T R T P: HE Eflexive Heory of Erception
T R T P: HE Eflexive Heory of Erception
T R T P: HE Eflexive Heory of Erception
ABSTRACT: The Reflexive Theory of Perception (RTP) claims that perception of an object or property X by an organism Z consists in Z being caused by X to acquire some disposition D toward X itself. This broadly behavioral perceptual theory explains perceptual intentionality and correct versus incorrect, plus successful versus unsuccessful, perception in a plausible evolutionary framework. The theory also undermines cognitive and perceptual modularity assumptions, including informational or purely epistemic views of perception in that, according to the RTP, any X-caused and X-directed dispositions are genuinely perceptualincluding affective, attitudinal, and immediately activated purely action-directed behavioral dispositions. Thus the RTP has the potential to provide the foundations for a broadly behavioral counter-revolution in cognitive science. Key words: behavioral theories of perception, reflexive theories, functionalism
I shall be arguing for a broadly behavioral theory of perception, to be called the Reflexive Theory of Perception, or RTP. Its advantages over previous behavioral theories of perception (e.g., Neisser, 1976; Pitcher, 1971; Taylor, 1962) could briefly be summarized as follows: A. It can integrate well with a broadly behavioral view of all psychological and cognitive activities, a view that is potentially fully competitive with non-behaviorist accounts. B. It can simultaneously satisfy all seven of the following requirements of an adequate behavioral theory of perception: 1. It involves reflexive elements, so that, for instance, perceptual aspects of behavioral conditioning can involve behavioral responses that are specifically directed toward the stimulus that caused them (Dewey, 1896; Gibson, 1950, 1966; Hull, 1943). 2. It can accommodate facts about the evolutionary development of perception in a wide range of species. 3. It is based on a functionalist dispositional structure, so it can both support behaviorist insights about perception and
AUTHORS NOTE: Please address all correspondence to John Dilworth, Department of Philosophy, Western Michigan University, Kalamazoo, MI 49008. Email: dilworth@wmich.edu. My thanks for very helpful comments to the Editor, Armando Machado, to anonymous referees for the journal, and to my WMU colleagues.
17
DILWORTH broader neuroscientific and cognitive science functionalist views (Bennett & Hacker, 2003). It can both accommodate and explain perceptual intentionality or representation in behaviorist terms via use of U. T. Places intentional analysis of dispositions, familiar to behaviorists, to develop a dispositional analysis of perceptual representation (Place, 1996). It can explain non-representational, active, or interactive aspects of perceptual activities, such as are argued for in recent sensorimotor accounts of perception (ORegan & No, 2001; Varela, Thompson, & Rosch, 1991). It can provide a non-modular theory of perception, required in a behaviorist theory because behavioral responses to a stimulus are external factors that could be of many different, module-crossing kinds. It can also explain common perceptual failures such as incorrect perception or unsuccessful perceptually based behaviors.
4.
5.
6.
7.
Thus a significant part of whatever novelty the RTP has is to be found in its integration of all of the above elements. In terms of specific elements, the emphasis on the primacy of reflexive factors in both perception and perceptual evolution seems to be new, as does also the dispositional analysis of perceptual representation and misrepresentation. In the first two sections of this paper a brief defense of behaviorist approaches to cognition will be supplied, along with a summary of the general theoretical methodology to be employed, before introducing and defending the theory in the remainder of the paper.
18
However, behaviorists can and must deny Blocks assumption that cognitive states are computational states. The whole identity of the kind of behavioral theory being considered by Block, according to which Behaviorism in one form is the view that two systems are mentally the same just in case they are the same in input-output capacities and dispositions, (p. 978) is bound up with its claim that cognitive states are dispositional states of input-output systems and not computational states of such systems. Hence the proper reply for dispositional behaviorists to make is that specifically computational states have as little to do with genuine cognition as do a variety of different neurological states, each of which might also realize a common dispositional behavioral state of an organism. Or, to put the point in terms of functional role (i.e., causal role in an inputdisposition-output model) the cognitive operation of calculating that two plus two equals four could equally be realized in either of Blocks different computational models, and arguably both of them can be best explained in terms of the functional role of the system in transforming, via its internal dispositional structure, two input copies of the number 2 into an output copy of the number 4. Thus, far from this kind of behaviorism or dispositional functionalism having been discredited, I would argue that much of the central core of our understanding of what is cognitively involved in such mathematical operations comes from a behavioral or functionalist model that applies equally well to abstract mathematical models of calculation, computer implementations, and relevant states of biological organisms (which may not even be capable of being in computational states for lack of the right kinds of physical structure). The above points constitute an initial sketch of how a broadly behavioral approach to cognition might be rehabilitated, even in what might be considered to be central or hard-core cognitive areas such as those involved in mathematical thinking. However, cognitivist opponents of behaviorism likely would switch at this point to an alternative strategy, as follows. Their claim likely would be that even if a behavioral theory or approach to much of cognition is theoretically viable or possiblerather than dead or discredited as usually assumednevertheless it would be scientifically redundant, or unnecessarily complicated, or currently too little investigated, and so on, to be able to adequately compete with entrenched standard cognitive approaches (this could be called the who needs it? strategy for rejecting behaviorist approaches). Or, to put the issue in another way, the challenge to the behaviorist likely would be to show, with respect to some important area of cognition, how a behaviorist or dispositional functionalist kind of explanation would be significantly
19
DILWORTH superior in explanatory power to competing, more standard cognitive approaches based on computational or other non-behavioral models which otherwise should, it would be assumed, win the theoretical contest by default because of their role as standard entrenched paradigms. It is this challenge that I shall take up in the rest of this paper, with respect to perception and perceptual activities throughout the biological kingdom.
20
REFLEXIVE THEORY OF PERCEPTION activities constructed upon this dispositional perceptual base, such as speculative thoughts of various kinds, may not themselves involve any direct behavioral dispositions in spite of their perceptual roots. One other element in the proposed account should be mentioned since it constitutes a vital part of an adequate, broadly behavioral reply to the attacks on radical behaviorism by Chomsky and others (e.g., Chomsky, 1959). That element is an evolutionary component, which could explain how characteristic kinds of genetically determined, and hence innately structured, perceptual dispositions could have evolved in a species. A broadly empiricist behavioral theory could still insist that all behavioral dispositions were initially perceptually acquired through the learning history of individual organisms, consistently with postulating evolutionary factors that preserved successful learning while extinguishing unsuccessful attempts, via the evolutionary genetic adaptation of species that included such individual learners. Thus there is still a place for behavioral concepts such as that of operant conditioning, or other kinds of perceptually based dispositional learning concepts applying to the individual history of an organism, while behaviorists may also happily accept a significant role for innate, speciesspecific dispositional structures such as language-acquisition skills in current human cognition.
21
DILWORTH acquired X-related dispositions as part of its perceptual contact with X. Also, negative X-related causal dispositions need to be considered too, for example a disposition to refrain from causal interaction with X when that interaction would otherwise occur, such as if Z is about to collide with object X, and its perception of X consists in its being caused by X to acquire a disposition, immediately activated, to avoid colliding with X. Thus overall, the best evidence that animal Z has perceived object X is if Z attempts to do something X-related, such as attempting to avoid X or to interact with it. At the same time, the best evidence that Z has not perceived X is if Zs behavior shows no manifestation of any X-related dispositions whatsoever. To be sure, this purely dispositional view of perception, and of the evidence for its occurrence, might initially seem intuitively questionable in that perception is widely regarded as being a process of information acquisition, with any associated behavioral dispositions, whether activated or not, being regarded as separable from, and subsequent to, the intake of such perceptual information.2 However, the RTP can immediately reply with a counter-challenge to such informational views, as follows: If a pure informational view were correct, it would be possible for an organism to perceive all kinds of things without ever engaging in any subsequent appropriate behavior. But such an intellectualist, pure acquisition of information view would empirically be completely empty in the absence of any concrete behavioral evidence that perception had actually occurred (Dilworth, 2004).3 Given that legitimate empirical perceptual theories must explain the role that behavioral evidence plays in establishing that perception has occurred, the simplest explanation of perception itself is that it consists in dispositions to behave in the ways that have been observed. Hence a dispositional theory of perception such as the RTP is the simplest available legitimate empirical theory, whereas a pure informational view has no comparable empirical credibility. Another initial intuitive roadblock to acceptance of a dispositional theory such as the RTP is that the category of dispositions, even when specifically limited to X-caused and X-related dispositions, might seem too unconnected with the standard perceptual and semantic issue of correct versus incorrect, or veridical versus non-veridical, perception. In what sense can some perceptually acquired, Xrelated disposition be correct or incorrect with respect to X, since any actual behavior toward X that manifests the disposition is simply a behavioral event, having no intrinsic semantic properties? Nevertheless, here too a supporter of the RTP can appeal to the common empirical currency of behavioral evidence and argue that the only actual evidence we can have as to the correctness or incorrectness of some particular perceptual episode in organism Zs history is broadly behavioral evidence, so that the RTP cannot be any worse off with respect to evidence of semantic correctness than any
See, e.g., Dretske (1981) and Fodor (1990). As emphasized by many psychologists from Dewey (1887) onward; see also Millikans point that informational views must consider consumer as well as producer aspects of information (e.g., 1989).
3 2
22
REFLEXIVE THEORY OF PERCEPTION other broadly empirical theory of perception. If at least some behavioral episodes do provide legitimate evidence of correct or incorrect perceptionas they must for those concepts to have any empirical contentthen they equally support the attribution of correctness or incorrectness to X-caused and X-related dispositions to thus behave, in conformity with the account of perception offered by the RTP. As a simple example, the perceptually acquired disposition for a hungry person to eat some nutritious food placed in front of him while being disposed to refrain from eating some rocks similarly placed would, when behaviorally manifested in either case under normal circumstances, provide adequate behavioral evidence of correct perception of the food and rocks on any theory of perception, including the RTP. Or a linguistic example: if, after gazing at a red object X, one says that is red while pointing at X, this would be clear behavioral evidence of correct perception of its color on any theory of perception. On the RTP, this case would involve X-caused, perceptually acquired correct dispositions with respect to the color of X, including a disposition to thus demonstratively utter the relevant sentence in appropriate circumstances (for a useful discussion of related behavioral issues in Quine and Davidson see George, 2004). To sum up this section, perhaps enough has already been said to show that the reflexive theory of perception has at least some initial viability in comparison with other perceptual theories. The following sections will seek to further demonstrate its theoretical strengths.
A recent account of complexities in the concept of adaptive evolution is given by Walsh (2003). 5 For a survey of relevant literature see Downes (2001).
23
DILWORTH those behaviors of individuals that tend to promote survival of their species, those that on average make no difference to its survival, and those which are statistically inimical to its survival). In this broad sense all behavior is adaptive in that each item of behavior has at least some minimal potential for changing the level of successful adaptation of its species to the relevant environment. For theoretical purposes, each item of adaptive behavior can usefully be regarded as being caused by some underlying behavioral disposition, initially in circumstances in which the activation conditions for that disposition are immediately realized. Then more subtle and powerful forms of dispositional causality could evolve gradually, given the evolutionary advantages of sometimes delaying a causal response until conditions are more optimalsuch as when a predator, primed with a disposition to eat its prey, waits until the prey is most vulnerable. But in order to achieve a more substantive explanation of adaptive behavior, issues of causality must be pushed back one stage further so that the primary issue regarding the causality of adaptive behavior is what caused those relevant dispositions themselves? In general terms, it is some environmental factor X that causes such a disposition, whether the factor is an object external to an organism Z or some internal part of Z itself. Now a second meaning of the term adaptive may be introduced that is closer to the everyday meaning of the term adapt. It involves some changes in disposition D, and hence behavior B, each of which results from some change in the environmental factor X. In this sense an organism Z adapts to, or is responsive to, changes in its environment X via a causal mechanism in which the changes in X (i.e., each succeeding changed state of X causes a corresponding change in Zs disposition D, and hence in its behavior B). Clearly this environmentally caused kind of responsive change in dispositions, and hence behavior, will often be required for successful adaptation, in the first sense, of the relevant kind of organism to an environment that is changing in significant ways. But we still are missing one crucial element that is needed in order to achieve a theoretically useful concept of perception as such. So far we have nothing but causal chains and causal correspondences relating Z and its environment. In order for genuine perception to occur in organisms of type Z they must be able to achieve some kinds of adaptively beneficial control or power over the environmental factors X that cause changes in their dispositions D, in addition to merely being responsive to them. In this manner the responsive changes in behavior in organisms of evolving type Z could become relevant (i.e., causally effective, in diminishing environmentally caused threats, or enhancing potential environmental benefits). But the only way in which this desired result of control over environmental factors can be achieved within the available naturalistic causal parameters is for the relevant X-caused dispositions D of organism Z to cause behavior that itself causally acts upon, or causally interacts with, those relevant environmental factors X. A typical controlling situation would be one in which an increasing value of X would have negative adaptive value for Z, but in which Z is caused by X to acquire a disposition D that, when activated, in turn leads to a reduction in the value of X
24
REFLEXIVE THEORY OF PERCEPTION (i.e., a negative feedback causal loop, in which an organism achieves adaptive stability with its environment by directly modifying the threatening changes in X).6 For example, much competitive behavior in animals can be explained thus: an increasing threat to animal Zs food supply from animal X will, if perception in Zs species has been adaptively successful, typically cause animal Z to fight off Xs attempt to eat the food needed by Z. The initial picture of perception that emerges from this account is of perception as one uniquely effective causal mechanism by which evolutionary adaptation can be achieved by a species, in which organisms of type Z are caused by some environmental item of type X to acquire X-related dispositions which, when activated, may improve the adaptive success of type Z organisms with respect to their interactions with items of type X. But this view of perception is none other than the RTP itself, as generalized to apply to adaptively relevant types of causal interactions between a species and its environment. To be sure, perception as thus characterized is not the only adaptively relevant causal mechanism, as the above account makes clear, such as a case in which an item X might cause organism Z to acquire non-X-related dispositions that nevertheless have adaptive value. For example, the scent of a certain plant X might lead to more reproductive behavior between members of the species Z, even though the scent-caused reproductive dispositions in such a case are not themselves scent-related or scent-directed. However, the evolutionary centrality or primacy of perception as a reflexively defined causal mechanism comes from the fact that in order to be maximally effective in evolutionary terms, the formation of such non-reflexive dispositions must itself be maximized by the perceptual acquisition of scent-related dispositions by members of species Z, such as a scent-caused disposition to seek out similar sources of the scent properties in the future, so as to ensure more cases of additional reproductive behavior. Thus in such a manner specifically reflexive, genuinely perceptual dispositions play a vital instrumental, facilitating, or catalytic role in potential adaptive successes, even when other causal mechanisms also have a significant role. To briefly summarize and explain the argument and broader context of this section, an uncompromising naturalist approach to perception demands that traditional epistemic approaches to perception, viewed as the only sensory, broadly empirical means of acquiring normatively correct information or knowledge about the world, be bypassed completely. In their place, a pluralist view of adaptively relevant causal factors or mechanisms must be postulated. However, one of those mechanisms, namely the reflexive causal mechanism that defines the subject matter of the RTP, is both theoretically and causally primary or central in that it is causally indispensable for some adaptive results while also uniquely facilitative of adaptive success for the other available causal mechanisms.
Such negative feedback analyses can also be used to explain purposive concepts in naturalistic terms (see Falk, 1995).
25
DILWORTH Also, this reflexive causal mechanism is the closest analog in evolutionary theory for the perceptual processes or mechanisms postulated in other, more traditional theories of perception, whether biological, psychological, or philosophical, and hence it deserves to be described specifically as a perceptual mechanism. As mentioned in the Introduction, all such theories must, if they are to have any substantive empirical content, accept the common currency of behavioral evidence that perception of an object X has indeed occurredevidence which can only be provided by X-caused cognitive activity that results in X-related kinds of behavior. The RTP is a minimalist perceptual theory that adds only a single factor to those central items of behavioral evidence for the occurrence of perception of X, namely that perception consists in the acquisition of X-caused dispositions to thus behave in an X-related way.
26
REFLEXIVE THEORY OF PERCEPTION perception-related dispositions, namely affective, emotive, or attitudinal dispositions (e.g., as manifested in desires) versus purely action-oriented dispositions that have no specific cognitive or affective components. But this high level, tripartite classification (e.g., as in Wundt, 1897) has no clear or theoretically principled application to more rudimentary cognitive systems. Hence, even if it is true that conscious human perception predominantly relies on a tightly circumscribed range of rational and epistemically relevant dispositions, this fact must not be allowed to bias, in a species-chauvinistic way, the account given of the basic nature of perception in an adequate naturalistic general theory of perception. Recall that the RTP claims that perception of an object X by an organism Z consists in Z being caused by X to acquire some (i.e., one or more) disposition D toward X itself. A characteristic feature of this view is that it does not limit in any way the dispositions D that might turn out to be thus acquired and hence count as genuinely perceptually acquired dispositions toward X. Thus, for example, the RTP has theoretical room for the possibility that dispositions grounding some desires, emotions, or attitudes toward X, or pure dispositions to act in some Xdirected manner, might be directly perceptually acquired in addition to dispositions providing a basis for knowledge or belief. In this the RTP is unlike other views of perception, which typically regard perception as exclusively involving epistemically relevant items such as information or beliefs about the state of the world, even if perception itself is not regarded as automatically being a justified process of knowledge acquisition.7 However, it seems not to have been realized that this narrow epistemic assumption about the nature of perception, as found in standard perceptual theories, introduces a serious and unwarranted theoretical bias into the very foundations of perceptual and cognitive theories, not just for lower or more rudimentary organisms as discussed above but also for higher mammalian (including human) perception as well, as will now be shown. The basic problem with such epistemic assumptions is that they foreclose on genuine empirical possibilities and force a hopelessly outdated faculty psychology on to the investigations and findings of contemporary cognitive science (see Fodor, 1983 for discussion of faculty psychology issues). On such views, a cognitive system is assumed to be divided into more or less rigid compartments, with an encapsulated perceptual system whose sole output is information about the world. It is then assumed that there must be independent, higher level cognitive units that further process such purely factual information about the world (including emotive or attitudinal units) that decide, on the basis of the perceptually acquired facts, what emotion, attitude, or value the organism should adopt to, or place upon, those facts, plus decision-making units that decide what actions should be taken in light of the perceptually discovered facts. Thus on such views, all emotions, values, or
7
For example, Armstrongs 1961 view of perception as belief acquisition does not require that all of the beliefs are true, but it does rule out any non-belief acquisitions, such as attitude or emotion-acquisition, as perceptual. Also, despite the evolutionary foundations of Millikans theory of perception she also views perception as primarily a matter of information acquisition (see, e.g., Millikan, 2004).
27
DILWORTH attitudes must involve higher-level cognitive interpretations or decisions about lower level perceptual facts. Similarly, any purely action-related dispositions are assumed to be exclusively the result of high-level rational decisions as to what it is best to do, all things considered, given the basic perceptual facts that are the low level input to high-level decision modules (also see Damasio, 1994 for useful discussion and criticism of such traditional views). But countervailing evidence concerning many of our emotions and attitudes is available in that they are often completely unreflective and instinctive, such as when one takes an instant liking, or dislike, to someone when one first meeting them or immediately hates, or loves, a painting on first seeing it. My claim is that the X-caused acquiring of such emotional dispositions toward X can be just as much a legitimate part of a low level, purely perceptual episode as can the acquiring of any other kind of more conventional epistemic perceptual disposition toward X. Similarly, one can instinctively or immediately react to something in an appropriate or inappropriate manner, which could equally be the manifestation of ones acquisition of a purely perceptual disposition to thus act. The assumption that all actions must be preceded by a high level decision to act on the basis of rationally evaluated facts rather than sometimes being an immediate manifestation of a pure, perceptually acquired disposition is just another distorting and unwarranted assumption implied by standard perceptual theories. In the next section an evolutionary argument will be given that offers empirical support for the claim that there is a wide incidence of such nonepistemic, but nevertheless genuinely perceptual, dispositions of such affective or purely action-oriented kinds.
28
REFLEXIVE THEORY OF PERCEPTION RTP, the relevant or salient aspect of ones perception of the vehicle might be nothing more than the immediate acquisition of a disposition to swerve so as to avoid it, a disposition which is, in turn, immediately activated so that one actually does thus swerve (Damasio, 1994, Ch. 8, gives a related account of such cases, with references). One main, evolutionarily significant difference between these two different methods of reacting to the oncoming vehicle is very simple: the reflexive disposition procedure will typically be significantly faster than the rational action procedure because it involves much less cognitive processing. The difference in reaction time might seem like a relatively insignificant difference, but its evolutionary significance is profound. This is because over the millennia species that generally organized their short-term reactions to worldly objects and events via reflexive perceptual dispositions would have a significant survival advantage over those that did not. Indeed, this factor alone virtually guarantees that any basic kind of decision-making in organisms that could be thus streamlined would actually be so streamlined. Hence, even if it is theoretically possible for perception to function in a purely epistemic modular way, as traditionally assumed, my claim is that evolutionary pressures alone would probably be sufficient to ensure the disappearance of such slower methods whenever the faster, more direct dispositional methods would be feasible. This argument is more powerful and versatile than it might seem at first because it potentially applies to any kind of disposition, not just to overtly survival-critical dispositions of flight or avoidance of dangers. Given the related evolutionary advantages of an efficient and simple cognitive structure in organisms, the streamlining procedures that optimize survival chances in danger avoidance are also highly likely to work similarly on any other potential cognitive structures. Any initial cognitive organization that functionally separated perception and recognition from various aspects of cognitive decision making would also be very likely to be streamlined into a purely perceptual disposition-acquisition process whenever possible. Thus, for example, in perceiving an interesting new book in a bookstore, my claim would be that probably at least part of that perceptual event was the acquisition of a disposition to read that book. However, if asked about the event, one is likely, in the grip of the traditional view, to produce a kind of rational reconstruction of the event and claim (in effect) that one first saw or perceived the book, then realized that it had the intrinsic property of being interesting, then one decided on that basis that one should read it, then one deliberately formed an intention to read the book that involved a disposition to read it. But in evolutionary terms such convoluted cognitive processes would have no chance of survival in cases where the relevant simple and immediate perceptual disposition-acquisition was also possible. At this point a possible line of criticism of the RTP handling of the distinction of purely perceptual versus higher-level cognitive decision-making should be discussed. The criticism is that acceptable decision-making requires, in addition to a speedy decision in some cases, also at least a minimum amount of rational
29
DILWORTH deliberation in all cases, including even in time-critical cases such as a decision to swerve to avoid an oncoming vehicle. Hence, it would be argued that the rational perceiver has to decide, for every perceptual situation with which she is confronted, whether to immediately act a certain way with respect to it or whether it would be better instead to engage in more prolonged deliberations (even if only for another second or so). But such decisions themselves involve a process of higher order cognitive deliberation based on prior, epistemically structured perceptual data, so there cannot be any cases of purely perceptual, immediately activated action-dispositions that are also acceptably rational. There is a standard kind of evolutionary answer to such questions that provides at least one kind of adequate response to them, as follows. The overall acceptability or practical rationality of a persons use of their perceptual mechanisms, including their differential usesometimes in a purely perceptual way and at other times in a more explicitly deliberative waydepends not on the details of their reasoning but instead on the general evolutionary success of the surviving gene pool of the species homo sapiens, the members of which have in fact successfully used such differential techniques. From this perspective, normative standards of rational decision-making, as opposed to those actual decision-making practices that have survived the evolutionary winnowing process, are simply causally irrelevant. As a coda to this section, it is important to note that the above evolutionary argument is sufficient by itself to refute traditional epistemic theories of perception along with their assumption that attitude formation and action-oriented decision making are always distinct stages of cognitive processing. Thus, whether or not the RTP is itself acceptable as an adequate perceptual theory, standard purely modular cognitive architectures cannot be even approximately correct because of the inevitable prevalence of evolutionarily effective shortcuts as discussed above (also see the following two sections for further discussions).
30
REFLEXIVE THEORY OF PERCEPTION that all cognition and perception are fundamentally modular, with each cognitive function, including perceptual functions, being performed by a specialized module. But for the reasons already given here, any genuinely behaviorist theory of perception invoking evolutionary factors plausibly should reject a modularity thesis for perception because, for instance, it is extremely unlikely that evolutionary forces alone could have trimmed down the indefinite range of possible reflexive, X-related behavioral responses (to a given perceptual stimulus X) to any specialized subset of a single particular kind, such as that of purely information-related responses. Thus the very strengths of an evolutionary approach condemn as highly unlikely any such purely modular approach to cognition (for related cautionary remarks see Fodor, 2000).
31
DILWORTH representational aspects of perception, which modular models cannot capture but which can be explained convincingly by the RTP). Here now is a second non-evolutionary argument for the non-modularity of perception. Support for the argument was, almost inadvertently, recently provided by Fodor (2000), who has been one of the leading advocates of computational views of cognition over the past thirty years or so but who argues in this work that the modular computational approach cannot capture the actual context sensitivity of many cognitive situations. But cases of context-sensitivity are, from a behavioral point of view, nothing more than the normal behavioral dependence of a response on specific external activation conditions rather than purely on factors internal to a given module, so that such kinds of cognitive context dependence are exactly what one would expect if a behavioral view is correct, while competing modular views are unable to explain them.
32
REFLEXIVE THEORY OF PERCEPTION incoming sensory data, but the guesses might be incorrect in some cases. For example, a perceiver could give evidence of having correctly perceived an object X to be red by her red-related behavioral response of putting the object into a bin reserved for red objects, or alternatively of incorrect perception if instead she puts the object into a green bin. Thus, on this account, perception does not involve sensorily acquired information concerning redness that is stored in a perceivers head; instead it involves a redness-caused perceptual state having the functional role of being dispositionally linked to possible red-related behavioral manifestations, whether of a correct or incorrect kind. An important implication of this view is that a perceptual state is not a computational state in that it is not, in and of itself, a pure informational state since it is only its dispositional links to possible red-related behavior that allow it to play its informational role. Thus the current RTP can also provide behaviorists with some principled reasons to reject the widely popular, computationally based representational theory of mind held by many cognitive scientists (on which see, e.g., Fodor, 1990 and Sterelny, 1990). Nevertheless, the everyday concept of a perceptual state being a representational one, or of an object being perceptually represented as being red, is still a useful one for many scientific purposesit is just that an alternative, dispositional (rather than computational) explanation of perceptual representation must be provided. Traditionally, the concept of representation has been closely linked with that of intentionality, or of thoughts or perceptions being about an object or property or of being directed toward such items. The current RTP already initially accommodates directedness toward actual objects or properties X via its X-directed dispositional account. However, because of the importance of the issue, some further discussion will now be provided. To begin, U. T. Place has argued that the relevant kind of dispositional aboutness (with respect to some possible situation that would actualize an Xrelated disposition) is an integral feature of any physical dispositional property, so that the relevant kind of aboutness is a purely physical feature of the relevant dispositions, and hence naturalistically unproblematic (see Place, 1996 and Armstrong, Martin, & Place, 1996 as well as the similar view of Molnar, 2003). Thus, as the title Intentionality as the Mark of the Dispositional (Place, 1996) suggests, ordinary dispositional properties themselves possess some of the main characteristics of intentional properties, including directedness toward an item X and the possibility that item X might not be actualized if worldly conditions are unfavorable. The novelty of the present use of Places theoretical discovery is that it invokes Places intentional analysis of dispositions in an analysis of perceptual intentionality itself. In other words, if a dispositional theory of perception and cognition can be provided, those dispositional structures will automatically inherit the intentional structures found by Place to be integral features of any dispositional properties. To be sure, some might be wary of Places analysis since it supports a view of the intentional objects of dispositional properties as possibly being nonexistent, since the possible objects or conditions with respect to which a
33
DILWORTH dispositional property might be activated or manifested might themselves not exist (e.g., see the chapters by Armstrong in Armstrong, Martin, & Place, 1996). However, in the perceptual case a weaker analysis of dispositional aboutness (i.e., one not requiring the possibility of nonexistent intentional objects) will suffice since the relevant object or property X of a perceptual state is guaranteed to exist because it has to cause the relevant perceptual state. Hence the present dispositional, naturalistic analysis of perceptual aboutness may be acceptable even to those who cannot stomach Places much stronger, unrestricted modal analysis of dispositions in general. Another important issue about perceptual representation is the possibility of misrepresentation, or of perceptually representing an object as having a property that it does not actually have, such as when someone perceives a red object to be green and hence puts it in a bin for green objects. A traditional problem in such cases is that ones perceptual representation seems to be about a property greennessthat is not present in the relevant object. But how can one have a disposition directed toward a non-present property (i.e., one that is not instantiated in the relevant perceptual situation)? The RTP has two replies to such concerns. First, the behavioral evidence in either correct or incorrect cases of perception is some actual behavior, a behavior that does manifest the relevant perceptually acquired disposition caused by the actual red color of object X. So only actual properties and actual behavioral responses are required to account for either correct representation or misrepresentation. Second, we must distinguish what a perceivers red-caused dispositions are actually directed towardnamely, in the present case, the actual red color of object Xfrom what the perceiver may think or believe they are directed toward because of her incorrect perception. If the perceiver puts red object X in a green bin this may show that she believes her action is directed toward a green color of X, which is how she perceptually represented X. But on the present analysis her action is directed toward the actual red color of X, with the evidence for her error being the incorrect sorting of the red object into the green bin. In other words, rather than the traditional explanation of perceptual misrepresentation as a kind of special mental correct aboutness with respect to a non-actual or non-present color, on the present analysis it is instead a matter of an incorrect aboutness as behaviorally evidenced with respect to an actual color. This completes the current dispositional analysis of the concepts of perceptual information and representation. As mentioned in the previous section, an important further implication of the present account of those concepts is that perception is non-modular in that the right external conditions would automatically trigger a behavioral manifestation of the relevant perceptual disposition without any passing of information among specialized modules, such as from a perceptual module to a decision module, then to a motor control module, and so on. However, to avoid any misunderstandings, my claim of non-modularity is specifically one claiming the non-independence or interactive behavioral functioning of perceptual, decision, and motor control cognitive functions. That
34
REFLEXIVE THEORY OF PERCEPTION issue is distinct from the issue of perceptual encapsulation (as claimed by Fodor, 1983, 1990), which instead concerns the extent to which perceptual processing is independent of prior beliefs and background knowledge possessed by a perceiver. The RTP is consistent with, and can be supportive of, perceptual encapsulation, which is one positive factor supporting a broadly realist view of cognitive and psychological information processing. On such an encapsulation viewin an RTP versionbehavioral interaction with the world can result in objectively correct scientific knowledge about it without any fundamental interference from prior personal beliefs or prejudices of scientists.
35
DILWORTH perception that can explain both its informational or representational aspects and its non-representational, worldly interactive elements within a single broadly behaviorist framework while also providing a novel account of how perceptual failures of various kinds are to be explained. The relevant non-representational, interactive elements could be explained as follows. First, the basic RTP theory already covers all X-caused and X-related behavioral dispositions of an organism so that many perceptual aspects of nonrepresentational interaction with the world are already explained by the basic theory. So rather than needing to define non-representational interactive cases, the theoretical situation is instead that representational cases need to be distinguished as a functionally distinctive subset of all cases, namely those to which standards of correct versus incorrect response may be applied. For example, a perception of food may involve an activated disposition to eat the food, to which normally no issues of correct versus incorrect response would apply, so that this would be primarily a non-representational perceptual interaction case. However, if an animal attempted to eat a pile of stones, the best explanation of its behavior might be that it had incorrectly perceived the stones as being food, in which case there is a salient representational or informational perceptual factor in the situation even though there are also non-representational interactive factors. An additional interactive and non-representational theoretical factor will now be introduced. The representational or informational functions of perception are closely linked to belief formation or knowledge acquisition. But organisms also use perceptual mechanisms as a way of satisfying their desiresbroadly speaking, to change the world or themselves, such as when a bird builds a nest or eats to satisfy its hunger. Now the fundamental functional difference of a desire from a belief is that the functional role of desire-related perceptual states is to behaviorally change a perceived object, which currently does not have property F, so that it acquires that property F. Fortunately, such desire-related perceptual states can still be explained in terms of their satisfying the basic RTP reflexive formula of being X-caused and Xrelated dispositions. The X in such cases would be some actual property of an object that one desires to be changed, such as a desire to change the red color of an object to a green color. Ones perceptual, red-caused perceptual state involves a red-related disposition to change that red color into the desired color, such as by repainting it. Thus, desires for new properties Y are behaviorally satisfied by Xcaused desires toward current property X, namely to change X into Y. In support of this analysis, it has the same theoretical virtue as the analysis of incorrect perception in the previous section (i.e., that no dispositions toward any nonexistent or non-present properties are required). Arguably, that actual-property feature is required in a genuine behavioral disposition account in that in order to be behaviorally effective there must be some actual conditions under which the disposition would be behaviorally manifested. But that causal condition cannot be satisfied unless there is some actual current property of an object toward which a perceivers disposition is directed. Now, just as belief-related perceptual dispositions can fail by being incorrect in misperception cases, so also can desires
36
REFLEXIVE THEORY OF PERCEPTION fail, or be unsatisfied, by being unrealizable by a given behavioral disposition. For example, just as one might misperceive a red object as being green, so also one might desire to change a red object into a green one, but manifest that desire with a behaviorally ineffective method which does not produce the desired result. Such failed-desire cases can be explained in a manner structurally similar to that in which incorrect perceptions or misperceptions were explained previously. In both cases the perceiver has false beliefs about her current dispositions. In the case of misperception of red as green, the perceiver falsely believes that her disposition is green-caused and green-directed when in fact it is red-caused and red-directed. In the ineffective desire case the perceiver acquires red-caused dispositions toward the current red property, which she falsely believes will change it into the desired green property, but her disposition in actuality manifests itself only as a failed attempt to do so. Thus, to summarize this section, the RTP now has available a comprehensive explanation of the basic elements of perceptual interactions with the world explanations which are also already closely tied in with other important cognitive concepts such as those of belief, desire, and intentionality, including the more specialized concepts of correct or incorrect perception or belief and of successful or failed perceptually integrated desires. As for other related perceptual theories, the RTP could provide some muchneeded theoretical foundations for recent, and increasingly influential, enactive, or sensorimotor perceptual theories (e.g., see Hurley, 1998 Ch. 10; ORegan & No, 2001), according to which perception should be understood in terms of the skilled interactions of perceivers with worldly items. Such theories currently deny that most perception is representational, but they have no behaviorally adequate interactive theory of perceptual representation with which to explain those cases or aspects, if any, in which perception might be genuinely representational.
37
DILWORTH X-caused and X-related perception itself from various other possible non-X-related dispositions that might accompany it. In such cases the scientist would still qualify as perceiving X as long as she acquires at least one X-related disposition Dsuch as a readiness to flee if the predator gets too closeeven if that disposition is almost never manifested because of careful selection of safe viewing conditions by the scientist. At the same time, other non-X-related dispositions accompanying that purely perceptual disposition D (e.g., a disposition to add further data to head count statistics) could also be considered to be part of the perceptual or observational situation in a broader sense (compare Holt, 1915). Such perceptually related, but not strictly reflexive, perceptual dispositions could also help to defuse a potential objection to the RTP, namely that we often perceive things, such as distant mountains or stars, to which apparently we acquire no dispositions at all. The current reply is to deny the claim in two ways. First, unmanifested reflexive dispositions may be invisible to casual observers, but they are still genuine, perceptually acquired dispositions. Second, often the purposes of perceiving distant objects are more related to ancillary, non-X-related dispositions (e.g., a desire to find out how far one still needs to drive in the case of perception of distant mountains) than they are to purely perceptual reflexive dispositions. But such dispositions, even if they only accompany X-related reflexive dispositions, may be counted as perceptually based in a broader sense as long as they are dependent on, or can only be acquired as part of, genuine X-related perceptual activities.
Conclusion
The first section of this paper briefly argued that a broadly behavioral approach to cognition is not dead after all, but instead just in need of a broader functionalist defense. A likely challenge in response from standard cognitivist views was subsequently identified, namely that it needs to be shown (with respect to some important area of cognition) how a behaviorist or dispositional functionalist kind of explanation would be significantly superior in explanatory power to competing, more standard cognitive approaches based on computational or other non-behavioral modelswhich otherwise should, it would be assumed, win the theoretical contest by default because of their role as standard entrenched paradigms. This challenge has been taken up in the rest of this paper for perception and perceptual activities throughout the biological kingdom, with an initial formulation and defense of a behaviorally based reflexive theory of perception (RTP). It has been argued, first, that standard modular accounts of cognition, whether of a traditional rationalistic or contemporary evolutionary or computational kind, are hopelessly unrealistic because the perceptual bases of all cognitive activities are fundamentally non-modular in the various ways that have been discussed. Second, standard information processing or computational models of cognition rely on a completely inadequate account of perceptual information or representation, whereas the current RTP can supply a novel and fully adequate behaviorist one,
38
REFLEXIVE THEORY OF PERCEPTION based on U. T. Places analysis of the intentionality of dispositions. Third, the RTP can supply an integrated account of all aspects of perception, whether representational or broadly interactive, whereas no competing theorieswhether standard cognitivist or behavioristhave so far been able to do so. Thus, in conclusion, it has been shown that the RTP is indeed significantly superior to standard cognitivist views, so that given the centrality of perception as a prime shaper of cognitive structure, the way is now open for a broadly behavioral revival in cognitive science, whose currently assumed modular, computational, and rationalistic structures have been shown to be significantly empirically inadequate.
References
Aristotle (350 BC). De anima. Available at: http://classics.mit.edu/Aristotle/soul.1.i.html Armstrong, D. M. (1961). Perception and the physical world. New York: The Humanities Press. Armstrong, D. M., Martin, C. B., & Place U. T. (1996). Dispositions: A discussion. Tim Crane (Ed)., London: Routledge. Barkow J., Cosmides, L., & Tooby, J. (Eds). (1992). The adapted mind. New York: Oxford University Press. Bennett, M. R., & Hacker, P. M. S. (2003). Philosophical foundations of neuroscience. Malden, MA: Blackwell. Block, N. (2001). Behaviorism revisited. Behavioral and Brain Sciences, 24, 977-978. Brandt, R. B. (1983). The concept of rational action. Social Theory and Practice, 9, 143164. Carruthers, P. (2003). Moderately massive modularity. In A. OHear (Ed)., Mind and persons. Cambridge: Cambridge University Press. Chomsky, N. (1959). Review of B. F. Skinners verbal behavior. Language, 35, 26-58. Chomsky, N. (1966). Cartesian linguistics. New York: Harper and Row. Cummins, R. (1989). Meaning and mental representation. Cambridge, MA: MIT Press. Cummins, R. (1996). Representations, targets and attitudes. Cambridge, MA: MIT Press. Damasio, A. (1994). Descartes error. New York: Harper Collins. Dewey, J. (1887). Psychology: The early works, Vol. 2. Carbondale, IL: Center for Dewey Studies. Dewey, J. (1896). The reflex arc concept in psychology. Psychological Review, 3, 357-370. Dilworth, J. (2004). Naturalized perception without information. The Journal of Mind and Behavior, 25(4), 349-368. Downes, S. M. (2001). Some recent developments in evolutionary approaches to the study of human cognition and behavior. Biology and Philosophy, 16, 575-595. Dretske, F. (1981). Knowledge and the flow of information. Cambridge, MA: MIT Press. Falk, A. (1995). Essay on natures semeiosis. Journal of Philosophical Research, 20, 297348. Fodor, J. (1983). The modularity of mind. Cambridge, MA: MIT Press. Fodor, J. (1990). A theory of content and other essays. Cambridge, MA: MIT Press. Fodor, J. (2000). The mind doesnt work that way: The scope and limits of computational psychology. Cambridge, MA: MIT Press. George, A. (2004). Linguistic practice and its discontents: Quine and Davidson on the source of sense. Philosophers Imprint, 4(1). Available at: http://www.philosophersimprint.org/004001/ Gibson, J. J. (1950). The perception of the physical world. Boston: Houghton Mifflin.
39
DILWORTH
Gibson, J. J. (1966). The senses considered as perceptual systems. Boston: Houghton Mifflin. Holt, E. B. (1915). Response and cognition. Journal of Philosophy, Psychology and Scientific Methods, 12, 365-373; 393-409. Hull, C. (1943). Principles of behavior. New York: Appleton-Century-Crofts. Hurley, S. (1998). Consciousness in action. Cambridge, MA: Harvard University Press. Mill, J. S. (1865). An examination of Sir William Hamiltons philosophy. London: Longmans. Millikan, R. (1989). Biosemantics. The Journal of Philosophy, 86, 281-297. Millikan, R. (2004). Varieties of meaning (2002 Nicod Lectures). Cambridge, MA: MIT Press. Molnar, G. (2003). Powers: A study in metaphysics. Oxford: Oxford University Press. Neisser, U. (1976). Cognition and reality. San Francisco: W. H. Freeman. ORegan, J. K., & No, A. (2001). A sensorimotor account of vision and visual consciousness. Behavioral and Brain Sciences, 24, 939-1031. Pitcher, G. (1971). A theory of perception. Princeton, NJ: Princeton University Press. Place, U. T. (1996). Intentionality as the mark of the dispositional. Dialectica, 50, 91-120. Sterelny, K. (1990). The representational theory of mind. Cambridge, MA: Blackwell. Taylor, J. G. (1962). The behavioral basis of perception. London: Yale University Press. Vanderbeeken, R., & Weber, E. (2002). Dispositional explanations of behavior. Behavior and Philosophy, 30, 43-59. Varela, F. J., Thompson, E., & Rosch, E. (1991). The embodied mind: Cognitive science and human experience. Cambridge, MA: MIT Press. Walsh, D. M. (2003). Fit and diversity: Explaining adaptive evolution. Philosophy of Science, 70, 280-301. Wundt, W. M. (1897). Outlines of psychology. C. H. Judd (Trans). Available at: http://psychclassics.yorku.ca/Wundt/Outlines/index.htm
40