The Perception and Production of Interdental Fricatives in Second Language Acquisition

This dissertation investigates the differential substitution of interdental fricatives ([theta¸,th]) by learners of English as a second language. Differential substitution (or transfer) occurs when learners whose native language does not include the "th" sounds, substitute different segments in their place, depending on the particular L1: [t,d], [s,z], or [f,v]. Throughout the past 50 years, various proposals have been forwarded to explain this phenomenon. The majority of these approaches have focussed on structural differences within the contrastive phonemic systems of various languages. This thesis examines two languages and two dialects of the same language: Japanese, Russian, European French, and Quebec French. Japanese and European French are known to substitute [s,z] in place of [theta¸,th], while for Quebec French and Russian, [t,d] are reported. Since European and Quebec French arguably have the same phonemic inventory of obstruents, this thesis explores the function of non-contrastive phonetic information in interdental substitution, in both perception and production. It is hypothesized that perception underlies production, such that those errors observed in production will be the sounds that are apt to be perceptually associated with the target. Furthermore, it is considered that non-contrastive features play a determining role in segmental transfer. In particular, the feature STRIDENT is hypothesized to be key in the choice of interdental substitute.;To account for how second language learners perceptually map target sounds to their internal representations, the Auditory Distance Model is developed. This model is based upon the following hypotheses. In the initial state of acquisition, learners potentially have access to all phonetic features provided by Universal Grammar. However, availability of this universal set of features is limited by the language-specific phonetic inventory, such that the absence of features or particular combinations of features in the L1 grammar forces the L2 learner to choose from among the phonetically closest L1 sounds. The selection process is implemented via the Auditory Distance Algorithm which compares the target intake features with those encoded on L1 segments. The Algorithm evaluates whether the intake and L1 features constitute matches or mis-matches and additionally calculates their relative weight. A feature's weight can be augmented if it stands in an enhancement relation with another feature. Enhanced features are preferred, but not mandatory, in linguistic systems. Thus cross-linguistic phonetic variation and the resulting diversity in feature weight is what determines differential substitution in perception and hence in production.;These hypotheses are empirically verified in five studies. The first two, the AXB-1 and AXB-2 perception tasks, were designed to tap phonetic and phonemic processing in separate conditions to demonstrate that the observed patterns of differential substitution emerge in phonetic, but not phonemic processing. The third perception experiment, Picture Identification, examines phonemic processing. The final studies analyze production. The results of one, a Word Production task, are compared with the perception findings. The other involves a Spectrographic Analysis of the L1 coronal fricative [s] to determine the degree to which the feature STRIDENT is acoustically manifested for each of the languages.The results from these studies largely support the hypotheses outlined above. To account for discrepancies between predictions and results, the role of visual information in lexical representations and the possibility of task-induced bias are discussed.