Abstract
Using metacontrast masking, we observed robust response priming from stimuli that could not be consciously identified. When the masked primes matched the imperative target, responses were faster and more accurate. Past research has explained these unconscious priming effects in terms of the overlap between prime and target response codes. However, using a 4:2 stimulus response mapping, we found that these priming effects were largely contingent on a physical match between the prime and target stimuli. Thus, unconscious priming effects were mostly stimulus-gated, and were only weakly influenced by stimulus-independent response codes. Further experiments suggest that stimulus-gating of unconscious priming was determined only by the response-relevant feature of the prime and target stimuli; the overlap between response-irrelevant features of primes and targets did not modulate priming. Finally, we examined the degree to which unconscious priming effects could be flexibly modulated by top-down settings. Top-down selection of the response-relevant features elicited trial-by-trial changes in unconscious priming effects. By contrast, top-down selection of the relevant stimulus-response rules had virtually no effect. We suggest that unconscious priming effects are caused by stimulus-driven retrieval of specific S-R associations. Top-down control can determine which stimuli will trigger this retrieval process, but not which S-R links will be expressed for a given stimulus.