Memory and persons

Philosophical Review

Volumn 112, Issue 3, 2003, Pages 289-337

  Burge, Tyler ^a  

^a UNIVERSITY OF CALIFORNIA   (United States)

Author keywords

[No Author keywords available]

Indexed keywords

EID: 34547180819     PISSN: 00318108     EISSN: 15581470     Source Type: Journal    
DOI: 10.1215/00318108-112-3-289     Document Type: Article

References (73)

1. One can, of course, have a memory presentation that p, but not believe it. My construal of "remembering that p" simply clarifies how I use the term. I believe that the less committal type of remembering-that and nonveridical memories are ultimately to be explained in terms of reliance on veridical memories.
2. I leave open whether video, television, or radio transmissions provide the relevant sort of direct awareness. I incline toward liberality. Grice in effect suggested that "I remember Napoleon's being defeated" more clearly requiresexperiencing the defeat first-hand. I find that intuitions vary even on this point. Some think that receiving direct reports and participating in the thrill of the event would warrant the gerund construction.
3. Tulving distinguishes episodic memories from semantic memories in "Episodic and Semantic Memory," in Organization of Memory, ed. E. Tulving and W. Donaldson (New York: Academic Press, 1972);
4. Elements of Episodic Memory (Oxford: Oxford University Press, Clarendon Press, 1983);
5. "Multiple Memory Systems and Consciousness," Human Neurobiology 6 (1987): 67-80. All experiential memories are instances of Tulving's episodic memories, with one caveat. Tulving explicates episodic memories as memories of events. I intend experiential memories to include a wider range. Some substantive content memories are Tulving's semantical memories. Those substantive content memories that do not involve de re constructions are semantical memories. Tulving does not distinguish substantive content memories from purely preservative memories. I also differ in allowing both to be nonpropositional. For recent discussion.
6. see A. Baddeley, M. Conway, and J. Aggleton, eds., Episodic Memory: New Directions in Research (Oxford: Oxford University Press, 2001). The generic types of memory that I discuss are further subdivided into working memory, shorter-term memory, and longer-term memory.
7. I use "purely preservative memory" here as I do in "Content Preservation" (Philosophical Review 102 (1993): 457-88.
8. reprinted in Content, ed. E. Villanueva (Atascadero: Ridgeview, 1995) and in Apriori Knowledge, ed. A. Casullo (Hampshire: Ashgate Publishers, 1999)), except that I am more explicit in applying the term to nonpropositional content preservations. In the earlier article I used "substantive memory" to cover both experiential memory and what I here call "substantive content memory." Substantive memory is distinguished by its role-and directly contrasts with purely preservative memory. It introduces subject matter and carries new warrant.
9. A de se form in memory must fix the framework of the context of the remembering. It will commonly retain a spatiotemporal framework from the remembered context, but may not. To be veridical, it must coordinate with the agent of the remembered context;
10. and to be a de se element in memory, it must normally function to further the needs, aims, perspective of the agent of the remembered context. There is evidence that animals that cache food also retain the time of the act of caching. Cf. N. S. Clayton, D. P. Griffiths, N. J. Emery, and A. Dickinson, "Elements of Episodic-like Memory in Animals," in Baddeley, Conway, and Aggleton, Episodic Memory. Animals like birds and dogs seem to have experiential memories with the two grades of de se involvement. It is less clear whether they remember acts from the inside (see below). This is a topic of conjecture and methodological debate in psychology.
11. This distinction comes from Shoemaker, "Persons and their Pasts," in his Identity, Cause, and Mind (Cambridge: Cambridge University Press, 1984), 27.
12. originally published in American Philosophical Quarterly 7 (1970): 269-85. I believe that the distinction was somewhat misformulated by Shoemaker, but this did not affect his primary points.
13. Cf. Gareth Evans, The Varieties of Reference, ed. John McDowell (Oxford: Oxford University Press, Clarendon Press, 1982), chap. 7.
14. As regards experiences, I think that we can (experientially) remember only our own, and only from the inside. So here the three grades are always filled. I think that we can remember another person's having an experience only from the outside. I remember your having pain, from the outside, because I remember your expressions of physical suffering. I cannot experience your pain, and I cannot remember from your perspective, as sufferer, on the pain, no matter how much I empathize and talk projectively about sharing your pain. I can remember my having an experience, say a pain, either from the inside-by remembering the pain-or from the outside. In the latter case my memory of having pain may derive from seeing myself in a mirror reacting to the pain. I can experientially remember an act as either my act or another's and either from the inside or from the outside.
15. A de se element in a representational content is distinct from a first-person concept, for reasons given above. The deep distinction between the two sorts of representational content and psychological capacity will surface only occasionally in this paper.
16. For discussion of indexes in the explanation of action, including preintentional action, see Marc Jeannerod, The Cognitive Neuroscience of Action (Oxford: Blackwell, 1997). The attribution of egocentric indexes is ubiquitous in perceptual and animal psychology.
17. Cf. my "Belief De Re," Journal of Philosophy 74 (1977): 338-62, sect. 2.
18. For an individual's states to have any representational content, some of this content must be associated with uses by and for the individual. This is one reason why thermometers and computers do not have autonomous representational content. Of course, they have a derivative kind of content in that they can be used for expressing and processing representational content.
19. For an argument on this sort of point that focuses on the mature first-person concept, but that is transferable with relatively obvious modifications to the contexts of this paper, see my "Reason and the First-Person," in Knowing Our Own Minds: Essays on Self-Knowledge, ed. Smith, Wright, and MacDonald (Oxford: Oxford University Press, Clarendon Press,1998).
20. For a discussion of these matters in somewhat greater depth, see my "Perceptual Entitlement," Philosophy and Phenomenological Research 67:503-48.
21. Sensation is an indicator that conduces to survival but is not genuinely perceptual. I discuss the distinction between sensation and perception in "Perception," International Journal of Psychoanalysis 84 (2003): 157-67. I intend to elaborate the distinction elsewhere. The argument that immediately follows in the text is also given in "Perceptual Entitlement."
22. I have discussed interlocution or testimony in "Content Preservation" and in my "Computer Proof, Apriori Knowledge, and Other Minds," Philosophical Perspectives 12 (1998): 1-37.
23. If the informant is not the ultimate source, there is still a difference in warrant. Although both informant and recipient are warranted by interlocution, the particular reliances are different, since each relies on different informants with different degrees of reliability in the causal chain. But the main point is that the ultimate source of warrant for the original instantiation of the step is different from the warrant of any given recipient through interlocution.There are, of course, arguments carried through jointly by many scientists, where no one scientist independently confirms all the steps. Such arguments are found convincing because they can be seen to be logically valid and because an individual can have meta-warrants that the steps he or she has not autonomously confirmed are warranted. For example, one is warranted in thinking that the scientists responsible for step 3 are reliable, hence that their advocacy is a sign of the step's being warranted. The meta-warrants must remain constant for a step from establishment of the step to its re-invocations, if the overall argument is to support its conclusion for an individual. Further, I think that one could not use meta-warrants unless one were capable of carrying out autonomous arguments that follow the scheme I am outlining.
24. As noted, section 10 elaborates and defends this line of reasoning. The argument does not claim that the agent of inference must have de se memories. But I think that any representational activity does require such memories in the psychological system. The argument's claim that the agent must have purely preservative memory yields the presupposition that the agent is the same through the inference, as well as the de se presumption that earlier instantiations of steps are the agent's own.
25. John Locke, Essay Concerning Human Understanding, ed. Peter H. Nidditch (Oxford: Oxford University Press, 1975), 2.27.9. It is not clear to me that this passage shows that Locke regarded memory as part of a definition or analysis of the concept of a person.
26. Joseph Butler, "Of Personal Identity," First Dissertation to the Analogy of Religion.
27. Shoemaker, "Persons and their Pasts," 24. I shall ignore the view that quasi-memory is a type of knowledge and regard it as a putative ability.
28. Derek Parfit, Reasons and Persons (Oxford: Oxford University Press, Clarendon Press, 1984), 220. Parfit sees himself as following Shoemaker in this characterization. In addition to the difference that I will highlight in this section, there are two other differences in their characterizations. One is Shoemaker's characterization of quasi-memory as a type of knowledge. The other is that Shoemaker writes of quasi-memory as being of an event that was experienced, whereas Parfit writes of quasi-memory as being of an experience. I think neither usage is sufficiently general to cover all types of memory and quasi-memory. This will not matter in what follows.
29. Shoemaker, "Persons and their Pasts," 24.
30. Shoemaker "Persons and their Pasts, "Ibid., 32.
31. I am not sure how aware either philosopher was of the features that I take to differentiate their characterizations. Parfit does not remark on the de se character of his characterization. Shoemaker never provides a completely sharp and straightforward characterization. Whether the distinguishing features of the characterizations were recognized will not matter.
32. Gareth Evans, The Varieties of Reference; Parfit, Reasons and Persons, 220-22, 516 n. 15.
33. Let me caution against a common mistake. Some infer from the possibility of systematic error in memory that systematic error is possible from the ground up. People do lose all accuracy in memory. Someone could also be systematically fooled in abnormal circumstances. These cases presuppose a memory competence that is subverted. It is impossible for an individual to have representational content, but never have an explanatorily fundamental memory capacity to preserve his own past representational states or events. In the case of many representational capacities, successful application may have occurred not in the life of the individual but only in the formation of the species' capacity. But purely preservative memory and de se aspects of memories are special. An individual whose activity was never correctly explainable in terms of successful preservation of its own contents from one moment to the next, or in terms of holding agency together with perception and aim over time, could not have a mind at all. Having attitudes with representational content is constitutively dependent on having de se and purely preservative memory competencies that issue in explanatorily relevant, successful holdings together, over time, of perceptions, needs, aims, and activities. In the absence of an explanatory core of veridical applications by the individual, there would be no representation.
34. The individual could learn inductively to distinguish presentations that connect to his own past. This would not suffice to give the presentations de se form. For that form necessarily yields motivations, norms, and acts that are immediate from the presentation itself.
35. Thus, I think that Parfit has a right to his de se conception of quasi-memory only if quasi-memory is, in explanatorily relevant ways, parasitic on normal memory. His conception is not strictly incoherent. We can imagine that a person connects to other people and cannot through any immediate phenomenological marker distinguish those cases from veridical memory. The competence could have a de se form (mistakenly applied in the cases of connections to others' past) only if fallible but successful coordination with the individual's own past states were the explanatorily fundamental case.
36. Several fine papers have cited practical agency as a problem for appeals to quasi-memory. Cf. Susan Wolf, "Self-Interest and Interest in Selves," Ethics 96 (1986): 704-20;
37. Christine Korsgaard, "Personal Identity and the Unity of Agency: A Kantian Response to Parfit," Philosophy and Public Affairs 18 (1989): 101-32;
38. John McDowell, "Reductionism and the First-Person," in Reading Parfit, ed. Jonathan Dancy (Oxford: Blackwell, 1997). These papers do not give the present arguments. Each contains large points that I do not accept. Yet each offers insights into what I consider to be sound and fruitful directions for thinking about the issues.
39. I always allow the use associated with aspects of psychological state-types corresponding to perceptual representations (contrast de se elements) to be established in an animal's conspecifics or species' ancestors (cf. note 25). One must, of course, not require that every representation have a use special to it. For example, representations of certain shapes that are of no use to an animal species might be individuated through their being constructed from principles governing representations of shapes that are of great interest or use.
40. An argument can be considered in the abstract, regardless of who, if anyone, put forward the steps, and regardless of what the warrant for the premises or steps are. We can construct (or abstractly, there is) an argument made of pieces extracted from the psychologies of different people. In some cases, warrants for the steps can perhaps be considered under impersonal standards of rationality. Moreover, the validity and soundness of any argument are personindependent. But we are considering argument and inference not as abstract sequences of propositions, but as psychological processes-as reasoning whose point is to support a conclusion. Such inferential reasoning is individuated not only through the logic of the argument but through the warrant of the steps. Deductive reasoning, considered as an inferential process that supports a conclusion, is meant not only to preserve truth and stay within the rules of deductive logic. It is meant to support or transfer warrant to a conclusion. A step in such reasoning must be considered to include its warrant. Reinvoking a step involves relying on its warrant. What I call "inferences" in this paper are norm-governed transitions in such reasoning. What I call "arguments" are chains of such inferences aimed at supporting a conclusion. The narrower, more formal conceptions of inference and argument presuppose these notions, I believe, in this sense: Any being that carries out inferences and arguments in any sense must be capable of carrying out inferences and arguments whose function is to provide warrant for a conclusion.
41. Epistemic internalists hold that warrants must be fully accessible to the warranted individuals, as reasons or justifications. I believe, by contrast, that not all warrants need be accessible to the warranted individual. Warrants are nevertheless always explained in terms of states and capacities of individuals, supplemented in some cases by relations to a subject-matter environment. I believe that epistemically externalist views join internalist views in either accepting the premise or yielding an analogous and equally effective premise that accords with their terminology and theory.
42. Insofar as there are de se elements in the content, of course, the referents of the indexes will shift. But the content can be type-identical.
43. The supposition that the content was warranted earlier through an exercise of quasi-memory will not change the situation. Each chain will have different warrant that depends on the nature of the chain from the quasiremembered event to the quasi-memory. (Cf. note 15.) In having different warrants, instances of quasi-memory are like instances of perception. Quasimemory is formally a new source of information. It is not purely preservative in my sense. Even supposing that an individual begins life with a set of reliable quasi-memories, these are sources of information and warrant for him. Although they derive from the states of another individual, and depend for their warrant on those earlier states' being warranted, they do not preserve warrant;
44. and they do not preserve information within the quasi-remembering individual. I should note that inference in practical reasoning, which is discussed near the end of section 4, is subject to the same-agent condition for two reasons. One has to do with the nature of steps in inference-by the argument just given. The other has specially to do with the fact that practical reasoning must preserve the practical commitments of the reasoner. In theoretical reasoning the commitment (to truth) is in a sense common to all inferrers. In practical reasoning the role of preserving the individual's own commitments-motivational elements-is additional. Individual practical reasoning would be doubly incoherent in the absence of the same-agent condition on purely preservative memory.
45. I believe that purely preservative memory clearly presupposes the presence of a de se memory competence in the same individual (as well as vice versa). But I do not rely on this belief in the present argument. It is enough that purely preservative memory constitutively involves an explanatorily relevant competence to preserve content from one's past. Purely preservative memory thus presupposes agent-identity and presumes the veridicality of de se attributions-whether or not the purely preservative memories are themselves de se.
46. I caution here about a special case of the mistake discussed in note 25. One might think that since we rely on others in interlocution, it does not matter whether a step of any argument comes from someone else. The broadbrush answer is that our reliance on others presupposes that our representational content and inferential abilities are already in place. If a recipient could not make intrapersonal inferences, taking lemmas or other propositional information from others would be impossible. Further diagnosis of this mistake can be derived from the discussion in this section.
47. For criticism of Descartes on memory see my "Content Preservation." Descartes was, of course, right to emphasize a role for noninferential comprehension in carrying out a proof. For a proof to be effective, there must be component steps that can be comprehended without inference.
48. This is the great lesson of Kant's emphasis on the role of inference in making judgment possible. Cf. Critique of Pure Reason, for example A 76ff./ B 102ff., B 130-31. It is also a deep theme in twentieth-century philosophy-stemming from work of Frege. I have in mind his assumption that meaning is to be understood only by understanding inferential structure, an assumption that suffuses his method in Foundations of Arithmetic, trans. J. L. Austin (Evan-ston: Northwestern University Press, 1968).
49. and "On Sense and Reference," in Translations from the Philosophical Writings of Gottlob Frege, ed. P. Geach and M. Black (Oxford: Blackwell, 1966).
50. Cf. my paper "Frege on Sense and Linguistic Meaning," in The Analytic Tradition, ed. David Bell and Neil Cooper (London: Basil Blackwell, 1990).
51. Cf. also Wittgenstein's Investigations, trans. Anscombe (New York: MacMillan, 1968). The picture of understanding as simply an immediate flash of insight has been repeatedly and convincingly undermined.
52. The point applies to norms as well as psychological competence, since relevant norms are grounded in such competencies and their functions. Normative standards for operating well in fulfilling the function of veridical representation, given the perspectival and operational limits of the individual, are the predecessors of epistemic norms. Applicability of such norms depends on certain processes' functioning to preserve fulfillment of them over time. A framework for these remarks is developed in "Perceptual Entitlement," sect. 1
53. Quasi-memories are like memories and interlocution in conveying information from prior representational states. They also resemble memories in that (presumably) one could reverse temporal order among them. Unlike all memories, they are formally or explanatorily sources of new information.Unlike purely preservative memories they are sources of new warrant
54. Of course, experiential memory and substantive content memory often embellish what was originally taken in. Some of this embellishment is distortion, and does not fulfill memory's representational function (though it may fulfill some biological or practical function). There is, however, the phenomenon of ordering, summarizing, drawing inferences, so as to extend the material of the original belief or experience
55. Cf. Daniel L. Schacter, Searching for Memory (New York: Basic Books, 1996). This may constitute an epistemic function of memory. I maintain, however, that insofar as it is an epistemic or representational function of memory, it must be a drawing out of elements implicit in material already present in memory
56. Shoemaker and Parfit seem aware of this threat. Each suggests that the first-personal elements in memory attributions derive from a "trivial" linguistic point (Shoemaker, "Persons and Their Pasts," 24;
57. Parfit, Reasons and Persons, 222-23). I have tried to show that de se- and first-person-presuming aspects of mental states are explanatorily fundamental. Fundamental psychological capacities are individuated in terms of de se-presuming competencies, which are in turn individuated partly in terms of the individual who is the agent of the competencies. Individuals are extended in time;
58. their perspectives, needs, aims, and activities are extended in time.
59. An alternative is that quasi-memory of another individual's past is not a nonveridical exercise of de se memory (that is, not the case I discussed in section 7), but a separate faculty, like perception. Such a faculty is still explanatorily and individuatively parasitic on genuine memory in the same psychological system. Explanation of its use, like explanation of perception, agency, and other representational capacities, requires attribution of memory proper. For being an individual agent requires being able to act out of an immediate sensitivity to one's own perspective, needs, and aims. This ability requires de se-presuming memories
60. Cf. my "Reason and the First-Person," 247ff.
61. Cf. my "Individualism and the Mental," Midwest Studies in Philosophy 4 (1979): 73-121.
62. and "Other Bodies," in Thought and Object, ed. Andrew Woodfield (London: Oxford University Press, 1982), 97-120.
63. Parfit's line is especially affected by the empirical success of individual psychology. One cannot eliminate the notions of individual agency, or of the individual, in giving an empirical psychological account of transactions in our world. Epistemic and practical norms are grounded in these empirical facts. Shoemaker's line is especially affected by the fact that the notions of agency, inference, and personhood are apriori connected to the explanatory centrality of de se presumptions, and their individuative presuppositions about individual identity. Memory is in on the explanatory ground floor of any account of the representational content of an individual agent. I have been emphasizing individuation and explanation in this section. Earlier sections, especially sections 4 and 10, also featured the role of representational norms (norms of veridicality) and epistemic norms in our understanding of presuppositions of agent identity in psychological states. Both reductionisms are vulnerable to criticisms based on normative considerations,but I will not undertake a separate diagnosis of these vulnerabilities here
64. Cf. Hume, A Treatise of Human Nature, 1.46;
65. G. C. Lichtenberg, Schriften und Briefe (Munich: Carl Hanser Verlag, 1971), 2.412.76;
66. Parfit, Reasons and Persons, 224-26. From the point of view of this project, the presence of a de se notion in the concept of quasi-memory would make that concept illicit. It is notable that Parfit's conception does not exclude de se elements from the concept of quasi-memory. Parfit tends to conflate the issue of reduction of the personal to the impersonal with two other issues. One is whether thinkers are "pure egos" that are "separately existing entities." The other is whether one can correctly and completely specify experiences and connections among them without presuming the existence of a subject that has these experiences. Materialists and many nonmaterialists would join Parfit in inclining toward a negative answer to the first question-pending clarification of its meaning. Parfit, the Hume before the appendix, Lichtenberg, Mach, and James maintain an affirmative answer to the second question. Our discussion of de se elements supports a negative answer
67. Cf. my "Reason and the First Person.
68. It is disputable whether the view is Locke's. Examples of this view are Shoemaker, "Self and Substance," Philosophical Perspectives 11 (1997): 283-304 (and references therein);
69. David Lewis, "Survival and Identity," in The Identities of Persons, ed. Amelie Rorty (Berkeley: University of California Press, 1976);
70. John Perry, "The Importance of Being Identical," in The Identities of Persons;
71. and H. P. Grice, "Personal Identity," Mind 50 (1941): 330-51.
72. Hume's skepticism about inference suggests this approach (cf. A Treatise of Human Nature, 1.4.1, 1.3.13). Another ersatz for inference might be regarded as warrant purely through credulity. If X seems to quasi-remember that p, X is warranted in believing p. Each surrogate for a step could have this sort of warrant. The view is incoherent if it omits purely preservative memory. A being that lacked a capacity for deductive inference could not have logical form for its representational states, hence no propositional attitudes. Such a view would also not account for the role of purely preservative memory in perception and agency. It would not account for use in representational transactions. So I think that this imagined case collapses into the case in which there is no inference and no representational content
73. This paper has an obviously Kantian flavor. It investigates necessary conditions on the possibility of having certain competencies. It also claims that psychological states presuppose a certain unity reflected in the form of the content of psychological states. Kant called his version of such unity "the transcendental unity of apperception." He saw such unity as necessary to applications of "I think" ( Critique of Pure Reason, B 131-42). By grounding his account in self-consciousness, Kant tends to ignore lower-level de se elements in mentality, present even in many lower animals. Such animals have perception but no thought, much less a first-person concept. At a more specific level, my arguments regarding inference are kin to Kantian suggestions about what is involved in holding a propositional thought together over time (cf. Critique of Pure Reason, A 98-110). I do not see in his work the specific argument I give. My arguments regarding perception are certainly congenial to Kant's view of empirical intuition (cf. A 98ff.). I know of no cousin of the argument regarding intentional agency in Kant. I did not develop these arguments by reflecting on Kant, but no doubt I was somehow influenced. Kant sees his account of the unity of apperception as "purely formal," with no immediate ontological significance for being in time. He bases this view on what I regard as an untenably restrictive epistemology. In the third Paralogism, which contains what must be the first appeal to quasi-memory, he imagines as an empirical possibility a series of persons who use "I think" in purported memories about their pasts, each having memory-like presentations that derive from the previous person (also applying "I think"). Each person mistakenly construes the previous one's past as his or her own. The case is primarily directed against the idea that we can through mere reflection have a certain scientific cognition that we are substances in the Cartesian sense. I accept Kant's rejection of this idea. But Kant also holds that only empirical experience could rule out such a series of erroneous self-attributions in an actual case. (Cf. Critique of Pure Reason, A 361-67 (esp. A 363 and A 363 n.), B 408.) I believe that this claim is mistaken. I think that such a case can be shown by apriori reflection alone to be impossible, or at least incompletely described. I think that Kant's target, "rational psychology" (which relies on certain features of Descartes's method of reasoning from the cogito) has more to be said for it than Kant allows. I do not accept Kant's restrictive epistemology of self-attribution, or his views about the cogito. Being fair to Kant's subtle position requires a much richer historical account, one I hope eventually to give