Fragmentary knowledge and the processing-specific control of structural sensitivity.

Perruchet (1994) suggested that "neither a specific-item retrieval process nor abstractive capacity" is required to explain Vokey and Brooks (1992) results, which instead can be accounted for in terms of knowledge of item fragments. The literature contains 2 definitions of abstractive. Rejecting abstractive in the sense of nonliteral units requires rejecting the results of studies on changed-letter transfer, some of which do not need the control group Perruchet claims and thus stand, confirming nonliteral (relational) abstraction. However, the rejection of pooled across items abstraction even in the form of our proposed retrieval-time pooling is feasible, which we had not previously appreciated. Perruchet's rejection of item retrieval ignores Vokey and Brooks' coding and retrieval variations. The coding and retrieval results cannot be ignored simply because they are nonstandard, however. Integrating this literature with known memory phenomena requires incorporating these processing factors, rather than assuming fixed, literal fragments.