24.961 Stress-3 Stress in Windows
[1] Kager (2012): many languages restrict stress/accent to a window of two or three syllables at
the right or left edge of the word.
R. Kager / Lingua 122 (2012) 1454--1493
R. Kager / Lingua 122 (2012) 1454--1493
1461
1461
Elías-Ulloa
thesepatterns
patternsusing
using
a strictly
disyllabic
adjusting
the weight
of an
initialsyllable
closed syllable
Elías-Ulloa(2005)
(2005)analyzes
analyzes these
a strictly
disyllabic
foot,foot,
adjusting
the weight
of an initial
closed
preceding
another
to(LˈH)
(LˈH)bybymeans
means
Weight-to-Stress
and Grouping
Harmony
1990).
This weight
preceding
anotherclosed
closedsyllable
syllable to
of of
Weight-to-Stress
and Grouping
Harmony
(Prince,(Prince,
1990). This
weight
• right
edge: Aklan
binary, by
Modern
Greek,
Italian,
Spanish
ternary
adjustment
isisindependently
suffixal
allomorphy
is sensitive
to odd/even
parity treating
count, treating
adjustment
independently supported
supported by
a asuffixal
allomorphy
thatthat
is sensitive
to odd/even
parity count,
closed closed
syllables
as
light
odd
positions
(see
2009
forfor
an an
analysis).
Assuming
that the
disyllabic
foot-based
analysis analysis
is
syllables
light
ininodd
positions
(seeVaysman,
Vaysman,
2009
analysis).
Assuming
that
the disyllabic
foot-based
is
• asleft
edge:
Onati
Basque
binary,
Choguita
Raramuri
ternary
correct
and
alsothat
thatititcan
canbe
be generalized
generalized totoCapanahua,
these
languages
no longer
constitute
evidence
for a mixed
correct
and
also
Capanahua,
these
languages
no longer
constitute
evidence
for adefault.
mixed default.
•thethe
accent
location
in in
window
is determined
by
syllable
weight,
lexical encoding,
as
asspurious.
thirdS/I
S/I
language
Mathimathi
(Hercus,
1969),
quantity-sensitivity
is likelyis tolikely
be well
spurious.
ForFor
third
language
in StressTyp,
StressTyp,
Mathimathi
(Hercus,
1969),
quantity-sensitivity
to be
Goedemans(1997)
(1997)shows
shows that
that Mathimathi
stress
is is
entirely
morphologically
governed;
the appearance
of typologically
Goedemans
Mathimathi
stress
entirely
morphologically
governed;
the
appearance
of
typologically
distance
from
word
edge
uncommonquantity
quantitycriteria
criteria is
is due
processes
targeting
stemsstems
of specific
lengths.lengths.
uncommon
due to
tohistorical
historicalphonological
phonological
processes
targeting
of specific
All
three
P/U languages
listed in StressTyp
areor
Austronesian:
Javanese,
Malay,
and Ngada,
and their status in stress
•
accent
may
be
realized
by
stress
pitch
and
hence
is
a
n
abstract
quantity
All three P/U languages listed in StressTyp are Austronesian: Javanese, Malay, and Ngada, and their status in stress
typology is problematic. For Javanese, the status of word stress is controversial and descriptions are not uniform (Horne,
typology
ismaximal
problematic.
For
Javanese,
the
status
ofbword
stress
controversial
and
not
uniform2002)
(Horne,
• Poedjosoedarmo,
window
three
syllables
ut found
toisbe
at descriptions
both
edges are
(cf.
Gordon
1961;
1982;is
Ras,
1985).
In the Indonesian
spoken
bysymmetric
native speakers
of Javanese,
no systematic
1961;
Poedjosoedarmo,
1982;
Ras,
1985).
the Indonesian
spoken
speakers
of Javanese,
no the
systematic
phonetic
correlates of stress
have
been
found.InMoreover,
stress seems
notby
to native
be perceived
by Javanese
listeners:
• correlates
final syllable
canhave
be obligatorily
unstressedstress
but not
thenot
initial
syllable
phonetic
of contours
stress
been
seems
to be
perceived(Goedemans
by Javanese
the
manipulation
of pitch
of final
and found.
prefinalMoreover,
syllables failed to result
in different
perceptions
andlisteners:
van
manipulation
of pitch
contours
final the
andsituation
prefinalappears
syllables
result
in different
perceptions
(Goedemans
Zanten, 2007).
For varieties
of of
Malay,
to failed
be no to
less
complex
(van Zanten
et al., 2003;
Roosman, and van
Zanten,
2007).
varieties
Malay,
the situation
appears
to be no lessamong
complex
(van Zanten
al., 2003;
Roosman,
2006;
ZuraidahFor
et al.,
2008). of
Finally,
Ngadˈa
is most probably
miscategorized
P/U systems
since et
schwa
does not
[2]
examples
occur
in the final
according
to Arndt
(1933).
For probably
these reasons,
it seems rather
premature
to claim the
existence
2006;
Zuraidah
etsyllable
al., 2008).
Finally,
Ngadˈa
is most
miscategorized
among
P/U systems
since
schwaofdoes not
P/Uin
systems
with
mixedaccording
defaults. Into
sum,
the(1933).
evidence
forthese
opposite
defaults
in the form
of S/I
and P/U systems
rather
occur
the final
syllable
Arndt
For
reasons,
it seems
rather
premature
to claim is
the
existence of
Macedonian:
antepenult
stress
withforthe
exceptional
lexical
stress
in with
loanwords
No compelling
arguments
theevidence
existencefor
of stress
windows
P/Uweak.
systems
with mixed
defaults.remain
In sum,
opposite
defaults
innon-uniform
the form ofdefault
S/I andsystems.
P/U systems is rather
consider languages
have for
hierarchies
for intrinsic
prominence
(e.g.
a non-uniform
syllable weightdefault
scale).systems.
Weight
weak.Finally,
No compelling
argumentsthat
remain
the existence
of stress
windows
with
hierarchies
are well-known
for that
unbounded
stress languages
(such as
Kelkhar's Hindi
Kashmiri;
Prince
and Weight
Finally, vodéničar
consider
languages
have hierarchies
for intrinsic
prominence
(e.g.
syllable
weight
scale).
‘miller’
vodeníčar-ot
‘miller
def.aand
Smolensky, 1993). In case a weight scale system also imposes a stress window, the primary stress falls on the heaviest
hierarchies are well-known for unbounded stress languages (such as Kelkhar's Hindi and Kashmiri; Prince and
syllable occurring
within the window
-- a weight peak.
In case there is not a single
weight peak,
but two or three peaks
konzumátor
‘consumer’
konzumatór-i-te
‘consumers
def.’
Smolensky,
1993). In case a weight scale system also imposes a stress window, the primary stress falls on the heaviest
compete for attracting stress, the choice among these is made per default, similarly to languages with a binary weight
syllable
occurring
within Pirahã
the window
a weight
peak.
In case
not three
a single
weight
peak,in but
twostress
or three
contrast.
For example,
(Muran;-- Everett
and
Everett,
1984)there
has aisfinal
syllable
window
which
is peaks
compete
for attracting
the choice
is made per default, similarly to languages with a binary weight
predictable
according stress,
to a syllable
weight among
hierarchythese
in (15):
contrast. For example, Pirahã (Muran; Everett and Everett, 1984) has a final three syllable window in which stress is
Piraha (Everett & Everett 1984)
predictable
according
syllable
in (15): G = voiced)
(15) CVV
> GVV to
> a VV
> CVweight
> GVhierarchy
(C = voiceless;
1
(15)
2
3
4
5
CVV > GVV > VV > CV > GV
(C = voiceless; G = voiced)
Stress
falls on
1
2 the heaviest
3 of the
4 last three
5 syllables in the word. In the representations below, the stress window has
been marked by square brackets.
Stress falls on the heaviest of the last three syllables in the word. In the representations below, the stress window has
(16) a. 4-4-[3-2-1] ko.so.ii.gai.ˈtai
‘eyebrow’
been marked
by square brackets.
b. 4-[1-4-4]
ʔi.ˈtii.ʔi.si
‘fish’
(16)
c.
1-1-[1-4-3]
a. d. 4-4-[3-2-1]
1-[1-5-5]
b. e. 4-[1-4-4]
4-[1-4-3]
c. f. 1-1-[1-4-3]
4-[5-5-2]
d. g. 1-[1-5-5]
4-[4-2-4]
e. 4-[1-4-3]
Inf.case4-[5-5-2]
the heaviest
g. 4-[4-2-4]
(17)
ʔoo.hoi.ˈhoi.hi.ai
‘caterpillar’
ko.so.ii.gai.ˈtai
‘eyebrow’
pii.ˈhoa.bi.gi
‘frog’
ʔi.ˈtii.ʔi.si
‘fish’fuel’
ʔi.ˈsii.ho.ai
‘liquid
ʔoo.hoi.ˈhoi.hi.ai
‘caterpillar’
ʔo.ga.ba.ˈgai
‘want’
pii.ˈhoa.bi.gi
‘frog’
ʔa.pa.ˈbaa.si
‘square’
ʔi.ˈsii.ho.ai
‘liquid fuel’
syllable
in the word falls
outside the window, it fails to attract the primary stress:
ʔo.ga.ba.ˈgai
‘want’
ʔa.pa.ˈbaa.si
‘square’
a. 1-1-5-[4-3-4] pia.hao.gi.so.ˈai.pi ‘cooking banana’
b. 1-[2-4-3]
poo.ˈgai.hi.ai
‘banana’
In case the heaviest syllable in the word falls outside the window, it fails to attract the primary stress:
c. 1-[3-5-5]
kao.ˈai.bo.gi
‘jungle spirits’
(17)
a.
1-1-5-[4-3-4]
pia.hao.gi.so.ˈai.pi
‘cooking banana’
c.
1-[3-5-5]
kao.ˈai.bo.gi
‘jungle spirits’
The default stress position in Pirahã is rightmost. This is evidenced by windows that contain more than a single
b. syllable
1-[2-4-3]
‘banana’
heaviest
(i.e. a tie),poo.ˈgai.hi.ai
in which case the rightmost
of these is stressed.
(18)
a. [2-1-1]
bai.toi.ˈsai
‘wildcat’
b. [1-1-1]
pao.hoa.ˈhai
‘anaconda’ This is evidenced by windows that contain more than a single
The default
stress position
in Pirahã is rightmost.
5-[1-5-1]
‘pig’rightmost of these is stressed.
heaviest c.
syllable
(i.e. a tie),ba.hoi.ga.ˈtoi
in which case the
d. 1-[2-5-2]
kao.bii.ga.ˈbai
‘almost falling’
4-1-5-[3-4-3] bai.toi.ˈsai
ka.pii.ga.ii.to.ˈii
‘pencil’
(18) a. e. [2-1-1]
‘wildcat’
4-4-[3-3-3]
ʔo.ho.aa.aa.ˈaa ‘searching
intensely’
b. f. [1-1-1]
pao.hoa.ˈhai
‘anaconda’
[4-4-4]
ko.ʔo.ˈpa
‘stomach’
c. g. 5-[1-5-1]
ba.hoi.ga.ˈtoi
‘pig’
h. [5-5-5]
gi.go.ˈgi
‘what about you’
d. 1-[2-5-2]
kao.bii.ga.ˈbai
‘almost falling’
e. 4-1-5-[3-4-3] ka.pii.ga.ii.to.ˈii
‘pencil’
Kager, René. “Stress in Windows: Language Typology and Factorial Typology.” Lingua 122, no. 13
f.
4-4-[3-3-3]
ʔo.ho.aa.aa.ˈaa ‘searching intensely’
(2012): 1454–93. © Elsevier. All rights reserved. This content is excluded from our Creative
g.
[4-4-4]
ko.ʔo.ˈpa
‘stomach’
Commons license. For more information, see http://ocw.mit.edu/help/faq-fair-use/.
h. [5-5-5]
gi.go.ˈgi
‘what about you’
1
words longer than two syllables (30f).
(30)
a.
b.
c.
d.
la.ˈqa.na
me.ˈloo.ni
ˈpaa.wits.ja
ˈsip.mas.mi
‘squirrel’
‘musk melon’
‘duck’
‘silver bracelet’
d.
e.
f.
ˈtaa.vo
ˈko.ho
ˈsi.kis.ve
‘cottontail’
‘wood’
‘car’
Azkoita Basque (Hualde 1998)
Two more examples of initial two syllable window systems that impose final unstressability are Oñati Basque, which
has lexical pitch accent (Hualde, 1999), and Aranda (Australian; Strehlow, 1942; Davis, 1988), which is based on syllable
- stress
rightmost
syllable
in trisyllabic window at left edge of word except that wordweight (the
presence
of syllable
onsets).
Turningfinal
now syllable
to initial three
syllable
windows constrained by non-finality, we already discussed a mixed lexical accent
is not
stressed
plus syllable weight system, that of Terêna, in section 2. An example of a purely accentual system is Azkoitia Basque
(Hualde,
accent in words longer than two syllables:
/gizon/
‘man’
‘the man-ABS’
(31) a. gi.ˈzo.na
c. /alargun/
‘widow’
gi.zo.ˈnai
‘the man-DAT’
a.lar.ˈgu.ne
‘the widow-ABS’
gi.zo.ˈna.na
‘the man-GEN+ABS’
a.lar.ˈgu.nei
‘the widow-DAT’
gi.zo.ˈna.kin
‘the man-COM’
a.lar.ˈgu.ne.na
‘the widow-GEN+ABS’
a.lar.ˈgu.ne.kin
‘the widow-COM’
gi.zo.'nan.tsa.ko ‘the man-BEN’
‘the widow-BEN’
a.lar.'gu.nen.tsa.ko
Kager, René. “Stress in Windows: Language Typology and Factorial Typology.” Lingua 122, no. 13
b. /on/
‘good’
(2012): 1454–93. © Elsevier. All rights reserved. This content is excluded from our Creative
ABS’
ˈo.na For more information,
‘the good
Commons license.
see onehttp://ocw.mit.edu/help/faq-fair-use/.
...
[3]Observe
analytical
to define
two
and over
three-syllable
window (/ˈo.na/ ‘the good one’, */o.ˈna/).
that options
strict non-finality
has
priority
(default) non-initiality
‘‘The accentual pattern of bisyllabic forms such as óna ‘the good one’ shows that when the application of both
•extrametricality
binary feet conditions
plus extrametricality
edge
syllable,
constraints
lapses, layered
would make ofthe
whole
domainrhythmic
unaccentuable,
initial on
extrametricality
is revoked.
Expressed
in below)
terms of ranked constraints, nonfinality is ranked above noninitiality.’’ (Hualde, 1998:108)
feet (see
•
problem noted by Green & Kenstowicz (1995) for foot-binarity plus extrametricality:
That is, on an analysis that attributes default third syllable stress to initial extrametricality (rendering the initial syllable
unparsable),
and final unstressability
to final extrametricality
(unparsability),
the former
the latter. This
extrametricality
must be “revoked”
in case final
syllable is strongest
inmust
wordoverrule
in Piraha:
ranking similarly applies to Hopi seen above, on the analysis that second syllable stress is due to initial unparsability.
(‘pii.ai).ia ‘scissors’ but ?o.gi.’ai ‘big’
•
same problem for OT: Non-Finality » Align-Ft-Right fails to license stress on ?o.gi.’ai
•
Green (1995) proposes Lapse constraint: *adjacent unstressed syllables not separated by
1476
a foot boundary: *(ss)ss#, *#ss(ss)
•
R. Kager / Lingua 122 (2012) 1454--1493
pathological
pattern:
in words(Eisner
of two or1997,
three syllables,
stress is free to fall on any syllable, whereas in longer words, stress is
entails
midpoint
pathology
Hyde 2008)
restricted
to fall on a non-peripheral syllable.
(49),
a candidate
with1454--1493
a right-aligned iamb satisfies LAPSE/PARSE-2, hence
1476
R. In
Kager
/ Lingua
122 (2012)
thedetermines
Lapse constraint
prevents
lexical
from
drifting
far from
the
lexical-accent
the outcome
because
FAITHstress
-ACCENT
dominates
thetoo
default
constraints.
pathological pattern: in words of two or three syllables, stress is free to fall on any syllable, whereas in longer words, stress is
edge of the word so that it remains within the window
(49)
Free
stress
in three syllable
forms
restricted
to lexical
fall on a
non-peripheral
syllable.
In (49), a candidate with a right-aligned iamb satisfies LAPSE/PARSE-2, hence
but determines
as the word
longer,
stressFwill
be
drawn
to the the
middle
ofconstraints.
the word
lexical- accent
thegets
outcome
because
AITH-A
CCENT
dominates
default
FAITH L APSE /
FT=
ALIGN /
/
to avoid a lapse on each
edge
word
-2 of the
PARSE
ACCENT
TROCHEE
WORD-L
(49)
*!
('
)
- [TD$INLE] no language attested has this bizarre system
*!
*
( ' )
*
*!
('
)
Free lexical stress in three syllable forms
*
*
(
'
)
[TD$INLE]
FAITH FT=
ALIGN L APSE /
/
/
-2
ACCENTof LAPSE
-L since any candidate with a strictly
PARSE
TROCHEE
WORD
In longer forms of four (or more)
syllables,
violation
/PARSE-2 will be
minimal:
) one more violation of LAPSE/P*!ARSE-2 than a candidate with a non-aligned foot, an ˈinternalˈ window
aligned foot('incurs
*! that heads a
* non-peripheral foot.
emerges that
stress to fall on any syllable
( restricts
' )
In longer forms
of LAPSE/PARSE-2 will be minimal:
since any candidate with a strictly
*!
*
(' )of four (or more) syllables, violation
(50)
aligned foot incurs
( ' ) one more violation of LAPSE/PARSE-2 than a*candidate with* a non-aligned foot, an ˈinternalˈ window
emerges that restricts stress to fall on any syllable that heads a non-peripheral foot.
Peripheral unstressability in four-syllable form: default overrules peripheral accent
(50) Peripheral unstressability in four-syllable form: default overrules peripheral accent
FAITH FT=
ALIGNL APSE/
/
/
FAITH L APSE/
FT=
ALIGN/
/
ACCENT
PARSE-2
TROCHEE
WORD -L
[TD$INLE]
PARSE-2
ACCENT
TROCHEE
WORD -L
*!
*
(' )
*!
*
(' )
*!
*
*
( ' )
*!
*
*
( ' )
*
*
(' )
[TD$INLE]
*
*
(' )
*
*!
*
( ' )
*
*!
*
( ' )
*!
*
*
(' )
*!
*
*
('
)
(51) Non-peripheral
*
*
( ' ) stressability*!in four-syllable form
*
*
*!
( ' )
L APSETypology
/
AITHFactorial
FT= LinguaA122,
LIGNno.
- 13
Fand
/
/ Windows: Language
Kager, René. “Stress
in
Typology.”
PinARSE
ACCENT
TROCHEE
WORD-L
-2 This
(2012):Non-peripheral
1454–93. © Elsevier.
All rights
reserved.
content
from our Creative
(51)
stressability
four-syllable
form is excluded
*
Commons license.
see http://ocw.mit.edu/help/faq-fair-use/.
(' For
) more information, *!
L APSE
F*T=
ALIGN FAITH
*! /
* (/ ' ) /
[TD$INLE]
PARSE-2
ACCENT
TROCHEE
WORD
*!
* -L
2
extrametricality-based models with binary feet cannot account for stress windows naturally leads to a shift in the
attention to different mechanisms: anti-lapse constraints and ternary foot models, which will be discussed in sections
4.2 and 4.3, respectively.
4.2. Symmetrical anti-lapse constraints
[4] grid-only with anti-lapse constraints also faces this problem (Gordon 2002)
A second model is based on symmetrical anti-lapse constraints. Anti-lapse
(58)
a.
x
xxxx#
extended lapse
b.
x
xxx#
lapse
c.
constraints ban sequences of unstressed
1983; Selkirk, 1984). In an edge-based model,
that target lapses specifically in word
2002). Below, three configurations of grids
x
xx#
no lapse
Kager, René. “Stress in Windows: Language Typology and Factorial Typology.” Lingua 122, no. 13
(2012): 1454–93. © Elsevier. All rights reserved. This content is excluded from our Creative
Commons license. For more information, see http://ocw.mit.edu/help/faq-fair-use/.
A final three syllable window system has a high-ranked constraint banning extended lapse at the right edge, while a
Final
three-syllable
window:
» Faith-Accent
» Align-maximal window
for the symmetrical
final two syllable window
systems
bans regular
lapse at *Extended-Lapse-R
the right edge. To account
size of three syllables,Head-L
anti-lapse constraints that target extended lapses must be assumed for both edges.
The specific implementation of this mechanism will be a pure grid (PG) model, adapted from Gordon (2002). Gordonˈs
initial
three-syllable
window:
*Extended-Lapse-L
» Faith-Accent
» Alignoriginal set was simplified
by eliminating
constraints
that are
exclusively relevant
for iterative stress.
The resulting model
has the following constraint
set.
Head-R
-
1480
R. Kager
(2012)too
1454--1493
Extended-Lapse keeps
stress/ Lingua
from 122
drifting
far away from edge and then
Faith orwhile
Head-alignment
in the stress
window
extended lapse constraints,
in words of three locates
syllablesstress
long, placing
on any syllable will satisfy both anti-lapse
constraints simultaneously. This predicts window systems in which stress is free to fall on any syllable in a three syllable
word, on
the second
or third in Lapse
four syllable
words,are
but top-ranked
may only occur
onthe
the best
third window
syllable inisfive
syllable
words. This
Problem:
when
both Extended
constraints
then
one
in
window pattern is a specimen of ˈMidpoint Pathologyˈ (Eisner, 1997; Hyde, 2008:4), ˈa collection of situations where an
the middle of a word with five syllables
object is drawn toward a medial position in a formˈ.
(61)
Window reduced in words of length 4 and 5 syllables
/
'
[TD$INLE]
*!
'
'
/
'
'
'
***
**
*
*E XTENDED *E XTENDED FAITH -A CCENT A LIGN -HEAD-L APSE-L
-L APSE-R
R
*!
*!
'
*
*!
*
*!
'
/
'
*E XTENDED *E XTENDED FAITH -A CCENT A LIGN -HEAD-L APSE-R
R
-L APSE-L
/
*!
*!
*
*
*
*
****
***
**
*
Kager, René.
“Stress
in Windows:
Typology.”
Lingua
122,the
no.two
13 high-ranking anti-lapse
Strikingly,
words
longer
than fiveLanguage
syllablesTypology
regain and
theirFactorial
accentual
freedom:
since
(2012):
1454–93.
© Elsevier.
rights reserved.
This
content
excluded
from
ourjumps
Creative
CCENT
into activity again. This restores
constraints
can
no longer
both beAll
satisfied,
one must
give
way,isand
FAITH-A
Commons
morewindow:
information, see http://ocw.mit.edu/help/faq-fair-use/.
the initial
three license.
syllableFor
stress
(62)
Window- regained
in words of words
length 6a syllables
in four-syllable
lexical accent on the second syllable can survive
[TD$INLE]
/
*E XTENDED
-A CCENT
A LIGN
-HEAD
*E XTENDED
FAITH
since/ both extended
lapse
constraints
are
satisfied;
but
when
the word is five
'
syllables there is an extended
so a violation occurs
* lapse on the
*! right and
*****
-L APSE-L
-L APSE-R
R
*
**** satisfies both
- ' the candidate with accent placed
in the middle of word
*!
***
*!
**
*!
*!
But ' the window is regained when the word is longer*than five syllables since
*!
*!
'
'
-
extended
lapse constraints
*!
'
*
now among the competing candidates every word will violate one or the
other of
of the
Extended
Lapse
constraints
and hence
is patterns
a tie that
will be
The factorial typology
constraint
set for
pure grids contains
254there
window
in which
stress in five syllable
words is fixed on theresolved
middle syllable,
with
shorter
words
and/or
longer
words
obeying
a
window.
Among
these,
175 show fixed
by the lower-ranked faithfulness or alignment constraints
midpoint stress in five syllable words, while shorter and longer words have accentual freedom within a window (of two or three
syllables, in initial or final position). The most severe cases are of the following kind: an (initial or final) window of two syllables
is imposed on longer words, i.e. a window size defeated by the fixed midpoint stress in five syllable forms.6
Additional pathological window patterns arise by specific interactions of anti-lapse constraints targeting lapses of different
3 which words
sizes. For example, the high-ranked combination *LAPSE-L » *EXTENDED-LAPSE-R generates a window system in
of all lengths are subject to an initial two syllable domain, except that four syllable words have fixed stress on the antepenult.
Table (42) displays window reductions for all logically possible combinations of anti-lapse constraints for words of lengths of
two to seven syllables. The stress windows are indicated by square brackets. In each cell, the stress patterns are listed that
*!
'
*
Strikingly, words longer than five syllables regain their accentual freedom: since the two high-ranking anti-lapse
constraints can no longer both be satisfied, one must give way, and FAITH-ACCENT jumps into activity again. This restores
the initial three syllable stress window:
/
(62)
'
[TD$INLE]
*E XTENDED *E XTENDED
-L APSE-L
-L APSE-R
/
'
'
'
'
*
*
*
*!
*!
*!
'
FAITH -A CCENT A LIGN -HEADR
*!
*!
*!
*!
*!
*****
****
***
**
*
Kager, René. “Stress in Windows: Language Typology and Factorial Typology.” Lingua 122, no. 13
(2012): 1454–93. © Elsevier. All rights reserved. This content is excluded from our Creative
Commons license. For more information, see http://ocw.mit.edu/help/faq-fair-use/.
The factorial typology of the constraint set for pure grids contains 254 window patterns in which stress in five syllable
words is fixed on the middle syllable, with shorter words and/or longer words obeying a window. Among these, 175 show fixed
midpoint stress in five syllable words, while shorter and longer words have accentual freedom within a window (of two or three
Thus,
free-ranking of natural constraints predicts an unattested stress system in the typology.
syllables, in initial or final position). The most severe cases are of the following kind: an (initial or final) window of two syllables
is imposed on longer words, i.e. a window size defeated by the fixed midpoint stress in five syllable forms.6
[5] The
problem
is not just
limited
to lexical
can arise
more generally
in any
system
Additional
pathological
window
patterns
arise byaccent
specificbut
interactions
of anti-lapse
constraints
targeting
lapses of different
sizes.
For
example,
the
high-ranked
combination
*L
APSE-L » *EXTENDED-LAPSE-R generates a window system in which words
where the two opposite-edge lapse constraints dominate alignment (Stanton 2014); stress is
of all lengths are subject to an initial two syllable domain, except that four syllable words have fixed stress on the antepenult.
drawn
middle
ofwindow
word just
in case
remove
lapses
at both edges;
but inconstraints
longer words
thisof lengths of
Tableto
(42)
displays
reductions
foritallcan
logically
possible
combinations
of anti-lapse
for words
two
to
seven
syllables.
The
stress
windows
are
indicated
by
square
brackets.
In
each
cell,
the
stress
patterns
are listed that
is not possible and so the default edge orientation will reassert itself
arise under a specific ranking of a pair of anti-lapse constraints. This ranking has anti-lapse constraint in the vertical
dimension as the higher-ranked one, and the anti-lapse constraint in the horizontal dimension as the lower-ranked one:
*Lapse-L » *Lapse-R » Align-L
6
Ss
sSs Ssss Sssss
Ss
Sss sSss ssSss Ssssss
To be sure, window systems are attested in which the degree of accentual freedom is correlated with syllable number, without showing
Midpoint Pathology. For example, the pitch accent in Kyungsang Korean (Kenstowicz and Sohn, 2001) is restricted by a final two syllable window
in words*Extended-Lapse-L
of two or three syllables, whereas
accent is fixed on the »penult
in words of four or more syllables. This is not a case of Midpoint Pathology,
» *Extended-Lapse-R
Align-L
as five syllable words are not accented on the middle syllables; moreover, longer words (exceeding four syllables) do not regain accentual
freedom. Thanks to an anonymous reviewer for pointing this out.
[6] Kager’s suggested solution is to represent the two and three-syllable windows with a weakly1482
/ Lingua
122foot
(2012)
1454--1493
layered foot: a ternary foot withR.a Kager
single
binary
inside.
Nonfinality is restricted to final
(64) unstressability.
Shapes of the Weakly Layered foot
[TD$INLE]
head + adjunct adjunct + head
no adjunct
binary head, trochee
([' ] )
( [' ])
(['
])
binarywhen
head,
iamb Layered
A stress window occurs
a Weakly
([ foot
' ] is
) obligatorily
( [ aligned
' ]) with a domain
([ ' ])edge. A two syllable window
unary
head
results when the adjunct
is disallowed,
and hence,
aligned with both
([' the
] ) head is doubly
( [' ])
([' ])foot edges.
(65)
Two syllable windows: adjunct disallowed
Kager,
René. “Stress in Windows:
Language
a. final
window
b. initial
windowTypology and Factorial Typology.” Lingua 122, no. 13
rights reserved. This content is excluded from our Creative
U ([σ (2012):
ˈσ]) # 1454–93.
([ˈσ])©#Elsevier.
I
#All([ˈσ
σ])
# ([ˈσ])
Commons license. For more information, see http://ocw.mit.edu/help/faq-fair-use/.
P ([ˈσ σ]) #
S # ([σ ˈσ])
Observe that the contrast of peripheral versus non-peripheral stress may arise either by manipulating the head's
Gen
is “hard-wired”
only
allowversus
a single
“adjunct”
syllable
either athe
left orof a DPS
dominance (trochaic
versus
iambic), or itstosize
(binary
unary).
Technically,
thison
involves
ranking
constraint above a right
foot form
constraint.
edge
of foot and the internal “head” foot is restricted to two syllables (cf. foot
A three syllable window emerges when an adjunct is allowed in the weakly layered foot, so that the head may be misbinarity)
aligned with one (but
not both) foot edges. As compared to two syllable windows, the number of representational
possibilities for- non-peripheral
increases.
Forateach
stress position, four representational options
amounts tostress
binary
feet with
mostnon-peripheral
one recursion
occur, as shown below.
-
(66)
Ito & Mester (2012) propose similar restrictions on the depth of recursion in phrasal
Three syllable
windows: adjunct allowed
phonology
a.
final window
U ([ˈσ])
#
([σ ˈσ])set:
#
Kager’s
(2012)
constraint
P
A
([ˈσ σ]) #
([ˈσ σ] σ)
b.
I
S
T
initial window
# ([ˈσ])
# ([σ ˈσ])
# (σ [σ ˈσ])
([σ ˈσ] σ) #
(σ [σ ˈσ]) #
(σ [ˈσ σ]) #
(σ [ˈσ])
([ˈσ] σ)
4
# ([ˈσ σ])
# (σ [ˈσ σ])
# ([ˈσ σ] σ)
# ([σ ˈσ] σ)
# ([ˈσ] σ)
# (σ [ˈσ])
Observe that the three-way stress contrast is representationally encoded as variation in four aspects of foot form: head
dominance, head size, head alignment, and foot size. Evidently, foot form variation is not random, but under control of
constraint interaction. Crucially, a weakly layered foot model of stress windows should not be combined with extrametricality
(of the final syllable unparsability type) as this would incorrectly predict four syllable window systems. (See Everett, 1988:233
for a similar argument based on Pirahã.) However, non-finality can be assumed as an unstressability requirement.
The following constraint set will be assumed:
(67)
Constraint set for the Weakly Layered Model
a.
b.
c.
d.
e.
f.
g.
h.
i.
j.
HD-BIN
ALIGN-HD-L
ALIGN-HD-R
HD=TROCHEE
HD=IAMB
PARSE-SYL
ALIGN-WORD-L
ALIGN-WORD-R
NON-FINALITY
FAITH-ACCENT
Heads are binary under syllabic or moraic analysis.
Heads are left-aligned with feet.
Heads are right-aligned with feet.
Heads begin with strong syllable.
Heads begin with weak syllable.
Syllables are parsed by feet.
Words are left-aligned with a foot.
Words are right-aligned with a foot.
Stress must not fall on the final syllable.
A lexical accent should be realized as primary stress.
HD-BIN is the counterpart of FT-BIN in the binary model, and it is evaluated analogously. Analogously to assumptions
Kager, for
René.
Windows:
Typology
and violates
Factorial H
Typology.”
Lingua
122,
13 . The Weakly Layered
made earlier
the“Stress
binaryinfoot
model,Language
a monosyllabic
head
D=IAMB, but
not H
D=Tno.
ROCHEE
(2012): 1454–93. © Elsevier. All rights reserved. This content is excluded from our Creative
Model has no constraints requiring peripheral syllables to be non-parsed by a foot, as in the binary foot model; the only
Commons license. For more information, see http://ocw.mit.edu/help/faq-fair-use/.
constraint referring to the final syllable requires its unstressability: NON-FINALITY. To determine the side of the adjunct
with respect to the head, a pair of head-to-foot alignment constraints (67b--c) are assumed. In case both constraints are
high-ranked, the adjunct is suppressed, in which case the size of the foot equals the size of its head (i.e. maximally two
-
no lapse and clash constraints
-
word-to-foot alignment constraints keep main stress foot at an edge while separate
constraints on foot form locate the stressed syllable within this maximally trisyllabic
foot domain; in the “foot-free” model all constraints are constraints on stress since
there is no foot or grouping by hypothesis
-
the mid-point pathology does not arise in Kager’s model since no consistent ranking
of the constraint set will produce this syndrome: to get third-syllable stress in a five
syllable word with lexical accent, the constraints defining the default must dominate
Faith for lexical stress; but then they will also override lexical stress within the
window; if Faith for lexical accent is top ranked then lexical accent can surface
independent of the window
[6] Stanton (2014)
-
over-generation in the typology of grammars predicted with lapse constraints does
not automatically entail weakly layered feet or more generally metrical grouping
-
the evidence needed to motivate the ranking with both lapse constraints
undominated does not appear in the data readily available to the learner
-
thus, such grammars are theoretically possible but not reachable for learnability
reasons (a familiar argument form; cf. Lightfoot’s 1979 Degree-0 learnability)
-
to learn the ranking where Extended Lapse constraints are top-ranked, long words
(six or seven syllables) are needed and they will be much less frequent in most (nonagglutinating) languages compared to shorter words
-
for binary-lapse dominant constraints the long-word argument does not hold and the
claim is that it is difficult to infer the appropriate ranking change (aka the “credit”
problem) with the Gradual Learning Algorithm
-
other pathologies: given Kager’s 2001 constraints licensing lapses at the stress peak,
a possible ranking will shift location of main stress in odd-parity words to license a
lapse. But this is also not attested. (not clear if this bears on the grouping question)
5
Norwegian University
Science
Technology
& Utrecht
University
Non-intervention
constraints
and theand
binary-to-ternary
rhythmic
continuum
VioletaofMartínez-Paricio
& René Kager
!
!"
!
VioletaofMartínez-Paricio
& René Kager
Norwegian University
Science and Technology
& Utrecht University
Norwegian
of Science
andhead
Technology
& UtrechtatUniversity
1. Introduction
The
IL foot respects
locality,University
due to adjacency
of the
and its dependent
both levels of structure.
Ternary
rhythm
the dependents
phenomenon
being((10)0)
placed and
rhythmically
on every
syllableadjacent
or mora.
1. Introduction
Although
the top is
level
ofof
thestress
IL dactyl
anapest (0(01))
arethird
not linearly
[7] weakly-layered feet have been proposed for ternary stress rhythms (Kager & Martinez-Paricio
paradigm
caseisisthe
Cayuvava
(Key
1961,
1967),
stress
occurs
third
Introduction
Ternary
rhythm
phenomenon
stress
being
placed
rhythmically
onevery
everyto
third
syllablecounting
orfoot.
mora.
toA1.
the
head
syllable,
what
matters
for of
locality
is
thatwhere
the dependents
are on
adjacent
thesyllable
innermost
2014)
1
backward
from
the
word
end:
Ternary
rhythm
isisthe
phenomenon
of1961,
stress
being where
placed
rhythmically
on every
every
third
or mora.
A paradigm
case
Cayuvava
1967),
stress
occurs
on
syllable counting
The
IL foot
is equally
local
in its(Key
vertical
dimension:
an upper
bound
of two
layersthird
of structure.
classical grammar: dactyl (Sss) anapest (ssS) amphibrach (sSs)
A
paradigm
casetheis word
Cayuvava
backward
from
end:1 (Key 1961, 1967), where stress occurs on every third syllable counting
Since
headedness
varies
at
two levels, four IL feet‘inside
arise: of
a dactyl
(1)
3n
!"#$"#$%&$'(&$)*$+*
cow’ ((10)0), two amphibrachs ((01)0) and
1
backward
from
the
word
end:
Cayuvava (dactyl) ((Ss)s)
(0(10))
(0(01)). Importantly, all four
typesof
arecow’
typologically attested. The dactyl ((10)0)
(1) and
3nan anapest
!"#$"#$%&$'(&$)*$+*
‘inside
3n+1
ma.!ra.ha.ha.'e.i.ki
‘their
blankets’
occurs
Tripura !"#$"#$%&$'(&$)*$+*
Bangla and Cayuvava:5
‘inside of cow’
(1) in 3n
3n+1
ma.!ra.ha.ha.'e.i.ki
‘their
blankets’
3n+2 i.ki.!ta.pa.re.'re.pe.ha
‘the
water
is clean’
(8)
The
dactylma.!ra.ha.ha.'e.i.ki
((10)0) in Tripura Bangla and Cayuvava
3n+1
‘their blankets’
3n+2 i.ki.!ta.pa.re.'re.pe.ha
‘the water is clean’
3n+1
((#bi.$%&.s'&.na
3n
((()*+)*&+,-&+..#/-+$0&+10&
Another 3n+2
classicali.ki.!ta.pa.re.'re.pe.ha
example is Chugach Alutiiq ‘the
Yupik
(Leer
1985a,b,c), where stress falls on every
water
is clean’
syllable
inclassical
position
3n!1 (e.g.is"Chugach
as3n+1
on
the ma.(((ra.ha).ha).((#e.i).ki)
final
syllable
of words
of length
Another
example
Alutiiq
Yupik
(Leer
1985a,b,c),
where
stress3n+1.
falls 2on every
2, "5), as well
3n+2
((#2'.ma).l!).((s'.na)
2
Chugach
Alutiq
(amphibrach)
(sS)s)well as on
Another
classical
example
is "Chugach
Yupik
(Leersyllable
1985a,b,c),
where
stress 3n+1.
falls on
every
syllable
in
position
3n!1
(e.g.
the final
of words
of length
2, "5), as Alutiiq
(2)
3n+2 ((#o.nu).k'&.(((ro.ni).j'&"
ta.'qa.ma.lu.'ni
‘apparently
getting done’
3n
"
3n+2
i.ki.(((ta.pa).re).((#re.pe).ha)
syllable in position 3n!1 (e.g. "2, "5), as well as on the final syllable of words of length 3n+1.2
(2)
3n+2 ta.'qa.ma.lu.'ni
‘apparently
getting done’
‘he
stopped eating
Chugach 3n
Alutiiq a.'ku.tar.tu.'nir.tuq
instantiates the ‘left-branching’ amphibrach
((01)0),akutaq’
to be supported in §4.
(2)
3n+2 ta.'qa.ma.lu.'ni
‘apparently getting done’
3n
a.'ku.tar.tu.'nir.tuq
‘hehestopped
akutaq’
.)eating
is going
to hunt porpoise’
3n+1
ma.',ar.su.qu.'ta.qu.'ni
(REFL
(9)
The
left-branching
amphibrach ((01)0) in‘if
Chugach
Alutiiq
3n
a.'ku.tar.tu.'nir.tuq
‘he stopped eating akutaq’
3n+1 ma.',ar.su.qu.'ta.qu.'ni
‘if he (REFL.) is going to hunt porpoise’
3n+2 ((ta.#qa).ma).(lu.#ni)
A third example
is
Tripura
Bangla
(Das
2001),
where
falls
on alltosyllables
in position 3n+1 (e.g.
is going
hunt porpoise’
3n+1 ma.',ar.su.qu.'ta.qu.'ni
‘if he stress
(REFL.)
3n
((a.#ku).tar).((tu.#nir).tuq)
Bangla
(dactyl)
((Ss)s)
"A1, third
"4),Tripla
except
thatisfinal
syllables
are(Das
always
unstressed.
Bangla
2001),
where stress falls on all syllables in position 3n+1 (e.g.
example
Tripura
A
example
is Tripura
Banglaare
(Das
2001),
where stress falls on all syllables in position 3n+1 (e.g.
"1,third
"4),3n+1
except((ma.#3ar).su).(qu.#ta).(qu.#ni)
that
final syllables
always
unstressed.
(3)
3n+2 '-..ma.l!.!s..na
‘criticism’
"1, "4), except that final syllables are always unstressed.
(3)anapest
3n+2
'-..ma.l!.!s..na
‘criticism’
The
(0(01))
matches the /accentual
of Winnebago (Miner 1979, 1981).6 However,
3n
'o.nu.k..!ro.ni.j./
// distribution
‘imitable’
(3) that binary
3n+2 and
'-..ma.l!.!s..na
‘criticism’
note
ternary alternation both occur,
and the conditioning factors are poorly understood
3n
'o.nu.k..!ro.ni.j.//
//
‘imitable’
3n+1
'o.no.nu.!da.)o.ni.j.
‘unintelligible’
(cf. Hale &
Eagle 1980, Hayes
3nWhite'o.nu.k..!ro.ni.j./
/ 1995:350).
//
‘imitable’
3n+1 'o.no.nu.!da.)o.ni.j.
‘unintelligible’
Winnebago
(anapest)
(s(sS))
Ternarity
has
been
reported
for
a
handful
of
other
languages: Estonian (Hint 1973), possibly Finnish
3n+1
'o.no.nu.!da.)o.ni.j.
‘unintelligible’
(10)
The
anapest
(0(01)) in Winnebago
3
non-stress
languages
with
(Carlson
Sentani
(Cowan
and of
Winnebago
(Miner 1979).
Ternarity1978),
has been
reported
for 1965)
a handful
other languages:
EstonianTwo
(Hint
1973), possibly
Finnish
3n+1 (hi.(d4o.#wi)).re
‘fall in’
3
ternary
groupings
arereported
Gilbertese
&ofWinnebago
Harrison
1999)
andEstonian
Irabu Ryukyuan
(Shimoji
2009).
Ternarity
has been
for (Blevins
a1965)
handful
other languages:
(Hintnon-stress
1973),
possibly
Finnish
Two
languages
with
(Carlson
1978),
Sentani
(Cowan
and
(Miner
1979).
3n+2 (ho.(ki.#wa)).(ro.#ke)
‘swing (noun)’
3
Two non-stress
languages
with
(Carlson
1978), Sentani
(Cowan 1965) and&Winnebago
(Miner
ternary
groupings
are Gilbertese
Harrison
1999)
and1979).
Irabu
(Shimoji
2009).Local
The
dominant
metrical
analysis of(Blevins
ternary rhythm
involves
binary
feet Ryukyuan
and underparsing
(Weak
ternary3ngroupings
are Gilbertese (Blevins & Harrison
1999)
and Irabu Ryukyuan (Shimoji 2009).
‘swing
(verb intr.)’
Parsing;
WLP; (ho.(ki.#wa)).(ro.(ro.#ke))
Hayes 1995).
WLP
a parsing
mechanism
derivational
theory(Weak
(WLP-DT)
The dominant
metrical
analysis
of as
ternary
rhythm
involves in
binary
feet andmetrical
underparsing
Local
isThe
well-constrained
(by 1995).
virtue
of
locality)
and
typologically
it hastheory
been(Weak
shown
that
dominant
analysis
of ternary
rhythm
involvessuccessful,
binary
feethowever
and metrical
underparsing
Local
Parsing;
WLP;metrical
Hayes
WLP
as a parsing
mechanism
in derivational
(WLP-DT)
[8]WLP;
feet as
contexts
for
segmental
(cf. Kenstowicz
1993,
Vaysman
2008,
Davis
2009,
WLP
analyses
couched
in
Optimality
suffer
from
severe
shortcomings,
to
two
Parsing;
Hayes
WLP
asTheory
a phonology
parsing
mechanism
in derivational
metrical
theory
(WLP-DT)
is well-constrained
(by1995).
virtue
of locality)
and (WLP-OT)
typologically
successful,
however
it has
been due
shown
that
and many more)
constraint
types:couched
Gradient
and
Rhythmic
Licensing.
Asfrom
we will
explain,
these
problems
is
well-constrained
(by virtue
of locality)
and typologically
successful,
however
it has
been
shown
that
WLP
analyses
inAlignment
Optimality
Theory
(WLP-OT)
suffer
severe
shortcomings,
due to can
two
be
grouped
under
two
headings:
theoretical
(non-locality)
and
typological
anddefined
under-generation).
- couched
Davis
& Alignment
Cho
(2003)
distribution
aspiration
in
English
can
be
asdue
foot-initial
WLP
analyses
in
Optimality
Theory
(WLP-OT)
suffer
from
shortcomings,
to two
4 ofLicensing.
constraint
types:
Gradient
and
Rhythmic
As
we severe
will(over
explain,
these problems
can
o(t áto)),
(Mèdi)(t
er(ránne))an,
if
a foot-initial
adjunct
postulated:
(p and
constraint
Alignment
and isRhythmic
Licensing.
As we
will(over
explain,
problems can
be groupedtypes:
underGradient
two
headings:
theoretical
(non-locality)
typological
andthese
under-generation).
h
h
h
(T àta)(ma(góuchi))
be grouped under two
headings: theoretical (non-locality)
and typological (over and under-generation).
1
h
-
Onset in the adjunct? athómathon, authómatha, octophus
1
1
6
shown
(57),arewhere
wepreferred
provide the
tonal patterns
two arise
wordsonly
withtoheavy
These
ternaryinfeet
overall
to binary
feet; theoflatter
satisfysyllables.
exhaustivity
andwords
avoid
have
similar
length
(both
trisyllabic)
andsupporting
stress occurs
exactlybased
the same
syllables
(i.e. the
first
unary
feet (46).
Here
we are
provide
evidence
thison
proposal
on the
distribution
of tones.
and
differthe
in tonal
their patterns
tonal melodies:
(57a)
syllable
bears a reports
low tone,
Thethird).
data inStill,
(56) they
illustrates
in wordsinwith
all the
lightsecond
syllables.
Leer (1985c)
that
whereas
thecan
second
syllable
(57b)different
does not.
syllables
exhibit
one ofinthree
tonal patterns: they can bear a high tone (H), a low tone (L)
or, under
specific
circumstances,
do not words
receive
a particular
tone;
pitch
of these syllables
depends
(57)
Tone
patterns
in Chugach
Alutiiq
light andPortuguese
heavythe
syllables
Kenstowicz
& Sandalo
(2014:with
in
Brazilian
intensity/duration
measures
of
on tones
neighboring
(Leer positions:
1985c: 164).
Chugach
tone
is not
lexically
various stress
tonic
> pretonic
posttonic
medial
> final:
a. of !an.
ci.vowels
qu!a insyllables
b. Importantly,
ta!a. ta. >
!qa
specified, but! its !(s(Ss))s
distribution
is determined exclusively by metrical
structure (Leer 1985c, Hewitt
!
!
!
HL
H
H
¡H L falls on adjunct
Martinez-Paricio
(2014): Chugach-Alutiq pitch-accents;
1991, Rice 1992,
Martínez-Paricio
2013).
‘my father’
‘I’ll go out’
(56)
Tone patterns in Chugach Alutiiq words with all light syllables
A strictly stress-based account of the distribution of tones in Chugach fails to capture the tonal
!
a.
ta.!"qu. ma.words
lu. "niwith heavy syllables are
b.
pi.!"su.
qu.
ta. "qu. !ni the stress-based
(56a-d).
However,
intosyllables,
consideration,
differences
betweenwhen
these two forms,
since stress falls on taken
the same
yet their pitch patterns
! !
!
! !
! !
account
proves
inadequate,
being
unable
to
predict
the
presence
or
absence
of Llowvs.tones.
is
are different. However,
inH (57a)
(57b) This
can be
H
Lthe tonal
H difference between unstressed syllables
H L
shown
in (57), where
we provide
tonal patterns
of two
words
heavy
syllables.
These
words
‘If hewith
is going
to hunt’
‘apparently
getting
done’
straightforwardly
captured
withinthe
a model
that allows
reference
to(refl.)
recursive
metrical
structure.
In
have
length (both
areChugach
trisyllabic)
andquantity-sensitive
stress
occurs
on exactly
the reader
same syllables
(i.e.
first
thesimilar
recursion-based
analysis
of Chugach
stress, the
is referred
to the
Martínezparticular,
— and
possibly
other
c.we propose
a. "ku. ta.that
"mek
d. languages
a."ta. ka as well (see §4.3)— exploits a
and
third).(2013),
Still,
they
differ
their
melodies:
(57a)
the
syllable
a dependent
low
tone,
Paricio
which
builds
heavily
on Leer
(1985),
Rice
(1992)
and
(1993).
is important
!two
!inof
! second
!of aKager
distinction
between
types
foottonal
dependents:
(i) in
the
dependent
FtMin
andbears
(ii)What
the
H
¡
H
H
L
whereas
second
syllable
in is(57b)
does
not. a low
thethepresent
discussion
thatin
(i)Chugach
Chugach
feettone
are only
moraic
andonto
iambic
and (ii) heavy
syllables
offor
a FtNon-min.
More
specifically,
docks
an unstressed
syllable
that
‘kind of food’ (abl sg)
‘my
father’
15
an
of their
own.
These
assumptions
explain to
why
forms with
the same
is constitute
directly
byfoot
a! FtNon-min,
whereas
no two
specific
tone
is assigned
immediate
dependents
! Tone
! dominated
! iambic
!
!!!!
(57)
patterns
in
Chugach
Alutiiq
words
with
light
and
heavy
syllables
can beof
seen
from and
(56),illustrated
stressed
syllables
are
always
high,of
and
somelike
of them
H
number
syllables
pattern,
but in
different
number
morae
[ta!a.ta.!qa]
‘my father’
(59a)
ofAs
FtMin.
analysis
below
(58)
and
§3 are
thatup-stepped
forms
with(¡H).
light
a. This
!an.
ci. qu!ais stress
b.(59). Remember
ta!a.
ta. !qafrom
is up-stepped
whenfeet
it isin
preceded
by another
andand
there
is nosyllable
intervening
low
the two
syllables
contain
3n+2 (58a),
3n (3b,H 3d)
3n+1
forms,
asbetween
long as the
latterhighs
do
! IL!‘I’ll
!go out’
! leading
! to differences
and [!an.ci.qu!a]
(59b), display differences in footing,
in their tones.
H L of
H high tones to stressed syllables (i.e. footHheads)
¡His
(56c).
The an
attraction
common
in mixed
not
contain
even number
of syllables (cf. 58c). Note that this analysis
correctly
predicts
thatprosodic
in the
‘my
father’
‘I’ll
go
out’
As
illustrated
in
(59a),
the
form
[ta!a.ta.!qa]
contains
an
initial
heavy
syllable,
which
projects
its own
systemsofwith
tonenoand
1987;(58c).
Bickmore 1995; de Lacy 2002, 2006). Likewise,
absence
IL feet,
lowstress
tones(e.g.
will Goldsmith
occur in a word
up-stepping
of afoot.
Haccount
when
by another
(e.g.
H¡H)
common
in the literature
of
minimal
bimoraic
Sincepreceded
thethe
second
and thirdH
light, isthey
are
into
another
A the
strictly
stress-based
of
distribution
ofsyllables
tonesHH"
inare
Chugach
fails
to grouped
capture
the tonal
(58)
Forms with light syllables
tone
and intonation
(Goldsmith
1976;
Pierrehumbert
1980;
2002;
2004,unparsed
etc.) and
bimoraic
foot. Thethese
construction
of asince
ternary
footfalls
would
have
instead
leftGussenhoven
the yet
finaltheir
syllable
or
differences
between
two forms,
stress
on
the Yip
same
syllables,
pitch
patterns
a. ((ta. !qu)FtMin. ma)FtNon-min (lu. !ni)FtMin
be easilyHowever,
captured
by atonal
simple
rule or via
constraint
interaction.
Whatin is(57a)
puzzling
in Chugach
arecan
the
difference
between
unstressed
syllables
vs. (57b)
can be is
a different.
degenerate
foot,
! but none
! of these options
! are allowed in Chugach (Leer 1985c). Next, [!an.ci.qu!a]
the distribution of
while
some H
unstressed
aretoassigned
a Lmetrical
(e.g. thestructure.
third syllable
straightforwardly
captured
a model
that allowssyllables
reference
recursive
In
H low tones:
Lwithin
(59b) contains two heavy syllables (the first and third), each projecting a bimoraic foot. To avoid
in 56a),b.other
unstressed
syllables
do
receive
a other
particular
tone, their
pitch
an interpolation
particular,
we propose
that. qu)
Chugach
— not
and!qu)
possibly
languages
as well
(seebeing
§4.3)—
exploits a
. ni)
FtNon-min. ((ta.
FtMin
FtNon-min
degenerate ((pi."!su)
feet andFtMin
non-exhaustivity,
the third
mora
is adjoined to the preceding foot, resulting in an
between neighboring
the dependents:
fourth!syllable
in 56a,b).
If one
considers
thethe
data
in (56), a
distinction
between !twotones
types!(e.g.
of foot
(i)!the
dependent
of only
a FtMin
and (ii)
dependent
IL ternary foot. HOnce their
metrical
structure
is
considered,
we
can
understand
why
these
forms
L
H
L
non-structural
stress-based
explanation
for
the
distribution
of
tones
could
be
proposed.
Namely,
of a FtNon-min. More specifically, in Chugach a low tone only docks onto an unstressed syllable that it
displayc. different
tonal. patterns:
[(ta!aH)(ta!qa¡H)] (63a) does not present any instance of recursion and,
(a. !ku)
(ta. tone
!mek)docks
FtMin
FtMin onto every post-stress syllable that is not immediately followed
be dominated
argued
that
a low
is could
directly
by
a FtNon-min,
whereas no specific tone is assigned to immediate dependents
!
!
hence,
no
low
tone approach
occurs;
inwould
contrast,
inin[((!an
)ciL)(qu!a
)]distributions
(63b), anfrom
ILinfoot
is aligned
with
the
Hand
HRemember
a high
tone.
This
derive
the
words
with
light
syllables
ofby
FtMin.
This
analysis
is illustrated
below
(58)correct
(59).
§3
that
forms
with
lightleft
H
¡H
!tonal
!
edge of
the((a.
prosodic
and,(58a),
consequently,
the and
dependent
of the non-minimal
foot
syllables
IL
3n+2
3n (3b, 3d)
3n+1 syllable
forms, as long
as surfaces
the latterwith
do a
d.contain
!ta)feetword
.inka)
FtMin
FtNon-min
tone.an even
notlow
contain
! number!of syllables (cf. 58c). Note
27 that this analysis correctly predicts that in the
H
L
!
!!!!!!!!!!!!!!!!
absence of IL feet, no low tones will occur in a word (58c).
(59)
Forms with heavy syllables
Heavy syllables
" light
in Chugach, those containing a diphthong or a long vowel, and a CVC syllable in
(58)
Forms
with
a. (ta
!a) (ta. syllables
!qa)
b. ((!a n) ci) (qu !a)
word-initial position"
(µ !qu)
µ) (µalways
µ) attract stress and hence, slightly
((µ alter
µ) µ)the
( µbasic
µ) footing. For details on
a. ((ta.
FtMin. ma)FtNon-min (lu. !ni)FtMin
!
!
!
!
!
!
!
!
28
H
¡H
H
L
H
H
L
H
of ternary
feet could propose an alternative analysis based on WLP, in which the second
Detractors
b. ((pi."!su)
FtMin. qu)FtNon-min. ((ta. !qu)FtMin. ni)FtNon-min
!
! would! then only dock onto unfooted syllables (e.g. Hayes
Low tones
syllable in (59b) !is left unparsed.
H
L
H
L
1995:345). However, as Hayes (1995:343) acknowledges, this proposal fails to explain why a form
7
c. (a. !ku)FtMin. (ta. !mek)FtMin
like (59b) favors a parsing [(!an) ci (qu!a)] over [(!an)(ci.qu!a)], given that the uneven iamb in the latter
!
!
H
!
form is less marked
from a¡Htypological perspective. !The recursion-based
account is preferable to a
Davis, Stuart & Mi-Hui Cho. (2003). The distribution of aspirated stops and/h/in American
English and Korean: an alignment approach with typological implications. Linguistics 41.
607–652.
Kager, René (2012). Stress in windows: language typology and factorial typology. Lingua 122.
1454–1493.
Martinez-Paricio, Violetta and Rene Kager. 2013. Non-intervention constraints and the binary-toternary rhythmic continuum. Utrech U ms.
Stanton, Juliet. 2014. Learnability shapes typology: the case of the midpoint pathology. MIT ms.
8
MIT OpenCourseWare
http://ocw.mit.edu
24.961 Introduction to Phonology
Fall 2014
For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms.