Professional Documents
Culture Documents
Linguistic perspectives
336
Now, in which cases will surface phonological representations differ from lex-
ical or phonological ones? Clementss (2001) answer is as follows: in all cases when
a given feature is active at the surface phonological level (cf. (4)).
(4) Activation Criterion:
In any language, redundant feature values are specified in all and only
the segments in which they are active.
A feature value is active in any segment or segment class which satisfies
a term in a constraint mentioning that feature. (Clements 2001: 17)
We assume that at the surface phonological level Bulgarian vowels maintain
the same features for constriction location ([coronal] and [labial]) as specified in
(2) but have full specifications (all features are active) for aperture (cf. (5)).
(5) Surface phonological representation for aperture:
[i, u] [, e, o] [a]
aperture aperture aperture
open: tier 1 +
tier 2 + +
2.3 Analysis
The prominence alignment approach to Vowel Reduction (Crosswhite 2000,
2001) takes into account the position (scale 1) and the absolute sonority of vowels
(scale 2):
Scale 1: Accentual prominence
Stressed
prom
> Unstressed
Scale 2: Vocalic prominence
a
prom
> E,
prom
> e, o
prom
> i, u
prom
>
Iberian and Slavonic Cultures: Contact and Comparison
337
(6) *Unstressed/a *Unstressed/e, o *Unstressed/i, u *Unstressed/
The alignment/markedness constraints in (6) are the result of crossing scale 1
with scale 2 where, for the present purposes, only unstressed position is considered
relevant. The constraint *Unstressed/a means the worst candidate for unstressed
position is [a]; *Unstressed/e, o means [e, o] are slightly better but are still bad
the last constraint says that [] is the best of the bad.
Constraint ranking for EB (Crosswhite 2000, 2001):
(7) Max[round], Max[+fron] *Unstressed/a *Unstressed/e, o *Unstressed/
i, u *Unstressed/ Max[-high]
In (8) we present Crosswhites explanation for the reduction of /e/ to [i] in EB.
The author uses traditional features like [round], [front] and [high].
(8) Reduction of /e/ in Eastern Bulgarian (Crosswhite 2000, 2001):
[sil]
[sel]
[sl]
[sal]
/sel/
villages
Max
[round]
Max
[front]
*Unstr-
a
*Unstr-
e,o
*Unstr-
i,u
*Unstr-
Max
[-high]
*!
*! *
*!
*
*
*
*
*
*
Our proposal for Western Bulgarian constraints and their ranking:
(9) Ident[labial], Ident[coronal] *Unstressed[+open
1
]
*Unstressed[+open
2
]&
7
[labial] Ident[open] *Unstressed[+open
2
]&[coronal]
*Unstressed[-open
2
]
8
7
The symbol & is used here to mean and; in OT it is used generally with the value & = Local
Conjunction of. Constraints.
8
For the linguists unfamiliar with the OT framework: the Ident constraints protect a given property
to survive at the phonetic level. For example, Ident[labial] says dont loose the labial property
of a vowel. The alignment/markedness constraints (those with *X) are constraints which point
out bad combinations. For instance, *Unstressed[+open
1
] means the most open vowels are bad
candidates for unstressed position. The symbol means dominates and allows us to establish
the ranking of constraints. For instance, Ident[coronal] dominates *Unstressed[+open] in the
Linguistic perspectives
338
For the sake of simplicity, in the tableaux that follow, we do not include con-
straints that do not interact (are not violated) and constraints which interact after a
fatal violation (*!) because their function is irrelevant for the final assessment. The
ranking of constraints is the same for all candidate evaluations.
(10) Preservation of /e/ in unstressed position:
sense that it states that to preserve the coronal property of a vowel is more important than to look
at the fact if a vowel is a bad candidate for unstressed position. The same happens in the ranking
of the alignment/markedness constraints *Unstressed[+open
1
] *Unstressed[+open
2
]
*Unstressed[-open
2
]. *Unstressed[+open
1
] dominates *Unstressed[+open
2
] which means it is
worse to be the most open vowel in unstressed position than to be the less open one, etc. Comma
(,) between two constraints means that they are unranked one to the other. And the symbol &
points out a conjunction of constraints, for example, *Unstressed[+open
2
]&[coronal] means the
combination of [+open
2
]&[coronal] is a bad combination to occur in unstressed position.
[dil]
[del]
[dl]
[dal]
[dol]
/del/
works
Ident
[coronal]
Ident
[open]
*Unstr-
[+open
2
]&[coronal]
*!
*!
*!
*!
*
(* = constrain violation; *! = fatal constraint violation; shading = any further violation is irrele-
vant for this candidate; F = the correct phonetic output; importance of constraints (ranking) =
from left to right)
In (10) one can see that the preservation of coronal in WB is essential and the
violation of Ident[coronal] eliminates at once three candidates. The winning candi-
date is [del] because it violates the lowest ranked constraint.
(11) Rising of /a/ in unstressed position:
[grd t]
[grud t]
[grod t]
[grad t]
/grad t/
the city
*Unstr-
[+open
1
]
*Unstr-
[+open
2
] & [labial]
Ident
[open]
*Unstr-
[-open
2
]
*!
*!
*
*
*!
Iberian and Slavonic Cultures: Contact and Comparison
339
In (11) we can see an interesting case where the winning candidate is deter-
mined at the second round, which is a possibility allowed in OT. The first two
forms violate Ident[open] but only [grudt] fatally violates the lowest ranked con-
straint *Unstr-[-open
2
].
(12) Rising of /o/ in unstressed position:
[dl t]
[dol t]
[dul t]
[dal t]
/dol t/
the city
Ident
[labial]
*Unstr-
[+open
2
] & [labial]
*Unstr-
[-open
2
]
*!
*!
*!
*
In (12) we show the rising of /o/ to [u], due to the [+open] value of the o-
candidate.
(13) Preservation of /i/ in unstressed position:
[vin]
[ven]
[vn]
/vin/
fault
Ident
[coronal]
Ident
[open]
*Unstr-
[-open
2
]
*!
*!
*!
(14) Preservation of /u/ in unstressed position:
[lun]
[lon]
[ln]
/lun/
moon
Ident
[lab]
*Unstr-
[+open
2
] & [labial]
*Unstr-
[-open
2
]
*!
*!
*
In (13) and (14) /i/ and /u/ do not shift because they are [-open
2
] and the
constraint *Unstr-[-open
2
] is the lowest ranked, i. e. its violation is irrelevant.
Linguistic perspectives
340
3. Vowel Reduction in European Portuguese
3.1. Data
It is consensual between linguists that European Portuguese has a system of
seven phonological vowels: /i, e, E/ palatal, /a/ central and /u, o, / rounded
9
. In
general, three degrees of height have been proposed for this system in the litera-
ture: /i, u/- high, /e, o / mid, and /E, a, / low (cf., e. g., Mateus & Andrade
(2000)). Phonetically, /a/ is lower than any other vowel so that one may consider
the existence of four degrees of vowel height in this language. However, from a
structural point of view, we may claim that three degrees of height are sufficient to
describe the Portuguese vowel system.
The phonetic EP vowel system also includes [, ] and a set of five nasalized
vowels [I ) , e , , o) , u)]. All phonological or phonetic vowels except [] may occur in
stressed position. Nasalized vowels in unstressed position do not undergo vowel
reduction although low nasalized vowels are raised by independent conditioning
into mid ones
10
.
The following examples of morphologically related words from Mateus &
Andrade (2000: 135) illustrate the VR in EP, for front vowels (a) and back vowels
(b):
(15) a. dedo [] dedada [] finger/fingerprint
bater [] bate [] to beat/you beat
festa [E@] festejo [] party/festivity
mel [E@] melado [] honey/sweetened with honey
fita [] fitinha [i] band/small band
b. fogo [] fogueira [u] fire/fireplace
porta [@] porteira [u] door/doorkeeper
furo [] furado [u] hole/pierced
gato [] gatinho [] cat/pussy
virar [] vira [] to turn/you turn
9
See Morais Barbosa (1965) for a structuralist view and subsequently for a different vowel counting.
10
As a matter of fact, there is a low nasalized vowel in EP, []. This vowel may appear as the result of
the fusion of the sequence [) ], as in a Antnia, which is pronounced [].
Iberian and Slavonic Cultures: Contact and Comparison
341
Unstressed /e, E/ change to [],
Unstressed /o, / change to [u],
Unstressed /a/ changes to [],
Unstressed /i, u/ remain unchanged.
Mateus & Andrade (2000: 135)
3.2 Representations
Let us start with the vowel []. It is a schwa or schwa-like vowel. Phonological-
ly, it exhibits the properties of a neutral or default vowel, it is never stressed, it is
the main epenthetic vowel, it is the most frequently deleted vowel, and it is the
result of the /E, e/ neutralization.
We assume that [] is underspecified for V-place and that it has not a full set of
aperture features. Bearing the specification [-open
1
], [] is thus identical to high
and mid vowels on the [open
1
] tier but distinct from them on the [open
2
] tier be-
cause the former are [-open
2
] and the latter are [+open
2
] (cf. (16)).
/i, e, E/ are [coronal], /u, o, / are [labial] and /, a/ are [dorsal].
We assume that two aperture tiers are necessary and sufficient for the descrip-
tion of three vocalic heights and all phonological contrasts in EP.
(16) Surface phonological representations for aperture:
[i, u] [] [e, o, ] [E, a, ]
aperture aperture aperture aperture
open: tier 1 +
tier 2 + +
3.3 Analysis
The Hierarchy of constraints we propose for EP VR is as follows:
(17) Ident[labial], Ident[dorsal] *Unstressed[+open
1
] *Unstressed[+open
2
]
Ident[open] *Unstressed[-open
2
] Ident[coronal]
Linguistic perspectives
342
(18) Preservation of /i/ in unstressed position:
Iberian and Slavonic Cultures: Contact and Comparison
The feature [coronal] survives in [i] of [vizi@u] because [] is penalized by
Ident[open], for [] differs in height value from the input vowel.
Notice that the form [vzi@ u] exists in EP but its first vowel is an effect of an
independent dissimilatory process. Generally unstressed /i/ does not change to [].
(19) Centralization of /e/ in unstressed position:
The vowel /e/ is penalized by *Unstressed[+open
2
]. The i-candidate fatally
violates *Unstressed[-open
2
]. /e/ looses [coronal] and becomes schwa because of
the low ranking of Ident[coronal] (remember the ranking in (17)).
(20) Centralization of /E/ in unstressed position:
[didd]
[dedd]
[dEdd]
[ddd]
[ddd]
[dadd]
/dedd/
fingerprint
*Unstr
[+open
1
]
*Unstr
[+open
2
]
Ident
[open]
*Unstr
[-open
2
]
*!
*!
*!
*!
*
*
!*
[mEldu]
[mildu]
[meldu]
[mldu]
[mldu]
[maldu]
/mEldu/
s. w. honey
*Unstr
[+open
1
]
*Unstr
[+open
2
]
Ident
[open]
*Unstr
[-open
2
]
*!
*!
*!
*!
*
*
!*
[vizu]
[vezu]
[vEzu]
[vzu]
[vzu]
[vazu]
/vizu /
neighbour
*Unstr
[+open
1
]
*Unstr
[+open
2
]
Ident
[open]
*Unstr
[-open
2
]
*!
*!
*!
*!
*!
*
343
The vowel /E/ is penalized by *Unstressed[+open] constraints. The i-candi-
date fatally violates *Unstressed[-open
2
] and /E/ looses [coronal] and becomes schwa
because of the low ranking of Ident[coronal] as in the case of /e/.
(21) Rising of /a/ in unstressed position:
Linguistic perspectives
[dorsal] is a strong feature in EP, so non dorsal candidates are all penalized.
The candidate [gatu] is penalized by *Unstr[+open
1
].
(22) Rising of /o/ in unstressed position:
(23) Rising of // in unstressed position:
[fog jr]
[fg jr]
[fug jr]
[fg jr]
[fg jr]
[fag jr]
Ident
[open]
/fog jr/
fireplace
Ident
[labial]
*Unstr
[+open
1
]
*Unstr
[+open
2
]
*!
*!
*!
*!
*!
*
[gatu]
[gtu]
[gtu]
[gutu]
[gotu]
[gtu]
[gitu]
[getu]
[gEtu]
*Unstr
[+open
2
]
/gatu/
little cat
Ident
[dorsal]
*Unstr
[+open
1
]
*!
*!
*!
*!
*!
*!
*!
*!
*
[port]
[prt]
[purt]
[prt]
[prt]
[part]
Ident
[open]
/prt/
little door
Ident
[labial]
*Unstr
[+open
1
]
*Unstr
[+open
2
]
*!
*!
*!
*!
*!
*
344
(24) Preservation of /u/ in unstressed position:
Iberian and Slavonic Cultures: Contact and Comparison
Ident
[open]
/pulr/
to jump
Ident
[labial]
*Unstr
[+open
1
]
*Unstr
[+open
2
]
*!
*!
*!
*!
*!
* [pulr]
[polr]
[plr]
[plr]
[pilr]
[plr]
From the evaluations presented in tableaux (22), (23) and (24), we can see that
[labial] is also a strong feature in EP. Ident[labial] penalizes all non labial candi-
dates. On the other hand, *Unstr[+open] constraints penalize forms with [o] or []
because, as stated before, these constraints do not tolerate open vowels.
Notice, that the ranking Ident[open] *Unstr[-open
2
] prevents the lowering
or centralization of high vowels so that /u, i/ preserve their original aperture.
4. Concluding remarks
With Feature Geometry features and Feature Activation we can easily express
combination between features or between features and positions, and make possi-
ble for lexically or phonologically redundant features to take part in the constraint
interaction.
If empirical data and future research require it, one may translate FG features
to, for example, an acoustic, and auditory or other articulatory ones.
In an OT framework it was possible to determine which features are strong
and which ones are weak; for instance, [coronal] is a weak feature in EP and it is
strong in WB: coronal vowels are decoronalized when reduced in EP, but not in
WB; [dorsal] plays no phonological role in WB but it is active in PE; [labial] be-
haves the same way in both languages it is strong since it preserves its underlying
value.
PE and WB have in common the fact that they have the same set and the same
ranking of constraints concerning aperture, penalizing more open (mid and low,
more sonorous) vowels in unstressed position. Both languages show some differ-
ences in the ranking of place constraints but nevertheless they belong to the prom-
inence reduction type.
345
Bibliography:
1. Bates, S. A. (1995). Acoustic Characterisation of Schwa: a Comparative Study. In Proceedings
of the XIIIth International Congress of Phonetic Sciences, Vol. 3: 230-233.
2. Browman, C. and Goldstein, L. (1992). Targetless schwa: An articulatory analysis In Docherty
e Ladd (eds.), Papers in Laboratory Phonology, II: Gesture, Segment, Prosody. Cambridge: Cam-
bridge University Press. 26-67.
3. Clements, G. N. (2001). The Representational Economy in Constraint-Based Phonology. In T.
Alan Hall (ed.), Distinctive Feature Theory. Berlin: Mouton de Gruyter.
4. Clements, G. N. and Hume, E. (1995). The internal organization of speech sounds. In J. Gold-
smith (ed), The Handbook of Phonological Theory. Cambridge USA/Oxford UK: Blackwell.
5. Crosswhite, K. (2000). Sonority-driven reduction. In BLS 26.
6. Crosswhite, K. (2001). Vowel Reduction in Optimality Theory. New York: Rutledge.
7. Hristovsky, G. (in press). Ocorrncia Mltipla de Traos e a Conjuno Local de Restries:
dados do Blgaro. In Actas do XX Encontro Nacional da APL. Lisboa: APL.
8. Hristovsky, G. (2000). Alternncias Voclicas e Consonnticas do Blgaro. PhD, Lisbon Univer-
sity.
9. Mateus, M. H. e Andrade, E. d (2000). The Phonology of Portuguese. Oxford: OUP.
10. Pettersson, T. e Wood, S. (1987). Vowel reduction in Bulgarian and its implications for theories
of vowel production: a review of the problem. In Folia Linguistica XXI/2-4: 261-279.
11. Prince, A, and Smolensky, P. (1993), [2002]. Optimality Theory: Constraint Interaction in Gen-
erative Grammar. New Brunswick, NJ: Rutgers University Centre for Cognitive Science.
12. Redenbarger, W. (1981). Articulator Features and Portuguese Vowel Height. Cambridge, Massa-
chusetts: Harvard University.
13. Steriade, D. (1995). Underspecification and Markedness. In J. Goldsmith (ed), The Handbook
of Phonological Theory. Cambridge USA/Oxford UK: Blackwell.
14. van Bergem, D. (1994). A model of coarticulatory effects on the schwa. In Speech Communica-
tion, 14(2):143-162.
15. Wood, S. e Pettersson, T. (1988). Vowel reduction in Bulgarian: The phonetic data and model
experiments. Folia Linguistica XXII/3-4: 239-262.
Linguistic perspectives