1. Long-‐distance
phonotactics
as
Tier-‐based
Strictly
2-‐Local
languages
Kevin
McMullin
and
Gunnar
Ólafur
Hansson
Department
of
Linguis:cs
University
of
Bri:sh
Columbia
Annual
Mee:ngs
on
Phonology
MassachuseCs
Ins:tute
of
Technology
September
19-‐21,
2014
1
2. Formal
language
theory
• Phonotac:c
paCerns:
stringsets
modeled
as
miniature
languages
• The
Chomsky
Hierarchy
(Chomsky
1956)
• Classifica:on
for
the
computa:onal
complexity
of
a
language
based
on
the
type
of
grammar
required
to
generate
it
2
(non-‐computable
languages)
Type
0:
recursively
enumerable
languages
(recursive
languages)
Type
1:
context-‐sensi:ve
languages
Type
2:
context-‐free
languages
(finite
languages)
Yoruba
copying
(Kobele
2006)
Bambara
noun
construc:on
(Culy
1985)
Swiss
German
crossing
dependencies
(Shieber
1985)
English
center
embedding
(Chomsky
1957)
Type
3:
regular
languages
3. Phonology
is
regular
• Virtually
all
phonological
mappings
are
regular
rela:ons
(Johnson
1972;
Kaplan
and
Kay
1994)
• Any
stringsets
generated
by
these
rela:ons
are
also
regular
(Rabin
and
ScoC
1959)
• Surface
phonotac:cs
are
regular
3
…
…
Type
3:
regular
languages
AND
PHONOTACTICS
(finite
languages)
Yoruba
copying
(Kobele
2006)
Bambara
noun
construc:on
(Culy
1985)
Swiss
German
crossing
dependencies
(Shieber
1985)
Consonant
dissimilaEon
(Payne
in
press)
Unbounded
consonant
harmony
(Heinz
2010)
…
English
center
embedding
(Chomsky
1957)
ALL
PHONOLOGY
4. Subregular
hierarchy
• Not
all
regular
languages
are
aCested
as
phonotac:c
paCerns
• Instead,
consider
a
proper
subset
of
the
regular
region
• We
know
a
lot
about
the
formal
proper:es
of
some
subregular
classes
(See
e.g.,
Heinz
2010;
Heinz,
Rawal,
and
Tanner
2011;
Rogers
and
Pullum
2011)
• Can
we
define
a
demand
for
(dis)agreement
within
the
Tier-‐based
Strictly
Local
class
of
formal
languages?
4
subregular
A
class
Learnable?)
(ACested?
Regular
languages
Finite
languages
Regular
Locally
Testable
Tier-‐based
Strictly
Local
Strictly
Piecewise
Star-‐Free
Locally
Threshold
Testable
Strictly
Local
Piecewise
Testable
(Adapted
from
Heinz
et
al.
2011)
5. Research
questions
• What
are
the
limits
on
the
complexity
of
phonotac:c
paCerns?
• Long-‐distance
+
Blocking
=
Complex?
5
ACested
languages
Possible
languages
Long-‐distance
consonant
agreement
with
blocking
Long-‐distance
consonant
disagreement
Accidental
gaps
6. Research
questions
• Can
this
boundary
be
defined
as
a
class
of
formal
languages?
• All
segmental
phonotac:c
paCerns
are
members
of
the
class
of
Tier-‐based
Strictly
2-‐Local
languages
(HYPOTHESIS)
6
7. Research
questions
• How
does
such
a
class
differ
from
the
predic:ons
of
other
theore:cal
approaches
to
long-‐distance
phonotac:cs?
• Specifically,
Agreement
by
Correspondence
7
9. Tier-‐based
Strictly
2-‐Local
languages
• Intui:vely
(illustrated
with
liquid
dissimila:on)
• Co-‐occurrence
restric:ons
against
con:guous
pairs
of
segments…
• Prohibited:
*ll
and
*rr
• PermiCed:
lr
and
rl
• …where
adjacency
is
assessed
only
among
a
subset
of
segments
• Liquids:
{l,r}
⟵
this
is
the
“Tier”
• Back
to
La:n
nav-‐alis
‘naval’
reg-‐alis
‘royal’
popul-‐aris
‘popular’
lun-‐aris
‘lunar’
flor-‐alis
‘floral’
plur-‐alis
‘plural’
9
blocking
by
[r]
(Heinz,
Rawal,
and
Tanner
2011)
10. Tier-‐based
Strictly
2-‐Local
languages
10
(Heinz,
Rawal,
and
Tanner
2011)
• The
parts
• Some
alphabet
Σ (a
segment
inventory)
of
permiCed
2-‐factors
(bigrams),
labeled
S
•
• Set
is
the
set
of
prohibited
2-‐factors
• A
:er
T
that
is
a
subset
of
Σ
• Tiers
can
be
defined
with:
• Features,
natural
classes,
arbitrary
subsets
of
the
inventory
• Example
strings
for
:ers
in
a
hypothe:cal
word
‘piɹeʃaʃolu’:
• Vowel
:er
–
ieaou
piɹeʃaʃolu
• Consonant
:er
–
pɹʃʃl
piɹeʃaʃolu
• Sibilant
:er
–
ʃʃ
piɹeʃaʃolu
• Liquid
:er
–
ɹl
piɹeʃaʃolu
• {ʃ,p,i,u}
:er
–
piʃʃu
piɹeʃaʃolu
S
11. Locality/blocking
as
a
consequence
of
the
T
• Illustra:ve
harmony
languages
• Σ =
{s,ʃ,p,t,a}
•
=
{*sʃ,
*ʃs}
*[+ant][-‐ant]
• ACested
proper:es
of
long-‐distance
phonotac:cs
arise
as
a
byproduct
of
the
segments
included
in
“T”
• These
languages
will
ban
sequences
of
sibilants
in
some
way
•
“Locality”
will
depend
on
the
contents
of
Τ
• i.e.
Which
segments
are
considered
when
assessing
adjacency?
11
[+consonantal]
|
[+coronal
–sonorant]
|
[+strident]
S
{s,ʃ,p,t}
{s,ʃ,t}
{s,ʃ}
Τ
=
{??}
12. Attested
properties
of
locality
• Unbounded
sibilant
harmony
• Σ =
{s,ʃ,p,t,a}
•
=
{*sʃ,
*ʃs}
• Τ
=
{s,ʃ}
(all
sibilants)
• Members
of
L:
{sasa,
satasa,
sasasasa,
ʃaʃa,
ʃataʃa,
ʃapapaʃa,
…}
• Not
in
L:
{saʃa,
sataʃa,
sasaʃaʃa,
ʃasa,
ʃatasa,
ʃapapasa,
…}
• Example:
Aari
(Omo:c;
Hayward
1990)
baʔ-‐s-‐e
‘bring-‐PFV-‐3SG’
ʧʼa̤ːq-‐ʃ-‐it
‘swear-‐PFV-‐1SG’
*ʧʼa̤ːq-‐s-‐it
ʃed-‐er-‐ʃ-‐it
‘see-‐PASS-‐PFV-‐1SG’
*ʃed-‐er-‐s-‐it
ʒa̤ːg-‐er-‐ʃ-‐e
‘sew-‐PASS-‐PFV-‐3SG’
*ʒa̤ːg-‐er-‐s-‐e
• Tier
contains
only
the
segments
included
in
• Σ =
{b,a,ʔ,s,e,u,ʃ,i,t,ʧʼ,a̤ː,q,d,r,ʒ,g}
12
S
S
[+cons]
|
[+cor
–son]
|
[+strid]
13. Attested
properties
of
locality
• Transvocalic
sibilant
harmony
• Σ =
{s,ʃ,p,t,a}
•
=
{*sʃ,
*ʃs}
• Τ
=
{s,ʃ,p,t}
(all
consonants;
assumes
CV
syllables)
• Members
of
L:
{sasa,
satasa,
sapaʃa,
ʃaʃa,
sataʃa,
sasapaʃaʃa,
…}
• Not
in
L:
{saʃa,
taʃasa,
ʃasaʃa,
…}
• Example:
Koorete/Koyra
(Omo:c;
Hayward
1982)
baːr-‐us-‐
‘lie.down-‐CAUS’
goːʧ-‐uʃ-‐
‘pull-‐CAUS’
ʃod-‐us-‐
‘uproot-‐CAUS’
• Tier
contains
everything
in
,
and
several
addi:onal
segments
• Σ =
{b,aː,r,u,s,g,oː,ʧ,ʃ,o,d}
13
S
S
[+cons]
|
[+cor
–son]
|
[+strid]
14. Attested
properties
of
locality
• Sibilant
harmony
with
blocking
• Σ =
{s,ʃ,p,t,a}
•
=
{*sʃ,
*ʃs}
• Τ
=
{s,ʃ,t}
(all
coronal
obstruents)
• Members
of
L:
{sasa,
sapasa,
ʃaʃa,
ʃataʃa,
sataʃa,
ʃatapasa…}
• Not
in
L:
{saʃa,
sapaʃa,
sasaʃaʃa,
taʃasa,
ʃapapasa,
…}
• Example:
Slovenian
(Jurgec
2011)
ʃpi-‐ʃ
‘(you)
sleep’
*spi-‐ʃ
ʒa-‐klɔn-‐iʃʧe
‘bomb
shelter’
*za-‐klɔn-‐iʃʧe
na-‐sit-‐iʃ
‘(you)
feed’
*na-‐ʃit-‐iʃ
zida-‐ʃ
‘(you)
build’
*ʒida-‐ʃ
• Tier
contains
everything
in
,
and
a
few
addi:onal
segments
• Σ =
{ʃ,p,i,ʒ,a,k,l,ɔ,n,ʧ,e,n,s,t,z,d}
14
S
[+cons]
|
[+cor
–son]
|
[+strid]
S
15. Summary
of
TSL2
patterns
15
• Long-‐distance
consonant
agreement
(harmony)
• Long-‐distance
consonant
disagreement
(dissimila:on)
• Unbounded
dependencies
• Transvocalic
dependencies
• Dependencies
with
blocking
• No
other
variants
of
locality
(e.g.
“separated
by
up
to
two
vowels”)
• Transvocalic
dependencies
can
be
thought
of
as
blocking
by
any
consonant
• The
rela:onship
between
T
and
determines
which
of
these
proper:es
the
language
exhibits
These
proper:es
all
come
for
free
by
allowing
varia:ons
in
T
S
16. Agreement
by
Correspondence
• Faithfulness
constraints
• IDENT[ant]-‐IO
• Correspondence
constraints
• Require
certain
sets
of
segments
to
be
in
correspondence
• Correspondence
is
facilitated
by
trigger-‐target
similarity
• CORR[X⟷Y],
CORR[G],
or
CORR[αG]
(e.g.,
CORR[+strid])
• ‘CC·∙Limiter’
constraints
(BenneC
2013)
that
impose
restric:ons
on
correspondents
• IDENT-‐CC
constraints
• Require
correspondents
to
agree
in
some
feature
• IDENT[F]-‐CC
(e.g.
IDENT[-‐ant]-‐CC)
16
(Hansson
2001,
2010;
Rose
and
Walker
2004)
17. Locality
in
Agreement
by
Correspondence
• Locality
constraints
(CC·∙Limiters)
• PROXIMITY
(Rose
and
Walker
2004)
• If
X
and
Y
are
in
correspondence…
• …then
X
and
Y
must
be
in
adjacent
syllables
• cf.
Transvocalic
(Hansson
2010)
• Makes
some
strange
predic:ons
• Harmony
that
fails
to
apply
to
a
sequence
of
targets
• /ʃasa/➝[ʃaʃa],
but
/ʃasasa/➝[ʃasasa]
• Harmony
that
applies
only
when
target
segment
is
preceded
by
an
even
number
of
segments
in
the
harmonizing
class
• /ʃasa/➝[ʃaʃa],
/sasaʃasa/➝[sasaʃaʃa],
/sasasasaʃasa/➝[sasasasaʃaʃa],
but
• /saʃasa/➝[saʃasa],
/sasasaʃasa/➝[sasasaʃasa]
17
18. Pathological
ABC
harmony
systems
• Harmony
that
fails
to
apply
to
a
sequence
of
targets
• This
is
not
a
TSL2
paCern
• A
word
with
‘ʃs’
on
the
sibilant
:er
is
only
gramma:cal
if
‘ss’
is
also
present
on
the
sibilant
:er
• Belongs
to
the
(Tier-‐based)
‘Locally
Testable’
class
at
best
18
!
/paʃasasa/! PROXIMITY! CORR!
[+strid]!
IDENT!
[:ant]:
CLCR!
IDENT!
[ant]:IO!
! a.!paʃasasa! ! ***!! ! !
! b.!paʃxasxasxa! *!! ! **! !
! c.!paʃxaʃxaʃxa! *!! ! ! **!
! d.!paʃxaʃxasa! ! **! ! *!!
☞! e.!paʃasxasxa! ! **! ! !
!
19. Pathological
ABC
harmony
systems
• Harmony
that
does
not
apply
when
the
trigger
is
preceded
by
an
odd
number
of
segments
in
the
harmonizing
class…
19
!
/pasaʃasa/! PROXIMITY! CORR!
[+strid]!
IDENT!
[:ant]:
CLCR!
IDENT!
[ant]:IO!
! a.!pasaʃasa! ! ***!! ! !
! b.!pasxaʃxasxa! *!! ! **! !
! c.!pasxaʃxaʃxa! *!! ! ! **!
☞! d.!pasxaʃxasa! ! **! ! !
! e.!pasaʃxaʃxa! ! **! ! *!!
!
20. Pathological
ABC
harmony
systems
• Harmony
that
does
not
apply
when
the
trigger
is
preceded
by
an
odd
number
of
segments
in
the
harmonizing
class…
• …but
does
when
there
is
an
even
number
• Such
a
paCern
is
computa:onally
complex
• Gramma:cality
is
determined
by
(unlimited)
coun:ng
of
sibilants
20
!
/sasaʃasa/! PROXIMITY! CORR!
[+strid]!
IDENT!
[9ant]9
CLCR!
IDENT!
[ant]9IO!
! a.!sasaʃasa! ! *****!*! ! !
! b.!sxasxaʃxasxa! *!**! ! *! !
! c.!sxasxaʃxaʃxa! *!**! ! ! *!
! d.!sxasxaʃxasa! *!! ***! ! !
☞! e.!sxasxaʃyaʃya! ! ****! ! *!
!
21. Another
approach
to
locality
in
ABC
• PROXIMITY
(Rose
and
Walker
2004)
• If
X
and
Y
are
in
correspondence…
• …then
X
and
Y
must
be
in
adjacent
syllables
• CC·∙SYLLADJ
(BenneC
2013)
• If
X
and
Y
are
in
correspondence…
• …and
there
is
no
segment
Z
in
the
output
such
that
X
<
Z
<
Y
and
X,Z
are
in
correspondence…
• …then
X
and
Y
must
be
in
adjacent
syllables
• This
effec:vely
undoes
the
transi:ve
property
of
the
correspondence
rela:on…
• Xx…Zx…Yx
would
not
violate
CC·∙SYLLADJ
(more
like
:ers)
• …but
s:ll
makes
problema:c
predic:ons
when
using
ABC
constraints
to
derive
dissimila:on
21
22. Problems
with
dissimilation
in
ABC
• Surface
Correspondence
Theory
of
Dissimila:on
(BenneC
2013)
• Effec:vely,
dissimila:on
is
the
avoidance
of
correspondence
• “Mismatch
predic:on”
for
typology
of
harmony
vs.
dissimila:on
• Dissimila:on
should
occur
in
exactly
the
opposite
contexts
• Predicts
paCerns
with
liCle
empirical
support…
• Beyond-‐transvocalic
dissimila:on
• …and
raises
complica:ons
for
descrip:vely
simple
paCerns
• Transvocalic
dissimila:on
• Dissimila:on
of
iden:cal
segments
22
25. Beyond-‐transvocalic
dissimilation
• Beyond-‐transvocalic
dissimila:on
is
not
a
TSL2
paCern
• Gramma:cality
cannot
be
determined
based
only
on
the
:er-‐based
2-‐factors
in
the
word
• Transvocalic:
CVpVbV,
CVbVpV
are
gramma:cal
• Beyond-‐:
*CVpVCVbV,
*CVbVCVpV
are
not
• This
is
a
(Tier-‐based)
Locally
Testable
language
• Gramma:cality
is
assessed
based
on
the
en:re
set
of
2-‐factors
on
a
:er
of
consonants
• No
words
with
both
‘Cp’
and
‘Cb’
25
26. Summary
of
argument
• Comparison
of
two
approaches
to
defining
the
space
of
possible
long-‐distance
phonotac:c
paCerns:
• as
Tier-‐based
Strictly
2-‐Local
languages
• as
a
factorial
typology
of
Agreement
by
Correspondence
constraints
• The
‘formally
grounded’
TSL2
approach
seems
to
do
a
beCer
job
of
capturing
the
aCested
typology
than
ABC
• except
when
the
CC·∙Limiter
constraints
are
made
to
treat
correspondence
sets
more
like
:ers
• Even
then,
other
pathologies
s:ll
remain
(Hansson
2014)
26
27. Further
issues
-‐
Similarity
• In
principle,
any
arbitrary
class
of
segments
can
be
a
:er
T
• Does
not
entail
any
no:on
of
similarity
• The
vast
majority
of
long-‐distance
dependencies
seem
to
be
facilitated
by
some
sort
of
similarity
• Though
some
seem
to
involve
unnatural
classes
• La:n
liquid
dissimila:on
can
also
be
blocked
by
labials
and
velars
(Roberts
2014)
• Tier
of
coronal
non-‐nasal
sonorants,
labials,
and
velars
• Poten:al
solu:ons
• Hard
restric:ons
on
the
sets
of
segments
that
can
be
a
:er
• Learning
bias
influencing
how
the
learner
navigates
the
space
of
possible
:ers
27
28. Further
issues
-‐
Multiple
TSL2
patterns
• Transvocalic
harmony
+
beyond-‐transvocalic
dissimila:on
• Transvocalic:
lVl,
rVr
are
gramma:cal,
*lVr,
*rVl
are
not
• Beyond-‐:
lV…Vr,
rV…Vl
are
gramma:cal
*lV…Vl,
*rV…Vr
are
not
• One
possible
case
is
aCested
(Sundanese;
BenneC
2013)
• Details
are
complicated
(infixa:on,
syllable
structure,
etc)
• Would
this
be
a
problem?
• The
surface
phonotac:cs
would
not
be
TSL2
• …but
could
be
modeled
as
two
TSL2
paCerns
(constraints),
where
harmony
overrides
dissimila:on
in
transvocalic
contexts
28
29. Acknowledgements
• Jeff
Heinz
and
Adam
Jardine
(University
of
Delaware)
• Douglas
Pulleyblank
and
Blake
Allen
(UBC)
• Audience
at
the
ABC
Conference,
UC
Berkeley
2014
• SSHRC
Insight
Grant
435–2013–0455:
“Long-‐distance
phonotac:cs:
learning
bias,
change,
and
typology”
(PI:
Gunnar
Ólafur
Hansson)
29
30. References
30
BenneC,
William
G.
2013.
Dissimila*on,
consonant
harmony,
and
surface
correspondence.
Doctoral
disserta:on,
Rutgers
University.
Chomsky,
Noam.
1956.
Three
models
for
the
descrip:on
of
language.
IRE
Transac*ons
on
Informa*on
Theory
2:
113–24.
Chomsky,
Noam.
1957.
Syntac*c
Structures.
Mouton
and
Co.,
The
Hague
Culy,
Christopher.
1985.
The
complexity
of
the
vocabulary
of
Bambara.
Linguis*cs
and
Philosophy
8:
345–51.
Hansson,
Gunnar
Ólafur.
2001.
Theore*cal
and
Typological
Issues
in
Consonant
Harmony.
Ph.D.
thesis,
University
of
California,
Berkeley.
Hansson,
Gunnar
Ólafur.
2010.
Consonant
harmony:
long-‐distance
interac*on
in
phonology.
Berkeley,
CA:
University
of
California
Press.
[Revision
of
2001
disserta:on,
U.C.
Berkeley.]
Hayward,
Richard
J.
1982.
Notes
on
the
Koyra
language.
Afrika
und
Übersee
65:
211–268.
Hayward,
Richard
J.
1990.
Notes
on
the
Aari
language.
Omo*c
language
studies,
ed.
by
Richard
J.
Hayward,
pp.
425–493.
London:
School
of
Oriental
and
African
Studies,
University
of
London.
Heinz,
Jeffrey.
2010.
Learning
long-‐distance
phonotac:cs.
Linguis*c
Inquiry
41(4):
623–661.
Heinz,
Jeffrey,
Chetan
Rawal
and
Herbert
G.
Tanner.
2011.
Tier-‐based
strictly
local
constraints
for
phonology.
Proceedings
of
the
49th
Annual
Mee*ng
of
the
Associa*on
for
Computa*onal
Linguis*cs,
pp.
58–64.
Associa:on
for
Computa:onal
Linguis:cs.
Jensen,
John.
1974.
Variables
in
phonology.
Language
50:
675–86.
Jurgec,
Peter.
2011.
Feature
Spreading
2.0:
A
Unified
Theory
of
Assimila*on.
Ph.D.
thesis,
University
of
Tromsø,
Norway.
Kaplan,
Ronald
M.,
and
Mar:n
Kay.
1994.
Regular
models
of
phonological
rule
systems.
Computa*onal
Linguis*cs
20:
331–78.
Kobele,
Gregory
M.
2006.
Genera*ng
Copies:
An
Investa*on
into
Structural
Iden*ty
in
Language
and
Grammar.
Ph.D.
thesis,
University
of
California,
Los
Angeles.
31. References
31
McNaughton,
Robert,
and
Seymour
Papert.
1971.
Counter-‐free
Automata.
Cambridge,
MA:
MIT
Press.
Odden,
David.
1994.
Adjacency
parameters
in
phonology.
Language
70(2):
289–330.
Payne,
Amanda.
2014.
Dissimila:on
as
a
subsequen:al
process.
To
appear
in
Proceedings
of
the
44th
Mee*ng
of
the
North
East
Linguis*c
Society.
Storrs,
CT:
University
of
Connec:cut.
Rabin,
Michael
O.,
and
Dana
ScoC.
1959.
Finite
automata
and
their
decision
problems.
IBM
Journal
of
Research
and
Development
3:
114–25.
Roberts,
Philip
J.
2014.
The
curious
incident
of
the
La:n
liquids.
Poster
presented
at
the
Agreement
by
Correspondence
Conference.
May
18-‐19,
2014,
University
of
California,
Berkeley.
Rogers,
James,
and
Geoffrey
K.
Pullum.
2011.
Aural
paCern
recogni:on
experiments
and
the
subregular
hierarchy.
Journal
of
Logic,
Language,
and
Inference
20:
329–42.
Rose,
Sharon
and
Rachel
Walker.
2004.
A
typology
of
consonant
agreement
as
correspondence.
Language
80(3):
475–531.
Shieber,
Stuart.
1985.
Evidence
against
the
context-‐freeness
of
natural
language.
Linguis*cs
and
Philosophy
8:
333-‐43.