It was tentatively suggested that a long unsuccessful search on Trial n of a paired associate task might be associated with a higher probability of recall on Trial n + 1 than would a short search. The reasons for thinking this included the evidence that several test trials can be associated with good recall and that items found with difficulty in a retrieval task were later well recalled in a free recall task. The prediction was not confirmed, however, in either of two studies. Data are presented on the reduction of search time as learning progresses and on the distribution of subjective feeling-of-knowing states during learning. Subjects appeared either to "know" or "not know" the answer and rarely gave responses indicative of intermediate states of knowing.In paired associates learning, there are two reasons for thinking that a long search, which does not end in success, nevertheless might result in better acquisition of the response term when the subject is told it. First, Gardiner, Craik, and Bleasdale (1973) have shown that, in a free recall task, a word that is difficult to find might be better recalled, later, from long-term memory. In one of their studies, subjects were given 50 definitions of words plus the initial letters of these words and had to find the appropriate words within 60 sec. After all 50 words had been tested, the subjects were given an unexpected recall test of the words. It was found that words that had been discovered only with difficulty had a higher probability of being recalled in the fmal free recall. In particular, words that had been associated with a tip-of-the-tongue (TOT) experience, as rated subjectively by participants in the experiment, were recalled. One might try to extrapolate from this free recall study to paired associates learning: If a word is preceded by a long and unsuccessful search, might it not then be better retained than a word not preceded by a long search? Second, Izawa (1967) has shown that even though two test trials are associated with the same accuracy of performance, a study trial following two such test trials can be associated with a higher level of acquisition than can a study trial follOWing only one such test trial. In later work (Izawa, 1970), the effect was successfully extended to cover the case of as many as 19 test trials intervening between study trials. Izawa's explanation for Preparation of this article was supported by Natural Sciences and Engineering Research Council of Canada Grant AO-126 to the first author. We should like to express our gratitude to Anne Chenier and Fatima Kheraj for research assistance. We are also grateful to Keith Norton for helpful advice and to David Dockendorff for assistance with equipment and programming. Experiment 1 was carried out while the author held Social Sciences and Humanities Research Council Leave Fellowship 451-780198. the advantage of test trials for future performance was that during a test trial, even though no new items are found, a larger repertoire of possible responses is sampled and thus ma...