Masked priming with graphemically related forms: Repetition or partial activation?

Form-priming occurs when a prime that is graphemically similar to the target word facilitates processing of the target. In an activation model (such as Morton's logogen model), such an effect can be interpreted as a partial-activation effect. A prime that shares letters with the target must inevitably produce activation in the detectors for both the prime and the target. Alternatively, form-priming could be seen as a special case of repetition-priming, in which the prime actually accesses the entry for the target. It is shown that masked-priming effects in the lexical decision task can be obtained for graphemically related pairs such as bontrast-CONTRAST, but not for four-letter pairs such as bamp-CAMP. It is suggested that the priming effect is controlled by neighbourhood density, short words usually having many neighbours, long words having very few. This hypothesis is supported by the finding that form-priming does occur for four-letter words if the prime and target are drawn from low-density neighbourhoods. For a partial-activation theory, an inhibitory mechanism that is sensitive to the number of prime-neighbours is required to explain the results. Of the several versions of a repetition account considered, the “best match” hypothesis appears to be the most promising: this assumes that priming is limited to the stimulus that best matches the prime. It is also shown that prime-target pairs that are related in form and meaning (e.g. made-MAKE) produce the same priming effect as identical pairs, as predicted by a repetition account that assumes a common entry underlying both forms.