The Representation and Processing of Vowels in Brazilian Portuguese and American English

This dissertation investigates how the abstract representation of vowels affects spoken word recognition in Brazilian Portuguese and American English by examining two issues in theoretical phonology and speech processing: underspecification theory, and underlying representation when there is alternation. Three experiments were conducted in Brazilian Portuguese (B.P.) in order to test the predictions of the Featurally Underspecified Lexicon (FUL) Model with regards to the specification of Articulator and Height features, while also testing the claim of the model that underspecification is a property of the lexicon itself and a predictor of complete lexical access. To do this, experiments tested the same B.P. vowel contrasts but provided evidence from different levels of linguistic representation and processing: semantic (cross-modal semantic priming; Experiment 1); phonological (cross-modal form/fragment priming; Experiment 2); and psychoacoustic perception (AX discrimination; Experiment 3). Evidence of the coronal-dorsal asymmetry predicted by FUL was found in Experiment 1, supporting the relevance of underspecification in processing, while also providing needed evidence for how mid vowel contrasts are represented in B.P. However, both features and acoustic distance between vowel categories were needed to explain the full set of findings. Additionally, because similar patterns were not replicated in Experiments 2-3, findings from Experiment 1 that support FUL did not seem to stem from lower-level perception or sub-lexical form activation. Collectively, the evidence from these experiments supports FUL's claim that listeners extract features from the signal and map them directly to the lexicon in achieving complete lexical access. The remaining B.P. experiment (cross-modal form/fragment priming; Experiment 4) investigated a well-studied case of mid vowel alternations within the B.P. verbal paradigm to address how speakers choose a UR (underlying representation), as well as the role of this UR in processing, for a pattern of alternation with a relatively abstract phonological analysis. This was the only experiment to test lexical items for which there is reason to posit an abstract UR that does not always match the surface representation (SR), given the alternation. Primarily, the experimental evidence showed that B.P. listeners primarily utilize the vowel heard in the signal to map to the SR of individual inflectional forms. Nevertheless, priming patterns also suggested that B.P. listeners have access to the UR of the alternating root vowel, as well as a floating height feature that sometimes spreads from the theme vowel (the vowel observed in the infinitive suffix) to the root. This finding indicates that B.P. listeners analyze the alternation pattern in terms of an abstract UR and phonological derivation (i.e. they do not just list the SRs of each inflection in the paradigm, even though the SR is what they access most strongly), and that the UR affects processing. Furthermore, the results of this experiment reinforce the arguments of previous authors about a host of factors that influence how speakers analyze the UR when there is alternation, and they support an analysis in the B.P. literature in which the lexical representation of B.P. verbs includes both the root and the theme vowel. As a typological comparison with B.P., an English experiment (cross-modal semantic priming; Experiment 5) found an asymmetry in processing between tense and lax high front vowels suggesting that lax vowels are phonologically specified for [RTR] while tense vowels are underspecified for [ATR]. This is a needed experimental contribution to an understudied part of the FUL feature geometry and is consistent with theoretical observations about the behavior of tense and lax vowels both in English and across the world's languages. Crucially, the typological comparison between languages with different phonological systems allows this dissertation to show that cross-linguistic differences in abstract representation have consequences in processing, while also providing experimental evidence in support of underspecification from multiple types of vowel features. As an additional contribution, findings from Experiment 5 also suggested that there is not always a 1-to-1 mapping between phonological representation and acoustic realization, reinforcing FUL's claim that the features represented and the features extracted from the speech signal are equally important in spoken word recognition. In summary, this dissertation contributes to issues about underspecification in the lexicon, abstract UR when there is alternation, and the individual representations of vowels in Brazilian Portuguese and American English. Together, these investigations highlight the importance of underlying representation in processing, making valuable contributions to the fields of psycholinguistics and theoretical phonology. Finally, the collective findings from multiple experimental paradigms allow this dissertation to make an important methodological contribution by showing how evidence from different levels of linguistic representation and processing can provide a deeper understanding of the linguistic system, thereby paving the way for future investigations of these and other languages. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page: http://www.proquest.com/en-US/products/dissertations/individuals.shtml.]