Proc. Nati. Acad. Sci. USAVol. 89, pp. 11113-11117, December 1992

Review

Implicit knowledge: New perspectives on unconscious processes

Daniel L. SchacterDepartment of Psychology, Harvard University, Cambridge, MA 02138

ABSTRACT Recent evidence fromcognitive science and neuroscience indi-cates that brain-damaged patients andnormal subjects can exhibit nonconsciousor implicit knowledge of stimuli that theyfail to recollect consciously or perceiveexplicitly. Dissociations between implicitand explicit knowledge, which have beenobserved across a variety of domains,tasks, and materials, raise fundamentalquestions about the nature of perception,memory, and consciousness. This articleprovides a selective review of relevant ev-idence and considers such phenomena aspriming and implicit memory in amnesicpatients and normal subjects, perceptionwithout awareness and "blindsight" inpatients with damage to visual cortex, andnonconscious recognition of familiar facesin patients with facial-recognition deficits(prosopagnosia). A variety of theoreticalapproaches to implicit/explicit dissocia-tions are considered. One view is that all ofthe various dissociations can be attributedto disruption or disconnection of a com-mon mechanism underlying consciousexperience; an alternative possibility isthat each dissociation requires a separateexplanation in terms of domain-specificprocesses and systems. More generally, itis concluded that rather than reflecting theoperation of affectively charged uncon-scious processes of the kind invoked bypsychodynamic or Freudian theorists, dis-sociations between implicit and explicitknowledge are a natural consequence ofthe ordinary computations of the brain.

Consider the following two clinical sce-narios. In the first, a patient with memoryproblems is shown a list of familiar wordsand several minutes later is unable toremember any of the list items whenasked to recollect them; indeed, he de-nies that a list of words had been pre-sented. But when he is required to per-form an incidental test that does notrequire conscious recollection of the list,the patient's performance indicates per-fectly normal retention of the previouslystudied words. In the second scenario, apatient with perceptual problems is ex-posed to a bright visual stimulus andclaims to see nothing. Yet when asked to"guess" in which of two locations thestimulus appeared, the patient performswell above the chance level, indicatingthat she has in some sense "seen"-

despite the absence of conscious experi-ence-the target stimulus.The foregoing scenarios may seem sur-

prising and even bizarre: How can apatient exhibit memory without remem-bering or perception without perceiving?It is tempting to suggest that the patientssuffer from psychiatric problems or, per-haps, are engaging in outright deceptionof the examiner. On the contrary, how-ever, these two scenarios represent ex-amples ofwhat have become almost com-monplace observations in the neuropsy-chological laboratory and are oftenreferred to as dissociations between ex-plicit and implicit knowledge (1). Explicitknowledge refers to knowledge that isexpressed as conscious experience andthat people are aware that they possess;the everyday uses of such terms as "see-ing" and "remembering" refer to explicitknowledge. Implicit knowledge, by con-trast, refers to knowledge that is revealedin task performance without any corre-sponding phenomenal awareness; im-plicit knowledge is often expressed unin-tentionally and tapped indirectly. Farfrom reflecting psychiatric symptoms ordissimulation, dissociations between ex-plicit and implicit knowledge are provid-ing important new insights into the fun-damental nature of perception, memory,and conscious experience.The terms explicit and implicit knowl-

edge are quite similar in meaning to con-scious and unconscious knowledge, andthe two sets of terms can be used inter-changeably. However, traditional con-ceptions of unconscious knowledge havebeen tied closely to Freudian and otherpsychodynamic constructs such as re-pression, drive, conflict, and the like. Asthis article should make evident, theseconcepts have little relevance to thekinds of phenomena that have been thesubject of recent neuropsychological andcognitive studies. Because the classical i

notion of the "unconscious" is so closelylinked to psychodynamic ideas, it seems Iprudent to use terminology that is not 1similarly burdened (for more extended Idiscussion, see ref. 2).The article provides a selective over- X

view of research that has documented End explored dissociations between ex- ilicit and implicit knowledge. It will fo- r-us primarily on explicit/implicit disso- c,iations in patients with memory disor- I

ders and perceptual disorders, althoughsimilar phenomena that have been docu-mented in other patient populations willalso be noted. In addition, some attentionwill be paid to analogous dissociationsthat have been produced in cognitivestudies of normal, non-brain-damagedsubjects. The article will conclude bysurveying theoretical accounts ofthe var-ious dissociations and by consideringwhether these diverse phenomena de-pend on similar underlying mechanisms.Taken together, the neuropsychologicaland cognitive evidence suggests that,rather than reflecting the operation ofaffectively charged psychodynamic pro-cesses, many implicit or unconscious ex-pressions of knowledge occur as a rela-tively routine consequence of the ordi-nary computations of the brain.

Memory Disorders

The most extensively studied neurologi-cal disorder of memory is known as theamnesic syndrome, which occurs as aconsequence of various kinds of patho-logical conditions (e.g., stroke, enceph-alitis, anoxia) that produce damage tomedial temporal and diencephalic brainregions (3, 4). Amnesic patients are char-acterized by a marked inability to remem-ber recent experiences together with nor-mal perception and intelligence. Theirmemory disorders are evident on a vari-ety of explicit memory tests, includingfree recall, where patients attempt toretrieve recently presented items withoutthe aid of experimenter-provided cues;cued recall, where various cues or hintsare provided to assist recollection; andrecognition, where previously studieditems are presented together with newitems, and subjects indicate which itemthey recollect from the study list.Despite their severe impairments in

explicitly remembering recently pre-sented information, it has been estab-lished beyond dispute that amnesic pa-tients can show intact implicit memoryfor aspects of the same information. Oneof the most intensively studied implicitmemory phenomena in amnesia is knownas repetition or direct priming: the facil-itated ability to identify, or make judg-ments about, target stimuli as a conse-quence of a recent exposure to them (5).In a typical priming experiment, subjects

11113

Proc. Natl. Acad. Sci. USA 89 (1992)

are shown a list of familiar words and are

later given an apparently unrelated testthat does not require explicit memory forthe study list. For example, on a stemcompletion test, subjects would be giventhree-letter word beginnings (e.g., T-A-B-) and asked to complete them with thefirst word that comes to mind. Half ofthestems could be completed with wordsthat had appeared previously on thestudy list (e.g., TABLE), and the otherhalf could not be completed with wordsfrom the study list. Priming is said tooccur when subjects provide the targetcompletion more frequently to stems thatrepresent studied words than to stemsthat represent nonstudied words. Earlystudies by Warrington and Weiskrantz(6) indicated that amnesic patients showwhat we would now call normal primingeffects on the stem-completion task de-spite poor explicit memory, althoughWarrington and Weiskrantz did not referto the phenomenon as priming. Subse-quent experiments confirmed and ex-tended the finding of normal stem-completion priming in amnesia and alsorevealed that amnesics exhibit normalpriming effects on a variety of other im-plicit memory tasks (refs. 7-12; for re-view, see ref. 13).Recent research has extended further

the boundaries of the priming phenome-non. In early experiments, the targetitems presented for study were well-learned materials, such as familiarwords, that exist in memory before theexperiment. Thus, it was possible to ar-

gue that priming effects in amnesic pa-tients reflect the temporary activation ofpreexisting memory representations(e.g., refs. 7 and 14). However, it has nowbeen established that amnesic patientscan show intact priming for novel infor-mation that does not have a preexistingmemory representation, includingpseudowords (e.g., "numdy"; refs. 15and 16) and nonverbal materials, such as

novel objects or patterns (17-20). It hasalso been shown that priming effects inamnesia can be quite long-lived, lastingacross retention intervals ofdays, weeks,or months (21-23).Research on priming effects in amnesic

patients has been complemented by a

large and ever-increasing literature onnormal subjects, indicating that primingcan be dissociated sharply from explicitmemory (for reviews, see refs. 24-26).One particularly important finding is thatpriming effects on various implicit mem-ory tests are relatively unaffected by ma-nipulations ofhow subjects encode targetmaterials during study-list presentation.For example, when subjects are inducedto process the semantic attributes ofwords at the time of study (e.g., makejudgments about a word's meaning), theirrecollection of the word on subsequentexplicit memory tests is generally much

higher than when they are induced toprocess nonsemantic physical features ofthe target words (e.g., count the numberof vowels in a word). But the magnitudeof priming effects is similar following thetwo kinds of study tasks (20, 27, 28).Moreover, the priming effect appears tobe modality-specific: It is reduced bystudy-to-test changes in visual or audi-tory modality of presentation (28, 29).Under certain circumstances, priming iseven reduced by study-to-test changes inthe particular type font or case in whicha word appears (30, 31) or the voice inwhich a word is spoken (32). These kindsof observations, taken together with theamnesia data, have led to a vigorousdebate concerning the psychological andneurophysiological processes and sys-tems that subserve priming and explicitmemory, respectively (cf. refs. 3, 25, and33).

Priming is not, however, the only ex-ample of preserved implicit memory inamnesic patients. It has been knownsince the classic studies of Milner andCorkin and their colleagues (34) that am-nesic patients can acquire new motorskills across numerous training sessions,and it is now clear that they can graduallyacquire perceptual and cognitive skills aswell (cf. refs. 35-37). Amnesics have alsoexhibited normal implicit learning of therules of an artificial grammar (38) and acomplex spatio-temporal sequence (39)and have proved capable of acquiringclassically conditioned responses (40,41). Severely amnesic patients have evenbeen able to learn (although not at anormal rate) the complex knowledge andskills needed to operate and program acomputer and have exhibited robust re-tention of such knowledge over delays of5-9 mo (42, 43)-despite little or no ex-plicit memory for their learning experi-ences.Some of this work on implicit memory

for complex knowledge and skills has a

parallel in-and, in fact, was inspiredby-studies of normal subjects. For ex-ample, Reber has reported a series ofstudies over the past 25 yr (refs. 44 and45) that have provided evidence for im-plicit learning ofgrammatical rules: Afterexposure to a list of consonant stringsthat are ordered according to the com-plex rules of an artificial grammar, sub-jects can later distinguish novel grammat-ical strings from novel nongrammaticalstrings, even though the subjects are un-able to articulate the nature of the rule(for a critique, see ref. 46). Similarly,Lewicki and colleagues (47) have pro-vided evidence that subjects can learncomplex patterns and contingencies de-spite poor explicit knowledge of them.Taken together, the data from amnesic

patients and normal subjects indicateclearly that memory for various kinds ofexperiences can be expressed indepen-

dently from, and in the absence of, con-scious recollection of those experiences.

Perceptual Disorders

At about the same time that early evi-dence was accumulating on implicitmemory in amnesic patients, there werereports of a puzzling and, in some re-spects, analogous phenomenon in pa-tients with disturbances of visual percep-tion. Initially documented by Poppel,Held, and Frost (48), the phenomenonwas referred to as "blindsight" by Weis-krantz and colleagues (49), in referenceto the seemingly paradoxical nature ofthe visual behavior exhibited by certainpatients with lesions to primary (striate)visual cortex. Such patients appeared tobe "blind" in the sense that they deniedseeing a stimulus presented in certainparts of the visual field. But when askedto guess about the location or other at-tributes ofthe same stimulus, the patientsexhibited "sight" in the sense that theirguessing performance was well abovechance and sometimes nearly perfect.The experimental paradigm used most

frequently to demonstrate blindsight in-volves requiring patients to localize astimulus presented in the blind field, ei-ther by pointing to or reaching toward thetarget location or by making a verbalresponse. Blindsight can also be exhib-ited when patients are asked to makevisual discriminations about other kindsof targets by guessing, including simplefigures (e.g., X vs. 0) and line orienta-tions (e.g., horizontal vs. vertical). How-ever, Perenin and Jeannerod (50) failed tofind evidence for pattern discrimination,and Weiskrantz (51) has suggested thatorientation discrimination is more fullypreserved than form discrimination inblindsight patients. Finally, blindsighthas also been demonstrated by using ex-perimental paradigms in which informa-tion presented in the blind field influ-ences patients' perceptions of, and re-sponses to, information presented in thesighted field (e.g., ref. 52).One of the most intriguing questions

about blindsight concerns exactly whatpatients experience when they respondaccurately to stimuli presented in theblind field and how these "perceptions"differ from those of conscious visualexperience. Although it is most oftenreported that patients claim to see noth-ing and are merely guessing, it has alsobeen reported that some patients occa-sionally describe a rather primitive visualawareness of a target. For example, apatient studied by Weiskrantz (51)claimed to sense "a definite pinpoint oflight" but when probed further insistedthat it did not "actually look like a light[but] ... nothing at all" (ref. 51, p. 378).It has also been reported that with exten-sive practice and training, patients can

11114 Review: Schacter

Proc. Natl. Acad. Sci. USA 89 (1992) 11115

develop a heightened awareness of stiuli in their blind fields (51, 53). In viewthis evidence, some critics have sigested that rather than representing iplicit or nonconscious perception of tget stimuli, blindsight can be attributedcautious responding on the basis of cgraded (but conscious) vision that is piduced by scattered light (54). Howevisuch explanations have difficulty accoimodating various kinds of evidence (fdiscussion, see refs. 1 and 55).Recent research has revealed implie

perceptual knowledge in other kindsbrain-damaged patients that is, in sonrespects, similar to that observedblindsight. One particularly striking eample comes from the study of visuform agnosia, a disorder in which ptients have difficulty perceiving and reiognizing virtually all kinds of visual oljects. Goodale, Milner, Jakobson, anCarey (56) described a patient who wEseverely impaired in making judgmentabout the width of three-dimensional oljects but made entirely normal motcadjustments of hand position as shreached toward a target object; the postioning of her finger-thumb grip variedirectly as a function of the object'width despite her impairment of conscious perception.There has also been a good deal a

recent research on implicit knowledge i]patients with prosopagnosia-the impaired ability to recognize familiar facesusually because of bilateral lesions tioccipito-temporal cortex. Although sucipatients typically deny any familiaritywith faces that ordinarily would be wetknown to them (e.g., a spouse or relative), there is now considerable evidencethat they possess implicit knowledge o:those faces. Early evidence was providedin a psychophysiological study by Bauer(57), in which a prosopagnosic patientviewed a familiar face and at the sametime listened to the experimenter read aseries of names; one belonged to the faceand the others did not. Despite failing torecognize the face explicitly, the patientshowed a maximal skin conductance re-sponse to the correct name. Tranel andDamasio (58) replicated and extendedthis phenomenon using a different para-digm for eliciting skin-conductance re-sponses to familiar faces that their patientfailed to recognize explicitly.Young and De Haan and their col-

leagues have provided a systematic se-ries of studies using behavioral measuresto demonstrate and explore implicit facialfamiliarity in prosopagnosic patients (forreview, see ref. 59). For example, theyreported the case of a patient who per-formed at chance levels when required tochoose which of two faces (one famous,one unknown) was familiar. However,when given a matching task in whichsubjects judged whether two simultane-

in- ously exposed faces were the sameiof different, the patient-just like normug- control subjects-responded moiim- quickly when the two faces were famotar- than when they were unknown. Simto larly, they also found that the patient wcle- slower to learn a name-face pairing whero- a familiar face was paired with an incoer, rect name than when it was paired withm- correct name, even though he claimefor that none of the faces were familiar; an

presentation of a famous face that wacit unfamiliar to the patient speeded up hiof ability to make judgments about verbsne information associated with the facin (e.g., seeing a photo of Prince Charlex- facilitated his response to the name Prinral cess Diana). Interestingly, Young anoa- colleagues have reported another case ihc- which the patient failed to show evidenceb- for implicit knowledge of unrecognizedid faces (see also ref. 60).as The data from blindsight, visual fornts agnosia, and prosopagnosia indicatib- clearly that conditions exist in whiclor some patients can show implicit knowle edge of visual stimuli that they fail eitheii- to perceive or to recognize explicitlyd Cognitive research with normal, nons brain-damaged subjects has long beeni- concerned with the possible existence of

"perception without awareness" oIif "subliminal perception". Early researchn in this area was fraught with methodolog-- ical difficulties and marked by contro-, versy (for reviews, see refs. 2 and 61).o Although some of the confusion still per-h sists in contemporary work, a number ofy findings and ideas have emerged during

the past decade that have helped to clar-,- ify the sense in which, and extent to

which, perception without awareness canf be said to exist.i In a typical paradigm for studying per-rception without awareness, two kinds oftevidence are provided: (i) explicit or di-

rect measures that are used to documentisubjects' failure to perceive a stimulus

consciously and (ii) implicit or indirect>measures that reveal an impact of the

undetected stimulus on some aspect ofperformance. For example, in a semanticpriming paradigm, the subject may claimthat he or she fails to detect the presenceofthe word chair when it is flashed brieflyand obscured by a visual mask; but thesubject will nevertheless be faster toidentify or make a judgment about therelated word table when it is presentedimmediately after the word chair thanwhen it is presented after a semanticallyunrelated word (e.g., ref. 62). The exis-tence of a semantic priming effect sug-gests that subjects have, indeed, regis-tered some features of the target stimulusdespite the apparent absence of con-scious perception.Why has research of this kind been

dogged by controversy? As Merikle (ref.63, p. 792) has stated, the controversy

or ". . . has centered on the issue, Whatal constitutes an adequate behavioral mea-re sure ofconscious perceptual experience?us Depending upon one's answer to thisi- question, the evidence for perceptionas without awareness is either overwhelm-,, ing or nonexistent." Cheesman and'r- Merikle (64) proposed a useful distinctiona between subjective and objective mea-d sures of conscious perception. Subjec-s tive measures typically involve a per-Is son's verbal statement that he or she doesisa, not detect the presence of a target stim-e ulus, as in the foregoing example. Objec-hs tive measures, by contrast, typically in-

volve such tasks as forced-choice judg-d ments in which subjects must choosen between the presented stimulus and ae nonpresented alternative, even whend they feel that they are just guessing. As

Merikle (63) points out, if one accepts asn valid subjective measures of consciouseperception, then the evidence for percep-h tion without awareness is strong (e.g.,- there is evidence for semantic primingr from stimuli that subjects claim that they

do not see); if one insists on an objectivemeasure, however, then the evidence is

lweak or nonexistent (e.g., there is littlefevidence of semantic priming from stim-ruli about which subjects are unable toImake accurate forced-choice discrimina-

tions).Although the intricacies of this debate

are beyond the purview of the presentarticle, it is worth noting that the distinc-tion between objective and subjectivemeasures of conscious perception hasimplications for our understanding ofneuropsychological phenomena, such asblindsight. In the blindsight literature,failures of conscious perception are typ-ically inferred from subjective measures;patients claim that they do not see atarget stimulus. Moreover, evidence forimplicit knowledge or nonconscious per-ception is frequently inferred fromabove-chance performance on a forced-choice test in which the patient claims tobe guessing-precisely the kind of testthat some would refer to as an objectivemeasure of conscious perception! Tomake matters even more complex,chance-level performance on forced-choice tests has been taken as evidencefor an absence of conscious perception insome studies of implicit or covert recog-nition in prosopagnosia (59). Merikle (63)has delineated several reasons why itprobably makes sense to accept datafrom subjective measures as evidence forfailure of conscious perception. Never-theless, it seems clear that careful atten-tion must be paid to possible differencesin underlying mechanisms when implicitknowledge is inferred from failures onsubjective or objective measure of con-scious perception, respectively.

Review: Schacter

Proc. Natl. Acad. Sci. USA 89 (1992)

Additional Neuropsychological Evidencefor Implicit Knowledge

Although most of the work on implicitknowledge in neuropsychological syn-dromes has involved disorders of mem-ory and perception, similar kinds of evi-dence have been gleaned from patientswith a variety of neuropsychological def-icits. Milberg and Blumstein and theircolleagues (65), for example, have stud-ied aphasic patients who exhibit severedeficits on explicit tests of language com-prehension and yet show robust semanticpriming effects for words that they fail tounderstand explicitly. Tyler (66) has de-scribed other kinds of aphasic patientswho are unable to make explicit judg-ments about the meaning and grammati-cality of sentences. But the performanceof these patients on a target-monitoringtask was disrupted by semantic andgrammatical violations, and the patternofdisruption was similar to that observedin normal control subjects. Evidence forimplicit knowledge has also been seen instudies of patients with reading disor-ders, who appear able to makejudgmentsabout properties of words that they can-not identify consciously (e.g., ref. 67).And observations suggestive of implicitknowledge have been reported in pa-tients who exhibit spatial neglect, asso-ciative agnosia, and unawareness of def-icit (for review, see ref. 1).

Theories and Mechanisms

The seemingly ubiquitous evidence forpreserved implicit knowledge despite im-paired explicit knowledge across a vari-ety ofpatient groups, experimental tasks,and knowledge domains is compelling.Moreover, the converging evidence inseveral instances from studies of non-brain-damaged subjects indicates that thebasic phenomenon is characteristic ofnormal cognitive function and is notsome sort of exotic curiosity that occursonly in pathological conditions. What arewe to make of these striking and coun-terintuitive phenomena? Although cur-rent theoretical understanding of them israther modest, several different ap-proaches can be distinguished.The "family resemblance" among the

various implicit/explicit dissociationsacross a variety of conditions has sug-gested to some that it is appropriate toseek a common explanation for them.For instance, Schacter et al. (1) specu-lated that a common mechanism mayunderlie conscious experiences of per-ceiving, knowing, and remembering-ahigh-level system that takes as its inputthe extensively processed output of per-ceptual and semantic representation sys-tems and that must be activated for phe-nomenal awareness in different domains(see also ref. 68). They suggested further

that this mechanism can become selec-tively disconnected from individual brainmodules that process and represent par-ticular kinds of information. If such mod-ules continue to function relatively nor-mally, then the information on whichthey operate could affect performanceand behavior implicitly, without any cor-responding phenomenal awareness. Adisconnection account of this kind seemsto fit well with the neuropsychologicaldata because patients do not suffer fromgeneralized impairments of consciousexperience; their problems with explicitknowledge are domain specific (see alsoref. 55). One difficulty with this ap-proach, however, is that it implies theexistence of a "'consciousness module"despite the paucity of experimental evi-dence for such a module. Another ap-proach to a common explanation for avariety of implicit/explicit dissociationshas been put forward by Edelman (69),who suggested that they may be attribut-able to selective dysfunctions of re-entrant loops-connections among brainregions that, when activated, are ordi-narily responsible for particular kinds ofconscious experiences of perceiving,knowing, and remembering. This ap-proach represents a parsimonious at-tempt to accommodate numerous implic-it/explicit dissociations without postulat-ing a consciousness module, but there is,as yet, no direct empirical support for it.

In contrast to these attempts to de-velop a single account for a variety ofphenomena, other researchers have fo-cused on individual implicit/explicit dis-sociations. For example, a number ofinvestigators have suggested that prim-ing, skill learning, and other manifesta-tions of implicit memory reflect the ac-tivity of memory systems that are sparedin amnesic patients. These systems canfunction independently of the memorysystem that ordinarily supports explicitmemory, depends on the integrity of thehippocampus and related structures, andis impaired in amnesia (cf. refs. 3, 4, 35,and 70). To illustrate, it has been sug-gested that priming depends on changesin early-stage perceptual representationsystems that preserve information aboutthe form and structure, but not the mean-ing and associative properties, of wordsand objects (e.g., refs. 5, 33, 70, and 71).Neuropsychological and neuroimagingevidence indicates that such systems de-pend on posterior cortical structures (cf.refs. 33 and 72), which is consistent withthe proposal that they can function nor-mally in amnesic patients. Although ex-perience-induced changes in perceptualrepresentation systems can provide a ba-sis for facilitated identification of de-graded words and objects, they do notprovide access to the kind of contextualand associative information that is impor-tant for conscious recollection and that

appears to depend on the hippocampusand related structures. Thus, by this viewpriming and explicit remembering de-pend on different underlying memorysystems.An important observation supporting

this kind of multiple memory systemsaccount is that amnesic patients typicallyexhibit normal levels of implicit memorydespite severely impaired explicit mem-ory; accordingly, it makes sense to pos-tulate that independent brain systemssupport the two forms of memory. Bycontrast, in the other neuropsychologicalsyndromes discussed in this article, pa-tients typically do not exhibit entirelynormal performance on tasks that tapimplicit knowledge, so it is more difficultto argue that independent brain systemsunderlie explicit and implicit knowledge(for discussion of the "multiple visualsystems" approach to blindsight, seerefs. 55, 56, and 73). For example, Wal-lace and Farah (74) have suggested that insome cases of prosopagnosia, residualimplicit knowledge may be a natural con-sequence of impairment to the facial pro-cessing system that normally supportsexplicit knowledge. They noted that insimulations of prosopagnosia with a neu-ral network, when "lesions" are made toa part of the network that supports facialrecognition, the network still showssome residual ability to "perform" tasksanalogous to those used to demonstrateimplicit facial knowledge in prosopag-nosic patients. This kind ofobservation isconsistent with the idea that when pa-tients exhibit some, but not normal, lev-els of implicit knowledge, the effect maybe attributable to the impaired function-ing of a damaged system that normallysupports explicit knowledge.Because research in this area is still in

its infancy, it is too early to state confi-dently whether a unified theoretical ac-count of different implicit/explicit disso-ciations will be possible or whether it willbe necessary to construct separate do-main-specific theories for each particularkind of dissociation. Current evidencedoes suggest, however, that demonstra-tions of fully intact implicit memory inamnesic patients probably demand a dif-ferent kind of explanation than do dem-onstrations of residual (but not normal)implicit knowledge in blindsight, proso-pagnosia, and other syndromes.Whatever the ultimate theoretical ac-

count of implicit/explicit dissociations,the fact that these phenomena can beobserved in normal subjects as well asneurological populations indicates thatthey are not exotic or unusual symptomsthat represent pathological consequencesof brain damage. Nor are these dissoci-ations intimately intertwined with emo-tional conflicts or psychodynamic pro-cesses (e.g., repression) that were crucialto postulation of the Freudian uncon-

11116 Review: Schacter

Proc. Nati. Acad. Sci. USA 89 (1992) 11117

scious (e.g., ref. 75). Rather, dissociations between implicit and expliciknowledge seem to arise as a naturaconsequence of the functional architecture of the brain and reflect the activity ocomputations that are routinely performed during the course of perceivingrecognizing, and remembering. A majoichallenge for future research is to understand more deeply the properties of thearchitecture and the nature of the com-putations responsible for dissociationsbetween implicit and explicit knowledge.

I thank Dana Osowiecki for help with prep-aration of the manuscript. The article wassupported by Grant RO1 MH45938-01A3 fromthe National Institute of Mental Health.

1. Schacter, D. L., McAndrews, M. P. &Moscovitch, M. (1988) in Thought With-out Language, ed. Weiskrantz, L. (Clar-endon, Oxford), pp. 242-278.

2. Greenwald, A. G. (1992) Am. Psychol.47, 766-779.

3. Squire, L. R. (1992) Psychol. Rev. 99,195-231.

4. Weiskrantz, L. (1985) in Memory Sys-tems of the Human Brain: Animal andHuman Cognitive Processes, eds. Wein-berger, N. M., McGaugh, J. L. &Lynch, G. (Guilford, New York), pp.380-415.

5. Tulving, E. & Schacter, D. L. (1990)Science 247, 301-306.

6. Warrington, E. K. & Weiskrantz, L.(1974) Neuropsychologia 12, 419-428.

7. Graf, P., Squire, L. R. & Mandler, G.(1984) J. Exp. Psychol. Learn. Mem.Cognit. 10, 164-178.

8. Cermak, L. S., Talbot, N., Chandler, K.& Wolbarst, L. R. (1985) Neuropsycho-logia 23, 615-622.

9. Jacoby, L. L. & Witherspoon, D. (1982)Can. J. Psychol. 36, 300-324.

10. Moscovitch, M. (1982) in Human Mem-ory and Amnesia, ed. Cermak, L. S.(Erlbaum, Hillsdale, NJ), pp. 337-370.

11. Shimamura, A. P. & Squire, L. R. (1984)J. Exp. Psychol. Gen. 113, 556-570.

12. Schacter, D. L. (1985) Ann. N. Y. Acad.Sci. 444, 44-53.

13. Shimamura, A. P. (1986) Q. J. Exp. Psy-chol. 38A, 619-644.

14. Rozin, P. (1976) in Neural Mechanismsof Learning and Memory, eds. Rosen-zweig, M. R. & Bennet, E. L. (MITPress, Cambridge, MA), pp. 3-48.

15. Cermak, L. S., Bleich, R. P. & Black-ford, M. (1988) Brain Cognit. 7, 145-156.

16. Haist, F., Musen, G. & Squire, L. R.(1991) Psychobiology 19, 275-285.

17. Gabrieli, J. D. E., Milberg, W., Keane,M. M. & Corkin, S. (1990) Neuropsycho-logia 28, 417-428.

18. Musen, G. & Squire, L. R. (1992) Mem.Cognit. 20, 441-448.

19. Schacter, D. L., Cooper, L. A., Tharan,

M. & Rubens, A. B. (1991) J. Cognit.it Neurosci. 3, 118-131.

20. Bowers, J. S. & Schacter, D. L., in Im-plicit Memory: New Directions in Cog-nition, Neuropsychology, and Develop-ment, eds. Graf, P. & Masson, M. E. J.(Academic, New York), in press.

21. Cave, C. B. & Squire, L. R. (1992) J.r Exp. Psychol. Learn. Mem. Cognit. 18,

509-520.22. MacAndrews, M. P., Glisky, E. L. &

Schacter, D. L. (1987) Neuropsycholo-s gia 25, 497-506.

23. Tulving, E., Hayman, C. A. G. & Mac-Donald, C. (1991) J. Exp. Psychol.Learn. Mem. Cognit. 17, 595-617.

24. Richardson-Klavehn, A. & Bjork, R. A.(1988) Annu. Rev. Psychol. 36, 475-543.

25. Roediger, H. L., III (1990) Am. Psychol.45, 1043-1056.

26. Schacter, D. L. (1987) J. Exp. Psychol.Learn. Mem. Cognit. 13, 501-518.

27. Graf, P., Mandler, G. & Haden, P. (1982)Science 218, 1243-1244.

28. Jacoby, L. L. & Dallas, M. (1981) J.Exp. Psychol. Gen. 110, 306-340.

29. Roediger, H. L. & Blaxton, T. A. (1987)Mem. Cognit. 15, 379-388.

30. Graf, P. & Ryan, L. (1990) J. Exp. Psy-chol. Learn. Mem. Cognit. 16, 978-992.

31. Marsolek, C. J., Kosslyn, S. M. &Squire, L. R. (1992) J. Exp. Psychol.Learn. Mem. Cognit. 18, 492-508.

32. Schacter, D. L. & Church, B. (1992) J.Exp. Psychol. Learn. Mem. Cognit. 18,915-930.

33. Schacter, D. L. (1992) Am. Psychol. 47,559-569.

34. Milner, B., Corkin, S. & Teuber, H. L.(1968) Neuropsychologia 6, 215-234.

35. Cohen, N. J. & Squire, L. R. (1980) Sci-ence 210, 207-210.

36. Saint-Cyr, J. A., Taylor, A. E. & Lang,A. E. (1988) Brain 111, 941-959.

37. Squire, L. R. & Frambach, M. (1990)Psychobiology 18, 109-117.

38. Knowlton, B. J., Ramus, S. J. & Squire,L. R. (1992) Psychol. Sci. 3, 172-179.

39. Nissen, M. J. & Bullemer, P. (1987) Cog-nit. Psychol. 19, 1-32.

40. Daum, I., Channon, S. & Canavar, A.(1989) J. Neurol. Neurosurg. Psychiat.52, 47-51.

41. Weiskrantz, L. & Warrington, E. K.(1979) Neuropsychologia 17, 187-194.

42. Glisky, E. L. & Schacter, D. L. (1988)Neuropsychologia 26, 173-178.

43. Glisky, E. L. & Schacter, D. L. (1989)Neuropsychologia 27, 107-120.

44. Reber, A. S. (1967) J. Verb. Learn.Verb. Behav. 6, 855-863.

45. Reber, A. S. (1989) J. Exp. Psychol.Gen. 118, 219-235.

46. Dulaney, D. E., Carlson, R. A. &Dewey, G. I. (1984) J. Exp. Psychol.Gen. 113, 541-555.

47. Lewicki, P., Hill, T. & Czyzewska, M.(1992) Am. Psychol. 47, 796-801.

48. Poppel, E., Held, R. & Frost, D. (1973)Nature (London) 243, 2295-22%.

49. Weiskrantz, L., Warrington, E. K.,Sanders, M. D. & Marshall, J. (1974)Brain 97, 709-728.

50. Perenin, M. T. & Jeannerod, M. (1978)Neuropsychologia 16, 1-13.

51. Weiskrantz, L. (1980) Q. J. Exp. Psy-chol. 32, 365-386.

52. Torjussen, T. (1978) Neuropsychologia16, 15-21.

53. Zihl, J. (1980) Neuropsychologia 18, 71-77.

54. Campion, J., Latto, R. & Smith, Y. M.(1983) Behav. Brain Sci. 6, 423-486.

55. Weiskrantz, L. (1986) Blindsight (Clar-endon, Oxford).

56. Goodale, M. A., Milner, A. D., Jakob-son, L. S. & Carey, D. P. (1991) Nature(London) 349, 154-156.

57. Bauer, R. M. (1984) Neuropsychologia22, 457-469.

58. Tranel, D. & Damasio, A. R. (1985) Sci-ence 228, 1453-1454.

59. Young, A. W. & De Haan, E. H. F.(1992) in The Neuropsychology of Con-sciousness, eds. Milner, A. D. & Rugg,M. D. (Academic, San Diego), pp. 69-90.

60. Humphreys, G. W., Troscianko, T.,Riddoch, M. J., Boucart, M., Donnelly,N. & Harding, G. F. A. (1992) in TheNeuropsychology of Consciousness,eds. Milner, A. D. & Rugg, M. D. (Ac-ademic, San Diego), pp. 39-68.

61. Holender, D. (1986) Behav. Brain Sci. 9,1-23.

62. Marcel, A. J. (1983) Cognit. Psychol. 15,197-237.

63. Merikle, P. M. (1992) Am. Psychol. 47,792-795.

64. Cheesman, J. & Merikle, P. M. (1986)Can. J. Psychol. 40, 343-367.

65. Milberg, W. & Blumstein, S. E. (1981)Brain Lang. 14, 371-385.

66. Tyler, L. K. (1992) in The Neuropsychol-ogy of Consciousness, eds. Milner,A. D. & Rugg, M. D. (Academic, SanDiego), pp. 159-178.

67. Shallice, T. & Saffran, E. M. (1986) Cog-nit. Neuropsychol. 3, 429-458.

68. Schacter, D. L. (1989) in Varieties ofMemory and Consciousness, eds. Roe-diger, H. L., III & Craik, F. I. M. (Erl-baum, Hillsdale, NJ), pp. 355-389.

69. Edelman, G. (1991) Bright Air, BrilliantFire: The Matter ofMind (Basic Books,New York).

70. Schacter, D. L. (1992) J. Cognit. Neu-rosci. 4, 244-256.

71. Schacter, D. L. (1990) Ann. N.Y. Acad.Sci. 608, 543-571.

72. Squire, L. R., Ojemann, J. G., Miezin,F. M., Petersen, S. E., Videen, T. 0. &Raichle, M. E. (1992) Proc. Natl. Acad.Sci. USA 89, 1837-1841.

73. Cowey, A. & Stoerig, P. (1992) in TheNeuropsychology of Consciousness,eds. Milner, A. D. & Rugg, M. D. (Ac-ademic, San Diego), pp. 11-37.

74. Wallace, M. A. & Farah, M. J. (1992) J.Cognit. Neurosci. 4, 150-154.

75. Kihlstrom, J. (1987) Science 237, 1445-1452.

Review: Schacter