Four experiments are described. The first three lend support to the assertion that retrieval from short-term storage (STS) is improved. possibly to a maximum, if items are recalled in their originally presented order. In the fourth experiment a modified recall condition was introduced in which written position of recall reflected order information. Although the subject was not constrained to recall the items in order under this modified recall condition. both item and order retention increased in comparison to both free and serial recall conditions. Within the theoretical framework adopted, the results indicate that retrieval from STS is improved by recalling in order; while long-term storage (LTS) is reduced by the constraint to recall in order. However, LTS is increased by the retention of order information when recalling in order is not required.Comparison of free and serial recall is illuminating. On the one hand, free recall has been used to advantage by Glanzer (1972) and others to study the effects of variables on long-and short-term storage (LTS and STS). A postlist reading task imposed between list presentation and recall is very effective in eliminating the STS (recency) effect (Glanzer, Gianutsos, & Dubin, 1969) and, therefore, permits the methodological isolation of STS and LTS components. This model operationally defines LTS by insertion of a postlist reading task designed to displace the to-be-recalled items from STS. Postman (1975) has been sharply critical of what he views as the circularity of this logic since it requires the assumption, or, to use Postman's term, "faith," that recall following a sufficiently long postlist task reflects LTS only. Specifically, the assumption is that similar items serve as displacers of items at the STS level. It is not necessary to make any assumptions as to the effectiveness of dissimilar items as displacers, which Postman rightly doubts.Generally, Postman (1975) is concerned that "the multiplicity and complexity of hypothetical mechanisms are outdistancing the observable facts" (p. 294). Sharing his concern about the reification of constructs, it should be noted that, although the present discussion is couched in terms of "stores" and "components," these are inferred processes and not things.Returning to the present analysis, the formula for computing and separating STS and LTS contributions is based on Raymond (1969) and is identical to that of Waugh and Norman (1965), except that Waugh and Norman estimated LTS recall from the middle positions of a list and did not utilize a postlist task. Various other authors have adopted the Waugh and Norman approach (viz., Craik & Levy, 1970;Glanzer, 1972), and it is in general use. Approaches to the problem of STS and LTS differentiation based on different assumptions and procedures than those adopted here have been proposed (Craik, 1968;Tulving & Colotla, 1970). However,the operational Raymond (1969) definition has the advantage that it can be generalized to serial and other kinds of recall tasks.On the other hand, orde...