Abstract
Unconscious response priming has been amply reported in studies adopting the dissociation paradigm, in which heavily masked prime stimuli modulate responses to targets (giving rise to a significant priming effect, mostly measured in reaction times, RT), whilst participants cannot classify the prime stimuli above chance level (e.g., measured in percent correct). The dissociation (significant priming effect vs. chance-level prime classification) is assumed to indicate that, even when participants are not consciously aware of the primes, they are nonetheless perceived and processed. Such purported dissociation, however, is predominantly based on a methodological problem of inequivalent metrics. We illustrated the problem by replicating and extending the behavioral part of Dehaene et al. (1998, Nature): In the masked priming experiment, one-digit numbers were visually presented as primes and targets. Participants judged whether target numbers were smaller or larger than 5. Subsequently, they were asked to ignore targets and classify heavily masked primes as smaller or larger than 5. The prime duration was manipulated by parametrically varying prime-mask stimulus onset asynchrony (SOA, from 10 milliseconds [ms] to 80 ms by 10 ms). At SOA=40 ms, a highly significant congruency priming effect (in RT difference) emerged, whereas prime classification (in percent correct) approximated chance-level. Upon closer inspection, however, no dissociation was found when the priming effect and prime classification performance was estimated in the same metric (e.g., in percent correct). Nor did a dissociation between priming task performance and prime classification performance arise at any other SOA. Furthermore, performance on prime classification increased monotonically along with increasing SOAs (from 20 ms upwards), which argues against chance-level prime classification at SOA=40 ms. In conclusion, performance dissociation, from which perception without awareness is to be derived, may purport to exist when the priming effect and prime classification is not compared in a shared metri
Meeting abstract presented at VSS 2015