| In this dissertation, an artificial lexicon and novel visual environment provide sufficient control to explore two unresolved issues in the spoken word recognition (SWR) literature - the nature of partially activated semantic representations and the locus of context effects.; In Experiment 1, participants learn names for novel objects and modifiers and use a computer mouse to modify the objects in accordance to spoken instructions. We verify that standard lexical effects are seen using this technique. We then use fMRI to investigate the nature of the neural representations that become partially activated during lexical competition. There is evidence that knowledge about objects and events is stored in brain areas used to perceive those objects and events. Furthermore, when participants perform tasks with words referring to objects or events, similar patterns of activity are seen. These findings dovetail with recent discussions about distributed lexical representations, in which any sort of information that co-occurs with words becomes part of the lexical representation. In Experiment 2, we observe increased BOLD activation anterior to MT for motion words relative to color words. This is the first demonstration of this effect with spoken language and with newly trained lexical items. Furthermore, the semantics of both the target and the cohort stimuli affect activation levels, suggesting that these areas may be part of distributed lexical representations. This activation is not strategic, since activation of multiple lexical candidates is not a conscious process.; In Experiments 3-5, the version of the visual world paradigm used in Experiment 1 is modified so that contextual information about possible object/modifier pairings is available before the onset of the target word. Anticipatory eye movements reveal that this contextual information is available and is used by participants before target onset. Examinations of cohort and neighborhood density effects reveal that although context does not have a pre-selective effect as predicted by early selection models of SWR, there is no discernable delay before context affects competition as an access-selection model requires. These data are an important contribution toward establishing what sorts of models may better capture the appropriate relationship between context and spoken word recognition. |
