A unified account of Turkish exceptional stress assigning

A unified account of Turkish exceptional stress
assigning suffixes
(Draft)
Deniz Özyıldız
Introduction
Data and terminology A subset of Turkish words feature fixed final stress and this
is generally assumed to be the ‘regular’ stress pattern of the language (Sezer 1983).
It affects both mono- and polymorphemic words, as illustrated in (1):
(1) a. ta.vá, ‘pan’
b. ta.va.cı́, ‘pan salesman’
c. ta.va.cı.sı́, ‘his pan salesman’
However, there are suffixes that cause stress to be assigned to non-final syllables
in two apparently distinct ways. In (2), a monosyllabic suffix causes stress to fall on
the syllable that immediately precedes it:
(2)
ta.va.cı.sı́.mi, ‘is it his pan salesman?’
In (3), a disyllabic suffix causes stress to fall on its initial syllable.
(3)
ya.pá.rak, ‘by doing’
Suffixes of the former kind are called prestressing and those of the latter stressed
suffixes (Inkelas 1994). In the following, I refer to both types as stress disrupting
suffixes and to those that do not affect the position of stress as stress neutral.
Proposal I will argue here that stress disrupting suffixes form a single natural class.
They are at most two syllables long and primary stress falls on the second to last
syllable counting from the right edge of the suffix. If Turkish feet are assumed to
be syllabic trochees and that their right edge aligns with the right edge of a stress
disrupting suffix, this pattern can be naturally captured.
1
I will implement an analysis based on McCarthy and Pruitt’s (2013) proposal
that stress disrupting suffixes are marked with an unpronounced diacritic, to which
specific constraints are allowed to refer. The system I propose, however, is simpler
than theirs and than that of Inkelas’s (1994). The two following properties will be
captured:
1. In the presence of a stress disrupting suffix and neutral suffixes, it is stress
assigned by the former that surfaces.
2. In the presence of two or more stress disrupting suffixes, it is the one assigned
by the leftmost that surfaces.
Finally, I will show that this analysis also extends to capture a second type of
exceptional stress pattern of the language: initial stress in adjectives that have undergone emphatic reduplication.
Omissions I will not mention Sezer stress (Sezer 1983), nor any form of secondary
stress. The latter omission can be formally implemented by assuming that some
undominated constraint (e.g. C ULMINATIVITY, Zuraw 2006) ensures that forms have
exactly one primary stress. I will not show any form that violates this constraint, nor
any form that is entirely unparsed.
Word final stress
Don’t stress the final syllable! Word final stress can be captured with the constraint F INAL S TR, which assigns one violation mark to a candidate that does not
have final stress. The optimal candidate according to this constraint might violate
N ON F INALITY, which assigns a violation mark to a candidate whose last syllable is
parsed.
Table 1: Word final stress
/ta.va.cı.sı/
→ a) ta.va.cı.(sı́)
→ b) ta.va.(cı.sı́)
c) ta.va.(cı́).sı
d) (tá.va).cı.sı
F INAL S TR
0
0
1W
1W
N ON F INALITY
1
1
0L
0L
In Table 1, the candidates a) with a final unary foot and b) with a final iamb
tie on both constraints. We do not yet have any motivation to distinguish between
the two parses, although stress disrupting suffixes will provide a reason to prefer a)
over b). Candidates c) and d) both lose because their final syllable is not stressed.
2
Comparing the two of them illustrates the fact that F INAL S TR is not designed as a
gradient constraint.
That N ON F INALITY is an active constraint in the language receives motivation
from the observation that the subset of words that are stressed according to Sezer’s
generalization are never stressed on their final syllable. I will not pursue this any
further.
’ R, W D , R), which assigns one vioNow, a gradient constraint such as A LIGN ( σ,
lation mark per intervening syllable between the right edge of the syllable bearing
primary stress and the right edge of the word, would successfully capture word final
stress. But F INAL S TR is given as a binary constraint because there is no evidence that
the specific position of primary stress relative to the right edge of the word makes a
difference. What matters is whether the last syllable is stressed or not.
Anticipation of a difficulty Regardless of the type of constraint used to derive
word final stress, a difficulty will arise for the analysis of stress disrupting suffixes.
Indeed, I will argue that the position of primary stress is calculated by counting left’ R, W D , R) will
wards starting from the right edge of any such suffix, but A LIGN ( σ,
exert a pressure to align stress towards the right. This effect will not be visible with
N ON F INALITY unless the stress disrupting suffix happens to be the final syllable of
the word. This opposing force will then have to be countered by allowing the chosen
constraint to be dominated.
Stress disrupting suffixes
An incorrect prediction
The analysis sketched out above makes an incorrect prediction for forms that contain stress disrupting suffixes. This is because, at this stage, forms with final stress
are more harmonic than forms with non-final stress, and stress disrupting suffixes
essentially cause stress to be non-final. This is shown in Table 2. The intended winner in a) loses to candidate b), with final stress.
Table 2: F INAL S TR applies to all forms of the language
/ta.va.cı.mı/
→ a) ta.va.(cı́.mı)
(!) b) ta.va.(cı.mı́)
F INAL S TR
1
0L
N ON F INALITY
1
1
This means that phonology must be made to see whether the input contains a
stress disrupting suffix or not. And, given that the position of primary stress is calculated relative to the position of the suffix, and not relative to some absolute edge
such as a root or a word boundary, phonology must make reference to such that
position.
3
Generalization
The two observations guiding the analysis to follow are that:
1. all prestressing suffixes are monosyllabic, and that
2. all stressed suffixes are stressed on their initial syllable and are at most disyllabic.
These observations undermine the claim that being a prestressing or a stressed
suffix are unrelated properties (Inkelas 1994). It suggests, on the contrary, that although these properties are descriptively adequate in highlighting a difference, the
two types of stress disrupting suffixes form a natural class. The definition of this
class obeys the generalization given in (4):
(4) Generalization: A stress disrupting suffix causes primary stress to fall on the
second syllable counting from the right edge of the suffix.
Indeed, for suffixes that obey this generalization, if the suffix is monosyllabic,
stress will be assigned to the syllable immediately preceding it, and if it is disyllabic,
it will be stressed on its initial syllable.
Motivating the generalization This claim bears upon what stress disrupting suffixes are attested in the language. In principle, then, it can be falsified by providing
a suffix that does not follow the generalization. I will examine two such counterexamples: one prestressing suffix that is disyllabic and another that is trisyllabic. (I
did not find any candidate for non-initially stressed polysyllabic suffixes, another
potential type of counterexample.)
These counterexamples are visualized in the second and third columns of the first
row of Table 3, an overview of the types of stress disrupting suffixes listed in Inkelas
(1994).
Table 3: Types of stress disrupting suffixes according to Inkelas (1994)
Number of σ:
Prestressing
Stressed
1
ko.mı́k.çe
funny- SUF
somewhat funnily
Unattested
2
ak.şám.le.yin
evening- SUF
during the evening
ge.lé.rek
come- SUF
by coming
3
gü.lér.ce.si.ne
laugh- SUF
as if he were laughing
Unattested
The three syllable ‘cA.sI.nA’ can be morphologically decomposed into the prestressing ‘-cA’ and two stress neutral suffixes, ‘-(s)I(n)’ and ‘(y)A’. The first suffix
in independent uses derives adverbs and adjectives, the second one is a possessive
4
marker and the last one marks the dative. That ‘-cA’ is prestressing is visible in the
first cell of Table 3. Given that stress neutral suffixes do not affect stress, the three
suffix sequence maintains the stress assigned by the stress disrupting suffix.
This claim can be challenged by the observation that this specific sequence of
suffixes has an idiosyncratic, non-compositional meaning and, thus, that it is a good
candidate for a polysyllabic prestressing suffix. Such an objection, however, would
be convincing if a polysyllabic suffix were found whose initial syllable did not correspond to an otherwise prestressing suffix. From the perspective of morphological
and phonological analyses the suffix ‘cA.sI.nA’ is transparent.
Turning now to the disyllabic prestressing suffix ‘-le.yin’. This suffix cannot be
decomposed in the same way as the previous one. However, it is not productive and
it occurs in only about the words listed in (5):
(5) sabahleyin, akşamleyin, ikindileyin, geceleyin
in the morning, in the evening, at tea time, at night
I will thus assume that forms containing this suffix are exceptionally stressed in
the sense that their stress pattern is learned.
I do not know whether all counterexamples can be dismissed by appealing to one
of these methods. But they provide a test that determines whether a counterexample
is truly challenging for the present analysis or not.
Analysis
A diacritic in the lexicon I assume that stress disrupting suffixes are marked by an
unpronounced diacritic. (Perhaps one could argue that the way these suffixes affect
stress is the way that the diacritic gets pronounced.) Example (6) provides a sample
lexicon corresponding to this assumption. Stress neutral suffixes are unmarked, (6a),
while both monosyllabic and disyllabic stress disrupting suffixes, in (6b) and (6c),
carry the diacritic, noted by the subscript d.
(6) a.
dI
b.
mId
c.
ArAkd
The grammar will then be able to derive non-final stress patterns by including
constraints that refer to the diacritic, thus applying only to forms that bear it.
The generalization above opens the possibility of a unified treatment for suffixes
like (6b) and (6c). Thus, I will not assume that there are differences in the way the
two types of suffixes are marked by the diacritic. One could imagine that there are
two distinct diacritics, for instance, or that a single diacritic may be placed on different segments. Such alternatives, it seems, undermine the perspective of a unified
treatment.
5
Aligning foot and suffix The right edge of a foot with the right edge of a d bearing
suffix can be aligned with the following constraint.
E XCEPTIONAL S UFFIX: Assign one violation mark per intervening syllable
between the right edge of a foot and the right edge of a d bearing suffix.
“A LIGN (F T, R, Xd , R)”
Given that in the presence of a stress disrupting suffix, stress assigned by it surfaces rather than final stress, this constraint must be ranked higher than F INAL S TR.
The constraint will rule out all candidates where the right edge of a foot is not aligned
with the right edge of a stress disrupting suffix. This is illustrated in the following
tableau by comparing candidates a), b) and c) to candidates d) and e).
Table 4: Feet are aligned with stress disrupting suffixes
/ta.va.cı.mıd .sın/
→ a) ta.va.(cı́.mıd ).sın
b) ta.va.(cı.mı́d ).sın
c) ta.va.cı.(mı́d ).sın
d) ta.va.cı.(mı́d .sın)
e) ta.va.cı.(mıd .sı́n)
E XC S TR
0
0
0
1W
1W
F INAL S TR
1
1
1
0L
0L
N ON F IN
0
0
0
1W
1W
But candidates a), b) and c) tie on all the constraints. Given that candidate a) is
the correct output, the binary syllabic trochee must be enforced.
Enforcing the trochee: a constraint ranking paradox
ducing F T B IN and T ROCHEE, defined as follows:
This can be done by intro-
F T B IN: Assign one violation mark per non-binary foot.
T ROCHEE: Assign one violation mark per right-headed foot.
F T B IN will rule out candidate b) where a unary foot includes the final syllable of
a stress disrupting suffix and T ROCHEE candidate c) where an iamb is aligned with
the right edge of the suffix.
Table 5: Enforcing the trochee
/ta.va.cı.mıd .sın/
→ a) ta.va.(cı́.mıd ).sın
b) ta.va.cı.(mı́d ).sın
c) ta.va.(cı.mı́d ).sın
E XC S TR
0
0
0
F T B IN
0
1W
0
T ROCHEE
0
0
1W
F INAL S TR
1
1
1
Now, although this move predicts the correct result for the set of stress disrupting suffixes, final stress is no longer predicted for forms that do not contain stress
disrupting suffixes. This is illustrated in Table 6.
6
Table 6: A constraint ranking paradox
/ta.va.cı/
→ a) ta.va.(cı́)
→ b) ta.(va.cı́)
c) ta.(vá.cı)
E XC S TR
0
0
0
F T B IN
1
0
0L
T ROCHEE
0
1
0L
F INAL S TR
0
0
1W
We have a constraint ranking paradox. To predict the correct form for stress disrupting suffixes, the trochee must be enforced by constraints that rank higher than
F INAL S TR (Table 5). But to derive final stress, the constraints that enforce the trochee
must be dominated by F INAL S TR (Table 6).
One way out is to keep the order F INAL S TR T ROCHEE and to introduce a
constraint, specific to diacritic marked forms, that dominates F INAL S TR.
This move can further be motivated by the observation that F INAL S TR must be
dominated by such a constraint in any case, because otherwise a difficulty arises
if a stress disrupting suffix happens to end a word. This is the difficulty that had
mentioned in the section on deriving word final stress.
Table 7: F INAL S TR
/ta.va.cı.mıd /
→ a) ta.va.(cı́.mıd )
b) ta.va.(cı.mı́d )
c) ta.va.cı.(mı́d )
E XC S TR
0
0
0
F INAL S TR
1
0L
0L
The following constraint is based on the generalization that no stress disrupting
suffix is stressed on its final syllable.
*F INAL S TR(d): Assign one violation mark per d-bearing suffix that is
stressed on its final syllable.
These constraints derive the correct pattern, as illustrated in Tables 8 and 9. We
do not need to include F T B IN any more, although it would need to be ordered below
*F INAL S TR(d).
Table 8: Deriving non-final stress
/ta.va.cı.mıd /
→ a) ta.va.(cı́.mıd )
b) ta.va.(cı.mı́d )
c) ta.va.cı.(mı́d )
E XC S TR
0
0
0
*F INAL S TR(d)
0
1W
1W
7
F INAL S TR
1
0L
0L
T ROCHEE
0
1W
0
Table 9: Deriving final stress
/ta.va.cı/
→ a) ta.va.(cı́)
b) ta.(va.cı́)
c) ta.(vá.cı)
E XC S TR
0
0
0
*F INAL S TR(d)
0
0
0
F INAL S TR
0
0
1W
T ROCHEE
0
1W
0
Adequate predictions This constraint set correctly predicts the position of primary
stress in inputs of the following type: a) with non-final monosyllabic stress disrupting suffix, b) with final disyllabic stress disrupting suffix and c) with non-final disyllabic stress disrupting suffix.
Table 10: Non-final monosyllabic stress disrupting suffix
/gel.me.di/
→ a) (gél.me).di
b) (gel.mé).di
c) gel.(mé).di
E XC S TR
0
0
0
*F INAL S TR(d)
0
1W
1W
F INAL S TR
1
1
1
T ROCHEE
0
1W
0
Table 11: Final disyllabic stress disrupting suffix
/ge.li.yor/
→ a) ge.(lı́.yor)
b) ge.(li.yór)
c) ge.li.(yór)
E XC S TR
0
0
0
*F INAL S TR(d)
0
1W
1W
F INAL S TR
1
0L
0L
T ROCHEE
0
1W
0
Table 12: Non-final disyllabic stress disrupting suffix
/ge.li.yor.du/
→ a) ge.(lı́.yor).du
b) ge.(li.yór).du
c) ge.li.(yór).du
E XC S TR
0
0
0
*F INAL S TR(d)
0
1W
1W
F INAL S TR
1
1
1L
T ROCHEE
0
1W
0
Multiple stress disrupting suffixes, leftmost wins When a word features multiple
stress disrupting suffixes, stress assigned by the leftmost suffix that surfaces. Under
the present analysis, for a word with two stress disrupting suffixes, the candidate
with a foot aligned with the leftmost suffix and the other with the foot aligned with
the rightmost suffix tie on all constraints.
The following constraint can be used to break the tie.
L EFTMOST: A LIGN (F T, L, W D , L): Assign one violation mark per intervening syllable between the left edge of a foot and the left edge of a word.
8
Table 13: Two stress disrupting suffixes, no winner
/gel.me.di.mi/
a) (gél.me).di.mi
b) gel.me.(dı́.mi)
E XC S TR
1
1
*F INAL S TR(d)
0
0
F INAL S TR
1
1
T ROCHEE
0
0
Table 14: Two stress disrupting suffixes, leftmost wins
/gel.me.di.mi/
→ a) (gél.me).di.mi
b) gel.me.(dı́.mi)
E XC S TR
1
1
*F INAL S TR(d)
0
0
F INAL S TR
1
1
T ROCHEE
0
0
L EFTMOST
0
2W
Expansion to emphatic reduplication
The constraint set developed for stress disrupting suffixes extends to a second set
of words that display an exceptional stress pattern. These are adjectives that have
undergone emphatic reduplication, illustrated in (7):
(7) a. ma.vı́, más.ma.vi, ‘blue, very blue’
b. in.cé, ı́p.in.ce, ‘thin, very thin’
If the word initial reduplicant is marked with the diacritic used for stress disrupting suffixes, the high ranking E XCEPTIONAL S TRESS will favor candidates with a foot
aligned with the right edge of the reduplicant. Given that the reduplicant is at most
monosyllabic, the foot to be monosyllabic as well: (másd ).ma.vi. This candidate,
the correct output, will trigger a *F INAL S TR(d) violation given that the d-bearing
morpheme is stressed. This makes the candidate (masd .má).vi a good competitor, violating E XCEPTIONAL S UFFIX but not *F INAL S TR. To rule it out, the following
ranking argument needs to be made E XC S TR *F INAL S TR:
Table: Ranking argument E XC S TR *F INAL S TR
/masd .ma.vi/ E XC S TR *F INAL S TR(d)
→ a) (másd ).ma.vi
0
1
b) (masd .má).vi
1W
0L
If these constraints are otherwise undominated, no other candidate is serious
competition. Candidates where the foot is not aligned with the reduplicant will be
ruled out by E XC S TR: masd .(má.vi). And candidate (másd .ma).vi, although
the position of stress is correct, will be ruled out given that it violates both E XC S TR
and *F INAL S TR(d).
9