the identification test, expressions for different animal sounds by 

speakers of 9 different languages were tested on 15 persons with 6 

different first languages. The main results are that some animals were 

identified by all listeners (e.g. dog, cow and cat), some less correctly 

identified and one animal not identified at all (e.g. duck, Hungarian 

expression). The interpretation test shows that, given the context of 

animal calls, it was quite easy for listeners to interpret animal calls from 

languages other than their own (cf. table 6.6). 

Thus, for these contrastive studies, the following general conclusion can 

be drawn. There are both similarities and dissimilarities between the 

256

expressions in the different languages 5. This is true for most phenomena 

in language, and sound symbolism is no exception here. The variation is 

greater for some semantic fields than for others; expressions for rough or 

smooth surface structure are more alike than expressions for pejoratives. 

A possible explanation of this is that 'surface structure' is closer to a 

potentially common phenomenon, namely sound imitation, either because 

stroking different surfaces give different sound effects or because of the 

auditory-tactile neural connection. Certain animals are imitated more 

consistently than others, possibly because certain animal cries are shorter 

and less complicated than others, cf. e.g. the cries of the cat and the 

rooster. 


Do non-Swedish speakers interpret Swedish phonesthemes in accordance 

with the semantic model of chapter 2 and the analysis of chapters 4 and 5? 

In the test of understanding Swedish sound symbolic words (and some 

neologisms constructed from phonesthemes) the general results are as 

follows: The subjects seldom interpreted the words correctly. However, 

the answers given almost all belong to categories within the semantic 

model of 2.6.2 (cf. table 6.7). It therefore seems that for certain 

categories, often related to sense impressions or emotions, there is a 

preference for sound symbolic expressions; it is more natural for 

speakers to imagine that an unknown word, where the expression is 

supposed to reflect the meaning, means e.g. 'quick or strong movement' 

than that it means e.g. 'my country'. 

The conclusion from the studies under questions 3 and 4 is that there are 

greater differences between these languages 6 for expression than for 

meaning. The expressions imitating animal sounds are easier to interpret 

than other sound symbolism for speakers of different languages. 

5 For crosslinguistic studies in other semantic fields, see e.g. Viberg (1999). 

6 Swedish, English, Icelandic, Polish, Hungarian, Finnish, Ososo, Malagasi, 

Slovenian, Korean, Japanese, Chinese, Estonian, Urdu, Persian, Kurdish, Arabic and 

Spanish. Persons who interpreted Swedish sound symbolic words spoke Arabic, 

Spanish, German, Dutch, Ibo, English. 

257

Interpretation of other sound symbolism often goes wrong (because 

expressions probably differ in different languages), but the semantic 

categories guessed at are most often within the semantic models of this 

thesis. 

8.2. Comparison of the studies 

The next table (8.3) is a comparison between the most frequent 

phonesthemes of the lexical studies with the most accurately interpreted 

and coded phonesthemes of the experimental studies. The most common 

clusters and the most common meanings are also shown. The data are 

from different tables and diagrams as indicated in table 8.3. 

Table 8.3 The most frequent 7 phonesthemes or the most successful 8 

phonesthemes (their consonant clusters and meanings), whichever is most 

appropriate, from the different studies. The examples are presented in 

frequency order. In some cases, however, consecutive examples might 

have the same frequency. 9 

7 According to the lexical analysis. 

8 In the case of the experiments where subjects produced or interpreted neologisms. 

9 Squares are empty for the experiments where there are no relevant data. 

258

lexically initial 

consonant 

clusters 

experimentally 

final consonant 

clusters 

forced choice 14 


expression 

phonesthemes consonant 

clusters 

pejorative: 

sl-, pjsound: 

kl-, kn-, fnlong 

thin form: 

sp-, spjquick 

or strong 

movement: 

flwetness: 

sl-, skv-10 quick or strong 

movement: -NNl, 

talking: -tr, 

sound: -st, 

pejorative: -fs, 

wetness: -sk, 


-Nk11 fj-: pejorative 

fn-: dryness 

str-: long thin 

form 

sp-: wetness 

vr-:bad mood 

pj- 

:pejorative15 259 

in absolute 

numbers: 

sl-, sn-, kn-, 

kr-, kl-, sp-, 

gl-, st-, tr- 

(cf. diagram 

4.1) 

in %: 

fn-, gn-, skv- 

, pj-, kn-, 

spr-, spj-, gl- 

, mj-, vr- (cf. 

diagram 4.4) 

-Nk, -sk,-fs, - 

nd, -tr (cf. 

diagram 5.1 

and table 5.21) 

-dr, -lr, -ml, 

-NNl 12 

10From table 4.8. 

11 From tables 5.3 and 5.18. 

12 From NFO4, diagram 5.2 (except for the 1 root clusters). 

13 From NFO4 and Svensk Baklängesordbok, section 5.5.3. 

meanings 

pejorative, 

sound, long 

thin form, 


movement, 

wetness, 

talking, light, 

diminutive, 

round form, 

walking (cf. 

diagram 4.5) 

sound, 

pejorative, 

talking, 


movement, 

long thin form, 

wetness13 wetness, 

hollow form 

separation 

dryness, 

pejorative (cf. 

table 7.3)

cross 

cultural 

interpretation 

forced choice 

expression to 

meaning16 free production 

expression to 


free production 


expression 

gr-: hollow 

form 

fn-: dryness 

sn-: talking 

pj-: pejorative 

sk-: quick or 

strong 

movement 

spr-: 

separation17 pj-: pejorative 

skr-: 

destruction 

skv-: 

wetness18 pejorative: fl-, 

fjbad 

mood: rclusters 

wetness: sl-, 

sv-19 fl20-: quick or 

strong 

movement 

skr- : 

destruction 

tr-: bad mood 

260 


movement21 14 The semantic feature 'sound' was not tested. 

15 From table 7.2. 

16 For both the production and interpretation experiments added together, the most 

successful phonesthemes were fn-: 'dryness' and pj-: 'pejorative' 




20Phonesthemes of words that were most successfully interpreted according to the 

Swedish norm. Cf. table 6.8. 

21 Most preferred meaning for unconventional but possible meanings. 'Pejorative' and 

'sound' were NOT among the more preferred meanings. Cf. table 6.9.

Table 8.3 shows, among other things, the following relations between 

different studies: The most successful initial phonesthemes of the 

experimental studies that correspond to initial phonesthemes of the most 

frequent semantic features are: pj-: pejorative (initial consonant clusters, 

forced choice meaning to expression, forced choice expression to 

meaning, free production expression to meaning); skv-:wetness (initials, 

free production expression to meaning); sl-:wetness (initials, free 

production meaning to expression) in this most condensed version of the 

results. 

Sl- is the percentally largest cluster for 'wetness' (and THE most sound 

symbolic cluster – for all semantic features – in absolute figures), pj- is 

the percentally largest 'pejorative' cluster and skv- is the percentally fifth 

largest wetness cluster 22. Pj- and skv- are both small but percentally 

highly sound symbolic clusters. 

An explanation for these results is that phonesthemes that are 

proportionally large (i. e. where a certain meaning has a proportionally 

large part of the (sound symbolic) roots of a cluster) are stored in such a 

way that they are more accurately accessed by the language user. 

However, all phonesthemes of the free production experimental results 

above match the lexical analysis. A possible explanation for most of these 

similarities could be that the phonesthemes of semantic features are 

relatable to neurological connections between the senses (in accordance 

with the model of 2.6.2). 

The most successful phonesthemes (which were in fact not very 

successful, cf. table 6.8) of the cross cultural interpretation experiment do 

not reflect the frequent Swedish semantic feature 'pejorative'. An 

explanation for the lack of success of this feature could be that the 

negative emotion that 'pejorative' is based on is a cultural trait of 

Swedish, but not of the other languages investigated here. However, of the 

phonesthemes that did succeed, one of the clusters, fl-, is the 

22 Cf. also table 4.8. 

261

proportionally most sound symbolic cluster for 'quick or strong 

movement', and the other two clusters skr- and tr- are the proportionally 

the most sound symbolic clusters for the semantic features 'destruction' 

and 'bad mood' respectively. These results are then in accordance with 

Swedish competence. One obvious explanation for the overall bad results 

is that non-native Swedes do not connect accurate expressions with the 

semantic features as a natural consequence of different phonological 

learning environments. 23 

The vowels were not taken into consideration in this table, since they 

were studied in a different way, which involved contrasting some 

meanings and sounds in minimal pairs. 

The cross-linguistic comparisons concerning interjections and the 

expressions for animal cries are not comparable to the data in table 8.3, 

because they concern semantic fields other than the Swedish lexical study 

and the experiments. The Thesaurus study, comparing English and 

Swedish, showed that, for the concepts 'stupidity', 'rough surface 

structure' and 'smooth surface structure', similar phonemes were used for 

the last two concepts, but not for the first. A possible explanation for this 

is that 'surface structure' is closer to a potentially simultaneous 

phenomenon, namely sound imitation, since stroking different surfaces 

give different sound effects. The comparison of different languages in the 

studies of interjections showed certain agreements in vowels. Most of the 

interjections consist of vowels and single consonants and are therefore not 

comparable with sound symbolism of initial or final consonant clusters. 

For the animal sounds, there were great expressive similarities between 

the different languages. It was quite easy for listeners of different first 

languages to correctly identify animals from the way in which their cries 

were represented in different languages, cf. summary in 8.1.3. 

Comparing these last two studies with the Thesaurus study (only involving 

the related languages Swedish and English but showing great differences), 

23 However, as can be seen in table 6.9, subjects mostly guessed within the expected 

semantic categories of the model. A potential explanation for these preferences has to 

do with neurological connections between the senses, i. e. the parts of the model in 

2.6.2 connected with innate predispositions. 

262

the conclusion is that the expressions where there is a more direct (mainly 

indexical or iconic) connection between expression and meaning are more 

alike in different languages. 

8.3 Possible explanations of onomatopoeia and 

sound symbolism 

Why is it that certain consonant clusters are connected with certain 

meanings? And why is it that we want certain meanings to be expressed 

sound symbolically, as well as with conventional morphemes; what is the 

function of sound symbolic morphemes as compared with full 

morphemes? 

The first question is easy to answer when it concerns sound imitation. The 

expressive side of words are articulated sounds and articulation can 

imitate sounds, in a more or less conventionalized way. One possible 

explanation to other sound symbolism is that phonesthemes were 

originally onomatopoeic and later developed into metaphorically (e.g. 

gn-: 'sound' and later also 'light') or metonymically (e.g. skv-: 'sound' 

and later also 'wetness') related meanings (cf. 2.6.1). Underlying this are 

the innate capacities for metaphor and metonymy. This is in accordance 

with the theories of e.g. Herder (cf. 1.6.) about the origin of language. 

This is not an impossible view since almost all initial consonant clusters 

have the feature 'sound' or 'talking' (which later feature is a sub-category 

of 'sound', cf. 4.2.3) to a greater or lesser extent. The only exceptions are 

pj-, spj-, spl- and spr-. 

There are some linguistic facts about Swedish and analyses of several 

languages by various linguists that show along what lines other sound 

symbolism in Swedish can partly be explained. A tendency is that lexically 

infrequent clusters (i.e. marked clusters) are exploited sound symbolically 

to a higher degree than lexically frequent clusters. It also seems as if a 

large proportion of the three-consonant s-clusters are used for sound 

symbolism in Swedish, cf. diagram 4.4. This is in accordance with the 

discussion of Hinton, Nichols, Ohala (1994) about marked sounds being 

used for sound symbolism. Also, sounds and sound combinations 

263

otherwise non-existent in the language (wild forms) occur in 

onomatopoeic and sound symbolic neologisms. 

Another aspect of markedness is that sounds that are new in a language 

are often used sound symbolically (cf. Austerlitz, 1994, on ø in Finnish). 

In Swedish, the latest great consonant change was the collapse of several 

consonant clusters (e. g. skj-, stj-, sj-, kj-, tj-) into the fricatives S and 

ç. A possible explanation is that the clusters containing j that were left 

over – spj-, bj-, fj-, pj- – became more unusual and marked in Swedish 

and therefore useful for sound symbolism. 

Classifying the different meanings of sound symbolism, it is clear that 

most of them fall under the perceptual category, but also under cognitive 

factors, viz. evaluation, but not under a social category. This hints at 

sound symbolism being in some way biologically grounded, rather than 

learned. 

Most of the meanings of the phonesthemes are relatable to the senses 

(probably so because of metaphorical or metonymical extensions from 

sound and/or neurological connections between the senses), apart from the 

'pejoratives' and 'mental feeling'. Thus the phonesthemes are related to 

hearing, seeing, touching, (but not to smelling, tasting) 24. The semantic 

features occurring over and over again are relatable to stable intermodal 

connections, i.e. one sense is connected to another. There are, however, 

many problems with an explanation in terms of synaesthesia (see 1.11.2). 

Neurological aspects relevant for a model (see chapter 2) are neurological 

connections between the senses, or a common ground for the senses. 

Modalities that have especially strong connections (according to 

Geschwind, 1965, according to Cytowic, 1989) are visual-auditory and 

tactile-auditory. 

Semantic features having to do with movements can ontogenetically be 

explained in terms of metonymy, i.e., simultaneity in time and space 

between sound and movement, and metaphor. In some cases 

24 There also exist, on a lexical level, meaning extensions from sensory modalities to 

mental phenomena (see Abelin, 1988). 

264

proprioception might be involved, as in an explanation of pejoratives 

where a feeling of disgust is experienced simultaneously with contractions 

of certain facial muscles, which contractions have an effect on articulation 

(cf. Darwin, "1965"). 

Ohala's (e.g. 1994) frequency code offers a solution for sound symbolism 

of diminutives and augmentatives. The common factor is high-low 

frequency for F0 (of intonation), noise or F2. High frequency is 

connected to smallness, low frequency to largeness. 

The second question – Why is it that we want certain meanings to be 

expressed sound symbolically, as well as with conventional morphemes? – 

can be answered with appeal to a kind of redundancy in the linguistic 

sign. In language, there is partly a "triple articulation". The "third 

articulation" is meaning bearing and motivated (cf. 1.2.5), and this makes 

sound symbolic words very effective. There are, in sound symbolic 

words, other ties between expression and meaning than merely the 

arbitrary conventional ones, namely those that are motivated but still to 

some extent conventional. We can still wonder if sound symbolism is a 

remnant from earlier stages, where it could have had a high survival 

value because of its connection with the senses and with things present at 

the moment something is spoken. An argument for this is, for example, 

Ohala's frequency code. Also, the connection between emotions and facial 

contractions possibly resulting in certain sounds for pejoratives could also 

be a remnant from earlier stages. On the other hand, it is possible, as 

Jespersen claimed, that languages grow richer and richer in sound 

symbolic words. 

8.3.1 Pejoratives 

The pejorative phonesthemes resist a simple explanation. Is there an 

ugliness code? An ugliness metaphor? Or do they have to do with a basic 

distinction between approval and disapproval? One of the 6 basic 

emotions, according to e.g. Ekman (1973) is disgust, which is an emotion 

underlying pejoratives. 

265

Darwin (1872) gives a possible explanation for sound symbolic words 

related to emotions, e.g. disgust (related to pejoratives) and interjections. 

The explanation builds on the instinctive contractions of facial muscles 

connected with a certain emotion. Pejoratives could be of a more 

indexical nature, the result of interjections which in turn could be 

conditioned by instinctive facial contractions. A similar type of word is 

one that could be termed truly iconic. These are words like pluta, truta 

and pussa, perhaps grina and a few more, where the pronunciation of the 

vowel in particular can be seen as being dependent on the shape of the 

face and sometimes connected to emotions. 

Hamano (1994) proposes an explanation in child directed speech. He 

studied palatalization in Japanese sound symbolism. Palatalization of 

alveolar stops and fricatives is associated with "childishness" and 

"immaturity". He connects this to studies on language acquisition 

reporting palatalization as one of the universal characteristics of early 

stages of children’s language acquisition. Palatalization is also reported as 

one of the commonest devices of baby-talk, i.e., child directed speech 

(Snow and Ferguson, 1977). Perhaps this is a way of explaining the 

Swedish pejorative clusters bj-, fj-, pj-, which do not fit into a 

synaesthetic and metaphoric pattern. 

The high frequency of pejoratives in Swedish phonesthemes may well be a 

cultural peculiarity, since it was not mirrored in the cross cultural 

interpretation experiment. The pejorative could be based on a general 

negative emotion, which could be favored in a certain social environment. 

Naturally, further cross linguistic research is needed before any 

conclusions can be drawn with respect to this. 

8.3.2 Summary 

To summarize: 

1) The unique feature of phonesthemes in relation to ordinary morphemes 

is that they are bound morphemes that are both meaning bearing and 

motivated (cf. 1.2.5). Words with phonesthemes are very effective since 

there are ties between expression and meaning other than merely the 

266

arbitrary conventional, namely the motivated and to some extent the 

conventional. 

2a) Certain meanings are suitable to be conveyed symbolically in sound 

e.g. meanings related to the senses because of the interaction between, on 

the one hand, innate capacities based on neurological connections between 

the senses and, on the other hand, learning in the present. The 

neurological connections which (according to Geschwind, 1965, 

according to Cytowic, 1989) are especially strong are the connections 

visually-auditorily and tactile-auditorily. They make it possible for sound 

imitative expressions and phonesthemes to lead to other sound symbolic 

connections between these sound sequences and e.g. forms and movements 

(which have to do with visual and/or tactile experience), and surface 

structures and consistency (which have to do with tactile experience). The 

learning process, of which a part is perceiving sounds and mimicking 

them, in combination with conventionalization, causes the sound symbolic 

expressions to vary (slightly) in different languages. 

2b) The connection between the expression and meaning of pejoratives is 

more difficult to relate to neurological connections between the senses and 

can be given various explanations, connected with e. g. markedness, child 

language or proprioception. The emotion of disgust might, however, be 

behind the semantic feature 'pejorative'. 

3) The result that final clusters predominantly end with a liquid or a 

voiceless obstruent could perhaps be explained by the auditory salience in 

the case of several voiceless obstruents; what is heard well is more useable 

in general. It could be that such sounds are chosen for sound symbolism 

since they could then fulfill functions other than merely the distinctive. It 

is more difficult to give an explanation of the predominance of liquids 

since this is a group based on both acoustic similarities and common 

phonological patterning. 

The usefulness of sound symbolism in general is connected with the 

strong tie between meaning and expression, discussed in 3.3. It is 

imaginable that peoples' reactions are quicker, stronger and more 

267

accurate to sound symbolic expressions, and therefore they can be e.g. 

more effective, more emotionally arousing, and more poetic. 

8.4 Predictions for sound symbolism in Swedish 

When a new non-arbitrary word is created or understood it is most likely 

to have the following characteristics: 

1. Its semantic content belongs to those described in chapter 4 (except for 

'light') and 5 25. 

Sound: Sound 

Talking 

Beat 

Movement: Movement 

Quick or strong movement 

Walking 

Falling 

Potential movement 

Quickness 

Light: Light 

Gaze 

Surface structure: Rough surface structure 

Smooth surface structure 

Consistency: Soft consistency 

(Plasticity) Hardness 

Slackness 

Stiffness 

Wetness: Wetness 

Adhesion 

25 The order of presentation follows the order in 2.6 and 4.2.3, which is based on the 

preliminary lexical frequencies of this study. After the lexical study of initial clusters 

was completed the order of the semantic features was somewhat changed, but the 

original order was kept in some lists. It therefore does not represent an analytical order. 

For the order of frequency of the completed lexical study of initial clusters, see diagram 

4.5. 

268

Dryness: Dryness 

Attitude: Attitude 

Secrecy 

Slang: Slang 

Jocular: Jocular 

Pejorative: Pejorative 

Destruction 

Mental feeling: Mental feeling 

Bad mood 

Bodily feeling: Bodily feeling 

Suffocation 

Separation: Separation 

Putting together: Putting together 

Diminutive: Diminutive 

Augmentative: Augmentative 

Form: Form 

Round form 

Short-wide form 

Thin form 

Hollow form 

Winding form 

Long thin form 

Narrow form 

Small end form 

Bent form 

Iterative: Iterative 

Fine grain 

2. Its initial and final consonant cluster is one that is appropriate for its 

semantic content, see chapters 4 (except for 'light') and 5. The most sound 

symbolic initial clusters in absolute numbers and percentally, for the five 

most frequent meanings, are shown in table 8.4. 

Table 8.4 The most sound symbolic initial clusters in absolute numbers 

and percentally, for the five most frequent meanings. 

269

meaning in freq. % in % % freq. 

freq. 

'pejorativ sl- 24 16 pj- 71 5 

e' 

'sound' kl-, 15 13, fn- 33 3 

kn- 

19 

'long thin sp- 23 17 spj- 33 2 

form' 

'quick or fl- 18 19 fl- 19 18 

strong 

movement 

' 

'wetness' sl- 12 8 skv- 40 4 

The most frequent semantic features for final clusters according to NFO 4 

are shown in table 8.5. Clusters and cluster frequencies are shown to the 

right (cf. tables 5.4, 5.18 and 5.19). 

Table 8.5 The most frequent semantic features for final clusters according 

to NFO 4. 

meaning in 

freq. 

freq. in % % 

'quick or 

Nl 4 Nl, 40, 

strong 

movement' 

dr, 

lr 

25, 

22 

'talking' tr, bl 5, 4 dr, 25, 

bl 24 

'sound' st, sl, 5 ,4, lr, sl 44, 

lr, tr 4, 4 

27 

'pejorative' fs, 10, fs, 59, 

sk, 5, 4 ms, 27, 

ms 

dr 25 

'wetness' sk 8 - - 

'long thin 

Nk, 7, 4 - - 

form' 

lk 

270

8.5 Main objective and further research 

The main objective of this thesis is the description of Swedish 

phonesthemes, which can constitute a basis for further research in this 

area. 

Another important result is the insight that sound symbolism is not an 

issue of whether or not phonesthemes exist. Instead, sound symbolism is 

present in consonant clusters in varying degrees. Some clusters are more 

sound symbolic than others while some meanings are more frequent than 

others. Some phonesthemes are more easily interpretable than others. 

Some phonesthemes are more productive than others and in slightly 

varying ways for different speakers. It is obvious that sound symbolic 

categories in Swedish show a conceptual organization that is compatible 

with both innate categories of thinking, such as similarity and contiguity, 

and with learning, in connection with sense impressions and emotion. 

An interesting continuation of this investigation is to study the effect of 

context on the interpretation of phonesthemes, e.g. in experiments with 

neologisms in different sentence contexts. 

Special spoken corpora would also be interesting to study, in search of 

neologisms (in context), e.g. corpora of child and adolescent language. 

It would also be interesting to study the stability of (the most common) 

phonesthemes over time, e.g. by analysis of "dead" words in the SAOB 

(Svenska Akademiens ordlista över svenska språket). 

Yet another interesting continuation would be to study the dispersion of 

(the most common) phonesthemes in dialect lexica. 

And, of course, it would be very interesting to expand the cross-linguistic 

comparisons to more semantic fields and to more languages. 

It is also very tempting to refresh e.g. the experiments of chapter 6 and 

the lexical decision experiment (Abelin, 1996) with auditive stimuli, with 

the aid of reaction time programs which were not available at the time 

271

when I initiated the experimentation. Naturally, the context effect should 

also be studied in this way. Connected with this is the aim to construct a 

psycholinguistic model for online processing for the perception and 

production of sound symbolic words. 

272

Bibliography 

Abelin, Å., Allwood, J. (1984) Tolkning av känsloprosodi - en kontrastiv studie, 

unpubl. paper, Göteborg University, Department of Linguistics. 

Abelin, Å. (1988) 'Patterns of synaesthesia in the Swedish vocabulary,' in Studies in 

Computer-Aided lexicology, Almqvist & Wiksell Int. Stockholm 

Abelin, Å. (1996) 'A lexical decision experiment with onomatopoeic, sound symbolic 

and arbitrary words', in TMH-QPSR 2/1996. Stockholm 

Adams, D., Lloyd, J. (1983) The meaning of Liff, Pan Books and Faber & Faber. 

London 

Allén, S., Eeg-Olofson, M., Gavare, R., Sjögreen, C. (1981) Svensk baklängesordbok, 

Esselte Studium. Stockholm 

Allén, S., Berg, S., Järborg, J., Löfström, J., Ralph, B., Sjögreen, C. (1980) 

Nusvensk frekvensordbok 4, Almqvist & Wiksell International. Stockholm 

Allott, R. (1973) The physical foundation of language, ELB Printers. Seaford Sussex 

Allwood, J. (1983) Om distinktionerna mellan semantik och pragmatik, Gothenburg 

papers in theoretical linguistics S3, Göteborg University, Department of Linguistics 

Allwood, J. (1983) 'Kan man tänka oberoende av språk' in U. Teleman (ed.) Tal och 

tanke, Liber förlag. Lund 

Allwood, J. (1985) 'Tvärkulturell kommunikation', in Papers in anthropological 

linguistics 12, Göteborg University, Department of Linguistics, pp. 9-61 

Allwood, J. (1989) Om begrepp – deras bestämning, analys och konstruktion (ms). 

Allwood, J., Andersson, L-G.(1976) Semantik. (Semantics). GULING 1. Göteborg 

University, Department of Linguistics 

Austerlitz, R. (1994) 'Finnish and Gilyak sound symbolism – an interplay between 

system and history'. in L. Hinton, J. Nichols, J.J. Ohala (eds) Sound symbolism, 

Cambridge University Press. Cambridge, pp. 249-260 

Bentley, M., Varon, E. (1933) An accessory study of phonetic symbolism Amer. J. 

Psychol. 45, 76-86 

Bloomfield, L. (1933) Language, Henry Holt. New York 

Bolinger, D. (1968) Aspects of language, Harcourt, Brace & World, Inc. New York 

Bolinger, D. (1950), Rime, assonance and morpheme analysis, Word 6: 117-136 

Bolinger, D. (1965) Forms of English: Accent, morpheme, order, Hokuou Publishing 

Company. Tokyo 

Bring, S. C. (1930) Svenskt ordförråd, Natur och Kultur. Stockholm 

Brown, R. W. (1958) Words and things, The Free Press. New York 

273

Bühler, K. (1969) "L'onomatopée et la fonction représentative du langage", in Essais sur 

le langage. J.-P. Pariente (ed.) Les Editions de Minuit. Paris (Partial translation of work 

originally published 1933 as Sprachtheorie ) Fisher. Jena 

Chastaing, M. (1958) "Le symbolisme de voyelles: significations des "i" I& II", Journal 

de Psychologie 55, 403-423 & 461-481 

Chastaing, M. (1965) "Pop - fop - pof - fof", Vie et Langage 159, 311-317 

Chastaing, M. (1966) "Si les r étaient des l", (Part 1) Vie et langage 173, 468-472; (Part 

2) 174, 502-507 

Childs, G.T. (1994) 'African ideophones'. in L. Hinton, J. Nichols, J.J. Ohala (eds.) 

Sound symbolism, Cambridge University Press. Cambridge, pp. 178-204 

Clark, H., Clark, E. (1977) Psychology and language, An introduction to 

psycholinguistics, Harcourt, Brace, Jovanovich. New York 

Cytowic, (1989) Synaesthesia - a union of the senses, Springer Verlag. New York 

Darwin, C. (1872) The expression of emotion in man and animals, John Murray. London 

Davis, R. (1961) The fitness of names to drawings: A cross-cultural study in 

Tanganyika, Brit. J. Psychol. 52, 259-268 

Diffloth, G. (1976) 'Expressives in Semai', Oceanic linguistics special publication no. 

13: Austroasiatic Studies I, Honolulu 

Diffloth, G., (1994) 'i:big, a:small', in L. Hinton, J. Nichols, J.J. Ohala (eds) Sound 

symbolism, Cambridge University Press. Cambridge 

Ekman, P. (1973) 'Cross-cultural studies of facial expressions' in P. Ekman (ed.) 

Darwin and facial expression , Academic Press. New York (pp. 169-222) 

Fano, G. (1962) Saggio sulle origine del linguaggio. Con una storia critica della dottrine 

glottogoniche, Guilin Einaudi Editore. Torino 

Fónagy, I. (1963) Die metaphern in der phonetik. The Hague 

Freud, S. (1891) On aphasia, a critical study, International Univ. Press. New York 

Gabelentz, G. v. d. (1891) Die Sprachwissenschaft, ihre Aufgaben, Methoden und 

bisherigen Ergebnisse. Leibzig 

Garlén, C. (1988) Svenskans fonologi: i kontrastiv och typologisk belysning, 

Studentlitteratur. Lund 

Garman, M. (1990) Psycholinguistics, Cambridge University Press. Cambridge 

Grammont, M. (1933) Traité de phonétique. Paris 

Hamano, S. (1994) 'Palatalization in Japanese sound symbolism', in L. Hinton, J. 

Nichols, J.J. Ohala (eds) Sound symbolism, Cambridge University Press. Cambridge, 

pp. 148-157 

Hellquist, E. (1966) Svensk etymologisk ordbok I-II. Gleerup. Lund 

274

Herder, J. G. (1772/1969) Essay on the origin of speech in P.H. Salus (ed.) On 

language: Plato to von Humboldt, Holt, Rinehart and Winston. New York, pp.147-166 

Hewes, G. W. (1977) 'Language origin Theories' in D. M. Rumbaugh (ed) Language 

Learning by a Chimpanzee, New York Academic Press. New York 

Hinton, L. (1986) 'Musical diffusion and linguistic diffusion'. in C. Frisbee (ed) 

Anthropology and Music: Essays in Honor of David P. McAllester. Detroit Monographs 

in Musicology 9. Detroit. Information Coordinators 

Hinton, L., Nichols, J., Ohala, J.J., (1994) 'Introduction: Sound symbolic processes', 

in L. Hinton, J. Nichols, J.J. Ohala (eds) Sound symbolism, Cambridge University 

Press. Cambridge, pp. 1-12 

Holland, M. K., Wertheimer, M. (1964) 'Some physiognomic aspects of naming, or, 

maluma and takete revisited', Percept. Mot. Skills 19, 111-117 

Householder, F.W. (1946) On the problem of sound and meaning, An English 

phonestheme (discussion in Language 23 feb 1946) 

Humboldt W. v. (1836/1907) Über die Verschiedenheit des menschlichen Sprachbaues 

und ihren Einfluss auf die geistige Entwicklung des Menschengeschlechts. Gesammelte 

Schriften, Vol. VII, Part 1. Königlich--Preussische Akademie der Wissenschaften. Berlin 

Hörmann, H. (1979) Psycholinguistics: an introduction to research and theory, Springer- 

Verlag. New York 

Ideforss, H. (1928) De primära interjektionerna i nysvenskan, 1, Primära impulsioner 

och imperationer, Carl Bloms boktryckeri. Lund 

Jacobsen Jr., W. H. (1994) 'Nootkan vocative vocalism and its implications' in L. 

Hinton, J. Nichols, J.J. Ohala (eds) Sound symbolism, Cambridge University Press. 

Cambridge, pp. 23-39 

Jakobson, R., Fant, G., Halle, M. (1957) Preliminaries to speech analysis, the 

distinctive features and their acoustic correlates, M.I.T. Press. Cambridge, Mass. 

Jakobson, R., Waugh, L. (1979) The sound shape of language, Harvester Press, Sussex 

Jespersen, O. (1918) "Nogle men-ord", studier tillegnade Esaias Tegner, Lund 

Jespersen, O. (1922 a) Language - its nature, development and origin, Allen and Unwin. 

London 

Jespersen, O. (1922 b) "Lydsymbolik", Nordisk tidskrift för vetenskap, konst och 

industri (II), Stockholm 

Jesperssen, O. (1933) "Symbolic value of the vowel i" [1922] in Linguistica, College 

Park. Maryland 

Jo'hannesson, A. (1949) Origin of language: four essays, Reykjavík, H.F. Leiftur, 

Blackwell. Oxford 

Katz, J.J., Postal, F. (1964) An integrated theory of linguistic descriptions, M.I.T. 

Press. Cambridge, Mass. 

Kotsinas, U. (1994) Ungdomsspråk, Hallgren & Fallgren. Uppsala 

275

Kukkonen, P. (1989) Från konst till vetenskap, PhD Thesis, Helsingfors university 

Källström, R. (1988) Morfologi, GULING 9, Göteborg University, Department of 

Linguistics. 

Köhler W. (1930) Gestalt psychology, Bell and Sons Ltd. London 

Lakoff, G., Johnsson, M. (1980) Metaphors we live by, The University of Chicago 

Press. Chicago and London 

Lakoff, G. (1987) Women, fire, and dangerous things: what categories reveal about the 

mind, University of Chicago Press. Chicago 

Lapolla, R.J, (1994) 'An experimental investigation into phonetic symbolism as it relates 

to Mandarin Chinese' in L. Hinton, J. Nichols, J.J. Ohala (eds.) Sound symbolism, 

Cambridge University Press. Cambridge 

Leech, G. N. (1969) Towards a semantic description of English, Longman. London 

Lehrer, A. (1974) Semantic fields and lexical structure, North-Holland. Amsterdam, 

London 

Lindblad, P. (1992) Rösten, Studentlitteratur. Lund 

Luria, A. (1977) Mannen med det obegränsade minnet, Wahlström & Widstrand. 

Stockholm 

Lyons, J. (1977) Semantics I-II, Cambridge University Press. Cambridge 

Malkiel, Y. (1978) 'From phonosymbolism to morphosymbolism', Fourth LACUS 

Forum, Columbia, S.C. 

Malkiel, Y. (1994) 'Regular sound development, phonosymbolic orchestration, 

disambiguation of homonyms' in L. Hinton, J. Nichols, J.J. Ohala (eds) Sound 


Malmgren, S-G, (1988) 'On regular polysemy in Swedish' in Studies in Computer-Aided 

Lexicology, Almqvist & Wiksell International. Stockholm 

Markell, N.N. & Hamp, E.P. (1960) "Connotative meanings of certain phoneme 

Sequences", Studies in linguistics 15, pp. 47-61 

Marks L.E. (1990) Synaesthesia: perception and metaphor in F. Burwick & W. Pape 

(eds) Aesthetic Illusion: theoretical and historical approaches, Berlin New York, W. de 

Gruyter 

Massaro, D. W., Tsuzaki, M., Cohen, M. M:, Gesi, A., & Heredia, R. (1993), 

'Bimodal speech perception : An examination across languages'. in Journal of Phonetics, 

21 

Meillet, A. (1931) Review of Wahlgren, 1930 Bulletin de la Société de linguistique de 

Paris 31 (2): 113-114 

Miller, G. A., Johnsson-Laird, P.N. (1976) Language and perception, Harward 

University Press. Cambridge, Mass. 

Müller, A. L., (1960) Experimentelle Untersuchungen zur stimmlichen Darstellung von 

Gefühlen, Dissertation. Göttingen 

276

Müller, M.F. (1861/1961) Three lectures on the science of language and its place in 

general education, repr. Indological Book House. Benares 

Nerman, T. (1954) Troll i ord, inrim och uddrim i svensk vers, Tidens förlag. Stockholm 

Newman, S. (1933) Further experiments in phonetic symbolism, American Journal of 

Psychology 45, 53-75 

Nordberg, B. (1986) 'The use of onomatopoeia in the conversational style of 

adolescents', FUMS rapport nr 132, Uppsala University. FUMS. Uppsala 

Ohala, J. J., (1994) The frequency code underlies the sound-symbolic use of voice pitch, 

in L. Hinton, J. Nichols, J.J. Ohala (eds.) Sound symbolism, Cambridge University 

Press. Cambridge 

Osgood, C. E. (1962 a) Psycholinguistic relativity and universality, Proc. 16th Int. 

Congr. Psychol. (Bonn 1960) North-Holland. Amsterdam 

Osgood, C. E. (1962 b) Studies on the generality of affective meaning systems, Amer. 

Psychologist 17, 10-28 

Osgood, C. E., G. J. Suci, and P. H. Tannenbaum (1957) The measurment of meaning, 

University of Illinois Press. Urbana, Illinois 

Oswalt, R. L. (1994) 'Inanimate imitatives in English', in L. Hinton, J. Nichols, J.J. 

Ohala (eds.) Sound symbolism, Cambridge University Press. Cambridge, pp. 293-306 

The Oxford dictionary of English etymology (1967), ed. by C. T. Onions, University 

press. Oxford 

Paget, R. (1930) Human Speech, Harcourt, Brace. New York 

Peirce, C.S. (1955) 'Logic as semiotic: The theory of signs', in Buchler, J. (ed) 

Philosophical writings of Peirce Dover. New York 

Peterfalvi, J.-M. (1965) 'Les recherches expérimentales sur le symbolisme phonétique', 

American Journal of Psychology 65, 439-473 

Plato (c. 428-348 B.C.) Cratylos dialogue (1962) Loeb Classical Library. Harward 

University Press. Cambridge 

Porzig, W. (1934/1973) 'Wesenhafte Bedeutungsbeziehungen'. Wortfeldsforschung 

1973. 78-103 

Révész, G. (1946) Ursprung und Vorgeschichte der Sprache, A. Francke. Bern 

Rhodes, R. (1994) 'Aural images', in L. Hinton, J. Nichols, J.J. Ohala (eds.) Sound 

symbolism, Cambridge University Press. Cambridge, pp. 276-292 

Roget's thesaurus of English words and phrases (1977) Penguin Books. 

Harmondsworth. Middlesex 

Rousseau, J.-J. (1822/1969) Essay on the origin of languages in P.H. Salus (ed.) On 

Language: Plato to von Humboldt, Holt, Rinehart and Winston. New York, pp. 138-146 

Samarin, W. (1978) 'Linguistic adaptation to speech function', in W. Mc Cormack, S.A. 

Wurm (eds) Approaches to language: Anthropological issues, Mouton. The Hague 

277

Sapir, E. (1921) Language, Harcourt Brace. New York 

Sapir, E. (1929) 'A study in phonetic symbolism', in D. Mandelbaum (ed.) Selected 

writings, Berkeley. California 

Saussure, F. de (1916/1983) Course in general linguistics, Duckworth. London 

Sereno J. A. (1994) 'Phonosyntactics', in L. Hinton, J. Nichols, J.J. Ohala (eds) Sound 


Sigurd, B. (1965) Phonotactic sructures in Swedish, Berlingska boktryckeriet. Lund 

Sigurd, B. (1993) 'Att tillverka varumärken', Språkvård 3, Svenska språknämnden. 

Stockholm, pp. 9-14 

Snow, C.E., Ferguson, C.A., (eds), (1977) Talking to Children: Language Input and 

Acquisition, Cambridge University Press. New York 

Stenström, C. (1984) Svenskans initiala konsonantkluster - en frekvensstudie, Unpubl. 

paper, Dept. of linguistics. Göteborg University, Department of Linguistics. 

Süssmilch, J.P. (1767), Versuch eines Beweises, dass die erste Sprache ihren Ursprung 

nicht vom Menschen, sondern allein vom Schöpfer erhalten habe, Akad. Vorl. Berlin 

Svensk Ordbok (1986) Esselte Studium. Stockholm 

Svenska Akademiens ordlista över svenska språket (1974) 10 uppl., P.A. Norstedt & 

Söner förlag. Stockholm 

Svenska Akademiens ordlista över svenska språket (1998) 12 uppl., Norstedts. 

Stockholm 

Ordbok över svenska språket, utgiven av Svenska Akademien (SAOB) (1926), 

Lindstedts universitetsbokhandel. Lund 

Thorndike, E. L. (1943) Man and his works, Harvard University Press. Cambridge, 

Mass. 

Thun, N. (1963) Reduplicative words in English: a study of formations of the types ticktock, 

hurly-burly, and shilly-shally. PhD Thesis, Stockholm University 

Traunmüller, H. (1996) 'Sound symbolism in deictic words', TMH-QPSR 2/1996. 

Stockholm 

Trier, J. (1934) Das sprachliche Feld, Neue Jahrbucher fur Wissenschaft und 

Jugendbildung 10, 428-449 

Tsuru, S., Fries, H. S. (1933) 'A problem in meaning', Journal of Gen. Psychology 8, 

281-284 

Ullman, S. (1972) Semantics An introduction to the science of meaning, Blackwell. 

Oxford 

Ultan, R. (1978) 'Size-sound symbolism', in J. Greenberg (ed) Universals of human 

language 2, 

278

Usnadze, D. (1924) 'Ein experimenteller Beitrag zum problem der psychologischen 

Grundlagen der Namengebung', Psychologische Forschung 5, 24-43. 

Werner, H. (1953/1957) Comparative psychology of mental development, International 

Universities Press. New York 

Werner, H., Kaplan, B. (1963) Symbol formation, Wiley. New York 

Wescott, R. (1975) 'Allolinguistics: exploring the peripheries of speech', in P. Reich 

(ed),The Second LACUS Forum, The Hornbeam Press, Columbia, South Carolina, 497- 

513 

Wescott, R. (1978) 'Lexical polygenesis: Words as resultants of multiple linguistic 

pressures' in Semantics and Lexicology 

Whorf, B. L. (1956/1967) Language, thought and reality: selected writings, ed J.B. 

Carroll (ed), The MIT paperback series 5, MIT Press, Cambridge, Mass. 

Viberg, Å. (1984) 'The verbs of perception: a typological study', Linguistics, 21, 1. 

Viberg, Å. (1993) 'Crosslinguistic perspectives on lexical organization and lexical 

progression' in K. Hyltenstam & Å. Viberg (eds), Progression and Regression in 

language, Cambridge University Press. Cambridge, pp. 340-385 

Wilde, K. (1958) 'Naive und kunstlerische Formen des graphischen Ausdrucks', Ber. 

21. Kongr. Dtsch. Ges. Psychol., Göttingen: Verlag fur Psychologie, Bonn, pp. 157- 

159 

Wisseman, H. (1954) Untersuchungen zur Onomatopoie, Part I: Die 

sprachpsychologischen Versuche, Winter. Heidelberg 

279

From SOB 

Word: Key words,SOB Category 

bjäbba uppnosigt prat pejorative, talking 

bjäfs överdriven pejorative 

bjällra klingande sound 

bjärt lysande light 

bjässe mycket stor augmentative 

black ljus, färg light 

bladdra prata strunt pejorative, talking 

blaffa större fläck mindr tilltalan pejorative 

blaj meningslöst prat pejorative, talking 

blank glansig yta light 

blarr struntprat pejorative, talking 

blaska röra i vatten plaska dålig pejorative, wetness 

blek färg light 

blemma blåsa round form 

blick seende gaze 

bliga titta envetet, dumt, fånstirra pejorative, gaze 

blind se gaze 

blinka ögonen gaze 

blixt ljus light 

block massivt stycke shortwide form 

blod vätska wetness 

blond ljus färg light 

blossa brinna, låga light 

bluddra struntprat pejorative, talking 

blunda ögonen gaze 

blunder misstag pejorative 

blåsa rundad round form 

bläck vätska färgad wetness 

bläddra blemma, blåsa round form 

blända ljus light 

blänga betrakta gaze 

blänka ljus light 

bläs ljus strimma light 

blöt vatten wetness 

braka ljud sound 

brassa skjuta , stora krafter sound 

brista sönder destruction 

brum ljud sound 

brus ljud sound 

bryta kraftig, spricka destruction 

bråte förbrukat, oanvändbart destruction 

bräck krosskada destruction 

bräka ljud sound 

bränning splitras destruction 

bröl ljud sound 

Appendix 1 Sida 1

drasut nedsättande pejorative 

dratta slarv, fumlig falling, pejorative 

dravel struntprat talking, pejorative 

dregel saliv, oavsiktligt wetness, pejorative 

dribbla rörelser movement 

drill ton sound 

droppa falla falling, wetness 

drulle ohyfsat, vårdslöst pejorative 

drummel klumpigt, vårdslöst pejorative 

dråsa falla klumpigt falling 

drägg botten, slödder pejorative 

drälla planlöst, vårdslöst, förflytta falling, pejorative 

drämma slå hårt beat 

dröse osorterad, ostrukturerad mängd pejorative 

drösa falla, tätt, okontrollerat falling 

fjant åtlöje pejorative 

fjolla saknar stadga i karaktär pejorative 

fjompig löjlig pejorative 

fjuttig futtig, torftig pejorative 

fjärta ljud, ofrivilligt sound 

fjäsa göra sig till pejorative 

fjäsk överdriven pejorative 

flabb skratt sound, pejorative 

flacka planlöst fara q or s movement, pejorative 

fladdra rörelse q or s movement 

flaga tunt stycke thin form 

flagga vifta q or s movement 

flak flat form, större stycke thin form 

flamma stark eld, orolig, plötslig light, q or s movement 

flamsa bullersam, tröttsamt, vårdslöst, stojigt pejorative 

flanera vandra utan bestämt mål q or s movement 

flarn bräckligt stycke poröst thin form 

flaxa slå häftigt, flygande q or s movement 

flimra dallrande ljus q or s movement, light 

flina mindre vackert, försmädligt pejorative 

flinga litet lätt tunt stycke thin form 

flink snabbt quickness 

flisa litet tunt vasst stycke thin form 

flitter oäkta, värdelös pejorative 

flopp snöplig, misslyckande pejorative 

floskel högtravande intetsäg. löjlig pejorative 

fluffa lätt luftig ruska airy consistency 

fluga flygande q or s movement 

flukta kasta snabb blick gaze 

fluktuation ständigt stigande och fallande q or s movement 

flummig intellektuellt oredig pejorative 

flyga förflytta sig q or s movement 

fly hastigt lämna q or s movement 

Appendix 1 Sida 2

flyta transporteras q or s movement 

flytta till annan plats q or s movement 

flåsa tungt snabbt q or s movement 

fläkta blåsa q or s movement 

flämta snabbt häftigt q or s movement 

flänga rusa framochtillbaka hetsigt q or s movement 

flöda rinna stor mängd q or s movement 

flöjel vindriktning lätt skiva thin form 

flöjt ton luftström skarp kant sound 

fnasig fjälla thin form 

fnask nedsättande pejorative 

fnatta springa planlöst q or s movement, pejorative 

fnissa skratta sound 

fnittra skratta sound 

fnoskig tokig pejorative 

fnurra osams pejorative 

fnysa häftigt ljudligt förakt sound 

fnöske porös dryness 

fradga skummande wetness 

frasa ljud sound 

frat söndergnagt skräp destruction 

frossa skakning köldkänslor q or s movement 

frotté poröst rough surface structure 

frusa häftigt välla fram q or s movement 

frusta ljud sound 

fryna rynka rough surface structure 

fräsa ljud sound 

fräta upplösa destruction 

glacé glänsande light 

glací jämn yta smooth surface str 

glaciär is smooth surface str 

glam högljudd sound 

glana stirra gaze 

glans jämnt sken blank yta light 

glas glänsande smooth surface str 

glatt slät blank hal smooth surface str 

glaubersalt glas smooth surface str 

glaukom syn gaze 

gli litet diminutive 

glida jämnt smooth surface str 

glimma lysa svagt skiftande light 

glimmer glänsande light 

glimra lysa svagt skiftande light 

glimt blänk ljus light 

glinder glänsande light 

glindra glimra light 

glipa smalt form, diminutive 

Appendix 1 Sida 3

glisa lysa light 

glissando glida ton sound, smooth surface 

glittra lysa starkt snabbt light 

glo stirra dumt gaze 

glob klotformig form 

glop slyngel pejorative 

glopp snö regn wetness 

gloria ljus light 

glosögd ögon gaze 

glosa glo gaze 

glufsa mycket ovårdat pejorative 

glugg liten form, diminutive 

glunkas sägas skvaller pejorative 

glupa glupskt pejorative 

glupsk nedsätt pejorative 

glutta titta gaze 

glytt barnunge diminutive 

glåmig blek light 

glåpord förolämpande yttrande pejorative 

gläfs svagt hundskall sound 

glänsa sken blank light 

glänta mindre öppen form, diminutive 

glöd sken light 

gnabb smågräl talking 

gnat småaktig upprepad talking, pejorative 

gneta småsaker, oväsentlig, snål pejorative 

gnida upprepade rörelser smooth surface str, q or s mo 

gnidare snål pejorative 

gnissla ljud sound 

gnista brinnande glödande light 

gnistra korta starka ljusglimtar light 

gno små snabba rörelser smooth surface str, q or s mo 

gnoding knorrhane sound 

gnola sjunga talking 

gnugga små korta rörelser smooth surface str, q or s mo 

gnutta mycket liten diminutive 

gny ljud talking 

gnägga kraftigt läte sound 

gnälla läte sound 

gnöla klaga talking 

gramse arg irriterad bad mood 

grav grävd hålighet i marken hollow form 

grift grav hollow form 

groll ovänskap bad mood 

grop djup hålighet hollow form 

grotta håltum hollow form 

grubba grop hollow form 

grubbla dyster bad mood 

Appendix 1 Sida 4

gruff gräl bad mood, talking 

grumlig ogenomskinlig oönskad pejorative 

grummel grums pejorative 

grums oönskad pejorative 

grumsa klagomål pejorative, talking 

grunka föremål vard slang 

grym samvetslöst tillfogar lidande bad mood 

grymta ljud talking 

gryt jordhåla hollow form 

gråta röst talking 

gräla högljudd respektlös diskussion bad mood, talking 

gräll syn alltför starkt light, pejorative 

gräma oönskad pejorative 

gräslig stark olust pejorative 

gräva djupare grop hollow form 

gröpa gräva mindre hål hollow form 

klabb kort tjockt stycke shortwide form 

klabb blöt wetness 

klack utskjutande, del shortwide form 

kladd blöt vidhäftande wetness 

klafs ljud våt klibbig sound, wetness 

klaga framföra missnöje talking 

klamp grovt stycke shortwide form 

klamp ljud sound 

klamra gripa adhesion 

klandra ogillande yttra talking 

klang toner sound 

klanka upprepat förebrå småaktigt talking, pejorative 

klanta dum klumpig pejorative 

klappa upprepat slå sound 

klappra ljud sound 

klatsch ljud sound 

klena kladdig wetness 

kleta blöt vidhäftande wetness, adhesion 

klibba kladdig wetness, adhesion 

klick klump smetig wetness, shortwide form 

klick ljud sound 

klifs ljud blött sound, wetness 

klimp mindre stycke shortwide form 

kling ljud sound 

klink spel sound 

klirra ljud sound 

klister klibbig adhesion 

klocka klang sound 

kloss grovt stycke shortwide form 

klot runt shortwide form 

klotter slarvig pejorative 

klubba kloss shortwide form 

klucka ljud sound, wetness 

Appendix 1 Sida 5

kludda dålig slarvig pejorative 

klump oformligt större stycke shortwide form 

klunk vätska wetness 

kluns klumpig pejorative 

klåpa dålig slarvig pejorative 

klämta ringa sound 

klänga hålla kvar adhesion 

kläpp ljud, klump sound, shortwide form 

klätt liten topp shortwide form 

klättra fästpunkt adhesion 

knacka upprepade slag lystring sound 

knackig dålig pejorative 

knagg pinne round form 

knaggla röra sig walking 

knaka ljud sound 

knal svag knapp pejorative 

knall ljud sound 

knalla gå långsamt vard walking 

knalle liten rund höjd round form 

knapp skiva, kula round form 

knapra knastrande sound 

knarr ljud sound 

knarrig vresig bad mood 

knasig tokigt dum pejorative 

knast utväxt round form 

knastra ljud sound 

knata springa vard walking 

knatte liten diminutive 

knatter ljud sound 

kneg nedsättande pejorative 

kneken förfallet pejorative 

knick tvär böj round form 

knipa ihop putting together 

knippa hophållna putting together 

knipsa nypa putting together 

knirka ljud sound 

knirra ljud sound 

knistra ljud sound 

knittra ljud sound 

knixa liten hastig q or s movement 

knocka slå beat 

knodd nedsättande pejorative 

knoga mödosamt gå walking 

knoge upphöjning round form 

knollra smålockig round form 

knop knut round form 

knopp liten kula round form 

knorr liten vriden form round form 

knorr muttrande kurrande sound 

Appendix 1 Sida 6

knorva veckig rynkig pejorative 

knot knorrhane sound 

knota muttrande klaga talking 

knota förtjockad round form 

knott liten mygga diminutive 

knottra liten upphöjning round form 

knubbig kort tjock round form 

knuff stöt beat 

knut sammanfogning putting together 

knutte nedsätt pejorative 

knyck plötslig rörelse q or s movement 

knyckla ihop putting together 

knyppla flätas putting together 

knyst ljud sound 

knyta samman putting together 

knåp smått obetydligt diminutive 

knäck brott slag beat 

knäppa ljud sound 

knödel bulle round form 

knöl mindre rundad round form 

knöl otrevlig pejorative 

knös nedsätt pejorative 

krabat litet diminutive 

kracka sönderdela destruction 

krackelera fina sprickor destruction 

krafs värdelösa småsaker pejorative 

krafsa ljud sound 

krake beklagansvärd pejorative 

krakel högljutt sound 

krakmandel tunt bräckligt skal thin form 

kram värdelösa pejorative 

kram våt wetness 

krams värdelösa småsaker pejorative 

krans ring sammanflätad winding form 

kras ljud sound 

krasch ljud sound 

kratsa rörelse q or s movement 

kratta fåra rough surface structure 

kratta dålig pejorative 

kravla mödosam förflytta walking 

kraxa ljud sound 

kreta karva dålig pejorative 

krimskrams värdelösa småsaker pejorative 

kringelikrokar mängd krökar winding form 

kringla form av en båge winding form 

krock häftig sammanstötning beat 

krok långsmalt halvcirkel winding form 

krokan krusiduller, snirklig winding form 

krokett frasig thin form 

Appendix 1 Sida 7

krossa sönder destruction 

kruka cylindrisk bukig round form 

krulla små lockar winding form 

krum böjd winding form 

krumbukt kraftfull slingrande rörelse winding form 

krumelur krokig linje winding form 

krumpen böjd winding form 

krusa full av små vågor winding form 

kruserlig tillgjord pejorative 

krusiduller form slingrande linjer winding form 

krusta skal thin form 

krustad bakverk i form av bägare thin form 

krycka böjt winding form 

krylla rörlig walking 

krypa förflytta sig walking 

krysta samtal, formulering sound 

kråka fågel sound 

kråma stolt vrida winding form 

krångel besvärligt pejorative 

krås veckad remsa rough surface str 

kräk klandervärd pejorative 

kräkla inrullat krön winding form 

kräla förflytta sig slingrande walking, winding form 

kräm tjockflytande wetness 

krämta harkla sound 

kränga vickande rörelser q or s movement 

kräpp tunt krusig yta rough surface str 

krök kraftig sväng winding form 

kvacka läte sound 

kvadda förstöra destruction 

kval lidande bodily feeling 

kvalm osund äckel bodily feeling 

kvarka sound 

kvarn nedbrytande destruction 

kvav kvalmig kväva bodily feeling 

kverulans klaga talking 

kvick små snabba rörelser q or s movement 

kvida ljud sound 

kvillra ljud sound 

kvirra kverulera sound 

kvissla liten blåsa diminutive 

kvist liten smal diminutive 

kvitter läte sound 

kväda sjunga sound 

kväka läte sound 

kvälja lukt äckla bodily feeling 

kväva svårighet att andas bodily feeling 

mjau läte sound 

Appendix 1 Sida 8

mjuk angenäm känsel jämn slät soft consistency 

mjäkig alltför vek pejorative 

mjäla mjölfin fine grain 

mjäll fin konsistens soft consistency 

mjäll fjäll fine grain 

mjöl finmald fine grain 

njugg onödigt snål pejorative 

pjatt obetydlig hållningslös pejorative 

pjoller meningslöst prat pejorative, talking 

pjosk överdriven pejorative 

pjunk gnäll veklighet pejorative 

pjåkig dålig pejorative 

pladask klumpigt ljudligt fall sound 

pladder oavbrutet innehållslöst prat talking 

pladuska stor utbredning oönskad pejorative 

planka fyrsidigt stycke form 

plask häftigt vatten ljud sound, wetness 

platt ytor form 

pligg kort spetsigt stift form, diminutive 

pling ljud sound 

plint lådformigt avsmalnande form 

plira titta halvt tillslutna gaze 

plissé regelbunden fin veckning rough surface structure 

plit starkt vard. slang 

ploj skämtsamt spratt slang 

plomb bly, fyllning form 

plottra spridda oväsentliga smådetaljer pejorative 

plufsig slappt fet pejorative 

plugg litet cylindriskt form, diminutive 

plump fläck pejorative 

plump grov ohyfsad pejorative 

plums ljud sound, wetness 

plundra hänsynslöst pejorative 

plunta vard. slang 

pluring starkt vard slang 

plurr vatten wetness 

plussig uppsvälld pejorative 

plutt liten diminutive 

plym fjäder soft consistency 

plymå dyna soft consistency 

plysa luckra upp ull soft consistency 

plysch sammetsliknande långhårigt soft consistency 

plätt litet runt form, diminutive 

plös kilformigt form 

pracka lura pejorative 

prassel ljud sound 

Appendix 1 Sida 9

prat tal talking 

prick mycket liten rundad platt form, diminutive 

prilla portion slang 

prillig småtokig slang 

propp föremål täppa hål öppning form 

propsa envist kräva talking 

prunka lysande färger light 

pruta övertala talking 

prutt fjärt sound 

pryl grovt nålformigt form 

pryl onyttig överflödig pejorative 

pryttel slang 

prål strålande light 

prång liten trång form, diminutive 

prångla tvivelaktigt knep pejorative 

pröjs vard. slang 

pröjsare vard slang 

skaka rörelse q or s movement 

skakel stång long thin form 

skal hårt tunt hölje hardness 

skalk hård kant hardness 

skall läte sound 

skalla ljuda starkt sound 

skalle hårdhet hardness 

skallra ljud sound 

skalm avlång rörlig long thin form 

skalpell rak long thin form 

skalv darra q or s movement 

skare hårt ytskikt hardness 

skarp spetsig kant hardness 

skava hårt tryck föras upprepat hardness 

skavank fel pejorative 

skippa hoppa q or s movement 

skopa mindre halvklotformigt round form 

skorpa hårdnad yta hardness 

skorr ljud sound 

skorvig yta hård ojämn hardness 

skorv gammal förfallen pejorative 

skott ljud sound 

skovel stor bred rundad round form 

skångra skaka starkt ljud sound 

skåra långsträckt fördjupning long thin form 

skrabba gammal utsliten destruction 

skraj vard. slang 

skral dålig pejorative 

skraltig dålig svag destruction 

skramla ljud sound 

skranglig ostadig destruction 

skrap ljud sound 

Appendix 1 Sida 10

skratta ljud talking 

skri läte starkt talking 

skrocka läte talking 

skrodera skryta talking, pejorative 

skrot skräp värdelöst destruction 

skrovlig grov ojämn yta rough surface str 

skrubb förvaringsutrymme pejorative 

skrubba gnida hårt grov ojämn rough surface str 

skrubbor bannor talking 

skrumpen förtorkad rynkig rough surface structure 

skrutt dåligt pejorative 

skrymsle litet utrymme diminutive 

skrymt spökeri troll pejorative 

skrynkla oönskat veck rough surface structure 

skryta tala överdrivet talking 

skrål pratande talking 

skråma mindre ytligt sår rough surface structure 

skrälla ljud sound 

skrälle gammalt nedslitet destruction 

skräna röster talking 

skräp värdelös destruction 

skräppa skryta talking, pejorative 

skrävla skryta talking, pejorative 

skröna lögnaktig talking, pejorative 

skröplig svag ömtålig destruction 

skval ljud vatten sound, wetness 

skvaller löst prat talking, pejorative 

skvalp ljud sound 

skvalta stänka wetness 

skvimpa skvalpa små rörelser movement, wetness 

skvätt liten mängd vätska wetness 

slabba vätska slarv wetness, pejorative 

slabbertacka sladdertacka talking, pejorative 

slack alltför stor slakhet slackness 

sladd tamp long thin form 

sladda okontrollerat glida q or s movement 

sladdra ointressant skvaller talking, pejorative 

sladdrig saknar styvhet slackness 

slafs kladdande wetness, pejorative 

slag hård träff ljud ton beat 

slak mjuk böjlig slackness 

slam vätska wetness 

slammer starkt buller sound 

slampa lösaktig slarvig pejorative 

slams värdelöst prat pejorative 

slamsa smal sladdrig remsa slackness, long thin form 

slana smal böjlig slackness, long thin form 

slang långt böjligt slackness, long thin form 

slank smal long thin form 

Appendix 1 Sida 11

slankig alltför mjuk böjlig slackness 

slant långt, spö long thin form 

slapp alltför mjuk slackness 

slarvig inte noggrann och ordentlig pejorative 

slas slö slarvig pejorative 

slasa gå tungt och hasande pejorative, walking 

slask blöt sörja sopor wetness, pejorative 

slatt bottenskvätt wetness, pejorative 

slattrig slapp, pratig talking, pejorative 

slejf band long thin form 

slem segt sekret wetness 

slicka flytande kletigt wetness 

slidder ointressant skvaller talking, pejorative 

sliddrig sladdrig slackness 

slimmad insydd long thin form 

slinga krök vindling long thin form 

slinka förflytta sig snabbt q or s movement 

slinka prostituerad pejorative 

slinkig sladdrig slackness 

slinta plötsligt glatthet q or s movement 

slipa yta slät gnida smooth surface str 

slipprig hal kladdig wetness, q or s movement 

slips band long thin form 

slira okontrollerat glida q or s movement 

slisk äckligt kladdigt wetness, pejorative 

slita häftigt rycka q or s movement 

slits smal öppen ränna long thin form 

slok odåga pejorative 

sloka slapp avlång böjlig slackness 

sludder otydligt tal talking, pejorative 

slugga boxas vilt q or s movement 

slum förfallet pejorative 

slunga kasta stor rörelse q or s movement 

sluring soppa wetness 

slurk liten vätska wetness, diminutive 

slusk mycket smutsig ovårdad pejorative 

sluss vatten wetness 

slutta luta nedåt yta potential movement 

slyna ouppfostrad slarvig pejorative 

slyngel ouppfostrad pejorative 

slå slag beat 

släde medar long thin form 

slägga stor tung hammare beat 

slänga kasta vårdslöst häftigt q or s movement 

slänt sluttning potential movement 

släntra gå långsamt utan mål walking 

slät fri från ojämnhet yta smooth surface str 

slätt vidsträckt plant smooth surface str 

slö kraftlös håglös pejorative 

slödder föraktad ouppfostrad pejorative 

Appendix 1 Sida 12

slösa för mycket misshushålla pejorative 

smack ljud sound 

smal ringa bredd narrow form 

smash snabbt hårt slag beat 

smask ljud sound 

smatt prång diminutive 

smatter ljud sound 

smegma sekret wetness 

smet kladdig röra wetness 

smetana tjock grädde wetness 

smicker överdrivet pejorative 

smink krämig konsistens wetness 

smisk aga beat 

smita avlägsna sig obehagligt secrecy 

smock rynkning i rutmönster rough surface str 

smocka hårt slag beat 

smolk smutspartikel pejorative 

smuggla olovligt secrecy 

smula liten partikel torr lös diminutive 

smussel i smyg olovligt secrecy 

smuts fläckar oönskade pejorative 

smutt liten klunk diminutive 

smutt prång diminutive 

smyga röra sig tyst och smidigt secrecy 

små ringa utsträckning diminutive 

smäck slag med handflata beat 

smäcker slank elegant narrow form 

smäda omdöme talking 

smäkta intensiv längtan mental feeling 

smäll aga beat 

smälta flytande wetness 

smärt smal smidig narrow form 

smärta plågsam känsla mental feeling 

smätta ljud sound 

smörj stryk beat 

smörja mjukt kräm wetness 

snabb quickness 

snabel mycket lång utdragen long thin form 

snacka prata talking 

snafs smuts pejorative 

snagga mycket kort small end form 

snappa kort snabb rörelse quickness 

snar snabb quickness 

snarka ljud sound 

snarpa läte sound 

snarra skorra sound 

snaskig sölig kladdig wetness, pejorative 

snatta mindre värde pejorative 

snattra läte sound 

Appendix 1 Sida 13

snava falla snubbla walking 

snegla titta ögon gaze 

snibb hörn trekantigt small end form 

snicksnack struntprat talking, pejorative 

(snigel) lång slemmig wetness, long thin form 

sniken girigt orättmätig pejorative 

snilja tråd long thin form 

snip spetsig form small end form 

snirkel starkt böjd linje spiral winding form 

snitsel smal remsa long thin form 

snitt liten finare diminutive 

sno långsmalt long thin form 

sno snabbt quickness 

snobb överdrivet fåfängt pejorative 

snodd garn long thin form 

snofsa vard. slang 

snok långt smalt long thin form 

snopen besviken mental feeling 

snopp smal ända small end form 

snor slem wetness 

snorkel rör long thin form 

snorkig oartig pejorative 

snubba tillrättavisning talking 

snubbe vard. slang 

snubbla falla walking 

snudig flink quickness 

snugga liten pipa diminutive 

snultra långsträckt long thin form 

snurra vrida cirkelformig winding form 

snusa hörbart andas sound 

snusk smuts illaluktande pejorative 

snut vard slang 

snutt liten kort diminutive 

snuva slem wetness 

snyfta stötvisa andetag sound 

snylta ytnyttja pejorative 

snyta fräsa sound 

snål överdrivet pejorative 

snäcka spiralvridet winding form 

snäppa ljud sound 

snärj högt tempo quickness 

snärp fågelläte sound 

snärt ända slag beat 

snärta nedsätt pejorative 

snäsa irriterad tillrättavisning talking 

snöd fåfäng tarvlig pejorative 

snöpa stumpa, vanställa pejorative 

snöre garn long thin form 

snörvla ljud sound 

Appendix 1 Sida 14

spackel degartat wetness 

spad vatten wetness 

spade platt thin form 

spagat åt var sitt håll separation 

spaljé spjälor long thin form 

spalt lodrät long thin form 

spana blicken gaze 

spant balk long thin form 

spark hård stöt beat 

sparre bjälke long thin form 

sparris stam long thin form 

spasm ryckning q or s movement 

spat klyvbart glansigt thin form, light 

spatel platt avlångt long thin form 

spatt tokig överdrift jocular 

spe illvilligt förlöjligande jocular 

spegel plan smooth surface str 

speja blicken gaze 

spektakel förargelse jocular 

spene utväxt long thin form 

spenslig tunn smal long thin form 

sperma vätska wetness 

speta litet tunt vasst long thin form 

spets tunn ända avsmalnande long thin form 

spett lång stång spets long thin form 

spex parodisk dråplig jocular 

spigg taggar long thin form 

spik spetsig pinne long thin form 

spila spjäla long thin form 

spilkum liten skål diminutive 

spill blir över pejorative 

spilla vätska wetness 

spindel långa ben long thin form 

spinkig mycket tunn long thin form 

spinna garn long thin form 

spinna ljud sound 

spira smal stav long thin form 

spiral kurva vriden winding form 

spola vatten wetness 

spole avlångt long thin form 

spoling ouppfostrad pejorative 

spont utstående parti long thin form 

spotsk hånfullhet attitude 

spott vätska wetness 

spurt ökning av farten q or s movement 

sputum slem wetness 

spydig elaka kommentarer attitude, talking 

spån tunn avhyvlad bit thin form 

Appendix 1 Sida 15

spång smal long thin form 

späd liten diminutive 

spänta stickor long thin form 

spö smal böjlig käpp long thin form 

spjut lång smal stång hård spets long thin form 

spjuver skämtar luras jocular 

spjäla ribba long thin form 

spjälka sönderdela destruction 

splinta klyva separation 

split oenighet separation 

splits ändar av tågvirke separation 

splitter små vass slagits sönder destruction, diminutive 

spraka knastra gnistra sound, light 

spralla kroppsligen livlig q or s movement 

spratt skämtsam lura jocular 

sprattla rörelser q or s movement 

sprej finfördelat vätska wetness, separation 

spreta åt olika håll separation 

spri smäckert rundhult long thin form 

spricka långsmalt brott long thin form 

sprida fördela utbredning separation 

springa lång mycket smal öppning long thin form 

springa förflytta sig walking 

sprinkler vatten, sprider wetness, separation 

sprint pinne long thin form 

sprits strut form long thin form 

spritta plötsligt rycka q or s movement 

sprudla bubblande välla fram q or s movement, wetness 

sprund smal öppning long thin form 

spruta häftigt stråle vätska q or s movement, wetness 

sprutt fart vard. q or s movement 

spränga stor kraft splittras separation 

sprätt överdrivet pejorative 

sprätta kringkastande separation 

spröjs list long thin form 

spröt långt, smalt long thin form 

stabbig kort kraftig short form 

stake lång käpp long thin form 

stamma tala stötigt talking, iterative 

stampa stöta hårt ljudligt sound, iterative 

stappla gå ostadig walking 

stav långt smalt long thin form 

stepp ljudliga slag sound, beat 

stick smärtsamt vass bodily feeling 

sticka tunn vass flisa long thin form 

Appendix 1 Sida 16

stift smalt spetsigt long thin form 

stig smal long thin form 

stilett stickvapen smal long thin form 

stim ljud röster talking 

sting smärtsam vass bodily feeling 

stint blick gaze 

stirra titta gaze 

stirrig upprörd virrig mental feeling 

stock stam long thin form 

stoft små partiklar fine grain 

stoj högljutt prat talking 

stolle tok pejorative 

stolpe påle long thin form 

stubb uppstickande strån short form 

stubbe uppstickande rest short form 

studsa rörelse q or s movement, iterative 

stulta tulta walking 

stump återstående bit short form 

stumpa liten diminutive 

stura tjura attitude 

stursk trotsig fräck attitude 

stybb finfördelat spill fine grain 

stycka hugga i bitar separation 

stylta stång long thin form 

stympa avhuggning short form 

styng smärtsam vasst bodily feeling 

styv ej lätt böjs stiffness 

stång avlängt rakt long thin form 

stånka läte talking 

stänka vätska wetness 

stöddig stor kraftig alltför självsäker attitude, pejorative 

stön ljud sound 

stöppla stöta q or s movement 

stör stång long thin form 

stöt rörelse kraftigt q or s movement 

stötta stolpe long thin form 

stöv hudavlagring fine grain 

strak rak stel stiffness 

stram spänd stiffness 

streamer långsmal remsa long thin form 

streck kortare linje long thin form 

stretch töjbar long thin form 

strigel läderrem long thin form 

stril ljud av vätska long thin form, sound, wetness 

strimla liten långsmal bit long thin form 

strimma långsmalt band ljus long thin form, light 

stripa rak hårtest long thin form 

stropp ögleformad round form 

Appendix 1 Sida 17

strosa lugnt promenera walking 

strul besvärligt pejorative 

strunt värdelöst pejorative 

strut konformigt long thin form 

strutta gå knyckigt walking 

strå stjälk long thin form 

stråk bandliknande long thin form 

stråle ljus long thin form, light 

stråt väg bana long thin form 

sträcka längd long thin form 

sträng spänd tråd long thin form 

sträv yta ojämnhet rough surface str 

strö finfördelat material fine grain 

ström rinnande vatten long thin form, wetness 

strössel avlånga korn fine grain, long thin form 

ströva långsamt gå walking 

svabba tvätta långa garnändar wetness, long thin form 

svacka större fördjupning bent form 

svada tala länge talking 

svaja svänga fram och tillbaka q or s movement 

svalla röra sig häftigt q or s movement 

svamla prata strunt talking. pejorative 

svank inåtböjt bent form 

svans långsmal long thin form 

svassa tillgjort pejorative, walking 

svepa rörelse q or s movement 

svett vätska wetness 

svicka tapp short form 

svida brännande smärta bodily feeling 

svikt böjlighet bent form 

sving rörelse q or s movement 

svirvel vridbar q or s movement 

svischa ljud sound 

svulstig alltför pejorative 

svämma vatten q or s movement, wetness 

sväng rörelse q or s movement 

svära yttra talking 

svärm flygande mindre q or s movement 

sväva rörelse q or s movement 

tradig långtråkig pejorative 

traggel ständig tröttsam upprepning pejorative 

trampa rörelse walking 

trams dumt prat pejorative, talking 

trasa sönderrivet destruction 

traska gå vårdslöst klumpigt walking, pejorative 

trassel virrvarr krångel pejorative 

tratt strutformig pip tratt long thin form 

trava röra sig walking 

Appendix 1 Sida 18

tredsk envis ovillighet bad mood 

trilla falla falling 

trilla forma runt round form 

trilsk enveten motsträvig bad mood 

trind klotrund form round form 

tripp kortare diminutive 

trippa gå små lätta hoppande walking, diminutive 

trips liten diminutive 

trissa liten rund round form, diminutive 

trist enformig nedstämd bad mood 

troll ful klumpig enfaldig pejorative 

tromb häftig luftvirvel q or s movement 

tromla roterande cylindrisk q or s movement 

tross lina long hin form 

trubbel besvär problem bad mood 

trubbig inte avsmalnande short form 

trudelutt glad melodi(stump) short form 

trulig trumpen bad mood 

truls liten slarvig pejorative 

trumla roterande q or s movement 

trumma slaginstrument sound 

trumpen butter och missnöjd bad mood 

trumpet blåsinstrument sound 

trunk stor augmentative 

truta skjuta fram läpparna long thin form 

tryne nos long thin form 

tråd tunt utdraget long thin form 

tråg avlångt fyrkantigt long thin form 

tråkig enformig negativ bad mood 

trål strutformig long thin form 

tråna känna stark längtan mental feeling 

trång minsta diminutive 

träda långsamma steg walking 

träns snöre long thin form 

träsk vatten wetness 

träta irriterad ordväxling bad mood, talking 

tving sammanhållning putting together 

tvinna sno ihop trådar putting together, long thin for 

vrak skadat odugligt destruction 

vrede stark känsla förorättad bad mood 

vrensk själsligt motsträvig bad mood 

vresig ovänlig missnöje bad mood 

vret liten diminutive 

vricka kraftigt vrida q or s movement 

vrida runt rörelse q or s movement 

vrål skrik talking 

vrång sur motsträvig bad mood 

vräka handlöst kasta q or s movement 

Appendix 1 Sida 19

vräkig överdrivet pejorative 

vränga vända fel pejorative 

vrövel struntprat talking, pejorative 

Appendix 1 Sida 20

Appendix 2 

Some interjections for 9 different languages. (Swedish and English are written in ordinary orthography) 

'Pejorative' 

Swedish Icelandic English Polish Hungarian Finnish Ososo Malagasi Slovenian 

usch [Oj] oh [O] [o:] [hï] [fjO] [dE] (nasal [fOI] 

hu [Ojbjak:] pooh [Oj] [u:] [h¨h¨h] [pçO] 

E) [fEI] 

"t" [|] [u:] phew [ox] [jOj] [hui] [œh] [ts] 

blä 

ugh 

[juj] [|] 

whistle 

ha 

ooh 

[fui] 

[uœk] 

håhå (jaja) 

tut [|] 

[|]+gesture whistle 

tvi 

boo 

öh 

äh 

bah 

asch 

äsch 

isch 

uh 

urrk 

fy 

bu 

tss

'positive' (appreciation, joy etc) 


tjo(hej) [vau] oh [O] [oh] [ah] [ah] - - 

ah [jipiO] ah [m:] kissing [m:] sigh 

åh 

wow 

sound + [jip:i] 

haha(ha) 

yippee 

gesture 

hihi(hi) 

[|] 

m: 

'surprise' 


oj [ha] oh [O:] [je:] [o:] [Ou] [a] [Oho:] 

åh 

ah 

[o:] [m:] [aj] 

(positive) 

ä 

whoops 

[h¨] 

[Ox] 

åhå 

[O:] 

hoppsan 

[OI] 

(negative) 

[jOI] 

(negative) 

'other bodily and mental feeling' 


aj [aj] oh [Ox] [au] [aj] [Ou] [aj] [aIs] 

('pain') [Ou] ow [O] [jaj] 

[euo:] 

[ux] 

[au] ouch 

[juj]

- - 

shivering 

sound 

[br:] shivering 

sound 

[O] 

[Oj] 

brrh 

ugh 

[ı] 

(voiceless) 

[br:] 

[hm] - [hm] - - - 

hm 

[ahem] 

hm: 

m: 

- - 

- - - - work 

songs 

hm 

(ingressive) 

brr 

burr 

('freezing') 

hm 

('thoughtfulness') 

åhej('åhå') 

('joint 

effort') 

[nu] 

[huhuh] expiration - - 

phew [ux] [´:] 

[h¨] 

[F] 

[´f:] 

sighs 

- - 

[ajE] - - - - [euo:] 

[Oia] 

- [aha:] - [aha:] 

- [aha] 

[jOj] 

- aha 

haha 

oho 

inhalation - - - [hï] silence - [Ox] 

[at·Cu:] atishoo [apCik] [haptsi] - [itC·imi] - [a·tCix] 

[tCixah] 

puh 

pust 

('exhaustion') 

ojojoj 

vojvoj 

håhåjaja 

('complain 

-ing') 

aha 

('sudden 

insight') 

iiih 

('fear') 

atjo 

('sneeze')

- [m:] (with 

rising intonation) 

[n´mn´m] 

yummy [njamnjam] - [m:] smacking 

noise 

[m:] 

[nam] 

[namEnamE] 

swearing 

- - - - maybe 

imitation 

of animal 

mums 

namnam 

m: 

('good 

taste') 

grr 

('anger') 

[gr:] 

[Oh] (with 

coarse 

voice) 

'commands to animals' 


schas [S:] shoo [S:] [hœS] (to [S:] [Cu:] [uS:] [pC:] 

('go +gesture 

insects) 

(common 

away') [tvi:tvi:] 

[Sits] 

command 

[hOuhOu] (to 

to several 

sheep) 

animals 

"march") 

pull pull 

- - - - - - - 

(to hens) 

('come 

here') 

[gibagiba gib] 

(voiceless 

[b]) (to 

sheep) 

[putaputaput] 

(to hens)

'commands to people' 


jaja [fOs:] now now [njE] (=no) [œlœ] (verb [kai] 

aja(baja) [svei] no no 

negation) 

ajaj [sveiDr] naughty 

(mild 

naughty 

warning 

to 

children) 

sch [¨s˘] sh [tC:] [tCit] [S:] gesture [S:] - 

(be quiet) 

[kuS:] +gesture 

pst [hai] here whistling [he:] [pst] (in family [e] - 

(soft call [ü] (=du) hey 

play) name 

for 

[EnE]=you 

attention) 

pst or 

whistling 

(to 

prostitute) 

bu [ba] boo [bu:] [hu] [b¨] [dja] [a] (coarse - 

(to scare 

[bux] [bu] 

voice) 

some- 

[´] (coarse 

body) 

voice) 

vyss vyss [bi¨mbi¨m] hushaby - - - [S:] - - 

lull lull [s¨s¨˘s¨s¨˘] 

(quietly) 

(putting singing 

sb to 

sleep)

_ - 

[ü] (=du) [EwE] (not 

to adults) 

[Ei] 

ohoj 

hoho 

(louder 

call for 

attention)

Studies in Sound Symbolism

Create successful ePaper yourself

Delete template?

Save as template?