Levels of Memory Processing Research Paper

1. Basic Ideas

Craik and Lockhart’s (1972) main objection to the memory stores perspective was that the defining characteristics of the various stores did not appear to be constant from one situation to another. For example, both the capacity and the rate of forgetting associated with the short-term store varied as a function of the meaningfulness of the material held in the store. As an alternative formulation, Craik and Lockhart proposed that the primary functions of the cognitive system are the perception and understanding of incoming material, and that the formation of the memory trace is an incidental by-product of these primary processing operations. In this formulation there is no special ‘faculty’ of memory, and no memory stores as such; memory is simply a function of the processing carried out on perceived stimuli—for whatever reason.

The further suggestion was that deeper levels of processing were associated with longer lasting memory traces. This notion has its roots in the idea that the input side of the cognitive system is organized hierarchically, with early sensory analyses gradually developing into analyses of meaning, association, and implication. Specifically, Craik and Lockhart based their levels of processing (LOP) view of memory on Anne Treisman’s (1964) levels of analysis formulation of the processes of selective attention. Treisman proposed that incoming stimuli are subjected to a series of ‘tests,’ organized hierarchically, with early tests being concerned with analysis of sensory features and later tests dealing with identification of words and objects. All incoming stimuli pass the early sensory tests, implying that sensory information is analyzed and therefore available for later memory, but only a progressively smaller proportion of stimuli penetrate through the successive tests to a full analysis of identification and meaning (the mechanism of selective attention). Craik and Lockhart capitalized on this general set of ideas and added the further notion that deeper (i.e., more meaningful) levels of analysis were associated with semantically richer and more durable memory traces. One of Craik and Lockhart’s main points was therefore that the processes of memory and attention are intimately interlinked. Indeed, in their formulation the processes of perceiving, attending, understanding, and remembering are all aspects of the overall cognitive system. Memory encoding processes are simply those processes carried out essentially for the purposes of perception and comprehension; memory retrieval processes may be thought of as a reinstatement or recapitulation of some substantial proportion of the processes that occurred during encoding.

Two further points discussed by Craik and Lockhart (1972) should be mentioned. The first is the distinction between two types of rehearsal: one type functions by continuing to process encoded material at the same level of analysis, whereas the second involves operations that enrich and elaborate the material by carrying processing to deeper levels. In the original paper Craik and Lockhart referred to these two functions by the somewhat uninspired names of Type I and Type II rehearsal, but preferable terms are ‘maintenance’ and ‘elaborative rehearsal.’ If later memory is simply a function of the deepest level of analysis obtained, then memory performance should increase as a function of greater amounts of elaborative processing, but should be independent of the amount of maintenance processing. This prediction was borne out in the case of subsequent recall (Craik and Watkins 1973, Woodward et al. 1973) but, interestingly, not for recognition. In this latter case, greater amounts of maintenance rehearsal are associated with increased levels of recognition memory (Woodward et al. 1973). Apparently recognition, but not recall, is sensitive to some strengthening aspect of maintenance processing.

The second point from the original paper is that the distinction between ‘short-term’ and ‘long-term’ memory was maintained, but not in the form of separate memory stores. For Craik and Lockhart, short-term or ‘primary memory’ was synonymous with active processing of some qualitative aspect of a stimulus or small set of stimuli. Material ‘held in primary memory’ was thus held to be equivalent to attention paid to that material. Given that attentional resources can be deployed flexibly to a large variety of different processes, this formulation solves the problem of why material held ‘in the short-term store’ can be of many different qualitative types, including phonemic, lexical, semantic, and even imaginal information. Primary memory was thus seen as a large set of possible processing activities, rather than as the single structural store envisaged in the Atkinson and Shiffrin (1971) model.

2. Empirical Evidence

If memory is a function of the deepest level of processing obtained, it should not matter how that level of analysis is produced; for example, intention to learn or memorize the material should be irrelevant. This thought led to a series of experiments by Craik and Tulving (1975) in which words were processed to various depths by preceding each word by a question. These types of question (or ‘orienting tasks’) were designed so that the following word need only be processed to a shallow level (e.g., ‘is the word printed in upper case letters?’), to an intermediate level (e.g., ‘does the word rhyme with train?’), or to a relatively deep semantic level (e.g., ‘is the word a type of animal?’). In a typical experiment, 60 concrete nouns were each preceded by one question of this type, 20 concerning case, 20 rhyme, and 20 semantic, and with half of the questions associated with a ‘yes’ answer and half with a ‘no’ answer (e.g., the word TABLE preceded by ‘does the word rhyme with fable?’ or by ‘does the word rhyme with stopper?’). The encoding phase was then followed either by a recall test for the 60 words or by a recognition test in which the 60 target words were mixed randomly with 120 new words of a similar type.

Table 1 shows the results from two experiments in the Craik and Tulving series. For both recall and recognition memory performance was a function of depth of processing, but unexpectedly ‘yes’ answers in the initial encoding phase gave rise to higher levels of performance than did ‘no’ answers, for rhyme and semantic questions at least. Craik and Tulving suggested that this latter result reflected the greater degrees of elaboration associated with ‘yes’ answers. For example, the word TIGER would be elaborated more following the question ‘is the word a jungle animal?’ than following ‘is the word a type of furniture?’ That is, the compatible question serves to specify and perhaps enrich the encoding of the target word to a greater degree.

Further experiments in the Craik and Tulving (1975) paper showed that processing time had little effect on subsequent memory performance; type (or depth) of processing was much more important. Studying the words under intentional or incidental learning conditions also made little difference to the pattern of results. Finally, motivation differences appeared to be unimportant in such experiments, as varying the reward associated with specific words had no differential effect on performance.

3. Criticisms and Rebuttals

The LOP notions attracted a lot of attention in the 1970s, probably because the process-oriented perspective had been in the back of many researchers’ minds at that time. The ideas also drew criticisms, however, and excellent critical reviews were published by Nelson (1977) and Baddeley (1978); replies to these critical points were made by Lockhart and Craik (1990).

The major criticism was that the LOP ideas were extremely vague, and difficult to disprove; the ideas did not constitute a predictive theory. A linked criticism concerned the circularity of the concept of ‘depth.’ Given that no independent index of depth of processing had been proposed, it seemed all too easy to claim that any event that was well remembered must have been processed deeply. One answer to this point is that the LOP ideas were always intended to provide a framework for memory research, rather than a tight predictive theory. Thinking of memory in terms of mental processes that vary in terms of the qualitative types of information they represent, suggests different concepts and different experiments than those suggested by a structural viewpoint. The absence of an independent index of depth is certainly a drawback, although judges show good agreement when asked to rank the relative depths of a set of orienting tasks. One possibility is that neuroimaging or neurophysiological techniques may provide such an index (e.g., Kapur et al. 1994, Vincent et al. 1996).

Baddeley (1978) cited evidence to show that pictures appeared to be relatively insensitive to LOP manipulations, and suggested that the LOP ideas may be restricted to verbal materials. However, another way to interpret this result is that pictures are simply very compatible with our cognitive analyzing processes, and are therefore processed deeply and meaningfully regardless of the ostensible orienting task.

Some criticisms seemed well founded, and the LOP framework was modified accordingly (Lockhart and Craik 1990). For example, Craik and Lockhart’s original formulation had implied that incoming stimuli were analyzed in a constant linear fashion from shallow to deep processing. This seems unnecessarily restrictive, and a more realistic account would allow for interactive processing throughout the cognitive system, but with the resulting memory trace reflecting those processing operations that were carried out, regardless of the order in which they were achieved. A second point concerns the durability of shallow, sensory processing operations. The 1972 paper asserted that the results of such analyses were quite transient, in line with observations that ‘sensory memory traces’ decayed within a matter of seconds. However, subsequent work using paradigms of implicit or procedural memory showed that shallow processing operations can have extremely long-lasting effects when tested in sensitive ways (e.g., Kolers 1979). Apparently the transient effects of sensory analyses are associated with explicit memory and conscious recollection.

A third critical point requiring acknowledgment stems from observations of amnesic patients. Such people can certainly process information deeply in the sense that they can comprehend incoming material and make appropriate meaningful responses, yet they have little or no recollection of the material at a later time. It therefore seems that deep processing is a necessary but not sufficient correlate of good memory performance. Some further set of operations (associated perhaps with neurophysiological processes taking place in the hippocampus and medial temporal lobes of the cerebral cortex) are also necessary, and it is these latter operations that are impaired in amnesic patients.

A final criticism modifies the point that deep semantic processing is inevitably best for subsequent memory. Several theorists have made the point that there is no such thing as a universally ‘good’ encoding condition; rather, a specific encoding condition is good to the extent that it is compatible with the cues available in the later retrieval environment. This idea is captured in the notions of transfer-appropriate processing, repetition of operations, and the encoding specificity principle. As one illustration, Morris et al. (1977) showed that when the retrieval test involves rhyming cues, rhyme encoding operations were associated with higher levels of recognition than were semantic encoding operations. It should also be noted, however, that the combination of semantic encoding and semantic retrieval was associated with a substantially higher level of recognition than the combination of rhyme encoding and rhyme retrieval (0.68 vs. 0.40 averaged over Experiments 1 and 2). It therefore seems that any complete account of the psychology of memory must involve principles that capture both notions of depth of processing and transfer-appropriate processing.

This suggestion is illustrated by an experiment by Fisher and Craik (1977). They had subjects encode words either by rhyme or semantically, and then provided retrieval cues that were either identical, similar, or different from the way the word was originally encoded. For example, if the word DOG was encoded by rhyme (rhymes with log—DOG), the respective retrieval cues would be ‘rhymes with log—,’ ‘rhymes with frog—,’ ‘associated with cat—.’ Table 2 shows that both LOP and transfer-appropriate processing affect the pattern of results. That is, semantic encoding is generally superior to rhyme encoding and the more similar the encoding and retrieval cues, the better is performance. Also, the two manipulations interact in the sense that the benefits of deeper semantic processing are greater with compatible cues (or alternatively the effects of encoding-retrieval compatibility are greater at deeper levels of processing).

4. Further Developments

Since the 1970s the LOP ideas have been used in a wide variety of theoretical and applied contexts, from social cognition and personality theory to cognitive neuroscience (Lockhart and Craik 1990). It has been shown, for example, that deep semantic processing is reliably associated with activation of the left prefrontal cortex (Kapur et al. 1994). It has also been shown that associating words with the self results in particularly high levels of recollection, arguably because a person’s ‘self-schema’ is richly detailed and meaningful.

The central observation that deep processing results in good memory, is undeniable. The theoretical reason for this finding is less certain, but a reasonable proposal is that deeper levels of processing result in a record that is distinctive from other encoded events, and therefore more discriminable at the time of retrieval. A further reason is that the interconnected schematic structure of deeper representations facilitates the processes of reconstruction at the time of retrieval (Lockhart and Craik 1990). In any event it appears that the concept of depth of processing, or some similar concept, is a necessary one for any complete account of human memory processes.

Bibliography:

1. Atkinson R C, Shiffrin R M 1971 The control of short-term memory. Scientific American 225: 82–90
2. Baddeley A D 1978 The trouble with levels: A re-examination of Craik and Lockhart’s framework for memory research. Psychological Re iew 85: 139–52
3. Cermak L S, Craik F I M 1979 Le els of Processing in Human Memory. Erlbaum, Hillsdale, NJ
4. Challis B H, Velichkovsky B M 1999 Stratification in Cognition and Consciousness. John Benjamins, Amsterdam
5. Craik F I M, Lockhart R S 1972 Levels of processing: A framework for memory research. Journal of Verbal Learning and Verbal Beha ior 11: 671–84
6. Craik F I M, Tulving E 1975 Depth of processing and the retention of words in episodic memory. Journal of Experimental Psychology: General 104: 268–94
7. Craik F I M, Watkins M J 1973 The role of rehearsal in shortterm memory. Journal of Verbal Learning and Verbal Beha ior 12: 599–607
8. Fisher R P, Craik F I M 1977 Interaction between encoding and retrieval operations in cued recall. Journal of Experimental Psychology: Human Learning and Memory 3: 701–11
9. Kapur S, Craik F I M, Tulving E, Wilson A A, Houle S, Brown G 1994 Neuroanatomical correlates of encoding in episodic memory: Levels of processing effect. Proceedings of the National Academy of Sciences of the USA 91: 2008–11
10. Kolers P A 1979 A pattern-analyzing basis of recognition. In: Cermak L S, Craik F I M (eds.) Le els of Processing in Human Memory. Erlbaum, Hillsdale, NJ, pp. 363–84
11. Lockhart R S, Craik F I M 1990 Levels of processing: A retrospective commentary on a framework for memory research. Canadian Journal of Psychology 44: 87–112
12. Morris C D, Bransford J D, Franks J J 1977 Levels of processing versus transfer appropriate processing. Journal of Verbal Learning and Verbal Beha ior 16: 519–33
13. Nelson T O 1977 Repetition and levels of processing. Journal of Verbal Learning and Verbal Beha ior 16: 151–77
14. Treisman A 1964 Selective attention in man. British Medical Bulletin 20: 12–16
15. Vincent A, Craik F I M, Furedy J J 1996 Relations among memory performance, mental workload and cardiovascular responses. International Journal of Psychophysiology 23: 181–98
16. Woodward A E, Bjork R A, Jongeward R M 1973 Recall and recognition as a function of primary rehearsal. Journal of Verbal Learning and Verbal Beha ior 12: 608–17