Abstract
Figure-ground perception arises from competition for ownership of a shared edge. Peterson et al. (2000) proposed that configural processes operating on opposite sides of an edge inhibit each other; inhibition of the more weakly cued side accounts for its status as a shapeless ground. Testing for inhibition, Peterson and Kim (2001) asked observers to decide if line drawing targets portrayed real or novel objects. Targets were preceded by unmasked novel silhouette primes. For half the real targets, the contours of the prime sketched a portion of the same basic level object on the ground side (primed targets). For the other half, the contours of the prime did not sketch a real object (unprimed targets). All the novel targets were unprimed. Object decision latencies were longer for primed than for unprimed real targets, consistent with the proposal that object memories matching the ground side of prime silhouettes were inhibited. However, it remained to be shown that when the same portions of real objects were seen as figures rather than as grounds in the primes, their memories were not inhibited, a necessary prediction of the model. We altered Peterson and Kim's (2001) prime silhouettes so that (1) regions seen as grounds were seen as figures and (2) half the primes portrayed portions of real objects, the other half portrayed portions of novel objects; prime type did not predict the target type. For both real and novel targets, primed trials were those on which prime type matched target type; unprimed trials were mismatches. For real targets, the match occurred at the basic level; for novel targets, the match occurred at the superordinate level. In Exps. 1 & 2, using different SOAs, object decisions for real targets were speeded on primed trials, whereas those for novel targets were unaffected by prime type, p < .05. In Exp. 3, we used masked primes and tested for priming when match and mismatch were defined semantically for real objects.
Support: NSF BCS 9906063 to MAP.