The whole rabbit: On the perceptual roots of Quine's indeterminacy puzzle

Abstract

In this paper I offer a novel analysis of Quine's indeterminacy puzzle and an unorthodox approach to its resolution. It is argued that the ultimate roots of indeterminacy lie not in behaviorism per se, but rather in Quine's commitment to a fundamental assumption about the nature of perceptual input, namely, the assumption that sensory information is strictly extensional. Calling this assumption the ‘principle of input extensionalism’ (PIE) I first demonstrate the fundamental role that it plays in generating Quine's argument for the indeterminacy of translation. It is then argued that a tacit acceptance of PIE is prevalent among contemporary theories of cognition and representation, making Quine's puzzle a living challenge for cognitive science and the philosophy of mind. A standard way of responding to the challenge consists of the idea that the indeterminacy generated at the perceptual level is canceled out courtesy of the contribution of higher cognitive processes. However, I argue that such a top-down solution leaves much to be desired. As an alternative, I offer a bottom-up solution, which consists of a systematic rejection of PIE in favor of an intensional view of perceptual input rooted in contemporary action-based theories of perception, and in particular in the notion of perceptual invariants.

Keywords:

• Behaviorism
• Cognitivism
• Extensional Entities
• Gestalt
• Indeterminism
• Input Extensionalism
• Intensional Entities
• Perceptual Invariants

Notes

Notes

[1] Two prominent philosophers whose writings on meaning and intentionality strongly assimilate Quine's ideas about radical interpretation are Davidson (1984) and Dennett (1987). However, it is a major claim of the current paper that one need not “sign in” to Quine's thesis in order to appreciate its significance and acknowledge its relevance. While I ultimately reject Quine's conclusion I shall nevertheless argue that it is one with which contemporary theories of meaning and representation must grapple.

[2] Donald Davidson is an odd exception in that he seems to affirm both indeterminacy and intentional realism. However, it must be observed that Davidson is not entirely comfortable with this combination, stressing that it is possible to find enough “reasonable and non-question-begging constraints” on a theory of meaning to ensure that it yields “correct interpretations” (1984, p. 152).

[3] In discussing Quine's puzzle reference is made to both ITT and the thesis of the inscrutability of reference (see section 2.1).

[4] It is worth noticing that Quine assumes that the natural bearers of meaning are whole sentences and that sub-sentential components acquire their meaning within the context of whole sentences (Quine, 1959, p. 152).

[5] Indeed, on Quine's view stimulations are our only first-hand source of objective knowledge. Correspondingly, the importance placed on observation sentences consists in their alleged transparency, namely, in the belief that such sentences reflect external stimuli reliably and without embroidery. In other words, Quine assumes that a certain class of sentences called ‘observation sentences’ consists of sentences which “wear their meanings on their sleeves” (Quine, 1960, p. 42).

[6] Recognizably, this is the reason why Quine insists on distinguishing his indeterminacy thesis from a mere underdetermination of theories by the facts (Quine, 1987; see also Gibson, 1987; Kirk, 2004; Searle, 1987).

[7] Some interpreters of Quine tend to emphasize the differences between ITT and the inscrutability of reference, and they may find the fact that I describe the two theses as tightly linked distasteful. That being said, I believe that my interpretation is consistent with the passages to which I refer. In particular, it is supported by a passage from Ontological Relativity, where, in introducing the inscrutability of reference, Quine states explicitly that “the indeterminacy of translation now confronting us… cuts across intension and extension alike…. Reference itself proves behaviorally inscrutable (1969a, p. 35). Finally, it may be mentioned that I am by no means alone in emphasizing the affinity between the two theses (see e.g., Kirk, 2004; Searle, 1987).

[8] Formulations of the extensionality principle frequently differ in their fine details. The present articulation, and in particular the distinction between logical co-extension and material co-extension, is meant to serve the discussion that follows. In any case, the principle so formulated is in line with Barcan Marcus's adequacy criterion, according to which a genuine extensionality principles is one that either “(a) directly or indirectly imposes restrictions on the possible values of the functional variables such that some intensional functions are prohibited or (b) it has the consequence of equating identity with a weaker form of equivalence” (Marcus, 1960, p. 57).

[9] The connection can be delineated as follows. First, as Cornman (1962) shows, two of Chisholm's three criteria for intentionality (the second and the third) are, as a matter of fact, criteria for intensionality. Second, as I argue elsewhere (Shani, 2007) the intensionality of propositional attitude reports (i.e., the familiar substitution failures characteristic of attitude reports such as ‘Frans believes that the morning star is divine’) is contingent on the fact that the attitudes and mental contents being reported are themselves intensional entities. Hence, what looms large behind Chisholm's criteria for intentionality is the fact that intentional discourse is irreducibly intensional, or to put it more directly, that intentional phenomena qualify as intensional entities.

[10] My use of the expression ‘primacy of the intentional’ involves a certain anachronism. As far as I know, it was only much after Quine's discussion in WO that Chisholm came to call his irreducibility thesis by that name (Chisholm, 1984); however, it seems to me that this name is considerably more accurate than ‘Brentano's thesis’, as it is often called.

[11] I am, of course, not oblivious to the popular attempt to explain away this apparent irreducibility of intensional entities using the conceptual apparatus of possible-worlds semantics (see, for example, Carnap, 1956; Chalmers, 2002; Lewis, 1986; Montague, 1960). However, as I argue elsewhere (Shani, 2007), a major problem with this strategy is that it presupposes that extensional entities are ontologically more fundamental than intensional entities (in particular, possible-worlds semantics sanctions the reduction of properties to an extensionalist, object-based ontology). As I argue there, there are good reasons to suspect that even if it were technically flawless (a disputable point in itself) the reductivist extensionalism of possible-worlds semantics fails to accord with our best contemporary science, making it ill-motivated from an ontological point of view.

[12] This is why I find Searle's (1987) treatment of the subject unsatisfactory. For even if we grant him the claim that our own conscious experience proves that we are the owners of intentional states with determinate semantic values, and hence bespeaks the incorrectness of Quine's thesis, the theoretical possibility of such semantic determinacy remains a mystery (see Shani, in press).

[13] For a classical discussion of this problem (albeit with different emphases than our present concerns) see Wittgenstein (1972, part II).

[14] I may mention that, in some important respects, my argument resembles similar arguments against nativism advanced by Bickhard (1991), Gibson (1979, p. 253), and Thelen and Smith (1994, p. 31). However, none of these authors apply their anti-nativist arguments to the problem of the intensional character of content.

[15] Note that an appeal to memory provides only for temporary refuge since the question remains what is the ultimate origin of the information thus supplied. If the memory is traced to past perceptual interactions input extensionalism reestablishes itself; if not, then the information must eventually be traced to innate knowledge.

[16] It is worth emphasizing that PIE is not a regular empirical hypothesis. At its background lie two deeply entrenched theoretical assumptions: (1) that perceptual input consists of purely physical entities (specifically, of discrete energy chunks goading our sensory receptors); and (2) that physical entities are necessarily extensional entities (i.e., satisfying the extensionality principle). Given these assumptions the possibility that some creatures might have developed, or might one day develop, perceptual organs sensitive to intensional nuances is precluded (for more on this problematic identification of ‘physical’ with ‘extensional’ see Shani, 2007).

[17] To illustrate, observe that weak nativism is entirely consistent with the possibility that equally intelligent and perceptually sophisticated creatures, perhaps even creatures much like us, could, in principle, perceive rabbits as undetached rabbit parts, or as indeterminate entities in the manner suggested by Quine, and yet their perceptions would be nonetheless adequate for that, and just as suitable for meaningful guidance of successful interactions. But this is just an indirect way of making the point that the intensional character of our own mental states is irrelevant to their representational adequacy.

[18] To be sure, Quine himself did think of his indeterminacy of translation thesis as a consequence of behaviorism (see, e.g., Quine, 1987, p. 5) but my point is that even though behaviorism is sufficient to entail Quinesque indeterminacy our discussion suggests that it is non-necessary.

[19] It is important to bear in mind that this does not exclude contribution by cognitive factors such as memory. Rather, what is denied is that such contribution applies to strictly extensional stimuli. Assuming that there never is such a thing as seeing without seeing as does not exclude top-down modification of the exact manner in which an object might be perceived.

[20] It is interesting to note that Quine discusses the notion of Gestalt on several occasions (e.g., 1969b, p. 84; 1973, chapter 1), concluding that it constitutes no serious threat to his atomistic theory of stimulus input (I thank an anonymous referee for bringing this point to my attention). However, Quine's dismissal of the gestaltist challenge depends entirely on the assumption that gestalt perception is necessarily conscious, from which he concludes that, like its arch rival, classical sensory atomism, gestalt theory belongs with an outdated “old epistemology.” But whatever the merits of Quine's critique may be (and I have my doubts about it) I make no assumption to the effect that gestalt perception (as understood in this section) necessitates conscious experience. Consequently I find nothing in Quine's discussion of gestalt perception that undermines my account, or indeed my critique of Quine's theory of stimulus input.

[21] For a critique of the concept of ‘stimulus’ as referring to static, momentary surface disturbances, see Gibson (1960, 1966, 1979).

[22] A transformation group on a set X is any set G of one-one transformations T on X which satisfies the group operations of inverse, identity, and composition (i.e., for each T in G there is T⁻¹ _; the identity transformation I= T · T⁻¹ is in G; for any two transformations S and T in G their composition S · T is in G).

[23] For a discussion of cross-ratios, see Cutting (1986), Michaels and Carello (1980). Topological invariants are discussed at length by Chen (2005), who argues that the most primitive and powerful invariants, i.e., those to whom attention is psychologically most natural and immediate, are topological.

[24] For more on perceptual invariants and their role in ecological psychology, see Cutting (1986), Gibson (1966, 1979), Michaels and Carello (1980).

[25] It may be noted that Gibson himself disapproved of the terms ‘form perception’ and ‘gestalt perception’ (Gibson, 1951, 1973). His disapproval was based on the presumption that a form is a concrete pattern, a visual image (say, of the face of an object). Rightly, Gibson stressed that it is the underlying invariant, not the overt perceptual patterns, that matters. However, as I show below, the Gestalt concept endorsed in the present paper has more to do with such underlying invariants than with concrete perceptual images. If we bear this in mind, and understand ‘form’ and ‘Gestalt’ in invariant terms, there is no harm in describing Gibson's own work on perceptual invariants as an important contribution to the study of form perception.

[26] The idea that the perception and representation of objects is to be associated with such complex network of reciprocal invariants can be traced to Piaget (1954), and is further developed by Bickhard (e.g., 1993; see also Shani, 2006).