Previous research by A. I. Schulman argued that an observed systematic decline in recognition memory in long word lists was due to the build-up of input and output proactive interference (PI). It also suggested that input PI resulted from process automatization; that is, each list item was processed or encoded in much the same way, producing a set of highly similar memory traces. Based on Schulman's conclusions about uniform processing of a word list, the research described in this paper began with the hypothesis that requiring subjects to periodically change their processing strategy would slow the build-up of input PI. In two experiments using lists of 100 and 140 common nouns, the decline in recognition memory was prevented by asking subjects to rate each word, during presentation, on familiarity, imagery, size, or pleasantness. In addition, one rating task was found to be as effective as four rating tasks in preventing the decline in recognition. Both types of rating tasks led to ceiling effects which rendered the processing automatization hypothesis ambiguous. An additional experiment on output PI--the influence of early recognition tests on later tests--indicated that output PI does not significantly account for the decline in recognition memory. (Author/GDC)