Graded lexical activation by pseudowords in cross-modal semantic priming:Spreading of activation, backward priming, or repair?

Spoken word recognition models treat mismatching sensory information differently. Mismatching information deactivates lexical entries according to one type of models. In another type of models, lexical entries are activated reflecting their degree of match with the input. This issue is mainly investigated with cross-modal semantic priming and lexical decision. This paradigm has been suspected of backward priming. I report three cross-modal priming experiments using naming and lexical decision to explore the sensitivity of word recognition to different degrees of match and to investigate the contribution of backward priming. Primes were pseudowords minimally (*baprika) or maximally (*zaprika) deviating from a word (paprika). A contribution of backward priming processes is likely. A repair of the mispronounced item as suggested by Marslen-Wilson (1993) seems untenable.