In two experiments, subjects read or generated items at both encoding and retrieval. At test, they were required to decide whether or not the targets were presented initially (recognition), and if so, whether they were initially read or generated (judgments of origin). Recognition for items that were initially generated was enhanced if they were once again generated at test in the same context, but not if they were generated at test without context. These results confirm that memory for occurrence is facilitated by repetition of the initial encoding operations at retrieval. Generating at test resulted in an increase in "generate" responses both for items that were initially generated and for items that were initially read. Overall, there was a decrease in the accuracy of origin discriminations. It is suggested that, when subjects generate at test, they are likely to mistakenly attribute these just-performed operations to be part of the memory trace for that item.Understanding the cogrutive operations, or mental processes, by which we perceive and imagine events is a central focus of cognitive psychology. More recently, it has been argued that repetition of cognitive operations is a critical determinant, or perhaps, the determinant of remembering (Kolers, 1975(Kolers, , 1979Kolers & Roediger, 1984). The focus of this paper is to examine further the effect of repetition of cognitive operations on memory for occurrence and to begin to explore the effects of repetition of cognitive operations on memory for the origin, or source, of a remembered event.Glisky and Rabinowitz (1985) reported that repetition of the same cognitive operations during retrieval as those that were performed during encoding resulted in an increase in recognition. We required subjects to read target words or generate words from word fragments (e.g., AL-OHO-). At test, they were again required to read or generate the words before making recognition decisions. In addition to the standard generation effect of better memory for the items presented as fragments (Slamecka & Graf, 1978), we observed what we called a specific generation enhancement effect. For those items that were initially generated, recognition performance was improved if they were generated again, rather than read, at test. There was no effect of reading or generating at test for those items that were initially read. We therefore argued that a complete account of the generation effect must include a specific processing component. Records of cognitive operations are stored as part of the memory trace of the stimulus event Kolers, 1975Kolers, , 1979Russo & Wisher, 1976). Repeating the same operations at test as those that were performed at encoding facilitates access to the memory trace of which those operations are a part. This view is supported by two additional findings. First, a generation enhancement effect was not observed when the letters that were missing from the target at retrieval were different from the letters missing at encoding (AL-OH-L). Thus, the generation enhancem...