A task analysis of the Lexical Decision Task (LDT) was performed. Several alternative explanations are explored, and experiments designed to investigate these models are reported. These experiments provide strong evidence for the use of multiple decision strategies in the LDT. Subjects responded to both visual and semantic attributes of targets. Lexical decisions to repeated nonwords resulted in both increased errors and decision latencies. Apparently, the familiarity of specific visual graphemes serves as one basis for performance in the LDT. A basis that is independent of an item's true lexical status. Also, lexical decisions were biased by typography. This was interpreted as confirming the visual basis of information in the LDT. Meaning in lexical decisions was studied by varying both the type of word referent (concrete or abstract), and the availability of meaning from a prior presentation. While meaning contributed to lexical decisions, its use depended upon both a stimuli's visual familarity and the nature of the task demands. It was concluded that the LDT does not measure a single process or memory structure (i.e., Lexical Memory), rather it reflects knowledge about the visual familiarity and the semantic uses of graphemes. An additional topic was the nature of the psychological mechanisms supporting recognition. Several general models of retrieval are contrasted. It is found that criterion bias models describe both the accuracy and latency of lexical decisions better than models which assume an ordered search of memory. An extension of the Random Walk decision process is presented. Within this model biases in decision often reflect differences in what is actually known about stimuli relative to a single criterion, rather than multiple and separate decision biases located at separate decision loci. The benefits of this analysis in explaining the effects of word frequency, stimulus repetition and typographic information in lexical decisions are discussed.

Doctor of Philosophy (PhD)