Psycholinguistic theories of lexical access have to define the properties of sublexical and lexical representations. They need to specify the mapping from the speech signal information onto higher processing levels and need to set the limits that a lexical representation may deviate from the signal. Put differently, how far may a target representation deviate from the input to still be counted as matching rather than mismatching the input?
We addressed this question by manipulating the phonological similarity of pseudowords to a real word. Lexical status of the target, the task, and the relationship of prime and target were manipulated to determine the level at which effects of mismatching information become evident.
The same set of primes was used in four experiments. The degree of phonological relatedness to the intended word was varied in terms of phonological features (*domato, *somato, tomato). Prime (*domato) and target (TOMATO) were phonologically related in phonological priming experiments. In semantic priming experiments, the prime (*domato) was phonologically related to a semantic associate (tomato) of the target (PAPRIKA). Similar conditions were present when targets were pseudowords. Lexical decision or naming latencies were measured in a cross-modal priming situation.
In both the semantic and the phonological priming experiment, priming effects were obtained for word targets. Priming effects were larger with phonological priming than with semantic priming, but did not vary with degree of mismatch. Pseudoword decisions were accelerated only in the phonological priming experiment. In addition, the pattern of results questions whether appropriate baseline conditions were used in earlier mediated semantic priming experiments (Marslen-Wilson, 1993; Connine, Blasko, & Titone, 1993).
The finding suggests that phonological mismatch of the kind used here is not carried over to the semantic level. One could assume that a semantic representation is activated if its phonological representation exceeds a critical activation level. Gradual phonological activation will not result in gradual semantic activation.