Abstract Lockhart and Craik (l990) interpreted results of Morris, Bransford, and Franks (1977) as showing a of level of encoding. Further consideration of data, however, reveals that this supposed of study was confounded with of Recognition of identity was better than recognition of a rhyme relation. In present studies, subjects encoded words semantically or by rhyme and then judged unstudied words were related to studied words. A semantic or rhyme criterion for relation defined retrieval tasks. There was no of retrieval task, and of encoding was not dependent on retrieval task. When words may share either a semantic or a rhyme relation with studied words, advantage of semantic encoding is not dependent on retrieval criterion. Lockhart and Craik (1990) addressed many conceptual and empirical challenges to levels of processing framework for memory. One recurrent theme focussed on relative contributions of encoding and retrieval contexts to performance on memory tests. Whereas Lockhart and Craik recognized importance of interaction of encoding and retrieval, they framed critical question as whether there are also effects determined by form of encoding ... when transfer appropriateness of initial processing is held constant (p. 107). The results of Morris, Bransford, and Franks (1977) were interpreted as an instance of such a effect. The present paper has two related aims. The first aim, as suggested by subtitle of this article, is to show that Lockhart and Craik (1990) failed to consider all data relevant to their conclusion regarding of level of encoding. Specifically, there was no of level of encoding in Morris et al. (1977) studies; misinterpretation offered by Lockhart and Craik was due to their failure to consider of types of used by Morris et al. The second aim of present paper is to present new data that are consistent with hypothesis that level of encoding has a across semantic and rhyme retrieval tasks. The critical difference between procedures used in present experiments and those used by Morris et al. is that present retrieval tasks used unstudied items for both semantic and rhyme conditions. Were Lockhart and Craik (1990) on level in interpreting data of Morris et al. (1977) as showing a of level of encoding? In Morris et al. experiments, subjects judged words relative to a semantic or rhyme context. The encoding task was followed by either a standard yes-no recognition or a rhyme recognition in which subjects judged new words rhymed with an item presented during encoding task. The rationale was to assumption of transfer appropriate processing framework that the value of particular types of acquisition activities must be defined relative to type of activities to be performed at time of test (Morris et al., 1977, P. 531). Whereas a levels of processing framework leads to prediction of a of encoding, transfer of processing framework leads to prediction that semantic encoding might be worse than rhyme encoding for rhyme judgement task. Lockhart and Craik (1990) pooled results across two experiments in Morris et al. (1977) and reported that correct performance on a standard recognition following semantic encoding (M = .68) was higher than performance on a rhyme-recognition task following rhyme encoding (M = .40). Those two means, however, do not constitute a of encoding where term main effect refers to differences among marginal means of an independent variable in a factorial design (Keppel, 1991, p. 232). Rather, means selected by Lockhart and Craik represent conditions where encoding context matched retrieval context. …