When a subject ( S ) is presented with a list of unrelated words for free recall, he usually retrieves the last few words in the list right away and then augments this terminal cluster with the recall of words from the beginning and middle of the list. The better recall of terminal items (the recency effect) can be handled by oneprocess models of memory (Melton, 1963) by postulating that an itemÊs strength or accessibility is very high immediately after presentation but falls off rapidly as further items are presented. In the last few years, however, several theorists have advocated two-process models of memory to describe the results of freerecall studies. For example, Waugh and Norman (1965) proposed that the last few words are retrieved from primary memory (PM) whereas earlier words are retrieved from secondary memory (SM) with greater difficulty. While generally accepting the PM/SM distinction, two-process theorists have themselves split into two camps: those postulating two stores (Atkinson & Shiffrin, 1968; Glanzer & Cunitz, 1966) and those who argue for one memory store but two retrieval processes (Tulving, 1968).