One of the most consistent findings in short-term memory (STM) research is that when a list of items is presented for immediate recall, and performance is measured as a function of input serial position, recall is characterised by a bow-shaped curve whereby the beginning of the list is well recalled (the primacy effect), and the end of the list is recalled best of all (the recency effect). These twin phenomena require to be explained by any theoretical account of STM function although the crucial role occupied by recency in current theoretical debate demands that it receives particular attention. A review of the relevant literature suggests that three broad classes of model can be identified which differ with regard to interpretations of one or other or both of the above effects. These are: (i) The modal model (eg Waugh and Norman 1965; Atkinson and Shiffrin 1968; Glanzer 1972) which explain the two effects in terms of the operation of two distinct memory stores; (ii) Levels of processing (Craik and Lockhart 1972; Craik and Tulving 1975) where the emphasis is placed on differences in the nature of the encoding operations devoted to early and late items; (iii) A recent sub-group of theories concerned only with recency, and which view the phenomenon as the operation of a specific retrieval strategy. The experimental investigation began by discussing the apparent invariance of immediate recall recency in, for example free recall, and suggested that this could be the result of an inbuilt bias in the processing of fixed length lists. In contrast evidence was presented from experiments using a running memory span paradigm (Experiments 1 and 2) which demonstrated that terminal item recall can be affected by such factors as memory load, presentation rate, and processing strategy. The results are interpreted as showing that recency effects can reflect the output of an active processor/rehearsal mechanism given appropriate task demands. A second series of experiments considered a second highly replicable phenomenon: the elimination of recency by interpolated activity. In particular attention was paid to the effects of varying the class of the distractor material (which did not have to be memorised) upon list recall. As the direction of any such influence was also considered important, a return to fixed length lists was demanded. The results showed that detrimental effects attributable to class similarity were uni-directional i.e. they were present when the distractor followed the memory list (Experiment 3) but absent when the paradigm was reversed (Experiment 4). When these occurred in the post-list delay condition however, they were observed to be uniform across all serial positions. In contrast, a special condition where both memory and distractor lists were identical showed enhanced recall regardless of whether the distractor followed (post-list delay1 or came before (pre-list condition), the memory items. These improvements to recall were restricted predominantly to primacy. It was argued that while the results of the delay task study are explicable in terms of most major theories ( especially the modal model) the findings from the pre-list experiment argue against positions which seek to explain all primacy effects by differential rehearsal or elaborate encoding. It was concluded that in order to account for the particular survival of the early portion of a set of stimuli which does not require to be memorised, some appeal to a concept of 'primacy salience' seems necessitated. A further series of experiments (Experiments 5, 6, 7 and 8) critically examined another aspect of the apparent invariance of serial position functions and demonstrated that when conceptually related stimuli are grouped together during input of a memory list, consistent deviations from the standard recall profile are obtained. These are characterised by durable, within-list primacy effects, contingent upon the beginning of each block, and were evident whether the block represented a brand new category or merely a subset of a previously-presented category (Experiments 5 and 6). The effects were considered particularly amenable to explanation in terms of versions of the modal model which incorporate control processes responsive to inbuilt organisational features of the material. While it could be argued that explanation is also possible via a levels framework, it was suggested that this would result in an account virtually indistinguishable from dichotomous positions it claims to supercede. Additional experiments with the same basic paradigm (Experiments 7 and 8), but where list length was increased to 60 items, revealed evidence of two co-occurring recency effects: (i) the typical 'immediate effect' spanning terminal items, and (ii) a parallel effect embracing terminal groups of related items. The presence of this second effect in immediate recall effectively ruled out explanation via either of the general models, ( a) because of the large number of intervening items involved which precludes the operation of a fixed capacity mechanism, and (b) because the effect's obvious dependence upon semantic features refutes any account of recency based on shallow phonemic processing. Only those models which view recency as the product of a retrieval strategy seemed capable of handling these data comfortably. Moreover such theories, which place emphasis on the 'appropriateness of the units' over which such a strategy can be applied when these are defined from the point of the recall test, received further support from tentative evidence of a list recency effect. The final set of experiments (Experiments 9 and 10) employed a final free recall paradigm with the dual purpose of examining the negative recency effect for items (Craik 1970) and simultaneously extending consideration of the recency effect for lists encountered in Experiment 8. While evidence on the former phenomenon indicated a somewhat unreliable phenomenon, the finding of list recency in final recall was confirmed as being both reliable and substantial. Furthermore, such effects were independent of the nature of the list material, and therefore presumably determined by the organisation imposed on each list during initial learning and the position the list occupied within the series. Further evidence from cued final recall emphasised retrieval aspects of the process involved. Based on all the evidence accumulated during the course of the investigations several conclusions were drawn: (i) Recency effects cannot be viewed as a unitary phenomenon: At least two processes require to be postulated to handle all the relevant data. One of these involves the output from an active processor/rehearsal mechanism, while the second reflects the operation of a retrieval strategy conducted on proximal units at the time of recall. As currently formulated, only one model accepts the need to explain recency effects by recourse to more than one mechanism (Craik and Jacoby 1975). (ii) By a similar token, an adequate account of primacy can only be given by assuming that two processes are responsible. The first of these again implicates an active processor which devotes differential attention to the initial items in any series of to-be-remembered material, while the second assumes that the beginning of any well-defined series which the subject is set to receive will achieve some privileged storage status. The basis of this second process however, is not fully understood. (iii) One inevitable corollary of conclusions (i) and (ii) is that no model of STM function currently in existence can encompass all the findings adequately. With reference to the effects obtained, the final section offers some comments concerning the particular weaknesses inherent in certain formulations.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:470276 |
| Date | January 1979 |
| Creators | Rejman, Michael Hunter |
| Publisher | University of Stirling |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
| Source | http://hdl.handle.net/1893/24391 |
Page generated in 0.0032 seconds