Abstract
The default view in the epistemology of forgetting is that human memory would be epistemically better if we were not so susceptible to forgetting—that forgetting is in general a cognitive vice. In this paper, I argue for the opposed view: normal human forgetting—the pattern of forgetting characteristic of cognitively normal adult human beings—approximates a virtue located at the mean between the opposed cognitive vices of forgetting too much and remembering too much. I argue, first, that, for any finite cognizer, a certain pattern of forgetting is necessary if her memory is to perform its function well. I argue, second, that, by eliminating “clutter” from her memory store, this pattern of forgetting improves the overall shape of the subject’s total doxastic state. I conclude by reviewing work in psychology which suggests that normal human forgetting approximates this virtuous pattern of forgetting.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
The default view should perhaps be understood as allowing that it is epistemically normative for a subject to forget information that she no longer accepts (information that has been left over after belief revision). See Sects. 2 and 3 for discussion of the role of forgetting in eliminating such “left over” information.
Though it is convenient to express the default view in this way, some might object to the formulation on the ground that only traits which are defects relative to those typical of the relevant larger population can count as vices. If the account of cognitive virtue set out in section 2 below is right, however, the formulation is acceptable, since the account does not require that cognitive vices be atypical.
The psychologist Gary Marcus has explicitly argued for something like the default view (2008); note that his discussion of memory and forgetting relies heavily on the computer model.
The focus on propositional records allows me to bring existing epistemological frameworks into contact with the psychology of memory in a relatively straightforward fashion, but the basic strategy of the argument can be extended to cover other types of records as well, including imagistic records. While considerations similar to those to which I appeal to show that forgetting of propositional records is necessary for epistemic virtue also suggest that forgetting of imagistic records is necessary for virtue, the argument for the latter conclusion will involve additional complexities, since frameworks for assessing the epistemic adequacy of cognitive processes involving such non-propositional representations are less well-developed.
There are psychological differences between the two types of non-encoding; see Schacter 2001 on transience vs. absent-mindedness. But (at least at the level of generality at which I am working here) they can be treated together for epistemological purposes.
The inaccessibility/unavailability distinction was first drawn by Tulving and Pearlstone (1966). Note that inaccessibility is often subjectively indistinguishable from unavailability. (The “tip of the tongue” phenomenon, in which a subject feels that she knows something but is presently unable to recall it (Metcalfe 1994), is an exception.) This probably accounts for our tendency to assume that forgetting is normally a matter of the complete elimination of a record.
This definition is less strict than that advocated by Wixted (2007), who proposes that we should say that forgetting occurs only if the cueing conditions in effect when retrieval is attempted are precisely the same as those that were in effect at an earlier time. While this strict definition might be appropriate for laboratory studies, where cueing conditions can be precisely reinstated, it has the consequence that forgetting can rarely be observed in non-laboratory contexts, where it is rare that precisely the same cues are used at different times.
Note that the recognition that forgetting is a matter of inaccessibility rather than unavailability should already render us less reluctant to accept the thesis that normal human forgetting is virtuous, for if forgetting is a matter of records becoming inaccessible, then it is not irreversible. See Sect. 2.
I emphasize that the vices in question are cognitive: there are cases in which forgetting might be prudentially but not cognitively appropriate (e.g., memories of personal trauma (Liao and Sandberg 2008)), cases in which forgetting might be cognitively but not morally appropriate (e.g., memories of one’s ancestors (Blustein 2008)), etc. The interactions among cognitive, prudential, and moral virtues and vices of memory is an interesting topic, but one which I cannot take up here.
Certain cases of persistence (in which a subject continues to recall a memory that she would rather forget) also seem to suggest that improved recall does not always result in an improvement to memory. See Schacter 2001 for discussion of persistence.
After the article in which her case is discussed was published, AJ identified herself publicly as Jill Price.
Though I rely here on examples of vicious remembering and forgetting due to abnormal memory systems, there can also be cases of vice due to abnormal uses of normal systems. (The well-known case of “S” (Solomon Shereshevsky), discussed by Luria (1987), might be of this type.) Given my focus on cognitive systems as virtuous or vicious—see Sect. 2—cases of the latter sort are less central here, though a complete treatment would have to take them into account (perhaps using a Zagzebski-style account (1996) of high-level virtues).
Whether a given subject is interested in information of a given type surely depends in part on her ability to retrieve information of that type; thus if we had a greater capacity to recall certain types of trivia, the information in question might cease to be trivia for us. I assume that we should assess a given cognitive faculty relative to the actual interests of the relevant subject rather than relative to the interests that she would come to have, were she to have the given faculty; thus the dependence of our interests on our retrieval capacities does not threaten my claim that a tendency to recall great quantities of trivia would not constitute an improvement to memory. AJ herself continues to be interested in the surplus information that she is able to retrieve (Parker et al. 2006, p. 39), so, strictly speaking, it is not trivia for her (at least not in the sense in which I use the term in Sect. 3). But presumably she would not continue to be interested in the information if she were not constantly and automatically retrieving it. In other words: most of the surplus information that a normal subject would be able to retrieve if she had AJ’s exceptional retrieval capacity is information in which she is not in fact interested and so, strictly speaking, is trivia.
This conception of virtue is particularly well-suited to my purposes here, as my focus is on the evaluation of a specific cognitive system. But, though Sosa’s conception contrasts with the virtue-responsibilist conception of Zagzebski (Zagzebski 1996), according to which virtues are certain acquired intellectual character traits, the virtue-responsibilist should be able to accept my basic conclusions (though she will want to rephrase them in other terms).
Lepock (2009) develops a similar approach in more detail; note that whereas I require speed for virtue, he replaces speed with the more general property of portability.
One might suspect that I have selected this nonstandard account of cognitive virtue precisely because it generates my desired conclusion that forgetting is necessary for virtuous memory. But, first, the account is independently plausible—power and speed are plainly desirable features in cognitive systems. Second, more standard accounts of virtue make it difficult even to ask the question about the normative status of forgetting, simply because they have nothing to say about the potential contributions of processes which eliminate but do not produce beliefs, and this is a legitimate reason for favouring my account. Finally, the account does not in fact by itself imply that normal human forgetting is virtuous but only that a certain pattern of forgetting is virtuous for creatures with finite computational resources; the extent to which normal human forgetting approximates this pattern is a further, empirical question.
It is plausible that the function of the system determines only a range of permissible levels, which implies that systems of the same type which employ somewhat different balances of reliability, power, and speed can all qualify as virtuous. My argument is compatible with this possibility, since it aims only to show that forgetting is necessary for the attainment of appropriate levels of reliability, power, and speed, without specifying a precise balance of these properties.
A question arises at this point about whether the reliability and power of memory are to be understood as conditional or as unconditional, belief-independent or belief-dependent. It will make sense to treat memory as belief-dependent if a record that P is normally stored in such a way that the subject is disposed to accept it when it is retrieved as a consequence of the subject’s accepting the content that P (believing that P) at the time of the encoding of the record, so that we can treat the earlier belief that P as an input to memory (despite the fact that memory does not literally store beliefs). I will, however, assume that the subject’s other cognitive systems are largely reliable, so that the majority of beliefs given to memory as inputs are true, which means that I will in general be able to ignore the distinction between conditional and unconditional reliability in what follows.
Bjork and Bjork use the metaphor of scaffolding to illustrate this feature of human memory: “We are fond of telling laypersons that our memories are not like a box in the sense that storing some information leaves less room for additional information. Rather, we say, a more appropriate analogy is that our memory is like a scaffolding structure of some kind such that the more developed (or elaborated) the structure the more additional ways there are to enter (or attach) new information” (1988, p. 285).
See McClelland, McNaughton, and O’Reilly (1995) for an influential explanation of this fact.
See the discussion of the relativity of the virtuousness of forgetting to the actual human memory system in Sect. 1 above.
It might be worried at this point that, depending on how the details of the default view are spelled out, the difference between that view and the view that I defend largely disappears. It we take the defender of the default view to be claiming that it is unfortunate that we have the finite computational resources which (I argue) render it necessary that we forget, the difference becomes one merely of emphasis: whereas the default theorist emphasis that it is unfortunate that we are in a situation that renders forgetting necessary, I emphasize that, given that we are in such a situation, it is fortunate that we forget. But the default view should not be understood this way. First, the default view on the epistemology of forgetting is plausibly taken precisely not to say anything about the implications of the fact of our finite computational resources—though there are a growing number of exceptions, most epistemological theories have not seriously taken the fact of finite computational resources into account. Second, a view which says that it is unfortunate that we have finite computational resources would anyway be strange, for the computational resources of any physical cognizer necessarily have epistemologically significant limitations.
I simplify by treating records as if they were discrete, an assumption challenged by views on which memories are stored only in a distributed, superpositional manner (Sutton 1998). A more realistic picture of the nature of memory traces would complicate my argument but should not affect the success of its basic strategy.
This is a simplifying assumption. See the discussion below of Cherniak’s description of the memory store as compartmentalized for a more precise statement.
It is natural to think of memory as storing and retrieving beliefs, but the thought involves a confusion. I take it that a subject has an occurrent belief that P when she has an activated representation that P that plays a certain role in her mental life—roughly: she accepts the representation as true (This is crude, but subtle differences among different conceptions of belief will not affect my argument here.). This can occur when a record is retrieved from LTM to working memory; but the record stored in LTM (obviously) is not an occurrent belief. I take it that a subject has a dispositional belief that P when she has a record that P stored in her LTM, she is disposed to retrieve the record in response to relevant stimuli, and she is disposed to form an occurrent belief with the record as its content (to accept the record as true) if the record is retrieved. (Not every record stored by memory would be believed if it were retrieved. Memory stores records stemming from imagining, fantasizing, etc. And memory normally continues to store the record that P even after the subject has abandoned her belief that P.) Though the subject will have a dispositional belief that P in part because she stores a record that P in LTM, long-term memory does not actually store the dispositional belief.
Memory also stores records of dreams, fantasies, etc., which the subject will normally not be disposed to accept. I bracket these in what follows.
It might be objected that I am not entitled to this assumption. This is not the place for the defence of such a general assumption, but I note that similar assumptions are made by many theorists; e.g., though they are interested specifically in misbelief, McKay and Dennett cite, in addition to Dennett’s work (1987), that of Fodor (1983) and Millikan (1984), as assuming that humans “have been biologically engineered to form true beliefs—by evolution” (2009, p. 493). It might be objected that the assumption starts to look particularly problematic when we consider the extent of our reliance on testimony; for a defence of the claim that formation of testimonial belief can be reliable despite our vulnerability to deception, see Michaelian 2010b.
The complete story about the reliability of memory will need to explain how memory can be reliable despite its constructive character; see Michaelian 2010a for an explanation of the compatibility of construction and reliability, emphasizing the role of metamemory.
Note that I am here using ‘outdated’ in a narrower sense than it often has in discussions of forgetting: the term often refers to information that is no longer relevant to the subject’s interests, whether or not it is still accurate.
The point is statistical: obviously, certain features of the environment are more or less fixed or at least invariant over long periods of time, so that not all records lose their relevance; the subject will continue to retrieve these records regularly.
The suggestion that forgetting is sensitive to retrieval history is not ad hoc; as we will see in Sect. 4, forgetting is indeed governed in part by retrieval history.
Recall the case of AJ: AJ’s memory is highly reliable, but it is remarkable especially in that it is exceptionally powerful: she is somehow able to retrieve far more records than those with normal memory systems can retrieve. This exceptionally powerful retrieval capacity comes at a cost: AJ spends a great deal of time simply processing her retrieved memories.
I have argued that a certain pattern of forgetting is virtuous in the sense that it is crucial for the performance of its function by memory. But forgetting can have additional cognitive advantages: given the pattern of forgetting associated with a virtuous memory system, other systems can exploit forgetting, for forgetting can itself serve as a source of information. Schooler and Hertwig have examined the benefits of forgetting for certain “fast and frugal” heuristics, including the recognition heuristic (Gigerenzer et al. 1999), the rule (in the case of a two-alternative choice) that “[i]f one of the two objects is recognized and the other is not, then infer that the recognized object has the higher value with respect to the criterion” (2005, p. 611). This simple rule is a useful heuristic “because lack of knowledge is often systematic rather than random”, so that “failure to recognize something may be informative” (2005, p. 611); the idea (counterintuitively) is that in certain contexts partial ignorance is better not only than total ignorance but even than lack of ignorance. Schooler and Hertwig suggest that forgetting can be beneficial because it enables us to use the recognition heuristic where we would otherwise be unable to use it: drawing on Anderson’s rational analysis of memory (Anderson 1990), they argue that forgetting enhances the performance of the recognition heuristic; the core idea is that if forgetting is a function of the frequency and recency with which information is encountered, then it can benefit cognition by making the recognition heuristic available for use.
Harman’s argument for CA does claim that there is “a limit to what one can retrieve” (1986, p. 41), which suggests that we might attempt to establish CA by an appeal to the second aspect of the finitary predicament. The idea would be that retrieval will not go well if one encodes clutter, presumably because then retrieval will often produce trivial beliefs. But the argument is not promising: given that forgetting is sensitive to retrieval history in the manner suggested in Sect. 4, forgetting will render trivial records inaccessible.
Of course, we care also about whether the beliefs are true; here, as throughout, I assume that we are dealing with subjects who acquire mostly true beliefs.
In fact, it is hard to see how we could accept CA without thereby committing ourselves to CE, whatever our reason for accepting CA.
This is not to say that virtuous forgetting necessarily implements CE perfectly: if the specific limitations on the subject’s computational resources mean that she must forget more than is required by clutter elimination, she will necessarily eliminate some interesting beliefs; but if her memory system is virtuous, she will preferentially forget clutter, thus minimizing the number of interesting beliefs that she forgets.
Though the view that human cognition (including memory) is adaptive is widespread, it is not uncontroversial. See, e.g., the responses to Anderson 1991 in the same issue.
See the discussion of the relativity of virtue to the normal human environment in Sect. 1 above.
Alternatively, we might content ourselves with the loose correspondence between adaptive memory and virtuous memory that obtains if we insist that virtuous memory preserves access to records that continue to interest the subject in whatever sense, as long as the correspondence is not too loose: if it can be argued—I will not propose such an argument here— that normal humans are after all fairly good at bringing their subjective interests into line with their dispositional and objective interests, the correspondence will be better than I have been assuming that it is, though still imperfect.
References
Anderson, J. R. (1990). The adaptive character of thought. Hillsdale: Erlbaum.
Anderson, J. R. (1991). Is human cognition adaptive? Behavioral and Brain Sciences, 14, 471–517.
Anderson, J. R., & Schooler, L. J. (2000). The adaptive nature of memory. In E. Tulving & F. I. M. Craik (Eds.), Handbook of memory (pp. 557–570). Oxford: Oxford University Press.
Bjork, E. L., & Bjork, R. A. (1988). On the adaptive aspects of retrieval failure in autobiographical memory. In M. M. Grueneberg, P. E. Morris, & R. N. Sykes (Eds.), Practical aspects of memory: Current research and issues (Vol. 1, pp. 283–288). New York: Wiley.
Bjork, R. A. (1989). Retrieval inhibition as an adaptive mechanism in human memory. In H. L. Roediger & F. I. M. Craik (Eds.), Varieties of memory and consciousness (pp. 309–330). Hillsdale: Erlbaum.
Bjork, R. A., & Vanhuele, M. (1992). Retrieval inhibition and related adaptive peculiarities of human memory. In J. F. Sherry & B. Sternthal (Eds.), Advances in consumer research (Vol. 19, pp. 155–160). Provo: Association for Consumer Research.
Blustein, J. (2008). The moral demands of memory. Cambridge: Cambridge University Press.
Boyer, P. (2009). What are memories for? Functions of recall in cognition and culture. In P. Boyer & J. V. Wertsch (Eds.), Memory in mind and culture (pp. 3–28). Cambridge: Cambridge University Press.
Cherniak, C. (1986). Minimal rationality. Cambridge: MIT.
Dennett, D. C. (1987). The intentional stance. Cambridge: MIT.
Fodor, J. A. (1983). The modularity of mind. Cambridge: MIT.
French, R. M. (1999). Catastrophic forgetting in connectionist networks. Trends in Cognitive Sciences, 3, 128–135.
Gigerenzer, G., Todd, P. M. and the ABC Research Group. 1999. Simple heuristics that make us smart. Oxford: Oxford University Press.
Goldman, A. (1992). Liaisons: Philosophy meets the cognitive and social sciences. Cambridge: MIT.
Goldman, A. (1999). Knowledge in a social world. Oxford: Clarendon Press.
Harman, G. (1986). Change in view: Principles of reasoning. Cambridge: MIT.
Kraemer, P. J., & Golding, J. M. (1997). Adaptive forgetting in animals. Psychonomic Bulletin and Review, 4, 480–491.
Lackey, J. (2005). Memory as a generative epistemic source. Philosophy and phenomenological research, 70, 636–658.
Lepock, C. (2009). Unifying the intellectual virtues. Philosophy and Phenomenological Research. (Forthcoming).
Liao, S., & Sandberg, A. (2008). The normativity of memory modification. Neuroethics, 1, 85–99.
Luria, A. R. (1987). The mind of a mnemonist. Cambridge: Harvard University Press.
Marcus, G. (2008). Kluge. Boston: Houghton Mifflin.
McClelland, J., McNaughton, B., & O’Reilly, R. (1995). Why there are complementary learning systems in the hippocampus and neocortex. Psychological Review, 102, 419–457.
McCloskey, M., & Cohen, N. (1989). Catastrophic interference in connectionist networks: The sequential learning problem. In G. H. Bower (Ed.), The psychology of learning and motivation (Vol. 24, pp. 109–164). San Diego: Academic Press.
McKay, R. T., & Dennett, D. C. (2009). The evolution of misbelief. Behavioral and Brain Sciences, 32, 493–510.
Metcalfe, J. (1994). Metacognition: Knowing about knowing. Cambridge: MIT.
Michaelian, K. (2010a). Generative memory. Philosophical psychology. (Forthcoming).
Michaelian, K. (2010b). In defence of gullibility: The epistemology of testimony and the psychology of deception detection. Synthese. (Forthcoming).
Michaelian, K. (2010c). Is memory a natural kind? Memory Studies. (Forthcoming).
Millikan, R. G. (1984). Language, though, and other biological categories. Cambridge: MIT.
Parker, E. L., Cahill, L., & McGaugh, J. L. (2006). A case of unusual autobiographical remembering. Neurocase, 12, 35–49.
Ratcliff, R. (1990). Connectionist models of recognition memory: Constaints imposed by learning and forgetting functions. Psychological Review, 97, 285–308.
Schacter, D. L. (2001). The seven sins of memory: How the mind forgets and remembers. New York: Houghton Mifflin.
Schooler, L. J., & Hertwig, R. (2005). How forgetting aids heuristic inference. Psychological Review, 112, 610–628.
Sosa, E. (1991). Knowledge in perspective. Cambridge: Cambridge University Press.
Suddendorf, T., & Corballis, M. C. (2007). The evolution of foresight: What is mental time travel, and is it unique to humans? Behavioral and Brain Sciences, 30, 299–313.
Sutton, J. (1998). Philosophy and memory traces: Descartes to connectionism. Cambridge: Cambrige University Press.
Tulving, E. (1999). On the uniqueness of episodic memory. In L. G. Nilsson & H. J. Markowitsch (Eds.), Cognitive neuroscience of memory (pp. 11–42). Gottingen: Hogrefe and Huber.
Tulving, E., & Pearlstone, Z. (1966). Availability versus accessibility of information in memory for words. Journal of Verbal Learning and Verbal Behavior, 5, 381–391.
Wixted, J. T. (2007). Forgetting: It’s not just the opposite of remembering. In H. L. Roediger III, Y. Dudai, & S. M. Fitzpatrick (Eds.), Science of memory: Concepts (pp. 329–335). Oxford: Oxford University Press.
Zagzebski, L. (1996). Virtues of the mind: An inquiry into the nature of virtue and the ethical foundations of knowledge. Cambridge: Cambridge University Press.
Acknowledgments
Thanks to Louise Antony, Charles Clifton, Hilary Kornblith, Chris Lepock, Joëlle Proust, Jonathan Schaffer, John Sutton, two anonymous referees, and audiences at an APIC seminar at the Institut Jean-Nicod and the 2010 meeting of the Dutch-Flemish Society for Analytic Philosophy for comments. The preparation of this article was supported in part by the Agence Nationale de la Recherche, under the contract ANR–08.BLAN–0205–01.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Michaelian, K. The Epistemology of Forgetting. Erkenn 74, 399–424 (2011). https://doi.org/10.1007/s10670-010-9232-4
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10670-010-9232-4