List-strength and list-length effects: Reply to Shiffrin, Ratcliff, Murnane, and Nobel (1993).

Abstract

R. M. Shiffrin, R . Ratcliff, K . Murnane, and P. Nobel (1993) claimed that TODAM (a theory of distributed associative memory) is unable simultaneously to predict an absent (or negative) liststrength effect (LSE) and a positive list-length effect (LLE). However, Shiffrin e t al. failed t o distinguish between situations in which lag (number of items intervening between study and test) is controlled and situations in which it is not. We stand by our previous conclusion; TODAM can explain why there is little or no LSE when at the same time there is an LLE when the LLE is studied under the standard conditions. To our knowledge there are no published studies where lag has been controlled. However, this simplified version of TODAM cannot explain an LLE when a scoring window i s used. Whether such a result would be inconsistent with a more complete version of TODAM remains to be seen. Murdock and Kahana (1993) derived TODAM (a theory of distributed associative memory) predictions for the liststrength effect (LSE), the list-length effect (LLE), and the d! from a study-test paradigm and showed that, with reasonable parameter values, the predicted results were in good agreement with the general pattern of data in the literature. They assumed that memory carried over from the preexperimental to the experimental situation and from list to list; they called this assumption the continuous memory assumption. Shiffrin, Ratcliff, Murnane, and Nobel (1993) claimed that they had analyzed this model and rejected it because it is unable, simultaneously, to predict an absent (or negative) LSE and a positive LLE. Shiffrin et al. reiterated this point with an informal analysis of TODAM, which was buttressed by a simulation using the Murdock and Kahana (1993) parameter values. The main problem with the analysis of Shiffrin et al. (1993) is that, for the LLE, they failed to distinguish between the situations where lag is controlled and the situations where it is not. (By lag we mean the number of items intervening between the study of an item and its test.) Shiffrin e t al.'s analysis applies to the former case, whereas our analysis applies to the latter case. It is important to understand that in studies of the LLE, lag is generally not controlled; instead, in the usual study-test paradigm all list items are tested, and there are an equal number of old and new items in the test phase. Items are randomly ordered in the test phase, so perforce lag and list length must be confounded. In fact, to our knowledge, there are no published studies of the LLE where lag is properly controlled.