SYLLABLE STRUCTURE IN THE MENTAL LEXICON TO EXPLAIN NONCONCATENATIVE MORPHOLOGY: A NEW MODEL FOR SPEECH PRODUCTION

Back to Page Authors: Dinesh Ramoo

Keywords: speech production, semitic morphology, language production model, nonconcatenative morphology

Abstract: Linguistics theory includes various units of production such as phonology (the study of the smallest units of sound) and morphology (the study of the smallest units of meaning) in a language (McCarthy, 1991). Most prominent speech production models operate within these domains to analyse empirical data to inform their hypothesis on how these units are organised and processed. Prominent speech production models include the spread activation model by Dell (1986; 1988) and the LRM model by Level et al. (1999) which are based on speech error analysis and chronometric data respectively. Dell stores phonemes based on their syllable position so that a /p/ in onset is stored separately from a /p/ in coda. This is supposed to account for the preservation of syllable position in speech errors as well as allophones (e.g., syllable-initial aspiration in English and syllable-final devoicing in German). The LRM model on the other hand stores phonemes according to serial position and only store a syllabic template to account for irregular stress patterns in some European languages. As these models often focus on data from European languages, they may not be able to account for language production in languages that do not share their morphological characteristics. The Lexicon with Syllable Structure (LEWISS) is a more recent model by Romani et al. (2011) which introduced the concept of storing syllable structure to organise phonemes based on data from Italian aphasic patients’ speech errors. While this model was initially based on data from European languages, this theoretical attempt will illustrate how it can better account for nonconcatenative morphology in Semitic languages better than other models. The Semitic language family includes Arabic, Hebrew, and Aramaic. The main characteristics of Semitic morphology is that of root and pattern. Most European languages tend to have regular morphology such as the plural morpheme in English being realised as [bɔɪz] from {boy}{s} and [wɔːks] from {walk}{s}. However, Semitic morphology superimposes roots usually consisting of three consonants onto a pattern or template that consists of consonants and vowels with slots for including the root. For example, the root {ktb} in Arabic may be slotted into C1aC2C2aC3 or maC1C2aC3 to produce /kattab/ “caused to write” or /maktab/ “office”. Dell cannot account for the storage of this pattern as the phonemes change syllable position for different morphological patterns: [k] is an onset in /kattab/ but a coda in /maktab/. The LRM model is also unable to store these roots as serial positions will ignore empty slots and cannot indicate germination (as in /kattab/ where [t] doubles). This study looked at how a model such as LEWISS might be able to account for these patterns by storing syllable structure that can adapt to these various morphological patterns efficiently. We will also present data from languages from Italian (Romani et al., 2011) and Hindi (Ramoo & Olson, 2012) to illustrate how this model can explain various linguistic phenomena in these languages as well. Further investigation is indicated to explore this model in Semitic languages.