RSS Feed (xml)
Nielsen, 1958). but at the time of my essay these discussions had not exerted much influence on the thinking of experimental psychologists.
Like other ideas in science that are of more than passing interest, the idea of multiple memory systems, too, was not created in its final form at the outset. Since 1972 it has been revised and modified, and undoubtedly it will undergo further revisions and modifications in the future. In a progress report on the distinction some 10 years after the original essay (Tulving, 1983), I ventured the opinion that sufficient evidence had become available to warrant taking seriously the hypothesis that the two systems arc differentiable not only in terms of their behavioral/cognitive functions but also in terms of brain structures or mechanisms sub serving these functions. Such a separation between the systems, if true, means that one system is capable of operating independently of the other, that the operations of one system can be globally enhanced or suppressed without comparable changes in the activity of the other, and that the functioning of the two systems **is governed at least partially by different principles" (Tulving, 1983, p. 66). In a further development of the hypothesis, I have (a) elaborated on the concept of "memory system" (Tulving, 1984a); (b) proposed that the episodic system is embedded, as a unique subsystem, in semantic memory (Tulving, 1984b,1985a); (c) suggested that episodic and semantic memories are characterized by different kinds of consciousness (Tulving, 1985b); and (d) related the episodic/semantic distinction to the broader problem of classification of learning and memory (Tulving, 1985c). As a result, what had been a "potentially useful" heuristic distinction in 1972 has now become a full-fledged hypothesis, or theory, of the basic nature of memory.
Criticism of the Hypothesis
The heuristic usefulness of the distinction seems to have been beyond dispute from the outset. As far as I know, not a single student of memory has raised any questions about it. However, the idea that the two systems arc functionally and structurally different—which henceforth in this note will be referred to as "the distinction"—has run into a good deal of opposition and has engendered a certain amount of controversy. In the opinion of opponents and critics, it seems to violate the principle of parsimony. The critics argue that the postulation of the distinction is unwarranted not only because there is no convincing empirical evidence supporting it but also because the phenomena seen as relevant by the proponents of the distinction can be explained within the framework of a unitary memory (e.g., Anderson & Ross, 1980; Baddeley, 1984; Craik & Jacoby, 1979; Hintzman, 1984; Kihistrom, 1984; Lachman & Naus, 1984; Roediger, 1984;
Roediger & Blaxton, in press)[1].
McKoon, Ratciiff, and Dell (1986) have now added a strong voice to the chorus of dissent. They have assembled a large number of objections and criticisms of the distinction in a single article and have elaborated them with zeal and vigor. Although their review does not contain anything startlingly new—many putative difficulties with the distinction have been mentioned by other critics before—the McKoon el al. article is important because of its comprehensiveness and systematicity. McKoon and her coauthors have raised a banner around which the forces of the unity of memory can rally, and under which they can make their stand.
The more general conclusions that McKoon et al. have drawn from their analysis are easy to concede. Thus, for instance. I agree that the evidence for the distinction is not strong, that many facts supportive of the distinction can be explained in a single-memory framework, and that the two systems are highly interdependent But such a state of affairs means only that the hypothesis of the distinction is still open rather than closed.
A large number of specific objections that McKoon et al. raise against the distinction are questionable and debatable. However, an attempt at a point-by-point rebuttal of their arguments would constitute a wrong strategy for advancing knowledge. Sorting out fact from fiction in a complex issue such as the distinction requires the best efforts of a large number of people. Yet, on the contemporary psychological scene, nothing makes potential participants lose interest in a-problem as effectively as does a flood of intricate technical and polemical arguments, charges and countercharges regarding points of logic, methods, procedures, data, and theory. The debate about episodic and semantic memory deserves a better fate than being drowned in such a flood. The specific points that McKoon and her coauthors have raised and that are deemed important by others will undoubtedly be dealt with in the course of the continuing debate, and the wheat will be separated from the chaff. •
In this note, I touch on some of the general issues that McKoon et al. have raised. These issues have to do with the metatheoretical nature of the distinction, the appropriateness of the hypothetico-deductivo method for the evaluation of the distinction, and relation between data and theory.
What Kind of a Hypothesis?
The idea of differentiable episodic and semantic memory systems, in its most rudimentary form, is simply a hunch that there is something basically different about the way people remember their own past and how they store and retrieve general information about the world. The hunch was originally based on casual observations made by the early, philosophically oriented memory typologists (Herrmann, 1982), and then reinforced by clinical observations of dissociations in memory pathology (e.g., Cla-parede. 1911/1951; Warrington & Weiskrantz, 1968, 1970). Thus, for instance, when Claparede's patient expressed her newly acquired "knowledge" that doctors sometimes hide pins in their hands, and refused to shake hands with him—without remembering the very recent episode in which she had acquired this knowledge—it seemed only logical for Claparede to conclude that remembering personal events is not quite the same thing as acquiring knowledge through such events, and that one of these forms of teaming can be impaired as a consequence of brain damage, whereas the other seems to be less affected. (For a recent experimental analysis of the kind of dissociation noticed by Claparede, see Schactcr, Harbluk, & McLachlan, 1984).
Dissociations observed as a consequence of brain damage, as well as those arising from experimentally controlled influences on the workings of the brain, imply the existence of different kinds, classes, or categories of learning and memory, sub served by different neural mechanisms that can be selectively impaired.
Identification of such classes is the objective of an as yet nonexistent scientific discipline—systematics of learning and memory—that deals with classification of the myriad forms of learning and memory (Tulving, 1985c). Classification as a research objective or accomplishment constitutes an important part of scientific activity in many mature disciplines, but in (lie psychology of human learning and memory it has usually been conducted somewhat haphazardly and informally[2]. The hypothesis of the distinction between episodic and semantic memories (Cermak, 1984; Kinsboume A Wood. 1975; Rozin, 1976; Schactcr & Tulving, 1982; Tulving, 1983; Wood, Ebert, A Kinsboume, 1982) can be regarded as part of an attempt to systematize the enterprise, as an operation that parallels similar attempts in the field of animal learning (e.g., Mishkin, Malamut, & Bachevalier, 1984; Oakley, 1981, 1983; Olton, 1985; Pribram, 1984; Ruggiero & Flagg. 1976).
Because of the relative novelty of classification research in learning and memory, the rules by which it is conducted and the methods and procedures for determining what kinds of empirical facts are relevant are still being worked out Lessons learned from other sciences, such as systematic biology (Tulving, 1985c), or from classification in psychology, exemplified by the distinction between short-term and long-term memory (McKoon et al., 1986), are only partly relevant Some other methods that may have been useful within the field for other purposes, including procedures that McKoon and her coauthors seem to favor, however, are even less so.
Appropriateness of the Hypothetico-Deductive Method
What McKoon and her coauthors find wrong with the distinction derives largely from their limited view of what the science of memory—at least at the present stage of its development—is all about, and how our understanding of memory is to be advanced. In their article they seem to be saying that there is only one way of "doing science": One constructs a theory, preferably an "explicit model" (p. 304), from which specific predictions can be derived, or "clearly testable hypotheses" (p. 296) deduced, and (hen one tests the theory by comparing the outcome of the test with the prediction, or the hypothesis. In the absence of such a theory, McKoon et al. say. "we see no way in which the status of the episodic-semantic distinction can be clarified" (p. 304).
By implicitly advocating the use of the hypothetico-deductive method for the study of a relatively new yet complex brain/ cognition problem, McKoon and her coauthors have made what Gilbert Ryle (1949) refers to as the category mistake: They treat a concept that belongs to one category as if it belonged to another. In McKoon's and her colleagues' case, the mistake is twofold : Not only do they treat the science of memory as if it were a science such as classical physics, but they also treat a problem in classification as if it were one of discovering the causal basis of a particular phenomenon. It is probably their unawareness of their dual category mistake that emboldens McKoon and her coauthors to expect that everyone should study the episodic/ semantic distinction in their way or not at all.
It is easy to agree that on rare occasions the hypthetico-de-ductive method can play a useful role even in today's science of teaming and memory. But it is unlikely to get us anywhere in me task of working out a natural classincatory scheme for teaming and memory, one that is based on neuropsychotogical criteria (Kinsbourne, 1976, in press; Tulving, 1985c). The rutes by which data are related to theory are somewhat different in classification than they are in research aimed at elucidating the processes and mechanisms underlying behavioral/cognitive phenomena. McKoon and her colleagues' criteria for judging the validity of theoretical ideas arc different, too, and make it difficult, if not impossible, to bring the empirical data to bear on the theoretical status of even those classificatory systems that are true by the standards of nature.
To illustrate the difficulty, let me use an analogy. Suppose a world exists in which people see and hear things as we do, but in which they have only a single sense organ for receiving both light and sound signals, and in which the higher brain structures mediating vision and audition have not been identified. The prevailing theory of sensation in this fictional world is a parsimonious one: Sensation is sensation is sensation (cf. Kihistrom, 1984). Whatever differences exist between seeing and hearing are ac¬counted for in terms of (a) differences in the kinds of stimuli, and (b) differences in the way the stimuli are internally processed. Suppose further that in such a world there lives a scientist who has a hunch that vision and audition represent separate sensory systems, each subserved by separate neural structures and mechanisms[3].
How would our fictional wayward scientist go about evaluating his or her hypothesis? What kind of a specific model would the scientist construct, and what kinds of falsifiable predictions would he or she make? (What kind of falsifiable deductions would we draw from what kind of a model of visual and auditory systems as we know them in our world?) More important, given that the systems are in fact anatomically and physiologically separable, what would it mean if the scientist succeeded in falsifying a deduction from a particular theory of the two systems? For us, knowing that the hypothesis is in fact true, the finding would mean that the particular model was wrong or that the experiment provided an inadequate test of the model. For our fictional scientist’s hypothetico-deductive colleagues, not privy to nature's secrets, the finding would mean that the hypothesis is weakened. After a number of such falsifications, they are likely to conclude that the hypothesis is completely untenable. They would have failed to discover the truth, because of their insistence on the application of inappropriate logic to experimental results spawned by incorrect particular models.[4]
By McKoon's and her colleagues' rules of the game, our mythical scientist with the idea of two sensory systems could not get away with something as simple and straightforward as "predicting" that it is possible for people to suffer brain damage that results in blindness and not in deafness, or vice versa. Such a state of affairs could readily be interpreted without invoking the concept of separate systems. (To find someone like Helen Keller would be downright embarrassing for our scientist, because simultaneous blindness and deafness lends itself to a ready interpretation as "similar" or "parallel" effects, supporting the "sensation is sensation" theory). Nor would our scientist get very far if he or she emulated Kihlstrom (1980) and did a study in which a reversible, selective loss of function of one system is created experimentally. Kihlstrom (1980) found that posthypnotic amnesia instructions for words teamed in a list affected his subjects' ability to remember the studied words as self-referential, temporally coded events (episodic memory) but did not affect their ability to recall or produce the same words in a semantic task. McKoon and her coauthors claimed such findings to be irrelevant to the episodic-semantic distinction, since the self-referential nature and temporal coding of events is true of episodic memory "by definition" and hence cannot "speak to the reality of the episodic-semantic distinction independently of this definition" (McKoon et al.. 1986, p. 296). Comparable data would probably be resisted by the hypothetico-deductive methodists in the case of the different sensory systems as well, on the same grounds.
Our imaginary scientist with the "unparsimonious" idea about different sensory systems would also have a hard time convincing critics like McKoon and her coauthors by other means. Thus, for instance, inspired by Wood, Taytor, Penny, and Stump (1980), the scientist might first predict the expected patterns and then measure the regional cerebral blood flow in people while they process visual or auditory information. If the scientist finds this prediction vindicated by the data, he or she runs the risk of being told that the hypothesis of two systems "is not well-formulated enough to present dearly testable hypotheses" (McKoon et al., 1986, p. 296), and also that because of the high cost of such studies, additional work of this kind would be unlikely and, therefore, the obtained finding merely "intriguing" (McKoon cl al., 1986. p. 302). The scientist is unlikely to be told, however, exactly which kinds of "predictions" count and which ones do not.
Given the difficulties of pleasing the critics, what could our fictional scientist do? What can a harassed and beleaguered pro¬ponent of the episodic/semantic distinction do?
Relation Between Theory and Data
In the pursuit of the episodic-semantic hypothesis, we obviously have to be guided by empirical facts. No one is interested in a theory that makes no contact with the world of data. Fortunately, the problem of judging the fit between nature and our ideas about it is no more difficult for the hypothesis of differentiable episodic and semantic memory systems than it is for many other theories and hypotheses. Even McKoon and her coauthors, despite their repeated complaints about the testability of the distinction, agree that empirical observations can speak for and against the theory. They make statements such as, "most evidence about processes does not support [the] distinction" (p. 298), "much of the research on amnesia seems to have important implications for the episodic-semantic distinction" (p. 303). and a certain pattern of results, from as yet undone experiments concerned with evoked potentials, "would provide quite strong evidence for a separation of systems" (McKoon et al., 1986, p. 302). One could not make claims of this kind about a theory that is impervious to empirical data. Clearly, then, adjudication of the debate about memory systems can be based on empirical evidence as much as on other relevant criteria.
But not all empirical evidence is equally valuable for the purpose of settling the dispute. For instance, the complex reaction-time experiments that McKoon and her colleagues, as well as others, have done (e.g., den Heyer, in press; McKoon & Ralcliff, 1979; Neely & Durgunoglu, 1985) are unlikely to turn out to be critical. The findings of dissociations between episodic and semantic tasks in such experiments may have stimulated thoughts about the two systems, and their interpretations can frequently be aided by the hypothesis of two kinds of memory. But because of the long chains of inferential reasoning that these experiments entail, and the sensitivity of their outcomes to apparently minor variations in procedure (Neely & Durgunoglu, 198.?), their theoretical significance remains doubtful. If one-tries to evaluate the distinction primarily in terms of the outcomes of these kinds of experiments, then McKoon and her colleagues' misgivings about the enterprise may well be justified[5].
[1] Curiously enough, few people seem to have bothered to ask the question of why the distinction is so useful in the heuristic sense. The hypothesis that episodic and semantic memory represent different systems would provide a perfect answer to the question: "Because the distinction is real by the standards of nature." The unitary-memory theorists could presumably only scratch their collective heads in deep perplexity when confronted by the query.
The author's research is supported by the Natural Sciences and Engineering Research Council of Canada Grant No. A8632 and by a Sjxx'iul Research Program Grant from the Connaught Fund, University of
The author is grateful to Janine Law for editorial assistance. Correspondence concerning this article should be addressed to Endel Tulving, Department of Psychology, University of Toronto, Toronto, Ontario, Canada MSS 1A1.
[2] The distinction between short-term and long-term memory may be thought of as an exception to the assertion made. but it is still true (hat few theorists concerned with that issue perceived it as an exercise in the sysmtematics of learning and memory.
[3] The analogy holds only roughly, in that the hypothesized distinction between episodic and semantic memory, or at least one version of it (Tulving, 1985a), postulates one as a subsystem of the other, rather than viewing the two as parallel systems.
[4] That such things can happen in real life is demonstrated by McKoon's and tier coauthors' conclusion that "the difficulties overcame the distinction" (McKoon et al., 1986. p. 296) between short-term (primary) and long-term (secondary) memory, in the sense that a number of particular models of it were found wanting. I think that neuropsychological evidence for the existence of separate primary-memory and secondary-memory mechanisms is good enough to keep the issue very much alive (e.g.. Badddey A Wan-melon. 1970:ShaIIice. 1979: Warrineton. 1982).
[5] For a more optimistic assessment of the value of experimental dissociations, see Shoben and Ross (in press).
Postingan
Tidak ada komentar:
Posting Komentar