James W. Cornman, Perception, Common Sense, and Science, Yale University Press, 1975.
7 Indirect Realism
With the rejection of naive realism because of its falsified view of perception, it is natural to turn to a theory that not only fits the falsifying facts, but explains them as well. We have seen that it is the claim that we are perceptually acquainted with external physical objects that should be rejected. It is tempting, however, to try to maintain that we are perceptually acquainted with something in perception. Indeed, it might be argued that if we are not perceptually acquainted with something, we would not perceive anything even indirectly. With the rejection of naive realism, the obvious candidates for objects of perceptual acquaintance are sensa. In such a way we might find ourselves coming to embrace indirect realism.
Indirect realism, like naive realism, can be construed as consisting of two subtheses, one about the nature of the external world and the other about our relationship to it when perceiving it. As I shall interpret the minimal position, however, it does not explicitly state those properties that external objects have or lack. Thus an indirect realist need not agree with Locke about primary and secondary qualities. He might hold that sensa of secondary qualities as well as those of primary qualities resemble them. What is essential to his position about external objects is what makes him a realist in common with a Lockean, a naive realist, a modified naive realist, and a Kantian realist. This is the thesis that there are external objects -- nonperceiving objects that exist independently of and unaffected by being perceived. However, unlike a Kantian realist, he holds that at least some of these external objects are perceivable physical objects. Thus we have:
Indirect Realism (first thesis) : There are perceivable physical objects that are not perceivers, that exist unperceived, and that are not affected either by being perceived or, generally, by changes in the usual conditions in which they are perceived.
The second thesis centers on the claim that those external objects that are perceived are indirectly, rather than directly, perceived.
[249]
This is because in all cases of perception of a physical object, a sensum is directly perceived. Yet this is not quite precise enough because it allows for the possibility that direct perception of, for example, a red sensum consists in red-sensing as a result of some effect on the perceiver by the sensum. I believe it is clear that no indirect realist would hold this odd view. He would insist on claiming that any perceiver of an external object is perceptually acquainted with at least one sensum. Indeed, all perceptual experiences a person has of physical objects consist in the person being perceptually acquainted with sensa. But this can be even further clarified, because as stated when discussing perceptual acquantaince, we need not mention the requirements that the perceiver be normal and that conditions be optimal when considering only individual sensa. Each perceiver is normal regarding his own sensa, and he perceives them under optimal conditions. Thus, according to an indirect realist, a perceiver is not only perceptually acquainted with his own sensa, he experiences them as they are, that is, as I shall say, he senses them. Following our discussion in the preceding chapter, we can define this as follows:
s senses sensum p at t = df
- s immediately perceives sensum p at t, and
- for any sensible phenomenal property, Q, if, at t, s has an experience of p as having Q, then at t he immediately perceives Q, which is a property of p at t.
We have, then, the following:
Indirect Realism (second thesis): Whenever an external physical object is perceived by someone, he indirectly perceives it and his perceptual experience of it consists in his sensing some one or more sensa.
We can see now where an indirect realist differs from other realists. Unlike all direct realists, and thus naive realists (including modified naive realists, as I have interpreted Galileo) he claims that no external objects are directly perceived. He differs from direct realists, then, because of his second thesis. He differs from a Kantian nondirect realist and a phenomenalist regarding both theses. But even with this definition that distinguishes an indirect realist in these ways, there remains much room for disagreements
[250]
among indirect realists. To see this in more detail, let us consider Locke, a classical example of an indirect realist who is also a "representative" realist, and a contemporary Sellarsian indirect realist.
A Lockean Representative Realism
What is the minimum statement that would distinguish Lockean representative realism from the other versions of indirect realism? We can, I find, ignore many of Locke's claims, such as those about material substance, and so ignore many of his confusions and mistakes. I would claim, however, that there are three distinctive features of a Lockean representative realism. One concerns the causal relationship between certain (for Locke, simple) sensa and the properties or qualities of external objects, one concerns the picture-like representational function of sensa, and one is an epistemological thesis about the grounds or foundation of empirical knowledge of the external world. These three theses are adumbrated in the following three quotations. First:
Whatsoever the mind perceives in itself, or is the immediate object of perception, thought, or understanding, that I call idea; and the power to produce any idea in our mind, I call quality of the subject wherein that power is.1
Second:
The ideas of primary qualities of bodies are resemblances of them, and their patterns do really exist in the bodies themselves; but the ideas produced in us by these secondary qualities have no resemblance of them at all.2
Third:
It is evident the mind knows not things immediately, but only by the intervention of the ideas it has of them. Our knowledge therefore, is real, only so far as there is conformity between our ideas and the reality of things. But what shall be here the criterion? How shall the mind, when it perceives nothing but [251] its own ideas, know that they agree with things themselves?3
I believe that the two theses about the causation of sensa and the resemblance of some sensa to their causes is clear in the first two passages. However, the epistemic thesis in the third selection requires some interpretation. According to Locke, it is not merely that only sensa are directly perceived, but also that a person has immediate or noninferential knowledge solely about himself and his own ideas, including sensa. Furthermore any knowledge someone has about either the existence or properties of external physical objects is inferential knowledge derived from knowledge of certain sensa. Thus a person's basic or ultimate premises for justifying claims about the external world refer exclusively to himself as a perceiver and his own sensa. Consequently, on this view, only deduction and induction by enumeration can be used to warrant inferences. What is called "hypothetical" or "hypothetico-deductive" induction is excluded, because it allows the assumption of the hypothesis to be proved as a premise in order to justify it by its experienced consequences. Using this interpretation of Locke's epistemological thesis, we can construct the following definition:
Lockean Representative Realism:
- Indirect realism is true, and
- all sensa a person senses when perceiving an external object are caused by some quality of the object, and
- some but not all sensa so caused resemble the qualities (that is, primary qualities) that cause them, and
- persons have knowledge of the existence and properties of some external objects, and each person's knowledge of this sort is inferred (deductively or inductively by enumeration) from knowledge solely about his own sensa and himself as perceiver.
It is widely held that the refuting objection to representative realism is that it leads to epistemological skepticism about the external physical world, and is, therefore, highly implausible. If this is true, then, as I have characterized the position with its claim that there is such knowledge, it leads to a contradiction. Nevertheless,
[252]
although this is a standard, even textbook, objection, I do not believe it has been firmly established. Indeed, I would claim that without clearly stating the various subtheses of the theory, no refutation succeeds, and no understanding results about which subtheses it would be best to abandon in order to avoid the objection with minimal change in the theory. The objection, of course, is recognized by Locke, as is clear from the second question he asks in the third quotation above. How, indeed, given this theory, are we to know that certain of our sensa agree with, or correspond to, let alone resemble, external objects? Locke's answer is that there are two sorts of ideas which we are justified in assuming "agree with things." The first are simple ideas, which, because they are not caused by the mind but by external objects,
represent to us things under those appearances which they are fitted to produce in us, whereby we are enabled to distinguish the sorts of particular substances, to discern the states they are in, and so to take them for our necessities, and apply them to our use.4
The second sort are all complex ideas, except that of substance, which are "not intended to be the copies of anything, nor referred to the existence of anything, as to their originals".5 Thus they are not supposed to represent or "agree with" any external objects, but only with ideas "of the mind's own making."
I believe we need not worry about whether only simple ideas represent external objects, but we do have to clarify what it is for a sensum to represent, be a copy of, agree with, and conform to some property of an external object. It clearly does not require representational picture-like resemblance, because Locke says,
The idea of whiteness, or bitterness, as it is in the mind, exactly answering that power, which is in any body to produce it there, has all the conformity it can, or ought to have, with things without us. And this conformity between our simple ideas, and the existence of things, is sufficient for real knowledge.6
Thus ideas of secondary qualities, which, for Locke, surely do not pictorially resemble those qualities, do represent or conform to them. What is essential to such representation is that there is a one-to-one correspondence between each distinct (simple) sensum we have at a certain time and a distinct property of an external object that causes the sensum. We can, then, use such sensa as proxies for properties of external objects in our attempt to obtain knowledge of external objects and the relationships among them. The relationships we can discover among simple sensa correspond to relationships among external objects and their properties. Thus we can justify generalizations relating properties of external objects by means of directly perceived constant conjunctions among those sensa that are correlated with, because caused by, those properties.
The crux of this Lockean rebuttal to the objection that his theory bars knowledge of external objects can be laid out as follows:
- Some of my (simple) sensa are not caused by me.
- If a sensum is not caused by me, then it is caused by an object external to me.
- For each distinct (simple) sensum, there is one corresponding distinct quality of the entity that causes it.
Therefore
- For some distinct (simple) sensa, there are corresponding distinct qualities of objects external to me.
I believe that Locke can plausibly argue for premises (1) and (2), even given the preceding restrictive interpretation of his epistemological theory. Premise (1) could be based on his awareness of his own passivity when receiving certain sensa. Premise (2) can be construed as an instance of the general claim that every entity has a cause, which, Locke might argue, is a conceptual truth and thus is merely about relations among ideas. Thus it is somewhat plausible to construe both (1) and (2) as statements solely about sensa and their perceivers, and so they could be known non-inferentially. However, (3) is more difficult for Locke to justify. Nevertheless, I believe he might also plausibly argue that it is an instance of another conceptual truth about causation, namely, [254]
for each distinct effect there is something with a distinct capacity which, given certain initial conditions, is exactly causally sufficient for just that effect.
Surely, much of the preceding attempt to reconstruct a plausible Lockean argument is debatable, but let us grant, for present purposes, that this argument succeeds. The question, of course, is just how much we have granted. It might seem that we have granted Locke all he needs to establish that on his theory there is inferential knowledge about external physical objects and at least some of their properties, namely, those that correspond to certain of our sensa. However, he has not established this, even with what we granted him. At most he has shown that there is at least one entity of some sort which is different from and independent of himself, and that it or some other entities have properties corresponding to certain of his sensa. He has not, then, found a way to justify that there are many external objects, let alone that they are physical objects. Furthermore, he has provided no way to justify anything about what properties these external objects have, except that they have capacities of some sort that correspond to certain of our sensa. Some properties correspond to sensed whiteness and sensed bitterness, but Locke has given us no way to justify what these properties are. In brief, he has done no more than justify a claim about the structure of what is external to himself. But such a claim is compatible with a Kantian realism with its realm of unknowable things-in-themselves, and also with a phenomenalism that postulates God as the one "external" object. In the latter case, it would be the structural relationships among God's idea (perhaps archetypes) that would correspond to the structural relationships that Locke discovers among certain of his sensa.7 But Locke claims to know more about what is external to him than this limited bit of information. The external world, he claims, consists of physical objects with certain primary qualities and certain imperceptible constituents. Can he justify this claim?
Rejection of the Lockean Theory: No Inference from Sensa to External Physical Objects
We have now reached the point at which we can make the [255]
decisive objection to the Lockean form of representative realism, which might be called "representational" realism. A Lockean's claim about the properties that external objects have is at best known inferentially, and his one way to justify the required inference is by means of his claim about the resemblance -- seemingly representational or pictorial resemblance -- between some of those properties and the sensa he is caused to sense because of those properties. Given the preceding Lockean epistemological theory, however, this resemblance claim is also at best justified inferentially, but the theory allows no way to do this. There is no way, either deductively or inductively, to infer that there are such resemblances from premises solely about oneself, one's own sensa, their sensible properties, and relationships among them. In other words, while a Lockean might be able to justify that certain sensa represent, in the sense of "stand proxy for," certain discriminable external properties, he has no way to justify any claim about what these properties resemble. It may be that our knowledge of these proxies and their relationships are suitable as a guide to life, that is, for making the predictions we need in order to know how to plan and what to expect in the future, because the lawlike structural relationships among sensa correspond to lawlike structural relationships among properties of external objects. For example, such a correspondence would obtain if substitution of sensum terms for all variables replacing physical-object terms in a Ramseyized theory should preserve truth. Nevertheless, no such structural knowledge provides grounds for knowledge of what constitutes the external world, which is one kind of knowledge a Lockean theory assumes. Let me lay out the argument that defeats this Lockean theory. It is, incidentally, similar to Stace's argument against realism.8 One difference between the two is that while the present argument works against this Lockean theory, neither it, nor Stace's argument succeeds against many other forms of realism. The argument is:
- If Lockean representative realism, that is, representational realism, is correct, then there is knowledge about what some external properties are only if some claim that some external properties (pictorially) resemble certain sensa is justified. [256]
- If representational realism is correct, then some such resemblance claim is justified, only if either such resemblances are directly perceived, or such resemblance claims are justified by a sound deductive or inductive argument with premises solely about sensa and their perceivers.
- If representational realism is correct, no case of such resemblance is directly perceived.
- If representational realism is correct, then such a resemblance claim is deductively justified, only if some statement solely about sensa and their perceivers entails the resemblance claim.
- No statement solely about sensa and their perceivers entails such a resemblance claim.
- If representational realism is correct, then such a resemblance claim is inductively justified, only if either it is inductively inferable from premises solely about sensa and their perceivers, or some samples of such a resemblance are directly perceived.
- No such resemblance claim is inductively inferable (by enumeration) from such premises.
Therefore
- If representational realism is correct, then there is no knowledge about what any external properties are.
- If representational realism is correct, then there is such knowledge.
Therefore
- Representational realism is incorrect.
Given our previous characterization of the Lockean theory, premises (2), (3), (4), (6), and (9) are clearly true, with the one priviso that statements solely about sensa and their perceivers include statements relating sensa to other sensa, and also statements expressing conceptual truths, that is, for a Lockean, relations among ideas. This leaves only premises (1), (5), and (7) available for him to reject.
It is easy to see that (5) is acceptable by recalling the discussion of
[257]
analytical phenomenalism in Chapter 3. The same sort of considerations that counted against any sensum statements, whether probabilistic or not, entailing physical-object statements, also can be used to refute any particular claims about sensum statements entailing these resemblance sentences. As in the previous case, this would show premise (5) to be justified inductively, from counter examples to attempts to refute it. Premise (7) is seen to be correct once it is realized that as construed here, induction is restricted to induction by enumeration. This requires that the premises mention entities that are samples of the kinds of entities mentioned in the conclusion. But no premises solely about a perceiver and his sensa mention any relationships between sensa and external properties.
There may be some disagreement about (1), primarily because neither Locke nor the Lockean theory characterized above make clear whether or not justification of the resemblance claim is crucial for knowledge of the external world. Nevertheless, it seems reasonable to interpret Locke, and our Lockean, in this way. But, if objections remain, the argument can be reformulated so it involves nothing about such a resemblance. Premise (1) can be omitted, (2) replaced as below, and corresponding adjustments made in premises (3) through (7). The following is surely acceptable:
(2a) If representational realism is correct, then there is knowledge about what some external properties are, only if either some external properties are directly perceived, or some claim about what property some external object has is justified by a sound deductive or inductive argument with premises solely about sensa and their perceivers.
What results is a sound argument, given the preceding characterization of Lockean representative realism.
The question immediately arises about what the minimal change would be that a Lockean could make in his theory to avoid this objection. He could renounce (9) and surrender to epistemological skepticism about all properties of whatever entity or entities cause our sensa. He would then have the following choice. He could embrace a Humean skepticism about the external world, which for Hume is a world of physical objects. Or he could adopt a version of Kantian realism and be skeptical about the external world, but
[258]
not about physical objects. He would, in the latter case, construe physical objects as "nothing but" appearances or appearings, and refrain from any knowledge claims about any (nonrelational) properties of whatever external objects there may be. Surely, neither alternative is very appealing to an erstwhile Lockean; yet there are no others available to him if he maintains the sensum-based epistemology expressed in the fourth defining condition of his thesis. Indeed, the amended argument shows that any indirect realist, whether or not Lockean, can avoid refutation only if he rejects either (9) or this restrictive epistemological theory. Premise (3a) -- that is, (3) amended to agree with (2a) -- is true for any indirect realist, (5a) and (7a) are true independently of the theory, and (2a), (4a), and (6a) are implied by the epistemological theory conjoined with indirect realism. To avoid skepticism or refutation, then, an indirect realist must either adopt a less restrictive epistemological theory or avoid all such theories. In essence, yet with many differences in emphasis and detail, we can construe the adoption of such an alternative theory as the crucial change in the move from the Lockean theory to Sellarsian scientific realism.
Sellarsian Scientific Realism
Generally two obstacles have been raised for an indirect realist who wishes to replace a sensum-based epistemology with a less restrictive, yet acceptable, theory. The first is the most difficult task of specifying a form of justification other than deduction and induction by enumeration which is acceptable, not only generally, but for the indirect realist's specific needs. The second is not basically an epistemological obstacle, but one supposedly having epistemologically disastrous consequences for an indirect realist. It is the claim that the correct theories of concept formation and concept learning are empiricist theories of the sort that restrict either the meaningful or justifiable use of concepts to what is directly perceived. Indeed, if the Kantian variety of a theory of concept learning is correct, then an indirect realist has only the choice between Humean and Kantian skepticism. Under such conditions Kantian realism might well seem more reasonable.
Empiricist Theories of Concept Formation and Learning
As I interpret it, a theory of concept formation is a theory of how
[259]
concepts of different sorts come to be meaningfully developed and formulated in a language. Thus it may be that the development of any language requires that the first sorts of concepts formulated are those with which a person responds to what he directly perceives. On this view, only after these basic concepts are developed might theoretical terms, including some psychological terms, come to be developed, perhaps by means of some sort of extensions or changes in the uses of the basic concepts. On the other hand, a theory of concept learning is basically a theory about the way or ways in which a human being can come to understand or acquire concepts, usually already formulated in some language. Furthermore, it is clearly possible that the required order, if there is one, in which concepts come to be formulated in a language is different from the required order, if any, in which a human being comes to understand or acquire these same concepts. I find that it is primarily theories about the order of concept learning or acquiring that have raised problems for indirect realism. In particular, there are three basically different varieties of empiricist theories which philosophers have proposed and which have bothersome epistemological consequences for indirect realism.
What makes these theories empiricist theories is that each implies that, if a person is to understand or to have acquired a nonlogical concept, then he must have had sensory experiences of a specified sort. They differ in how they specify this role of sensory experience. The first theory, which stems from Aristotle and Aquinas, I shall call "abstractionism;" the second, associated with the British empiricists, can be called "associationism;" and let us dub the third "Kantian constructionism." Let me roughly define one species of each sort as follows.
First:
- Abstractionism: A person acquires a nonlogical concept, C (for example, redness), if and only if:
- The universal, U (for example, redness) corresponding to C is exemplified by, and thus exists in, observable particulars in such a way that the entities with which the person is perceptually acquainted in perceiving the particulars contain either U (redness) itself or something suitably similar to U (redness), and he abstracts this common universal element from these entities, or [260]
- he constructs the concept ultimately from concepts acquired in accordance with (a).9
Second:
- Associationism: A person acquires a nonlogical concept, C (for example, redness), if and only if
- repeated association of a term, T ('red'), with instances of property P, (being red) with which the person is perceptually acquainted has resulted in the person understanding that T ('red') is to be used to express the concept C (redness), or
- (as above in I)
Third:
- Kantian Constructionism: A person understands a non-logical concept (for example, redness), if and only if he has developed it in attempting to impose, in the way required by certain of the rules, that govern all human understanding, some order and unity on the relevent sorts of (spatio-temporal) data with which he is perceptually acquainted.
For our purposes, there is no appreciable difference between I and II. It would seem that one of them is capable of explaining the acquisition of a concept just in case the other is also. The crucial question for both theories is to discover what additional sorts of concepts they allow us to acquire by construction, given the sorts of concepts they allow us to acquire from experience, that is, by abstraction or association. Generally, restrictive limitations have been put on our constructive ability so that a concept is acquired by construction only if it can be completely characterized by previously acquired concepts, and, thus ultimately by
[261]
non-constructed concepts. This view of the construction of concepts corresponds to the view that any cognitively meaningful term must either be explicitly defined ultimately by terms that refer to what is directly experienced, or be such a term itself. For example, there have been attempts to show that theoretical terms, including nonsensory psychological terms such as 'believe,' are definable by physical-object observation terms, and that such observation terms are in turn definable by sensation terms. But, as is well known, all such attempts at such logical constructionism have failed.10 Consequently, given these theories of concept learning, if indirect realism is correct, and if our constructive ability is so limited, then only sensation concepts are acquired from experience. Thus no physical-object concepts, whether observation concepts or theoretical concepts, and no nonsensory psychological concepts would be acquired at all. Even for a naive realist, however, no theoretical concepts and no nonsensory psychological concepts would ever be acquired. As a consequence, no form of scientific realism would be viable, because all its varieties are reasonable only if we are capable of understanding the world using theoretical terms and thus theoretical concepts. But if we are unable to acquire them, we are obviously unable to understand the world in terms of them. Clearly, then, the Sellarsian theory requires a different theory of concept learning.
The obstacle supposedly raised by the Kantian theory is different. As previously stated, it provides no reason to think that we do not come to understand and justifiably use theoretical concepts. Surely these are important tools we use in our attempts to organize and unify the myriads of sense impressions impinging on us. On this view, these concepts like all other empirical concepts including psychological concepts, are developed for this purpose, under the restrictions governing all human thought. The Kantian obstacle arises, however, in the further Kantian claim that nonpsychological concepts developed as conceptual responses to sense impressions can meaningfully apply at most -to physical objects, and these are "nothing but" appearances (or appearings). Thus while this theory differs from I and II as previously interpreted, because it allows
[262]
us to understand theoretical concepts, it is importantly like both of those theories. Given indirect realism, none of these three theories allows the meaningful application of any empirical concepts of any sort to any external objects.
Neither a Lockean nor a Sellarsian can consistently maintain any of these three theories. It is clear, however, that he can justifiably reject the first two theories, because it is obvious that we do understand some theoretical concepts, contrary to these theories. The problem is that the requirements for the construction of concepts are too restrictive, and should be relaxed to allow these theories to explain our acquisition of these concepts. Once this is done, neither theory raises any obstacle for an indirect realist unless that crucial, additional Kantian stricture on the range of application of concepts is acceptable. But I find no reason to believe that concepts derived from, or developed in response to, what is directly experienced are meaningfully applicable at most to what is directly experienced. This claim is even more restrictive than the positivistic verifiability criterion of meaningfulness. That a sentence is in principle empirically verifiable does not imply that it and any concepts its terms express are meaningfully applicable only to what is directly experienced. Surely nothing more restrictive than meeting the verifiability principle should be required of empirical terms and concepts. Thus it is reasonable to hold that this Kantian obstacle to Sellarsian scientific realism can be surmounted.
A Sellarsian Theory of Concept Formation
It might seem at this point that it is incumbent on any indirect realist to propose or at least suggest plausible theories of concept formation and learning which his theory can accommodate. Sellars has partially met this challenge by sketching briefly a theory of concept formation that allows for the meaningfulness and learnability of terms that apply to theoretical entities. It is Kantian in an important respect because Sellars holds that certain empirical concepts are developed in our attempts to make conceptual responses to what is directly experienced, that is, sense impressions. One of his unique theses, however, is that only physical-object observation concepts are developed in this way (see PP, pp. 351-58). Thus no theoretical concepts of science and no psychological concepts, including sensation concepts, are developed in this way.
[263]
What results from this sort of conceptual response is what he calls "the manifest image," or picture, or conception of the world. It is the world so pictured, a world that does not exist according to this scientific realist, that corresponds to the Kantian intersubjective phenomenal world of physical objects.
The second unique feature of Sellars' theory is his thesis about how theoretical and psychological concepts come to be developed. He obviously must avoid that restrictive construction theory for these nonobservation concepts. He contrues both sorts of concepts, including sensation concepts, as theoretical, and claims that just as the role of models and analogies is crucial in the development of the theoretical concepts of science, so also is analogy basic in the development of our more ordinary psychological concepts (see PP, pp. 344-50 on models). As he says,
The interpretation of the framework of sense impressions as a theoretical framework suggests that the analogy between the attributes of impressions and the perceptible attributes of physical objects is but another case of the role of analogy in theoretical concept formation (SM, p. 21).
Is Sellars' thesis reasonable? Its present rudimentary and somewhat vague state makes it difficult to decide. I find this to present no problem, however, because I believe that it need not be decided in order to evaluate the Sellarsian theory, or any other version of indirect realism. My reason is this. Any adequate theory of concept formation and learning must account for our acquiring, or developing, or coming to understand, theoretical concepts of science. Such a theory, when fully specified, would be a scientific theory, and the ability to explain such acquisitions would be one test of its adequacy. Thus the theory would not raise an obstacle by prohibiting our understanding of these concepts.
There is also no reason to think that such a scientific explanatory theory would require or even justify any claim that limits the range of applicability of these concepts. The problem of what they apply to, if anything, corresponds to the problem of what theoretical terms refer to, if anything. As we saw when examining the argument for scientific instrumentalism based on the theoretician's dilemma, a question about the reference or nonreference of these terms is not itself a scientific question of how to interpret the theoretical
[264]
terms of a theory that explains some observed data. Because of this, successful explanation by a scientific theory generally does not, by itself, provide reason to interpret the theory or its subject matter realistically, or instrumentally, or in terms of what I call "Kantian scientific phenomenalism". This last is the theory that the theoretical terms of science refer at most to phenomena, or mere appearances. It should be noted, however, that, as will be discussed in Chapter 8, the problems that arise in attempting to interpret quantum theory realistically, do provide a strong argument for construing that theory and its subject matter instrumentally. This is primarily because of certain peculiarities in the observable effects of its purported subject matter, that is, subatomic particulars. But there is no reason to think that human learning of concepts or the development of concepts within a language have peculiarities that will force a scientific theory that explains these facts to restrict the range of applicability of certain of the concepts that are developed and learned.
Examination of Sellarsian Indirect Scientific Realism
We have found no reason to believe that some viable scientific theories of concept formation and learning will force an indirect realist to choose between Humean and Kantian skepticism. The way is clear, then, to discover whether an indirect realist with Lockean proclivities can reasonably avoid both these varieties of epistemological skepticism by devising a less restrictive epistemology than Locke's sensation-based theory. I shall approach this task by considering what is essential to a Sellarsian version of realism that is both indirect and scientific However, one of my primary aims is to discern whether what is essential to the Sellarsian scientific realism and epistemology is as reasonable when conjoined with direct realism as it is when conjoined with indirect realism. Because of this, I shall begin by characterizing the scientific realistic theory independently of the two subtheses of indirect realism.
Scientific Realism
It might seem that there is just one version of scientific realism, and it is roughly captured by Sellars' claim "that in the dimension of describing and explaining the world, science is the measure of all things, of what is that it is, and of what is not that is not" (SPR,
p. 230). [265] Sellars also claims that what explains best describes best, and science explains best. On this view, then, those theoretical terms of scientific theories that best explain what is observed provide the most accurate and comprehensive description of everything there is. To see that this is not the only version, nor, I believe, even Sellars' version, let us compare this thesis with one opposed to it, namely, scientific instrumentalism. As we saw in Chapter 4, this is the view that any theoretical term of science, T,' for which 'There is a ' entails 'There is a P,' where 'P' is a pure theoretical term, is merely a nonreferring symbolic device that is used to warrant inferences from observation premises to observation conclusions. It is clear that the thesis we derived from what Sellars said is much stronger than the denial of scientific instrumentalism. It is also true, however, that the mere denial is not strong enough to warrant being called even a minimal version of scientific realism. It is not enough that some theoretical terms are referring terms. Those terms, such as 'electron' and 'electron spin,' that are needed for the best scientific explanation of what is observed, must succeed in referring to particular objects and properties.
Even more is needed. It must turn out that the properties and constituents that are referred to are just those theoretical entities that the appropriate theoretical sentences would ascribe to the objects if the theoretical terms in the sentences should function referentially and descriptively. For example, if the theoretical term of psychoanalytic theory, 'repressed desire,' were needed in psychology to refer to some cause of abnormal behavior, but this cause was actually a supernatural demon that possesses people, then scientific realism would be incorrect. Also, if a Kantian scientific phenomenalism should be correct, then theoretical terms would refer to mere appearances (or appearings), and scientific realism would be false. Thus, unlike the way I have previously characterized the minimal position, we need something like the following definition:
Minimal Scientific Realism: Each nonsentient physical object and each sentient being has the constituents, properties, and relations ascribed to it by the theoretical scientific terms that are required for the best (nontranscribed) scientific explanations of its perceivable behavior and effects.11
[266]
Let us abbreviate this definition as stating that each nonsentient physical object and each sentient being is at least a scientific object.
It is clear that this position is not as strong as the one quoted from Sellars. There are two basic steps from this minimal view to the other, which I call "extreme" scientific realism. The first step is to adopt "moderate" scientific realism. This thesis, at first approximation, limits the constituents, properties, and relations of nonsentient physical objects to just those ascribed to them by science. For example, if no scientific theory that explains the behavior of such objects ascribes sensuous color to them, then none of the objects are colored. To be more accurate, however, some properties (perhaps, being aesthetically pleasing) and some relations (such as being owned by some person) should not be banned from these physical objects just because they are not necessary for scientific explanations of the behavior of the objects. Science is not the measure of all the properties and relations of all things. Let us say, however, that on this second species of scientific realism, science is the measure of any "a posteriori" property of or any relationship among nonsentient physical objects. By an "a posteriori" property or relationship I mean, roughly, one that it is reasonable to claim entities have or lack, only if there is some experiential evidence or theoretical scientific reason sufficient to justify the claim.
The second step is to claim that science is also the preceding sort of measure of all sentient beings, including persons. Thus if only physicalistic terms are needed to explain the perceivable behavior and effects of persons, then this extreme position, but not the more moderate position, implies that all persons are purely physical objects. We can state the moderate thesis, as follows:
Moderate Scientific Realism: Minimal scientific realism is true, and each nonsentient physical object has only the constituents and a posteriori properties and relationships with other nonsentient physical objects that are ascribed to it by the required theoretical scientific terms.
[267]
Let us abbreviate this as stating that all nonsentient physical objects are merely scientific objects, and all sentient beings are at least scientific objects. The corresponding abbreviation for the extreme view is:
Extreme Scientific Realism: Each nonsentient physical object and each sentient being (including each person) is merely a scientific object.
It is clear that if we were to locate Locke in this spectrum of positions, he would qualify as at least a moderate scientific realist. This is also true of Galileo, Newton, Descartes, and, today, Sellars. However, there is doubt about whether these men would be extreme scientific realists. It is not clear that ideas for Locke and Descartes, sensations of secondary qualities for Galileo, and sensa for Sellars are either introduced or justified solely on the grounds that they are required for some explanatory function of science. Indeed, for Locke and Descartes, the existence of ideas is required primarily because of their particular sensum-based epistemological
theories. That this particular reason for sensa is not one acceptable to an extreme scientific realist can be seen by realizing that no requirements of science necessitate such an epistemology. For example, based on the result of Chapter 2, a sensing-based epistemology would seem to be at least as acceptable to an extreme, scientific realist. Furthermore, the Sellarsian epistemology, which does not require sensations for a foundation, is surely compatible with what science requires. Interestingly, however, as we shall see, Sellars seems not to hold the extreme theory. And his reasons for this are, like those of Locke and Descartes, importantly epistemological. Indeed, in spite of his basic epistemological disagreement with both men, these reasons lead him to agree with them that there are sensa.
SELLARSIAN SCIENTIFIC REALISM
Sellars' basic difference from extreme scientific realism is best seen by concentrating on one sort of human behavior, namely, that verbal behavior that is said to express propositional attitudes about the world. Often when such overt behavior is in response to perceptual stimulation, as when some one utters, " There is a red and triangular object here,"
[268]
the verbal response is said to express a perceptual propositional attitude. Such overt verbal behavior must be explained as well as nonverbal behavior (see SSS, p. 400), and, according to Sellars,
perceptual propositional attitudes are "elements in a 'theory' designed to explain human behavior" (SSS, p. 398), in particular, verbal behavior. But such explanatory entities must themselves be explained. Consequently, it is not enough to explain merely observable behavior. The nonbehavioral facts that persons have perceptual propositional attitudes about their conceptual representations of the world around them, must also be explained. According to Sellars, then, a person being "under the visual impression that (visually taking it to be the case that) there is (or of there being) a red and rectangular physical object in front of one" (SM, p. 14) is used to explain certain overt behavior. But such "impressions that" must also be explained, says Sellars, if science is to provide a full and unified explanation and description of everything that occurs. According to him,
It is therefore crucial to my thesis to emphasize that sense impressions or raw feels are common sense theoretical constructs introduced to explain the occurrence not of white rat type discriminative behavior, but rather of perceptual propositional attitudes, and are therefore bound up with the explanations of why human language contains families of predicates having the logical properties of words for perceptible qualities and relations (IA, sec. 49).
It seems, then, that Sellars disagrees with someone such as Quine about what is required to explain human behavior, that is, he disagrees with Quine's implication that "correlative physical states" will be able to take over the explanatory roles of all mental entities.12 More specifically, he appears to claim that they will not take over the role of certain theoretical "inner" particulars, sensa, in the scientific explanation of propositional attitudes that are in turn used to explain certain verbal behavior.
It is not clear, however, why he should not agree with Quine that, although states of human beings that are purely physical [269]
cannot assume the explanatory roles of propositional attitudes in the present, intermediary stage of scientific development, they can do so once neurophysiology and other physical sciences develop more fully.13 If he did accept this, then he would slide into the "crude" physicalism he wishes to avoid (see SM, p. 22). But if he were an extreme scientific realist, and that theory taken with future scientific advances leads to "crude" physicalism, that is, materialism, then he should accept the materialistic consequences.
How might we best construe Sellars' resistance to a Quinean slide towards materialism? He might merely be making a prediction that science will never be able to uncover a physical correlate for each mental state, and so some mental states will always, as a matter of fact, be required (see SSS, pp. 397-99). His resistance, however, seems too strong to be explained in this way. A more reasonable construal is that he believes that no explanation of why someone has a particular conceptual representation of an observable object is satisfactory unless it includes reference to appropriate sensations. He says,
Even in normal cases there is the genuine question, 'Why does
the perceiver conceptually represent a red (blue, etc.) rectangular (circular, etc.) object in the presence of an object having
these qualities?' The answer would seem to require that all the
possible ways in which conceptual representations of colour and
shape can resemble and differ correspond to ways in which
their immediate non-conceptual occasions, which must surely
be construed as states of the perceiver, can resemble and differ.
Thus, these non-conceptual states must have characteristics
which, without being colours, are sufficiently analogous to
colour to enable these states to play this guiding role (SM, p. 18).
But, according to Sellars, sense impressions in common-sense explanations (the manifest image), and certain states involving sensa in scientific explanations, are the nonconceptual states that are appropriately analogous to color, that is, analogous regarding both their structural and nonstructural properties. Indeed, the concepts
[270] corresponding to these entities have developed analogously from observation concepts. Nevertheless, one thing is far from clear in this account. Why is it that this guiding function is to be performed by something analogous to observable properties in nonstructural ways, as Sellars seems to think? It is true that our manifest explanations and many explanations in psychology presently rely on entities analogous in these ways. For example, red sensa are nonstructurally analogous to red physical objects. But I find no reason at this early stage of theory development to rule out the thesis that the nonconceptual states, which in future scientific explanations will have this guiding role, will be merely structurally isomorphic to "all the possible ways in which conceptual representations can resemble and differ." These states, then, might well be purely physical states of brains, which like many other sorts of guidance systems need not guide by means of nonstructural resemblances. Indeed, as was mentioned when discussing Gibson's theory, the perceptual systems he proposes seem to have no need for sensations at all. I see no reason why such systems could not also be used, again without sensations, to explain perceptual propositional attitudes, and thereby certain observable verbal behavior.
There is one more reason I believe Sellars might have for resisting materialism. He can be interpreted as believing that materialism omits certain basic features of the world that any adequate ontological theory must include in its account of what there is. These are certain of the sensuous entities we seem to experience constantly. He says, "We must find a place in the world for color in the aesthetically interesting sense with its ultimate homogeneity" (SSS, p. 408). It may be objected that this is inconsistent with Sellars' often enunciated opposition to "the" given, because he seems to say here that homogeneous, sensuous color is given. However, he says that he rejects "the" given in one and only one sense. According to him,
To reject the myth of the given is not to commit oneself to the idea that empirical knowledge as it is now constituted has no rock bottom level of observation predicates proper. It is to commit oneself rather to the idea where even if it does have a rock bottom level, it is still in principle replaceably by another conceptual framework in which these predicates do
not, strictly speaking, occur (PP, p. 353).
[271]
Thus he can consistently maintain that color is given in some form or other, but in what form it actually occurs -- whether as properties of entities (such as physical objects, light, or sensa), or as internal specifications of sensings, or as something else -- is to be decided by what explains best scientifically. On this view, then, science must interpret sensuous features of the world, such as sounds, smells, and tastes; if cannot eliminate them or transform them so much that there remains nothing significantly like them. Such a Sellarsian can, then, eliminate sensuous color from the external world in a Lockean way, but if he does so he must locate something much like it in perceivers. In this respect, Sellars accepts one sort of given.14
While I believe the third reason for resisting materialism is the strongest because the species of "the" given on which it is based is quite reasonable, we need not decide which, if any, of these proposed reasons Sellars actually accepts. All three support the following construal of Sellarsian (and I hope Sellars' own) scientific realism:
Sellarsian Scientific Realism: Minimal scientific realism is true, and each nonsentient physical object and each sentient being has only the constituents, and a posteriori properties and relationships that are ascribed to it by the theoretical scientific terms that are required for the (nontranscribed) scientific theory that best explains its perceivable behavior and effects in a way that also explains and interprets the sensuous features of the world and the responses of perceivers to them.
SELLARSIAN INDIRECT REALISM AND EPISTEMOLOGY
In the preceding discussion there is some hint that Sellars is an indirect realist and that his epistemological theory differs importantly from Locke's sensation-based theory. It is true that he interprets the manifest
image or common-sense view of perception as a direct realism on the grounds that it understands perception as involving sense impressions construed as objectless states of perceivers, for example, red-sensings (see SPR, p. 100). He claims, however, that [272] the scientific image or description requires perception to involve inner, theoretical particulars, which he calls quot;sensa," the concepts for which are derived analogically from observation concepts. When discussing the scientific conceptual framework, he says that it is true "from the standpoint of this sophisticated framework that when a person sees that a physical object is red and triangular on the facing side, part of what is 'really' going on is that a red and triangular sensum exists where certain microtheoretically construed cortical processes are going on" (SPR, p. 103). Thus, given his scientific realism, he embraces the scientific framework as providing the best description of what there is, and so it is reasonable to interpret him to hold that indirect realism with sensa is one part of the best theory of perception and the external world. It should be noted that Sellars might claim that in the scientific image it is not correct to say that a perceiver senses a sensum. Rather there is some theoretical relationship between certain of the microparticles that are among the constituents of his brain and the sensum (see SSS, p. 428). This, however, does not imply that Sellars is not an indirect realist, because this relation is the theoretical counterpart in science of the sensing relationship. Strictly speaking, the definition of the second thesis of indirect realism should be amended to allow for such counterparts, but I shall ignore this complication because nothing essential to our discussion is lost and simplicity is gained.
Sellars' epistemological theory is not elaborated in detail, but what is essential for our purposes is captured, roughly, in his claim that what explains best describes best. This can be stated more precisely, I believe, as the principle that whatever statements are the most satisfactory as an explanation of perceivable phenomenaalso provide the most accurate and comprehensive description of the phenomena and of the entities explaining the phenomena. Given this somewhat plausible principle and Sellars' additional claim that a scientific explanation is the most satisfactory explanation of any perceivable phenomenon, we can justify the claim that the theoretical statements of science provide the most accurate and comprehensive description of what there is. Thus, contrary to Locke, we can justify a certain description of the external world without inferring it from premises solely about sensa and their relationships to their perceivers; and, contrary to Kant, "the real or 'noumenal' world which supports the 'world of appearances' is not a metaphysical [273] world of unknowable things in themselves, but simply the world as construed by scientific theory" (SPR, p. 97).
I do not wish to pursue here whether it is reasonable to adopt Sellars' basic epistemological principle, and his claim that science explains best. I have argued briefly elsewhere that both claims are clearly debatable.15 While scientific explanations are best for certain purposes, they do not seem best for other purposes. Which of these purposes, if any, is relevant to what is the best description? Indeed, why consider any sort of best explanation as sufficient for the best description? Surely the argument for scientific instrumentalism from the theoretician's dilemma opens that claim to doubt. But even if all arguments for instrumentalism are refuted, I believe that at most it is reasonable to hold the principle that the scientific theory that explains best is at least part of the most accurate and comprehensive description of what there is. That is, we are at most justified in accepting minimal scientific realism unless there are reasons to think, as Sellars believes, that if minimal scientific realism is correct, then minimal common-sense realism is mistaken. If this should be true, and if minimal scientific realism should be more reasonable than minimal common-sense realism, as we shall consider in Chapter 8, then there might be reason to accept at least moderate scientific realism and perhaps also the stronger Sellarsian principle.
For our purposes of discovering whether Sellarsian scientific realism combines at least as reasonably with direct as with indirect realism, we shall see that it does not matter which of the two principles a Sellarsian accepts. Let us, then, use the weaker, more reasonable principle in our final attempt to characterize Sellarsian indirect scientific realism with its accompanying epistemology made explicit. We have the following:
Sellarsian Indirect Scientific Realism (and epistemology):
- Indirect realism (with counterparts) is true, and
- Sellarsian scientific realism is true, and
- the statements that provide the best explanations of perceivable phenomena are part of the most accurate and comprehensive description of what there is, and [274]
- theoretical statements of science provide the best explanations of perceivable phenomena.
This theory contains everything central to Locke's theory, except the defeating sensum-based epistemology. And, on the initially quite plausible assumption that the Sellarsian epistemology is reasonable, we can draw the preliminary conclusion that Sellarsian indirect realism is more reasonable than Lockean indirect realism. Thus, if we can show that the last three defining clauses of the preceding characterization, which specify Sellarsian scientific realism and epistemology, provide no more reason to accept sensa and indirect realism than to accept some form of direct realism, we shall have some reason to conclude that there is a form of direct realism that is at least as reasonable as Lockean and Sellarsian indirect realism.
AN ARGUMENT FOR SELLARSIAN INDIRECT REALISM
I have elsewhere examined what I had found to be the most plausible Sellarsian attempts to show that features of Sellarsian scientific realism and epistemology make indirect realism more reasonable than direct realism. I believe that I showed that all of them fail, and so I shall not reconsider them here.16 However, Sellars replied to my first attempted refutations (see SSS), and, although I believe I adapted my arguments to avoid his replies, I would like to extract from his reply, as it illuminates other of his writings, what I now believe is the strongest argument to show that Sellarsian scientific realism provides reason to postulate sensa.
We have seen that the unique feature of Sellarsian scientific realism is its statement that there are sensuous features of the world which can be interpreted in different ways, but are not to be eliminated completely. The crucial issueis whether, given this version of scientific realism, one way of interpreting these features is more reasonable than all others. Let us make an assumption that is surely plausible: either some persons red-sense if the manifest image or description of the universe is correct, or there is some true scientific counterpart of this fact if the Sellarsian scientific image is correct. Furthermore, let us also assume that if there is to be the scientific image, and thus one complete scientific description of
[275]
what there is, then science must become unified in the sense that there is one complete scientific description of what there is which is to compete with the manifest or common-sense description. If no such unification occurs, then each separate scientific theory will present a partial description that will not be integrated into one overall picture. Consequently, the unified manifest image with scientific instrumentalism will prove to be more reasonable than partial scientific images with realism. According to Sellars, this unity of science requires that each description of objects that is true at some nonbasic level of science, such as the scientific counterpart in psychology of the description of a person red-sensing, must have a corresponding true description at the basic level of science. To accomodate this requirement, Sellars adopts a principle of ontological reduction, namely:
R. If an object is in a strict sense a system of objects, then every property of the object must consist in the fact that its constituents have such and such qualities and stand in such and such relations or, roughly, every property of a system of objects consists in properties of, and relations between its constituents (SPR, p. 26).
It is Sellars' claim that his reduction principle requires that the scientific counterpart of an event of a person red-sensing is reduced to, that is, consists in, an event of at least some of the scientifically basic constituents of the person individually having certain properties or relationships. And, in order that the sensuous features captured in the manifest description by 'red-sensing' not be eliminated by such reduction, some of these properties of basic entities must be scientific counterparts of sensuous manifest characteristics, such as redness (see SSS, pp. 413-14). Thus, according to this argument, we must, as required by the scientific image, postulate some scientifically basic entites as entities with such sensuous scientific theoretical properties. And, the most reasonable candidates for this role are sensa, rather than elementary particles of physics or scientific counterparts of Cartesian egos. Thus, given Sellarsian scientific realism, we should postulate sensa, and indirect realism is justified.
One way to lay out the details of the premises of this argument is as follows: [276]
- If Sellarsian scientific realism is correct, then there is a scientific counterpart of what in the manifest image is a person red-sensing.
- If Sellarsian scientific realism is correct, then the scientific counterpart of a person is a system of scientifically basic entities.
Therefore
- If Sellarsian scientific realism is correct, then some system, s, of scientifically basic entities, which is the counterpart of a person, has group property, G, which is the scientific counterpart of the property of red-sensing.
- If Sellarsian scientific realism is correct, then all scientific group properties are reducible, that is, a system having a scientific group property consists in certain scientifically basic constituents of the system individually having certain properties or relations.
Therefore
- If Sellarsian scientific realism is correct, then s having G (the scientific counterpart of red-sensing) consists in certain scientifically basic constituents of s individually having certain properties or relations.
- If Sellarsian scientific realism is correct and system, s, having group property, G, consists in basic constituents of s individually having certain properties or relations, then science should postulate that there are individual basic entities with sensuous properties.
- Any basic entity that science postulates to have a sensuous property is either an individual elementary particle, or a counterpart of a Cartesian ego, or a sensum.
Therefore
- If Sellarsian scientific realism is correct, then science should postulate that there are either elementary particles with sensuous properties, or counterparts of Cartesian egos, or sensa.
[277]
- It is more reasonable for science to postulate that there
are sensa than that there are counterparts of Cartesian egos or that elementary particles have sensuous scientific properties.
Therefore
- If Sellarsian scientific realism is correct, then science should postulate that there are sensa with sensuous properties, such as "red" sensa.
Premises (1) and (2) are clearly acceptable. Premise (2) is true because persons are sentient beings, and thus if Sellarsian scientific realism is true, they have just the constituents ascribed to them by the appropriate scientific theory. It might be objected, however, that (1) should not require there to be a scientific counterpart of a person red-sensing. As we have seen, however, Sellars insists on counterparts because, I believe, he thinks that even if 'red-sensing' should appear in scientific explanations, it is likely its scientific explanatory use would differ from its manifest explanatory use. And Sellars holds that difference in use implies difference in meaning. We need not, however, debate here about a criterion for difference in meaning. We need only allow that any entity is one of its own counterparts, and then nothing essential to any of the preceding premises depends on whether or not this Sellarsian view about meaning is correct. Indeed, nothing essential depends on whether the scientific image requires different entities from those of the manifest image.
It might also seem that premise (6) is debatable, but although it is surely not an entailment, it is justified by the requirement of the Sellarsian theory that the theoretical terms that describe best must include an adequate interpretation of the sensuous features of the world. Thus unified science should somehow include in its description of basic entities some terms that in some way ascribe some such features to the basic entities. And so science should postulate that some basic entities have certain sensuous properties, if the Sellarsian theory is correct.
OBJECTION: SCIENTIFIC REALISM WITHOUT REDUCTION
There remain premises (4), (7), and (9), none of which I find to be clearly acceptable. Indeed, I find (4) clearly unacceptable. The crucial problem is whether Sellarsian scientific realism can permit [278] irreducible group properties, contrary to (4). As previously stated, the Sellarsian theory does not entail such reduction, nor does it do so when the requirement is explicitly conjoined with it that science be unified so that there is one, unified scientific description of what there is. That is, the scientific image does not require the reduction of group properties even with the reasonable assumption that it is correct only if every macro-object is identical with a system of scientifically basic entities, each of which individually has only the properties ascribed to it in the basic science (cf. SSS, pp. 414-15). There is no reason to think that the scientific image requires the reduction of group properties to properties that are part of the subject matter of the basic science. At most it would seem to require that the basic science can explain, by means of correspondence rules that link individual basic entities with systems, why those systems have such group properties. Nor is there any reason to think that, even if all scientific properties must be included in the basic science, it could not be done by postulating properties in the basic science that apply to groups of basic entities but not to individual basic entities. For example, temperature might be postulated in the basic science as a property that groups of basic particulars have but no individual basic particulars have. Both of these ways to accommodate group properties clearly seem to be open to the Sellarsian theory. But since neither consists in the reduction of the properties, premise (4) is mistaken.
Sellars' reply at this point would seem to be to invoke R, his principle of ontological reduction, but, because of a basic ambiguity in R as stated, it is not clear it supports (4). It can be construed in either of two ways. First:
Ra. If an object is a system of objects, then the object having a property consists in some of the constituents of the object, either as a group or individually, having certain properties or relations.
Second:
Rb. If an object is a system of objects, then the object having a property consists in some constituents of the object individually having certain properties or relations.
It is clear that the Sellarsian scientific image requires at least Ra, but
[279]
it is also clear that Sellars needs Rb to justify (4). But what reason might there be to think a principle as strong as Rb is also required? Unfortunately, Sellars provides little to help answer this. At one point he says he accepts it without argument (SPR, p. 35). At another place he replies to my suggestion that he need not adopt such a strong principle for his version of scientific realism by saying that "the acceptance of the 'strong' principle of reducibility is what makes the 'version' mine" (SSS, p. 415). Surely something more is required to justify the use of Rb.
I think that we can uncover in Sellars' writings a better reason, but even this one does not justify Rb. Sellars holds ontological theses, distinct from his scientific realism, which he says "can be lumped together under the heading 'logical atomism' -- thus my acceptance of a strong principle of reducibility concerning the attributes of wholes, and 'Tractarian' approach to ontology (SSS, p. 396). It is, I believe, this "Tractarian" approach that is crucial and requires him to adopt Rb, indeed, an extremely strong interpretation of Rb. All nonbasic objects are systems of basic objects in the sense that they are logical constructions out of basic objects. Thus all predicates, such as 'temperature,' that truly apply only to scientifically nonbasic entities, whether theoretical, observational, or phenomenal predicates, are to be contextually defined by those predicates that truly apply to individual basic entities. Thus not only is Rb required, but the reduction of a group property, such as temperature, expressed in Rb, is to be construed in terms of the definition of the corresponding predicate by predicates applicable to individual basic entities. As I have argued elsewhere, however, it is clear by now that neither scientific realism, nor any competing ontological theory, should be encumbered with what there is good reason to think are the unsatisfiable and unreasonable demands of logical constructionism.17 Thus this, which I find to be the strongest defense of (4), fails, and so, with a Sellarsian having at least two other ways to accommodate group properties, we can reject (4) as too strong a requirement for Sellarsian (if not Sellars' own) scientific realism.
Having rejected (4), we need not consider premises (7) and (9). However, I shall examine in depth issues directly relevant to both
[280]
premises when I discuss in Chapter 8 whether Eddington's scientific table consisting of discrete imperceptible particles might be identical with a common-sense table having a continuous red surface. Regarding (7), I shall say here only that it is not clear its disjunction is exhaustive. Might not the basic entities that have sensuous properties be something like fields, rather than particles, egos, or sensa'? Or perhaps, contrary to (9) and Sellars' strong denial (see , SSS, p. 437), it might be most reasonable to claim that elementary particles themselves, when -- and seemingly only when -- they are in certain systems, come to be sensuously colored. Again, contrary to (9), it might be no less reasonable to postulate scientific counterparts of Cartesian egos as the bearers of sensuous properties, as it is to assume sensa. I see no reason to think that the former are scientifically less respectable than the latter. Thus, even if (4) were acceptable, a direct realist might well find ways to rebut the justification of indirect realism by way of Sellarsian scientific realism.
Conclusion about Sellarsian Indirect Realism
With the rejection of (4) and the resulting different ways for a Sellarsian to accommodate sensuous qualities without postulating sensa, we can conclude that Sellarsian scientific realism sans sensa is at least as reasonable as Sellars' own version with sensa. This, of course, is compatible with Sellars' version being as reasonable as Sellarsian direct realism, and, in fairness to Sellars, that may be all he wishes to establish. If premise (7) were amended to include all alternatives and (9) changed to express that the postulation of sensa is at least as reasonable as the other alternatives, then both would be quite plausible. Nevertheless, the objection to (4) would still remain, and so (10) modified to reflect the change in (9) would still be unjustified. Of course, (10) further amended so it refers to what Sellars calls "his own" version of scientific realism would be justified, because (4) with its antecedent referring to this version would be true. But then we would clearly not have the most reasonable species of indirect realism. Because of the unreasonably restrictive Tractarian requirements that Sellars places on ontological theories, his own version is, like Locke's theory, less reasonable than what I have called the "Sellarsian" version. As the preceding discussion indicates, there is reason to think that
[281]
Sellars' Tractarian requirement of logical construction neither will, nor need be met. All else being equal, a theory without these restrictions is more reasonable than one with them.
Preliminary Conclusion about Indirect Realism
We have examined what I have claimed to be the most reasonable form of scientific realism that accommodates an indirect realist such as Locke. We have seen that a Lockean requires at least moderate scientific realism because of his scientific conception of the external world. Whether it is more reasonable for him to hold, in addition, the extreme or rather the Sellarsian version of scientific realism is settled, I would argue, by a claim that seems to be as close to being irrefutable as any empirical claim. This is the claim that there are sensuous features of some entities of the world, and so these features should be interpreted rather than eliminated. Thus external constraints to force interpretation, rather than elimination, should be put on any version of scientific realism which itself limits the properties entities can have. These constraints, which are included in the Sellarsian theory but not in the extreme theory, falsify the extreme theory, if science becomes and remains behavioristic in subject matter, and physicalistic in method and theoretical postulations. And it is surely plausible that science will develop in this way. Thus, I conclude that the Sellarsian theory is more plausible than the extreme theory, and that if moderate scientific realism proves to be acceptable, then the Sellarsian theory results in the most reasonable form of indirect realism when it is combined with the plausible Sellarsian epistemology and the general thesis of indirect realism. However, we have seen that Sellarsian scientific realism when conjoined with the Sellarsian epistemology does not require or even justify the postulation of sensa. It combines at least as reasonably with the view that perceivers have the nonreduced group property of red-sensing as with the thesis that they sense red sensa. Thus, we can also conclude that if moderate scientific realism is acceptable, then Sellarsian direct realism is at least as reasonable as any form of indirect realism.
Reaching this last conclusion is an important step in the attempt to discover whether some form of direct realism is at least as reasonable as any metaphysical theory of perception and the
[282]
external world that requires sensa. It might seem that the one remaining step is to compare this Sellarsian version of direct realism with sensum phenomenalism on the grounds that we have found the most reasonable form of realism. This would be premature, however. There are other realistic theories we should examine first, and there are objections to all forms of scientific realism we have not refuted. We have seen that naive realism should be rejected because of its acquaintance view of perception. Common-sense realism, however, as the view of the nature of the external world contained in naive realism, might be combined with a different theory of perception to result in a more reasonable form of direct realism than the Sellarsian version. There will be extremely strong support for such a theory if, in the next chapter, we fail to find a way to neutralize the argument for scientific instrumentalism based on the theoretician's dilemma, or we find new objections to scientific realism sufficient to refute it. Another realistic theory we should also consider has so far been only mentioned. It is a form of nondirect realism which, unlike both direct and indirect realism, denies that we perceive external objects in any way. Its strongest version is Kantian realism. Let us consider it now.
Provisional Conclusion about Kantian Realism
In the Introduction, I discussed three versions of Kantian realism. One version reduces physical objects to appearances, or sensa. Two versions state that physical-object terms and concepts actually apply to nothing at all, but they are the terms and concepts that we develop, in accordance with inter subjectively applicable rules governing concept learning, to categorize the flux of our sensory experience in an understandable way. These two versions of the theory differ concerning whether these experiences are objectless sensings or include sensa. All three interpretations, however, agree that experiences are caused by external things-in-themselves, and that no physical objects are external objects. Indeed, for the last two theories, there are no physical objects at all, and thus there is nothing denoted by physical-object terms, whether they are observation terms or theoretical terms. Of course, there are noumenal objects, that is, external things-in-themselves, but, on all three versions, none of these objects are perceived in any way and none of their nonrelational properties are known.
[283]
One consequence of these three theories is what I called Kantian epistemological skepticism, which, as we have seen, receives independent support if -- and, it seems, only if -- it is reasonable to accept either the restrictive Kantian theory about the learning and range of applicability of concepts, or the similarly restrictive Lockean epistemology. But I have tried to indicate that such a Kantian theory of concepts is clearly overly restrictive, and I have said that the Sellarsian epistemology is at least as reasonable as the Lockean variety. It is true that I have not argued for this second claim. To do so would require a much deeper probing into epistemology than can be done here. Justification of this claim, then, must be put aside until another time. Yet, even without such justification, I believe that it is provisionally as reasonable to accept the less restrictive Sellarsian theory as to accept the Lockean theory. Consequently, on this provisionally plausible assumption that we can reject these Kantian and Lockean restrictions there is reason to reject Kantian nondirect realism with its skepticism for the Sellarsian claim that science uncovers precise information about things-in-themselves, provided, of course, that there are no reasons to reject scientific realism. We must, then, search for such reasons. We shall do so in the next chapter in the context of a discussion of reasons for and against the form of common-sense realism that I call "compatible" common-sense realism because it combines the common-sense theory with minimal scientific realism.
Notes
1 Locke, Essay Concerning Human Understanding, Book II, Ch. 8, sec. 8.
2 Ibid., Book II, Chap. 8, sec. 15.
3 Ibid., Book IV, Chap. 4, sec. 3.
4 Ibid., Book IV, Chap. 4, sec. 4.
5 Ibid., Book IV, Chap. 4, sec. 5.
6 Ibid., Book IV, Chap. 4, sec. 4.
7 See Berkeley, Principles, Dialogues, and Philosophical Correspondence, C. M. Turbayne, ed. (Indianapolis: Bobbs-Merrill, 1965), Dialogue III, 21.
8 See W. Stace, "The Refutation of Realism," Mind 43 (1934): 145-55.
9 Compare W. Sellars' discussion of abstractionism in SPR, pp. 41-50. Throughout this chapter, I shall use abbreviations for Sellars' works as follows:
IA: "The Identity Approach to the Mind-Body Problem," Review of Metaphysics 18 (1965); reprinted in PP, Chap. 15.
PP: Philosophical Perspectives (Springfield, 111.: Charles C. Thomas, 1967).
SM: Science and Metaphysics (London: Routledge and Kegan Paul, 1967).
SPR: Science, Perception, and Reality (New York: Humanities Press, 1963).
SSS: "Science, Sense Impressions and Sensa: A Reply to Cornman." Review of Metaphysics 23 (1971): 391-447.
10 The latter attempt is discussed in Chapter 3 and the former in my Materialism and Sensations, pp. 132-40.
11 For my previous characterization of three species of scientific realism, see Materialism and Sensations, p. 230. One change here is the reference to nontranscribed theories. There are no theoretical terms in a Craigian or Ramseyan version of a theory.
12 See W. V. Quine, Word and Object (Cambridge, Mass.: Massachusetts Institute of Technology Press, 1960), pp. 264-65; and The Ways of Paradox (New York: Random House, 1966), pp. 208-14.
13 By 'purely physical' I mean something like what Sellars means by 'physical2' See IA, sec. 45. For my definitions of 'physical property' and 'physical object,' see Materialism and Sensations, pp. 12 and 14. I reject Sellars' definition on pp. 10-11.
14 I have discussed a wide variety of givens in "Materialism and Some Myths about Some Givens", The Monist 56 (1972): 215-33.
15 See Materialism and Sensations, pp. 179-84.
16 Ibid., Chap. 7.
17 Ibid., pp. 243-45.
|