How ‘ ‘ semantic ’ ’ is response priming restricted to practiced items ? A reply to Abrams & Grinspan ( 2007 ) q

Abrams and Grinspan (2007) provide some of the best evidence so far for the existence of unconscious perception. Indeed, they show that priming is still reliable even when partial awareness is impossible, that is when subjects cannot be conscious even of fragments of the prime. This demonstration adds to the growing body of work that provides evidence for the genuine existence of subliminal influences (see Kouider & Dehaene, 2007 for a review). This article by Abrams and Grinspan is of particular interest because past research has not adequately considered the problem of partial awareness. In several papers (Kouider & Dupoux, 2001, 2004, 2005), we have argued that although there exists compelling evidence for subliminal priming that involves perceptual, lexical and motor levels of processing, the existence of semantic priming remains unproven. Moreover, we have shown that some examples of masked semantic priming previously thought to be subliminal could rather be attributed to partial awareness (Kouider & Dupoux, 2004). In a Stroop priming task with masked color words as primes (e.g., GREEN) and colors as targets (rows of blue or green hashes), we used a situation in which priming occurs in the absence of global awareness (e.g., subjects could not discriminate between minimally different primes like GREEN and GENER). Though this phenomenon has previously been cited as proof of the existence of unconscious semantic priming (Cheesman & Merikle, 1986; Merikle & Joordens, 1997), we found in this paradigm that subjects were still partially aware of the primes (i.e., they could discriminate letters from pseudo-letters). Moreover, we also added a priming condition with pseudo-word primes (e.g., GENER) and showed that Stroop priming occurs equally as well as for real color words, demonstrating that priming was driven by fragments or letters. We hence hypothesized that when the stimulus set is small and over-repeated, subjects reconstruct the entire prime on the basis of fragments/letters; perceiving a few letters (G, R) is sufficient to reconstruct the entire word GREEN, and hence induces an apparent subliminal ‘‘semantic’’ priming effect. Abrams and Grinspan used an affective evaluation task in which subjects classified pleasant (e.g., smile) vs. unpleasant (e.g., tumor) target words preceded by congruent or incongruent prime words (i.e., from a same or different category). In earlier studies, Abrams and colleagues (Greenwald, Draine, & Abrams, 1996) showed congruity priming under conditions where participants could not perform the evaluation task on the prime, suggesting reliable semantic priming without awareness. However, Abrams and Greenwald (2000) later found that their own past findings could be reinterpreted as requiring no semantic mediation. Indeed, congruity priming occurred only if the prime words had been over-repeated as targets and thus extensively practiced.