| Summary: | In this thesis we show that lexical decisions (LDT) to isolated words are slowed when a target word has either, many orthographic competitors (defined as words sharing outer letter frames), or a single strong competitor (defined as words sharing all letters, e.g. from/form). This result is not found for naming latencies. It is shown that naming is more sensitive to variables which concern the mapping of orthography to phonology, than to purely orthographic measures. In contrast to these findings Andrews (1989, 1992) found that the existence of orthographic competitors in the form of neighbours (Coltheart, Davelaar, Jonasson and Besner (1977)), facilitated both LDT and naming responses. This suggests that neighbours conspire rather than compete for recognition. We repeat Andrews' experiments and replicate the conspiracy effect of neighbourhood size for low frequency words. Andrews' results replicate however, only for LDT and not for naming. In a further investigation of neighbourhood effects we measure eye fixations to target words in neutral sentence contexts. Here we find an effect of neighbourhood size for both high and low frequency words. The effect is, however, one of conspiracy for high frequency words, and one of competition for low frequency words. We describe an activation time course model of word processing which we argue can account for both competition and conspiracy effects, and for the differences in effects for high and low frequency words. Neighbourhood influences are less robust for naming than for LDT or sentence reading. This may be because isolated word naming relies less on lexical processing than the other two tasks. Whether naming responses are produced lexically, or by strict grapheme-phoneme correspondence rules, will depend on the proportion of spelling-sound regular and irregular words in the stimulus list. We demonstrate the influence of stimulus list structure on naming by comparing data from the Seidenberg and Waters (1989) Mega Study with data from smaller studies. For tasks which demand more lexical processing, such as LDT, we demonstrate that the influence of stimulus structure will depend on the disparity between lexical knowledge and local information obtained from the stimulus list. Stimulus list structure is most influential when it gives little, or no, reflection of real lexical experience.
|
|---|
