STEVEN YALOWITZ
SEMANTIC DETERMINANTS AND PSYCHOLOGY AS A SCIENCE
ABSTRACT. One central but unrecognized strand of the complex debate between W. V. Quine and Donald Davidson over the status of psychology as a science turns on their disagreement concerning the compatibility of strict psychophysical, semantic-determining laws with the possibility of error. That disagreement in turn underlies their opposing views on the location of semantic determinants: proximal (on bodily surfaces) or distal (in the external world). This paper articulates these two disputes, their wider context, and argues that both are fundamentally misconceived. There is no special tension between error and strict semantic-determining laws; moreover, the purported bearing of that issue on the dispute over the location of semantic determinants depends upon a mistaken conception of the relation between the nomic status of generalizations and degree of distance between explanans and explananda. Finally, the wider significance of these conclusions for related contemporary debates is noted. And independent considerations about the possibility of communication, also present in Quine’s and Davidson’s thinking, are brought to bear on the question of the location of semantic determinants.
It has seemed intuitive to many people that the possibility of strict laws on the basis of which mental events can be exceptionlessly predicted or explained stands in tension with fundamental features of our commonsense picture of persons. Strict psychological and psychophysical laws have been thought to be inconsistent with the ideas of both free and rational agency – the notions of choice, deliberation and reason seem to make conceptual demands that the existence of such laws would contradict. This paper explains and critically examines presuppositions and purported consequences of one version of this kind of worry. That worry concerns a putative conflict between the idea of strict psychophysical laws and a basic semantic requirement of the concept of rationality: the possibility of error in the applications of one’s concepts and expressions to the world (Sections 1 and 2). One purported consequence is that this conflict bears directly on the dispute between internalists and externalists over the location of semantic determinants – whether the objects and events that determine the contents of psychological states and meanings of utterances are located in speakers’ heads, on the surfaces of their bodies, or in the external world. I argue in this paper that the worry about a conflict between strict psychophysical laws and the concept of error is mistaken (Section 3), and that in any case it does not bear directly on the question of the location of semantic determinants (Section 4). I
Erkenntnis 49: 57–91, 1998. © 1998 Kluwer Academic Publishers. Printed in the Netherlands.
58
STEVEN YALOWITZ
explain why the two issues – psychology’s scientific status and the location of semantic determinants – are conceptually distinct, and sketch a set of considerations for deciding the question concerning semantic determinants that are completely independent of the question of the scientific status of psychology (Section 5). The worry about error can initially be put as follows (see further Sections 2 and 3). To be rational requires, minimally, the capacity to apply concepts. Concepts are individuated by their conditions of correct application – by their conditions of correctness. There must, then, be a difference between applying concepts and expressions correctly and incorrectly. Thus, the possibility of mistaken application – of error – is built into the idea of rationality. Now, one influential approach to determining the conditions of correctness for a concept or expression is a causal theory, which in broad stroke holds that the objects or events that cause an utterance to be tokened by an agent determine the meaning of that utterance. (For simplicity, I shall formulate the problem linguistically, and I shall ignore complications concerning the primacy of thoughts/sentences over concepts/expressions by using ‘utterance’ as a catch-all term). Many of the laws issued by such theories are psychophysical, in covering relations between physical causes – the conditions of correctness – and mental effects – utterance-tokenings. (I focus in this paper on psychophysical semantic laws, which have physical conditions of correctness, only to simplify and sharpen discussion. All semantic laws have psychological properties (i.e., utterance-tokenings) as explananda; the properties in the explanans vary according to the conditions of correctness in question. The considerations in this paper carry over without remainder to these other sorts of semantic laws.) A first pass at a causal, semantic-determining law might hold that if x’s cause an utterance y to be tokened, then y means x. But a strict formulation of such a law would state that ‘for all x’s, if x causes y to be tokened, then y means x’. And in holding that whatever causes a y be tokened determines the meaning of y, no room would be made for cases of error – cases where some z causes y to be tokened, but y nonetheless counts as mistakenly applied to z. Following Jerry Fodor, we can say that the problem faced by these sorts of strict laws is due to the robustness of meaning; many things cause utterances to be tokened that do not determine their meaning (Fodor 1990b, 91). In response to this robustness, one might weaken the strict semanticdetermining law to ‘ceteris paribus, if x’s cause y’s to be tokened, then y means x’. This would avoid the problem of allowing everything that causes the tokening of the utterance to determine its meaning; only certain of those causes would be thought of as meaning-determining.1 And
SEMANTIC DETERMINANTS AND PSYCHOLOGY AS A SCIENCE
59
it would reflect the not implausible assumption that people generally use their language correctly, in the idea that, under normal conditions, what causes an utterance to be tokened are its conditions of correctness. At the same time, it would allow for error – when conditions are not normal, things other than an utterance’s conditions of correctness cause it to be tokened. But besides the difficult problem of distinguishing between the meaning-determining and other causes of tokenings – a problem which continues to exercise these theories – this move leaves the initial worry about the compatibility of error and strictness untouched. And that can suggest a dilemma: either these ceteris paribus laws cannot be sharpened into strict laws without clashing with the robustness of meaning and thus facing the problem of error; or we still need to be shown how they can be so sharpened without that problem arising. The worry about error sees the ceteris paribus feature of those formulations as ineliminable, and forced by the inability of strict causal laws relating conditions of correctness and utterance-tokenings to account for the possibility of error. If, as many people hold, the scientific status of an explanatory framework depends upon its potential for delivering strict laws, then a science of human behavior – and, in particular, a scientific semantics – would be impossible, due to the demands of rationality; a (strict) scientific semantics would be incompatible with the possibility of error. I suggest in this paper that the dilemma stated above forms the basis of a complex debate between W. V. Quine and Donald Davidson over the scientific status of psychology and semantics (see Section 2 for justification of this unorthodox picture of their debate). Davidson takes the first horn, rejecting strict semantic-determining laws, while Quine takes the second, confronting the worry about error head-on. Davidson’s position is motivated by what he perceives as a failure of Quine’s view adequately to account for error while holding on to strict laws. That debate in turn bears directly on their dispute over the location of semantic determinants – whether the objects and events that determine meanings and contents ought to be located in speakers’ heads (internally), on the surfaces of their bodies (proximally), or in the shared public world (externally). Quine’s citation of proximal semantic determinants and Davidson’s citation of external determinants are each motivated in large part by their opposing reactions to the claim about strict laws and error.2 Their reactions trace back to a shared assumption that I shall call the nomic proportionality principle (Section 2). This holds that the degree of distance between events related as ‘cause’ and ‘effect’ is inversely proportional to the nomic status of the covering law: the greater the distance, the less strict the law. In this paper I want to carefully develop and evaluate the claim about error, their reactions to it,
60
STEVEN YALOWITZ
the proportionality principle, and its bearing on the dispute over the appropriate location of semantic determinants. I will argue that the dilemma is misconceived: the possibility of error in and of itself does not conflict with the idea of strict laws. I will also reject the purported connection between this issue and that concerning the location of semantic determinants; the proportionality principle on which that connection rests is false. Following Quine and Davidson, much recent philosophy of mind and language has operated on the assumption that the two issues are intimately linked – that views about the scientific status of psychology constrain or are constrained by views about the location of semantic determinants. For instance, many proponents of the view that semantic determinants ought to be located internally or proximally have argued for this on the basis of the claim that it is required by the demands of a ‘genuinely scientific’ psychology (e.g., Fodor 1981). And that claim typically depends upon the demand that psychological properties supervene on physical properties, or that ‘genuine’ science requires at least potentially exceptionless laws. External determinants are then claimed to be incompatible with both of these demands, while proximal or internal determinants are held to cohere with them. In each case, the idea of strict lawful relations between psychological and physical properties is held to dictate views on the location of semantic determinants. Further, a crucial background assumption is typically that, without the possibility of a scientific psychology, the door to the outright elimination of psychology is opened (e.g., Stich 1983). So the positing of an intimate link between the questions about psychology’s scientific status and its semantic determinants has been held to have critical consequences for the metaphysical status of psychology. It is thus a matter of some urgency to determine the actual relations between the questions. The issues to be examined in what follows, then, are whether error prevents semantic-determining laws from being strict; the bearing of that question on the location of semantic determinants; and the wider debate between Quine and Davidson within which these questions arise. Though I will focus on that debate, I will try to draw out the significance of my discussion for the general issues. I will conclude by sketching a distinct set of concerns about communication, also present in Quine’s and Davidson’s thinking, and explain their bearing on the question of the location of semantic determinants.
1.
For Quine and Davidson, semantic attributions are a species of psychological explanation. Such explanations invoke causal notions at a number
SEMANTIC DETERMINANTS AND PSYCHOLOGY AS A SCIENCE
61
of different levels. First, the reasons cited in explaining an agent’s action explain it by virtue of causing it. Second, psychological predicates are causally defined: attitudes (belief, desire etc.) are causally individuated – the difference between believing and desiring that p is due to the different causal roles that these attitudes play in the etiology of thought and action. And the contents these attitudes subtend are individuated causally as well; generally, what causes the tokening of an utterance determines its content. Quine and Davidson’s debate over the location of semantic determinants concerns the proper causal individuation of contents. Now, questions about causation and causal explanation are pursued, for both Quine and Davidson, within the framework of the Humean and Millian tradition (see further Section 3), which requires strict covering laws for singular causal statements and emphasizes the interest-relativity of causal explanation (Davidson 1990, 77–8, and Davidson 1991, 163; Quine 1974, 7, and Quine 1969, 132ff.).3 It is within that framework that questions are to be formulated concerning the status and determination of linguistic meaning and psychological content. The causes that are cited in commonsense psychological explanations must make the agent’s behavior rational or intelligible – this is the ‘interest’ relative to which such explanations function. And both Quine and Davidson hold that the possibility of mistaken belief is a condition for attributing psychological states to an agent, and thus for making her intelligible (see Section 2).4 I want to focus here on the purported consequences of this requirement of possible error for psychological ascription. For Davidson has in effect argued that mental anomalism – the doctrine that there can be no strict laws upon the basis of which mental events can be predicted or explained (Davidson 1980c) – is a condition for the possibility of mistaken belief. And since this latter possibility is a condition of seeing a creature as rational, mental anomalism is purportedly required if the concept of rationality is to have application. Genuine psychological explanation, then, is held to be incompatible with the idea that psychology can be a science, at least in the way that ultimate physics might be – yielding exceptionless predictive laws that explicitly cite all causally relevant conditions for an effect-type.5 Furthermore, externalism is purportedly required by mental anomalism. Only states whose contents are externally determined can have the capacity for genuinely rationalizing behavior; the possibility of error, a condition for the application of concepts and therefore for being a rational agent, is held to force externalism (Davidson 1984b, 169–70; Davidson 1985, 479– 80; Davidson 1992). So externalism is claimed to be a condition for the possibility of mistaken belief, and therefore of rationality. Mental anomalism thus requires externalism. (See Section 3 for discussion of more
62
STEVEN YALOWITZ
traditional readings of the argument for mental anomalism, which focus on the normativity of practical and theoretical reason, and Section 4 for an alternative reading, which focuses on the background conception of causation). Thus, the external location of semantic determinants is dictated by the (un)scientific status of psychology. And, as we shall see in Section 2, Quine’s opposing semantic views are motivated by opposing views on the scientific status of psychology. In the next few sections, I will lay out and examine the reasoning behind the Davidsonian chain of thought sketched above. However, one caveat is in order at this point. It will be evident that the picture I have sketched is unorthodox when taken as an interpretation of Davidson’s views, in particular his view of mental anomalism, and thus as an interpretation of the debate between him and Quine. Moreover, though Davidson endorses each link in the chain of thought, he never explicitly puts together the chain as a whole. Because of this, its attribution to him is somewhat problematic, and requires justification (see Section 2). However, the core of that chain of thought – that externalism and mental anomalism are conceptually linked – is explicitly endorsed in a very recent discussion (Davidson 1995, 5–6, 12). I shall also offer other textual grounds, as well as some methodological ones, for the attribution. So I think there are plausible reasons for attributing the view to Davidson, and for seeing it as underlying his dispute with Quine. Ultimately, however, I think it is less important whether Davidson would endorse the argument, and more important that it provides a suggestive and illuminating perspective on his discussions of each of these issues, as well as a novel and useful structure for thinking about his disputes with Quine. It is also an interesting and suggestive line of thought in its own right. Indeed, it is a prima facie plausible and novel way of arguing for semantic externalism, and it is worth consideration on this ground alone. In fairness to Davidson, however, I must acknowledge some looseness of fit between the argument under consideration and Davidson’s texts.
2.
The requirement of possible error, which grounds the argument sketched above and plays such a central role in Davidson’s thinking, has its source in Quine’s notion of stimulus meaning in Word and Object, and it is crucial to consider that background in order to come to terms with Davidson’s views on semantics and psychology. Quine holds that while working with that notion as the evidential ground for radical translation, there is no basis for attributing error to an individual considered in isolation from any community. A condition for attributing error to an individual is that a distinction
SEMANTIC DETERMINANTS AND PSYCHOLOGY AS A SCIENCE
63
can be made between what goes into learning the meaning of an utterance and what goes into learning supplementary matters concerning the relevant objects. This distinction, for Quine, requires a communal context. Stimulus meaning, Quine’s basic, objective semantic building block, cannot by itself provide for that distinction (Quine 1960, 47).6 Only by looking for invariances between the stimulus meanings of different speakers can we go on to isolate what might reasonably be called a belief component – ‘collateral information’ – in any particular speaker’s psychology, which is potentially false. The invariances between speakers’ stimulus meanings provide a stable background against which we can pick out idiosyncratic features of particular individuals’ stimulus meanings. It is those idiosyncratic features that form the basis of the concept of belief, which in turn then allows for the applicability of the notion of a mistaken belief. Without relativizing to a community, all we ‘objectively have’ with respect to individuals is, as Quine puts it, ‘just an evolving adjustment to nature, reflected in an evolving set of dispositions to be prompted by stimulations to assent to or dissent from sentences’ (Quine 1960, 39).7 That is, whatever conditions prompt assent and dissent to an utterance will constitute its stimulus meaning; all usage will be in accordance with that meaning, and thus by definition appropriate to its prompting conditions. In the terms of stimulus meaning, then, there is no basis, in an individual’s behavior considered in isolation from her fellows, to attribute linguistic or empirical error. Such error only emerges within a shared linguistic practice: an individual must speak the same language as some others if she is to be attributed mistaken belief. When considering this result, one can reasonably wonder: what is to be gained from working within the constraints that the notion of stimulus meaning imposes on content-ascription? For both Quine and Davidson, one thing that it provides is tight connections between behavior and what, assuming stimulus meaning, is taken to be its cause. Stimulus meaning is, after all, defined in terms of proximal causes – stimulation patterns of nerve endings – and on their shared view, the possibility of tight, predictive laws increases when the causal chains leading from stimulus to behavior are shorter. As Davidson observes, the further out one goes in the causal chain extending beyond a person’s skin, the greater chance there is that the chain can be interfered with, with the result that the appropriate behavior is not triggered (Davidson 1992, 262). The proximal cause thus ‘has the best claim to be called the stimulus, since the more distant an event is causally the more chance there is that the causal chain will be broken’ (Davidson 1989a, 197).8 For Quine, it is precisely this point which motivates the use of stimulus meaning and the appeal to proximal causes; they are
64
STEVEN YALOWITZ
required for a serious science of human behavior, one which approaches strict laws as nearly as possible (Quine 1960, 31).9 This bears out their shared commitment to a nomic proportionality principle, which holds that the degree of distance between cause and effect is inversely proportional to the nomic status (strict or ceteris paribus) of the covering law – thus, laws covering distantly related events can only be ceteris paribus laws. In the present context, that principle derives from the following sorts of familiar considerations. Distal causes in an individual’s immediate environment, such as rabbits, can often go unregistered by her, in which case ‘rabbit’ will not be tokened; conversely, ‘rabbit’ can be tokened though no rabbit is actually present. In the latter case, there is something that causes her to token ‘rabbit’, which is lacking in the former case. In the Cartesian tradition, these two points are generalized and codified in the idea that behavior depends upon – is produced by – what seems to be the case to the individual rather than what is the case. What explains tokenings of ‘rabbit’ is that there seems to be a rabbit rather than that there is a rabbit. For if we want to be able to predict as well as possible when the individual will token ‘rabbit’, we should track the former rather than the latter; we want to correlate the behavior with those events which always bring it about. The semantic version of this point is that whatever exceptionlessly brings about her tokenings gives us the best understanding of what her tokenings mean, since it tells us to what she is ‘actually’ responding. In this respect, Quine’s proximal causes play the same explanatory role as Cartesian ‘seemings’; their proximity to the relevant effects accords them a more exceptionless predictive relation, and thus a more scientific semantic function. And that is precisely what the nomic proportionality principle states.10 Given Quine’s scientific aspirations, this principle forces him to stay within the confines of stimulus meaning when confronting the problem of the robustness of meaning (according to which some of the conditions that cause utterances to be tokened do not determine their meaning, and can thus lead to mistaken usage). Without the background of communal linguistic practice by which to sort through stimulus meaning and isolate ‘idiosyncratic’ stimuli, it is impossible, given those aspirations and that principle, to hold individual usage to be mistaken when considered ‘in isolation’. The subsequent Quinean move to communal embedding, however, does not alter the nomic status of the relation between stimulus and utterance. It just provides an artificial device for distinguishing meaning-determining stimuli from stimuli which provide the basis for error – ‘collateral information’. Rather than, say, one strict law relating individualistic stimulus meaning (i.e., of individuals considered in isolation) to some utterance-
SEMANTIC DETERMINANTS AND PSYCHOLOGY AS A SCIENCE
65
type, there will now be a number of strict laws. One of these (the ‘semantic’ law) is arrived at through comparison of individualistic stimulus meanings of ‘fellow speakers’ and covers stimulus-response patterns which are invariant across that population with respect to that utterance-type. Other strict laws concerning that utterance-type will be individualistic, covering the idiosyncratic features of individualistic stimulus meaning which vary across individuals. A ‘serious’ science of human behavior, in confining itself to proximal stimuli, will still deliver (or endeavor to deliver – see Sections 4 and 5) strict psychophysical laws. I want to focus on the relationship between Quine’s four claims here: the impossibility of assigning error to individuals’ usage considered in isolation, the demand for strict predictive psychophysical laws, the nomic proportionality principle, and the citation of proximal causes in the determination of meaning and explanation of behavior. I will first explain how these features of Quine’s picture can form the background for a move to mental anomalism as well as to distal causes – externalism – in the determination of linguistic and mental content. I will then turn to the question of the extent to which such a move can plausibly be attributed to Davidson. The initial point is to observe that Quine’s communal response to the problem of squaring error with nomic causes is unsatisfactory. There are a variety of reasons for thinking this: it fails to make sense of the objectivity of error, in equating communal usage with correctness; it also appears to lead to a form of relativism. These and other worries have been expressed about similar readings of Wittgenstein (Kripke 1982), and have led to widespread rejection of communal solutions to the problem of error. The immediate problem this creates for Quine’s overall view is that it is difficult to see how else one might square nomic semantic determinants with the possibility of error within his framework. As we saw, it is their nomic status that generated the problem of error for Quine’s proximal semantic determinants. This suggests that only anomic semantic determinants will allow for error. With the proportionality principle in place, this in turn appears to push semantic determinants further out into the world than Quine’s proximal causes, so that semantic determinants do not have such an exceptionless relation to utterances. We end up with the following chain of reasoning: rationality demands the possibility of error; it thus requires anomic semantic determinants; anomic semantic determinants are externally located; thus, externalism is a necessary condition for rationality. Let me expand this line of thought a bit. The basic idea is that to the extent that we want to be in a position to hold agents to be mistaken in their application of utterances and their beliefs, laws on the basis of which
66
STEVEN YALOWITZ
their linguistic and mental contents are determined cannot be strict. Only ceteris paribus generalizations have the sort of slackness that can allow for the concept of error to take hold. For it is purportedly only when there can be cases where the antecedent of a semantic-determining psychophysical generalization relating objects or events to utterances or thoughts is met, but the consequent is not, that we can be in a position to say that error has occurred: cases where, even though z caused y to be tokened, y did not mean z but rather x. On this view, the semantic-determining law ‘if x’s cause tokenings of y, then y means x’ could only have a ceteris paribus status. The further claim is that given the proportionality principle, it is only by appealing to distal causes that semantic-determining laws will be sufficiently slack.11 That principle links the questions of the location of semantic determinants and psychology’s scientific status. Further, although the issue here is formulated with respect to a particular kind of psychophysical law – governing the determination of linguistic content by causes – one can see how these considerations can extend to non-semantic psychophysical laws, governing, for instance, the relation between ‘primary reasons’ and actions. And it seems plausible to think that the impossibility of strict semantic laws would transfer over to, and thus restrict the scope of, any laws in which states with content (propositional attitudes) figure. Indeed, this intermeshing of semantic with other types of psychological and psychophysical laws should be unsurprising given the Quinean rejection of the analytic-synthetic distinction endorsed by Davidson, and the subsequent inextricability of meaning and belief.12 The ceteris paribus status of all laws in which mental properties figure can thus be claimed to be a condition for satisfying the demand for the possibility of error. However, I will continue to focus only on semantic-determining psychophysical laws. I have noted that this line of argument is quite unorthodox as a reconstruction of Davidson’s picture, especially his views on mental anomalism. I have also conceded that Davidson never asserts such an argument at any one place in his writings. I therefore want to pause here and consider the question to what extent it can plausibly be attributed to him.13 First and foremost, Quine and Davidson agree that error cannot be made sense of when considering an individual’s linguistic usage in isolation (Quine 1960, 39, 47; Davidson, 1992). Second, as we have seen, for Quine the possibility of error requires embedding an individual’s usage within a shared linguistic context. For Davidson, however, this sort of communal solution to the problem of error leads to an objectionable sort of relativism which is inadequate for a genuinely objective notion of error (Davidson 1984, xix; Davidson 1990a, 309).14 The possibility of error cannot be due
SEMANTIC DETERMINANTS AND PSYCHOLOGY AS A SCIENCE
67
merely to the relationship between different individuals’ overlapping linguistic dispositions. It thus must be reflected directly in the individualistic semantic-determining laws themselves. Third, Quine and Davidson also agree in their endorsement of the nomic proportionality principle; as we have seen, this entails that it is by virtue of their proximal location that Quine’s semantic determinants are nomic (Quine 1960, 31, 28–30; Davidson 1992, 262 and Davidson 1989a, 197). These three points constitute the central moves in my reconstruction, and are clearly present in Davidson’s thinking. So, more generally, are all of the claims in my reconstruction: the need for possible error; the failure of solitary individuals’ proximal causes to allow for error; the rejection of the Quinean communal solution to this problem; the rejection of nomic semantic-determining laws; the endorsement of the proportionality principle; and, finally, the endorsement of distal semantic determinants. The only interpretive issue is how to conceive the relationship between them, and, more generally, how to conceive the relationship between Davidson’s rejection of the move to shared communal language, and his endorsement of mental anomalism and externalism. That relationship, in my reconstruction, is fixed by highlighting the proportionality principle. This principle plays the crucial role in mediating the relationship between Quine and Davidson’s shared rejection of the possibility of a solitary individual being in error, on the one hand, and their divergent attitudes towards anomalism and externalism on the other. Given the principle, Quine can hold on to a scientific psychology only by retaining proximal determinants; he is thus forced to the communal solution to the problem of error. Relative to Quine’s position, Davidson’s rejection of communal accounts of error leaves no nomic alternative for solving that problem. Anomalism follows directly, and this together with the proportionality principle leads to externalism. The proportionality principle thus links the questions of the location of semantic determinants and psychology’s scientific status. Perhaps it would tell against highlighting this principle and using it to link these various views together in this way if there was other evidence that made it clear that no connection existed for Davidson between these two questions. But a recent discussion of his actually provides direct textual support that he endorses such a connection (Davidson 1995). He agrees with the view, held by Fodor and others, that narrow (i.e., nonexternally determined) ‘content’ is required for the idea of a scientific ‘psychology’ in which lawlike connections are made between mental and physical properties. Davidson then explicitly claims that externalism stands ‘in the way of a serious science of psychology’ (Davidson 1995, 6). He notes that externalism introduces an irreducibly causal element into psy-
68
STEVEN YALOWITZ
chology, as a result of which psychology ‘cannot hope to emulate physics, which has striven successfully to extrude all causal concepts from its laws. Externalism sets limits to how complete psychological explanation can be . . . ’ (Davidson 1995, 12; emphasis added). This makes it clear that he sees an explicit connection between the questions concerning the location of semantic determinants and the scientific status of psychology.15 I offer one final reason for my reconstruction before moving on to critically evaluate its lines of argument. Consider first the starkest disagreement that Davidson has with Quine: whether stimulus meaning should be the starting point in semantics and psychology. Stimulus meaning, as I have emphasized, is inextricably linked for Quine with a communal conception of language, a scientific conception of psychology, and a proximal conception of semantic determinants. Now, it is uncontroversial to see Davidson’s overall philosophical views as emerging out of a sustained reflection on this Quinean background. Consider now the fact that in rejecting stimulus meaning Davidson also ends up rejecting communal conceptions of language, scientific conceptions of psychology and proximal conceptions of semantic determinants. My reconstruction provides an extremely plausible explanation of this dialectical situation: Davidson is recognizing the tight connection between these issues in Quine’s thinking, and this gets reflected in his views on anomalism and externalism. For this and previously stated reasons, I will attribute the line of argument in my reconstruction to Davidson, while acknowledging the somewhat speculative status of this attribution. Davidson’s commitments and explicit statements, then, suggest the intriguing claim that both mental anomalism as well as externalism are required for the sort of slackness that the requirement of the possibility of error imposes on semantic-determining laws. Rationality, anomalism and externalism thus stand in a chain of presupposition; the initial point concerns the possibility of error, a presupposition of rationality. Suppose Davidson is right about this initial point and its bearing upon the question of the possibility of strict laws. Suppose that he is also right about the relations between citations of proximal causes and the possibility of strict laws, on the one hand, and citations of distal causes and the impossibility of such laws, on the other. Then a plausible connection will have been made between normativity – in the semantic sense of error focused upon here – mental anomalism and externalism. Now, the connection between rationality and error seems unproblematic, given the idea that conditions of correctness, which individuate contents, entail conditions of incorrect usage; I shall not question it here. But the two further claims are extremely problematic. These concern the relation between slackness in predictive
SEMANTIC DETERMINANTS AND PSYCHOLOGY AS A SCIENCE
69
accuracy and the possibility of error, on the one hand, and the relation between strictness of laws and location of causes, on the other. In Section 3, I will argue against the purported connection between error and predictive slackness that would ground mental anomalism in considerations regarding the normativity of rationality. In Section 4, I will then argue against the proportionality principle’s claim that there is a direct connection between strictness of laws and locations of causes of a kind adequate to underwriting either Davidson’s external or Quine’s proximal semantic determinants – or, for that matter, other philosophers’ more internal ones.
3.
The semantic argument for mental anomalism that I have laid out is quite different from the sort of argument typically associated with this Davidsonian doctrine. Before turning to the purported connection between the possibility of seeing agents as potentially mistaken and the impossibility of strict psychophysical laws, it is worth briefly noting the relationship between these different argumentative strategies. According to what has by now become the orthodox reading of Davidson’s doctrine, it derives from the constitutive, normative role of rationality in psychological attributions. That role is claimed to block the possibility of strict lawful relations between mental and physical properties. However, the orthodox reading is not concerned with the semantic sense of normativity – the possibility of error – that was the focus of the argument of Section 2 at all. It is concerned exclusively with a sense of normativity tied to practical and theoretical reason. For instance, Jaegwon Kim’s rendition turns on the idea that the existence of psychophysical bridge laws would threaten the different essential features of psychological and physical properties and principles (such as explanatory scope and modal status), by importing incompatible aspects of each to the other (Kim 1985). Though it is clear from Kim’s discussion, it is worth emphasizing that the psychological principles in question here are not semantic-determining ones. William Child’s rendition turns on the idea that principles of rationality which govern the ascription of mental properties are ‘uncodifiable’ – incapable of delivering a detailed answer to every question concerning what an agent ought to believe or do in particular circumstances – due to the nature of practical and theoretical reasoning (Child 1993). This is claimed to be unlike the codifiable principles that govern the ascription of physical properties, which do deliver such detailed answers. The anomalism of the mental results from its uncodifiability. Thus, both Kim’s and Child’s accounts focus on the sense of normativity that attaches to practical and theoretical reasoning, rather
70
STEVEN YALOWITZ
than on the semantic sense of normativity at issue in the argument laid out in Section 2. These two notions of normativity are no doubt related in important and complex ways, but it is nonetheless worthwhile distinguishing them; especially so, since Kim’s and Child’s arguments fail both on internal, interpretative grounds and on external grounds of philosophical plausibility. I have argued this at length elsewhere, and do not propose rehearsing the relevant issues here (Yalowitz 1997a). But given this situation, the semantic argument for mental anomalism that is the subject of the present paper takes on added importance; it seems unlikely that any further, more promising arguments based on the notion of rationality are forthcoming. I have, however, argued elsewhere that a different argument can be garnered for mental anomalism within Davidson’s (and, curiously, Quine’s) framework, from considerations having nothing particularly to do with rationality (Yalowitz 1998a). After arguing in this section against the semantic strategy for grounding mental anomalism of Section 2, I will outline this alternative argument (Section 4), since it will be useful in evaluating the claim about the relation between location of causes and strictness of laws. I turn now to the semantic argument for mental anomalism. I noted earlier that Quine shares Davidson’s view that the better we can predict an individual’s behavior (i.e., the stricter the laws we use to do so) the less reason we have for holding her mistaken. The idea is that if the stimulus conditions which cause an agent to token an utterance determine its content (ignoring negative stimulus conditions for the sake of simplicity), then (given that causal-types are sufficient for their effect-types) a strict psychophysical law would enable us to predict, without exception, when a speaker will token that utterance, and a tokening of that utterance will always be true since its content is determined by what causes it. I have noted that Davidson rejects Quine’s subsequent move to embedding the speaker within a shared linguistic community as a way of artificially resuscitating the notion of error,16 and therefore claims that predictive slackness – the lack of strict psychophysical laws — is a necessary condition for error.17 Davidson must then hold that it is only what ‘normally’ causes tokenings of such types that determines their content; in other words, it can only be a selection from the full sufficient condition for tokenings of an utterance-type that determines its content. John Mackie’s useful notion of an ‘INUS’ condition (Mackie 1974, 62) – a condition that is an insufficient but necessary part of an unnecessary but sufficient condition for some effect-type – provides a rough idea of what is at issue here (though see further below).18 The INUS conditions appealed to in causal explanations reflect the particular interests that those explanations serve. Thus,
SEMANTIC DETERMINANTS AND PSYCHOLOGY AS A SCIENCE
71
what determines an utterance’s content is not really what, strictly speaking, causes its tokening – in the sense of providing a sufficient condition for it – but rather what, in the interests of psychological explanation, makes it intelligible (on this point see also Yalowitz 1997b, Section 4). This line of reasoning boils down to the following. The citation of an INUS condition (and not the full, sufficient condition – ‘the’ cause) for the tokening of an utterance-type is a condition of determining that utterance’s content; otherwise, misuse of that utterance-type will not be possible, and thus a condition for rationality will not be met. Given Davidson’s assumption about the connection between conditions for content-ascription and conditions of rationality, an utterance only has content if it is only an INUS condition that is appealed to in explaining its tokening. Explaining the tokening of an utterance by appeal to a full, sufficient condition would be an obstacle to determining what content it has, and, indeed, preclude seeing it as having content. That is the legacy of Quine’s views on these matters which Davidson is concerned to avoid, since Quine’s own compensatory move to communal language seems to entail epistemological relativism. At this point, we need to focus on the purported connection between this claim about the need for INUS conditions in explaining content, on the one hand, and mental anomalism, on the other. In particular, it is not clear why an INUS condition cannot be appealed to in explanation of the content of a tokening even when that INUS condition is nonetheless flanked, or potentially flanked, by other INUS conditions that, together, form a fully articulated sufficient condition for a tokening of that type. (Full articulability of the antecedent is what distinguishes strict from hedged laws.) Mental anomalism denies the possibility of strict laws on the basis of which mental events can be predicted or explained, while what Davidson is claiming here suggests something weaker: that only INUS conditions (and not full, sufficient conditions) can constitute contents. It does not seem to follow from this latter point that there cannot be strict laws on the basis of which an agent’s tokening some utterance can be predicted or explained. All that follows is that the content of that utterance is determined, not by all the conditions which together form the antecedent of such a law, but rather by a selection from that set of conditions. And the idea that causal explanation is interest-relative, of which this point is simply an instance, cannot by itself rule out strict laws; selection occurs relative to those interests, but it can, for all that, be selection from a fully articulable sufficient condition.19 So the point about content-determination does not seem to be in the appropriate sort of conflict with strict psychophysical laws in order to underwrite mental anomalism. We have not been given a reason for thinking that there cannot be a fully articulable sufficient
72
STEVEN YALOWITZ
condition for the tokening of utterance-types. And that is what a strict, semantic-determining psychophysical law would be. I will now consider two possible Davidsonian responses to this line of thought. First, Davidson might object that ‘tokening an utterance’ is not a mental (description of an) event at all – though holding an utterance true is – and that not all utterance-tokenings are instances of holding an utterance true. At issue here is the dispute between Quine and Davidson over the basic evidence for the ascription of content (i.e., radical interpretation). Quine holds that this evidence must be understood in terms of assents and dissents to utterances, which are behaviorally scrutable and non-intentional (‘surface-assent’), while Davidson claims that the holding-true of an utterance, an explicitly intentional attitude, is the basic evidence (Quine 1975, 91; Davidson 1984a, 135). According to Davidson, it is the relations between the latter and particular conditions in the world that determine the utterance’s content. These conditions themselves fall short of forming a full sufficient condition for holding an utterance true; but they are not INUS conditions either, since (unlike INUS conditions) there purportedly is no fully articulable sufficient condition of which they are a part. This is what anomalism amounts to generally – the claim that certain properties (conditions) cannot figure in generalizations in which the antecedents can be fully articulated so as to yield an exceptionless predictive law. There can be occurrences of those antecedent conditions which are articulated that are not followed by the consequent of the law. Laws in which anomic properties figure must then be qualified by an ineliminable ceteris paribus label. In contrast to INUS conditions, then, I shall therefore call anomic properties ceteris paribus (CP) conditions. CP conditions in the antecedent of a law are insufficient and non-redundant parts of an unnecessary and non-articulable sufficient condition for some effect. But effects of laws can also be CP conditions, if they are anomic – if there can be no fully articulated antecedent on which they exceptionlessly follow. For Davidson, then, mental properties such as holding an utterance to be true are CP conditions. Therefore, a strict law in which a distal-type figures as an INUS condition for the effect of some ‘utterance-tokening’ would not be a counterexample to mental anomalism, since it would not be a psychophysical law; no mental event-type would appear in its consequent. On this line of reasoning, a condition for a mental event-type occurring in the consequent of a semantic-determining law is that it be a ceteris paribus law, with its antecedent formulated in terms of CP conditions. To give a fully articulable sufficient condition (the antecedent of a strict law) for the ‘tokening of an utterance-type’ is to preclude assigning that utterance any content; indeed, it precludes describing it as the holding-true of an utterance.
SEMANTIC DETERMINANTS AND PSYCHOLOGY AS A SCIENCE
73
This response, however, begs the very question at issue. That issue is whether there is a connection between a condition of rationality – error – and mental anomalism; but mental anomalism cannot be presupposed in establishing that connection without making the argument circular. And to deny that tokenings of utterances are type-identical with mental events such as holding utterances true can in the end be based on nothing other than a denial of the possibility of laws connecting these two sorts of eventtypes. Davidson’s rationale for insisting, against Quine, that the basic evidence for content-ascription must be intentionally described (i.e., holdingtrue vs. surface assent) can ultimately be nothing other than the doctrine of mental anomalism. For it is only if not all instances of surface-assent (tokening an utterance) are instances of holding an utterance true – and thus that the latter type is not reducible to the former – that Davidson is in a position to reject Quine’s method of content-ascription and insist upon his own. Quine’s method depends upon a reductionist conception of contentascription – that it can be grounded in evidence that is described nonintentionally. But what reason can Davidson have for rejecting Quine’s method except the doctrine of mental anomalism? It is circular to appeal to the claim that holding-trues cannot be reduced to surface-assents in order to support that doctrine and its purported basis in rationality. That claim assumes mental anomalism, which then requires independent support. In order to avoid the circularity of this first response, Davidson might consider a second way of connecting mental anomalism’s insistence that only CP conditions can constitute contents with the requirement of possible error. According to this more general line of thought, there is a problematic sort of relativism that plagues the idea of ‘selecting’ explanatory conditions from a fully articulable sufficient condition. It is only if the ‘selection’ is in some sense forced on one that one has the right to say that the factors ‘selected’ constitute contents that can underwrite an adequate notion of error. In illustration of this, consider again Quine’s communal solution to the problem of error. The stimulus meaning for an individual considered in isolation from any communal context is the fully articulated set of conditions that will prompt assent and dissent to an utterance-type. In order to account for the possibility of mistaken utterances, Quine has us isolate, from this fully articulated stimulus meaning, a class of stimulustypes which are shared between a particular group of individuals. The leftover, idiosyncratic stimulus-types are labeled ‘collateral information’, and can trigger fallible utterances. Now, different classes of stimuli could be isolated as ‘the meaning’ in this way relative to different communities, either counterfactually or actually. Counterfactually, we can imagine the individual interacting with a different group of interlocutors whose over-
74
STEVEN YALOWITZ
lapping class of stimulus-types is different than the actual case. In this counterfactual case, some of the individual’s utterances that are true ‘by definition’ in the actual world could be false, even though nothing about that individual herself, or the world to which she is reacting (other than her interlocutors), had changed.20 And even in the actual world, there can be cases where there are different classes of overlap (different ‘dialects’) in different sectors of a community of interlocutors. Relative to one sector and ‘meaning’, an individual’s utterance would be true by definition, while relative to another it might be false. The question then arises: how could an individual’s utterance on a given occasion be a genuine error, mistaken about the objective world (relative to Quenglish) when it is also a genuine truth (relative to English)? Perhaps some indeterminacy is unavoidable, but this is too much. If all that ‘truth’ and ‘error’ amount to is going on, or failing to go on, in the same way as others – and one could just as easily be seen as going on in one way (relative to one community) as another – then an unsatisfying epistemological relativism seems to follow directly. The notion of ‘error’ that emerges from relativizing an individual’s usage to a community of interlocutors seems inadequate for capturing our intuitive conception of objectivity (Davidson 1984, xix; Davidson 1990a, 309 – for discussion, see Yalowitz 1998b). The leading thought now is that this result could be avoided only if there was no fully articulable sufficient condition for the tokening of an utterance-type. In this case, only conditions insufficient for such types could be correlated with them to any extent at all; these correlations would thus hold only for the most part, not exceptionlessly. It would then be required that we think of those conditions as constitutive of content; no selection (in anything like the Quinean fashion) of INUS conditions would be occurring at all. This argument, then, holds that in order to make sense of the possibility of genuine error, conditions constitutive of content must be CP and not INUS conditions; there can be no fully articulable sufficient condition of which they form a part. This line of thought, then, would justify the central claim at issue, that predictive slackness – and therefore mental anomalism – is a condition for the possibility of genuine, objective error. This is a powerful line of argument. The philosophical intuition it reflects is deep and tenacious, and it demands explanation (see further below). But there are two points to make in response to it. First, if the worry about relativity of selection is the central objection to thinking of INUS conditions as determining contents, then the argument is too strong, since it would rule out all causal explanation that deals in INUS conditions; all such explanation, within Davidson’s Humean-Millian framework, involves a selection from the total sufficient condition for some effect-type, one
SEMANTIC DETERMINANTS AND PSYCHOLOGY AS A SCIENCE
75
relative to our interests and conceptions of ‘normal conditions’. All such explanation would end up being anomic, with the conditions appealed to being CP and not INUS conditions.21 But this cannot help Davidson, since the point of focusing on error was supposed to be in the service of showing that conditions unique to rationality – the possibility of error – and therefore psychology are responsible for mental anomalism. But the considerations offered too easily generalize. It is the purported ‘relativism’ associated with the selection of INUS conditions that conflicts with ‘genuine’ error and thus supposedly rules out strict semantic-determining laws; but that relativity afflicts all explanation, not just psychology. It is the notion of ‘genuine’ explanation that is really at odds with selectivity; the conflict with ‘genuine’ error is just an instance of this general problem. Suppose there have been a series of fires in an inexpensive and unsavory apartment complex; the tenants have all been informed of low-level gas leaks which, while not dangerous to inhale, could lead to fires – no smoking, or open fire generally, is allowed. As it happens, serious cigarette smokers could not resist lighting up. Now, in the explanation of the fires, appeals to either the gas leaks or the lighting of cigarettes could be claimed to be in a sense arbitrary: relative to the landlord’s interests, it was the latter; relative to the board of health it was the former. Neither, strictly speaking, is a true citation of ‘the cause’, since each cites something which falls short of a full sufficient condition. Does that mean that neither explanation is ‘genuine’, that both are illusory? If that is so, then there is very little (if any) ‘genuine’ explanation that goes on in the sciences. This points up the completely general nature of the selection problem. Therefore, invoking this problem in the argument from the possibility of error to mental anomalism cannot do the work required, of grounding mental anomalism in a property unique to psychology. If there is a problem here, it is due to the interest-relativity of causal citation and its supposed conflict with the notion of ‘genuine’ explanation; it is not due to error in particular. There is a second problem with this attempt to ground mental anomalism in the possibility of error, related to the previous one. The selection problem arises even with ceteris paribus laws and CP conditions. It is rare that we are aware of only one CP condition roughly co-varying with some effect-type, so that we are forced to appeal to it in explanation because of lack of alternatives. The fact that we are hardly ever forced by the evidence to cite one explanatory condition is a completely ubiquitous phenomenon. So mental anomalism cannot count as a solution to the selection problem being posed for the possibility of strict semantic-determining laws, since the grounds offered for rejecting such laws would apply equally to their
76
STEVEN YALOWITZ
ceteris paribus counterparts. And Davidson has no objection to these latter sorts of laws (Davidson 1980c, 216). The selection problem, and thus the relativism objection, therefore fails to do the work that is required if it is to ground mental anomalism in rationality as opposed to more generic features of psychological generalizations such as mere interest-relativity (see Section 4). How serious is it as a general challenge to the explanatory force of INUS (or CP) condition citation? Besides the fact, noted above, that it calls into question most if not all of science, more importantly, it completely fails to make contact with the idea that causal citation is relative to explanatory aims and interests – an idea endorsed by both Quine and Davidson (Section 1). Failure to see ‘genuine’ explanation due to selectivity of causal factors depends almost entirely on having not related the INUS or CP conditions selectively focused on to the interests that the proffered explanations are addressing. In any case, the notion of ‘interest-free’ explanation is at best an ideal; without it, a harmless underdetermination (and perhaps some residual indeterminacy) remains, but this can hardly sustain a compelling challenge to actual explanations. Now, it is true that there also remains the pressing question which interests should be in the picture – what the aims of enquiry ought to be. I do not propose entering into a discussion of this thorny problem here, though I do doubt that there is, or needs to be, an informative general answer. However, I shall be suggesting, in Section 5, that the question of how communication is possible – a perfectly legitimate aim of enquiry – is an essential ‘interest’ which informs commonsense psychological explanation, and that it is crucial for settling the debate concerning the location of semantic determinants. In this section I have argued against the attempt to ground mental anomalism in the possibility of error, and thus in the semantic notion of normativity. I argued that the plausible point that content is determined by a selection from the full sufficient condition for an utterance-tokening (and not the full sufficient condition itself) is not by itself in conflict with the idea of strict semantic-determining laws. And I rejected possible replies to this argument deriving from Davidson’s views on radical interpretation and relativism. I did concede a powerful intuition behind the semantic argument that requires some explanation. If mental anomalism is indeed true, then it will be the case that many features of psychology are at odds with the idea of strict psychophysical laws; error is a case in point. But it is one thing for a particular feature constitutive of psychology to stand in tension with that idea, and quite another for that feature to be the basis for the rejection of strict psychophysical laws. The tension may be due to some more basic property which is presupposed by that particular feature.
SEMANTIC DETERMINANTS AND PSYCHOLOGY AS A SCIENCE
77
One has to be careful in sorting out what is doing the explanatory work in grounding irreducibility, and not confuse ground and consequence. In the present case, this is illustrated by the fact that the purported conflict between error and strict semantic-determining laws is due to the notion of a ‘genuine’ explanation presupposed by the notion of ‘genuine’ error. And as we shall see in Section 4, on my alternative story of how mental anomalism is grounded in the causal definition of mental properties, the possibility of error will conflict with strict psychophysical laws because the notion of error presupposes certain causal notions – in particular, the notion of something determining a content by causing it.
4.
The semantic notion of normativity cannot, then, ground mental anomalism. Moreover, as I noted in Section 3, the notion of normativity attaching to practical and theoretical reasoning looks no more promising. I do, however, think that a compelling argument for mental anomalism can be garnered on Davidson’s behalf. This makes it worthwhile to examine the issue of the bearing of mental anomalism on the question of the location of semantic determinants, even if the argument from error fails. Since this alternative argument for mental anomalism turns on issues quite different than normativity, it will be less confusing if I come back to it after discussing the question of the location of semantic determinants. For the purposes of this discussion, then, I will grant Davidson’s general claim that psychological properties are anomic, and put aside the question of what the basis of that claim is. The present issue is not the grounds for this claim, but its consequences: can the claim motivate Davidson’s citation of distal semantic determinants, and its denial motivate Quine’s citation of proximal ones? I have emphasized that Davidson’s rationale for distal types is rooted in the proportionality principle he shares with Quine. It is expressed by the thought that one has to take into account much more in the way of potential interrupters when attempting to formulate laws relating spatially distant event-types than laws relating more spatially contiguous ones. I shall now argue that this thought is inadequate for justifying Davidson’s claim that distal causal types are necessary in order to account for the anomic status of a property. Three points show that distal types are neither sufficient nor necessary for anomicity. And they have a correspondingly negative effect on Quine’s tying of proximal causes to strict laws. First, complexity of factors is not sufficient for anomicity; the fact that more potential interrupters need to be taken into account when citing distal
78
STEVEN YALOWITZ
types does not prevent the formulation of a strict law, only facile formulation. There is no reason to think that citation of a distal type in the antecedent of a generalization is by itself incompatible with articulating a set of accompanying conditions which together provide a sufficient condition for the effect. And such an articulation would give us a strict law citing distal causes.22 This point becomes transparent when one considers the question of the status of causal chains in the debate between Quine and Davidson, which has been suppressed in the previous discussion. Suppose first that saying that some distal event-token causes an utterance-token amounts to positing intervening causal chains in virtue of which that causal relation holds. Then, according to the requirement of strict covering laws for singular causal statements (Section 1), one must posit a strict law covering each of these intervening chains, since each is a singular causal relation. However, since causation is a transitive relation, to say that there is a strict law relating each link of the overall chain to its successor entails that there is a strict law relating the initial and terminal events. Therefore, distance is compatible with (indeed, requires) a strict-law relation when one considers tokens; this follows directly from the strict coveringlaw requirement. When we now consider this issue from the perspective of types and not tokens, the situation is no different. To say that some distal-type causes an utterance-type, on the causal chain view, requires that causal relations obtain between any types that mediate that distance. If each of these mediating type-relations are strict – if, that is, they express a strict law – then, once again, transitivity will entail that the initial distal and terminal utterance-types are related by a strict law. And there is no reason to deny that these mediating links can be strict at present.23 So the mediation of distal causal relations by chains, at both the token and type levels, is compatible with strict covering laws. Suppose, instead, that one rejects the requirement of mediating chains for causal relations, and allows for genuine causation at a distance. Then my initial point holds: while the initial distal type is not by itself a sufficient condition – it is at best an INUS condition – there is no reason to think that its inclusion in the antecedent of a generalization is by itself incompatible with articulating a set of accompanying conditions that together provide a sufficient condition for the effect. And such an articulation would give us a strict law citing distal causes. The distal location of the causal determinant-types of contents cannot, therefore, provide a sufficient condition for anomalism and the possibility of error, whatever one’s views on causal chains. Second, distal types are not necessary for anomalism either. This can be seen by noting that Quine’s proximal causal-types are not spatially contiguous with their effects, assent-dissent behavior. I will be discussing this
SEMANTIC DETERMINANTS AND PSYCHOLOGY AS A SCIENCE
79
point in more detail in Section 5, so here I will be brief. Quine rejects internal, neural causes in favor of proximal ones because he supposes that the latter, but not the former, can account for the public accessibility of semantic determinants (Quine 1960, 31; see also 5–8). The refusal, from a semantic perspective, to cite types further inside the head means that any requirement of ‘direct’ contiguity (however that might be made sense of) between semantic explanans and explananda is given up. And that means that there is no relevant qualitative difference between proximal and distal types with respect to nomic status; both depend upon mediating causal chains, and thus require that potential interrupters be taken into account. So even if, contrary to my first point, the need for such accounting were sufficient for anomicity, it would not motivate Davidson’s distal causal-types over Quine’s proximal ones. Third, contiguity by itself is in any case clearly not enough to guarantee an effect-type. And given the central assumption about causation – the requirement of strict covering laws – such a guarantee is what is needed in order for the generalization to be a strict one. As we saw in Section 3, the antecedent of a strict law must fully articulate conditions that guarantee that the effect-type will occur. And the typical kinds that are cited in explanations and generalizations, even when constrained by a contiguity requirement, do not play this sort of role; they do not state genuinely sufficient conditions. Therefore, even if sense could be given to the idea that certain internal events did not depend upon mediating chains in causing behavior,24 they would not be sufficient for picking out ‘the causes’ of utterances. And my discussion of causal chains makes it clear that the citation of such internal events is not necessary for strict laws. Quine’s proximal causes are, then, not in conflict with Davidson’s demand for anomic slackness. To that extent, Davidson’s move to distal causes is unmotivated – it is not required in order to satisfy the demands of mental anomalism. That by itself is sufficient for rejecting their shared thought that the nomic status of psychology bears directly on the dispute over the location of semantic determinants, and the proportionality principle on which it is based. I will return to the question what else might settle that dispute in Section 5. But I want first to sketch an alternative argument for mental anomalism that surfaces at certain points in both Quine’s and Davidson’s writings; this argument has independent interest, but it also serves to solidify what I have so far argued in this section. Given space limitations, I can only lay out the argument’s contours briefly and dogmatically (see Yalowitz 1998a). This alternative argument focuses on the conception of causation and causal explanation that forms the background for both Quine’s and David-
80
STEVEN YALOWITZ
son’s views about science. It holds that the causal nature of psychological properties and the strict nature of the types of laws being denied by Davidson is sufficient for the doctrine of mental anomalism. As noted in Section 1, psychological predicates are causally defined: both attitudes and contents are individuated causally, and reasons explain behavior by causing it. Causally defined predicates, however, are not fit for inclusion within strict laws (recall Davidson’s remark, noted in Section 2, that external semantic determination is sufficient for anomalism). The application of such predicates is necessarily bound up with ceteris paribus conditions.25 Since such predicates are unfit for inclusion within strict laws, there can therefore be no strict laws (purely psychological or psychophysical) on the basis of which mental events can explain, be predicted or explained. Mental anomalism is thus to be traced back to generic features (i.e., causal definition) of psychological properties; features that they share with many other sorts of properties. Thus, no route through controversial conceptions of rationality is needed to ground mental anomalism. The formulation of strict laws, then, requires very different kinds of predicates – they must be non-causally defined. We may not know any such predicates, but the commitment to the principle of the nomological character of causality entails that we know that there must be such predicates.26 In their own ways, this conception of a strict-law property is expressed by Hume (1888, 86), Quine (1974, 14; 1979, 165–66) and Davidson (1989b, 5) (for discussion, see Yalowitz 1998a). And Quine and Davidson agree that psychological predicates are causally defined, and thus unfit for strict-law inclusion (Davidson 1991, 162; Quine 1986, 397–98, read alongside Quine 1974, 12). With this conception of strict-law properties in mind, I now want to turn back to the purported connection between location of cause and strictness of law. The central point to notice is that the properties that are used to pick out Quinean proximal causes (e.g., stimulation of ocular nerves) are causally defined: defined in terms of the sorts of events which they typically result from and produce. Therefore, given the story I glossed above about mental anomalism being due to the generic, causal features that psychological properties share with many other properties – a story that surfaces in both Quine’s and Davidson’s writing – the properties in Quine’s picture will themselves not be fit for inclusion within strict laws. That causal definition is shared by the concepts of the special sciences generally, including Quine’s behavioristic psychology; it distinguishes them all, as a group, from ‘ultimate physics’ (Davidson 1987, 45–6). On this view, it cannot be mere location itself, but rather the properties in terms of which events and objects in a location are picked out, that decides the question of nomic status. The vocabulary suitable to strict laws can only refer to properties
SEMANTIC DETERMINANTS AND PSYCHOLOGY AS A SCIENCE
81
which are non-causally defined, and this rules out Quine’s description of the proximal causes. Stimulus meaning, Quine’s basic, objective semantic building block, is anomic; it is a CP and not an INUS condition.27 Given their causal individuation, then, Quine’s proximal causes are once again not in conflict with Davidson’s demand for anomic slackness. To that extent, Davidson’s move to distal causes is again unmotivated – it is not required in order to satisfy the demands of mental anomalism. It is only if proximal causes could be individuated in non-causal terms that they could figure in the strict laws that Davidson insists we must reject and Quine insists we must accept. Similarly, it is only if distal causes could not be individuated in non-causal terms that they could not figure in strict laws. Now, it is important to be clear that the issue here concerns types and not tokens. I already noted, in the above discussion of causal chains, that for causally-related tokens distance is irrelevant to nomic coverage because of the transitivity of causation – all tokens in a causal chain are covered by strict laws, no matter how distant they are from each other. At the level of types, the situation is different, though no better for Quine’s and Davidson’s purposes. If, as I have noted, both Davidson’s distal types as well as Quine’s proximal types are causally defined, then they cannot be accounted for in nomic, non-causally defined terms. Particular cows and the ocular irradiations they cause will have nomic descriptions, but the types ‘cow’ and ‘ocular irradiation’ will not (the same, of course, holds for more internal determinants); that is what it means to say that causally-defined properties are anomic. One could perhaps replace the type ‘ocular irradiation’ with one that is not causally defined and only picks out proximal causes; but one can provide an analogous replacement for causally-defined distal types just as well. So for both tokens and types, distance is irrelevant to nomic status. These considerations show that in the framework within which Quine and Davidson operate, the question of the location of a causal-type generally – its distance from the effect – is completely independent of the question of its suitability for inclusion within strict laws. Location by itself, whether internal, proximal or distal, is neither a necessary nor sufficient condition for inclusion (or not) within strict psychological or psychophysical laws. The falsity of the nomic proportionality principle entails this whether or not one holds the further thesis concerning the anomic status of causally-defined properties. Davidson’s semantic externalism thus cannot be motivated by considerations deriving from the need to make sense of semantic normativity. And Quine’s invocation of proximal causes cannot be motivated by the demand for strict laws in a scientific psychology. The scale on which measurements such as ‘distal’ and ‘proximal’ are made is quantitative – it is concerned with the mere degree of distance between
82
STEVEN YALOWITZ
‘cause’ and effect – while, within this framework, what is required to guarantee strictness is something (i.e., non-causally defined properties) of an altogether different quality. It should be noted that this criticism of both Quine and Davidson is purely internal to their debate. Both attempt to motivate their position on semantic determinants by appeal to the need to establish or avoid strict psychophysical laws. But both also hold that causally defined predicates are unsuitable for such laws. And their citation of their respectively favored semantic determinants is done in causally-defined terms. Neither pays attention to the crucial question of the relation between non-causal definition – a necessary condition of strictlaw inclusion – on the one hand, and degree of distance between causes and effects, on the other hand. As I have argued in this section, attention to this question and, independently, to the relation between INUS/CP conditions and sufficient conditions, as well as the nature of causal chains, shows that the question of the location of semantic determinants cannot be decided simply by the stance taken on the scientific status of psychology.
5.
I have argued for two central conclusions in this paper. First, there is no special conflict between the normative notion of error and the idea of strict psychophysical laws. Any apparent conflict is to be traced to the more generic feature of interest-relativity which is constitutive of psychological and semantic explanation, and the special sciences generally. This is further evidence (Sections 3–4) that any truth there is to mental anomalism will be due to a more general truth about the special sciences, and is not owed to particular, idiosyncratic features of psychology such as normativity or rationality.28 The second central conclusion of this paper is that the semantic and causal-explanatory framework within which Quine and Davidson operate cannot by itself decide the question of the location of semantic determinants. In the debate I have been discussing, both Quine and Davidson hold that it is primarily the nature of strict laws – the demands of a serious science – that, together with favored conceptual demands that are thought to lead to or away from them, tell us where those determinants reside. The proportionality principle plays the central role in this shared picture. For Quine, the demands of scientific respectability pull semantic determinants in the direction of strict-law inclusion, and this is taken to motivate proximal location (communal embedding is thought to take care of the possibility of error). For Davidson, the demands of the notion of genuine error pull semantic determinants away from strict-law inclusion, thus
SEMANTIC DETERMINANTS AND PSYCHOLOGY AS A SCIENCE
83
motivating distal location. I have argued that neither demand adequately motivates either subsequent semantic move, because of the falsity of the proportionality principle. Before turning to the question of what might adequately motivate such a move, I want to emphasize the general significance of this conclusion that there is no intrinsic relationship between the questions of the location of semantic determinants and the status of psychology as a science. It might initially seem that this conclusion is confined to a specific and, indeed, idiosyncratic dispute between Quine and Davidson. But the conclusion easily generalizes. Recall the central considerations from which it derives: first, a broad semantic thesis to the effect that the meanings of utterances and contents of thoughts are determined by objects and events which cause their tokening; second, a requirement of the possibility of error; third, the assumption of the nomological character of causality; fourth, the insufficiency of either distal, proximal or internal types on their own to guarantee effect-types; and, finally, the falsity of the nomic proportionality principle. I argued that one’s attitude toward the nomic status of psychophysical generalizations is both unnecessary and insufficient for deciding where semantic determinants ought to be located because the location of an object or event-type in a causal chain is neither sufficient nor necessary for determining its nomic status. While the semantic and causal assumptions driving this claim are not indisputable, they have, individually and collectively, received widespread support, and to that extent the conclusions reached here on their basis have general significance that extends beyond Quine’s and Davidson’s debate. The nomic proportionality principle continues to play a central role in discussions of these topics, and recognition of its falsity should have important consequences for the internalist-externalist debate, by shifting the focus of that debate away from concerns with the scientific status of psychology.29 In concluding this paper, I want to explore briefly one plausible direction that this shift of focus could take, which is independently present in both Quine’s and Davidson’s work. Both respect the constraint of the public accessibility of semantic determinants; I want to sketch some considerations about this constraint that are relevant to the debate about the location of semantic determinants. In Quine’s own case, as already noted, the constraint is taken to rule out more inner, neural semantic determinants (Quine 1960, 31; see also 5–8); proximal causes are intended simultaneously to meet the demands of both scientific respectability (in the direction of contiguity and thus nomic strictness) and public accessibility (location on observationally-available bodily surfaces). I have observed above, and argued elsewhere in detail,
84
STEVEN YALOWITZ
that this is a very unsatisfactory compromise; neither demand is met by Quine’s proximal causes, since they are both anomic and not publicly accessible in the relevant sense. They cannot explain the possibility of communication (Yalowitz 1997b; see also Stenius 1969, 28). For Davidson, the constraint of public accessibility plays a more subtle role; though it is clearly prominent in his account of radical interpretation, he rarely if ever argues directly that it bears the central responsibility for semantic externalism (but see Davidson 1989a, 195). The most explicit connection drawn between the two occurs in the course of an argument that the determinacy of meaning and content requires both external determination and actual linguistic interaction, and thus public accessibility (Davidson 1992). I have argued elsewhere that this argument fails, and furthermore is inconsistent with other central features of Davidson’s view, in particular his rejection of the possibility of alternative conceptual schemes. Actual linguistic interaction is not a necessary condition for meaning, and we are therefore still without a justification for the weaker requirement of public accessibility.30 One thing that I think we should learn from this failure is that this requirement must be treated as an assumption (or, better, presumption) rather than as a conclusion demanding justification; indeed, this is something that should have already been learned from the history of failed attempts to read Wittgenstein’s private language argument as a transcendental argument for the requirement.31 My own view is that Quine’s emphasis on the requirement of public accessibility is on the right track for fixing the location of semantic determinants, and that it is only his misguided conception of the demands that a science must meet that prevents him from following through on the insight. Davidson’s distal semantic determinants provide the only way to actually satisfy that requirement; contrary to Quine, semantic externalism is required in order to make sense of the possibility of our being able to understand each other – only the correlation of utterances with public, and therefore external, objects and events can explain how communication is possible.32 But contrary to Davidson, this correlation does not itself depend on denying the status of psychology as a science, nor does it presuppose actual communicative interaction. Acknowledgement of the falsity of the proportionality principle, and the wedge that this drives between questions concerning the nomic status and semantic determination of commonsense psychology, should make it easier to recognize the centrality of the presumption of communicability for the foundations of externalism.33
SEMANTIC DETERMINANTS AND PSYCHOLOGY AS A SCIENCE
85
NOTES 1 For a survey of such responses, see Fodor 1990a. Fodor’s own asymmetric-dependence
theory also delivers only ceteris paribus semantic-determining laws (Fodor 1990b, 95); but this is due to his insistence on the ceteris paribus nature of special science laws generally, and not to some particular worry about error (see Fodor 1989). Fodor would thus reject the dilemma in the text below. As will become apparent, I am in agreement with Fodor about this, though our reasons differ; for Fodor, multiple realizability forces rejection of strict special-science laws, while on my view that rejection derives from the interest-relativity and causal definition of special-science predicates. See discussion in Section 4 below. 2 The dilemma also bears directly on Quine and Davidson’s debate over the communal nature of language – whether shared language is a necessary condition for error. I have discussed this part of their debate at length in Yalowitz 1998b. See also note 14, and Sections 2, 3 and 5 below. 3 Davidson has a somewhat awkward relation to the Humean tradition, in that he resists analyzing singular causal statements in terms of subsumption by laws (Davidson 1980b, 158). That difference is unimportant for my concerns in this paper. 4 Davidson of course also holds that in ascribing intentional states to an agent, a condition for seeing her as rational is that not only must interpreter and agent agree on much in their view of the world, but also that their view must, in large part, be correct (Davidson 1984c). Quine (1984) does not endorse this further requirement. For discussion, see Yalowitz 1998b. 5 Note that this definition is compatible with the possibility that the ultimate, strict physical laws are indeterministic. Indeterministic laws can be exceptionless, in the sense that they cite all causally relevant factors; their predictive limitations are not owed to a failure to cite any possible interfering conditions – if the effect does not occur when the antecedent conditions obtain, there is nothing else to be cited in explanation of this failure other than the fact of an indeterministic universe. It also ought to be noted that this definition is not satisfied by Jerry Fodor’s suggestions about the capacity of ceteris paribus generalizations to guarantee their effects (Fodor 1989, 73–76). Though such generalizations are in a sense exceptionless – when the antecedent is satisfied and the ceteris paribus conditions are met, the consequent must occur – they are not strict in the sense of having fully articulable antecedents. For related discussion, see Section 3 below. 6 Quine writes that ‘[a]s long as the linguist does no more than correlate the native’s observation sentences with his own by stimulus meanings, he cannot discount any of the native’s verdicts as false . . . ’. For related discussion, see Yalowitz 1998b. 7 Compare Kripke’s claim, in Kripke 1982, that for an individual considered in isolation, there can be no distinction between correct and incorrect usage. The similarity between Quine and Kripke that I am pointing to here is different than that noted by Kripke, 55–57. 8 For more on causal chains, see Sections 4 and 5 below. 9 Quine speaks in approving terms of the fact that ‘[s]timulation can remain the same though the rabbit can be supplanted by counterfeit’. See also 28–30. For further discussion of this point, see Sections 4 and 5 below. 10 It is worth drawing out the full Cartesian analogy here. The Cartesian move to ‘ideas’ is motivated in large part by the desire for epistemological certainty – to avoid error. And that is precisely what produces the difficulty facing Quine’s proxy for ideas, stimulus meanings. By gaining predictive accuracy (‘certainty’) they stand in tension with the possibility of error. Davidson, in criticizing Quine’s proximal theory (Davidson 1990, 73–76), focuses on
86
STEVEN YALOWITZ
the familiar epistemological point that Cartesianism entails skepticism about the external world, and thus the possibility of massive error. But Davidson curiously fails to notice the semantic implications of the proximal theory, on which skepticism is actually avoided because distal reference is unintelligible; the proximal theory leads to a form of idealism. I have discussed this issue at length in Yalowitz 1997b; see also Stenius 1969, 30–31, and note 32 below. 11 Though I do not propose entering into a detailed discussion here, it should be pointed out that availing oneself of distal objects in this way in order to account for the possibility of error does not fully solve the problem for which Quine’s communal strategy is offered as a solution. Just as if anything which stimulates an utterance goes into its stimulus meaning, then the possibility of error becomes problematic, just so one needs a method for determining which of the distal objects and events that cause an utterance-type actually semantically determine it in order to prevent the same problem from arising for externalism. So appealing to the distal environment does not completely solve the problem by itself, even if it does adequately ground the sheer possibility of error. I have offered a solution to this problem in Yalowitz 1998c. I might also note a suggestive connection here between the claim in the text that predictive slackness is a condition for error, and therefore rationality, and Davidson’s claim that it is a necessary (though of course not sufficient) condition for an action’s being free (see Davidson 1980c, 225). 12 This point comes out clearly in Davidson’s attack on social theories of semantic competence, and defense of an idiolectal approach, which emphasizes the lack of any border between semantic knowledge and knowing one’s way around the world generally. See Davidson 1986, 445–46. 13 Though I focus in the text on textual grounds for the attribution, there are also several important philosophical and methodological grounds. The first two concern Davidson’s avowed philosophical aims of establishing mental anomalism and externalism. It is clear that he wants to argue for mental anomalism on the basis of normative aspects of rationality (Davidson 1980c, 222). The argument for this is typically seen as focusing on norms of theoretical and practical reason (Kim 1985; Child 1993; see Section 3 below). But I have argued at length elsewhere that such arguments are not compelling (Yalowitz 1997a and 1998a). That suggests looking closely at other features of rationality, such as the idea of semantic normativity that I have been focusing upon here, which might be more promising. Second, I have also argued that Davidson’s recent attempts to establish externalism (Davidson 1989a and 1992) are inconsistent with important aspects of his overall view, in particular his rejection of conceptual relativism, and in any case are highly problematic (Yalowitz 1998b). This leaves externalism ungrounded, and again suggests that other avenues be explored that might provide grounding for it. A central motivation for the reconstruction developed in the text is that it explores a strategy for grounding both mental anomalism as well as externalism that both fits neatly into Davidson’s general picture and does not rely on these dubious sets of considerations. Third, and related to these first two reasons, error is a normative notion, and it has also played an important role in Davidson’s thinking about externalism (Davidson 1984b, 170; Davidson 1985, 479–80). So the idea of a unified argument for both mental anomalism and externalism which is grounded in a normative notion like error reflects important aspects of Davidson’s thinking about the normative basis of these two doctrines. This leads to a fourth, methodological reason for the line of argument I have developed. It provides a coherent structure for Davidson’s views, and sees them as derived from what is surely the least controversial aspect of his picture – the requirement of the possibility of error. Moreover, it also allows for a neat formulation of the structure of his debate with Quine over the location of semantic determinants and
SEMANTIC DETERMINANTS AND PSYCHOLOGY AS A SCIENCE
87
the status of psychology as a science. I might add that the fact that the argument is never explicitly stated by Davidson is not a conclusive consideration against attributing it to him. Davidson’s Quinean background may well exert subtle pressures on his thinking that he is not fully aware of (see Yalowitz 1998b). An argument may be implicit yet unacknowledged. This sort of interpretive strategy is commonplace and fruitful in historical studies. 14 Davidson does in fact require a community of interpreters in order for the notion of error to take hold; but there need be no shared language in Quine’s sense. He holds that determinate meaning and content depends upon social interaction, for otherwise there can be no rationale for locating semantic determinants at any one point on the causal chain of events and objects that extends from distal objects to a speaker’s brain (Davidson 1992). (It is worth noting that this claim stands in tension with his endorsement of the proportionality principle.) The connection between this claim about determinacy and the one focused on in the text about error is straightforward: error is only possible if a determinate content is in place, and this, it is claimed, presupposes social ‘triangulation’ and the distal causes it produces. In Yalowitz 1998b I have explained how Davidson’s argument for this requirement of community fails once one mixes into the Davidsonian brew of anomalism and externalism his other major departure from Quine – his rejection of the possibility of alternative conceptual schemes. In that paper I also elaborate on the point about objectivity noted in the text. See further Sections 4 and 5 below. 15 Davidson’s claim here is that externalism is sufficient for anomalism, which is the reverse of the sufficiency claim in my reconstruction. But the important point is the conceptual connection being made between the two issues. For related discussion, see Section 4 below. 16 It is an artificial way of accounting for the notion of error because it in effect stipulates certain INUS conditions (on which see below in the text) – a subset of the full stimulus meaning – as determinative of content, against the background of which certain uses can be deemed mistaken. In Yalowitz 1998b I have discussed unsatisfying aspects of this ‘skeptical solution’ to the problem of error; see also related discussion below. 17 Necessary, but not sufficient; social interaction (though not shared language in Quine’s sense) is required as well. See note 14 above. 18 INUS conditions only provide a rough model for the sorts of conditions selected, because on Davidson’s view these are supposed to be anomic, and thus do not fit the sort of part-whole relation that exists between INUS conditions and fully articulable sufficient conditions. See discussion below of ‘CP’ – ceteris paribus – conditions. 19 Indeed, the Quinean strategy for constructing the notion of an individual’s being in error relative to other speakers described in Section 2 seems intended to be a clear instance of this; however, see discussion in Section 5 which tells against the fulfillment of this intention. 20 This is precisely the case in certain ‘Twin Earth’ scenarios imagined by Tyler Burge (1979). 21 It should be noted that my alternative argument for mental anomalism, glossed in Section 4 below, embraces these consequences, and that they also show up in aspects of both Quine’s and Davidson’s thinking about causal explanation (see references cited in Section 4). 22 I am here bracketing the question of whether other, non-psychological properties are anomic. The situation is different, though no better for Quine’s or Davidson’s purposes, if all causally-defined properties are anomic. See note 23 and further discussion below.
88
STEVEN YALOWITZ
23 Unless, of course, one assumes that the types are causally defined, and that no nomic
relations can hold between such types. But I am bracketing this assumption here; see discussion below. 24 Such a view is put forth by Myles Brand in the course of his discussion of the problem of deviant causal chains: how behavior must be caused by intentions if that behavior is to count as intentional. On Brand’s view, intentions must be ‘proximal’ causes of actions, and thus action must begin just where intentional causation leaves off – in the brain. See Brand 1984, 20–21. 25 For explanation of this point, see Yalowitz 1998a, Section 4. For relevant discussion in Davidson, see Davidson 1987, 41ff; and Davidson 1991, 162ff. Davidson is not entirely consistent on the bearing of causal definition on nomic inclusion; for discussion and criticism, see Yalowitz 1998a, Section 5. 26 This is a corollary of Davidson’s well-known point about the relationship between singular causal statements and strict covering laws – that true instances of the former entail the latter, even if we do not know any such laws. See Davidson 1980a, 16–18, and Davidson 1980b, 160. 27 For a response to the possible rejoinder that Quine’s proximal causes can nonetheless be ‘picked out’ in non-causal terms, see below in the text. I have discussed the general question of whether causally-defined properties can be reduced to non-causally defined ones (and Quine’s own views on this question) in Yalowitz 1998a. 28 I have argued elsewhere (Yalowitz 1998a, Section 5, and 1997b) that the normative and rational features of psychology are responsible, not for psychology’s irreducibility, but rather its ineliminability. This is what captures the intuition that psychology stands apart from the other special sciences. 29 The principle is presupposed in the influential doctrine of ‘methodological solipsism’, articulated in Fodor 1981, which underlies many internalist positions. Note that the point in the text does not directly address the claim that requiring the supervenience of mental on physical properties forces proximal or internal semantic determinants (because TwinEarth cases are thought to show that external determinants violate supervenience). I do not propose going into this issue here, but will simply note that this claim depends upon assuming a version of ‘strong’ supervenience that demands strict cross-world relations between specific mental and physical properties. Other conceptions of supervenience – in particular, ‘global’ supervenience – do not have this result. More importantly, there is conceptual space for ceteris paribus forms of supervenience, which have unfortunately been unexplored in the literature. Both ceteris paribus supervenience and, on one reading, global supervenience are compatible with mental anomalism. (For discussion of varieties of supervenience, see Kim 1990). 30 See note 14 above. 31 See Stroud 1983. Wittgenstein’s rule-following considerations have often been read as arguing that a condition of grasping one’s own meanings is that they be grasped by others (e.g., Kripke 1982; McDowell 1984). In Yalowitz 1998c, I argue, against these sorts of considerations, that the reverse is true; third-person understanding depends upon self-understanding – on first-person authority. 32 For useful discussion of this point, see Bilgrami 1992, chapter 5. I might add here that another important element in motivating externalism is also present in Quine’s thinking: that we are committed to the posits of our theory of the world. In Yalowitz 1997b, I have explained how that commitment forces external objects and events to play the role of semantic determinants. See note 10 above.
SEMANTIC DETERMINANTS AND PSYCHOLOGY AS A SCIENCE
89
33 I am especially indebted to Sandy Mitchell for valuable advice on earlier drafts of this
paper, as well as David Brink. I would also like to thank Pat Kitcher, Philip Kitcher, Jessica Pfeifer, Oron Shagrir, Gila Sher and an anonymous referee for helpful comments.
REFERENCES
Bilgrami, A.: 1992, Belief and Meaning, Blackwell, Cambridge. Brand, M.: 1983, Acting and Intending: Towards a Naturalized Action Theory, MIT Press, Cambridge, MA. Burge, T.: 1979, ‘Individualism and the Mental’, in P. French, T. Uehling and H. Wettstein (eds.), Midwest Studies in Philosophy, Vol. IV, University of Minnesota Press, Minneapolis, pp. 73–121. Child, W.: 1993, ‘Anomalism, Uncodifiability, and Psychophysical Relations’, The Philosophical Review 102(2), 214–45. Davidson, D.: 1980, Essays on Actions and Events, Clarendon Press, Oxford. Davidson, D.: 1980a, ‘Actions, Reasons and Causes’, reprinted in Davidson 1980, pp. 3– 19. Davidson, D.: 1980b, ‘Causal Relations’, reprinted in Davidson 1980, pp. 149–62. Davidson, D.: 1980c, ‘Mental Events’, reprinted in Davidson 1980, pp. 207–25. Davidson, D.: 1984, Inquiries into Truth and Interpretation, Clarendon Press, Oxford. Davidson, D.: 1984a, ‘Radical Interpretation’, reprinted in Davidson 1984, pp. 125–39. Davidson, D.: 1984b, ‘Thought and Talk’, reprinted in Davidson 1984, pp. 155–70. Davidson, D.: 1984c, ‘On the Very Idea of a Conceptual Scheme’, reprinted in Davidson 1984, 183–98. Davidson, D.: 1985, ‘Rational Animals’, in LePore and McLaughlin 1985, pp. 473–80. Davidson, D.: 1986, ‘A Nice Derangement of Epitaphs’, in E. LePore (ed.), Truth and Interpretation: Perspectives on the Philosophy of Donald Davidson, Blackwell, New York, pp. 433–446. Davidson, D.: 1987, ‘Problems in the Explanation of Action’, in P. Pettit, R. Norman and J. Sylvan (eds.), Metaphysics and Morality, Blackwell, Oxford, pp. 34–49. Davidson, D.: 1989a, ‘The Conditions of Thought’, in J. Brandl and W. Gombocz (eds.), Grazer Philosophischen Studien: The Mind of Donald Davidson, Vol. 36, 193–200. Davidson, D.: 1989b, ‘Reply to Tyler Burge’, unpublished. Davidson, D.: 1990, ‘Meaning, Truth and Evidence’, in R. Barrett and R. Gibson (eds.), Perspectives on Quine, Blackwell, New York, pp. 68–79. Davidson, D.: 1990a, ‘The Structure and Content of Truth’, The Journal of Philosophy 87(6), 279–328. Davidson, D.: 1991, ‘Three Varieties of Knowledge’, in A. P. Griffiths (ed.), A. J. Ayer: Memorial Essays, Cambridge University Press, New York, pp. 153–56. Davidson, D.: 1992, ‘The Second Person’, in P. French, T. Uehling and H. Wettstein (eds.) Midwest Studies in Philosophy, Vol. XVII, University of Notre Dame Press, Notre Dame, pp. 255–67. Davidson, D.: 1995, ‘Could There be a Science of Rationality?’, International Journal of Philosophical Studies 3(1), 1–16. Hume, D.: 1888, A Treatise of Human Nature, L.A. Selby-Bigge (ed.), Clarendon Press, Oxford.
90
STEVEN YALOWITZ
Fodor, J.: 1981, ‘Methodological Solipsism Considered as a Research Program in Psychology’, reprinted in his Representations, MIT Press, Cambridge, pp. 225–53. Fodor, J.: 1990, A Theory of Content and Other Essays, MIT Press, Cambridge. Fodor, J.: 1990a, ‘A Theory of Content, I: The Problem’, in Fodor 1990, pp. 51–87. Fodor, J.: 1990b, ‘A Theory of Content, II: The Theory’, in Fodor 1990, pp. 89–136. Fodor, J.: 1989, ‘Making Mind Matter More’, Philosophical Topics 17(1), 59–79. Kim, J.: 1985, ‘Psychophysical Laws’, in LePore and McLaughlin 1985, pp. 369–86. Kim, J.: 1990, ‘Supervenience as a Philosophical Concept’, Metaphilosophy 21, 1–27. Kripke, S.: 1982, Wittgenstein on Rules and Private Language, Harvard University Press, Cambridge. LePore, E. and McLaughlin, B. (eds.): 1985, Actions and Events: Perspectives on the Philosophy of Donald Davidson, Blackwell, New York. Mackie, J.: 1974, The Cement of the Universe, Clarendon Press, Oxford. McDowell, J.: 1979, ‘Virtue and Reason’ The Monist 62(3), 331–50. McDowell, J.: 1984, ‘Wittgenstein on Following a Rule’, Synthese 58(3), 325–63. McDowell, J.: 1985, ‘Functionalism and Anomalous Monism’, in LePore and McLaughlin 1985, pp. 387–398. Quine, W. V.: 1960, Word and Object, MIT Press, Cambridge. Quine, W. V.: 1969, ‘Natural Kinds’, in his Ontological Relativity and Other Essays, Columbia University Press, New York, pp. 114–38. Quine, W. V.: 1974, The Roots of Reference, Open Court Press, La Salle. Quine, W. V.: 1975, ‘Mind and Verbal Dispositions’, in S. Guttenplan (ed.), Mind and Language, Clarendon Press, New York, pp. 83–95. Quine, W. V.: 1979, ‘Facts of the Matter’, reprinted in R. Shahan and C. Swoyer (eds.), Essays on the Philosophy of W. V. Quine, University of Oklahoma Press, Norman, pp. 155–69. Quine, W. V.: 1984, ‘Relativism and Absolutism’, The Monist 67(3), 293–95. Quine, W. V.: 1986, ‘Reply to Parsons’, in L. Hahn and P. Schilpp (eds.), The Philosophy of W. V. Quine, Opencourt Press, La Salle, pp. 396–403. Stenius, E.: 1969, ‘Beginning with Ordinary Things’, in D. Davidson and J. Hintikka (eds.), Words and Objections, Humanities Press, New York, pp. 27–52. Stich, S.: 1983, From Folk Psychology to Cognitive Science: The Case Against Belief, MIT Press, Cambridge, MA. Stroud, B.: 1983, ‘Wittgenstein’s “Treatment” of the Quest for “A Language which Describes my Inner Experiences and which only I Myself can Understand” ’, in P. Weingartner and J. Czermak (eds.), Epistemology and Philosophy of Science: Proceedings of the 7th International Wittgenstein Symposium, Holder-Pichler-Tempsky, Vienna, pp. 438–45. Yalowitz, S.: 1997a, ‘Rationality and the Argument for Anomalous Monism’, Philosophical Studies 87(3), 235–58. Yalowitz, S.: 1997b, ‘Normativity, Externalism and the Status of Content’, unpublished manuscript. Yalowitz, S.: 1998a, ‘Causation in the Argument for Anomalous Monism’, Canadian Journal of Philosophy 28(2), 183–226. Yalowitz, S.: 1998b, ‘Davidson’s Social Externalism’, Philosophia 27(1–4).
SEMANTIC DETERMINANTS AND PSYCHOLOGY AS A SCIENCE
91
Yalowitz, S.: 1998c, ‘Individualism, Normativity and the Epistemology of Understanding’, Philosophical Studies (forthcoming). Manuscript submitted July 21, 1997 Final version received December 30, 1997 Manuscript accepted March 24, 1998 Department of Philosophy The University of California at San Diego 9500 Gilman Drive La Jolla, CA 92093-0119 U.S.A.