Level of processing and the process-dissociation procedure: elusiveness of null effects on estimates of automatic retrieval.

We describe two experiments that used the process-dissociation procedure to investigate the effects of level of processing on estimates of controlled and automatic retrieval processes in word-stem completion tasks. Despite our best endeavours, we found the null effect of level of processing on estimates of automatic retrieval reported by Toth, Reingold, and Jacoby (1994) elusive. Estimates of automatic retrieval were not independent of level of processing but inversely related to it. In part, the reason was that, following deeper levels of processing, instructions to exclude recollected words led to floor effects. But the inverse relationship persisted even when floor effects were avoided. Only participants who were not given strict instructions in the exclusion task-and who also qualified as lax responders based on answers in a structured post-test interview-showed no effect of level of processing on estimates of automatic retrieval. This null effect apparently occurred because these participants failed to exclude words that they in fact recollected from the study list. This finding violates the critical assumption that in this task participants exclude recollected words. The results are therefore paradoxical. Successful replication of the null effect occurred only under conditions that preclude the very use of the procedure. This paradox has important implications for views on how consciousness should be conceived in relation to memory.