| In this dissertation I present the results of three experiments as a case study in combining phonological theory and neurolinguistic methods. The text commences with a review of linguistic approaches to 'mentalism', noting that linguistics has alternately embraced and rejected 'mentalism' a number of times. Ultimately I find grounding in the notion that linguistic theory seeks to model the human language capacity, and in doing so, must address the psychological instantiations of theoretical principles, particularly given the recent rise of cognitive science, and the multitudinous methods for investigating cognitive processing now available. I argue that linguistic theory must be integrated into the current cognitive dialog in order to truly make strides towards understanding language as a human system.;The specific approach taken is to examine the role predictable alternations play in the processing of two English prefixes: 'in-' and 'un-'. These two prefixes provide a convenient pairing because 'in-' alternates depending on what stem it is attached to (for instance, 'i[n]decent', 'i[m]balanced'), whereas 'un-' typically does not (compare 'u[n]detered', 'u[n]becoming'). Several theories which address the phonological context of lexical items have suggested that items which regularly alternate may be underspecified. That is, that the alternating features of these items (in this case, the place of articulation) may be omitted in the stored form of these words, as it is predictable given the context. If this is the case, 'in-' should be underspecified, 'un-' should not, and thus, each prefix is suggested to have a separate representational status.;To test this hypothesis, a set of three auditory lexical decision experiments were carried out (two behavioral, one electrophysiological) using a set of 'in-' and 'un-' words, some of which were mispronounced. These items were of three types: (1) UMP forms ('u[m]predictable'), with a modified 'un-' prefix attached to a labial stem, (2) UMT forms ('u[m]tenable'), with a modified 'un-' prefix on a non-labial stem, and (3) IN forms ('i[n]balanced', 'i[m]tangible') with a modified 'in-' prefix on either labial or non-labial stems. Subjects were asked to identify whether a given stimulus item was properly pronounced, or was mispronounced / unfamiliar. Following from the above theory, the prediction is made that the IN items would be indistinguishable from real words, by virtue of the underspecified lexical form.;The results showed a fascinating pattern of responses, in terms of latencies, classification accuracy, and electrophysiological responses. First, UMT nonwords, showed better nonword discrimination accuracy, fast response latencies, and generated an N400 nonword response. UMP nonwords, on the other hand, showed poor nonword discrimination accuracy, somewhat fast response latencies, and lacked an N400 response. Crucially, the IN nonwords also showed poor nonword discrimination, slow response latencies, and also lacked an N400 response. In short, subjects did not appear to be able to distinguish either the phonetically-licensed UMP nonwords, nor the IN nonwords, from real words. This result provides evidence not only for alternation-based underspecification, but in the case of UMP, also suggests that assimilation context is evaluated during online processing.;These experiments point to a rare area where theoretical linguistics and psycholinguistic theories provide competing yet complementary predictions about language processing, highlighting the utility of an interdisciplinary approach to speech perception. I conclude by suggesting a model of speech perception which can incorporate insights from both lexical access models (in terms of addressing assimilation context on-line) and theoretical phonology (in terms of alternation-based underspecification), in service of modeling the obtained results. |
