Craik_Lockhart_1972

JOURNAL OF VERBAL LEARNING AND VERBAL BEHAVIOR 11, 671-684 (1972)Levels of Processing: A Framework for Memory Research 1 FERGUS I. M. CRAIK AND ROBERTS. LOCKHART University of Toronto, Toronto 181, Ontario, Canada This paper brieflyreviewsthe evidencefor multistore theories of memory and points out some difficultieswith the approach. An alternative framework for human memory researchis then outlined in terms of depth or levelsof processing.Some current data and argumentsare reexamined in the light of this alternative framework and implications for furtherresearch considered.Over the past decade, models of human store models, (b) question their adequacy, andmemory have been dominated by the concept (c) propose an alternative framework in termsof stores and the transfer of information of levels of processing. We will argue that theamong them. One major criterion for distin- memory trace can be understood as a by-guishing between stores has been their different product of perceptual analysis and that traceretention characteristics. The temporal pro- persistence is a positive function of the depthperties of stored information have, thus, to which the stimulus has been analyzed.played a dual role: Besides constituting the Stimuli may also be retained over short inter-basic phenomenon to be explained, they have vals by continued processing at a constantalso been used to generate the theoretical depth. These views offer a new way to interpretconstructs in terms of which the explanation existing data and provide a heuristiciframe-is formulated. The apparent circularity has work for further research.been avoided by the specification of additionalproperties of the stores (such as their capacity MULTISTOREMODELSand coding characteristics) thereby character- The Case in Favorizing them independently of the phenomenato be explained. The constructs, thus formu- When man is viewed as a processor of in-lated, have been used to account for data formation (Miller, 1956; Broadbent, 1958), itacross a variety of paradigms and experimental seems necessary to postulate holding mechan-conditions. The essential concept underlying isms or memory stores at various points in thesuch explanations is that of information being system. For example, on the basis of histransferred from one store to another, and the dichotic listening studies, Broadbent (1958)store-to-store transfer models may be distin- proposed that information must be heldguished, at least in terms of emphasis, from transiently before entering the limited-capacityexplanations which associate different reten- processing channel. Items could be held overtion characteristics with qualitative changes the short term by recycling them, afterin the memory code. perception, through the same transient storageIn the present paper we will do three things: system. From there, information could be(a) examine the reasons for proposing multi- transferred into and retained in a more a This research was supported by Grants A8261 and permanent long-term store. Broadbent's ideasA0355 from the National Research Council of Canada have been developed and extended by Waughto the first and second author, respectively.We thank and Norman (1965), Peterson (1966), andour colleagues who read a preliminary version of the Atkinson and Shiffrin (1968). According topaper and made many helpful suggestions. the modal model (Murdock, 1967), it is now 671\"2opyright © 1972 by Academic Press, Inc.All rights of reproduction in any form reserved.

672 CRAIK AND LOCKHARTwidely accepted that memory can be classified guished from sensory memories by virtue ofinto three levels of storage: sensory stores, its limited capacity (Miller, 1956; Broadbent,short-term memory (STM) and long-term 1958), by the finding that information is lostmemory (LTM). Since there has been some principally by a process of displacementambiguity in the usage of terms in this area, (Waugh & Norman, 1965), and by the slowerwe shall follow the convention of using STM rate of forgetting from STS: 5-20 seconds asand LTM to refer to experimental situations, opposed to the 1-2-second estimates forand the terms \"short-term store\" (STS) and sensory storage. While most research has\"long-term store\" (LTS) to refer to the two concentrated on verbal STS, there is evidencerelevant storage systems. that more literal \"representational\" inform- ation may also be held over the short term Stimuli can be entered into the sensory stores (Posner, 1967), although the relationshipregardless of whether or not the subject is between such modality-specific stores and thepaying attention to that source; that is, verbal STS has not been made clear.sensory stores are \"preattentive\" (Neisser,1967). The input is represented in a rather The distinctions between STS and LTS areliteral form and can be overwritten by further well-documented. Whereas STS has a limitedinputs in the same modality (Neisser, 1967; capacity, LTS has no known limit; verbalCrowder & Morton, 1969). Further features items are usually coded phonemically in STSwhich distinguish the sensory registers from but largely in terms of their semantic featureslater st~bres are the modality-specific nature in LTS (Baddeley, 1966); forgetting fromand moderately large capacity of sensory STS is complete within 30 seconds or lessstores and the transience of their contents. while forgetting from LTS is either very slow or the material is not forgotten at all (Shiffrin AtteOtion to the material in a sensoryregister & Atkinson, 1969). In the free-recall paradigm,is equivalent to reading it out and transferring it is generally believed that the last few itemsit to STS. Here, verbal items are coded in are retrieved from STS and prior items aresome phonemic fashion (Shulman, 1971) or retrieved from LTS; it is now known thatin auditory-verbal-linguistic terms (Atkinson several variables affect one of these retrieval& Shiffrin, 1968). The STS is further distin- TABLE 1COMMONLYACCEPTEDDIFFERENCES BETWEENTHE THREE STAGESOF VERBALMEMORY(SEE TEXT FOR SOURCES)Feature Sensory r e g i s t e r s Short-term store Long-term storeEntry of information Preattentive Requires attention Rehearsal Continued attention RepetitionMaintenance of information Not possible Rehearsal Organization Phonemic Largely semanticFormat of information Literal copy of input Probably visual Some auditory and visual Possibly semanticCapacity Large Small No known limitInformation loss Decay Displacement Possibly no loss Possibly decay Loss of accessibility orTrace duration k-2 SecondsRetrieval Readout Up to 30 seconds discriminabilityby inter- Probably automatic ference Items in consciousness Minutes to years Temporal/phonemic cues Retrieval cues Possiblysearch process

LEVELS OF PROCESSING 673components without affecting the other Capacity(Glanzer, 1972). Further persuasive evidencefor the STS/LTS dichotomy comes from Although limited capacity has been a majorclinical studies (Milner, 1970; Warrington, feature of the information flow approach,1971). The distinguishing features of the three and especially a feature of STS in multistorestorage levels are summarized in Table I. models, the exact nature of the capacity limitation is somewhat obscure. In particular, The attractiveness of the \"box\" approach it has been unclear whether the limitation isis not difficult to understand. Such multistore one of processing capacity, storage capacity,models are apparently specific and concrete; or is meant to apply to some interaction be-information flows in well-regulated paths tween the two. In terms of the computerbetween stores whose characteristics have analogy on which information flow modelsintuitive appeal; their properties may be are based, the issue is whether the limitationelicited by experiment and described either refers to the storage capacity of a memorybehaviorally or mathematically. All that register or to the rate at which the processorremains, it seems, is to specify the properties can perform certain operations. The notion ofof each component more preciselyand to work a limited-capacity channel (Broadbent, 1958)out the transfer functions more accurately. appears to emphasize the second interpreta- tion while later models of memory, such as Despite all these points in their favor, when that of Waugh and Norman (1965), appear tothe evidence for multistore models is examined favor the storage interpretation. Both inter-in greater detail, the stores become less pretations are present in Miller (1956) but thetangible. One warning sign is the progressively relationship between the two is not explicitlygreater part played by \"control processes\" in worked out.more recent formulations (for example,Atkinson & Shiffrin, 1971). In the next section Attempts to measure the capacity of STSwe consider the adequacy ofmultistore notions have leant towards the storage interpretation,more critically. and considered number of items to be the appropriate scale of measurement. SuchThe Case Against attempts have provided quite a range of values. For example, recent estimates of The multistore approach has not been primary memory size (Baddeley, 1970; Mur-without its general critics (Melton, 1963; dock, 1972) have yielded values between twoMurdock, 1972). Other workers have objected and four words. However, measures ofto certain aspects of the formulation. For memory span (which have been said to reflectexample, Tulving and Patterson (1968) argued the limited capacity of the STM box) areagainst the notion of information being typically between five and nine items, depend-transferred from one store to another. ing on whether the items in question are words,Similarly, Shallice and Warrington (1970) letters or digits (Crannell & Parrish, 1957).presented evidence against the idea that Finally, if the words in a span test form ainformation must necessarily \"pass through\" sentence, young subjects can accuratelySTS to enter LTS. reproduce strings of up to 20 words (Craik & Masani, 1969). Thus, if capacity is a critical In our view, the criteria listed in the previous feature of STM operation, a box model has tosection do not provide satisfactory grounds account for this very wide range of capacityfor distinguishing between separate stores. estimates.The adequacy of the evidence will be con-sidered with reference to the concepts of The most widely accepted explanation ofcapacity, coding, and finally, the retention this variation is that capacity is limited infunction itself.

674 CRAIK AND LOCKHARTterms of chunks, and that few or many items an important characteristic by which one storecan be recoded into a chunk depending on the is distinguished from another.meaningfulness of the material. Apart fromthe difficultyof defining a chunk independently We will argue that the coding question isfrom its memorial consequences, this view more appropriately formulated in terms of theentails a rather flexible notion of STS as a processing demands imposed by the experi-storage compartment which can accept a mental paradigm and the material to bevariety of codes from simple physical features remembered. In some paradigms and withto complex semantic ones. certain material, acoustic coding may be either adequate or all that is possible. In From the standpoint of the present paper, other circumstances processing to a semanticthe concept of capacity is to be understood in level may be both possible and advantageous.terms of a limitation on processing; limitationsof storage are held to be a direct consequence Forgetting Characteristicsof this more fundamental limitation. If memory stores are to be distinguished inCoding terms of their forgetting characteristics, a minimal requirement would seem to be that Working with verbal material, Conrad the retention function should be invariant(1964) and Baddeley (1966) provided one across different paradigms and experimentalplausible basis for distinguishing STS and conditions. While this invariance has not beenLTS. They concluded that information in STS rigorously tested, there are cases where itwas coded acoustically and that coding was clearly breaks down. We will give twopredominantly semantic in LTS. Further examples. First, in the finite-state models ofresearch has blurred this distinction, how- paired-associate learning, the state commonlyever. First, it has been shown that STS coding identified as STS shows forgetting character-can be either acoustic or articulatory (Levy, istics which are differentfrom those established1971; Peterson & Johnson, 1971). Second, for STS in other paradigms (Kintsch, 1970,recent papers by Kroll and his colleagues p. 206). In the former case, STS retention(Kroll et al., 1970) have demonstrated that extends over as many as 20 intervening itemseven with verbal material, STS can sometimes while in the free-recall and probe paradigmsbe visual. Apparently STS can accept a variety (Waugh & Norman, 1965), STS informationof physical codes. is lost much more rapidly. As a second example, the durability of the memory trace for Can STS also hold semantic information? visual stimuli appears to depend on the mat-The persistence of contradictory evidence erial and the paradigm. According to Neissersuggests either that the question has been (1967), the icon lasts 1 second or less, Posnerinappropriately formulated or that the answer (1969) and his colleagues have found evidencedepends on the paradigm used. When tradi- for visual persistence of up to 1.5 seconds,tional STM paradigms are considered, the while other recent studies by Murdock (1971),answer seems to be \"no\" (Kintsch & Buschke, Phillips and Baddeley (1971) and by Kroll et1969; Craik & Levy, 1970), although Shulman al. (1970) have yielded estimates of 6, 10, and(1970, 1972) has recently presented persuasive 25 seconds, respectively. Estimates are evenevidence in favor of a semantic STS. While longer in recognRion memory for picturestype of coding may originally have seemed a (Shepard, 1967; Haber, 1970). Given that wegood basis for the distinction between short- recognize pictures, faces, tunes, and voicesterm and long-term memory, the distinction after long periods of time, it is clear that weno longer appears satisfactory. A defender of have long-term memory for relatively literalthe multistore notion might argue that STS nonverbal information. Thus, it is difficult tocoding is flexible, but this position removes

LEVELS OF PROCESSING 675draw a line between \"sensory memory\" and with matching the input against stored abstrac-\"representational\" or \"pictorial\" memory. tions from past learning; that is, later stages are concerned with pattern recognition and We will argue that retention depends upon the extraction of meaning. This conceptionsuch aspects of the paradigm as study time, of a series or hierarchy of processing stages isamount of material presented and mode of often referred to as \"depth of processing\"test; also upon the extent to which the subject where greater \"depth\" implies a greater degreehas developed systems to analyze and enrich of semantic or cognitive analysis. After theparticular types of stimuli; that is, the familiar- stimulus has been recognized, it may undergoity, compatibility, and meaningfulness of the further processing by enrichment or elabor-material. ation. For example, after a word is recognized, it may trigger associations, images or stories Although we believe that the multistore on the basis of the subject's past experienceformulation is unsatisfactory in terms of its with the word. Such \"elaboration coding\"capacity, coding, and forgetting character- (Tulving & Madigan, 1970) is not restricted toistics, obviously there are some basic findings verbal material. We would argue that similarwhich any model must accommodate. It seems levels of processing exist in the perceptualcertain that stimuli are encoded in different analysis of sounds, sights, smells and so on.ways within the memory system: A word may Analysis proceeds through a series of sensorybe encoded at various times in terms of its stages to levels associated with matching orvisual, phonemic, or semantic features, its pattern recognition and finally to semantic-verbal associates, or an image. Differently associative stages of stimulus enrichment.encoded representations apparently persist fordifferent lengths of time. The phenomenon of One of the results of this perceptual analysislimited capacity at some points in the system is the memory trace. Such features of the traceseems real enough and, thus, should also be as its coding characteristics and its persistencetaken into consideration. Finally, the roles thus arise essentially as byproducts of percep-of perceptual, attentional, and rehearsal tual processing (Morton, 1970). Specifically,processes should also be noted. we suggest that trace persistence is a function of depth of analysis, with deeper levels of One way of coping with the kinds of in- analysis associated with more elaborate,consistencies we have described is to postulate longer lasting, and stronger traces. Since theadditional stores (see, Morton, 1970; Sperling, organism is normally concerned only with1970). However, we think it is more useful to the extraction &meaning from the stimuli, it isfocus on the encoding operations themselves advantageous to store the products of suchand to consider the proposal that rates of deep analyses, but there is usually no need toforgetting are a function of the type and depth store the products of preliminary analyses.of encoding. This view is developed in the It is perfectly possible to draw a box aroundnext section. early analyses and call it sensory memory and a box around intermediate analyses called LEVELS OF PROCESSING short-term memory, but that procedure both oversimplifies matters and evades the more Many theorists now agree that perception significant issues.involves the rapid analysis of stimuli at anumber of levels or stages (Selfridge & Neisser, Although certain analytic operations must1960; Treisman, 1964; Sutherland, 1968). precede others, much recent evidence suggestsPreliminary stages are concerned with the that we perceive at meaningful, deeper levelsanalysis of such physical or sensory features before we perceive the results of logicallyas lines, angles, brightness, pitch, and loud- prior analyses (Macnamara, 1972; Savin &ness, while later stages are more concerned

676 CRAIK AND LOCKHARTBever, 1970). Further elaborative coding does limited-capacity central processor which maynot exist in a hierarchy of necessary steps be deployed in a number of different ways.and this seems especially true of later pro- If this processing capacity is used to maintaincessing stages. In this sense, \"spread\" of information at one level, the phenomena ofencoding might be a more accurate descrip- short-term memory will appear. The processortion, but the term \"depth\" will be retained as itself is neutral with regard to coding char-it conveys the flavor of our argument. acteristics: The observed PM code will depend on the processing modality within which the Highly familiar, meaningful stimuli are processor is operating. Further, while limitedcompatible, by definition, with existing cogni- capacity is a function of the processor itself,tive structures. Such stimuli (for example, the number of items held will depend upon thepictures and sentences) will be processed to a level at which the processor is operating. Atdeep level more rapidly than less meaningful deeper levels the subject can make greater usestimuli and will be well-retained. Thus, speed of learned rules and past knowledge; thus,of analysis does not necessarily predict reten- material can be more efficiently handled andtion. Retention is a function of depth, and more can be retained. There is apparentlyvarious factors, such as the amount of atten- great variability in the ease with which inform-tion devoted to a stimulus, its compatibility ation at different levels can be maintainedwith the analyzing structures, and the pro- in PM. Some types of information (forcessing time available, will determine the depth example, phonemic features of words) areto which it is processed. particularly easy to maintain while the main- tenance of others (such as early visual analyses Thus, we prefer to think of memory tied to --the \"icon\") is apparently impossible.levels of perceptual processing. Although theselevels may be grouped into stages (sensory The essential feature of PM retention isanalyses, pattern recognition, and stimulus that aspects of the material are still beingelaboration, for example) processing levels processed or attended to. Our notion of PM is,may be more usefully envisaged as a con- thus, synonymous with that of James (1890)tinuum of analysis. Thus, memory, too, is in that PM items are still in consciousness.viewed as a continuum from the transient When attention is diverted from the item,products of sensory analyses to the highly information will be lost at the rate appropriatedurable products of semantic-associative to its level of processing--slower rates foroperations. However, superimposed on this deeper levels. While PM retention is, thus,basic memory system there is a second way in equivalent to continued processing, this typewhich stimuli can be retained--by reeirculat- of processing merely prolongs an item's highing information at one level of processing. In accessibility without leading to formation ofour view, such descriptions as \"continued a more permanent memory trace. This Type Iattention to certain aspects of the stimulus,\" processing, that is, repetition of analyses\"keeping the items in consciousness,\" \"hold- which have already been carried out, may being the items in the rehearsal buffer,\" and contrasted with Type II processing which\"retention of the items in primary memory\" involves deeper analysis of the stimulus. Onlyall refer to the same concept of maintaining this second type of rehearsal should lead toinformation at one level of processing. To improved memory performance. To the extentpreserve some measure of continuity with that the su~ect utilizes Type II processing,existing terminology, we will use the term memory will improve with total study time,primary memory (PM) to refer to this oper- but when he engages in Type I processing, theation, although it should be noted that our \"total time hypothesis\" (see Cooper & Pantle, usage is more restricted than the usual one. 1967) will break down. Stoff and Eagle (1971) We endorse Moray's (1967) notion of a

LEVELS OF PROCESSING 677have reported findings in line with this experimenter has a control over the processingsuggestion. the subject applies to the material that he doesTo summarize, it is suggested that the not have when the subject is merely instructedmemory trace is better described in terms to learn and uses an unknown coding strategy.of depth of processing or degree of stimulus We will consider several examples whichelaboration. Deeper analysis leads to a more illustrate this point. Tresselt and Mayznerpersistent trace. While information may be (1960) tested free recall after incidental learn-held in PM, such maintenance will not in itself ing under three different orienting tasks:improve subsequent retention; when attention crossing out vowels, copying the words, andis diverted, information is lost at a rate which judging the degree to which the word was andepends essentially on the level of analysis. instance of the concept \"economic\". Under the last condition, the number of words re- EXISTING DATA REEXAMINED called was four times higher than that of the first and twice that of the second condition.Incidental Learning Similar results using the free-recall paradigmWhen memory traces are viewed as the have been obtained by Hyde and Jenkinsproduct of a particular form of processing, (1969), and Johnston and Jenkins (1971).much of the incidental learning literature The experiments by Jenkins and his colleaguesacquires a new significance. There are several showed that with lists of highly associatedreviews of this literature (Postman, 1964; word pairs, free recall and organizationMcLaughlin, 1965), and we will make no resulting from an orienting task which requiredattempt to be comprehensive. An important the use of the word as a semantic unit, wascharacteristic of the incidental learning para- equivalent to that of an intentional controldigm is that the subject processes the material group with no incidental task, but both werein a way compatible with or determined by substantially superior to an incidental groupthe orienting task. The comparison of whose task involved treating the wordretention across different orienting tasks, structurally (checking for certain letters ortherefore, provides a relatively pure measure estimating the number of letters in the word).of the memorial consequences of different These results are consistent with those ofprocessing activities. According to the view of Mandler (1967) who showed that incidentalthe present paper, and in agreement with learning during categorization of wordsPostman (1964), the instruction to learn yielded a similar recall level to that of a groupfacilitates performance only insofar as it leads who performed the same activity but whothe subject to process the material in a manner knew that their recall would be tested.which is more effective than the processing Experiments involving the incidental learn-induced by the orienting task in the incidental ing of sentences (Bobrow & Bower, 1969;condition. Thus, it is possible, that with an Rosenberg & Schiller, 1971) have shown thatappropriate orienting task and an inappro- recall after an orienting task that requiredpriate intentional strategy, learning under processing the sentence to a semantic level wasincidental conditions could be superior to that substantially superior to recall of words fromunder intentional conditions. equivalently exposed sentences which wereFrom the point of view of this paper, then, processed nonsemantically.the interesting thing to do is to systematically Schulman (1971) had subjects scan a liststudy retention following different orienting of words for targets defined either structurallytasks within the incidental condition, rather (such as words containing the letter A) orthan to compare incidental with intentional semantically (such as words denoting livinglearning. Under incidental conditions, the things). After the scanning task, subjects were

678 CRAIK AND LOCKHARTgiven an unexpected test of recognition degree. When the orienting task involved thememory. Performance in the semantically production of mediating responses, perform-defined target conditions was significantly ance was equal to that of unhindered inten-better than that in the structurally defined tional learning and superior to when theconditions although scanning time per word orienting task was rating words for pleasant-was approximately the same in most cases. ness. In single-trial free recall, this latter orienting task produces performance equal to These results support the general conclusion that of intentional learning (Hyde & Jenkins,that memory performance is a positive func- 1969). Identical orienting tasks do not seem totion of the level of processing required by the have equivalent effects across different para-orienting task. However, beyond a certain digms. The interaction between initial en-stage, the form of processing which will prove coding and subsequent retrieval operations isoptimal depends on the retrieval or trace worth emphasizing. Although the distinctionutilization requirements of the subsequent between availability and accessibility (Tulvingmemory test. There is clear evidence in the & Pearlstone, 1966) is a useful one, theincidental learning literature that the relative effectiveness of a retrieval cue depends on itsvalue of different orienting tasks is not the compatibility with the item's initial encodingsame for all tests of memory. or, more generally, the extent to which the retrieval situation reinstates the learning This conclusion is supported by compari- context.sons of the differential effects of orientingtasks on recognition and recall. Eagle and Selective Attention and Sensory StorageLeiter (1964) found that whereas free recallin an unhindered intentional condition was Moray (1959) showed that words presentedsuperior to that of an incidental group and to the nonattended channel in a dichoticto a second intentional group who had also to listening test were not recognized in a laterperform the orienting task, these latter two memory test. Similarly, Neisser (1964) hasconditions showed superior recognition per- shown that nontarget items in a visual searchformance. Such a result poses no: difficulty task left no recognizable trace. Thus, if stimuliprovided it is assumed that optimal processing are only partially analyzed, or processed onlydoes not take the same form for both memory to peripheral levels, their record in memorytests. In the Eagle and Leiter (1964) experi- is extremely fleeting. This point was neatlyment, the orienting task, while almost demonstrated by Treisman (1964). When thecertainly involving some degree of semantic same prose passage was played to both earsanalysis, might have served to prevent the dichotically, but staggered in time with thekind of elaborative processing necessary for unattended ear leading, the lag betweenlater access to the stored information. On the messages had to be reduced to 1.5 secondsother hand, such elaborative coding might before the subject realized that the messageshinder subsequent discrimination between were identical. When the attended (shadowed)target words and the associatively related ear was leading, however, subjects noticed thedistractors used in this experiment. Results similarity at a mean lag of 4.5 seconds. Thus,consistent with this kind of analysis have also although the subjects were not trying tobeen reported by Dornbush and Winnick remember the material in either case, the(1967) and Estes and DaPolito (1967). further processing necessitated by shadowing was sufficient to treble the durability of the While the orienting tasks used by Wicker memory trace. Treisman also found thatand Bernstein (1969) in their study of in- meaningfulness of the material (reversedcidental paired-associate learning all required speech versus normal speech, and randomanalysis to a semantic level, they did notfacilitate subsequent performance to the same

LEVELS OF PROCESSING 679words versus prose) affected the lag necessary term retention as well. We would also suggestfor recognition, but only when the attended that it is processing level, rather than inform-channel was leading. If the message was ation content, which determines the rate ofrejected after early analyses, meaningfulness decay.played no part; but when the message wasattended, more meaningful material could be The STS/LTS Distinctionprocessed further and was, thus, retained The phenomenon of a limited-capacitylonger. The three estimates of memory per- holding mechanism in memory (Miller, 1956;sistence in these experiments (1.5 seconds for Broadbent, 1958) is handled in the presentall nonattended material, 3 seconds for attend- framework by assuming that a flexible centraled reversed speech and attended strings processor can be deployed to one of severalof random words, and 5 seconds for attended levels in one of several encoding dimensions,prose) can be attributed to the functioning of and that this central processor can only dealdifferent stores, but it is more reasonable, in with a limited number of items at a given time.our view, to postulate that persistence is a That is, items are kept in consciousness or infunction of processing level. primary memory by continuing to rehearseWhile further studies will not be reviewed in them at a fixed level of processing. The naturesuch detail, it may be noted that the findings of the items will depend upon the encodingand conclusions of many other workers in the dimension and the level within that dimen-area of sensory memory can also be accom- sion. At deeper levels the subject can makemodated in the present framework. Neisser more use of learned cognitive structures so(1967, p. 33) concluded that \"longer exposures that the item will become more complex andlead to longer-lasting icons.\" Studies by semantic. The depth at which primaryNorman (1969), Glucksberg and Cowen memory operates will depend both upon the(1970), and Peterson and Kroener (1964) may usefulness to the subject of continuing toall be interpreted as showing that non- process at that level and also upon the amen-attended verbal material is lost within a few ability of the material to deeper processing.seconds. Thus, if the subject's task is merely to repro-Massaro (1970) suggested that memory for duce a few words seconds after hearing them,an item is directly related to the amount of he need not hold them at a level deeper thanperceptual processing of the item, a statement phonemic analysis. If the words form awhich is obviously in line with the present meaningful sentence, however, they areproposals, although his later arguments compatible with deeper learned structures(Massaro, 1972), that echoic memory inevit- and larger units may be dealt with. It seemsably lasts only 250 milliseconds, are probably that primary memory deals at any level withovergeneralizations. Shaffer and Shiffrin con- units or \"chunks\" rather than with inform-cluded from an experiment on picture recogni- ation (see Kintsch, 1970, pp. 175-181). Thattion that \"it might prove more fruitful to is, we rehearse a sound, a letter, a word, anconsider the more parsimonious view that idea, or an image in the same way that wethere is just a single short-term visual memory. perceive objects and not constellations ofThis short-term visual memory would decay attributes.quickly when the information content of the As pointed out earlier, a common distinc-visual field was high and more slowly when the tion between memory stores is their differentinformation content was greatly reduced\" coding characteristics; STS is said to be(Shaffer & Shiffrin, 1972, p. 295). Plainly this predominantly acoustic (or articulatory) whileview is similar to our own, although we would LTS is largely semantic. According to theargue that the continuum extends to long- present argument, acoustic errors will pre-

680 CRAIK AND LOCKHARTdominate only insofar as analysis has not The Serial Position Curveproceeded to a semantic level. There are at leastthree sources of the failure of processing to Serial-position effects have been a majorreach this level; the nature of the material, source of evidence for the STS/LTS distinctionlimited available processing capacity, and (see Broadbent, 1971, pp. 354-361; Kintsch,task demands. Much of the data on acoustic 1970, pp. 153-162). In free recall, the recencyconfusions in short-term memory is based on effect is held to reflect output from STS whilematerial such as letters and digits which have previous items are retrieved from LTSrelatively little semantic content. The nature (Glanzer & Cunitz, 1966). Several theoreticalof this material itself tends to constrain pro- accounts of the primacy effect have beencessing to a structural level of analysis and it given, but perhaps the most plausible is thatshould be no surprise, therefore, that errors initial items receive more rehearsals and are,of a structural nature result. Such errors can thus, better registered in LTS (Atkinson &also occur with meaningful material if process- Shiffrin, 1968; Bruce & Papay, 1970). Weing capacity is diverted to an irrelevant task agree with these conclusions. Since the(Eagle & Ortoff, 1967). subject knows he must stop attending to initial items in order to perceive and rehearse A further set of results relevant to the subsequent items, he subjects these first itemsSTS/LTS distinction are those that show that to Type II processing; that is, deeper semanticin free recall, variables such as presentation processing. Final list items can survive onrate and word frequency, affect long-term but phonemic encoding, however, which gives risenot short-term retention (Glanzer, 1972). Our to excellent immediate recall (since they areinterpretation of these findings is that in- still being processed in primary memory)creasing presentation rate, or using unfamiliar but is wiped out by the necessity to processwords, inhibits or prevents processing to those interpolated material. In fact, if terminallevels necessary to support long-term reten- items have been less deeply processed thantion, but does not affect coding operations of initial items, the levels of processing formula-the kind that are adequate for short-term tion would predict that in a subsequent recallretention. It follows from this interpretation attempt, final items should be recalled leastthat diverting processing capacity as in the well of all list items. The finding of negativeEagle and Ortoff (1967) experiments should recency (Craik, 1970) supports this prediction.result in a greater decrement in long-term An alternative explanation of negative recencythan in short-term retention and, indeed, could be that recency items were rehearsedthere is good evidence that such is the case fewer times than earlier items (Rundus, 1971).(Murdock, 1965; Silverstein & Glanzer, However, recent studies by Jacoby and Bartz1971). (1972), Watkins (1972), and Craik (1972) have shown that it is the type rather than the amount Conversely, manipulations that influence of processing which determines the subse-processing at a structural level should have quent recall of the last few items in a list.transitory, but no long-term, effects. Modalitydifferences (Murdock, 1966) provide a clear In serial recall, subjects must retain theexample. Finally, long-term recall should be first few items so that they can at least com-facilitated by manipulations which induce mence their recall correctly. The greatlydeeper or more elaborative processing. We enhanced primacy effect is thus probablysuggest that the encoding variability hypo- attributable, in part at least, to primary-thesis as it has been used to account for the memory retention. The degree to whichspacing effect in free recall (Madigan, 1969; subjects also encode initial items at a deeperMelton, 1970) is to be understood in these level is likely to depend on the material andterms.

LEVELS OF PROCESSING 681the task. Using a relatively slow (2.5 seconds) trace or merely postpones forgetting dependspresentation rate and words as visually on what the subject is doing with his rehearsal.presented stimuli, Palmer and Ornstein (1971) Only deeper processing will lead to anfound that an interpolated task only partially improvement in memory.eliminated the primacy effect. However,Baddeley (1968) presented digits auditorily CONCLUDING COMMENTSat a 1-second rate and found that primacy wasentirely eliminated by the necessity to perform Our account of memory in terms of levelsa further task. of processing has much in common with a number of other recent formulations. CermakRepetition and Rehearsal Effects (1972), for example, has outlined a theoretical framework very similar to our own. Per- One suggestion in the present formulation ceptually oriented attribute-encoding theoriesis that Type I processing does nothing to such as those of Bower (1967) and Norman andenhance memory for the stimulus; once Rumelhart (1970) have a close affinity with theattention is diverted, the trace is lost at the present approach as does that of Posner (1969)rate appropriate to its deepest analyzed level. who advocates stages of processing withThus, the concept of processing has been split different characteristics associated with eachinto Type I or same-level processing and Type stage.II processing which involves further, deeperanalysis of the stimulus and leads to a more If the memory trace is viewed as the by-durable trace. Similarly, the effects of repeated product of perceptual analysis, an importantpresentation depend on whether the repeated goal of future research will be to specify thestimulus is merely processed to the same level memorial consequences of various types ofor encoded differently on its further present- perceptual operations. We have suggestedations. There is evidence, both in audition the comparison of orienting tasks within the(Moray, 1959; Norman, 1969), and in vision incidental learning paradigm as one method(Turvey, 1967), that repetition of an item by which the experimenter can have moreencoded only at a sensory level, does not lead direct control over the encoding operationsto an improvement in memory performance. that subjects perform. Since deeper analysis will usually involve longer processing time, Tulving (1966) has also shown that repeti- it will be extremely important to disentangletion without intention to learn does not such variables as study time and amount offacilitate learning. Tulving's explanation of the effort from depth as such. For example, timeabsence of learning in terms of interitem may be a correlate of memory to the extent thatorganization cannot easily be distinguished time is necessary for processing to some level,from an explanation in terms of levels of but it is possible that further time spent inprocessing. Similarly, Glanzer and Meinzer merely recycling the information after this(1967) have shown that overt repetition of optimal level will not predict trace durability.items in free recall is a less effective strategythan that normally used by subjects. Although Our approach does not constitute a theoryboth Waugh and Norman (1965), and Atkin- of memory. Rather, it provides a conceptualson and Shiffrin (1968) have suggested that framework--a set of orienting attitudes--rehearsal has the dual function of maintaining within which memory research might proceed.information in primary memory and trans- While multistore models have played a usefulferring it to secondary memory, the experi- role, we suggest that they are often taken tooments by Tulving (1966) and by Glanzer and literally and that more fruitful questions areMeinzer (1967) show that this is not necessarily generated by the present formulation. Ourso. Thus, whether rehearsal strengthens the position is obviously speculative and far from

682 CRAIK AND LOCKHARTcomplete. We have looked at memory purely COOPER,E. H., & PANTLE,A. J. The total-time hypo- thesis in verbal learning. Psychological Bulletin,from the input or encoding end; no attempt 1967, 68, 221-234.has been made to specify either how items are CRAIK,F. I. M. The fate of primary memory items in free recall. Journal of Verbal Learning anddifferentiated from one another, are grouped Verbal Behavior, 1970, 9, 143-148.together and organized, or how they are CRAIK,F. I. M. A 'levels of analysis' view of memory. Paper presented at the 2nd Erindale Symposiumretrieved from the system. While our position on Communicationand Affect, March, 1972.does not imply any specific view of these CRAIK,F. I. M., & LEVY,B. A. Semantic and acoustic information in primary memory. Journal ofprocesses, it does provide an appropriate Experimental Psychology, 1970, 86, 77-82.framework within which they can be CRAm, F. I. M., & MASANI,P. A. Age and intelligence differences in coding and retrieval of word lists.understood. British Journal of Psychology, 1969, 60, 315-319. REFERENCES CRANNELL,C. W., & PARRISH,J. M. A comparison of immediate memory span for digits, letters, andATKINSON,R. C., & SHIFFRIN,R. M. Human memory: words. Journal of Psychology, 1957, 44, 319- A proposed system and its control processes. In 327. K. W. Spence and J. T. Spence (Eds.) Thepsycho- logy of learning and motivation: Advances in CROWDER, R. G., & MORTON, J. Precategorical research and theory, Vol. II. New York: Academic acoustic storage. Perception and Psychophysics, Press, 1968. Pp. 89-195. 1969, 5, 365-373.ATKINSON,R. C., & SHIFFRIN,R. M. The control of DORNBUSH, R. L., & WlNNICIC,W. A. Short-term short-term memory. Scientific American, I971, intentional and incidental learning. Journal of 224, 82-89. Experimental Psychology, 1967, 73, 608-611.BADDELEY, A. D. Short-term memory for word EAGLE, M., & LEITER, E. Recall and recognition sequences as a function of acoustic, semantic, and in intentional and incidental learning. Journal formal similarity. Quarterly Journal of Experi- of Experimental Psychology, 1964, 68, 58-63. mental Psychology, 1966, 18, 362-365. EAGLE,M., & ORTOEF,E. The effectof leve/ofattentionBADDELEY, A. D. How does acoustic similarity upon \"phonetic\" recognition errors. Journal of influence short-term memory? Quarterly Jour- VerbalLearning and VerbalBehavior, 1967, 6, 226- nal of Experimental Psychology, 1968, 20, 249- 231. 264. ESTES,W. K., & DAPOLITO,F. Independent variationBADDELEYA, . D. Estimatingthe short-term component of information storage and retrieval processes in free recall. British Journal of Psychology, 1970, in paired-associate learning. Journal of Experi- 61, 13-15. mental Psychology, 1967, 75, 18-26.BOBROW,S. A., & BOWER,G. H. Comprehension and GLANZER,M. Storage mechanisms in recall. In G. H. recall of sentences. JournalofExperimentalPsych- Bower (Ed.) The psychology o f learning and ology, 1969, 80, 455-461. motivation: Advances in research and theory. Vol. 5. New York: Academic Press, 1972. Pp.BOWER, G. H. A multicomponent theory of the 129-193. memory trace. In K. W. Spence and J. T. Spence (Eds.) The psychology of learning and motivation: GLANZER,M., & CUNITZ,A. R. Two storage mechan- Advances in research and theory, Vol. 1. New isms in free recall. Journal of Verbal Learning and York: Academic Press, 1967. Pp. 230-325. Verbal Behavior, 1966, 5, 351-360.BROADBENT, D. E. Perception and communication. GLANZER,M., & MEINZER,A. The effects of intralist New York: Pergamon Press, 1958. activity on free recall. Journal of Verbal Learning and Verbal Behavior, •967, 6, 928-935.BROADBENT,D. E. Decision and stress. New York: Academic Press, 1971. GLUCKSBERG, S., & COWEN, G. N. Memory for nonattended auditory material. Cognitive Psycho-BRUCE,D., & PAPAY,J. P. Primacy effect in single-trial logy, 1970, 1, 149-156. free recall. Journal of Verbal Learning and Verbal Behavior, 1970, 9, 473-486. HABER,R. N. How werememberwhat we see. Scientific American, 1970, 222, 104-112.CERMAK,L. S. Human memory. Research and theory. New York: Ronald, 1972. HYDE,T. S., & JENKINS,J. J. The differential effects of incidental tasks on the organization of recall of aCONRAD, R. Acoustic confusions in immediate list of highly associated words. Journal of Experi- memory. British Journal of Psychology, 1964, mental Psychology, 1969, 82, 472-481. 55, 75-84.

LEVELS OF PROCESSING 683JACOBY, L. L., & BARTZ,W. H. Encoding processes MORAY, N. Attention in dichotic listening: affective and the negative recency effect. Journal of Verbal cues and the influence of instructions. Quarterly Learning and Verbal Behavior, 1972, 11, 561-565. Journal of Experimental Psychology, 1959, 9, 56-60.JAMES,W. Principles of psychology. New York: Holt, 1890. MORAY,N. Where is capacity limited ? A survey and a model. In A. Sanders (Ed.) Attention andperform-JOHNSTON,C. D., & JENKINS,J. J. Two more incidental ante. Amsterdam: North-Holland, 1967. tasks that differentially affect associative clustering in recall. Journal of Experimental Psychology, MORTON,J. A functional model of memory. In D. A. 1971, 89, 92-95. Norman (Ed.) Models of human memory. New York: Academic Press, 1970. Pp. 203-254.KINTSCH, W. Learning, memory, and conceptual processes. New York: Wiley, 1970. MURDOCK,B. 13. JR. Effects of a subsidiary task on short-term memory. British Journal ofPsychology,KdNTSCH, W., & 13tJSCHKE, H. Homophones and 1965, 56, 413-419. synonyms in short-term memory. Journal of Experimental Psychology, 1969, 80, 403-407. MURDOCK, B. B. JR. Visual and auditory stores in short-term memory. QuarterlyJournalofExperi-KROLL,N. E. A., PARKS,T., PARKINSON,S. R., 13IEBER, mental Psychology, 1966, 18, 206-211. S. L., & JOr~NSON,A. L. Short-term memory while shadowing. Recall of visually and aurally pre- MURDOCK, B. B. JR. Recent developments in short- sented letters. Journal of Experimental Psychology, term memory. British Journal of Psychology, 1967, 1970, 85, 220-224. 58, 421-433.LEvv, B. A. Role of articulation in auditory and visual MtJRDOCK,]3. B. JR. Four channel effects in short-term short-term memory. Journal of Verbal Learning memory. Psychonomic Science, 1971, 24, 197- and Verbal Behavior, 1971, 10, 123-132. 198.MACNAMARA,J. Cognitive basis of language learningin MVRDOCK, B. B. JR. Short-term memory. In G. H. infants. Psychological Review, 1972,79, 1-13. Bower (Ed.) Psychology of learn#Tgand motivation, Vol. 5. New York: Academic Press, 1972. Pp.MADIGAN, S. A. Intraserial repetition and coding 67-•27. processes in free recall. Journal of VerbalLearning and Verbal Behavior, 1969, 8, 828-835. NEISSER, U. Visual search. Scientific American, 1964, 210, 94-102.MANDLER,G. Organization and Memory. In K. W. Spence and J. T. Spence (Eds.) The psychology of NEISSER,U. Cognitive psychology. New York: Apple- learning and motivation: Advances in research and ton-Century-Crofts, 1967. theory, Vol. 1. New York: Academic Press, 1967. Pp. 328-372. NORMAN,D. A. Memory while shadowing. Quarterly Journal of Experimental Psychology, 1969, 21,MASSARO,D. W. Perceptual processes and forgetting 85-93. in memory tasks. Psychological Review, 1970, 77, 557-567. NORMAN,D. A., & RVraELHART,D. E. A system for perception and memory. In D. A. Norman (Ed.)MASSARO, D. W. Preperceptual images, processing Models of human memory. New York: Academic time, and perceptual units in auditory perception. Press, 1970. Pp 21-64. Psychological Review, 1972, 79, 124-145. PALMER,S. E., & ORNSTEIN,P. A. Role of rehearsalMcLAUGHEIN, B. \"Intentional\" and \"incidental\" strategy in serial probed recall. Journal of Experi- learning in human subjects: The role of instruc- mental Psychology, 1971, 88, 60-66. tions to learn and motivation. Psychological Bulletin, 1965, 63, 359-376. PETERSON, L. R. Short-term verbal memory and learning. Psychological Review, 1966, 73, 193-MELTON,A. W. Implications of short-term memory 207. for a general theory of memory. Journal of Verbal Learning and Verbal Behavior, 1963, 2, 1-21. PETrRSON, L. R., & JOHNSON, S. T. Some effects of minimizing articulation on short-term retention.MELTON,A. W. The situation with respect to the spac- Journal of Verbal Learning and Verbal Behavior, ing of repetitions and memory. Journal of Verbal 1971, 10, 346-354. Learning and Verbal Behavior, 1970, 9, 596-606. PETERSON,L. R., & KROENER,S. Dichotic stimulationMILLER, G. A. The magical number seven, plus or and retention.Journal ofExperimentalPsychology, minus two: Some limits on our capacity for 1964, 68, 125-130. processing information. Psychological Review, 1956, 63, 81-97. PmLLIVS, W. A., & BADDELEY,A. D. Reaction time and short-term visual memory. PsychonomicMILNER,13. Memory and the medial temporal regions Science, 1971, 22, 73-74. of the brain. In K. H. Pribram and D. E. Broad- bent (Eds.) Biology of memory. New York: POSNER, M. I. Short-term memory systems in human Academic Press, 1970. Pp. 29-50. information processing. Acta Psychologica, 1967, 27, 267-284.

684 CRAIK AND LOCKHARTPOSNER, M. I. Abstraction and the process of recogni- ation. In A. Young and D. B. Lindsley (Eds.) tion. In G. H. Bower and J. T. Spence (Eds.) Early experience and visual information processing The psychology of learning and motivation: in perceptual and reading disorders. Washington: Advances in research and theory, Vol. III. New National Academy of Sciences, 1970.Pp. 198-215. STorr, M., & EAGLE,M. N. The relationship among York: McGraw-Hill, 1969. Pp. 152-179. reported strategies, presentation rate, and verbalPOSTMAN, L. Short-term memory and incidental ability and their effects on free recall learning. Journal of Experimental Psychology, 1971, 87, learning. In A. W. Melton (Ed.) Categories of 423-428. human learning. New York: Academic Press, 1964. SUTHERLAND,N. S. Outlines of a theory of visual pattern recognition in animals and man. Proceed- Pp. 145-201. ings of the Royal Society. Series B, 1968, 171,ROSENBERG,S., & SCHILLER,W. J. Semantic coding and 297-317. TREISMAN, A. Monitoring and storage of irrelevant incidental sentence recall. Journal of Experimental messages in selective attention. Journal of Verbal Psychology, 1971, 90, 345-346. Learning and Verbal Behavior, 1964,3, 449-459.RUNDUS, D. Analysis of rehearsal processes in free TRESSELT,M. E., & MAYZYER,M. S. A study of inci- recall. Journal of Experimental Psychology, 1971, dental learning. Journal of Psychology, 1960, 50, 89, 63-77. 339-347.SAVlN, H. B., & BEVER, T. G. The nonperceptual TULVINC, E. Subjective organization and effects of reality of the phoneme. Journal of VerbalLearning repetition in multi-trial free-recall learning. and Verbal Behavior, 1970, 9, 295-302. Journal of Verbal Learning and Verbal Behavior,SCHULMAN,A. I. Recognitionmemory for targets from 1966, 5, 193-197. a scanned word list. British Journal of Psychology, TULVING,E., & MADIGAN,S. A. Memory and verbal learning. Annual Review of Psychology, 1970, 21, 1971, 62, 335-346. 437-484.SELFRIDGE,O. G., & NEISSER,U, Pattern recognition TULVIN6,E., & PATTERSON,R. D. Functionalunits and retrieval processes in free recall. Journal of by machine. Scientific American, 1960, 203, 60-68. Experimental Psychology, 1968, 77, 239-248. TULVIN6, E., & PEARLSTONE,Z. Availability versusSHAFFER,W. O., & SHIFFRIN,R. M. Rehearsal and accessibility of informationin memory for words. storage of visual information. Journal of Experi- Journal of Verbal Learning and Verbal Behavior, mental Psychology, 1972, 92, 292-296. 1966, 5, 381-391. TURVEY, M. T. Repetition and the preperceptualSHALLICE, T., & WARRINGTON,E. K. Independent information store, Journal of Experimental functioning of verbal memory stores: A neuro- Psychology, 1967, 74, 289-293. psychological study. Quarterly Journal of Experi- WARRINGTON, E. K. Neurological disorders of mentalPsychology, 1970, 22, 261-273. memory. British Medical Bulletin, 1971, 27 243-247.SHEPARD, R. N. Recognition memory for words, WATKINS,M. J. The characteristics and functions of sentences, and pictures. Journal of VerbalLearning primary memory. Unpublished Ph.D. thesis, and Verbal Behavior, 1967, 6, 156-163. University of London, 1972. WAUCH,N. C., & NORMAN,D. A. Primary memory.SHIFFRIN, R. M., & ATKINSON,R. C. Storage and Psychological Review, t965, 72, 89-104. retrieval processes in long-term memory. Psycho- WICKER,F. W., & BERYSTEIN,A. L. Association value logical Review, 1967, 76, 179-193. and orienting task in incidental and intentional paired-associate learning. Journal of ExperimentalSHULMAN,H. G. Encoding and retention of semantic Psychology, 1969, 81, 308-311. and phonemic informationin short-term memory. Journal of Verbal Learning and Verbal Behavior, (Received June 3, 1972) 1970, 9, 499-508.SHULMAN, H. G. Similarity effects in short-term memory. PsyehologicalBulletin, 1971,75, 399~15.SrlULMAN,H. G. Semantic confusion errors in short- term memory. Journal of Verbal Learning and Verbal Behavior, 1972, 11, 221-227.SILVERSTEIN,C., & GLANZER,M. Concurrent task in free recall: Differential effects of LTS and STS. Psychonomic Science, 1971, 22, 367-368.SPERLING,G. Short-term memory, long-termmemory, and scanning in the processing of visual inform-