Recoding strategies and the retrieval of information from memory

Abstract The slope of the function relating reaction time to the size of a positive target set in a recognition memory experiment increases when the probe is different in form from the remembered items. In an experiment in which S s were trained in a letter-number translation scheme, the slope was 150% greater when the positive set items were digits and the probe item was a letter or vice versa than when the positive set and the probe items were in the same form. In an experiment using S s who knew the binary—octal translation scheme, the slope was 130% greater when the probe was a binary triple than when it was an octal digit, indicating that both binary and octal numbers were retained as octal numbers. Other experiments allowed the rejection of an hypothesis that S s' memory of the positive set was disrupted by the act of translating the probe into the form of the remembered items, and the tentative rejection of an hypothesis that each item in a positive set is stored in two forms, one of which is compared with a probe more quickly than the other. The hypothesis most successful in accounting for the data claimed that S s translate each of the remembered items to match the probe item in form.