El-Dakhs, D. (2011). Lexical access in second language speech production: The case of Arabic- English bilinguals. In Achievements and perspectives in SLA of speech: New Sounds 2010 Volume 1 (eds. Magdalena Wrembel, Malgorzata Kul and Katarzyna Dziubalska-Kolaczyk)- Frankfurt am Main: Peter Lang Verlag. Lexical Access in Second Language Speech Production: The Case of Arabic-English Bilinguals Dina El-Dakhs Prince Sultan University ddakhs@pscw.psu.edu.sa / dinadakhs@yahoo.com ABSTRACT The present paper reports the findings of a study that examined the effects of L2 word frequency, morphological complexity, sentential context and proficiency level on the activation of L1 lexical nodes during L2 speech production. The participants, 90 less proficient and 90 more proficient unbalanced, late Arabic-English bilinguals with L1 dominance, performed an oral repetition task of English primes in isolation and at the end of semantically constraining and neutral sentences, and then performed a lexical decision task on individual Arabic targets that were either translation equivalents of the English primes, unrelated words or non-words. The English primes varied in frequency (low vs. high) and morphological complexity (monomorphemic vs. compound). The reaction time to the Arabic targets was measured to the nearest millisecond with longer times interpreted in terms of additional processing load, and, hence, less L1 activation. The study results were statistically analyzed using a T-test and discussed in relation to models of the bilingual lexicon, lexical access and L2 speech production. Keywords: bilingual speech production, lexical access, word type, sentential context, L2 proficiency 1. INTRODUCTION Most speech production models (e.g., Levelt 1989) divide the speaking process into three main stages: conceptualization, formulation and articulation. Conceptualization involves the generation of a pre-verbal message that represents the speaker's communicative intentions. This is followed by the formulation stage which retrieves relevant lexical items from the mental lexicon. The retrieval process, known as lexical access, consists of two major steps: lexical selection and word-form encoding. Lexical selection entails the choice of target items from the pool of nodes stored in the lexicon whereas word-form encoding specifies the morphological, phonological, and phonetic characteristics of the selected items. On the basis of this information, an articulatory plan is executed at the final stage. It is generally accepted in the literature that lexical access depends on the principle of 'spreading activation'. During lexical access, concepts spread activation to items in the mental lexicon that represent full or partial matches due to their conceptual (e.g., synonyms) or formal (e.g., rhyming pairs) similarity, and all the activated items compete for lexical selection. Only the nodes with the highest level of activation are finally selected for articulation. Needless to say, the ability to select appropriate items from a vast lexicon with remarkable speed and accuracy has long amazed scholars, especially that during normal conversations a speaker is estimated to produce up to five words per second with less than one whole-word error per 2000 words (Levelt 1989). What is even more surprising is how bi- and multilingual speakers manage to perform this daunting task with minimal processing cost or error rate. It seems hard enough to select the word chair