Opaque Allomorphy in OT
Allomorph distribution in Optimality Theory (Prince & Smolensky 1993) is accounted for by the
properties of the output (see Anttila 1997, Kager 1999, Kiparsky 2000, Mascar\ 1996, Mester 1994,
Oostendorph 1998, Rubach & Booij 2001, among others). Opaque allomorphy occurs when the
distribution of the allomorphs cannot be determined in that way. This is problematic for a surface-based
approach to phonology such as OT. This talk argues that opaque allomorphy can be accounted for in
terms of preserving contrasts (see Kenstowicz 1995, Steriade 1997, Urbanczyk 1998).
In Polish, there is a process of Coronal Palatalization (Rubach 1984) by which anterior coronals
(dentals and alveolars) become prepalatals before front vowels (/t d n s z/[ dï ½ Ñ ï]/_i e j). There are
also underlying prepalatals in Polish. In the locative and vocative singular of masculine and neuter nouns,
underlying and derived prepalatals take different allomorphs. Derived prepalatals take the front suffix [e]
while underlying prepalatals take the back suffix [u] (see (1)).
This is an example of opaque allomorphy (Kenstowicz 1995, Oostendorp 1998). The properties of
the output in the locative do not explain the distribution of the allomorphs. Given the outputs, both sets of
prepalatals should take the same allomorph. Since prepalatals are front consonants (Rubach 1984), like
other front consonants in Polish they should take the front allomorph [e] (see (2)).
My key proposal is that allomorphy in the locative (shown in (1)) preserves the original contrast
between dental/alveolar vs. prepalatal sounds /list/ vs. /liÑ/ despite palatalization. The original contrast in
consonant quality is manifested as different allomorphs, /list/ vs. /liÑ/ map onto [liÑ + e] vs. [liÑ + u],
respectively. In other words, the original contrast in the stem-final consonant is transformed into a
surface contrast in the choice of the locative suffix (see (3)).
To implement this observation, I propose that contrast exists as an imperative in the grammar (see
Lubowicz 2003, cf. Flemming 1995, Padgett 1997) which is formulated as a family of rankable and
violable constraints on preserving contrasts, called PC constraints. PC constraints are defined in (4). PC
constraints demand that input contrasts be preserved in the output in some way but not necessarily in the
same way as in the input. Unlike standard faithfulness, PC constraints allow for contrast transformation.
To evaluate constraints on contrast, a candidate must be a set of input-output mappings, called a scenario.
In case of allomorphy, the inputs of a scenario contain a set of stems that are identical for all languages
(Lubowicz 2003) and a set of allomorphs which are language-particular. An example is given in (3).
Following Rubach (1984), I assume that dentals/alveolars vs. prepalatals contrast in tongue
height. Thus, the relevant contrast constraint in Polish is PC(high). This is defined in (5). The PC
constraint demands that forms that contrast in height need to map onto distinct outputs. The main idea is
that palatalization can compel neutralization of the height contrast (PAL >> PC(high)) but allomorphy
preserves the contrast despite palatalization. Furthermore, I assume that the unmarked allomorph after
prepalatals is [e] (*Front/u) but to preserve contrast a marked allomorph [u] is selected (PC(high) >>
*Front/u). The relevant tableau is in (6).
Three scenarios are compared: the actual scenario (scenario A), a contrast-neutralizing scenario
(scenario B), and a non-palatalizing scenario (scenario C). Scenario A wins since it preserves the contrast
in height, thus satisfying PC(high), despite palatalization. Scenario B neutralizes the contrast in height
and thus fails on PC(high). Scenario C fails on PAL.
To conclude, it is proposed that locative allomorph distribution in Polish is opaque and as opaque
allomorphy it can be accounted for in terms of preserving contrasts (this proposal will be also compared
to LPM-OT). The key idea is that the different allomorphs of the locative suffix keep apart forms that the
regular phonology would otherwise neutralize. This contributes to the body of work on morphological
opacity and the role for paradigmatic contrast. The contrast account has far reaching consequences. The
main implication is that unlike other faithfulness constraints, constraints on contrast preservation can
activate a phonological process including allomorph selection.
(1) Allomorphy in the locative after prepalatals
a. Derived Prepalatals b. Underlying Prepalatals
nominative sg. locative sg. nominative sg. locative sg.
lis[t] liÑ[] + e `letter' liÑ[] liÑ[] + u `leaf'
obia[d] obie[dï] + e `dinner' narz"[dï] + e narz"[dï] + u `tool'
ok[n] + o ok[½] + e `window' ko[½] ko[½] + u `horse'
»obu[z] »obu[ï] + e `troublemaker' pa[ï] pa[ï] + u `butterfly'
(2) Allomorphy in the locative after other consonants
a. Front consonants b. Back consonants
nominative sg. locative sg. nominative sg. locative sg.
ch»o[p] ch»o[p'] + je `peasant' so[k] so[k] + u `juice'
do[m] do[m'] + je `house' ko[Ó] ko[Ó] + u `basket'
gra[f] gra[f'] + je `graph' stra[x] stra[x] + u `fear'
(3) The role for allomorphy
Input Output
list, {+e, +u} liÑ + e
liÑ, {+e, +u} liÑ + u
(4) PC(P)
For each pair of inputs contrasting in P that map onto the same output in a scenario, assign a
violation mark. "If inputs are distinct in P, they need to remain distinct (not necessarily in P)."
(5) PC(high)
For each pair of inputs contrasting in height that map onto the same output in a scenario, assign a
violation mark. "Preserve the contrast in height (in some way)."
(6) The role for contrast
Scenarios PAL PC(high) *Front/u
/list, {+e, +u}/ liÑ + e
A. Actual *
L /liÑ, {+e, +u}/ liÑ + u
B. Contrast-neutralizing /list, (+e, +u}/ liÑ + e *!
/liÑ, {+e, +u}/ liÑ + e
/list, (+e, +u}/ list + e
C. Non-palatalizing *!
/liÑ, {+e, +u}/ liÑ + e