Is Narrow Content Superfluous?

Jesse J. Prinz

jjprinz@midway.uchicago.edu

University of Chicago

1994

Introduction

Not too long ago, Jerry Fodor was widely regarded as the most ardent crusader for narrow content[1] All this changed with the publication of The Elm and the Expert: Mentalese and Its Semantics (Fodor, 1994). In the second chapter of that book, Fodor announces that we can do without narrow content. Fodor's retraction has been perceived as an act of treachery by those of us who have befriended of narrow content. It is sometimes said that a cause reaches its darkest hour when its greatest champions defect to the side of its opponents. If that is the case, narrow content seems to be approaching its darkest hour. Fortunately, friends of narrow content need not be disheartened. I will argue that Fodor's most recent arguments against narrow content are unsuccessful. There are many arguments in The Elm and the Expert that do not directly address the individuation of mental states. These will undoubtedly be widely discussed by those interested in mind and meaning. I will limit my discussion to the arguments in the second chapter.

To understand the significance of Fodor's arguments, we must situate them within a more general campaign that has been waged against individualism over the past twenty years. The term 'individualism' has not been given a uniform definition in the literature. Fodor (1987) defines it as the view that mental states should be individuated with respect to their causal powers. This definition might be overly restrictive (cf., e.g., Loar, 1987). Individualism is more frequently defined in terms of an Autonomy principle:

Autonomy. There is a way of individuating the mental states of an individual that does not depend in any deep way on the physical or social environment in which that individual is situated.

This principle does not necessarily entail that we must pick out mental states without mentioning the environments in which they are situated. Rather, it entails that no particular environment is essential to the identity of a mental state. That is what is meant when we call individuation that respects Autonomy narrow. But narrow individuation does not purchase narrow content. For that, another principle is needed:

Semanticity. Narrowly individuated states possess semantic properties and can be differentiated on the basis of those properties.

This principle insures that narrowly individuated states are not merely syntactic states; they have semantic properties. This should not be confused with the claim that narrowly individuated states are semantically evaluable (i.e., that they have truth conditions). Instead, the idea is that they make a certain kind of contribution to the determination of truth conditions and that they can be distinguished from one another on the basis of those contributions. The semantic properties of narrowly individuated mental states can be called their 'narrow content.' Still, Semanticity does not fully motivate individualism. The proponent of narrow content must also demonstrate that narrow individuation is not only possible, but also useful for studying the mind. This gives us a final principle:
Explanatory Value. States individuated without reference to the environment are relevant to, needed for, or privileged by scientific psychology.

I will define individualism as the sum of these three principles. Each principle has been the subject of debate. For example, Burge (1979) and Putnam (1975) challenge Autonomy, Stich (1983) and Millikan (1993) challenge Semanticity, and Burge (1986) challenges Explanatory Value. Fodor's (1994) arguments fall into the last category. Those arguments do not challenge Autonomy or Semanticity. Fodor does not want to deny the existence or coherence of narrow content. Instead, he wants to demonstrate that broad psychological laws (i.e., laws that individuate mental states extensionally) are sufficient for psychological explanation. This poses a threat to the Explanatory Value clause. If psychological laws are broad, then it is not clear how narrow content could have any use for science.

To defend his thesis, Fodor (1994, chapter 2) discusses four kinds of cases that have been used to motivate narrow content. He argues that broad psychological laws can handle some of these cases, and that the others fall outside of the explanatory domain of scientific psychology. Because Fodor's arguments have yet to be widely discussed, I will outline them in some detail and respond to each in turn. In each case, I will argue that Fodor fails to support his claim that psychological laws do not need narrow content.

1. Twins

Fodor's first argument addresses Twin Earth cases. I will not rehearse the all-too-familiar Twin Earth thought experiments here. Those familiar with the twin cases will know that they are often cited as evidence against narrow content. But proponents of individualism have argued that twins can be more successfully enlisted in defense of narrow content. Here's one version of the argument:

(1)

Fodor goes after the second premise. He argues that, even if 1.P1 is true, psychological laws need not generalize over twins[3] The argument runs as follows:

(2)

This argument highlights a particular example, but the conclusion can be generalized. The first idea is that Twin-Earth cases are nomologically impossible, because Twin-Earth has different natural laws. If this premise is accepted, 2.P4 does all the work. If psychological laws only need to subsume nomologically possible cases, then Twin-Earth cases are irrelevant to psychology. Consequently, broad psychological laws are not impugned by their failure to generalize over twins.

Twins can be annoying. Lots of people would like to rule out Twin-Earth cases tout court. Unfortunately, it does not seem to be as easy as Fodor thinks. Both of the crucial premises ÷ that twins are nomologically impossible and that psychological laws are restricted to nomologically possible cases ÷ are vulnerable. That some twin cases do not violate natural laws is evidenced by the fact that we can construct twin cases here on Earth. Fodor concedes this point, and such cases are the subject of his second argument. However, he firmly maintains the truth of 2.P4. By my lights, this premise is too strict. There are nomologically impossible cases that should fall under our psychological laws. The make this point more plausible, it is important to clarify what 2.P4 really says. Importantly, it does not say that there could not be laws which subsumed both Jones and Twin-Jones. Elsewhere, Fodor (1974) argues that it is characteristic of the special sciences that their taxonomies do not reduce to the taxonomies of the sciences on which they supervene. Consequently, laws of special sciences need not be reducible to laws of the sciences on which they supervene. It seems to be consistent with this view that the laws of one science could be preserved even if the laws governing its underlying substrata were different. For example, the laws of economics could be preserved if the laws of chemistry were different. That is, economic laws could subsume worlds that are governed by different chemical laws than we have on Earth. Or, in Fodor's idiom, economic laws can subsume nomologically impossible worlds. By analogy, if psychological states are not type-identified with physical states, it is reasonable to conclude that psychological laws could be designed to generalize over both nomologically possible and nomologically impossible cases. For example, we could design laws that would subsume both Jones and Twin-Jones.

This last observation is perfectly consistent with 2.P4. That premise does not allege that there are no laws subsuming nomologically impossible cases. It alleges that the laws we make use of here on Earth, the laws sought by a scientific psychology, need not subsume such cases. This claim requires further defense. When we turn our attention to the laws of physics and chemistry, the modality is much stronger: physical and chemical laws cannot subsume nomologically impossible cases. Laws of special sciences can subsume such cases, but Fodor thinks they are not required to. Although he doesn't defend this claim explicitly, the following reasoning might be what he has in mind. First he might observe that special sciences are designed for our use. Consequently, they do not need to be capable of predicting the activities of impossible individuals. To move from this relatively uncontroversial claim to 2.P4, another assumption is required. In particular, Fodor must show that the only reason to subsume impossible individuals under our laws is that doing so would allow us to predict their activities. If predicting Twin-Jones' activities are the only motivation behind subsuming her under our laws, then we need not subsume her under our laws. The problem with this defense of 2.P4 is that we might have other reasons for constructing laws that subsume twins. It is not enough to show that we can get by with laws that don't subsume twins. To defeat the Explanatory Value clause, Fodor must show that laws that subsume twins make no contribution to scientific psychology. Even if laws that subsume twins have no more predictive power that laws that fail to subsume twins, they might have more explanatory power. If so, 2.P4 is unjustified.

Fodor's argument might convince us that, if there is a psychological generalization that subsumes both Jones and Twin-Jones, it will have no predictive advantage over one that just subsumes Jones. Because Twin-Jones cannot exist, we never need to predict her actions. Nevertheless, there is reason to believe that a psychological generalization subsuming both Jones and Twin-Jones would have explanatory power. Laws that do not subsume twins are assumed to be broad. Broadly individuated mental states advert to aspects of the environment. It is a property of each broadly individuated state that it is related to some aspect of the environment. For example, it is a property of broadly individuated water-beliefs that they are related to water. From the fact that we can construct laws subsuming twins, we come to realize that these relational properties are not causally efficacious. Jones and Twin-Jones have mental states that are narrowly equivalent and broadly distinct. We can construct generalizations which attribute the same causal powers to these mental states in spite of the differences in their broad content. This indicates that the broad relational properties of those states are not contributing to their causal efficacy. Put differently, looking at the behavior of twins allows us to determine which properties of our mental states are causally efficacious. This undoubtedly contributes to our understanding of how the mind works. Therefore, the study of twins has explanatory value. Broad psychological laws wantonly fail to distinguish the causally efficacious properties of mental states from the causally inert relational properties. When we accommodate twins in constructing psychological laws, we remedy this problem. Thus, even if broad laws can successfully predict which effects our mental states will cause, they fail to explain how those effects are caused; they fail to tell us which properties of our mental states are doing the work. If we construct psychological laws that accommodate twins, we can fill that explanatory lacuna.

2. Neighbors

Everyone is well aware that twins can occur on Earth. I will call Earth-bound twins neighbors. Imagine the following possibility. Our planet contains both water and some extremely similar substance, neighbor-water. There is one isolated community that has only water and another isolated community that has only neighbor-water. Now imagine that Jones lives in the first community and her molecular duplicate, Neighbor-Jones, lives in the second. Because of their molecular equivalence, we can assume that Jones and Neighbor-Jones share the same narrowly specified concepts including a concept we can call C. C-thoughts (i.e., thoughts containing tokens of the concept C) are caused by stuff with the same macro-properties as water (e.g., wet, thirst-quenching, transparent, etc.). Accordingly, neither Jones nor Neighbor-Jones would be able to distinguish water from neighbor-water; both substances would trigger their C-thoughts. Nevertheless, because the two Joneses never leave their respective communities, it just so happens that Jones' C-thoughts are only caused by water samples, and Neighbor-Jones' C tokens are only caused by neighbor-water samples. Many individualists think that this kind of case tells against broad psychology. To make this point, they often argue as follows:

(3)

Fodor's main qualm is with 3.P1. He thinks that broad contents are determined by The One True Informational Semantics (hereafter, OTIS). As it stands, the setup given above does not contain enough information to determine whether OTIS would render 3.P1 true. In particular, the truth of that premise depends on whether it is nomologically possible for Jones and Neighbor-Jones to distinguish water from neighbor-water. If it is not, Fodor will argue that 3.P1 is false; if it is, Fodor will argue that 3.P4 is false. Either way, argument (3) is unsound.

Let's begin by considering the first case in which it is not possible to distinguish water from neighbor-water. According to Fodor, OTIS entails the following principle:

The Principle of Disjunctive Content (PDC): If it is nomologically necessary that some creature cannot distinguish Xs from Ys, then that creature cannot have a concept that applies to Xs but not to Ys. (Fodor, 1994, p.32)

If we apply PDC to case of Jones and Neighbor-Jones, we get the following argument:

(4)

If the conclusion of this argument is true, the first premise of argument (3) must be rejected. That premise asserted that the C-tokens of Jones and Neighbor-Jones differed in broad content. To see what's wrong with Fodor's argument it is important to understand why we adopted that premise in the first place.

It turns out that the intuition behind 4.P1 rest on a tacit assumption about content determination. This assumption can summarized as follows:

The Actual Causes Assumption (ACA): C refers to Xs only if C-tokens are actually caused by Xs.

In its contrapositive form ACA entails that C could not refer to Ys if C-tokens were never actually caused by Ys, even if C-tokens would be caused by Ys in counterfactual situations. This brings us back to the case described above. Ex hypothesi it just so happens that Jones' C-tokens are caused only by water, and Neighbor-Jones' C-tokens are caused only by neighbor-water. Therefore, if ACA is correct, Jones' C-tokens and Neighbor-Jones' C-tokens cannot have the same broad content. In disputing this claim, Fodor must reject ACA. Indeed, PDC entails that ACA is false. If PDC is correct, then a class of things can belong to the broad content of one of my concepts even if members of that class have never caused me to token that concept[4] ACA says that this is impossible. The question is, Which of these two principles of content determination is correct?

Interestingly, Fodor discusses ACA in an earlier work[5] He argues that ACA is necessary if we want to avoid a certain kind of verificationism. If OTIS does not include ACA, it will be committed to the view that, if we can't distinguish Xs from Ys, we can't refer to Xs without referring to Ys. Thus, reference is significantly constrained by epistemology. To avoid this embarrassing commitment, Fodor recommends ACA. If ACA is correct, reference is constrained by the causal history of a concept (a fact about the world) and not our ability to distinguish things and properties (an epistemological fact). Given the sordid history of verificationism in semantics, the first kind of constraint seems preferable. Fodor (1990, p.122) notes that ACA does not avert verificationism completely; a concept can only refer non-disjunctively to the class of one of its actual causes if the members of that class can be distinguished from other actual causes. He also notes that ACA complicates the semantics of non-instantiated natural kinds: how can we have UNICORN concepts if we never actually encounter unicorns? In light of these last considerations, Fodor remains undecided about whether we should include ACA in OTIS. Apparently, Fodor has now overcome his ambivalence. As we have seen, his recent endorsement of PDC constitutes a repudiation of ACA.

Should we follow Fodor in rejecting ACA? I submit we should not. The reason is that, without ACA, OTIS seems to invite a veridical and pernicious sorites argument. Because my limited perceptual abilities are limited, for any given one of my concepts, there will be nomologically possible objects that I cannot discriminate from the objects that actually cause tokenings of that concept. For example, my concept HORSE is actually caused by horses, but there are nomologically possible things that I could not distinguish from horses. For example, the species that immediately precedes horses on the evolutionary chain would look just like a horse to me. This species is not only possible, it actually existed right here on Earth. Thus, OTIS sans ACA is committed to the view that my HORSE concept refers to both horses and these proto-horses. This commitment is bad on its own, but it is not the end of our troubles. I would be equally unable to distinguish proto-horses from their ancestors, proto-proto-horses. If OTIS sans ACA is correct, I cannot have a concept that refers proto-horses and not proto-proto-horses. Therefore, my HORSE concept refers to horses, proto-horses, and proto-proto-horses. This same reasoning can take all the way back through evolutionary history. It leaves us with the preposterous conclusion that my HORSE concept refers to every entity in the horse lineage including the single-celled creatures that gave rise to all living things on Earth.

Here, Fodor might be tempted by the following reply. PDC says if it is necessary that some creature cannot distinguish Xs from Ys, then that creature cannot have a concept that applies to Xs but not to Ys. This allowed us to conclude that my HORSE concept refers to horses and proto-horses. However, we might not be entitled to the next step down our slippery slope. If proto-proto-horses are indistinguishable from proto-horses, but they are not indistinguishable from horses. Fodor could argue that, PDC is only committed to the view that my HORSE concept refers to proto-proto-horses if they are indistinguishable from both proto-horses and horses. Unfortunately, his argument would be uncompelling. My HORSE concept is disjunctive: it refers to horses or proto-horses. If it were conjunctive, it would be vacuous because nothing is both a horse and a proto-horse. Stated using a disjunctive concept, PDC amounts to the view that if it is necessary that some creature cannot distinguish (Xs or Zs) from Ys, then that creature cannot have a concept that applies to (Xs or Zs) but not to Ys. If we attend to the logic, it is plain that this principle is applicable even if Ys are not indistinguishable from both Xs and Zs. Even if Ys are only indistinguishable from Zs, there will still be no concept that applies to (Xs or Zs) without applying to Ys. I take this to be a reductio on OTIS sans ACA, what Fodor calls pure informational semantics. If I am right, Fodor's current semantic program collapse under a disjunctive explosion.

In response, Fodor might argue that this sorites argument underestimates our discriminative capacities. It might be that we are capable of discriminating horses from their immediate ancestors. Of course, to block the sorites, a stronger claim is required. Fodor must show that there is no nomologically possible species that is indistinguishable from horses. It is hard to imagine that this claim is defensible given the fact that our capacities are finite. For the sake of argument, let's grant that our discriminative capacities are virtually unbounded. If we let Fodor avail himself of this escape route, he will have to withdraw the first premise of argument (4). If humans have extraordinary discriminative capacities, Jones and Neighbor-Jones should be able to distinguish water from neighbor-water. The reader will recall that Fodor thinks that the strategy we adopt in response to argument (3) will depend on this issue of distinguishability. We have seen that the strategy Fodor adopts under the assumption that water and neighbor-water are indistinguishable exposes deep problems in his semantic program. We now turn to the strategy he adopts under the opposite assumption.

The first thing Fodor notes is that we can have disjunctive concepts that subsume distinguishable things. Common examples are heterogeneous kind concepts, functional kind concepts, and subject-relative functional kinds (1994, p.30f). To generate a real neighbor case, we must assume that the concepts in question are natural kind concepts. Such cases are possible, argues Fodor, but psychological laws are not required to generalize over neighbors. Here's why:

(5)

Depending on the way the case is set up, OTIS might entail that Jones' C-tokens refer to water and Neighbor-Jones' C tokens refer to neighbor-water. If so, their C-thoughts will not be subsumed by the same broad psychological laws. The moral of argument (5) is that this missed generalization doesn't matter. If water and neighbor-water are distinguishable, any failure to distinguish them would be accidental. From this it follows that any similarities in the way Jones and Neighbor-Jones behave with respect to C would be accidental. Psychology is not required to capture accidents. Therefore, broad psychological laws suffice.

My primary worry concerns 5.P4. This premise brings us back to an issue we touched on earlier in discussing twins. We saw that Fodor is not arguing that we are unable to construct laws that subsume twins; he is merely arguing that scientific psychology can get by without such laws. Here, he seems to be making the same claim about neighbors. It would be absurd to deny that there could be laws subsuming both Jones and Neighbor-Jones. Ex hypothesi, they behave in the same ways and they are molecular duplicates. But, once again, Fodor claims that psychology does not need to accommodate such laws. And, once again, we are left with a puzzle about the basis of the normative terms in this claim. What does is mean to say that a science is required or not required to do something? More specifically, what does it mean to say the science of psychology does not need to explain something that it can explain (e.g., the behavioral similarities of neighbors)?

To answer such questions, it is important to understand the relationship between scientific ability and scientific responsibility. Scientific ability is the ability of a science to explain and predict a range of phenomena. Scientific responsibility is a precept determining what phenomena a given science should explain or what patterns it should account for. Under normal circumstances, scientific responsibility is determined by scientific ability: a science should generally explain whatever phenomena it can explain. There is no obvious reason why we should restrict the explanatory breadth of a science beyond the pragmatic constraints imposed by the current state of that science. Applied to the present example, this principle provides an argument for narrow content. If we construct psychological laws that advert to narrowly individuated states, we will be able to explain the behavioral similarities of Jones and Neighbor-Jones. To reject such generalizations because their explanatory success is 'accidental' seems quite rash. As we have seen, accidents do not preclude lawlike similarities in behavior. As long as such similarities occur, laws are well motivated. This conclusion is particularly apt in the present case. Unlike nomologically impossible twins, neighbors can occur here on Earth. Thus, unlike laws subsuming nomologically impossible twins, laws subsuming neighbors can actually be useful in predicting behavior. This suggests that we should seek and embrace such laws even if their success is accidental. To appreciate the truth of this claim, we must avoid being fooled by the negative connotations of the word 'accidental'. One might adopt the view that science should ban successful generalizations when their success is completely fortuitous. However, the success of generalizations subsuming Jones and Neighbor-Jones is not fortuitous. It results from the fact that Jones and her neighbor are molecular duplicates. Banning successful generalizations when their success is not fortuitous amounts to bad scientific practice.

3. Experts

Deferential concepts comprise a third class of cases that have been used in defense of narrow content. It is argued that if a person has two concepts that are narrowly identical, they should fall under the same laws even if they correspond to concepts that experts are able to distinguish. To use Putnam's example, the fact that some experts can distinguish elms from beeches should not have any bearing on the causal powers of my ELM and BEECH concepts. If the causal powers of my two concepts are the same, they should fall under the same laws. Against this reasoning, Fodor offers the following argument:

(6)

Before we can evaluate this argument, we must resolve an ambiguity in 6.C1. The difficulty rests in the phrase 'distinct concepts.' The term 'concept' is ambiguous between broadly and narrowly individuated concepts. Given this ambiguity, it is unclear from my reconstruction whether Fodor is arguing that my ELM concept and my BEECH concept are broadly distinct or narrowly distinct. I will contend that on the former reading, the argument is question begging, and, on the latter reading, his argument can lend support to individualism. Let's begin by assuming the former. If Fodor is arguing that deferential concepts are broadly distinct, he is in agreement with just about every author in the literature. His discussion of experts as instruments becomes nothing more than a colorful way of describing Putnam's linguistic division of labor. On this reading, 6.C1 is completely uncontroversial. But, this reading of 6.C1 does not entail 6.C2. To claim that my deferential concepts are subsumed by different laws solely on the basis of the fact that they differ in broad content flagrantly begs the question against the individualist.

This strongly suggests that Fodor is referring to narrow contents when he says my ELM and BEECH concepts are distinct. That is an interesting and controversial claim. The idea seems to be that the internal mechanisms which cause me to defer to experts result in the fact that my ELM concept covaries with elms and my BEECH concept covaries with beeches under ideal circumstances. My ELM and BEECH concepts are not tokened in all the same counterfactual situations, because, in some of those situations, I am using an expert as a tool. If this argument is correct, them my ELM and BEECH concepts cannot be subsumed under the same narrow laws. Fodor thinks this conclusion is destructive to individualism. If he can prove that deferential concepts are narrowly distinct, he will deprive the individualist of one argument for narrow content. Consequently, the individualist might be inclined to rebut Fodor's argument for the narrow distinctness of deferential concepts.

Two reasons suggest that this rebuttal is not worth the effort. First, this argument has a different structure than Fodor's arguments against twins and neighbors. In those arguments, Fodor focuses on cases where individualists and anti-individualists would individuate differently, and he argues that laws based on anti-individualistic individuation will suffice. If he is right, then individualists individuate content at the wrong grain. Therefore, the individualists must try to refute these arguments. In the present argument, Fodor tries to demonstrate that individualist and anti-individualist individuation are equally fine grained; both distinguish my ELM and BEECH concepts. If he is right, experts yield a stalemate between individualism and anti-individualism. If both camps agree that my deferential concepts are distinct, then we have no way of deciding whether to subsume deferential concepts under narrow laws or broad laws. The jury is still out. Therefore, argument (6) does not threaten individualism as directly as the preceding arguments. The second reason why the individualists should not bother rebutting argument (6) is more substantial. In the literature, deferential concepts are generally used to argue against individualism, not in favor. Pre-philosophical intuitions are sceptical of theories that fail to distinguish my ELM concept from my BEECH concept. Consequently, individualists typically appeal to other kinds of cases to defend their program. Fodor's argument allows individualists to accommodate pre-philosophical intuitions. If he is right, the individualist can maintain that deferential concepts are narrowly distinct. From this perspective, Fodor's argument is grist to the individualist's mill.

4. Frege Cases

Frege cases are frequently marshaled in defense of narrow individuation. To use Fodor's favorite example, consider Oedipus. Oedipus does not realize that Jocasta is his mother. He believes that Jocasta is his wife and that his mother is not his wife. Extensionally, these beliefs are inconsistent, but there also seems to be a sense in which they are consistent. This consistency resides in the fact that Oedipus' beliefs are mediated by distinct internal states; in a word, his beliefs have different narrow contents. The appeal to distinct narrow contents also explains Oedipus' behavior. If Oedipus wants to see his mother, he will not go into Jocasta's room; if he wants to see his wife, he will. If we had to individuate Oedipus' desires extensionally, we would not be able to explain this behavioral difference; the desire to see his mother and the desire to see his wife are extensionally type identical for Oedipus. This indicates that extensionally individuated mental states are inadequate.

Apparently, Fodor is no longer convinced by the Oedipus example. He thinks that it is inconsistent with a basic tenet of belief/desire psychology. He calls this tenet the Principle of Informational Equilibrium, which can be put as follows:

(PIE) If an agent knew all the relevant information, she would not normally act other than she does. (Fodor, 1994, p.42)

This principle is quite powerful. It prevents the proliferation of Frege cases. Recall that Frege cases generally involve individuals who are not in 'epistemic equilibrium' with respect to the facts on which they act. If Oedipus had known all the relevant information, he would have known that Jocasta was his mother. Had he known this, he would have acted very differently. If PIE is correct, then Oedipus' situation is very unusual. But PIE cannot be the end of Fodor's story. It is one thing to claim that Frege cases are rare, and quite another to claim that they need not be addressed by a good psychological theory. To defend the latter claim, Fodor must cash out the word 'normally' in his statement of PIE. If normality were merely a statistical measure of relative frequency, abnormal cases could not be banished from the domain of scientific psychology. Scientific theories should explain exceptional cases if those cases are not unprincipled. Fodor realizes this. In defending PIE, he indicates that normality is deeply tied to rationality. Agents who are not in epistemic equilibrium (and thereby fail to satisfy PIE) are generally irrational. The laws of belief/desire psychology are designed to subsume rational agents. Therefore, agents who fail to satisfy PIE are not within the explanatory domain of belief/desire psychology; they are unprincipled exceptions.

I will turn to Fodor's defense of PIE presently, but first I will summarize its role in the argument against Frege cases:

(7)

That 7.P2 follows from 7.P1 can be explained in the following way. When agents fail to realize that their co-extensional concepts are co-extensional, they tend to make certain kinds of errors. For example, in not realizing that his JOCASTA and MOMMY concepts were co-extensional, Oedipus inadvertently committed incest. Hence, agents who fail to identify their co-extensional concepts do not satisfy PIE. If they were savvy to all they relevant facts , they would not act as they do. As indicated, Fodor seems to regard agents who do not satisfy PIE as irrational. Consequently, agents who fail to identify their co-extensional concepts are irrational. In short, if PIE is true, then 7.P2 is true.

As can tell, argument (7) rests on the truth of PIE. Fodor thinks this principle can be derived from three psychological truisms:

T1 You cannot choose A over B unless you believe you would prefer A to B if all the facts were known to you,

T2 The success of an action is accidental unless the beliefs that the agent acts on are true, and

T3 No belief/desire psychology can view the normal success of rational actions as accidental[6]

Assuming that belief/desire psychology is correct (a big assumption!), PIE is a direct consequence of T1-T3. If T1 is true then, in choosing to do A, I believe I would choose A if I know all the relevant information. If T3 is true and my performance of A is successful, then its success is not accidental. If T2 is true, the success of A could only have been non-accidental if the beliefs motivating A were true. Among those beliefs was the belief that I would have done A if I had maximal knowledge of the relevant information. Therefore, if A is successful, it is true that I would have done A if I had maximal knowledge. Thus, PIE is confirmed. Unfortunately, not all of these 'truisms' are true.

T1 strikes me as too strong, but I will assume that it is correct for the sake of argument. The real problem is with T2. 'Accidental' can be a slippery term, but here it is being clearly constrained by T3. As I understand belief/desire psychology, the success of rational actions can be accidental in one sense. This sense can be brought out by constructing a Gettier-like case. Imagine that I am walking through the forest, and I suddenly notice that I am surrounded by what I take to be wolves. Because I know that wolves inhabit this forest and because I can clearly see a bunch of things that look just like wolves, my belief that I am surrounded by wolves is justified. I also have a justified belief that wolves are dangerous. These two beliefs causes me to run out of the forest. This seems to be a paradigmatically rational act. But now suppose the following: first, all the objects I took to be wolves were actually Hollywood props set up for a movie scene that was being filmed in the forest unbeknownst to me; second, some real wolves were hidden out of sight in the bushes just behind those props. Given these two facts, we can conclude that my rational act was successful (I avoided a real danger), but its success was accidental (the objects that I took to be dangerous were not the real source of the danger). T3 cannot rule out this kind of accidental success, because such cases are consistent with ordinary intuitions about rationality.

This leaves us with a question: What kind of accidental success is inconsistent with ordinary intuitions about rationality? As we have indicated, rationality seems to involve justification. We act rationally when the reasons for our actions are justified. However, justified beliefs are not essential for successful action. For example, many people regard playing the lottery as an irrational act. The chances of winning are so small that investing in a lottery ticket is an unjustified expense. Nevertheless, people do win lotteries. This is an example of an accidental success that is not the result of rationality. Belief/desire psychology cannot account for the relationship between believing you will win the lottery and actually winning, because chances of winning are not increased by that belief. In contrast, belief/desire psychology can account for the relationship between having a justified belief that you are in danger and your success in avoiding danger, because the likelihood that a justified belief will be true is significantly better than chance. The upshot of all of this is that the use of the word 'accidental' in T2 differs from the use of that word in T3. On my reading, belief/desire psychology requires that the beliefs underlying our rational actions be justified, not that they be true. My belief that the objects before me are wolves is false, but justified. The lucky lottery player's belief that she will win if she buys a ticket is true, but unjustified. Both of us are successful in performing the actions based on those beliefs, but only one of us acts rationally. Thus, from the vantage point of belief/desire psychology, rationality does not preclude successful actions that are based on false beliefs; it only precludes successful actions that are based on unjustified beliefs. If we are to define the term 'accidental' with reference to the way belief/desire psychology conceives rationality (i.e., with reference to T3), then T2 is too strong. Instead we should say:

T2' The success of an action is accidental unless the beliefs that an agent acts on are justified.

It should be clear that T1, T2', and T3 do not entail PIE. Behavior based on justified beliefs is a weaker requirement than behavior based on true beliefs. Our beliefs can be justified even when we don't have all the relevant information. Justification is generally defined relative to some corpus of information. We say: Given what I know and what I can reasonably be expected to find out, I am justified in believing P. If we define accidental behavior in terms of justification, we might be able to derive a principle that is similar to PIE:

(HUMBLE-PIE) A rational agent is justified in believing that she would not act differently if she knew all the relevant information.

This principle is much weaker than Fodor's principle. The fact that an agent is justified in believing she is in epistemic equilibrium does not entail that she is in epistemic equilibrium. Her belief might be justified simply in virtue of being implied by the information at her disposal. Given the information at Oedipus' disposal, it seems that he behaved rationally in marrying Jocasta. He had no reason to suspect that he would not have married her if he had all the relevant information; in particular, he had no reason to suspect that Jocasta was his mother. Indeed, the fact that Oedipus acted rationally is what makes the story interesting[7]

Now we can take stock. The foregoing considerations indicate that PIE cannot be derived from ordinary intuitions about rationality. Without PIE, 7.P2 has no foundation: there is no reason to think that rational agents generally identify co-extensional beliefs. In other words, the agents involved in Frege cases cannot be written off as irrational. All parties agree that rational agents should be subsumed by psychological laws. Therefore, psychological laws should subsume Frege cases. Q.E.D.

In response, Fodor might opt for a strategy that we dismissed earlier. HUMBLE-PIE says that rational agents are justified in believing they are in epistemic equilibrium. The likelihood that a justified belief is true is better than chance. Thus, HUMBLE-PIE entails that the likelihood that a rational agent is in epistemic equilibrium is better than chance. If this is true, the likelihood that a rational agent will identify co-extensional concepts is better than chance; i.e., Frege cases will be relatively contained. Given this, Fodor might conclude that, even if the agents involved in Frege cases are rational, those cases are so rare that psychology does not need to accommodate them. Two responses. First, we cannot infer that Frege cases are rare from the fact that the likelihood that a rational agent will identify co-extensional concepts is better than chance. 'Better than chance' could mean that we identify co-extensional concepts 70% of the time. Thus, Frege cases could still occur 30% of the time. If so, psychological laws had better subsume them. Otherwise, psychology will turn out to be an explanatorily barren science.

My second response takes the opposite course. Let's assume that rational agents identify their co-extensional beliefs 95% of the time. In that case, Frege cases will be quite rare. However, as I have argued, their rarity does not testify to their irrationality. Under certain circumstances, rational agents will fail to recognize crucial identities. Fodor would have us believe that these exceptional cases fall outside the explanatory domain of psychology, but he does not deny that psychological laws could be constructed to accommodate them. In fact, he seems to concede that we could easily construct such laws if we were to admit narrow contents into our explanatory arsenal. Why should these narrow laws be excluded? Fodor does not provide a satisfying answer. He has no theory-independent reasons for barring the employment of narrow laws for the explanation of Frege cases. As with other sciences, a good psychological theory should explain both normal cases and rare exceptions (unless those exceptions are completely unprincipled). This brings us back to our claim that scientific responsibility should be determined by scientific ability. If we can construct lawlike generalizations subsuming Oedipus, and those generalizations advert to narrow content, then we ought to embrace narrow content.

4. Conclusion

I have argued that Fodor's critique of the case for narrow content is not compelling. In each instance, Fodor fails to demonstrate that broad psychological laws are sufficient for psychological explanation. Here's a recap. First, Fodor told us that twin cases should be barred from psychology because they are nomologically impossible. In response, I argued that barring twin cases would leave us with laws that were explanatorily inadequate because they would rely on a method of individuation that fails to distinguish those properties of mental states which are causally efficacious from those which are not. Next, Fodor turned to neighbors. He argued that some neighbors will be subsumed by broad laws, because they share the same broad contents. Neighbors who do not have the same broad contents should not be subsumed under the same laws, because the similarities in their behavior are accidental. In response to the first case, I argued that we can only assign the same broad contents to neighbors if we adopt a semantic theory that renders garden variety concepts wildly disjunctive. In response to the second case, I argued that accidental similarities in behavior should be captured by laws, if those laws are principled and predictive. After that, I examined Fodor's account of deferential concepts. He argues that broad psychology does not miss any generalizations involving deferential concepts, because such concepts are both broadly and narrowly distinct. I indicated that individualists would openly embrace this conclusion because they have no stake in insisting that concepts like ELM and BEECH have the same narrow content. On the contrary, such a view has been an embarrassment to those individualists who have accepted it. Finally, I presented Fodor's treatment of Frege cases. He contends that psychology does not need to accommodate people who suffer from Fregean conclusions, because such people are irrational. I argued that the failure to identify co-extensional terms is not a symptom of irrationality. Perfectly rational agents do it all the time. A psychological theory that excluded Frege cases would be unable to account for the behavior of agents within its explanatory jurisdiction. In sum, Fodor has not undermined the arguments motivating narrow psychology. Indeed, our discussion of those arguments has reconfirmed the explanatory value of narrow content. Individualists do not need to throw in the towel.

References

Burge, T. 1979: Individualism and the Mental. In P. A. French, T. E. Uehling Jr., and H. K. Wettstein (eds.), Midwest Studies in Philosophy II: Contemporary Perspectives in the Philosophy of Language. Minneapolis: University of Minnesota Press.

Burge, T. 1986: Individualism and Psychology. Philosophical Review 95: 3-45.

Fodor, J. A. 1974: Special Sciences, or The Disunity of Science as a Working Hypothesis. Synthese 28: 77-115.

Fodor, J. A. 1980b: Methodological Solipsism Considered as a Research Strategy in Cognitive Psychology. Behavioral and Brain Sciences 3: 63-110.

Fodor, J. A. 1987: Psychosemantics. Cambridge, MA: MIT Press.

Fodor, J. A. 1989: Substitution Arguments and the Individuation of Beliefs. In G. Boolos (ed.), Method, Reason and Language. Cambridge: Cambridge University Press. Reprinted in Fodor 1990.

Fodor, J. A. 1990: A Theory of Content and Other Essays. Cambridge, MA: MIT Press.

Fodor, J. A. 1991a: Replies. In B. Loewer and G. Rey (eds.), Meaning in Mind: Fodor and His Critics. Oxford: Basil Blackwell.

Fodor, J. A. 1991b: A Modal Argument For Narrow Content. Journal of Philosophy 88: 5-25.

Fodor, J. A. 1994: The Elm and the Expert: Mentalese and Its Semantics. Delivered at Stanford University as the 1993 Jean Nicod Lectures. Cambridge MA: MIT Press.

Loar, B. 1987: Social Content and Psychological Content. In R. Grimm and D. Merrill (eds.), Contents of Thought. Tucson: University of Arizona Press.

Millikan, R. G. 1993: White Queen Psychology; or, The Last Myth of the Given. In White Quine Psychology and Other Essays for Alice. Cambridge, MA: MIT Press.

Putnam, H. 1975: The Meaning of 'Meaning'. In Philosophical Papers, vol. 2: Mind, Language and Reality. Cambridge: Cambridge University Press.

Stich, S. P. 1983: From Folk Psychology to Cognitive Science: The Case Against Belief. Cambridge, MA: MIT Press.

Notes

1. Cf. Fodor (1980), Fodor (1987), Fodor (1991a), and Fodor (1991b).

[return to main text]

2. This is really a simplification. Some individualists have argued that Twin's behave differently, but that the causal determinants of their behavior are type identical because they supervene on the same brain states. Either way, P2 seems to follow. Cf. Burge (1986) and Fodor (1987).

[return to main text]

3. I will use the notation '1.P1' to represent the first premise in argument (1).

[return to main text]

4. Fodor likes to say that concepts refer to properties and their instances and not to classes of things and their members. My arguments will do not depend on this distinction.

[return to main text]

5. Cf. the title papers in Fodor (1990).

[return to main text]

6. Cf. Fodor (1994, p.42). Fodor only labels T1 and T2 truisms. He characterizes T3 as a basic principle of belief/desire psychology.

[return to main text]

7. In earlier works, Fodor seems to be equally convinced of Oedipus' rationality. For example, see Fodor (1989, p.176, note 10).

[return to main text]