August 15, 2000



No Experience Necessary:

Empiricism, Non-inferential Knowledge, and Secondary Qualities



Empiricism can be a view in epistemology: without perceptual experience, we can have no knowledge of contingent matters of fact.  Empiricism can be a view in semantics: propositional or more generally conceptual content is unintelligible apart from its relation to perceptual experience.  Empiricism can be a view in the philosophy of mind: "experience must constitute a tribunal, mediating the way our thinking is answerable to how things are, as it must be if we are to make sense of it as thinking at all."[1]  The most important development in empiricism in recent decades[2] is McDowell's Mind and World.  The view he puts forward there is empiricist in all these senses.  In this essay I want to highlight certain features of his concept of experience, first by showing how he avoids some pitfalls that notoriously ensnare traditional attempts to work out empiricist intuitions, and second by comparing and contrasting it with two other ways of construing perceptual experience—one less committive than McDowell's and the other more so—that also avoid the classical difficulties.  The stripped-down view is the one I endorse.  It is epitomized by my title: No Experience Necessary.  Indeed, though the word "experience" is mentioned in the 750 pages of Making It Explicit, it is never used.  I want to say, with Laplace [check this!], "Je n'ai pas besoin de cette hypothèse la."  The explanatory work done by what Sellars has taught us is the theoretical concept experience can be done without postulating a layer of potentially evidentially significant (hence conceptually articulated) states in between purely causally occasioned and physiologically specifiable responses to environing stimuli and full-blown perceptual judgments.  Even were he to grant this claim, the realist McDowell would, I think, insist that nevertheless, perceptual experiences that are not yet judgments are there, and therefore deserve theoretical recognition.  I think that such realism about perceptual experiences commits its adherents to there being answers to the sorts of questions I will raise under the heading of the more committive views, which McDowell also resists. 




McDowell's empiricism is distinguishable from classical versions in at least two fundamental ways.  First, with Kant and Sellars, McDowell understands experience as a thoroughly conceptual achievement.  Thus he insists that anything that does not have concepts does not have perceptual experience either.  Because he does, McDowell counts also as endorsing the fundamental rationalist insight: that to be aware of something in the sense in which such awareness can serve as evidence for beliefs amounting to knowledge is to bring it under a concept.  This principle dictates that one must already have concepts in order to have experience in the sense he is addressing—a sense that in view of its fealty to the rationalist principle deserves to be seen as a successor of Leibniz’ notion of apperception.


McDowell also insists that anything that does not have perceptual experience does not have concepts either.  That is, he endorses the view I called semantic empiricism above.  Concept use and perceptual experience are two aspects of one achievement.  This view was emphatically not a feature of traditional rationalism.  In his synthesis of these themes of classical rationalism and classical empiricism, as in so many other respects, McDowell is a kantian. 


Second, for McDowell perceptual experience is generally (though not in every case) immediately and essentially revelatory of empirical facts.  That is, it is essential to McDowell's concept of perceptual experience that the fact that things are thus and so can be the content of a perceptual experience.  When things go well, the fact itself is visible to us.  It is the content we experience.  The perceiving mind includes what it perceives. 


Because he understands perceptual experience as requiring the grasp of concepts, McDowell avoids the Myth of the Given, which afflicts all classical versions of epistemological empiricism.   The Myth of the Given is the claim that there is some kind of experience the having of which does not presuppose grasp of concepts, such that merely having the experience counts as knowing something, or can serve as evidence for beliefs, judgments, claims, and so on, that such a nonconceptual experience can rationally ground, and not just causally occasion, belief.  In "Empiricism and the Philosophy of Mind"[3], Sellars shows to McDowell's satisfaction (and to mine) that the project of making intelligible a concept of experience that is in this way amphibious between the nonconceptual world and our conceptually structured thought is a hopeless one.  By contrast, McDowell is clear in taking perceptual experiences to have the same sort of content that perceptual judgments have—and hence to be conceptually structured. 


Since McDowell also takes concept use to be a linguistic achievement (in line with Sellars' doctrine that to grasp a concept is to master the use of a word), he takes it that we learn to have perceptual experiences only when we come to have a language.  Thus perceptual experience is not something we share with nonlinguistic animals such as cats and chimpanzees.  No doubt there is some sort of broadly perceptual attunement to things that we do share with our primate and mammalian cousins.  We might call it ‘sentience’.  But it will not qualify as experience, according to McDowell's rationalist usage.  We might call the capacity for experience in this sense ‘sapience’.  As a consequence, McDowell insists that we cannot understand what we have, perceptual experiences, by construing it as the result of starting with what we share with our sentient but not sapient animal relatives, and then adding something (say, the ability to use concepts).  For what we would need to 'add' is not itself intelligible apart from the notion of perceptual experience.[4]


Other thinkers who are careful to avoid the Myth of the Given do so by placing the interface between nonconceptual causal stimuli and conceptual response at the point where environing stimuli cause perceptual judgments.  That is, they avoid the Myth by seeing nothing nonjudgmental that could serve to justify perceptual judgments, rather than just to cause them.  Davidson notoriously takes this line, endorsing the slogan that nothing but a belief can justify another belief.  I would argue that Sellars himself has a view of this shape.[5]  And it is the line I take in my book.[6]  McDowell, however, construes perceptual experiences as not involving the sort of endorsement characteristic of judging or believing: perceptual experiences have judgeable, believable contents, but they are not judgments or beliefs.  When a perceiver does advance from perceptual experience to judgment or belief, however, the experience can serve to justify the resulting commitment. 


I said above that the second feature that distinguishes McDowell's view of perceptual experience from those appealed to by empiricists of a more traditional stripe is his view that in favored cases, when perception is veridical, the content of perceptual experience just is the fact perceived.  McDowell endorses the Fregean approach, which construes facts as true thoughts—'thoughts' not in the psychological sense of thinkings, but in the semantic sense of the contents that are thought, or better, thinkable.  The obvious pitfall in the vicinity of such a view is the need to deal with the fact that we make perceptual mistakes.  That is, we sometimes cannot tell the difference between the case in which we are having a perceptual experience whose content is a fact and cases where there is no such fact to be perceived.  Traditionally, the explanatory strategy for addressing such phenomena had the shape of a two factor theory: one starts with a notion of perceptual experience as what is common to the veridical and the nonveridical cases, and then distinguishes them by adding something external to the experience: the truth of the claim, that is, the actual existence of the fact in question.  Epistemologically, this strategy sets the theorist up for the Argument from Illusion, and hence for a skeptical conclusion.  McDowell's objection to the two factor strategy is not epistemological, however, but semantic.  It is not that it makes the notion of perceptual knowledge unintelligible (though it does that, too).  It is that it makes unintelligible the notion of objective purport—our experiences (and therefore, our thoughts) so much as seeming to be about the perceptible world.  He thinks that constraint can only be met by an account that is entitled to endorse what is perhaps his favorite quote from Wittgenstein:  “When we say, and mean that such-and-such is the case, we—and our meaning—do not stop anywhere short of the fact; but we mean: this-is-so.”[7]  McDowell's perceptual realism is his way of explaining how this can be so.  Extending the doctrine of semantic empiricism, he thinks that if we can't make this feature of our thought and talk intelligible for perceptual experience, then we can't make it intelligible for any claims or beliefs. 


On his view, the only thing a veridical perceptual experience and a corresponding hallucination have in common is that their subject can't tell them apart.  There is no experience in common.  We just are not infallible about the contents of our experiences, and can confuse being in the state of having one for being in the state of having another—for instance by responding to each by endorsing the same perceptual judgment.  Once again, he insists, we cannot understand veridical experience by construing it as the result of starting with a notion of what is common to the state that prompts a veridical perceptual judgment and the state that prompts a corresponding mistaken perceptual judgment, and then adding something (say, the truth of the claim in question).[8] 


The various features of McDowell's view that I have focused on are related.  The revelation of perceptible fact in perceptual experience is 'immediate' in the sense that the conceptual abilities required (by the first condition above) are exercised passively in perception.  They are the very same conceptual abilities exercised actively in, say, making a judgment as the result of an inference, but differ in that the application of concepts in perceptual experience is wrung from us involuntarily by the perceptible fact.  The way in which concepts are brought passively into play falls short of judgment or belief, however.  The content is presented to the potential knower as a candidate for endorsement.  But an act of judgment is required to endorse it.  So what is wrung from us by the facts is not judgments, but only petitions for judgments. 


McDowell thinks it is important to maintain this distinction in order to make intelligible the sense in which we are rationally responsible for our perceptual judgments.  I agree that it is essential to make sense of that responsibility.  But I do not see that doing that requires postulating in the standard case an advance from merely entertained judgeable-but-not-yet-endorsed content to endorsement or judgment.  The fact that the passively arrived at judgments, once they are on board, are open to criticism in the light of collateral commitments—which is what being "on board" in the relevant sense means—seems to me entirely sufficient as an account of what our rational responsibility consists in.  Indeed, liability to this sort of criticism is the primary sense McDowell himself associates with the kantian concept of spontaneity in Mind and World.  I worry that in this regard he has fallen in unnecessarily and incorrectly with the pre-Kantian tradition that saw a prior, independently intelligible act of will as required prior to assessments of responsibility—a picture of a cloud of merely entertained judgeable contents awaiting the exercise of an act of will by which we plump for some of them.[9] 




I want to situate McDowell's notion of perceptual experience by placing it with respect to two other notions, one broader than his and one narrower.  The broader notion is non-inferential knowledge acquired in response to environing stimuli.  The narrower notion is that of immediate awareness of secondary qualities. 


I said above that thinkers such as Davidson, who reject the Myth of the Given, have typically rejected also the idea of any conceptually structured intermediary between causal stimuli and full-blown observational judgments.  McDowell thinks that we need to postulate perceptual experiences, which are such intermediaries—though we must be clear that they are intermediaries only in the straightforward causal sense of being brought about by environing stimuli and bringing about observational judgments, not in the sense of the sort of epistemological intermediaries that give rise to the picture of a “veil of ideas”.  His view is clearly coherent, and is not subject to the objections Davidson forwards against epistemological intermediaries as classically conceived.  But we might still ask what explanatory ground is gained by countenancing perceptual experiences, since we can avoid the Myth of the Given without them.  One part of McDowell’s answer is that his notion of experience lets us distinguish cases of genuine perception from other cases of responsively acquired noninferential knowledge.  I want to sketch an account of this broader class, and then say why McDowell thinks we must also distinguish a privileged species within this genus. 


Quine suggests[10] that what distinguishes specifically observational knowledge is that observation reports are reliably keyed to environing stimuli in a way that is widely shared within some community—so that members of that community almost always agree about what to say when concurrently stimulated in the same way.  This suggests that we think of there being two elements one needs to master in order to be able to make a certain kind of observation report, two distinguishable sorts of practical know-how involved.  First, one must have a acquired a reliable differential responsive disposition: a disposition reliably to respond differentially to some kind of stimulus.  Which stimuli we can come differentially to respond to depends on how we are wired up and trained.  Humans lack the appropriate physiology to respond differentially to different radio frequencies, for instance, without technological aids.  Blind mammals cannot respond differentially to colors.  These capacities are something we can share with nonconceptual creatures such as pigeons—or as far as that goes, with photocells and thermostats.  Second, one must have the capacity to produce conceptually articulated responses: to respond to red things not just by pecking at one button or closing one circuit rather than another, but by claiming that there is something red present.  I think we should understand this latter capacity as the ability to take up a certain kind of stance in the space of reasons: to make a move in what Sellars calls “the game of giving and asking for reasons” of a sort that can both serve as and stand in need of reasons.  A parrot could be taught to respond to red things by uttering the noise “That’s red,” but it would not be saying or claiming that anything was red.  I think we can understand what it is lacking as the ability to tell what it would be committing itself to by such a claim, and what would entitle it to that commitment—that is, what follows from the claim that something is red (for instance, that it is colored and spatially extended) and what would be evidence for it (for instance that it is scarlet) or against it (for instance, that it is green).  But nothing in what follows depends on this particular way of understanding the dimension of endorsement that distinguishes observational reports from mere differential responses. 


If it turns out that I can reliably differentially respond to a certain sort of state of affairs by noninferentially reporting the presence of a state of affairs of that sort, and if I know that I am reliable in this way, then I think that true reports of this kind deserve to be called observationally acquired knowledge.  This is in some ways a fairly radical view—though, I think, a defensible one.  For one consequence of thinking of observation this way is that there is no particular line to be drawn between what is in principle observable and what is not.  The only constraints are what a reporter can be trained under some circumstances reliably to differentiate, and what concepts she can then key the application of to those responsive dispositions.  Thus a properly trained physicist, who can respond systematically differently to differently shaped tracks in a cloud chamber will, if she responds by noninferentially reporting the presence of mu mesons, count as genuinely observing those subatomic particles.  She may start out by reporting the presence of hooked vapor trails and inferring the presence of mu mesons, but if she then learns to eliminate the intermediate response and respond directly to the trails by reporting mesons, she will be observing them.  “Standard conditions” for observing mu mesons will include the presence of the cloud chamber, just as standard conditions for observing the colors of things includes the presence of adequate light of the right kind.  And the community for whom ‘mu meson’ is an observation predicate will be much smaller and more highly specialized than the community for whom ‘red’ is one.  But these are differences of degree, rather than kind. 


Again, it may be that if challenged about a noninferential report of a mu meson, our physicist would retreat to an inferential justification, invoking the shape of the vapor trail that prompted her report.  But we need not understand that retreat as signifying that the original report was, after all, the product of an inference.  Rather, the claim of the presence of a mu meson, which was noninferentially elicited as a direct response to a causal chain that included (in the favored cases) both mu mesons and vapor trails (but which was a report of mu mesons and not vapor trails—or retinal irradiations—because of the inferential role of the concept that was applied in it) can be justified inferentially after the fact by appealing to a safer noninferential report, regarding the shape of the visible vapor trail.  This report is safer in the dual sense that first, the physicist is more reliable reporting the shapes of vapor trails than she is the presence of mu mesons (since the latter are more distal in the causal chain of reliably covarying events that culminate in the report, so there is more room for things to go wrong) and second, the capacity reliably to report the presence of vapor trails of various shapes is much more widely shared among various reporters than is the capacity reliably to report the presence of mu mesons (even in the presence of a cloud chamber).  The practice of justifying a challenged report by retreating to a safer one, from which the original claim can then be derived inferentially, should not (certainly need not) be taken to indicate that the original report was itself covertly the product of a process of inference.


As I would use the terms, following out the rationalist principle that I take McDowell also to endorse, to be aware of something (in the sense relevant to assessments of sapience) is just to apply a concept to it—that is, to make a judgment or undertake a doxastic commitment regarding it.  Awareness deserves to be called ‘immediate’ just in case it is not the product of a process of inference.  Thus, beliefs acquired noninferentially, by the exercise of reliable dispositions to respond differentially to stimuli of a certain sort by making corresponding reports (‘corresponding’ in the sense that what is reported is some element of the causal chain of reliably covarying events that culminates in the report in question) embody immediate awareness of the items reported.  The first contrasting view (null hypothesis) with respect to which I want to place McDowell’s view is then that this is the only sense of ‘immediate awareness’ we need in order to understand our perceptual knowledge of the world around us.


If we press this picture of observation as consisting just in the exercise of reliable differential responsive dispositions to apply concepts[11], even more outré examples present themselves.  Suppose that at least some people can be conditioned to discriminate male from female newly hatched chicks, just by being corrected until they become reliable.  They have no idea what features of the chicks they are presented with they are responding differentially to, but they not only become reliable, they also come to know that they are reliable.  When one of them noninferentially responds to a chick by classifying it as male, if he is correct, I think he has observational knowledge of that fact.  (And I think McDowell is prepared to agree.)  This can be so even if it is later discovered (I’m told that this is true) that the chicken sexers are wrong in thinking that they are discriminating the chicks visually—that in fact, although they are not aware of it, the discrimination is being done on an olfactory basis.  According to this way of thinking about observation, what sense is in play can only be discriminated by discovering what sorts of alterations of conditions degrade or improve the performance of the reliable reporters.  If altering light levels does not change their reliability, but blocking their noses does, then they are working on the basis of scent, not of sight.[12] 


McDowell thinks that although there can be cases of what is in a broad sense observational knowledge like this [if even that seems too generous, mark this very special sense by calling it "Bobservation"], they must be sharply distinguished from cases of genuine perceptual knowledge, for instance being able to see shapes or colors.  That is, he rejects the suggestion that the latter be assimilated to the former.  When we see colors and shapes, we have perceptual experiences corresponding to the judgments we go on to make or the beliefs we go on to form.  The chicken sexers in my example do not have perceptual experiences of chicks as male or female.  They just respond blindly, though they have learned to trust those blind responses.  There is for them no appearance of the chicks as male or female. 


Put another way, McDowell is committed to there being two kinds of beliefs acquired noninferentially by the exercise of reliable dispositions to respond differentially to stimuli by reporting elements of the causal chain that culminates in the report.  In genuine perception, the belief is the result of endorsing the content of a perceptual experience.  In the other sort (what might be called ‘mere observation’) the belief is acquired blindly, that is in the absence of a perceptual experience with the same content.  Under the right circumstances, one just finds oneself with the belief in question.  But this sort of belief formation is not a case of facts becoming visible (or more generally, perceptible) to us.  Although these beliefs are noninferentially elicited from the believer by environmental stimuli, the warrant for those beliefs is in an important sense inferential.  The believer’s justification for beliefs of this sort depends on drawing conclusions from an antecedent claim of reliability.  In this respect, the believer herself is in no different position than a third-person observer would be.


There is certainly an intuitive appeal to this distinction.  But I worry whether its appeal is merely intuitive, or whether there is important explanatory work for the distinction to do.  After all, we have lots of residually cartesian intuitions.  This worry is a pragmatic one, in the spirit of Quine’s query in “Two Dogmas of Empiricism” concerning what features of our linguistic practice—the way we actually use language—reflect or are explained by the distinction of claims into analytic and synthetic.  Intuitions that are quite possibly (I’m inclined to say necessarily) infected by prior theoretical commitments are not to the point here.  Once properly trained, we just find ourselves responding to visible red things by calling them red.  And in this usage, ‘visible’ need mean no more than ‘in standard conditions for visual observation’, that is, in good light, on an unoccluded sight line to the observer, and so on. 


Once we have relinquished the Myth of the Given, we must be careful not to assimilate the making of such noninferential judgments to the identification of something by criteria.  I may apply the concept ‘white oak’ to a tree because I have noticed the characteristic bilaterally symmetrically roundly-lobed leaves.  It makes sense to ask how I knew that it was a white oak, and an answer can be given.  But in the case of red things, there is no set of features I am noticing, from which I conclude that they are red.  I can just tell red things by looking at them.  If there weren’t some features like this, there couldn’t be any empirical knowledge of the sort exemplified by my white oak judgment either.  I can say that the patch looks red, in a sense of ‘looks’ that is no more committive than that involved in saying I can tell red things by looking at them.  That is the only sense in which the world need appear to me as anything.


In “Empiricism and the Philosophy of Mind”, Sellars offers a recipe for introducing ‘looks’ or ‘appears’ talk, wherever there is a noninferential reporting practice.   Whenever a reporter suspects her own reliability under certain conditions of observation, she can express her usually reliable disposition to report something as being f, but withhold her endorsement of that claim, by saying only that it looks (or appears) f.  The chicken sexers are certainly able to introduce ‘looks’ and ‘appears’ talk in this way.  But McDowell’s claim is then that there is an important difference between such uses of these locutions and their use to report perceptual experiences.  He thinks that the capacity to have perceptual experiences is different from, and more fundamental than, the capacity to make noninferential observations of mu mesons in cloud chambers and of the sexes of chickens.  Unless we could have perceptual experiences, we could not make any observations at all—even though not all observations of a state of affairs involve perceptual experiences of those states of affairs.  That is, the capacity to become noninferentially informed about the world by learning blindly to respond differentially to it depends upon a more basic capacity for states of affairs to become immediately apparent in perception.  Thus it is important to McDowell to distinguish a notion of conceptually structured perception that is narrower than the merely responsive notion of conceptually structured observation I have sketched. 


I asked above, in a pragmatic spirit, about what explanatory work such a distinction does for us.  McDowell has a response, of course.  It is that without the notion of conceptually articulated perceptual experience that distinguishes genuine perception from merely responsively acquired noninferential belief, we cannot understand the empirical content of any of our claims.  For without that notion, we are doomed to embrace one horn of the dilemma from which Mind and World sets out to free us.  In McDowell’s view, the picture of observation I have been suggesting may be all we need substitutes mere causal constraint by the world for the genuinely rational constraint that is required for us to make intelligible to ourselves the idea of our beliefs as about the world without us.  For that notion of aboutness requires that our beliefs answer rationally for their correctness to the facts they purport to present, not merely that they are causally occasioned by them.  This way of working out semantic empiricism presents deep and important issues, which I cannot pursue here.  (I have addressed some of them in a preliminary fashion elsewhere.[13]) 


Rather than press that set of global philosophical issues, I focus here on the distinction McDowell is obliged to draw between genuine perception and what I have been calling ‘mere observation’ ["Bobservation"] to raise a much more local and limited issue.  What sort of a fact is it that in some cases where we noninferentially acquire a true belief by exercising a reliable disposition noninferentially to respond to the fact in question by acquiring the belief there is a perceptual experience present, while in others there is not?  How would we go about settling the question of whether the physicist has genuine perceptual experiences of mu mesons?  Is there any way in principle to tell other than asking her?  And if we do ask her, is there any chance that she is wrong, because she has been taught a bad theory?  Could I think I was having perceptual experiences of mu mesons or the maleness of chickens when I was not, or vice versa?  Do we know just by having a perceptual experience what sensory modality it corresponds to (so that the—supposed—fact that the chicken sexers get this wrong is decisive evidence that they do not have genuine perceptual experiences)?  The answers to questions such as these determine just how classically cartesian McDowell’s notion of perceptual experience is—and so, from my point of view, just how suspicious we should be of it.  I do not assert that his answers to these questions will be Cartesian ones.  I don’t know how to answer them, and do not find much help in Mind and World. 





 Putting things in terms of how the world appears to us raises a danger of getting McDowell wrong in the other direction, however.  For a natural response to the sort of distinction of cases on which I am claiming McDowell insists—at least for philosophers familiar with the empiricist tradition McDowell is extending—is to think that what sets off mere observation of the sort epitomized by the mu meson and chicken sexing cases from genuine perception is that the physicist and the chicken sexer are not reporting their awareness of any secondary qualities.  Being a mu meson or a male chick are primary qualities, and so not directly or immediately experienceable in the sense in which secondary qualities such as red are.  For traditional empiricism took it that our awareness of the perceptible world is, as it were, painted in secondary qualities: qualities that nothing outside the mind can literally have, purely experiential properties more or less reliably induced in minds as the effects of external bodies.[14]  These secondary qualities correlate with, and so represent features of perceptible objects.  But since they are merely the effects those features have on suitably prepared and situated minds, they do not present properties literally exhibited by the objects themselves.  Phenomena of this sort, the secondary qualities of things,  are all that is directly or immediately perceivable.  Coming to know about anything else is the result of making inferences from the occurrence of the experiences of secondary qualities they occasion in us.


Following Gareth Evans, McDowell has endorsed a pragmatic account of the distinction between secondary and primary qualities.  (By calling it ‘pragmatic’ I mean to indicate that it defines the distinction in terms of differences in the use of expressions for—predicates used to attribute the occurrence of—secondary and primary qualities.)  According to this way of understanding things, to take f to express a secondary quality concept is to take it that one cannot count as having mastered the use of ‘f’ talk[15] unless one has also mastered the use of ‘looks-f’ talk.  This criterion distinguishes predicates such as ‘red’, which express secondary qualities, from those such as ‘square’, which express primary qualities.  For one does not count as fully understanding the concept red unless one knows what it is for things to look red.  While a blind geometer can count as fully understanding the concept square even if she cannot discriminate one by looking at it.  According to the minimally committive account of observation sketched above, one can learn ‘looks-f’ talk just in case one has mastered the noninferential circumstances of appropriate application of the concept f—that  is, just in case one has both mastered the inferential role of the concept, and has been trained into the reliable differential responsive dispositions that key its noninferential application to the apparent presence of the reported state of affairs.[16] 


Since McDowell’s ‘minimal empiricism’[17] seeks to rehabilitate what was right in the appeals to experience that motivated classical empiricism, it is tempting to understand his distinction between genuine perceptual experience and mere noninferential observation of environing circumstances in terms of the role of secondary qualities in the former.  Perceptual experience, the thought would run, is always experience immediately of secondary qualities.  That is what is missing in the mu meson and chicken sexing case.  (Not that there are not secondary qualities involved in those cases, but rather that what is reported in those cases is not the occurrence of secondary qualities.)  But this would be to misunderstand McDowell’s position.  For he thinks we  can have perceptual experience of some primary qualities, not just secondary ones.  Thus shapes, for instance, can be visible and tangible—genuinely the subjects of perceptual experience.  Where there are perceptual experiences, there are appearances, which can be reported by the use of  ‘looks’ talk.  And since McDowell admits that a certain attenuated form of ‘looks’ talk applies even to mere observation, without corresponding perceptual experiences, it should be marked that in these cases it will be ‘looks’ talk in the stronger sense.  But the existence of perceptual experiences that are being reported by such ‘looks’ talk does not require that the mastery of such talk is an essential feature of mastery of the concepts being applied.  Talk of perceptual experiences is not a way of talking about secondary qualities.  All immediate awareness of secondary qualities involves perceptual experiences, but not necessarily vice versa.


Here it is important to keep in mind a distinction between two different ways in which one might understand the Evans-McDowell characterization of secondary qualities in terms of ‘looks’-talk.  I claimed above that Sellars gives us a recipe for introducing a use of ‘looks-j’ (or, more generally, ‘appears-j’) corresponding to any predicate j that has a noninferential reporting use.   According to this understanding, there is no problem with the physicist talking about things looking like mu mesons, or the chicken sexer talking about things looking like (or appearing to be) male chicks.  If the Evans-McDowell criterion for being a secondary quality concept is combined with this understanding of the use of ‘looks’ (or ‘appears’), then what results is the notion of concepts that are essentially observable—in the sense that in order fully to master the concept, one must have mastered its noninferential circumstances of application.  Red is pretty clearly like this, and mu meson is pretty clearly not like this.  But just as we can introduce a use for ‘looks to be a mu meson’, we could also introduce another concept, which is just like mu meson except that mastery of the noninferential use of the expression, and of the corresponding ‘looks’ locution, is required for certification as having mastered that concept.  And similarly for any merely observational property.  This fact may suggest that the notion of essentially observable [Bobservable] concept should not be identified with the classical notion of secondary quality concept.


In fact, this is McDowell’s view.  For this reason, he does not understand the appeal to ‘looks’ in the definition of secondary qualities in the minimal Sellarsian sense identified above.  For in this sense, there could be experience of secondary qualities where there are no corresponding perceptual experiences in his sense.  And he is committed to perceptual experiences being necessary, though not sufficient, for awareness of secondary qualities.  McDowell understands the responsive use of ‘looks-j’ locutions as genuine reports—not, as on the minimal Sellarsian line rehearsed above, merely expressions of dispositions to make endorsements one is not in fact making.  What ‘looks’ claims report (at least in the central cases) is just perceptual experiences. 


I have situated McDowell’s notion of perceptual experience between a broader notion and a narrower one—between the concept of knowledge noninferentially acquired by applying concepts as the result of reliable differential responsive dispositions, and the concept of immediate awareness of secondary qualities.  As I pointed out above, McDowell denies that the broader concept of merely noninferential knowledge is independent of that of perceptual experience: if we could not have perceptual experiences, then we could not know things noninferentially at all.  (Indeed, he thinks we could not know anything at all.)  I would like to end this discussion with a question, his answer to which I have not been able to determine from McDowell’s writings:  Could there be perceptual experience, for McDowell, if there were no secondary qualities?  That is, could anyone have perceptual experiences of primary qualities if she could not also have perceptual experiences of secondary qualities?  If not, why not?  If so, what would it be like?  And once again, what sort of questions are these?  How ought one to go about addressing them?  Is it a matter for introspection, or for empirical investigation?  If purely philosophical argumentation is needed, what are the criteria of adequacy according to which we should assess the answers?




In closing, I would like to add a further query.  If we look at the end of Mind and World, we see that we can have non-inferential knowledge of normative facts: of meanings, for instance, and of how it is appropriate to act.  Coming to be able to make such non-inferential judgments is part of being brought up properly, part of acquiring our second nature.  Along something like the same lines, in his earlier writings, McDowell has urged (in opposition to Davidson’s interpretational view) that fully competent speakers of a language do not infer the meanings of others’ utterances from the noises they make, rather they directly or immediately hear those meanings.  Coming to speak the language is coming to be able to perceive the meanings of the remarks of other speakers of it.  The connection I have in mind between these claims is that claims about what someone means are normative claims.  They have consequences concerning what she has committed herself to, what she is responsible for, what it would take for her claim to be correct, and so on.  So McDowell’s view is that normative facts are noninferentially knowable.


It has always seemed to me to be one of the great advantages of the account of observational knowledge in terms of reliable differential responsive dispositions to apply concepts noninferentially that it makes perfect sense of these claims.  If I have mastered the use of some normative vocabulary (whether pertaining to meanings, or to how it is proper to behave nonlinguistically), and if I can be trained reliably to apply it noninferentially, as a differential response to the occurrence of normatively specified states of affairs, then I can have observational knowledge of those normative states of affairs: I can see (or at least perceive[18]) what it is appropriate to do or say.  Normative concepts are no worse off than concepts like mu meson in terms of their capacity to acquire observational uses.  So here is my final question for McDowell:  is this mere non-inferential knowledge?  Or are the normative statuses also perceptually experienceable, for McDowell?  I don't think he commits himself on this, any more than he does on the question of whether secondary qualities are necessary for experience.  Indeed, one could ask further:  are there (can there be) secondary qualities corresponding to essentially normative states of affairs that are noninferentially knowable?


I think that these are important questions in their own right.  It seems to me a virtue of McDowell’s writings about sense experience that it brings such questions into view.  I am also inclined to think that I do not fully understand McDowell’s concept of perceptual experience until I know how it bears on this sort of question.  McDowell does not address himself to these questions, and I do not know what he would say about them if he did.  I do suspect[19] that his response will be in the form of deflection:  Interesting questions though these might be, the project of Mind and World does not require that they be addressed.  For the project of that work is heavily diagnostic and lightly therapeutic, but not at all theoretical.  Its task is not, as its author understands it, to present a theory of content, or of perceptual experience.  It is rather to make evident to its readers the common presuppositions powering an oscillation between two equally unsatisfactory ways of talking about the role of experience in empirical thought.  It aims further at giving us some instruction in how we might talk once we have freed ourselves from attachment to those fatal philosophical assumptions that structure so much of the tradition by which we have been shaped.  Doing that does not, the claim would be, require taking a stand on every potentially controversial issue that could arise in the vicinity once we have thrown off the fetters in which commitment to defective (though after Mind and World intelligibly and forgiveably tempting) ways of thinking about nature and the relation between causal and rational constraint have bound us. 


The response I am putting in McDowell’s mouth has considerable force.  But in the end, I do not find it satisfactory.  It seems to me that the therapeutic dimension of the enterprise of Mind and World involves a commitment to there being at least some satisfactory way of extending the things he has said about observation, perception, and sensory experience so as to answer the sort of questions our following out of his remarks has raised.  For instance, he is committed to there being a distinction between two sorts of noninferentially acquired knowledge of states of affairs: in one kind there is an experience of that state of affairs, and in the other not.  But, we should ask, does this distinction manifest itself in any way or explain anything outside the confines of the theory?  (Compare Quine’s corresponding question about the analytic/synthetic distinction.)  Or is it real only in the way the question of whether socinianism is a heresy once had to be taken seriously, because until it was settled we wouldn’t know who the true Pope is?  I think the issue of whether the ways McDowell has recommended we talk about perceptual experience can be extended so as to afford sensible answers to the sorts of questions I have argued his discussion implicitly raises delineates a fair dimension along which the adequacy of his story should be appraised.


McDowell’s bold and ingenious rehabilitation of the empiricists’ concept of experience requires us to make conceptual distinctions far subtler than any the tradition worried about.  He also gives us the conceptual raw materials to make those distinctions clear.  This is all pure advance.  I have sought here to rehearse some of these distinctions, and to use them to invite McDowell to commit himself in the terms he has provided on issues that he has not yet formally addressed.


Bob Brandom

University of Pittsburgh

[1]   Mind and World [Harvard University Press, 1994] p. xii (from the Introduction, added in the paperback edition of 1996].

[2]   I would say, at least since van Fraassen's sophisticated version.

[3]  Originally published in 1956, this classic essay has recently been reprinted, with an Introduction by Richard Rorty and a section by section Study Guide by Robert Brandom [Harvard University Press, 1997].

[4]   In this respect he parts company with the picture I present in Making It Explicit [Harvard University Press, 1994], as he makes clear in his comments on the book in “Brandom on Inference and Representation,” Philosophy and Phenomenological Research Vol. LVII, No.1, March 1997, pp. 157ff.

[5]   What he calls 'sense impressions' are causal antecedents of perceptual judgments, but do not serve to justify them.

[6]   Making It Explicit.  See especially the first half of Chapter Four.

[7]   Philosophical Investigations §95.

[8]   On this point, see his “Knowledge and the Internal”, and my companion piece “Knowledge and the Social Articulation of the Space of Reasons,” both in Philosophy and Phenomenological Research 55 (4), December ’95.

[9]   Of course, as pointed out at the outset, McDowell breaks with this picture in rejecting the idea that we could learn merely to entertain conceptual contents first, without at the same time having endorsed many of them.  But as I see it, this avoids only one of the objectionable features of this cartesian picture.

[10]   In “Epistemology Naturalized”, in Ontological relativity, and other essays, [New York: Columbia University Press, 1969]

[11]  I develop and defend such an account in Chapter Four of Making It Explicit.

[12]   I owe this way of thinking about the difference between sensory modalities in terms of conditions of reliability to Lionel Shapiro.

[13]  “Perception and Rational Constraint” in Philosophical Issues 7  1996: Perception (Sociedad Filosofica Ibero Americana—edited by Enrique Villaneuva) pp. 241-260.  Abbreviated version published in Philosophy and Phenomenological Research, LVIII No. 2 (June 1998), pp. 369-374.

[14]  Berkeley is the paradigmatic defender of such a view, but as an implicit theme, this way of thinking about secondary qualities was pervasive in pre-Kantian empiricism.

[15]  Sellars glosses grasping a concept as mastering the use of a word. 

[16]  McDowell will insist that a richer notion of mastering ‘looks’ (or, more generally, ‘appears’) talk—one that involves the reporting of perceptual experiences, not just the conceptually structured exercise of reliable differential responsive dispositions—should be brought to bear in defining secondary qualities.  But this qualification does not make a difference for the use I am making here of the Evans-McDowell characterization of secondary qualities.

[17]  His characterization, in the new Introduction to the paperback edition of Mind and World.

[18]  Perhaps not ‘see’ or ‘hear’, since these terms are committive as regards sense modality—commitments to be cashed out, as I indicated above, in terms of the nature of the conditions that degrade or improve reliability.  Notice that in this sense it is appropriate to talk about hearing the meaning of someone’s oral utterance, and seeing the meaning of her written remark. 

[19]   Based on our conversations on the matter in connection with our joint seminar on Perception (University of Pittsburgh, Spring 1998).