Date of Degree


Document Type


Degree Name



Speech-Language-Hearing Sciences


Valerie Shafer

Committee Members

Michelle MacRoy-Higgins

Martin Gitterman

Subject Categories

Speech and Hearing Science


Haitian Creole


The distributional patterns of heritage speakers’ reanalyzed first language are often grammatically divergent from native speakers. Irrespective of the heritage language, there is converging evidence that the cognitive process of reanalysis of heritage languages is often evident in less salient properties of the language, such as inflectional morphology (Polinsky & Kagan, 2007). Therefore, Haitian Creole (HC) is a candidate for restructuring. In this investigation we compared the morphophonological patterns of the definite article system at the production and perception level between two groups of early learners of Haitian Creole (i.e. native speakers vs. United States (US) born heritage speakers).

The study examined the extent to which heritage speakers showed variability versus systematicity when employing the various forms of the definite articles (a, ã, nã, la, lã). Participants performed a translation production task, and two perception tasks administered through a forced-choice grammaticality judgment experiment using real noun phrases and non-word phrases. Dependent measures were (i) accuracy, and (ii) morphophonological patterns of reanalysis.

At the production level, the results of this investigation indicated significant divergence between heritage and native speakers in their translation production of the morphophonological form of the definite article system. Although variation was predominately noted among the experimental group, an emerging trend of systematicity at the production level was also observed. Omissions followed by the overgeneralized determiner “la”, had advantage in selection among the competing alternatives. Thus “la” served as the dominant default property within the determiner inventory. The highest degree of destabilization occurred when the target form was “ã” and stabilization was most visible with “a/ã”. Thus, it was evident that sizeable populations of heritage speakers in this production task were reinterpreting the morphophonological rules of the definite article system.

Sociological variables such as self-rated proficiency and literacy were also correlated with the translation production experiment. No significant correlation was found between the two sociolinguistic variables and the production of the morphophonological form of the definite articles. Conversely, a significant negative correlation between phonemic cuing and accuracy was reflected among this cohort. Higher phonemic cuing (to facilitate access, if a participant could not find a word) was observed with lower definite article production scores. This occurrence indicated that heritage speakers who had difficulties with lexical access exhibited less native-like definite article proficiency.

Overall, in the production task seven heritage speakers yielded scores of 30 and above (out of 40 items). Measures from the social-language questionnaire indicated more robust input in HC during their childhood for these higher performers. Specifically, these seven speakers resided with grandparents who had limited English proficiency during their childhood.

In the perceptual domain the native speakers outperformed the heritage speakers by a wide margin in grammaticality judgment. The perception experiments consisted of two grammaticality judgment tasks. One involved real noun phrases and the second reflected non-word phrases. Both tasks yielded significant group differences. In both the production task and perception tasks the definite article “ã” was the most unstable form. However, “la” and “a/ã” were also grammatically reinterpreted, but with slightly lesser frequency than “ã”.

A significant correlation was found between the self-rated proficiency and grammaticality judgments of the real noun phrases, and between self-reported literacy and grammaticality judgments of both real and non-word noun phrases. Heritage speakers who reported that they were literate as a result of attending liturgical services in the heritage languages (HC and French) yielded higher scores when compared to those who did not develop literacy skills in the heritage language. Thus, it appears that reading facilitates some level of native–like grammatical sensitivity among heritage speakers.

Among the U.S. born heritage speakers of HC, the variability in the data affirms the phenomenon of language evolution or the “indigenization” or adaptation process of the language operating in a new locality (Mufwene, 2008). Thus, the emerging trend of overgeneralization observed with the definite article “la” at the expressive level is arguably an indication of an unmarked property having advantage selection over marked features. It is plausible that high frequency features are less taxing on the working memory of heritage speakers and, thus, are easier to acquire (Anderson, 1977). According to our proposed theory, which we reference as the contact vernacular adaptation hypothesis, unmarked features in the heritage language are more adaptable to the process of language shift in a noveau habitat, and have advantage in selection. Moreover, the flexibility of unmarked properties creates an openness for the inheritance of new features in a language during the “indigenization” or adaptation process in a new locality.