Bilingual speakers present an intriguing puzzle for selective attention research because of the co-existence of two languages in the brain. Research (e.g., Marian & Spivey, 2003) shows that both languages of the bilingual are activated in parallel when bilinguals intend to use one language alone. What are the cognitive mechanisms that enable bilinguals to select one language for use in the midst of competing languages in the brain? How do bilinguals regulate influences from the nontarget language when one language is in use? These are some of the major questions investigated by cognitive psychology and psycholinguistic researchers. To understand bilingual language processing, it is important to uncover the structure and organisation of language representations in memory, as well as the processes involved in coordinating the two different languages. This dissertation presents a series of studies that examine the cognitive mechanisms underlying bilingual lexical processing, using a priming technique, and further test predictions from the two rival theories of conceptual negative priming - the inhibition based and episodic retrieval accounts.

The tasks involved the presentation of pairs of items, one in uppercase letters and one lowercase letters. Those in lowercase letters served as targets and those in uppercase served as distractors. Participants were asked to ignore the uppercase items in prime-probe couplets, which comprised a trial. During presentation of the prime stimuli, participants were required to name the lowercase target word aloud, followed by the probe display, which required them to make a lexical decision (word/nonword judgement) to the lowercase item. The key relationships between prime and probe stimuli were manipulated in three ways. In the attended repetition (AR) condition the attended lowercase prime word was the same as, or the noncognate translation equivalent of the attended probe target item. In the control (CO) condition the prime word pairs had no relationship with the probe items, and in the ignored repetition (IR) condition the prime distractor word was the same as, or the noncognate translation equivalent, of the attended probe item. Each word was used no more than twice in the experiments, and then only to satisfy either the AR or IR manipulations.

In the within-language experiments (all English stimuli, or all Twi stimuli), lexical decisions to probe target words were facilitated when the word was identical to the preceding prime target word, whereas delayed lexical decisions to probe target words occurred when the word was identical to the preceding prime distractor word. The within language experiments thus produced positive (facilitatory) and negative priming (NP) effects respectively. The NP effect that emerged in the within-language experiments showed that NP can occur with experimentally novel stimuli, that is, stimuli that are seen no more than twice in an experiment, once as the distractor and then as the target in the prime-probe couplet.

In the bilingual (Twi-English) experiments, between languages rather than within language priming manipulations were used. Ignored repetition NP effects were found across languages but cross-language AR positive priming effects disappeared. This dissociation of priming effects in the within and between-language priming conditions is inconsistent with episodic retrieval predictions. Instead, the results support inhibition-based accounts by showing that bilingual language selection is achieved by active inhibition and that inhibition can flexibly operate at both the local and global levels of abstraction. In the between-language (Twi-English) task where participants were categorised according to second language (L2 = English) less and more proficient bilinguals, the more proficient produced robust NP effects coupled with no positive priming. Less proficient bilinguals, however, showed a trend towards positive priming, but nonsignificant NP. These results indicate that more proficient bilinguals relied on inhibition of the Twi (first) language in order to perform the task in the English (second) language, thereby accounting for the elimination of AR positive priming and the preservation of IR NP. Less proficient participants, in contrast, seem to rely on the first (Twi) language as a form of crutch to perform the task in the second language.

Finally, in the between-language (English-Twi) experiment where participants were required to name prime targets in their L2 followed by making lexical decisions to letter strings in their L1, AR positive priming effects were obtained, but no IR NP effects. A plausible conclusion suggested that there is inhibition of the dominant (L1) language when bilinguals perform the prime task in the weaker (L2) language. However, the weaker (L2) language does not have to be inhibited to the same magnitude (or perhaps at all) in order to perform the subsequent probe task in the dominant (L1) language; hence accounting for the AR positive priming. Due to the relative weakness of the language used in the prime, the prime distractor also elicits comparatively weaker competition with the target, thus reducing the degree of inhibition applied to it, which could potentially account for the absence of NP. Taken together, the cross-language priming effects reported here seem to fit single-store models of bilingual language representation, wherein conceptual representations are deemed to be integrated across languages in bilinguals. Because of the processing mechanisms involved in juggling the languages and words within them, however, they operate as if they were functionally separated (Neumann, McCloskey, & Felio, 1999).