ABSTRACT: When one recalls that P, how is one justified in believing that P? The three most natural answers to this question are found inadequate: a memory belief is not justified by a belief in the reliability of memory; a memory experience does not provide a new, foundational justification for a belief; and memory does not merely preserve the same justification a belief had when first adopted. Instead, the justification of a memory belief is a product of both the initial justification for adopting it and the justification for retaining it provided by seeming memories. This view captures our intuitions about justification in several cases, while none of the alternative views can.
The sun is about 93 million miles away from the earth. How do I know that? Well, I learned it once. I don't know when or how I learned it, but I did, and I now remember it. I couldn't tell you how the distance to the sun was calculated either, but it's something that scientists have discovered. How do I know that scientists have discovered it? Well, I don't know how I learned that either, but I remember it too.
Even granting the reliability of scientists and other experts, this does not sound like a very impressive justification. Yet arguably, most of our knowledge is like that. A few more examples: there is a 3-hour time difference between Los Angeles and New York; Abraham Lincoln was President of the United States during the Civil War; the word "tree" refers in English to a certain kind of plant; the square on the hypoteneuse of a right triangle equals the sum of the squares on the other two sides; wood is a poor conductor of heat and electricity; China is in Asia. I don't know how I learned any of those facts, but however I learned them, I kept them in memory since then (doubtless I gained numerous confirmations of them since the first time I learned them, and I can't specifically remember any of those occasions either), and I have no serious doubt about any one of them.
What justifies me in believing that the sun is 93 million miles from the earth? The fact that I don't remember my original reason for adopting that belief suggests that whatever that reason was, it can not be considered a reason I now have for my belief.(1)
In general, when S remembers that P, what kind of justification does S have for believing P? Three possible answers to this question naturally come to mind:
First, perhaps my justification is inferential. And perhaps it is something like this: I now seem to remember that the earth is 93 million miles away from the sun. In the past, I have generally found that expectations formed on the basis of my seeming memories have been borne out. For example, I seemed to remember my address, and when I went to that address, I found an apartment of just the sort I was expecting. This strongly confirms that my seeming memories are highly reliable. Therefore (probably), it is true that the sun is 93 million miles away from the earth.
The most obvious problem here is one of circularity. How do I know that in the past, my seeming memories have been corroborated? Well, I seem to remember that that's generally been the case. But, on the present theory of memory knowledge,(2) I cannot trust that until I first prove the reliability of my memory. Therefore, I cannot use my past experiences in this way -- nor, in fact, in any other way -- in my argument for the reliability of memory.
Thus, if any inferential account is to work, the premises of my argument must rely solely on my present experiences and/or a priori insights. I can't use any previously-gained knowledge. It seems unlikely that I could derive the reliability of memory from premises of this kind; at any rate, I have no idea how such an inference would go. Additionally, an inferential theory would face two further constraints that increase its difficulties. First, the argument would have to be short and simple, such that one could hold it all in mind at once. Otherwise, completion of the argument would depend on one's remembering that the earlier stages of the argument had been correctly executed, and this would illicitly presuppose the reliability of memory.
Second, I would have to be in some sense using the argument every time I had a justified memory belief. It would not be enough for me to go through the argument once, and thenceforth merely remember that I had demonstrated the reliability of memory. For if merely remembering that my memory is reliable were enough for me to be justified in believing my memory is reliable, then merely remembering that the sun is 93 million miles away from the earth should be sufficient for me to be justified in believing that the sun is 93 million miles away from the earth -- contrary to the present theory, but in accord with the theory to be considered in section 2.
Given that my belief that the sun is 93 million miles from the earth is continuously present (it remains as a dispositional belief even when I'm not thinking about it), I will apparently need to be employing the argument for the reliability of memory continuously, if I am to keep my justification. The defender of the inferential account may claim that I am using this argument (whatever it is) for the reliability of memory only unconsciously, but it remains implausible that I am using it all the time, even unconsciously. Indeed, there is no evidence that I have ever employed any such argument at all, so skepticism seems to be the price of the inferential account.
Perhaps, then, my justification is non-inferential. Perhaps memory experiences create the same sort of foundational justification that (some epistemologists argue) sensory experiences do. Just having an experience of seeming to perceive that P makes one prima facie justified in believing that P, and similarly, having an experience of seeming to remember that P makes one prima facie justified in believing that P.(3)
This view has counter-intuitive results. Suppose I initially learn that P by means of an a priori proof of it (the proof is short, so I can hold it all in mind at once and do not need to use memory). So I have an adequate justification for believing P from the start, although the possibility of mistakes, even in short proofs, makes my justification less than completely conclusive. However, a few moments pass, and I now am able, in addition, to recall that P. If I entertain the proof while also remembering that P, I will now have two justifications for P, one inferential and one foundational. Thus, my warrant becomes more secure with the passage of time.
Here's another case. Suppose that I initially adopt the unjustified belief that P (perhaps by wishful thinking or some such irrational process). The next day, however, my belief is adequately justified, because I now seem to remember that P. The passage of time has transformed my irrational belief into a rational one.
It might be argued that in this latter case, I have a defeater for P, since I can recall that I adopted P by wishful thinking.(4) Therefore, modify the case as follows: a number of years pass, and I no longer recall how I initially "learned" that P, but I still clearly "remember" that P.(5) For example, suppose that I initially accepted the existence of life after death by wishful thinking. I now no longer remember where I got that belief, but I just seem to remember that that's something I know. On the other hand, my brother Pete adopted the same belief in exactly the same way. However, his memory is better than mine, so he also remembers how he got the belief. As a result, my belief system is rational and his is not. That seems wrong.
To further confirm that this result is wrong, Thomas Senor (1993) asks us to consider an analogy to moral philosophy. Suppose there is a certain ruthless tyrant, call him "Saddam", who decides to viciously attack a neighboring country. Suppose that at the time he makes this choice, Saddam's character is such that it would be psychologically impossible for him to behave in any other way. Suppose, however, that Saddam, having begun as a normal boy, acquired this deplorable character as a result of a series of evil choices that he made of his own free will. In this case, we would surely not excuse Saddam's present actions on the ground that he could not do otherwise. Rather, Saddam's culpability in his past choices follows him to the present day, rendering him culpable for the present evil actions that flow from them. Similarly, argues Senor, a person's previous epistemic irresponsibility follows him, making him epistemically blameworthy for any present likely-to-be-false beliefs that result from his previous irrationality. A present-day belief cannot be rendered epistemically justified by the fact that, doing the best one can do now results in acceptance of the belief, if this situation results from previous epistemic irrationality -- just as a present-day action can not be rendered morally blameless by the fact that, doing the best one can do now results in performance of the action, if this situation results from previous immorality.(6)
Both of the above cases -- the case where memory would increase one's justification for a belief, and the case where memory would convert an unjustified belief into a justified one -- point up the following general, intuitive constraint on a theory of memory justification: the justification for a belief cannot be increased by its passing into memory; it can only be lowered. The foundational theory fails to account for this.(7)
Here is a third view. When I remember that P, my justification for believing P is whatever it was to begin with. Memory just preserves the justification (or lack of it) of my beliefs.(8) So my justification for thinking that the sun is about 93 million miles away is, perhaps, that Mrs. Kim in second grade told me that it was -- even if I don't know that that is my justification. On this view, the fact that I don't remember what my original justification for P was does not prevent me from still having that justification for P. This seems more natural than the preceding two theories.
But now recall Russell's five-minute hypothesis. Suppose God created someone five minutes ago in exactly the state that I was in five minutes ago, surrounded by exactly the same kinds of things. Call this person Mike2. Mike2 was created complete with false memories of his past life, identical to my memories of my past life. He thinks his name is "Mike" and is presently writing a paper about the problem of memory knowledge. His situation would be (to him) indistinguishable from my actual situation. Usually, this scenario is mentioned for the purpose of asking, How do I know I'm not actually in that situation? But here I mention it to make a positive point. What sort of things would it be rational for Mike2 to believe? Pretty clearly, just the same things that it is rational for me, now, to believe (modulo appropriate changes in indexical references). Most of Mike2's beliefs about his own past are false, but he has no way of knowing that, and no more reason for suspecting it than I have for suspecting that my beliefs about my past are false. So if I am justified in believing that I ate a bagel this morning, Mike2 is justified (though mistaken) in believing that he ate a bagel this morning. Furthermore, it seems that he has the same degree and kind of justification that I now have.
But of course, this contradicts the present theory of memory knowledge. According to the present view, memory merely preserves one's initial justification for a belief. Therefore, I am adequately justified in believing that I ate a bagel this morning, on the basis of sensory evidence. Mike2 has no such justification, since he never had any sensory experience of eating a bagel. At minimum, he does not have the same sort of justification that I have, and it appears that he has no justification at all. On this theory, then, Mike2 is highly irrational (unlike myself), even though he is intrinsically identical to me.
Thus, there is an interesting problem of memory knowledge. The three most obvious theories of the justification of memory beliefs are all unacceptable. How can we find a theory that is not subject to any of the preceding objections? Our verdict on the case of Mike2 seems to demand that the justification of memory beliefs depend only on the current state of the believer, and not on his past; otherwise, Mike2 would be found to be drastically less rational than myself. But if the past history of a memory belief is thus irrelevant to its justification, won't this allow us to construct cases where memory transforms an irrational belief into a rational one (as in our objection to the foundational theory)? How, that is, can we reconcile the principle that the degree of justification of a remembered belief can never exceed the original degree of justification one had for its adoption, with the apparent lesson of the five-minute hypothesis, that the past history of a belief is irrelevant to its present justification? It seems that our intuitions are simply contradictory.
Not so. There is a theory that accommodates our intuitions about all of the cases, incorporating elements of both the foundational view and the preservation view. I call it the "dualistic theory" because it holds that the question, "What is my justification for believing that P?" requires a two-part answer: first, why I was justified in adopting the belief that P; and second, why I was justified in retaining it.(9) On this view, a belief is justified full stop if and only if it was adopted in an acceptable way and thenceforward retained in an acceptable way. The normal functioning of memory, in the absence of specific reasons for revising a belief, constitutes an epistemically acceptable manner of retaining beliefs.
So far, this sounds exactly like the preservation theory. However, we will see in a moment how, having distinguished two parts of a belief's justification, the dualistic theory is in a position to make an appropriate concession to the foundationalist account that avoids the major objection to the preservation theory.
It is already clear that the present view avoids the foundationalist's main problem. The dualistic view does not allow an initially irrational belief to become rational merely by passing into memory, since a rational belief, in the full sense, requires both rational acquisition and rational retention.
How, then, can the dualistic theory avoid the objection from the five-minute hypothesis -- how can it secure Mike2's epistemic rationality? Simply by this posit: coming to believe something by seeming to remember it (in the absence of specific grounds for doubt) constitutes an epistemically rational way of acquiring the belief. This posit captures the foundationalist intuition, that I am rational in believing something I seem to remember even if on this particular occasion, unbeknownst to me, my memory is deceiving me -- even if, that is to say, I never really had that belief before. From the standpoint of epistemic responsibility, this is surely correct. The unfortunate Mike2 has not committed any epistemic wrongs; he has done the best that could be expected of him. Our theory credits him this: since Mike2 acquired his belief that he ate a bagel this morning by seeming to remember it, he is rational in accepting it.(10)
But this posit does not introduce the possibility of memory's converting an irrational belief into a rational one. For the principle only applies to a case in which having a seeming memory that P was actually one's way of acquiring the belief that P. Recall the case where I believe P by wishful thinking and later seem to remember that P. Having a seeming memory in this case is not my method of acquiring the belief; wishful thinking is. Apparent memory is only my way of retaining the belief. Since a justified belief must have both a rational acquisition method and a rational retention method, this belief is unjustified.
It must be admitted that this view can not maintain the supervenience of epistemic justification on the current, intrinsic state of the believer. That, seemingly desirable characteristic is genuinely inconsistent with the conjunction of two other principles we have been assuming: first, that memory can not convert unjustified belief to justified belief; and second, that in typical circumstances our remembered beliefs are justified. For it is possible to have two people who are in the same state presently, each having forgotten his original reason for adopting P, one of whom did and the other of whom did not originally have a good reason for accepting P. The one person must be counted justified in his present belief (else we have memory skepticism), and the other must be counted unjustified (else we have an unjustified belief converted to a justified belief by the passage of time). It follows that the justificatory status of the belief that P does not supervene on the current, intrinsic state of the believer. Of course, it may still supervene on the total history of intrinsic states of the believer.
To illustrate, return to the case of myself and Mike2. Let's suppose that, among many beliefs I have for which I do not remember my original reasons for adopting them, there are some rational beliefs and a few irrational ones. I am justified in believing P, say, but unjustified in believing Q. Mike2, likewise, will be justified in believing P. But unlike me, on the present theory, Mike2 will also be justified in believing Q, since he, unlike me, acquired the belief through apparent memory. So there is one way in which the victim of the five-minute hypothesis would be epistemically better off than we actually are -- he has no fewer, and possibly more, justified beliefs.
On reflection, we can see that this result is correct and that the principle of current time-slice supervenience is therefore wrong. For Mike2, there is no relevant difference between his belief that Q and his belief that P. Both are adopted in the same way, so if we grant that his belief that P is justified, we have to allow his belief that Q to be justified similarly. Recall Senor's analogy with moral philosophy. Suppose that a person (call him "Saddam2") were created and placed at the head of a country, with a compulsion to invade a neighboring country. Saddam2 is born lacking free will, his decision to invade already predetermined. In that case, Saddam2 could not be morally blamed for his action. We have already said that Saddam, who acquired a similar psychological compulsion through earlier bad choices, can be blamed for the same action. So the moral culpability of a decision does not supervene on the internal state of the agent at the time of decision-making; it depends, too, on the agent's past choices. Likewise, we should not be surprised that the epistemic status of a belief depends in part on the believer's past thought processes.
So far, I have stated the dualistic theory as a theory of when a memory belief is justified or unjustified. But we can generalize the theory to give an account of the degree of justification that a belief has, and this generalization provides a further demonstration of the superiority of the dualistic view. The natural extension of the simple dualistic view would be to say that there are two degrees of justification involved in any belief -- a degree of justification associated with the adoption of the belief, and a degree of justification associated with its retention -- and that the overall level of justification of a belief is the product of those two quantities. The first of these two quantities is simply a matter of the conclusiveness of the grounds one originally had (again, this holds true even if one has forgotten those grounds). We can think of it as a number between 0 and 1, with 1 representing infallible justification for a belief, and 0 representing infallible justification for rejecting the belief. The second quantity is a matter of the credibility of one's memory, and it, too, can be thought of as a number between 0 and 1. If one has a relatively faint memory, such that one is quite unsure whether one really remembers that P or not, then this number will be close to ½. If one has a very firm and clear memory, the number will be close to 1. If one has special reason for doubting the reliability of one's memory (e.g., one has misremembered similar things in the past), this can lower the second number further.
One of our objections to the foundationalist account was based on the principle that the justification of a belief can be lowered through its passing into memory but can not be raised. The foundationalist could not accommodate this fact, because for him, the past justification of a belief is irrelevant to its present justification. But the generalized dualistic view easily accommodates the principle -- when one multiplies the original degree of justification by a number less than or equal to 1 representing the credibility of the memory, one necessarily gets something less than or equal to the original degree of justification.
The dualistic view also surpasses the straight preservation theory in the treatment of degrees of justification. Under the straight preservation view, the justification I now have for P when I remember that P is the same as the justification I had for P originally. Given this, the only natural view to take as to the degree of justification I now have for P is that it is identical to the degree of justification I originally had, on the principle that the degree of one's justification for P supervenes on what one's justification for P is. But this result is mistaken -- one should not be as confident that P ten years after learning it as one was when it was fresh in one's mind. One should not have 100% confidence in one's memory. The passage of time introduces new possibilities of error; therefore, it lowers one's justification for believing a proposition. Here, as elsewhere, the dualistic view succeeds in accommodating our intuitions about justification, escaping the objections that tell against the two main alternatives.(11)
One of two main theories of memory knowledge locates the justification of a memory belief solely in the memory impression. The other locates it solely in the original acquisition of the belief. As a result, one theory implies that memory can raise a belief's justification, while the other implies that memory cannot lower a belief's justification. The solution is to locate a belief's justification both in the circumstances of its initial acquisition and in the nature of the present memory experience.(12)
1. So argues Ginet (1975), pp. 153-6.
2. I am assuming that knowledge is a kind of justified belief, where justification is understood in terms of epistemic responsibility. If this is not what knowledge is, then what I am looking for should be described as "a theory of justified memory belief" rather than "a theory of memory knowledge".
3. Pollock's (1986) view, pp. 50-2.
4. It is not even clear that this reply is available. Consider a similar case: suppose I initially adopt P by wishful thinking, but later I perceive that P. At this point, my belief becomes justified. The fact that I initially adopted P through wishful thinking is merely irrelevant to the truth of P -- it does not count against P. It is unclear why, on Pollock's view, things should work any differently if memory is substituted for perception, since memory likewise provides foundational justification.
5. Annis (1980), pp. 325-6, uses this kind of case to refute Pollock's view.
6. Senor (1993), pp. 468-9.
7. Malcolm (1963), pp. 230-1, almost says this, except he does not seem to allow the possibility of one's justification being lowered.
8. This view is defended by Malcolm (1963), pp. 229-30; Annis (1980); Naylor (1983); and Owens (1998).
9. Owens (1998) makes use of this distinction, but he does not take advantage of the opportunity it provides to avoid the five-minute-hypothesis objection.
10. The following objection could be pressed: Mike2 is created with a host of dispositional beliefs implanted in him. Having never consciously entertained the propositions that these beliefs are about, he has not had any (occurrent) experiences of seeming to remember them, so the present theory cannot account for Mike2's justification for his dispositional beliefs. An obvious response would be to say that Mike2's beliefs are justified by virtue of his dispositional seeming-memories (quasi-memories). But a more interesting response, and the one I favor, is that Mike2 does not (can not) have the same dispositional beliefs that I have. A detailed discussion of this point would take us too far afield, but briefly: in order to genuinely believe that P, it is not enough that one would occurrently believe P if one considered it -- "dispositional belief" does not merely mean "disposition to believe". To believe P, a person must either (i) occurrently believe it, (ii) have once believed it, having never changed his mind about it, (iii) believe something else which presupposes it, or (iv) believe something else which obviously entails it. So there are three ways of dispositionally believing something, but each presupposes another belief or belief at an earlier time. For this reason, Mike2 cannot be created with dispositional beliefs already implanted in him; he can't have any dispositional beliefs prior to his first occurrent belief.
11. The inferential theory is not one of the two main theories; I do not know of any philosopher who actually defends it.
12. I would like to thank David Owens for stimulating my interest in and initial thoughts on this topic.
Annis, David B.: 1980, 'Memory and Justification', Philosophy and Phenomenological Research 40, 324-333.
Ginet, Carl: 1975, Knowledge, Perception, and Memory, D. Reidel, Boston.
Malcolm, Norman: 1963, Knowledge and Certainty, Prentice-Hall, Englewood Cliffs, NJ.
Naylor, Andrew: 1983, 'Justification in Memory Knowledge', Synthese 55, 269-286.
Owens, David: 1998, Reason without Freedom (unpublished ms.), chapter 9.
Pollock, John L.: 1986, Contemporary Theories of Knowledge, Rowman & Littlefield, Totowa, NJ.
Senor, Thomas D.: 1993, 'Internalistic Foundationalism and the Justification of Memory Belief', Synthese 94, 453-476.