On Unclear and Indistinct Ideas Ruth Garrett Millikan Philosophical Perspectives, Vol. 8, Logic and Language. (1994), pp. 75-100. Stable URL: http://links.jstor.org/sici?sici=1520-8583%281994%298%3C75%3AOUAII%3E2.0.CO%3B2-S Philosophical Perspectives is currently published by Blackwell Publishing.
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/about/terms.html. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/journals/black.html. Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission.
JSTOR is an independent not-for-profit organization dedicated to and preserving a digital archive of scholarly journals. For more information regarding JSTOR, please contact
[email protected].
http://www.jstor.org Wed Jun 6 07:04:14 2007
Philosophical Perspectives, 8, Logic and Language, 1994
ON UNCLEAR AND INDISTINCT IDEAS*
Ruth Garrett Millikan
University of Connecticut
...for it is scarcely conceivable that we can make a judgment or entertain a supposition without knowing what it is we are judging or supposing about ...the meaning we attach to our words must be something with which we are acquainted ...[but] Julius Caesar is not himself before our minds. (Russell, The Problems of Philosophy, p. 58.) The difficulty with Russell's Principle has always been to explain what it means. (Gareth Evans, The Varieties of Reference, p. 89) Introduction Evans believed that there should be a way of explaining "what Russell's Principle means" that makes it come out true. To this end, he vigorously reinterpreted the notion "acquaintance," but his analysis of the varieties of this relation bears scant resemblance to anything in Russell. I will argue that under a very plausible interpretation which, I fancy, is considerably closer in one way to Russell and in another way to Evans than either is to the other, knowing what one is judging about turns out to be a matter of degree. The lowest degree may indeed be necessary for having thought at all, but the highest degree is never realized. As Evans saw, grasping the identity of the object of one's thought requires having a concept of that object, which requires, in turn, conforming to what Evans called the "generality constraint" (1982, p. 100). But a concept is an ability, a know-how, and unlike either know-thats or dispositions, know-hows come in degrees. One can know how but still fail. Indeed, it is common actually to be mistaken about the object of one's thought. But there is, on the other hand, a tendency for the object of one's thought td become whatever one takes it to be. The result is that thought can easily become equivocal. Misidentifying the object of one's thought is not, in central cases, an innocent act of false judgment, but an act that muddies the very content of the thought involved, corrupting the inner representational system. Reciprocally, the development and maintenance of relatively clear and distinct ideas is a substantive ongoing activity indispensable to the progress of thought. Descartes
76 1Ruth Garrett Millikan went astray only in failing to see how much more this activity involves than mere armchair reflection. A brief remark on methodology before the plunge. I will embrace without argument the Sellarsian thesis (Sellars 1956) that theories about the nature of thought are theories. Minimally, they are not mere descriptions, certainly not descriptions or "analyses" of concepts, but constructions. Hence I will not tolerate arguments of the form, "[ilnsistence on such cases involves an overthrow of our notion of what it is to possess a concept" (Evans 1982, p. 119) or "...the suggestion subverts the very logic or grammar of the concept of knowing what it is for it to be true that... (p. 116.)."
Part One: Concepts I. Knowing That I am Thinking of Alice
What then does it mean to claim that a person "cannot make a judgment about something unless he knows which object his judgment is about"? First, what kind of a knowing is this, a variety of knowing that, or a variety of knowing how? Suppose that it is a variety of knowing that: I cannot make a judgment about, say, Alice, unless I know that my judgment is about Alice. Is the claim then that knowing I am thinking of Alice is ontologically prior to or involved in thinking of Alice? Or is the connection epistemological, a sort of K-K thesis: thinking of Alice entails thinking that I am thinking of Alice? Russell espoused both of these connections. Thinking of Alice and knowing one is thinking of Alice are ontologically identical; within the mind there is no distinction between being and knowing; Russell's Cartesian mind is transparent to itself. But as soon as we depart from Russell's view it appears likely that neither of these connections holds. Knowing I am thinking of Alice is surely posterior rather than prior to thinking of Alice. I cannot know that I am thinking of Alice unless I first think of Alice, any more than I can know that I am hungry unless I am first hungry. Nor is knowing I am thinking of Alice the same thing as thinking of Alice. Knowing requires thinking, so knowing that I am thinking of Alice would require the capacity to think that I am thinking of Alice, hence the capacity to think about thoughts. But this is a capacity there is scant reason to suppose a thinker must have. For example, there is evidence that children don't have it until well after they acquire fluent speech. There is scant reason to suppose then that thinking of Alice must even be accompanied by thinking that I am thinking of Alice. That thinking of a thing cannot require knowing that I am thinking of it can be brought out in another way. Consider what it would be to know that I was thinking of Alice. Barring Russell's view of thought as direct confrontation of mind with object, this knowing could not involve directly comparing my thought with Alice. Rather, I would have' to think of my thinking and I would have to think of Alice and perhaps also of the relation that made the one a thought of the other. In any event, I would surely have to think of Alice. But if thinking of Alice involves knowing that I am thinking of Alice, and this
On Unclear and Indistinct Ideas / 77 requires thinking of Alice again, we are in a regress. It is conceivable, though false, I believe, that it should be necessary to have the capacity or the disposition, whenever my thought turns to Alice, to think that I am thinking about Alice. But it is not conceivable that actualizing this capacity should be constitutive of having thoughts about Alice. There is a second perhaps more familiar way of interpreting the notion "knowing what one is thinking of' which is not Russellian, and which takes this knowing to be a kind of knowing that. Thinking of Alice might involve thinking that the object of one's thought was that which bore certain uniquely distinguishing properties which in fact distinguished Alice. This route is not Russellian, of course, because Russell takes a thought with this kind of structure to be not a thinking of but an existential judgment. Nor would this solve the problem at hand, for it presupposes that we can think of properties. To think of properties, if Russell's Principle is right, we must know what properties we are thinking of. And this knowing cannot be analyzed in the same way again without a regress. Must we then think of Properties in the Russellian way, placing them directly before the mind?
II. Knowing How to Think of Alice If thinking of Alice, making judgments about her, is to require knowing what one thinks of, it seems that this cannot be a knowing that. Is it some kind of knowing how, then, with regard to Alice? Knowing how to do what with regard to Alice? Suppose we first ask the perfectly general question: what does one need to know how to do in order to think of Alice? And let us first explore the most general answer there is to this question: one must know how to think of Alice. Now that, you may say, is a perfectly silly, an entirely vacuous, answer. But I think it is not a silly answer. It is a true and important one. For the answer to the question might instead have been that one need not know how to do anything at all in order to think of Alice. Consider: in order to depress the carpet under your feet do you have to know how to depress the carpet under your feet? In order to trip and fall do you have to know how to trip and fall? In order to win the State Lottery do you have to know how to win the State Lottery? In order to do anything at all one must have had, of course, a disposition, under some combinations of internal and external conditions, to do that thing. But it is crucial that knowing how is not at all the same thing as just having a disposition. Know-hows reside in the order of purposes, not in the causal order bare. Know-hows are expressed only in purposeful doings. To say that thinking of Alice requires knowing how to think of Alice is to claim that this act must take place within the purposive order. One may go on then to explain how the purposive order takes its place within the causal order, but that leaves the original point intact. Thinking of does not occur in the causal order bare. But we must be very careful here. Having claimed, in this spirit, that thinking of Alice requires knowing how to think of Alice, what we must not do is to proceed to an analysis of purposiveness that rests it on intentionality. We must not take it, for instance, that residing in the purposive order requires that
78 / Ruth Garrett Millikan thinking of Alice should rest on intending to think of Alice. Not, at least, if one supposes that intending to do a thing requires thinking of or about what is to be done. That would lead back to supposing that thinking of Alice requires knowing that we are thinking of Alice-the regress discussed above. What sticks a thought onto its object cannot be an act of intending. This is not the place to present a full-fledged theory of purposiveness and intentionality. But I should like to advertise one. My own proposal has been that the purposiveorder is the order in which there exist historically fashioned "teleo-" or "proper" functions, of which biological purposes or functions are one variety, and that the intentionality of our thought rests on biological function (1984, 1986a, 1986b, 1989, 1990, 1991a, 1991b). Consonant with this, for an organism to know how to do x is for it to possess an intact mechanism that is biologically designed to do x, or that is designed to be tuned to do things like x and has been tuned to do x as designed. But biological design is not performance. Biological designs have, in general, a high rate of failure, not because of breakage, but because they often require quite special conditions to operate right (Millikan 1992). I will not press this particular issue forward here, though it will have to make cameo appearances. But let me point out that there does not seem to be any other proposal than the biological proposal on the horizon that makes sense of the fact that it clearly is true in general that knowing how is not merely having a disposition (1986a, 1990)l. But to observe that thinking of Alice requires us to know how to think of Alice does not cast much light on what more restricted ability, that might sensibly be called "knowing what we are thinking of', is presupposed. Evans held that this was "a capacity to distinguish the object of [one's] judgment from all other things" (1982, p. 89) -given our example, the capacity to distinguish Alice from all other things. Having this capacity, he said, was what made the difference between being capable only of judging, say, that a person had such and such attributes and being capable of judging that Alice had them (1982, pp. 1278). And Evans was for the most part clear that this capacity was some kind of ability or know-how, not a kind of knowing that. He was also clear that this ability to discriminate Alice could not be merely the ability to call to mind an idea that was, in some manner inaccessible to the thinker, externally (e.g., causally) hooked to Alice and Alice only. Rather, Evans thought, its being hooked to Alice must "reside in facts about what the [thinking] subject can or cannot do at that time" (1982, p. 116), facts determining that the thinker has a "concept" or "adequate Idea" of Alice. I will return to Evans soon. But first, I will propose my own reading of "knowing what one is thinking about." This will be another reading on which Russell too required us to know what our thoughts are about, and a reading on which, without doubt, we generally do know what our thoughts are about. This knowing is a sort of ability or know-how, and one that is naturally interpreted as the having of a concept. After some discussion of this notion, I will show how something very like this view of what it is to have a concept is implicit in Evans' "generality constraint" on concepts, hence in his notion of an "adequate Idea".
On Unclear and Indistinct Ideas / 79
III. Coidentifying What seems to be yearned for in the notion that I must know which object my thought is about is a sort of confrontation of thought, on the one side, with the object bare, on the other, taking place, per impossible, within thought itself. Indeed, Russell's view is that exactly this sort of confrontation is possible-the object bare is part of the thought. But, despite contemporary hyperbole that speaks of thoughts that require real objects in order to be thoughts as "Russellian thoughts," no thought actually consists in part of its object-any more than a mother, though she has to have a child to be a mother, consists in part of her child. The closest thing to the yearned-for ideal that actually makes some sense, I suggest, is a confrontation of one thought of an object with another thought of that same object, taking place within thought itself, and effecting a recognition of the sameness of the object. Putting this picturesquely, if you imagine the various thoughts that you have about, say, Saul Kripke, as a sort of story that you tell yourself using various thought tokens of him (including perhaps perceptual indexical thoughts of him), then knowing who you are thinking of in this story corresponds to your ability to make what Strawson called "storyrelative identifications" of the person in the story (1959, p. 18). There is no way that you can cut through the stories that you tell yourself about Saul Kripke in order to tack them inside your mind directly onto Saul, in order to know in any more direct way than that who you are thinking of. Suppose that knowing what one is thinking of is, just, having the capacity to recognize when two of one's thought tokens are thoughts of the same. Call the acts that realize this capacity acts of "coidentifying." In the first instance it is thought tokens that are coidentified, but thought types may be said to be coidentified when the thinker knows to coidentify their tokens, that is, the thinker is disposed to coidentify these tokens in accordance with an ability. Next let us loosen the notion to coidentify a bit so that it is not a success verb but only a verb of trying. As believing is to knowing, so coidentifying is to recognizing real sameness in objects of thought: coidentifying can take place mistakenly. Let us further add a harmless equivocation. When a thinker coidentifies thoughts, she also coidentifies the objects of these thoughts. This will be a slightly different sense of "coidentify", of course, but one that is closely analogically related. Last, just as thinking of an object requires knowing how to think of it, correctly coidentifying thoughts hence objects requires knowing how to coidentify them; coidentifying takes place in the purposive order. Should it occur to you to wonder why I call what seems to be, merely, identifying, by the awkward name "coidentify", then you are right on my track. I believe, indeed, that not only everything properly called "identifying", but also "reidentifying" (Strawson 1959, pp. 31-8) and "re-identifying" (Evans 1982, p. 126) also has the structure described? The,point of the term "coidentify" is to emphasize, first, that whereas this act involves only one object (when it is correctly executed), it always involves two thought tokens. Second, the point is to emphasize the symmetry of this act, which is of considerable importance.
80 1Ruth Garrett Millikan Third, the point is to make it possible to talk of coidentifying not merely pairs but larger sets of thoughts. Suppose, for example, that I see that that woman, the one just ahead up the block, is walking, and suppose that I take her to be my friend Alice. That is, I identify her. I identify her as Alice. In doing so I coidentify one thought, that woman, with another thought, Alice. Or if a percept is a sort of representation and can represent a person, perhaps what I do is to coidentify the object of a percept with that of a thought-the details are not important. In any event the result of coidentifying is that I take Alice to be now walking. Later I see Alice again in the market and reidentify her. In so doing I coidentify another thought, that woman, with the thought Alice, hence also with my earlier thought that woman. The next day I may hear that the city mayor is in Washington and, coidentifying my thought the city mayor with my thought Alice, hence also with my two earlier thoughts that woman, take Alice to be now in Washington., The proposal on the table is that if these acts of coidentifying are correct, and correct not by accident but as the result of a genuine ability, then I have manifested my knowledge of the identity of the object of the various thought tokens coidentified. I have manifested my knowledge that I was thinking, each time, of Alice. Knowing what I am thinking of is being capable of coidentifying my thought with other thoughts of the same. It is being able to distinguish thinking of a thing again from thinking of some different object. Russell required that a person know the object of his thought in this manner. Why is it important for Russell that the object of thought should be "itself before our mind"? Because since all aspects of what appears before the mind are transparent to mind, the identity or difference of objects directly before our minds cannot possibly fail to be recognized. Why, for Russell, do definite descriptions not express thoughts of the objects they denote? Since Russell takes it that two thoughts of the same object cannot fail to be grasped as of the same, but since two definite descriptions of the same certainly can fail to be grasped as of the same, Russell cannot take definite descriptions to be thoughts of objects.
IV. Coidentifying and Evans' Generality Constraint Let me now sketch the connection between this way of "knowing what the object of one's thought is" and Evans' view of what it is to have a concept of an object. Detailed exegesis of Evans' text is not the point, but rather its suggestiveness. Mainly I hope to make clear how "knowing what the object of one's thought is," when interpreted this way, can very plausibly be taken to constitute having a concept of an object. Suppose that I think to myself Alice is slim and then I think Alice is trim. If having done this I also grasp that, despite their different sentential contexts, these two thought tokens of "Alice" have the same object, that the same thing is both slim and trim, that is what it is fqr me to coidentify these two tokens of "Alice". Now perhaps that strikes you as an odd example, or at least as an odd way to put things. For perhaps it strikes you that thinking two tokens of the same thought type, as it were, beside one another, or in the same breath of consciousness, surely constitutes an act of coidentifying, rather than requiring to
On Unclear and Indistinct Ideas / 81 have one superimposed. Later I will attempt to disillusion you about this. But I propose that we make the simplifying interim assumption that tokens of the same idea type are, under specifiable and quite ordinary conditions, automatically coidentified. If we also assume that to coidentify ideas of different types disposes one, perfectly generally, to substitute one in for the other in judgments, then we can model the act of coidentifying as iterating, or as becoming generally prepared to iterate, a new token of an old idea type in the context of a new judgment.3 For example, each of the various cases mentioned above in which tokens of the thought Alice are coidentified with tokens of other types, can also be viewed as acts that iterates a new token of the old idea type, Alice, in a new judgment: Alice is walking; Alice is in the market; Alice is in Washington; etc. Now Evans takes it that thinking of a thing requires having a "concept" of it or, in the case of objects as distinguished from properties, an "Idea" of it. A concept or Idea is a general ability that "makes it possible for a subject to think of an object in a series of indefinitely many thoughts, in each of which he will be thinking of it in the same way" (1982, p. 104). For example, to have an Idea of Alice, I must be able to think of Alice not only, say, in the context of the thought that she is slim, but also (given appropriate predicate concepts) in the context of the thought that she is trim, that she is walking, that she is city mayor, etc. More precisely, I must understand what it would be for Alice, as distinguished from all others, to have each of these various attributes. Evans calls this constraint on concepts "the generality constraint" (1982 section 4.3). Evans' "generality constraint" is not just the familiar contemporary view that thought must be compositional. The verificationist background from which Evans' thought emerged lends it quite another flavor and use. It implies, rather, a general capacity to reiterate the thought Alice in other evidenced or grounded judgments about her. According to Evans, "[Iln order for a subject to be credited with the thought that p , he must know what it is for it to be the case that p" (1982, p. 105). To be sure, this kind of knowing what "is hard to give any substance to...when this is not to be equated with an ability to determine whether or not [p] is true" (1982, p. 106). But Evans thinks we are obliged to try. He tries by unpacking "know what it is for it to be the case that" (say) it is Alice who has this or that property, by reference, first, to possession of a "fundamental Idea" of Alice, 6, based on the criterion of identity associated with her defining category -the criterion for same personhood, presumably. The notion of a criterion of identity is relational; it tells what constitutes being the same one again. The unpacking proceeds, next, to the requirement that one understand "what it is for it to be the case that ...Fa= a]" for various other kinds of ideas, a, such as descriptions and indexical thoughts. Thus the problem is reduced, in large part, to the question what it is to "know what it is for it to be the case that" various kinds of identity equations hold. Evans' discussions of these equations all exemplify a common theme. Understanding what it would be for A to equal B is always described such that it involves having the capacity to coidentify A with B, or involves having a more general capacity under which this capacity falls. For example, my concept of Alice is expressed in my ability to recognize her on sight ("recognition based
82 / Ruth Garrett Millikan identification" (Evans 1982, chapter 8)), which is a way, of course, of iterating grounded thoughts of her in new judgment contexts-a way of coidentifying her. Similarly, where P is an indexically indicated position in egocentric space and p a position in public space, that] in which knowledge of what it is for identity propositions of the form P = p l to be true consists" in "the capacity to discover...where in the world one is" (1982 p. 162), that is, "[the] ability to locate [one's] egocentric space in the framework of a cognitive map" (p. 163). Such "locating" clearly is another act of coidentifying. Evans returns several times to an example of a man who retains the memory of a steel ball he once saw, but retains no information as to when or where he saw it, nor concerning any other characteristic that would distinguish it from an identical ball he also once saw but forgot. Evans claims that this man has no Idea of the remembered ball, that he is not, in any theoretically interesting sense, thinking of it. This is because "our subject's supposed idea of that ball is completely independent not only from any possible experience, but also from everything else in his conceptual repertoire. There is no question of his recognizing the ball; and there is nothing else he can do which will show that his thought is really about one of the two balls (about that ball), rather than about the other" (p. 115). The point, I take it, is that this man seems to be debarred in principle from ever making another grounded judgment about that ball-from ever reasonably coidentifying his supposed thought of it with any other thought of it. For this reason, his supposed thought does not meet the generality constraint, he has no concept of the ball, he does not know which ball his memory is of. And Evans concludes, also, that he cannot, properly speaking, be said to have the capacity to think of that ball at all, but only of a ball. Exegesis of Evans to one side, I propose that a foundational cognitive capacity is the ability correctly to coidentify one's thoughts, a capacity that is not a kind of know-that, but rather a capacity for a certain kind of movement in thought.4 And I propose that having this capacity, with regard to a thought, is what it is to have a concept of the object of that thought. Let me now spell out certain consequences of this view.
'I
V. As Abilities, Concepts are Fallible and Rest on Alternative Means Evans' man harboring a memory trace of the ball doesn't realize that he ever saw any other such ball. So he might wrongly take certain descriptions of it to be uniquely identifying and as a result, later coidentify his memory of the ball with another ball he is told of. But if he coidentifies it that way correctly, this will be only by accident, not in accordance with a genuine ability. So this doesn't help him to have a concept of the ball. Because abilities belong to the purposive order, not all dispositions are abilities. Conversely, abilities are not merely dispositions, not even dispositions to succeed in one's purposes, say, to succeed when one tries. I know how to walk yet may trip this time when I try; I know how to cook, but may bum the dinner tonight. One can have an ability to do what one is not able-at the moment-to do. Still, to have an ability does require that there be some conditions under which trying will bring success. Namely, I have suggested, one will succeed under conditions required for normal
On Unclear and Indistinct Ideas / 83 correct functioning of the tuned biological mechanisms that are responsible, as so tuned, for one's ability. What having an ability does not require, however, is that one be unfailing in recognizing when these required conditions obtain or even, indeed, that one understand what these conditions are. Accordingly, to have an ability (correctly) to coidentify a thought does not require that one never misidentify (the object of) that thought. It does not require that one always recognize conditions under which one's acts of coidentifying will go wrong. Equally important is that abilities typically rest on alternative means. My ability to get from home to school rests on many alternative means. So does my ability to tie my shoes, though there is one way at which I am most practiced and best. Similarly, the capacity to coidentify a thought typically rests on a variety of alternative conceptual means. The number of ways I can coidentify thoughts of each of my daughters is nearly innumerable-through appearance of body or body parts, postures, clothes, sounds of voices and feet, characteristic activities, handwriting, various nicknames and dozens and dozens of descriptions. But although it is clear that there are many such ways at my disposal, there seems to be no way to say exactly how many. How many ways can you identify. your best friend by looking? There is no clear principle by which to count these ways. Thus concepts may include numerous means of conception, without there being clear demarcations among these means.5 VI. Conceptual Know-how vs. Knowing About World Structure
If the above sort of description of having a concept is right, however, we must depart from Evans' views on one very crucial point-on the role of "fundamental Ideas of objects." These Ideas, Evans said, are based on grasping "criteria of identity" for objects. I am not sympathetic to any form of linguistic idealism, hence not to the notion that there are "criteria of identity" that we employ which determine object identity. Elsewhere I have defended a thoroughly realist notion of the structure of both object and property identity (1984, chapters 16 and 17). But sanctioning "criteria of identity" would not give help where we need it anyway. What, for example, would the relevance be of having a "grasp of the criterion of identity for persons over time" or of "places over time" to a practical ability to recognize the same person or place again over time, hence to iterate thoughts of this person in new judgments? We don't reidentify persons by following their space-time worms around. Besides, dogs are quite good at recognizing their masters, babies at recognizing their mothers, even though it is quite certain that neither conceives of a criterion of identity for persons over time. Research in child development suggests that children don't even acquire concepts of time until about age four, though they certainly can recognize their mothers. Just as I do not have to be able to describe or even to recognize the conditions required for exercise of other abilities of mine, for example, just as I can ride a bicycle without understanding the laws of dynamics, neither do I have to understand the ontological structure of the world upon which my (of course fallible) coidentifying abilities depend. Nearly the whole history of philosophy to the contrary, analysis of the world structures that account for the possibility of human knowing is not the
84 / Ruth Garrett Millikan same thing as analysis of the content or structure of what is known. For example, a venerable tradition argues that the possibility of identifying or having concepts of other individuals depends upon the fact that each such individual is uniquely located relative to us in space-time. This is surely a valid point, but not because conceiving of other individuals requires us to think of their relations to us, anchoring our thoughts of them beginning with thoughts of ourselves. The valid point is that having a concept of anything at all involves the capacity to coidentify it, which in turn means interacting with it, actively collecting together various manifestations of it that impinge upon our senses, that appear in our thought over time. And obviously one cannot collect together manifestations of something not in one's space-time system. What is true and important is that the activity of collecting new truths about any individual can be accomplished only in so far as our world has a certain space-time and causal structure in which we too are ingredient and to which we are attuned. That is, for the most part we can find our way about in it. This should not be confused with the assumption that knowledge of or thoughts about this structure are required for success in this activity. The capacity to reiterate Alice in new grounded judgments is a high level skill exercised in the world. It is not something done inside one's mind. VII. How Concepts Grow
A concept, we have noted, may have many conceptual means or "ways of being applied". If we use the latter familiar mode of expression, however, we must be careful not to forget the symmetry of the act of coidentifying. "Applying a concept" strongly suggests that a thought or mental term is directly applied to the world, whereas in fact coidentifying always involves two thoughts, and involves them equally. If I coidentify this person before me with Peter, I coidentify Peter with this person, learning about Peter from my observation of this person, but at the same time, perceiving and interpreting what this person is doing in the light of his being Peter. Using the act of iterating a thought type in new judgments as a model for the act of coidentifying has a related disadvantage. To speak of iterating the thought type Peter focuses attention upon one term only of the act of coidentifying, whereas every iteration of Peter in fact involves both the thought Peter and some other thought or percept. Concepts are really best thought of as determining networks of thoughts, joined by sets of coidentifying capacities. These specific capacities are means of the concept, and they can increase in number. Networks of coidentified thoughts, i.e., concepts, can grow, characteristically becoming more adequate over time, such that coidentifying is performed more often, under more conditions, with less chance of error. This is but one more respect in which concepts are like other abilities, which we are continually improving, adding more means to, getting better at employing old means. And as we often have systematic ways of improving our abilities, we have systematic ways of developing our concepts (Millikan 1984, Chapters 15, 18, and 19). For example, having met a new person in a crowd, first you coidentify by tracking with your eyes, head and feet, coidentifying various
On Unclear and Indistinct Ideas / 85 demonstrative thoughts. As you track you memorize the face or stance or, perhaps for the moment, just the clothes. You thus extend your tracking abilities, so that you can keep track of the person over temporary and then longer breaks in your perception of them. You add to your tracking ability by collecting identifying information and by learning the person's name. This enables you to keep track via less direct observations and via the information-bearing medium that is language. Identifying information helps in the tracking, however, only in so far as you are able to coidentify the properties, relations or kinds mentioned in the identifying descriptions. Some portion of these abilities is surely innate, just as the basic ability to track with the eyes is not learned but matures. Recognition of color constancy over various lighting conditions and of shape constancy over various viewing angles are examples, recognition of phoneme constancy over speakers of different sizes and sexes is probably another, and so forth. But recognition of many constancies is learned in whole or part by employing prior tracking abilities (What do voices sound like under water? What do persimmons taste like?), and by learning names and discovering identifying descriptions.6 Discovery processes of this sort are never completed, surely cannot be in principle. You could never obtain such an adequate concept of anything that there would be no circumstances, no guises, under which you would fail to recognize it-in the flesh, by indirect observation, by name, or by description. Knowing the object of one's thought then is a matter of degree, with the highest degree never realized. VIII. Temporary Conceptual Means
Not only are concepts imperfect abilities, the core conceptual means for many of them are only temporary abilities. Sometimes this is because our memories are short. For example, although I can usually remember C sharp for five minutes, so as to coidentify it when I hear it again, I always forget it overnight-I don't have "absolute pitch". Luckily my concept of C sharp has other conceptual means. I know how to coidentify it by going to the piano, or the flute, or by asking my friend Brian, who has completely infallible pitch. Because I know C sharp's name, I also know how to iterate the thought of C sharp in new grounded judgments by hearing or reading about it. I may read what its frequency is, or about the difficulty of playing this or that instrument in its key. New faces can be, for me, a bit like C sharp. I can coidentify them over the next hour, but not always over the next week. If I have forgotten both the new face and the name, I may have to act very fast to pick up the scent. Descriptions are sometimes useful, but their usefulness may be short lived. I may need to call on the memory of others quite soon, before they forget who fits the description, before they forget, say, who all was present on the relevant occasion. Sometimes there may be no way to pick up the scent, and I will never know who that was that I met, who that is I am remembering. For a short time once I had a useable concept of that person, but I don't have one any more. Another sort of temporary conceptual means consists in tracking methods that are easily disrupted. Consider the concept you have of the disposable glass
86 / Ruth Garrett Millikan that is "your glass" at a cocktail party. You track it by keeping it in your hand, or by setting it down somewhere that you remember. But if you turn your back and someone should straighten up a bit, though the glass itself may not have wandered far, that may be the end of the tracking trail for that glass. Similarly, concepts of individual dishes in a matching set in one's cupboard are likely to have only temporary conceptual means. If these dishes have no individual salient distinguishing marks, and you have no cause to remember special happenings concerning any of them, every time the dishes are done and put away again, all your individual concepts of cups, glasses, and plates disappear, and new concepts of the same old individual dishes must be born again next meal. As an experiment, try to think, right now, serially, of each individual fork in your silver drawer. IX. Unusable Concepts
However, the loss of a practical ability to track an object does not necessarily mean total loss of its concept. For instance, I believe that Evans was subtly mistaken about even the man and the ball. According to Evans' original story, the man fails to remember the second ball he saw because of a blow on the head. Now imagine Evans' story as truly describing the realization of a perverse philosopher's thought experiment. Years later the philosopher shows up, pulls the remembered ball out of his pocket, and explains the whole episode to his victim, who then correctly coidentifies the ball of his memory with the ball he sees. So he was not debarred in principle from ever making another grounded judgment about that ball after all. True, Evans does stipulate that the man does not think of his remembered ball as the one that caused his memory. But this stipulation is totally implausible. No adult but only an animal or an infant could fail to recognize a description such as "the ball you are remembering", given the right context. Normal adult humans, then, always have concepts of what they remember, so long as the memories remain distinct. The trouble with memories of the individual forks in one's silver drawer is that they tend not to remain distinct but to merge. But now you may ask, is it really fair to say that the man has an ability to reidentify the ball, on the assumption that there is no perverse philosopher involved, no person to show up bearing again the scent? Let us first look at some other examples of this sort of problem. Earlier I mentioned Strawson's notion of "story-relative identification". Strawson explains this notion with an example. A speaker is telling a factual story which begins "A man and a boy were standing by a fountain... The man had a drink...". The hearer identifies the references of the two tokens of "man" as being to the same man, but does not identify this man with anyone outside the story. Strawson says of this kind of identification that it is "identification within [the] story; but not identification within history," hence that it is not "full identification" (1959 p. 18). Yet given that the story is factual, the hearer surely knows means of further tracking. He can, for example, ask the speaker who the man was. Thus identification of people in stories is not necessarily isolated from "history". But perhaps the speaker himself does not know who the story is really
On Unclear and Indistinct Ideas / 87 about, or the hearer does not ask and later forgets who told him the story. On the other hand, perhaps these links can be reestablished. The hearer knows a way to find out who told him the story, and a way to find out who told the story to the hearer, hence who the man was. But now suppose this method is chancey. He will be lucky if it works, but he can .try. Does he still "have an ability" to coidentify the man in the story? Clearly "having an ability" is not a very precise notion as ordinarily used. It is worth noting here that there are cases that Strawson would consider to involve "full identifications" that are more tenuously connected than many stories to any practical capacity actually to reiterate tokens of a thought in grounded judgments. Consider my thought, the person who wove this part, as I inspect a particularly intricate tiny section of a medieval wall hanging of unknown origin. This thought contains a perfectly clear definite description based, as Strawson prefers identifying descriptions to be, on an unambiguous demonstrative. But almost undoubtedly there is no way at all of my ever iterating the thought of this wonderful weaver in any new grounded judgments. Certainly having knowledge of the ontological structures in the world that relate that weaver to me, say, understanding the "criteria of identity" for pieces of tapestry and for persons over time, is insufficient when it comes to the practical business of actually tracking this weaver. Now except where forgetting was involved, in each case above where an ability to coidentify seemed to be questionable or to be fading away, the cause of the problem lay in the world rather than in the thinker. Here is an example up close. Suppose that you have a casual friend whom you sometimes see at parties and sometimes on the street. You have a good number of ways of tracking this friend. You recognize his looks, know his voice in a crowd, know his name, address, occupation and place of work. Surely you have an adequate concept of him. Then suppose through an accident he loses his voice, his face is disfigured, he leaves his job without notice, moves and takes on an alias. Despite your best efforts, you never coidentify him again. Now compare another sort of ability. Suppose that you know how to type. You learned on an old 1930s Underwood that your brother found under the sink at the cabin. But this Underwood is now unrepairable and you can't find another, indeed, there are no pre-electronic machines about anywhere, they have all been turned in as scrap metal, and you don't understand how the modern machines work. Besides, their keyboards are completely differently arranged, as a result of modem efficiency studies. Do you still know how to type? One reasonable thing to say, it seems to me, is that you do still know how to type, it is just that the particular means you need to employ to that end make your skill useless in practice. Similarly, I may know exactly how to invest my money to make a million, except that I haven't got any money. Or I may know how to make a marvelous gourmet dish for which the ingredients are completely unavailable. If there was a time when people,knew how to make tasty dodo stew, they didn't suddenly stop knowing how on the expiration date of the last dodo. One has an ability if one has a mechanism inside one that has been correctly tuned for doing a certain kind of task and that has not been damaged. How often the external conditions presupposed for proper exercise of the ability actually
88 / Ruth Garrett Millikan hold, indeed, whether they ever hold at all, does not affect the matter. At any rate, this is how we count conceptual abilities. We do consider ourselves capable of thinking of things the thoughts of which we have no chance, practically speaking, of ever coidentifying. We count ourselves as sometimes having completely unusable concepts. Let us return then to the man harboring the memory trace of the ball. What makes it true that he has a concept of the ball is that he knows he remembers it, hence knows an identifying description of it, and he has the general capacity to make use of identifying descriptions for purposes of coidentifying. Most relevantly, perhaps, he knows how to make use of definite descriptions by leaning on other people's knowledge. But whether he in fact manages to locate a person with knowledge relevant to this particular case, indeed whether such a person exists at all, reflects in no way on his conceptual abilities. X . Another Regress?
Evans made a stronger claim than just that concepts conform to his generality constraint. He claimed that where there is no concept there is no thinking of at all. Now I argued earlier that for thinking of to require "knowing what one is thinking of' would produce a regress if this knowing is interpreted as knowing that. But on the interpretations that Evans and I have given of concepts, if thinking of requires concepts, and having concepts requires identifying what one is thinking of, aren't we again in a regress? If I am to "know what it is for a proposition of the form ra=al to be true" (Evans 1982 p. 110) then there must be something that it is for it to be true, and this would seem to require that both 6 and a already have semantic values. This should be so, moreover, for concepts of properties, relations or kinds as well as concepts ("Ideas") of individuals. One can succeed or fail in recognizing the same color or shape again, or the same musical pitch again, or the same mineral or plant again, as well as the same individual again. Further, each of these latter items may be grasped indexically through perception, or by name, or by description. For properties, relations and kinds, as well as for individuals, I must be able to determine when what comes into my head in the way of representing has the same semantic value again and when it does not. But such an ability seems to imply that these semantic values exist prior to identification of them, hence prior to conception. Elsewhere I have argued that aboutness or intentionality is a far simpler phenomenon than conception. It is involved, for example, in perception even in the case of the most primitive animals. The capacity to coidentify elements of inner representations is intimately related to the capacity to make mediate inferences, which is surely confined to relatively few species, but there are lots of intentional items in organisms that don't participate in inferences (Millikan 1984, 1989, 1993a). However, in the case of human conception or thought, as opposed to certain aspects, at least, of human perception, I think that intentionality does depend on conception. This is because the intentional content of any inner sign, what it is about, always rests heavily on how it is used by the organism. Thus the contents of thoughts, whose primary functions are to
On Unclear and Indistinct Ideas / 89 participate in inferences, must be relative to this primary function. To be capable of conception is to have the ability to be guided by perceptual input in tracking activities that succeed in mapping certain features of one's distal environment consistently into patterns of thought. And what "consistently" has to mean here is that the same distal item is mapped, in different complete thoughts, with thought-terms or elements that one knows to coidentify. That is the same thing as to say that the mapping achieved is one the inferencing systems know how to use. What makes a mental token Alice be a thought of Alice and nothing else is, in part, that one has the ability (under appropriate conditions of the sort for which one's conceptual capacities were designed or tuned) to iterate other tokens to be coidentified with this token, each such token being conjoined with a different coidentifiable predicate thought, in such a fashion that the whole pattern abstractly mirrors actual configurations in one's world involving Alice. Correctly coidentifying in thought has the logical form, then, of contributing to the solution of a coordination problem or, say, of directing one light ray toward a focus with others. Having a concept is like focusing the eyes. When both eyes are open but not focused on the same thing, neither eye sees anything, but when both are focused on the same thing, then both see, both see the same, and both are correct in their focus. It is not that first one has a thought of something and then later one may or may not figure out how to coidentify it, how to discover what it is one is thinking about. Rather, one is not thinking of something at all unless one already has some means of coidentifying it (though as we have seen, that means may happen to be unusable given the actual arrangements in one's world).7 XI. Identity Statements: the Strawson-Lockwood View
Given this sort of description of what a concept is, it seems clear that different people's concepts of the same object will not usually be the same. Each person will have a unique network of coidentifying abilities pertaining to each conceptualized item. Moreover, there seems to be nothing to bar a person from having several concepts of the same thing but that are not joined, the ideas within these nets not being coidentified. This accords with the view of Strawson (1974) and Lockwood (1971) that the role of identity sentences involving proper names is to effect concept joinings, to hook together isolated concept nets that are concepts of the same.8 Lockwood puts it this way. "[Tlhe hearer possess[esl a body of information, or mental 'file', relating to [an] individual in question, to which [a] name, as it were, gives access... . In the making of an (intendedly informative) identity statement, the speaker...assume[s]...that the hearer has more than one file on the subject of his assertion... . The purpose of an identity statement, which will be fulfilled if it is accepted as true, is precisely to get the hearer to merge these files or bodies of information into one" (pp. 208-209). Lockwood's analogy of concepts with,file folders is good, but can perhaps be improved. For whereas one thinks of a file folder as having some one central name or designation under which it is filed, concept nets involve multiple thoughts of different types, none of which is more central, more the label of the folder, than any other. A name gives access to a whole concept net, not to any
90 / Ruth Garrett Millikan particular thought type in the net. We should note also that it is not just thoughts of individuals that have "folders", but other denotative thoughts as well, so that the information in each of the folders is in many other folders as well. The cross referencing system required is not easy to imagine on this analogy. But Strawson's and Lockwood's basic idea is sound. The function of identity statements is to cause the joining of previously isolated concept nets. There is a reason why the Strawson-Lockwood thesis has not been more widely accepted, however. Frege thought that there was a single sense or thought type9 attached to each unambiguous denotative public language term, the same for every person who understands the term. But the Strawson-Lockwood thesis suggests that there need be nothing except reference in common to the concept nets accessed by different speakers via a public language term. So it is hard to see how any notion of public language sense, at least for uncompounded words, could be reconstructed after this move. (Of course there are always exceptions. There are occasional words that almost everyone in a language community learns to associate with certain facts about their referents, such as that Hesperus is seen low on the horizon in the evening, Phosphorus in the morning, or that Mark Twain was an author-"Mark Twain" was, after all, his pen name. Also, occasionally uncompounded words are passed down from generation to generation by explicit conventional definition. For example, nearly every English speaking child is taught that a circle has all points equidistant from its center.) There is also an even more disturbing threat to Fregean thought lurking in the background which we would be delinquent not to mention. This concerns the notion of sense as that which is grasped by the mind and which determines the referent of thought. It would be natural to say, following the StrawsonLockwood move, that "Cicero" and "Tully" do not have public senses, but only senses relative to individual speakers, and to call these "private" senses. The private senses of these names, for an individual speaker, would then be said to differ only if the statement "Cicero is Tully" would actually inform that speaker, that is, only if the coidentity nets to which "Cicero" and "Tully" gave access were different for this speaker. But if senses, as what is grasped by the individual thinker's mind, are equated with identity nets, then it is likely that no two people ever think of a referent by means of exactly the same sense. Is there then some other way to reconstruct the notion of Fregean sense? For example, might various thought tokens in the same coidentity net have different senses? Might these senses determine referents independently of the nets in which individual graspings of them are enmeshed? The remainder of this essay moves toward a negative answer to these questions.10 Part Two: Confused Ideas XII.' The Passive Picture Theory of Coidentifying
It is time now to complete the description of concepts being offered by characterizing the act of coidentifying. But this completion will prove unsettling. Actual acts of coidentifying will emerge as partial determiners of
On Unclear and Indistinct Ideas / 9 1 reference. The result is that mistaken acts of coidentifying tend to corrupt thought, producing unclarity or indistinctness. Brace then for deconstruction. Some paragraphs back I declined to disabuse the reader who might suppose that merely harboring two identical thoughts within a single act of consciousness would constitute recognition of the sameness of their objects. Indeed, I have been modeling the act of coidentifying as iterating the same thought type in the context of new grounded judgments, as though mere repetition of a thought constituted a recognition of the sameness of its object. It is time to make amends. That the act of identifying is not accomplished by the mere occurrence in the mind of two like thought tokens is clear in the case of tokens widely separated in time. Thinking of a thing one day that it is red and thinking of it the next that it is sweet, no matter how similar the manners of thinking of it are, does not constitute grasp that the same thing is both red and sweet. One might suppose, however, that placing these thought tokens of the same beside one another within the same act of consciousness or judgment would constitute such an act. But unfortunately what might lead one to suppose this would be a confusion-a confusion between representing or thinking of sames and representing or thinking of sameness. This confusion results, in turn, from a subtle application of the "passive picture theory" of mental representation. Nowadays the passive picture theory is not held explicitly, indeed, when actually stated it strikes us as pretty crude. None the less it not uncommonly figures as an important implicit assumption, especially in theories of perception. The passive picture theory supposes that perception, thought, or mental representation is accomplished by means of items appearing before the mind that actually have the properties they represent to the mind. The properties claimed by perceivings or thinkings to characterize the world exist in "objective reality" (Descartes), or they, or doubles of them, are true of sense data, or percepts, or phenomenal objects, or visual fields, etc. When and only when the world resembles the inner picture, then the perceiving or thinking is veridical, showing how things really are. The passive picture theory has an influence today mainly with regard to formal or abstract kinds of resemblance such as resemblance in temporal properties, continuities,determinateness, number, etc.11 Some have suggested that the main problem with the passive picture theory of perception and thought is that it invokes a regress: how will the inner eye or mind then perceive the inner picture? In the same way that the outer eye does? 12 I think there is a more fundamental error involved in the passive picture theory. For it is surely possible to give a coherent answer to the question how the inner eye or inner mind works that does not invoke a regress. The inner eye or mind can be taken to understand what is before it merely by reacting appropriately, by being guided appropriately for purposes of thought and action. Thus the regress can be avoided. A bigger trouble with the passive picture theory, I suggest, is that it produces a facade of understanding that overlooks the need to give an account of the way the inner understander works, an account of the mechanics of inner representation or of thought, of what kind of reacting is comprehending. Having projected the properties perceived or thought of to the inside of the mind, the assumption is that there can be no problem about how they manage to move
92 / Ruth Garrett Millikan the mind so as to constitute its grasp of them as what is represented. Their mere reclining in the mind constitutes the mind's envisioning of them. That is why I call this theory the "passive picture theory" of inner representation. What's wrong with it in the first instance is the passive part. And once you see that it must be the mind's reaction that constitutes understanding of an inner representation, you see that the picture part is also suspect. Why would a picture be needed to move the mind appropriately? At the very least, wouldn't something more abstractly isomorphic do as well? 13 And even if a picture was needed to move the mind appropriately, it certainly doesn't follow that only a picture is needed, that appropriate movement is guaranteed by the mere presence of a picture. XIII. Repeating vs. Coidentifying
The passive picture theory results in confusing properties of what is represented with properties of the vehicle of representation and, of course, vice versa. An abstract classical example, suggested by Kant in the Parallogisms, was Hume's confusing a succession of mental representations (of "impressions") with a representation (an "impression") of succession. Even more abstract, but just as much a confusion, is to take the occurrence of sameness or repetition in the vehicle of representation to be a representation of sameness. Or, if you are uneasy about thoughts being representations or having vehicles, then the confusion is, just, taking recurrence of a thought or sameness of thought for an understanding of object sameness. That this would be a mistake does not mean, necessarily, that the occurrence of identical thoughts within the same mental breath-Alice is slim next to Alice is trim-might not, under certain circumstance, make available, or even psychologically necessitate or precipitate, an act of coidentifying. That possibility we can leave open. The point is only that such a co-occurrence could not, in and of itself, constitute a recognition of sameness. Recognizing a sameness must be something over and above mere co-occurrence of two like thoughts of the same. Conversely, notice that it is perfectly clear that a grasp of sameness does not require the simultaneous presence of identical thoughts representing these sames. If it did, then one couldn't grasp that the same was both red and round if the redness had been perceived a week ago. And one couldn't grasp that the city mayor was the same as Alice or the same as the woman one sees in the market. Suppose that identical thought tokens always possess identical contents.14 Then sameness in thought type will be an indication of sameness in content, perhaps it will contain the fact of this sameness as natural information, and so forth. But compare: two bee dances danced side by side simultaneously by two bees may jointly be an indication, or between them contain the natural information, that there are two sites of nvtar which are forty yards apart. It does not follow that the watching bees can read this information off the pair of dances. Not everything that falls out of a representational system is necessarily read or readable even by its primary interpreters. And it is even more clear that not every bit of natural information harbored in an organism's system is
On Unclear and Indistinct Ideas / 93 understood by the system. If we have rejected the passive picture theory of thought we should also be able to see that the mere being the same of two thoughts cannot accomplish anything all by itself even when the fact of this sameness is a natural indication of a sameness in content, or when this sameness falls out by implication of the inner representational system. The fact of sameness must be read somehow if it is to represent, rather than just be, a sameness. The sameness must appropriately interact with or move the thinking system in some way if it is to represent itself. The question arises then how it must move the thinking system in order to represent itself. In the same way, presumably, as the system is moved when nonidentical thoughts are understood as representing the same. In the same way, for example, as when this person is understood to be Alice, or Alice understood to be the city mayor. When we consider these cases, it is clear that identity of thought type is entirely incidental to the thinker's act of recognizing content sameness. The act of coidentifying is not dependent in any way on sameness of thought type. Nor should we fall into another nearby error. The way that the system must move or be moved in order to be grasping a sameness is not just in-the-sameway-again. Consider the story of Zak. Zak is a patient at the Bell Neurological Institute, a victim of stroke, suffering selective amnesia. Each morning Dr. Helm comes in to see Zak, wearing a white coat and a name tag that says "Dr. Helm, MD". Each morning Zak greets him with "good morning Dr. Helm," and when asked if he knows who Helm is, being no fool, Zak unhesitatingly answers "my doctor". In brief, the appearance is that Zak always identifies Helm the same way and correctly. Nor, we suppose, does Zak have problems articulating a theory of the identity of persons over time; he used to be a philosophy professor. Upon further questioning, however, each morning Zak reveals that he does not remember ever having seen Helm before, nor does he show any signs of familiarity with the routine Helm puts him through each morning. That is, it appears that Zak does not recognize Helm after all. Though he appears to have an individuating idea of him, even what Evans would call a "fundamental Idea" of him, he is incapable of coidentifying him, he has no concept of this person that lasts over time. Compare this story with a much simpler case: does the frog that reacts the same way each time its retinal bug-detector fires thereby cognize a sameness among the bugs it eats? Something like the opposite is true, I suggest. A creature's perception that it is encountering the same thing again shows up, characteristically, in its reacting differently this time, differently in accord with what was learned last time. That the baby recognizes you is exhibited not in its crying again-that is how it reacts to strangers-but in its smiling, or exhibiting other behaviors apparently based on its earlier experience with you. And of course the notion that reidentifying a thing involves "applying the same concept again9'-that is, attaching the same thought or mental name to it-is precisely the version of the passive picture'theory we have been discussing.ls
94 / Ruth Garrett Millikan XIV. The Act of Coidentifying
In what kind of way does one's thinking have to move, then, in order to grasp an identity? We can approach the matter this way. What is the function, the point, of grasping an identity? What does one do with a knowledge of identity, say, with an identity judgment? What changes take place when I recognize for the first time that Cicero is Tully? Well, if I knew before that Cicero was bald I now also know that Tully was bald. And how does that change anything, that I now know that Tully was bald? After all, I already knew that Cicero was bald and that was exactly the same thing to know. Why not be satisfied with knowing some things about Cicero under one idea, other things under others, even if I don't know that these ideas grasp the same? So long as I pack all the right information in one way or another, why does it matter what notation I use? It matters, of course, because if I don't recognize the identity of Cicero with Tully, then I cannot combine the various things that I know about the object Cicero so as to yield anything new. Every mediate inference pivots on a middle term of some kind, on the ability to recognize that two thought tokens share a content. Similarly, only through recognizing the identity of an item currently perceived with an item perceived earlier can what was learned earlier be joined with what is perceived now to yield informed action. Recognizing incompatible cognitive attitudes hence avoiding contradiction also turns on a grasp of identity. In short, grasp of identity is the pivot on which every exercise of thought that collects together information, effects its interaction, or applies it, must turn. Indeed, to coidentify two ideas (tokens) consists in no more and no less, I suggest, than that the thoughts-that (tokens) in which these are embedded interact or combine, or that there is preparation for them to combine, in this manner.16 It follows that acts of coidentifying can occur not only over tokens of different thought types, but over tokens widely separated in time. For ideas had yesterday or last year to be coidentified with ideas had today requires only that the former, taken with context, causally affect the uses of the latter, taken with context, in ways that produce new information or informed action. At this point we can safely admit that there are some occasions when reacting the same way again is a manifestation of having coidentified a thing. When the reaction recurs because it earlier proved a useful one, then the transference of the action disposition from the context of the relevant earlier percept or thought (that (kind of) animal, say, or that (individual) man) to the content of a later one coidentifies (parts of) the two percepts or thoughts. Thus the phenomena "stimulus generalization" and "stimulus discriminization," should these sometimes occur as classically understood, would correspond to principles by which concepts are formed. Contrast this case of reacting the same with the case of a duckling that has imprinted upon its mother. The result of imprinting is that whenever the duckling sees its mother, a certain set of behavioral dispositions emerges. The Duckling has stored away a "template" matching its mother's appearance so as to "recognize" her. Despite our natural use of "recognize" in this context, it does not follow (though it may be true) that the duckling coidentifies its mother or that it has a concept of her. A repeated set
On Unclear and Indistinct Ideas / 95 of reflexes, even if triggered by an acquired "template," is not a concept. Only in so far as the duckling is capable of learning things about its mother, which it applies on future occasions, does the duckling have anything like a concept of its mother. (But, you may say, surely the duckling uses its template to help it track its mother; it tracks her by following her around. Yes, but is its following her around the right kind of tracking, namely, tracking for thc purpose of picking up a variety of information about her, about one and the same thing, to apply later?)
XV. Identifying Acts vs. Identity Judgments So we can admit that sometimes similarity of reaction is an indication that acts of coidentifying are occumng. We can also admit that samenesses among thought tokens might dispose the thinker to coidentify these tokens. But there is no need to admit any kind of analytical connection between type sameness in thought tokens and the disposition to coidentify. (By "type sameness" we can mean, depending on how the ontology of thoughts is conceived, sameness in what presents itself to consciousness, in what is "grasped by the mind" in thinking or, on a mind-brain identity theory, in the physical make up of the thought vehicle-in the "shape" of the thoughts.) It is possible that sameness among thought tokens does dispose a thinker to acts of coidentifying, but might there be other kinds of relations among tokens that have this effect too? The model of rational thought as an unfolding logical system lies deep in the philosophical tradition, reaching back at least as far as Aristotle. According to this model, the act of coidentifying is legitimated only by sameness of type. For example, when one adds P to P 3 Q to obtain Q, it is (generally assumed to be17) the shape similarity of the two tokens of P that allows them to be coidentified. Compatibly, the inference from "Tully wrote books" and "If Cicero wrote books he could probably read" to "Cicero could probably read" is treated as requiring both (1) the additional premise "Cicero was Tully" and also (2) the translation, "If Cicero wrote books then Cicero could probably read". In traditional logic, one must translate and add identity premises until all mediate inferences turn only on symbol type samenesses. But consider: how does one get from "If Cicero wrote books he could probably read" to "If Cicero wrote books Cicero could probably read"? That movement, after all, is a movement that requires coidentifying. It accomplishes grounded iteration of "Cicero" in a new sentence context. Nor is there a way to unpack this move by supplying an identity premise. The sentence "He is Cicero" can't be a mediator because he isn't, in general, Cicero. Similarity of shape cannot be made to be the hinge upon which this kind of mediate inference turns. Anaphoric reference devices occur in all natural languages. Why should they not occur also in thought? Acts that coidentified anaphorical references in thought would neither coidentify in accordance with similarity of thought type, nor would they rest on identity premises that could be made explicit in thought. But notice that they might nonetheless be mistaken. Suppose, for example, that I moved in thought from Cicero wrote books and If Cicero wrote books then he could read to Cicero could read, but it then turned out that the "he" referred to in the consequent of the
96 / Ruth Garrett Millikan conditional was not Cicero but some other man. Then my inference would be invalid. But it would not necessarily have had a false premise, not even an implicit one. Mistaken identifications are not then the same as false beliefs. Consider also coidentifications involving indexical tokens. Tokens of natural language indexicals are coidentified when they have suitable relations to the same external objects, not when they are alike. "This" cannot, in general, be coidentified with "this". Similarly, when we accumulate information about an object by means of perceptual tracking, the coidentifying of consecutive thought indexicals is mediated by relations that loop through the external world. It is through contiguity of effects of the object being tracked upon the perceptualmotor mechanisms that the coidentifying is accomplished. These coidentifications also rest neither on sameness of thought type, nor on identity premises that could be made explicit in thought. But, since one can unknowingly lose track when perceptually tracking, mistakes in coidentifying resulting in mistaken inferences can result.18 There is then a kind of error in thought that is not false belief or judgment but error of a distinct kind. Erroneous coidentification is failure on the level of know-how. It is not failure to know that but failure in an activity. XVI. Confused Thoughts
One result of mistaken coidentification may be invalid inference. For example, suppose that I see that something in the tree is a squirrel and then I see that it takes off and flies. I may conclude that some squirrels can fly. But the inference will be invalid if I had unknowingly lost track of the squirrel and some other "it" did the flying. Since invalid inferences can lead from true premises to false conclusions, erroneous acts of coidentifying can of course give rise to false beliefs. More interesting however, is that erroneous acts of coidentifying can give rise to equivocal thoughts. That is the theme alluded to in the title of this paper, and we are at last in a position to explore it. Let us first be clear exactly what the claim is (lest the title of this section be too apt). To lack the ability always to coidentify correctly, to be disposed sometimes to error, is part of the human condition. And to be in this condition is not necessarily to harbor any confusions; it is not (yet) to have made any kind of mistake. Similarly, it is not a mistake to fail to know how to ride a bicycle, or to fail to know how very well. But to lack an infallible ability is to be subject to failure, and should a failure occur, that of course is an error. It is when actual errors occur in coidentifying that conclusions may turn out false and, more interesting, that thoughts may become equivocal or "confused." When this occurs, it is the thoughts, the very thought tokens, the ideas themselves, that become equivocal. Equivocation in thought is a kind of corruption or error that is quite other than falsity in a proposition believed. It attaches to terms rather than propositions, and it is not reconstructible & a propositional error. Further, since it is not constituted by but merely results from the fallibility of one's ability to coidentify, it is not really a conceptual error either. A corrupted thought is not a
On Unclear and Indistinct Ideas / 97 faulty concept anymore than it is a false belief. Calling on Bishop Butler, "it is its own thing." What then is a confused thought? A person who has very basic misinformation about a thing may be said to be confused about it, but that sort of confusion is not what I have in mind. "Do you see that woman?" Jane says, pointing to Ann, and then she tells you a whole kettle of lies about Ann, all of which you believe. Perhaps you never learn anything else much about Ann, but once in a while you do see her again on the street, and then you review in your mind all those dreadful lies. You are woefully confused about Ann, but not confused in the way that I mean. For it is definitely about Ann that you are confused, and not about something ambiguous. Perhaps you even believe a lot of wrong individuating descriptions of Ann, but that is all right too. So long as you don't actually use any of these descriptions in such a way as to result in actual misidentifications of Ann you will not yet have a confused thought. Suppose, on the other hand, that you mistake Carol whom you see on the street for the city mayor. Having heard that the Mayor was in Washington just this morning, you conclude that she must have taken a plane home. That WHO must have taken a plane home? That is what I mean by a confused thought. It is an equivocal thought. It isn't a thought of the one woman any more or less than it is of the other, but hovers between the two. Suppose that I have Bill and Biff confused, that I take each to be the other. Returning to Lockwood's image, I have only one file folder into which I put all the information gleaned from perceptual encounters with both. This is not a false belief about identity that I hold. That is, I hold no belief, explicit or implicit, of the form "X is Y". To do that I should have to have a thought of Bill and another of Biff, which thoughts I was disposed to coidentify. But a thought of Bill that is other than my thought of Biff is exactly what I don't have. Nor is what I believe composed of a whole collection of identity beliefs of the form this man is that man, each of these equating a pair of indexical thoughts, memory indexicals or current indexicals, each member of each pair referring to just one or just the other of these men. If there ever was a philosopher's myth it is this, that I should have kept separate memory indexicals stored away for every one of my past encounters-with everybody? Surely many or most of my memory traces are not of specific encounters with Bill or with Biff at all, but amalgams of traces of both men. Indeed what, if not just such amalgams of indexicals, are your non-indexical thoughts of most of your friends, your thoughts of them that you would naturally express just by using their names? From whence do the reference relations for non-indexical thoughts typically come? Typically you begin with "that woman," then "that woman" again, as you track, collecting initial information about her. Then you put a thought-name to that woman, thinking of her as (public-nameless) a, or if you have learned her public name, as Alice. Is there any more to it than that? Suppose you then learn an identifying description that applies to Alice. Do the indexical origins of the thought Alice suddenly become irrelevant? Does the description take over, leaving no trace of indexicality behind? And what if the description should turn out to be wrong, or not to identify uniquely?
98 / Ruth Garrett Millikan The point may be made clearer using the following examples. John Campbell, in "Is Sense Transparent" (1987-88). mentions three kinds of cases in which perceptual indexicals can in principle be wrongly coidentified, yet he claims that when these coidentifications are made correctly the coidentified indexicals count as having the same Fregean senses. I will not open the can of Fregean senses here, but Campbell's examples are perfect for illustrating the phenomenon of equivocation in thought. The first example is simple eye tracking, where you take it that what you see is the same object from moment to moment. The second example is simultaneous visual and tactual perception of what you take to be the same object. The third is simultaneous perception with your two eyes of what you take to be the same object. In each of these cases it is possible (under easily constructable laboratory conditions, in fact) that the object is not actually the same. Then one will have taken two representations that are not of the same, for example, the representations presented to the brain by the two eyes, and used them jointly to derive a new representation, in a way that wrongly presupposes they do represent the same. That is, one will then have miscoidentified. And in each case it seems clear that what remains after the merger is a simple or uncompounded indexical thought that points to two objects at once. One ends with a confused or ambiguous thought.19 To have two things or more confused in one's mind is surely a common condition. For example, much of the history of science might be told in these terms. What is astonishing is not that it happens, but that in dealing with common objects, properties and kinds it doesn't happen more often. What is astonishing is how good we usually are at keeping track of those ordinary things in our world that (unlike which glass is which in the cupboard) matter to us -at not mixing them all up together. What is astonishing is how good our concepts tend to be, despite the fact that they must operate, as must our other abilities, on principles resting not just on the character of our minds, but on the structure of the world outside. Notes
*
Much of the first half of this paper is a lightly revised version of "Knowing What I'm Thinking Of", Aristotelian Society Proceedings Supplementary Volume XLVII (1993). I am grateful to the Aristotelian Society for permission to use this material. Parts of sections XII, XI11 and XIV are drawn from "Perceptual Content and Fregean Myth", Mind 100.4, for which permission has kindly been granted by Mind. The paper was completed while the author was a Fellow at the Center for Advanced Study in the Behavioral Sciences. I am grateful for financial support provided by the National Endowment for the Humanities #RA-20037-88 and the Andrew W. Mellon Foundation. 1. In the austere terminology of (Millikan 1984), an ability corresponds to the proper function of a biological device when the device has normal structure, that is, when the device is intact. Most human abilities, being learned rather than hardwired, correspond to derived proper functions of our behavior-producing systems. There are for each kind of ability, normal conditions for its successful exercise, but often these conditions are not average conditions. 2. Accordingly, in (Millikan 1984) I called this act "the act of identifying". See, especially, Chapter 15.
On Unclear and Indistinct Ideas / 99 3. This definition is a very temporary expedient. There are problems with the notion that there are such things as idea types. See (Millikan 1993a). 4. I describe the act of identifying in sections XII-XIV below. 5. In (1984) I rather confusingly called means of conception "intensions". 6. A discussion of concept formation, on the assumption that it works in this sort of way, is in Chapters 9, 18 and 19 of (Millikan 1984). 7. Clarification of this position is in (Millikan 1993a). 8. Lockwood says he got the general idea from Strawson's lectures at Oxford. See also (Millikan 1984, chapter 12). 9. I am not using "thought" in Frege's technical way but more loosely. 10. A more detailed discussion occurs in (Millikan 1993a). 11. Examples from both the theory of perception and the theory of cognition are discussed in (Millikan 1991~). 12. See for example (Evans 1985a p. 397). 13. On numerous occasions I have taken the position that thinking and perception likely both involve inner representation and that representation involves abstract mappings by which representations are projected onto representeds. But this claim does not entail that any particular concrete properties and relations are shared by representation and represented. Nor is it likely to be open to merely philosophical demonstration which abstract mathematical relations are shared. 14. This is not a trivial supposition: see my (1991, 1993a). Indeed, in (1993a) I argue that it is unlikely that an individual's thought tokens are grouped into types by similarity of any kind. 15. For a qualification of this last remark, see Millikan 1993a. 16. See my (1984), Chapter 15 for a fuller discussion of acts of coidentifying. 17. But see Millikan (1993a) section IX). 1 8 . This point is kindred to Evans' thesis that the thought indexicals when one is successfully tracking an object all have the same Fregean sense (Evans 1981), and to John Campbell's thesis (1987-88) that no implicit judgments of identity are involved during such tracking. But the final intent of my arguments is to undercut rather than to reconstruct Fregean sense. 19. According to Gareth Evans, in the case of losing track (the other cases are discussed only by Campbell), what one is left with is no thought at all (e.g., 1982 p. 173). This is because Evans supposes that unknowingly losing track of an object shows that one didn't have an "adequate Idea" of it, that is, that one didn't have an ability to keep track of it. But I have argued that all abilities are fallible, that failure is not necessarily an indication of lack of ability. Surely another reason for Evans' conclusion was, simply, that the philosophical tradition has never seriously considered that there might be such a thing as a posteriori known equivocation in thought.
References Campbell, J. 1987188: "Is Sense Transparent?". Aristotelian Society Proceedings, New Series Vol.LXXXVII1, pp. 273-92. Evans, G. 1982: The Varieties of Reference. 0xford:Clarendon Press. Evans, G. 1985a: "Molyneux's Question", in (Evans 1985b), pp. 364-399. Evans, G. 1985b: Collected Papers. Oxford: Clarendon Press. Evans, G. 1981: "Understanding Demonstratives", in H. Parrett and J. Bouveresse eds., Meaning and Understanding (Berlin: deGruyter). Reprinted in Evans (1985b). Lockwood, M. 1971: "Identity and Reference", in M. K. Munitz ed., Identity and Individuation, New York University Press, pp. 199-211. Millikan, R. G. 1984: Language, Thought, and Other Biological Categories. Cambridge:Bradford BooksIMIT Press. Millikan, R. G. 1986a: "The Price of Correspondence Truth". Nous 20 pp. 453-68.
100 /Ruth Garrett Millikan Millikan, R. G. 1986b: "Thoughts Without Laws". The Philosophical Review 95, pp. 47-80. Reprinted in Millikan (1993b). Millikan, R. G. 1989: "Biosemantics". The Journul of Philosophy 86.6, pp. 281-97. Reprinted in Millikan (1993b). Millikan, R. G. 1990: "Truth Rules, Hoverflies, an the Kripke-Wittgenstein Paradox". The Philosophical Review 99.3, pp. 323-53. Reprinted in Millikan (1993b). Millikan, R. G. 1991a: "Compare and Contrast Dretske, Fodor and Millikan on Teleosemantics". Philosophical Topics 18.2, pp. 151-61. Reprinted in Millikan (1993b). Millikan, R. G. 1991b: "Speaking up for Darwin", in G. Rey and B. Loewer eds., Meaning in Mind; Fodor and his Critics, London: Blackwell, pp. 151-64. Millikan, R. G. 1991c: "Perceptual Content and Fregean Myth". Mind 100.4, pp. 439-59. -. - . Millikan, R. G. 1992: "Explanation in Biopsychology", in J. Heil and A1 Mele eds, Mental Causation, Oxford: Clarendon Press, pp. Reprinted in Millikan (1993b). Millikan, R. G. 1993a: "On Mental Orthography", in Bo Dahlbom, ed. Dennett and his Critics. Oxford: Blackwell. Millikan, R.G. 1993b: White Queen Psychology and Other Essays for Alice. Cambridge: Bradford BooksNIT Press. Russell, B. 1912 The Problems of Philosophy. New York: H. Holt Inc. Sellars, W. 1956: "Empiricism and the Philosophy of Mind", in H. Feigl and M. Scriven eds., Minnesota Studies in the Philosophy of Science vol. I , Minneapolis: University of Minnesota Press, pp. 253-329. Strawson, P. F. 1959: Individuals. London: Methuen & Co. Strawson, P. F. 1974: Subject and Predicate in Logic and Grammar. London: Methuen & Co.