Andrew Chrucky, Critique of Wilfrid Sellars' Materialism, 1990

CHAPTER 10

ANIMAL REALISM

      The position I want to defend will be called Animal Realism. Since it has an analogy to theory construction, it could also be called Theoretical Realism. It could also be called Representative Realism, but I avoid calling it so because of the misleading connotations of the term 'representation' suggesting a copy theory. Animal Realism is the view that the 'cognitive' building blocks of a conceptual framework are, what have been variously called, 'ideas', 'sensa', 'sense data', 'sensibilia', 'qualia', 'essences', and probably in some uses 'stimuli'. Specifically my claim is that the Common Sense Framework is a rather rigid theory (in an extended sense) about sensa. From a structural point of view, Animal Realism has a common component with the positions which Sellars calls Hypothetico-Deductive Realism and New Phenomenalism{1}, and which Cornman calls Theoretical Phenomenalism.{2} The distinction between my position and these is that these positions ultimately take an instrumentalist view of theories, while I take a realistic one. I claim that not all theoretical entities are either eliminable or reducible to sensa. This entails that the postulates of a theory may name and describe reality.

      However I also want to say that the Common Sense Framework, though an adequate 'theory' for practical purposes, is not an adequate theory for truth. This is exactly Sellars' position except for the use of "theory" (Sellars would say here "conceptual framework"). The more adequate theory for describing and explaining what exists -- and here Sellars' and my view coincide -- is the conceptual framework of the Scientific Image. But immediately I will qualify my claim by saying that from my perspective the Scientific Image gives a more adequate description of the materials of the world, and a better non-teleological explanation of the world. This leaves out of consideration human artifacts which include such things as tables and social institutions -- all of these are understood through a knowledge of the purposes they serve; and are, therefore, teleologically constituted. But perhaps Sellars is conscious of this when he writes cryptically that the Scientific Image must be supplemented with practical discourse. I take this to imply that a teleological framework has to supplement the non-teleological framework of the Scientific Image. If that is what Sellars means. then there is a superficial coincidence between my view and Sellars'.

      The difference between us then becomes this. Sellars refuses to call the Common Sense Framework (his Manifest Image) a theory . He does this because the only possible candidate for the Manifest Image to be a theory about would be the manifold of sensa. But these are epistemically unavailable in Sellars' view prior to the possession of a public language; consequently, for Sellars, there is nothing for the Manifest Image to be a theory about.

I. CORRESPONDENCE RULES

      I find Sellars' discussion of correspondence rules to be a key to the relation of the Manifest and the Scientific Image, but unfortunately his discussion of correspondence rules is both sketchy and obscure. However the difficulty in understanding Sellars may be due to a failure on my part, but it also could be an obscurity on Sellars' part -- resulting from trying to be consistent with other principles of his philosophy. In any case, I am also surprised that Sellars' commentators on correspondence rules write as if Sellars' position were clear. When reading them I get no further understanding than reading Sellars himself because all they seem to be doing is repeating Sellars' obscure formulas without further explanation. This seems to be true of the expositions of Gary Gutting, Joseph Pitt, and C. F. Delaney.{3} The other type of commentary that does offer an interpretation of Sellars' correspondence rules is that of William Rottschaeffer.{4} The trouble with his interpretation is that it saddles Sellars with what seems to me to be a false consequence, which is

Sellars is, indeed, still burdened with an empiricist doctrine that what-is is what can be or is perceived. As a result in the long-run there are no inferred or unobservable entities in Sellars' ontology. There are no factual truths other that perceptual truths, and meaning is ultimately gained in perception. For in the long run there is only one mode of knowledge, perception.{5}

I believe that this is a misunderstanding not only of Sellars but of empiricism as well. As far as empiricism is concerned, it is true that for this view it is sufficient for saying that something exists if it is sensed, but this is not a necessary condition for existence. Recall that the necessary conditions for something being a material existent for Sellars is that the thing be in space and time and have causal properties.{6} These conditions do not mention perception. Now if it were true that that which was sensed did not have causal properties, then by the empiricist doctrine it would still be proper to say that the sensed thing exists, but it would be improper to say on this basis alone, by Sellars' definition, that a material things exists. If the sensed entity is taken not to have causal properties, it would then be more appropriate to become an Epiphenomenalist rather than an Extreme Materialist.

      However, the issue of what exists is a different problem from how knowledge of existents is obtained or tested. And it is true that, in some sense, both acquisition of and testing of our knowledge of what exists does require perception. But Rottschaefer has somehow scrambled the ontological and the epistemic distinctions, with the consequence that he attributes to Sellars Berkeley's idealistic formula: esse est percipi. And the root of the misunderstanding is, I suspect, his misunderstanding of Sellars' writings on correspondence rules.

      Thomas Russman has also misinterpreted Sellars by concluding that Sellars resolves the mind-body problem by "eliminating the mind or introspectible mental occurrences."{7} But I will not examine his reasoning for this paradoxical conclusion.

      I don't know if I understand Sellars' position on correspondence rules well enough, but I will try to go beyond rehashing Sellars, on the one hand; and I will try to avoid saddling Sellars' with an absurdity, on the other hand.

      Talk of correspondence rules is introduced to bridge the language of observation and the language of theory. Correspondence rules in this broader sense are rules correlating theoretical predicates with empirical predicates. Within this broader classification we can distinguish . . . substantive correspondence rules and methodological correspondence rules.{8}

      The very first thing to note is that a correspondence rule is some kind of correlation. Now a correlation, if it is not accidental, will either be a functional relation as in the Charles-Boyle Law which relates the three properties of gases: volume, pressure, and temperature; or the correlation may be a (broadly) causal law which relates events in time. And these two types of correlations may be expressed by either statistical or deterministic laws.

      In the case of methodological correspondence rules, for Sellars, there seems to be a correlation between a theoretical cause and an observable effect. The example Sellars gives is:

Spectroscope appropriately related to gas shows such and such lines <----> Atoms in region R are in such and such a state of excitation. ['<---->' stands for a correlation]

It is clear that Sellars intends us to understand that the state of atoms is region R is a cause of the appearance of the lines as seen or photographed through the spectroscope. Sellars points out that the details of the causal relation may not be known in detail.{9}

      Substantive correspondence rules correlate an empirical law with a theoretical law. This is illustrated by:

Temperature of gas in the region R is such and such <----> Mean kinetic energy of molecules in region R are such and such {10}10

Here the correlation is between the Charles-Boyle Law and the kinetic energy of molecules. Sellars does not explain himself, but I take it that temperature is operationally defined by a thermometer or a thermocouple reading. If the temperature is being measured by a thermometer, then we have a situation where the liquid in the thermometer as a result of being placed in a gas that has higher kinetic energy is causing the volume of the liquid in the thermometer to expand. And there is direct calibration of the relationship between the volume as registered by the thermometer and the temperature.

      I have a difficulty with this example because 'temperature' seems to be a theoretical term. When I look at a thermometer all I see is where the top of the liquid is relative to the scale on the glass of the thermometer. And on the basis of a theory of the instrument I infer that whatever the thermometer was placed in had such and such a temperature. But the word 'temperature' is here playing the part of a variable which stands for the cause of the thermometer acting as it does. And, of course, I know also what this unknown cause does to me -- the higher the temperature the more uncomfortable I get, and at some point in some circumstances, it causes me to sweat, or to feel pain, or to be burned. Now if a substantive correspondence rule is meant to correlate observational and theoretical predicates, and this is an example of this, then the example is a poor one because what it does is correlate two theoretical predicates.

      However my interpretation is trading on an ambiguity in the term 'observation' as is pointed out by R. Carnap

A philosopher would not consider a temperature of, perhaps, 80 degrees centigrade, or a weight of 93 1/2 pounds, an observable because there is no direct sensory perception of such magnitudes. To a physicist, both are observables because they can be measured in an extremely simple way.{11}

Sellars is mindful of this ambiguity when he writes that "it is not absurd to speak of observing viruses and protein molecules through an appropriately constructed electron microscope."{12} However if this use of 'observation' is not restricted in some way, then any perception of some symptom or effect of some event can be construed as seeing the event itself. For example, a person, by this usage, could say that he sees a fire because he has spotted smoke; or that he sees an airplane because he sees something in the sky which looks like a dark speck.

      Regardless of the use of 'observation' to include both 'primary' and 'secondary epistemic seeing' (as Dretske calls them){13}, Sellars also distinguishes a subset of substantive correspondence rules -- inter-theory correspondence rules -- which relate theoretical laws with other theoretical laws, e.g., chemical and physical laws. I take it the correspondence is between a molecule in chemistry and the constituent atoms and/or subatomic particles of physics. So if the term 'observation' is taken in the narrow sense, as I have, then the temperature/kinetic energy case is an example of this kind of theoretical substantive correspondence rule. But the intention in introducing substantive correspondence rules is to correlate the objects of the observational domain with the objects of the theoretical domain. And although I believe the above example fails to illustrate this for the narrow sense of 'observation', identifying water with H(2)O, however, does illustrate it. Substantive correspondence rules are then interpreted by Sellars as formulas for a possible replacement of the observational language by a theoretical one. He thinks it is possible to make "observational predicates into definitional abbreviations of complex theoretical locutions."{14} His prescription applies to 'water' as an abbreviation for 'H(2)O';{15} and even more so it applies to 'aspirin' as an abbreviation for 'acetylsalicylic acid', which in turn is an abbreviation for the formula 'CH(3)COOC(6)H(4)COOH'. Sellars goes on to say that "substantive correspondence rules are anticipations of definitions," and the "only alternative to this conception in the instrumentalist conception."{16} He does not prescribe adopting these definitions yet -- not because it is impossible to do so, but because science is still developing; which means that better formulas are still possible. So his recommendation is to wait:

substantive correspondence rules are anticipations of definitions which it would be inappropriate to implement in developing science, but the implementation of which in an ideal state of scientific knowledge would be the achieving of a unified vision of the world in which the methodologically important dualism of observation and theoretical framework would be transcended, and the world of theory and the world of observation would be one.{17}

As far as I am aware, saying 'H(2)O' or 'sodium chloride' instead of 'water' and 'salt', respectively, is not only a possible linguistic practice, but it is one which is used now to some extent, and it seems to be appropriate in technical contexts, such as in a chemistry lab. And as far as I can see, this is all that Sellars could possibly mean by viewing substantive rules as candidate redefinitions of observational terms. Replaceable observation terms must be restricted to talk about the materials which organisms and artifacts are composed of. It cannot be a language which can be substituted for artifacts such as cars and tables -- to say nothing of social institutions. The language of artifacts is tied essentially to functions and human purposes. Another way of saying this is that the language of artifacts involves practical talk. And practical talk is for Sellars, as was mentioned, something that has to be added -- not substituted for -- theoretical scientific talk.

      Given this construal of what Sellars means by the replaceability of theoretical language for the observational language, this leaves out of account the methodological correspondence rules as rules for inferring the presence of theoretical entities. In other words substantial correspondence rules allow for replacement of a segment of observational language, while methodological correspondence rules sanction inference to the existence of theoretical entities.

      Given this reading of correspondence rules, very little follows about actual or possible replacement of the observational language. On my reading, only the materials of the Manifest Image are subject to replacement. But everything else remains intact. If Sellars intends to claim more than this, then he hasn't made it clear to me what he does mean.

      Let me ask the following question. What is the relation between a substantive and a methodological correspondence rule? And it strikes me that the two can be made equivalent. This is to say that for every observation of a physical thing, there is some causal explanation which a methodological correspondence rule will express; and for every methodological correspondence rule there is some observation (in the broad sense) that will be about the cause.

      However, the core issue about observations is being evaded by talk of correspondence rules while using the term 'observation' in this theory-laden sense. There are two more basic uses of the term 'observation' than this. The first more basic sense is the case of 'primary epistemic seeing' in which we simply report seeing a physical object as having such and such a primary or secondary quality (in Locke's sense). And the other kind of observation is the rock bottom sense of 'observation' in which, as Sellars would put it, I see a volume of pink stuff. When the question is raised whether observational reports can be replaced, I think of these two types of observations, and not the cases of 'secondary epistemic seeing' which Sellars seems to be considering. Now, although theoretically it is possible that we could react in some way differently to stimuli or sensa other than by seeing physical objects, if that happened such perception would be disastrous for survival. So from the evolutionary perspective I don't think it is possible to perceive the world other than composed of physical objects. This applies, I believe, to all higher animals. As concerns the perceptions of persons, I would say that Kant's transcendental argument for the possibility of objective experiences requires that we perceive physical objects. As to the perception of sense data, which are no more than Sellars' volumes of colored stuff, Sellars himself talks about the concepts of 'occurrent sensible properties' as 'constant'.{18}

      My strongest reason for thinking that our perception of physical objects cannot be altered is that I believe our concept of a physical object is supervenient on an rs-concept of a physical object. And rs-concepts are involuntary products determined by an innate and environmental interplay.

      But if what I claim about the perception of physical objects is the case, then there is no way to alter our rs-concept of a physical object. This does not mean that we cannot alter the conceptual accretions to our rs-concept of a physical object. But Sellars writes that (i) we can alter or eliminate our conceptual framework:

According to the view I am proposing, correspondence rules would appear in the material mode as statements to the effect that the objects of the observational framework do not really exist -- there are no such things. They envisage the abandonment of a sense and its denotation.{19}

. . . one framework is, with appropriate adjustments in the larger context, replaceable by the other -- eliminable in favor of the other. The replacement would be justified by the greater explanatory power of the new framework.{20}

Specifically, (ii) we can alter our conceptual framework by a conceptual retraining. I find both these claims difficult to understand, unless the scope of his discussion covers only the sorts of cases I have considered above. So let me get to the root of the difficulty. The difficulty can be expressed by the problem of locating colors in the scheme of things.{21} Colors are, on Sellars final view, identified with sensa. And the explanation for why we think we see colors as properties of physical things is that we, so to say, superimpose sensa on physical objects. The result is a visual illusion.

physical objects aren't really colored; colors exist only in the perceiver', and that 'to see that he facing surface of a physical object is red and triangular is to mistake a red and triangular sense content for a physical object with a red and triangular facing side.

Sellars is correct in describing this as a mistake for an ordinary person who would, if asked to explain where he thinks the color of a red apple really is, reply 'out there where the apple is'. But there need be no mistake for someone who is aware that he is suffering an illusion. Such a person may say, for example, while looking at a drawing of the Muller-Lyer illusion, "The lines do not look parallel, but I know they are." He has made no mistake, despite seeing an illusion. By parity of reasoning, in the case of colors, a person may say, "the red looks like it is where the apple is, but I know that cannot be the case". And no mistake has been made.

      Sellars is claiming that the Manifest Image is false simply because in it secondary qualities like colors are attributed to physical objects. And from a theoretical perspective we know that physical objects are out there and colors are somewhere in the region of our brains. So that the superimposition or supervenience of colors on physical objects is a miscategorization. This is clear. And I accept this. The result is that we live in a sort of illusion -- though an illusion that serves well our practical purposes. I also take it that Sellars thinks that this illusion is somehow a product of linguistic training. And if it is then it can be replaced by a different linguistic training -- a training to be specified, I take it, by correspondence rules.

      I disagree. I don't see how a recategorization or reconceptualization can get rid of the illusion. I don't think that it is the case that our illusion is the result of linguistic training; it is the result of a pre-linguistic and innately guided response to the environment which results in an animal representational system. Sellars anticipates something like this move and calls it

the bastard notion of a pre-conceptual awareness of items as being red and triangular which is supposed to be the language independent foundation on which the meaningfulness of language rests.{22}

I do interpose a pre-conceptual representational system, but it does not explain the conceptuality of a conceptual system (as Sellars suggests), it rather imposes a core to conceptualization, but otherwise leaves conceptualization unrestricted as to further -- shall I say -- overlaying of concepts.

      Sellars at this point in his philosophical development had apparently not made the reflective distinction between a representative system and language (as he does in "Mental Events"); or if he did, he made nothing of it -- so that my view is not entertained at all.

II. THE GAP: SEEING PHYSICAL OBJECTS

      It is significant that Sellars assumes the existence of a Behaviorese language for the mythical Ryleans, and does not try to explain how it was possible for them to acquire it. This approach hides the existence of a problem. The problem is that Behaviorese takes it for granted that people can recognize physical objects. But how is it possible to recognize them? Sellars' answer is that a necessary condition is the possession of a concept of a physical object. But let's push it one step further by asking: How is it possible to acquire the concept of a physical object in the first place?

      The question I am raising is really the question: How is language possible? And Sellars' broad answer is: by a process of natural and social evolution.{23} But more specifically, what would such a process require? His answer here is that one first has "To learn pattern governed behavior" by being "conditioned to arrange perceptible elements into patterns and to form these, in turn, into more complex patterns and sequences of patterns."{24} What are we to make of this? What are these "perceptible elements"? Does he really mean "sensible elements"? And is the learner being passively impressed, or is the learner actively selecting or constructing a structure? And if there is an evolutionary process at work, what are the species specific capabilities or tendencies that are at work?

      To get a handle on an appropriate approach, let us ask ourselves what Sellars' answer could possibly be from the perspective of his Scientific Image. Sellarsian theory of Methodological Behaviorism must surely be a part of his Scientific Image. This is to say that a sophisticated behaviorism will take stock of the best ontology available to science, and use it to provide an account of learning about physical objects. Sellars introduces 'sensa' as theoretical entities to explain (conceptual) perception. They are, however, introduced as non-conceptual states, and are not the objects of sense cognition.

It is, indeed, 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 micro-theoretically construed cortical processes are going on.{25}

      Assuming that this is the case, my question is: If sensa constitute the elements of the sensorial field, and all empirical knowledge is somehow causally grounded on them, still how is it possible to learn Behaviorese? As far as I know, Sellars never addresses himself to this question. At best we can speculate on a possible answer.

      So how can we breach the gap between the ontology of the Process Ontology, which is a Heracleitean flux of absolute processes, and the Behaviorese of the Rylean language? But the problem is even more basic. We attribute to animals the ability to rs-see physical things. How do we explain this? The gap must be filled by some theory of Representational Systems, RS.

      Learning (unconsciously) about physical objects, for Sellars, consists minimally of patterning sensa by 'logical construction' in such ways as schematically specified by the 'logic of complex particulars' of the Process Ontology.{26} But if animals are credited with rs-seeing physical objects, either they too use 'logical construction' on sensa, or something else -- let's call it Humean (hypothesis) construction by analogy with Humean inferences. I will assume the latter. So if animals are able to rs-see physical objects, this suggests that our concept of a physical object is a sophisticated supervenience on an rs-concept. But an rs-concept cannot be a logical construction since animals, by assumption, lack a logical vocabulary. So there must be a more sophisticated construction by human beings which employs logical operators. But if this logical construction of a physical object out of sensa is introduced as a theory to explain perception, then Sellars is committed to the postulation of not only rs-concepts, but more sophisticated ur-concepts to pre-linguistic people, prior to the acquisition of a linguistic concept of a physical object. The upshot is that Sellars seems to be committed to an innate mechanism of perceptual structuring in terms of ur-concepts or rs-concepts independently of language. I don't see how Sellars can avoid these implication.

      This outcome dovetails with the position of Richard Gregory (who cites Helmholtz as his predecessor). His position is that perception is an unconscious inference process that takes sensorial stimulations as data for the construction of a perceptual 'hypothesis.'{27} This is also G. Harman's position.{28} And it seems to have been also the position of Paul Feyerabend -- a position which Sellars criticized in "Scientific Realism or Irenic Instrumentalism: A Critique of Nagel and Feyerabend on Theoretical Explanation." {29} The claim of these writers is that stimuli or sense data can be viewed as data requiring explanation, and perception is the hypothesis which explains the data.

      The main objections to viewing perceptions as hypotheses is that there is a disanalogy between this use and the normal or paradigm use of the word 'hypothesis'. In the paradigm use, a 'hypothesis' is a statement which has a semantical relation to other statements which express awareness of data and evidence. This is to say that 'data', 'evidence', and 'hypotheses' are in the paradigm case all consciously entertained epistemic statements. Also, there is this disanalogy. If a hypothesis is disconfirmed it can be withdrawn; whereas there is no analogue for withdrawing a perception on the basis of disconfirming evidence. The conclusion reached, then, is that a perception cannot be a 'hypothesis' in the ordinary use of the term.

      But it can be replied that perception is not literally a conscious hypothesis, but something analogous which is sufficiently similar to warrant being called a 'hypothesis' in an extended sense. So what would be analogous if perceptions are viewed as hypotheses? The answer is that there are sense data and other perceptions to act as data and evidence analogues, and there is an analogous unconscious process of data recognition, hypothesis formation, and anticipation of new data as evidence. Viewing perception as hypothesis formation helps to explain how seeing as switches occur when viewing ambiguous figures such as, for example, Wittgenstein's drawing of a duck-rabbit. It also helps to explain how visual illusions are possible.{30}

      Sellars would caution that such a use of 'hypothesis', which is an analogical extension of our ordinary term, requires a commentary as to which respects perception is like a hypothesis and in which respects it is not. But such a commentary has been provided above, so what would be further objections to calling perceptions hypotheses?

      The objection to viewing perceptions as hypotheses may be the fact that such a thesis requires the positing of unconscious mental processes. But this is compatible with the Sellarsian theory of concept formation which requires a pre-conceptual period of conditioned learning in which observational sentences are paired with sensorial stimuli, which occurs at an unconscious level. Furthermore, if we reflect on how it is possible to distinguish items within the manifold of sense, a distinction must be made between relevant and irrelevant stimuli, and within the relevant between the essential and accidental ones. The relevant-irrelevant distinction may be identified with the Gestaltist foreground-background distinction, or an umbra-penumbra distinction. And then within the foreground another distinction can be made between essential and accidental stimuli. If these distinctions are not present, then it is inconceivable, for example, how for a perceiver a tree can be distinguished from the forest, and how a maple tree can be distinguished from an oak. In other words, a behavioristic theory must posit stimuli discrimination and stimuli generalization. The results of these innate mechanisms is a perceptual response. One reason for calling this response a hypothesis is, I take it, because it is malleable. As a result of more or new data perception changes. So the analogy between perceptions and hypotheses is that both are a function of new data. All this occurring on the unconscious level.

      Sellars provides further clues to his position in "Scientific Realism or Irenic Instrumentalism: A Critique of Nagel and Feyerabend on Theoretical Explanation." While commenting on a paper by Paul Feyerabend, he categorically says that the Manifest Image is not a theory:

And while I agree with Feyerabend that the conceptual framework of "common sense" is, in the last analysis, false, I find it confusing . . . to speak of this framework as a false theory .{31}

He adds, "the conceptual framework of common sense has no external subject-matter and is not, therefore, in the relevant sense a theory of anything."{32} He goes on to ask:

What would the conceptual framework of common sense be a theory of ? [and answers] A trivial answer is, of course, available, namely, 'Common sense objects and events. ' But this would be like answering the question 'What is atomic theory a theory of?' by saying 'Atoms. ' A false answer is also available, namely, 'Sense Impressions'. {33}

I see no reason for Sellars to object to this as a causal theory of perception. Commenting on Feyerabend's thesis that "the framework of physical objects is a theory with respect to sense impressions," Sellars writes: "In this sense, though not in the ordinary sense, the framework of physical objects is a theory with respect to sense impressions."{34} His only reservation is this: "This conceptual framework, though brought into play by sense impressions is not, at least primarily, about sense impressions."{35} But this is no objection. After all a microtheory that posits electrons to explain oscilloscope patterns is not primarily about the illumination of the oscilloscope -- that would be explained by a theory of the oscilloscope -- it is rather a theory about the patterns of illumination. By analogy, the perceptual hypothesis would not be primarily about sense data but about the patterns of sense data.

      The fundamental difference between Sellars' position and mine goes back to the issue of rs-concepts. Does a nonconceptual Representational System allow for the construction of a theory of the common sense framework? The consideration of this topic is undertaken in Sellars' essay "Phenomenalism". Since I am not defending Phenomenalism and am in agreement with Sellars' basic objections against it, the only part of this essay that concerns me here is his attempt to refute the initial argument for, what he calls, New Phenomenalism.

      His counterargument is:

(A)

If the relation of sensa to the common sense framework is analogous to the relation of observations to a scientific theory, then since the necessary condition for the latter is the availability of inductive (observational) generalizations, the necessary condition for the former is the availability of inductive generalizations of sensa.

(B)

But such inductive generalizations about sensa are not available without the use of the language of physical things.

(C)

Therefore, the relation of sensa to the common sense framework is not analogous to the relation of observations to a scientific theory.

      The argument is valid. But it relies on the assumption that the premises and theoretical conclusion are to be formulated in a conventional language. Given this assumption, I have no quarrel with Sellars' position.

      But Sellars is operating on a sophisticated level of communication, where it is appropriate to locate the object of reference in objective space. And he is thinking of securing this reference relative to a speaker. So he is assuming here that for objective generalizations and communication self-reference is required in discriminating and organizing sense contents. I am, however, concerned with a less sophisticated level of awareness, which we may share with animals, and which makes language and communication possible in the first place. It is a level at which discriminations occur not only of sensory qualities but also of groupings of sensory qualities. Quine refers to this as "qualitative spacing", and recognizes its innate status: "Needed as they are for all learning, these distinctive spacings cannot themselves all be learned; some must be innate."{36} Such a postulation of discrimination and grouping of sensorial qualities is a necessary condition for understanding how animals can get about in their environment and learn anything at all.

      A Behavioristic theory assumes the innate capacity for forming stimulus associations and for stimulus generalization, and this is the type of generalization that suffices for my purposes. In addition I require an innate capacity for something like theory construction, and again a Behavioristic theory must assume the capacity of animals to respond to physical objects (or at least to sets of complex Gestalt phenomena) and not merely as phenomenalism would have it to atomic units. A 'theory' of physical things is, from this primitive perspective, a bundle theory in which qualities more or less coalesce, and in which such bundles interconnect. All of this is had, I surmise, at the animal level through a Representative System.

      Sellars has an argument for premise (B). It is, however, couched on the conceptual level. After I state it I will transpose it to the Representational System level. The argument is this:

(1) Either the inductive generalizations about sensa are accidentally autobiographical (A-generalizations) or they are essentially autobiographical (E-generalizations).

(2) If they are A-generalizations, then the generalization:

Whenever (or for the most part whenever) I have such and such a pattern of sense contents, I have a sense content of the kind in question remains true if 'anybody' is substituted for "I".

(3) If they are E-generalizations, then the above generalization does not remain true if 'anybody' is substituted for "I".

(4) If they are A-generalizations, then they must include a reference to physical circumstances.

(5) If they include reference to physical circumstances, then appeal to A-generalizations begs the question.

(6) Therefore, only E-generalizations are available.

      Let me first ask what is the point of this distinction between A- and E-generalizations? Although Sellars is not quite explicit on this, the way it makes sense for me is to interpret it along the following lines. The E-generalization is meant to apply to those cases where something is abnormal either about me or my circumstances. For example, suppose I have jaundice, then my visual generalizations about what I perceive may have many descriptions of things appearing yellow. In that case, although it may very well be the case that something yellow will be followed by something else being yellow for me, I cannot universalize my experience to be true for everyone; though I could generalize that this will be true for the most part for anyone who has jaundice. However such a generalization would require relevant public information. An A-generalization, by contrast, is made on the assumption that I and my circumstances are normal, and that what is true for me will be true for any normal person in normal circumstances. So, in any case, both generalizations presuppose knowledge of the public circumstances, and both kind of generalizations are implicitly relativized to states of persons and their circumstances.

      Seen in this light, Sellars point is that any experiential generalization is determined by various factors about the person and his circumstances, and that a full understanding of why experiences come in the patterns that they do depends on understanding these conditioning public factors.

      Any Reductive or Analytical Phenomenalism suffers shipwreck against this argument. The ability to predict which sensa will be consequent on which prior sensa requires a mention of the circumstances in a physicalistic language. Phenomenalism thus must either beg the question by appealing to a physicalist language which it claims can be eliminated; or, if it does not make such an appeal, it will fail to provide reliable generalizations.

      I agree with all of this. But it is beside the point. The issue of whether "anyone" can be substituted for "I" does not arise for me because I am working at a level at which this distinction does not even have to be present.

      Sellars, in his examination of Phenomenalism, interprets the claim that there are possible sense contents as involving a terminal statement of the kind: 'If I (who am in state T) do action A in circumstance C, then I may have sense content S'. This is too sophisticated. Certainly actions and circumstances will have an effect on the passing show, but abstracting the actions and circumstances, the passing show will still have some modicum of regularity. Of course taking into account actions and circumstances enhances the understanding of the regularity, but it is not essential to the discernment of minimal regularity. The Phenomenalist step need not be concerned with the conditions of the passing show, he need only discern fleeting or relatively stable patterns, and on this basis to anticipate further features.

      Notice that Sellars does not deny the presence of generalizations. Without a knowledge of how circumstances determine the presence of experiences, the generalizations may not be very reliable. Suppose some person, S, is looking at a rolling red ball. The experience will (if it is objective) have some kind of regularity about it, despite unfavorable circumstances. Suppose that S is blinking or moving his head, then the experience will be of something disappearing and appearing or jumping from one area of one's field of vision to another. Granted. Still there is regularity. This point was made by James Cornman in his essay "Theoretical Phenomenalism."{37} Although the so-called E-generalizations are not strictly universal generalizations, it suffices that they can be probabilistic generalizations. The phenomenology of this kind of awareness is captured by Berkeley's and Russell's "bundle theory" of physical objects. What perhaps should be said is that physical objects are seen as bundles of sense qualities remembered, had, or anticipated. Initially, we may conjecture, these bundles have no determined regularity about their exact make-up: the bundles come in family-resemblances, and their interconnections have, if we use our perspective to describe them, at best, a statistical regularity.

      Sellars response would be that all of this begs the question. The point at issue is whether it makes sense to speak about concepts without presupposing a battery of concepts which include those of physical objects. Sellars' attack rests on the assumption that we are working with conceptual cognitions, and that conceptual cognitions entail a language, and that a language can be learned only in a social context on the basis of a behavioristic learning theory.

      Since I do not want to dispute this, I paraphrase the whole discussion to the level of rs-concepts. I am not claiming that physical objects are reducible to actual and possible sense contents, I am claiming that we have an rs-concept of a physical objects which is functioning analogously to a theoretical concept.

      The question may then be raised: How is such a hypothesis constructed? One response is that this is an irrelevant question. How theoretical hypotheses in science arise is varied and at times quite mysterious. The claim is sometimes made that we 'infer' a hypothesis, but this is not quite right, or if the word 'infer' is to be retained, then different types of inferences have to be distinguished. It seems to me that theoretical hypothesis formation is akin to poetry in relying on an inspiration of an appropriate metaphor or analogy. Philosophers of science, including Sellars, talk about the heuristic value of models for theories. In other words, theoretical hypothesis formation is, if we are going to retain the word 'inference', inference to a model. But instead of 'inference', the words 'insight', 'intuition', 'illumination' seem more appropriate. Some cognitive psychologists describe pre-linguistic awareness as hypothesis formation in this sense. When, for example, R. Gregory talks about perception as hypothesis formation, he too is using 'hypothesis' in this sense. I suspect that a Gestalt is also a hypothesis in this sense. Now how are hypotheses of this sort constructed? I don't think anyone knows. If it were known, this would provide foundations for a logic or algorithm for discovery. But, for all I know, discovery is ultimately, as K. Popper contends, a trial and error sort of affair. But the ignorance of the manner of genesis of hypotheses does not preclude hypotheses from being justified for acceptance. So in the case of talking about the formation of physical object 'rs-hypotheses' or rs-concepts of physical objects, ignorance of their genesis is just as mysterious as is the genesis of any theoretical hypothesis.

      Another response, one which may or may not be implicit in Sellars' position, is that the genesis of an rs-hypothesis is impossible -- which is quite a different claim than saying it is mysterious.

III. THE COMPUTER ANALOGY

      I will approach this issue of rs-hypotheses of physical objects from the perspective of computer simulation. Sellars occasionally uses the computer as a model of language behavior.{38} He provides computer analogues for language entry, intra-language, and language-departure transitions, and he also provides a computer analogue for induction. We know that the difference between the conceptual and the non-conceptual is the availability or the unavailability of logical words for inferences, which make acting in accordance with rules possible. And once rules are mentioned we have to introduce a meta-language which duplicates the observation-inference-action transitions on its level. Also we must include various 'auxiliary positions'. He has also provided for names via a four-dimensional coordinate language, for spatial and temporal demonstrative expressions, and for various predicates in his computer analogy.

      The computer analogy, however, is inadequately formulated by Sellars to account for the perception of physical objects; and he has omitted from his computer analogy the distinction between two roles of logic. On the one hand the hardware of the computer is such that it has built-in logic gates. Without these there would be no computer. On the other hand there are various software programs, appropriately called 'languages', which contain their own logics, for example, the language BASIC. This distinction suggests two different ways that logic can play a role in behavior. On the one hand, the logic can exist without any language, and have a role in the unconscious and involuntary integration of sensations (sensa, qualia) resulting in the perception of physical objects. On the other hand, a logic (of a program language) can exist in a language, and be available to conscious and voluntary use.

      Is this acceptable to Sellars? I don't know. Some of his claims are open to interpretation. Sellars distinguishes Humean inference from Aristotelian inference, and Humean inference can be viewed as weakened or truncated Aristotelian inference. But is there theoretical room for something in between -- perhaps a less truncated form of Aristotelian inference? Why not something like Intuitionistic Logic? As I understand Sellars position, he grants animals at best simple associative mechanisms, such as associating a signal with food. But perhaps animal behavior could be understood in terms of acting in conformity to some specified logic (as opposed to being rule governed). This would enable an animal to ape much more human conceptual abilities.

IV. ORDER OF BEING AND ORDER OF KNOWING

      Sellars often makes use of the Aristotelian distinction between what exists and what is known to us. And he applies this distinction to the relation of thoughts to knowledge. He states that thoughts exist without being known; knowledge of them requires the conceptual apparatus of a language. If this is all that Sellars claims then there would be no essential controversy with innatism. But this formulation blurs the distinction between conceptual thoughts and rs-thoughts. It is clear that Sellars would grant human beings rs-thoughts independent of language, but it is not equally clear that he would not grant them conceptual thoughts independently of language. The claim that thoughts are ontologically prior to language may mean no more than that after a language is learned, a person acquires conceptual thoughts, and these as internalized are prior in the order of being from then on. On this interpretation we have to assume that the Ryleans in learning a public language at the same time (concomitantly) are learning to think conceptually. From this it would follow that there is no conceptual thinking prior to the learning of a language. The distinction between an ontological priority of thoughts becomes, at best, the claim that thoughts (qua neural patterns) are causally (though not genetically) prior to overt utterings. It becomes the claim that such neural patterns embody the dispositions to overt utterings. What Sellars would not be claiming, on this interpretation, is that in the order of being conceptual thoughts exist prior to the acquisition of a public language.

      From this interpretation many of Sellars' claims follow: his claim that "the individual as a conceptual thinker is essentially a member of a group"{39} ; and that "It is no accident that one learns to think in the very process of learning to speak."{40}

      An alternative interpretation would be that Sellars is to be taken quite literally in saying that thoughts exist prior to language acquisition. But it may also mean that ur-conceptual thought exist independently of language learning. Perhaps this is all that Chisholm was claiming when he wrote to Sellars:

Thoughts would be intentional even if there were no linguistic entities. (This is a sentence about psychology. I concede that if we had no language, our thoughts would be considerably more crude than they are. {41}

In his correspondence with D. Rosenthal, Sellars tries to reconstruct Chisholm's position in this way:

My guess is that Chisholm so uses 'psychological' that it includes items which I would deny to be intentional (e. g., sense impressions, tickles, pains, etc. ) -- cf. my discussion of the pseudo-intentionality of sense impressions in EPM. Unless I am very much mistaken, he thinks that they have intentionality, or confuses a broader sense of 'intentionality' in which a non-conceptual items can have intentionality, with a narrower sense in which they cannot.{42}

It is not clear what Sellars is alluding to by 'intentionality in the broader sense. And it is still not clear to me what kind of thinking Sellars grants to pre-linguistic people and animals. My best guess is that he has in mind something like the 'logical' construction of physical objects out of sensa.

V. LOGICAL CONSTRUCTION

      The aim of logical construction was to provide a material equivalence between the analysandum and the analysans. But when we read Sellars on these matters he seems to be after something else. Suppose now that the analysandum is a 'theoretical construct' that is not equivalent to the analysans, but a theoretical superstructure that unifies the so called elements -- what may be called a 'theoretical hypothesis'. If that is the case, then we can at most search for elements and unifying principles, and say they are the ingredients of the analysandum in some sense, but which do not exhaust its constitution.

      On this alternative view, logical construction is not either a matter of same-level or different-level analysis, simply because it is not a matter of analysis, but of something else. It is more akin to saying that a cube of pink ice is identical to a swarm of molecules. It is a matter of cross-categorical correspondences.

VI. UNALTERABILITY OF RS-CONCEPTS

      I believe that there is a stratum of the Manifest Image consisting of the rs-seeing of physical objects which is an invariable achievement of normal higher animal life. If this is the case, then this may be an unalterable way of seeing. If we consider the relation of belief to seeing, we can note certain independencies. First consider the cases of visual illusions. Despite my sophistication and beliefs to the contrary, when I look at the various illusory pictures (Muller-Lyer, Hering, Pozo, Poggendorf, Necker, etc.{43} things still look as they are not. This suggests that there are ways of seeing that are independent of what one believes. This same independence is illustrated by the constancies of perception, which is the phenomenon that things tend to look normal despite various abnormal alterations (to a degree) in our visual environment. This ranges from seeing a white object as white in yellow illumination to the extreme cases like those experiments of Ivo Kohler with distorting goggles as, for example, inverting the visual field, so that everything initially looks upside down. What happens is that the wearer of such goggles eventually gets adjusted to the distortion and sees the world almost normally. The existence of these two classes of phenomena -- visual illusions and visual constancies -- suggests that our visual endowments are independent of our conscious beliefs.


[Go to Citations]