Jack Marley-Payne, MIT
Belief has long been held to be connected with both speech and action. However, cases of conflicting behaviour show that only one of these connections can be constitutive. Intellectualism is the view that the connection between belief and speech (and also conscious judgement) is to be prioritized. And, therefore, subjects with conflicting behaviour believe what they say. A prima facie compelling motivation for the view is the claim that beliefs are intelligent states, and only states involved in linguistic and conscious processes are sophisticated enough to meet the appropriate standard of intelligence. In this paper I plan to examine this line of thought and argue that it is mistaken.
We say what we believe to be true. Moreover, our beliefs guide our actions. As rules of thumb these claims are beyond reproach. Indeed they seem so plausible, it’s tempting to infer that they form constitutive conditions for belief. However, this can’t be so: the two are incompatible, when taken as fully general principles. This is revealed by cases of conflicting behaviour in which a subject is disposed to assert one thing, and indeed endorse it upon reflection, while behaving as though she believes the opposite – for example, an elite sportsplayer who makes incorrect statements about how one should play. Therefore, there must either be some beliefs that are not manifested in verbal behaviour or there must be some action guiding states that are not beliefs. We must choose whether our conception of belief is primarily connected with language and deliberation, or with action.
Intellectualism is the view that the connection between belief and speech (and also conscious judgement) is to be prioritized. And, therefore, subjects with conflicting behaviour believe what they say. A prima facie compelling motivation for the view is the claim that beliefs are intelligent states, and only states involved in linguistic and conscious processes are sophisticated enough to meet the appropriate standard of intelligence. In this paper I plan to examine this line of thought and argue that it is mistaken. In short, the reason for this is that verbal states are less sophisticated than might initially be supposed and non-verbal states are more so: both meet criteria for sophistication imperfectly. Therefore, both look equally good candidates to qualify as beliefs.
My argument will procede as follows. First I will describe some cases of conflicting behaviour and the account of them offered by intellectualism (§2+3). Then I will look at two ways to make the argument for intellectualism precise, defended by Gendler (2008a) and Stich (1979) – who focus on evidence sensitivity and inferential integration resepctively as criteria for sophistication (§4+5). My response will be to look at a number of examples that illustrate the imperfect ways in which our mental states meet these conditions. I will conclude by suggesting that we should move away from the intellectualist view entirely and adopt an action based account of belief.
2. Cases of Conflicting Behaviour
The examples of concern are ones in which a subject’s verbal behaviour and conscious judgment indicates belief in a proposition, while her non-verbal non-conscious behaviour indicates a belief in that proposition’s negation. Note that conscious judgment is naturally grouped together with linguistic behaviour as the internal analogue of assertion – a subject will consciously affirm a proposition to herself when she’s unwilling or unable to assert it aloud. I’ll call verbal behaviour and counscious judgment intellectual behaviour, and other kinds non-intellectual behaviour. Now, consider the following cases:
Skywalk: Over the Grand Canyon there is a giant glass bridge that members of the public may walk out onto. The average subject will be well informed of the strength of glass of this thickness; moreover she will sincerely assert the sentence ‘I am in no danger, the bridge will support my weight’. And yet despite this, while on the bridge she trembles and sweats and is eager to get across it as soon as possible.1
Implicit Bias: Psychological studies reveal that many white Americans who profess to being committed to racial equality discriminate against black people in their unconscious behaviour. They sincerely assert that black people deserve equal treatment, are equally trustworthy etc. And yet they are less willing to make eye contact, stand further away and display other ‘micro-aggressions’ during social interactions; they are also less likely to hire black candidates relative to the quality of their CV and discriminate in numerous other ways.1
Roadworks: Ben is told that due to road works, the bridge he normally takes on his way to work will be closed. Upon hearing this, he thinks to himself (consciously) that he will have to take the roundabout route. He is disposed to sincerely assert this to anyone who asks him how he’ll be travelling to work in the next week. However, when he drives to work, he’s disposed to set off on the old route, not leave extra time etc.1
The subject on the skywalk has intellectual behaviour suggesting she believes that she is safe, while her non-intellectual behaviour (her trembling and sweating) suggests a belief that she’s in danger. The implicitly biased subject’s intellectual behaviour indicates belief that black people are equally trustworthy and deserve equal treatment to white people, while the unconscious behaviour indicates a belief that black people are worthy of suspicion. Ben’s intellectual behaviour indicates a belief that the bridge is closed while his non-intellectual behaviour indicates a belief that it is open. The question, in light of this is: what do the subjects actually believe?
To put things neutrally, the subjects have two ‘belief-like states’ – the one guiding intellectual behaviour and the other guiding non-intellectual behaviour. I’ll refer to these as intellectual states and non-intellectual states respectively. So our question can be rephrased as asking: which of these belief-like states are actually beliefs? Both the intellectual and non-intellectual states have an action guiding role appropriate for belief, while (by construction) only the intellectual states have contents the subjects can articulate.4 Therefore, deciding what is to be said about cases of conflicting behaviour requires taking a stand on whether belief is primarily tied to action, or to speech.
3. Introducing Intellectualism
Intellectualism offers a way of understanding these cases. It is the thesis that the subjects believe only what they say (and consciously judge) in these examples – or more generally the claim that only intellectual states are beliefs.5 So according to the view, the subject on the skywalk believes that she’s safe, the implicitly biased subject believes that all races are equal, Ben believes that the bridge is closed – and none of the subjects believe the proposition indicated by their non-intellectual behaviour. Though there are many dissenters,6 intellectualism has been influentially advocated by a number of philosophers. Here are some notable endorsements of the view:
[I]f a subject is psychologically (and physiologically) normal, inclined to be cooperative and has no motivation to deceive us, then if she believes that p and is asked whether p is the case, she will generally say that it is… [In some cases] the subject may be temporarily paralyzed and thus unable to assent to anything. Or he may have a strong desire to mislead his questioner, or simply wish to say nothing. Still, under these circumstances, if we ask a subject whether p is the case, he will generally have a certain sort of characteristic experience…One might also describe the experience as being aware that p or being conscious that p. [Stich (1978) pp 40-41]
[In cases of conflicting behaviour] our beliefs and desires mandate pursuing behaviour B and abstaining from behaviour A, but we nonetheless find ourselves acting – or feeling a propensity to act – in A-like ways… It seems misleading to describe these A-like behaviours as fully intentional [i.e. as manifestations of belief]: their pursuit runs contrary to what our reflective commitments mandate. [Gendler (2012) p 799.]7
[B]eliefs and other cognitive states are analyzed in terms of their dispositions to cause phenomenally conscious episodes of judgment, rather than their dispositions to cause physical behavior. [Smithies (2012) p 348]
[T]he most straightforward expression of a belief is an assertion…beliefs, recognitions and so on, are going to be ascribed to animals in an impoverished and … a somewhat conventionalised sense. [Williams pp 139-140]8
Since intellectualism requires denying a plausible seeming principle about the connection between belief and action, the view requires defense. One might simply rely on intuitions about cases but this would not be a particularly promising strategy since such intuitions would be highly contestable. Much more interesting is if a principled argument can be made for the position. As was mentioned in the introduction, a compelling motivation for the view is the idea that only intellectual states are sufficiently intelligent to qualify as beliefs. Clearly some behaviour guiding states, such as our reflexes, are too crude to be beliefs – so it’s highly plausible to think belief entails a certain degree of sophistication. If one can show that only intellectual states meet this condition, one will have a compelling argument for intellectualism.
Sophistication is a loose notion, so one needs to work with a more precise criterion for belief if one is to assess this line of argument. Two promising strategies here have been to appeal to evidence sensitivity and inferential integration as necessary conditions for belief.9 These seem like a reasonable gloss on sophistication, since responsiveness to the environment and to previously obtained information distinguish an intelligent process from an automatic one. It also seems prima facie plausible that intellectual behaviour has a privileged connection with these features. For example, it is through reading and writing that scientific opinions are expressed and communicated. This linguistic behaviour is sensitive to incredibly complex experimental evidence, as well as the detailed testimony of other researchers; and, it is in writing that researchers tend to work through long chains of inferences when, for example, doing mathematical proofs.
Despite their plausibility, I think both versions of this argument fail, as I’ll now argue. Both intellectual and non-intellectual states are imperfectly evidence sensitive and inferentially integrated. Thus, considerations of sophistication do not count in favour of intellectualism.
4. Evidence Sensitivity
The argument centred on evidence sensitivity is given by Gendler, who writes:
[W]hatever belief is—it is normatively governed by the following constraint: belief aims to ‘track truth’ in the sense that belief is subject to immediate revision in the face of changes in our all-things-considered evidence… In each of the cases we have been considering, only one of the competing tendencies is evidence-sensitive in this way. The man [on the skywalk] believes that he is safe because if he were to gain evidence to the contrary, his attitude would change accordingly.10
To state the argument explicitly:
1) All beliefs must be appropriately evidence sensitive
2) Only intellectual states are appropriately evidence sensitive
C) All beliefs are intellectual states
For this argument to be successful, the intellectualist must specify what level of evidence sensitivity is ‘appropriate’ for belief. A promising starting point is the fact that in all the cases of conflicting behaviour we’ve discussed, only the intellectual state is sensitive to linguistic evidence. For example, the subject on the skywalk’s disposition to assert she’s safe is sensitive to whether the staff tell her that she is; while her trembling and sweating is unaffected by such testimony. One might argue, therefore, that beliefs must be sensitive to all the evidence, and since the non-intellectual states are not sensitive to linguistic evidence they cannot be a manifestation of belief.
Sensitivity to all evidence is a very stringent condition for belief so one might be tempted to weaken it. One option would be to go with the minimal constraint that beliefs must be sensitive to some evidence – and argue that non-intellectual states do not even satisfy this. Another option would be to adopt an intermediate condition: that beliefs must be sensitive to evidence in a sufficiently uniform and consistent manner; it might be thought that our ability to grasp linguistic evidence in a systematic fashion indicates this. Or, following Gendler, one could emphasise immediacy – she claims that belief is ‘subject to immediate revision’: that it can ‘turn on a dime’.11 This is meant to contrast with the slow habitual changes in non-conscious behaviour. Here, then, are four alternative theses that could be used to flesh out the argument for intellectualism:
- Beliefs must be sensitive to all evidence.
- Beliefs must be sensitive to evidence in a unified and systematic manner.
- Beliefs must be immediately revisable in response to evidence.
- Beliefs must be sensitive to some evidence.
The intellectualist must then argue that the intellectual states meet the condition of choice while the non-intellectual states do not, thus securing premise two of the argument. None of these theses can be upheld however: even intellectual states fail to meet the criteria of E1/E2/E3, while many non-intellectual states satisfy E4.12 I will first concentrate on the relationship between linguistic behaviour and evidence and then move onto other types of intellectual behaviour.
4.1 Linguistic Behaviour
Consider the following examples:
Bob the baseball fielder: Bob is waiting to catch the baseball. As the ball is struck he is unable to say where it is going to land; however, if a reliable source were to tell him the coordinates of where it was heading, he would then be disposed to relay this information when asked. On the other hand, the visual evidence of seeing the ball’s trajectory after it’s hit does not help; he is still unable to state the location – though he is able to move himself to the right place in order to catch it.13
This illustrates that often our intellectual states are sensitive to linguistic evidence but not to perceptual evidence. When it comes to the issue of where the ball is going to land, only Bob’s non-intellectual states are sensitive to the relevant visual evidence.
Sam the teenager: Sam makes assertions about what work he must do to get a good job. He says things like “I don’t need to work much at all to get a job: my friends’ older brothers all did far less work than me and they’re all lawyers and bankers now.” However, the intellectual states underlying these assertions are not sensitive to all linguistic evidence. If his friends were to tell him that the job market has changed and he needs to get some internships if he wants to succeed, he would revise what he said. However, if his family were to tell him the very same thing he would not change at all.14
I take it that this type of example is not unfamiliar. It illustrates that sensitivity to linguistic evidence is not an all or nothing matter even amongst intellectual states.15
Rich the carnivore: Rich asserts that there’s nothing objectionable about eating meat, and indeed he eats lots of it. No amount of testimony as to how cruel the meat industry is will make him change what he says. However, if he were to visit a slaughterhouse and see the cruelty (the very same processes that he has already been told about) what he’d be disposed to assert would change.
This shows that what we are willing to say and endorse may not be sensitive to testimony but only perceptual evidence. Again, I take it to be a common phenomenon that seeing something will change our behaviour whereas being told about it will not.
Thomas the slow learner: Thomas is taking a set theory course and struggling a little bit. He’s been told some strange things that he didn’t believe before, such as that there are ‘different sizes of infinity’; and he’s been shown watertight proofs that demonstrate the claims. When he hears these things first time round he’s just confused and doesn’t believe what he’s told. However, he is studious and reads through the proofs multiple times until eventually he ‘gets it’ and believes the theorems.
This shows that linguistic behaviour is not always immediately revisable in the face of evidence which mandates a change. Sometimes (especially in cases where the fact being indicated is especially radical, surprising or strange) it can take a while before what we’ve seen or been told sinks in and we finally change what we are disposed to assert. This phenomenon has been identified and studied in social psychology, and is known as belief perseverance.16
These examples together show that intellectual states fail to satisfy the criteria set by E1-E3 – at least when we restrict our attention to their manifestations in linguistic behaviour. Sam the teenager, Rich the carnivore and Bob the baseball player all illustrate how intellectual states are not sensitive to all evidence (E1). Moreover, Rich shows how intellectual states may not be sensitive to linguistic evidence, and Sam shows that our interactions with linguistic evidence may be messy (E2). Finally, Thomas shows how linguistic dispositions need not be immediately be sensitive to evidence (E3).
4.2 Other Intellectual Behaviour
A natural response to these examples is to point out that intellectual behaviour does not just mean linguistic behaviour. Recall that the various defenders of intellectualism differ as to which types of behaviour they take to be central to belief. First, as has already been discussed, there is the distinction between conscious judgments and overt linguistic behaviour. Second, one can distinguish simple assertion from reflective endorsement – that is, an assertion which is made on the basis of a deliberative process and such that one could offer reasons in support of it. Perhaps if we look at a broader range of phenomena, a connection between intellectual behaviour and evidence will emerge.
It might be suggested that only one type of intellectual behaviour is sensitive to evidence in the requisite sense and that it alone is the mark of belief. Alternatively one might claim that the various types of behaviour tend to come as a package. Though any of these taken in isolation might not seem to have a special kind of evidence sensitivity, taken together they do. I think, though, that neither of these strategies can succeed. None of the types of behaviour have a sufficiently strong connection with evidence and they can all come apart from each other. Turning first to reflective endorsement, consider the following example:
Malcolm the snap-judger: Malcolm is a successful businessman who each day has to make dozens of assertions about people’s character, fitness for a particular job, susceptibility to particular forms of persuasion etc. on the basis of quick decisions. If he were to reflect on any particular assertion he had made, he would often not endorse it; his on balance judgment would go against his unreflective assertion. However, at least in the sphere of business decisions, Malcolm’s snap-assertions are more accurate than his reflective judgements.
This shows how our sincere assertions can come apart from what we reflectively endorse; and that, at least some of the time, what we say unreflectively can be more sensitive to the evidence than what we are disposed to endorse after deliberation. There is empirical evidence backing up the claim that cases like this are quite commonplace. For example, Halberstadt and Levine (1999) asked a group of basketball experts to predict the outcome of a basketball game: half were asked to give and analyse reasons before making their prediction and half were not; those who had to give reasons were less accurate on the whole that those who did not do so. Similarly, Wilson et al (1984) asked subjects who were in a relationship to predict how long it would last, with half giving reasons and half not. Again, it turned out that the predictions of the control group were significantly more accurate.
Of course there are also times when judgments arrived at reflectively are better attuned to the evidence than unreflective assertions. For example, when the subject matter of the assertion is some sophisticated area of science. The correct conclusion is that both types of behaviour are imperfectly sensitive to the evidence, and which fares better varies depending on the circumstances.
The next thing to consider is the relationship between assertion and conscious judgment. It might seem that these two are tricky to pull apart since we are almost always aware of what we are saying. Perhaps there are cases of talking on complete autopilot where we are totally oblivious of the words coming out of our mouth, but the separation here is shallow – we could quite easily become conscious of what we were saying by refocusing our attention. However, this only demonstrates a correlation between assertion and linguistic conscious judgment – internal ‘sayings to oneself’. There are other conscious occurrences that have a claim to being a manifestation of belief and that do come apart from what we say. Consider the following case:
Maya the navigator: Maya lives in Brussels – a city with winding disorderly streets. She is able to navigate it effectively by consulting a mental image she has of the city’s layout. This is a process that occurs reflectively and that she endorses as reliable. However, much of the information that she draws upon, she would be unable to put into words: she could not describe the features that she is aware of when visualising the city.
This illustrates what should be an uncontroversial point: we sometimes have conscious representations – for example mental images – whose content we are unable to fully articulate. These mental images are, moreover, evidence sensitive: Maya’s mental map is formed, and can be updated, as a result of her perceptions of Brussels. Moreover, her verbal behaviour is not as sensitive to this perceptual evidence, since she cannot articulate such a rich description of the city. It should be noted again, though, that in some respects linguistic behaviour will do better at responding to the evidence. For example, if Maya receives a bunch of linguistic evidence about the city’s layout in an Urban Studies course, she will be able to make assertions about the ratio of public space to private space, the percentage of green space etc. And she will not be able to alter her mental image of the city in a way that encodes this information. So again, conscious states and verbal behaviour are both partially sensitive to the evidence, and which fares better depends on the circumstances.17
4.3 Evidence and Non-Intellectual States
We can conclude that intellectual states are sensitive to evidence in an imperfect and complex way. There’s no straightforward method for stating in what ways such behaviour is evidence sensitive; one has to make reference to the type of evidence in question and the circumstances in which the behaviour is elicited. This means that the only plausible evidence based constraint on belief is the minimal E4.
This level of evidence sensitivity is also exhibited by non-intellectual states, however. Though such states are often not sensitive to linguistic testimony, they often are sensitive to perceptual evidence. Consider, for example, Bob the baseball player: his disposition to move to the appropriate location to catch the ball is sensitive to perceptual evidence regarding the ball’s trajectory through the air. Similar things could be said about much action in sport. More generally, we have dispositions to navigate environments correctly that we are not conscious of. For example, there are many buildings that I have visited just a few times and that I am able to find my way around when I return, but that I am unable to visualise. Moreover, in a familiar environment I will be able to reach for door handles, light switches and elevator buttons without looking, despite being unable to picture their location. These dispositions too are shaped by perceptual evidence.
A crucial point is that even the non-intellectual states of the subjects in the original cases may display this level of evidence sensitivity. Note that the examples as presented do not specify whether the non-intellectual states are evidence sensitive. However, there are plausible ways of filling in the details that make this the case. First take the trembling and sweating of the subject on the skywalk. This behaviour is clearly sensitive to perceptual evidence since it’s seeing through the glass bridge that causes it. Moreover, it might be that if the subject were to vigorously jump up and down, throw herself against the walls etc. and see that everything still held firm, her trembling would cease – sometimes this kind of visceral demonstration is an effective technique for getting over unnecessary fear. Second, Ben’s disposition to drive along the wrong road is sensitive to visual evidence – he’ll turn back when he sees signs for diversions around the bridge; moreover, he’ll lose the disposition (we may stipulate) once he’s made the mistake a few times. Delayed sensitivity to the evidence is enough to satisfy E4: we’ve seen that intellectual states may not be immediately sensitive to evidence either, as with Thomas the slow learner.
The case of implicit bias is more complicated since the non-intellectual behaviour in question is manifested in a vast range of situations and, moreover, there is much uncertainty over how it is formed and how it can be changed. There does seem to be evidence that a subject’s behaviour will become less biased if she is exposed to people who do not fit with the stereotype. For example, Dasgupta and Greenwald (2001) found that exposing subjects to admired black exemplars and disliked white exemplars, significantly reduced implicit bias behaviour. And Shook and Fazio (2008) found that white students who shared a dorm room with a black student, saw a reduction in implicit bias over the time they spent together. This suggests that even though testimony that black people are not inferior does not influence implicit bias behaviour, more direct forms of evidence may do – such as witnessing up close a black person who does not conform to the stereotype. This is comparable to the partial evidence sensitivity of Rich the meat eater’s linguistic dispositions.18
Together, these considerations show that the fact that beliefs are evidence sensitive does not speak in favour of intellectualism. In fact, an examination of the relationship between various belief-like states and evidence points in the opposite direction since it reveals important similarities between intellectual and non-intellectual states.
5. Inferential Integration
The second argument for intellectualism appeals to the connection between belief and inferential integration. It seems a plausible general principle that beliefs are inferentially integrated – in a sense to be spelled out below – and this might be thought to support intellectualism. The canonical presentation of this idea is given by Stich (1978), who says: ‘a person’s body of beliefs forms an elaborate and interconnected network with a vast number of potential inference patterns leading from every belief to almost any other.’19
He claims, moreover, that the states which have this property are exactly those states that we have ‘conscious access’ to – where the mark of conscious access to a state with the content p is the ability to say that p when asked, or to consciously judge that p when considering the question.20 More recently, this line of argument has been pursued by Neil Levy, specifically to argue that implicit biases are not beliefs. He puts it as follows:
Implicit attitudes are not beliefs. They do not feature often enough and broadly enough in the kinds of normatively respectable inferential transitions that characterize beliefs… they exhibit some of the kind of inference aptness that characterize beliefs. They do so in a patchy and fragmented manner, which indicates they have propositional structure. They are patchy endorsements. [Levy (2014 a) p. 18]21
To be explicit, the intellectualist can be seen as making the following argument:
1. All beliefs are inferentially integrated.
2. Only intellectual states are inferentially integrated.
C. Only intellectual states are beliefs.22
There certainly seems something right about the idea that beliefs must be inferentially integrated (premise 1): intuitively, beliefs are the kinds of thing that can result from and produce inferences. Moreover, if behaviour results from an inferentially integrated network of states, so that a large body of information is brought to bear on it, this seems like a good reason to think of it as intelligent, rather than a reflexive response to a given stimuli. And as was mentioned above: beliefs produce intelligent behaviour.23
Stich motivates the claim that only intellectual states are inferentially integrated (premise 2) by example. Suppose that we cognitively encode certain propositions about the syntax of our language – let p be such a proposition (e.g. one providing some constraint on anaphora binding). This would be a paradigm case of a non-intellectual state. Stich notes that a subject might, as a result of her research in linguistics, believe that if p holds then Chomsky is mistaken – she would assert such a thing in conversation – and she might also cognitively encode p, employing it in her processing of language. However, she would not be in a position to conclude that Chomsky is mistaken: if you asked her whether Chomsky was mistaken, she’d say she didn’t know, perhaps adding that she didn’t know whether p held.24 This shows that the subject’s representation of p is not inferentially integrated with her belief that if p then Chomsky is mistaken. On the other hand, if a subject was disposed to assert both ‘p’, and ‘if p then Chomsky is mistaken’ then she would presumably be able to conclude that Chomsky is mistaken. I think that Stich is right about this example but I don’t think all non-intellectual states are like cognitive encoding of linguistic rules.25
Before assessing the argument, though, a clarification is in order about how to think about inference. Stich’s main example of it is modus ponens. This might suggest that inference is a relation that can hold only between entities with a sentential structure, since the most natural way to think of modus ponens is as a relation between sentences (you need a conditional, after all). However, it’s clear that our ordinary conception of inference doesn’t have this requirement – when it comes to imagistic reasoning for example. Suppose I have a Klee painting that I want to hang in my kitchen; I may visualise both the room’s layout and the painting in order to form a belief as to whether it is too big to fit between the fridge and the oven. This process seems a perfectly good example of an inference – I draw upon information I have about the painting and my kitchen to see whether a particular action is possible. Moreover, it’s an open question whether such processes involve analogue as well as sentential representations – the imagery debate is not yet resolved.26 We are happy to classify processes as inferences without presuming they involve only sentential entities.
I think, therefore, that we should work with a minimal conception of inference which makes no assumptions about the types of representation it involves – sentential, analogue, or whatever. I do not propose to offer an analysis of the concept, since I think we have a decent intuitive grip on it. In what follows, I will work with this intuitive understanding.
Now returning to the argument, it’s important to distinguish two theses about inferential integration – strong and weak.
Strong integration: A state is a belief only if it is inferentially integrated with all other beliefs.
Weak integration: A state is a belief only if it is inferentially integrated with a sufficient number of other beliefs (but not necessarily all of them).
My strategy will be to argue that the strong integration thesis is false, and that the weak integration thesis, though plausible, does not favour intellectualism since non-intellectual states also satisfy the conditions it sets.27 To put it in the terminology of Levy, intellectual states are also patchy endorsements, so being a patchy endorsement does not preclude a state from being a belief.
Strong integration is too strong since not even paradigm cases of belief by the intellectualist’s lights meet it – some intellectual states cannot be integrated with others. First, it’s clear that the information manifested in linguistic behaviour cannot always be integrated with that represented in conscious mental images. Recall the example of Maya the navigator: she cannot verbally articulate the information represented in her mental map, and she cannot use the theoretical information she asserts to modify the map. In fact, experiments have shown that in certain cases, attempts to articulate the contents of visual memory – e.g., to describe a remembered face – actually degrade it.28
The intellectualist might respond by denying that mental images are manifestations of belief – that all belief must be manifested in linguistic behaviour. As well as leading to an unattractive picture of belief, this move fails to rescue strong integration. For example, sometimes whether we are able to provide a given piece of information verbally depends on what question we are asked. Suppose you are asked ‘is there a four letter English word ending E-N-Y?’ – you might well not be able to answer. However if you are asked ‘how do you spell deny?’ you will correctly answer ‘D-E-N-Y’. So you can access the information ‘the English word deny is spelt D-E-N-Y’. Moreover, you might well be able to give examples of words ending, A-N-Y, I-N-Y, and O-N-Y on demand (‘many’, ‘tiny’, ‘pony’). And presumably you could say that ‘puny’ is spelt P-U-N-Y. So you have information that together entails that for every vowel x, there is an English word that ends x-N-Y. However, (prior to reading this paragraph) you were not capable of putting this information together to draw this inference.29
In response to this, the defender of strong integration might appeal to cognitive architecture. My examples show that sometimes subjects are not, intuitively speaking, able to bring together in inference certain of their intellectual states. One could argue that this is not how the claim that ‘all beliefs can be inferentially integrated with all other beliefs’ is to be interpreted. Instead what might be relevant is some sort of in principle accessibility – that there are no barriers in virtue of cognitive architecture, only ‘performance limitations’.
I don’t think this reply is adequate. First, it’s not at all clear that all and only intellectual states are inferentially integrated in this sense. If the mind is ‘massively modular’, as is argued by Carruthers (2006), then intellectual states will themselves be encapsulated from each other. Thus, there would be no beliefs at all according to this version of strong integration.
Moreover, when we switch from an intuitive to a theoretical notion of inferential integration, premise 1 of the argument loses its intuitive plausibility. Though it might be the case that the nature of belief is determined by unexpected findings in empirical psychology, how the two are connected is tied up with wide ranging and controversial methodological questions. This is a matter I’ll discuss briefly in the final section, but for now I want to put aside strong integration and turn to weak integration. My contention is that many non-intellectual states are weakly integrated. One of the most compelling examples of this is, I think, behaviour in sport. Consider the following case:
Federer: In the 2006 Wimbledon final between Federer and Nadal, the following rally takes place. Federer hits three backhands down the line, causing Nadal to stay planted on that side – lulled into a false sense of security. He then hits two hard shots to the opposite side, taking advantage of Nadal’s flat-footedness and forcing him to scramble. This leads Nadal to play a weak short shot, allowing Federer to return at a very sharp angle so that he wins the point.30
In this example, Federer acts in an intelligent way, executing a complex and difficult plan to win the point. It is not habitual or reflexive behaviour, since it is tailored to Nadal’s specific abilities – against other players Federer could have tried to win the point sooner, but Nadal is exceptionally quick. It’s also open to modification depending on exactly what Nadal throws back, and when and how he gets wrong-footed. This strongly suggests it is the product of a network of inferentially integrated states since it is sensitive to information received from a variety of sources over an extended period of time – background knowledge of Nadal and of Federer’s own abilities, and perceptual information about what the ball is doing, the court conditions etc. It is not, though, the product of a conscious process of deliberation; it happens far too fast for that.31
There is also this level of inferential integration in the cases mentioned in previous sections. Bob the baseball player’s disposition to move to catch the ball may have been calibrated on the basis of a wide variety of information such as his visual perception of the ball, his sense of the wind strength and direction, and his kinaesthetic sense of the state of his own body (how fast he can run, dive etc.). Ben is not only disposed to drive on the old route to work, but will take his dry cleaning even though he only passes the dry cleaners on the old route; and if he fancies a bagel, he will change the way he is driving to go past the deli etc. Recall also the example of a building I have visited a few times and can remember my way around as I go – though I can’t articulate or visualize its layout. When I enter it, I will alter my behaviour depending on what I aim to achieve there, and also if I get new perceptual evidence that the layout has changed. With implicit bias the characteristic behaviour manifests itself in a variety of ways – body language, speech, workplace decisions etc. It also appears to draw on background information in its activation – hiring decisions are unconsciously influenced by the racial connotations of the name on the CV and so would appear to be mediated by background beliefs about which names are typical of which races.32
I conclude that the appeal to inferential integration to establish intellectualism fails.33 In the next section I’ll briefly look at where this leaves us when it comes to understanding belief.
6. The Way Forward
We have seen that intellectual states cannot be singled out as states of particular interest by considerations of sophistication. Indeed, intellectual and non-intellectual states form a unified kind in virtue of both being imperfectly evidence sensitive and inferentially integrated. I think this makes a more liberal account of belief attractive: one on which all sufficiently sophisticated states with the appropriate action guiding role count as beliefs.
It is of course open to the intellectualist to deny such a claim, but I think at this point the burden of proof is on them to give an argument for restricting what qualifies as a belief. There are, moreover, methodological grounds for resisting such a move. By ensuring a tight link between belief and intelligent behaviour, the action-based account picks out a practically significant category. If states are evidence sensitive then we are able to work out when an agent is in them on the basis of her environmental setting – I know you believe that it’s raining because you are sat in front of a window and can see the rain coming down. Moreover, this feature allows us to influence such states by presenting new evidence. If states are inferentially integrated, they will influence behaviour in a systematic way over a range of situations. Thus, knowing when subjects possess such states allows us to predict, explain and influence their behaviour in a systematic manner. This has been held to be a central feature of belief (and belief attribution) by philosophers as otherwise opposed as Dennett (1971) and Fodor (1987).
Thus an action-based account of belief accords with the central role of belief ascription, while an intellectulist account hampers it. I think this gives us good reason to prefer the action account.
Anderson, C. A. (2007). Belief perseverance. Encyclopedia of Social Psychology. Thousand Oaks, CA: Sage, 109-110.
Brownstein, M., & Madva, A. (2012). Ethical automaticity. Philosophy of the Social Sciences, 42(1), 68-98.
Carruthers, P. (2006). The architecture of the mind. Oxford University Press.
Dasgupta, N., & Greenwald, A. G. (2001). On the malleability of automatic attitudes. Journal of personality and social psychology, 81(5), 800.
Davidson, D. (1975). Thought and talk. Mind and language, 7-23.
Davies, M. (1989). Tacit knowledge and subdoxastic states.
Dennett, D. C. (1971). Intentional systems. The Journal of Philosophy, 68(4), 87-106.
Fodor, J. A. (1983). The modularity of mind: An essay on faculty psychology. MIT press.
Fodor, J. A. (1987). Psychosemantics. The MIT Press.
Frankish, K. (2004). Mind and supermind. Cambridge University Press.
Gendler, T. S. (2008a) Alief and belief. Journal of Philosophy, 105(10), 634.
Gendler, T. S. (2008b). Alief in action (and reaction). Mind & Language, 23(5), 552-585.
Gendler (2012) Intuition, Imagination, and Philosophical Methodology: Summary. Analysis 72 (4): 759-764
Gertler, B. (2011). Self-knowledge and the transparency of belief. Self-Knowledge. Oxford: Oxford.
Halberstadt, J. B., & Levine, G. M. (1999). Effects of Reasons Analysis on the Accuracy of Predicting Basketball Games1. Journal of Applied Social Psychology, 29(3), 517-530.
Kosslyn, S. M. (1996). Image and brain: The resolution of the imagery debate. The MIT Press.
Levy, N. (2014a). Neither Fish nor Fowl: Implicit Attitudes as Patchy Endorsements. Noûs.
Levy, N. (2014b). Consciousness, implicit attitudes and moral responsibility. Noûs, 48(1), 21-40.
Mandelbaum, E. (2012). Against alief. Philosophical Studies, 1-15.
Powers, L. H. (1978). Knowledge by deduction. The Philosophical Review, 87(3), 337-371.
Pylyshyn, Z. W. (1981). The imagery debate: Analogue media versus tacit knowledge. Psychological review, 88(1), 16-45.
Ryle, G. (1949). The concept of mind. Routledge.
Schooler, J. W., & Engstler-Schooler, T. Y. (1990). Verbal overshadowing of visual memories: Some things are better left unsaid. Cognitive psychology,22(1), 36-71.
Schwitzgebel, E. (2010). Acting contrary to our professed beliefs or the gulf between occurrent judgment and dispositional belief. Pacific Philosophical Quarterly, 91(4), 531-553.
Shoemaker, S. (2009). Self-intimation and second order belief. Erkenntnis,71(1), 35-51.
Shook, N. J., & Fazio, R. H. (2008). Interracial Roommate Relationships An Experimental Field Test of the Contact Hypothesis. Psychological Science,19(7), 717-723.
Smithies, D. (2012). The Mental Lives of Zombies. Philosophical Perspectives, 26(1), 343-372.
Stalnaker, R. (1984). Inquiry.
Stalnaker, R. (1991). The problem of logical omniscience, I. Synthese, 89(3), 425-440.
Stich, S. P. (1978). Beliefs and subdoxastic states. Philosophy of Science, 499-518. In his (2011) Collected Papers: Volume 1 (page references to the latter).
Tye, M. (2000). The imagery debate. Mit Press.
Wallace, D. F. (2006). Federer as religious experience. New York Times, 20.
Williams, B. (1973). Deciding to believe. Problems of the Self, 136(151), 401-419.
Williamson, T. (2007). The philosophy of philosophy. John Wiley & Sons.
Wilson, T. D., Dunn, D. S., Bybee, J. A., Hyman, D. B., & Rotondo, J. A. (1984). Effects of analyzing reasons on attitude–behavior consistency. Journal of Personality and Social Psychology, 47(1), 5.
Zimmerman, A. (2007). The nature of belief. Journal of Consciousness Studies, 14(11), 61-82.
- This example comes from Gendler (2008a).
- See, e.g., Gendler (2008a), Schwitzgebel (2010), Gertler (2011) for discussion.
- This example comes from Schwitzgebel (2010). Gendler’s (2008a) case of the lost wallet, and Zimmerman’s (2007) case of Hope and the Dustbin are similar in form.
- I’m assuming that assertion is itself a type of action.
- My arguments apply equally to the broad and narrow version of the thesis so I will not dwell on which is the more plausible.
- There are many ways one could make such a denial. For example: Shoemaker (2009) claims that both intellectual and non-intellectual states are beliefs, so that the subjects in the examples above have contradictory beliefs; Schwitzgebel (2010) argues that intellectual and non-intellectual states are both parts of a single complex belief state, so that subjects with conflicting behavior are in a state of ‘in between belief’; Gertler (2009) claims that the non-intellectual states alone are beliefs.
- Gendler claims that the non-intellectual states I’m considering along with many others form a distinctive mental kind she calls ‘alief’. It should be noted that my arguments do not entail that all cases Gendler classifies as alief are since many of them – such as priming effects – lack the appropriate degree of sophistication or action guiding role. Thus I leave room for a less expansive notion of alief.
- See also Davidson (1975), Brownstein & Madva (2012) and Zimmerman (2009) for similar ideas.
- There have been other attempts to capture the complexity of belief similar in spirit to the ones I’m considering – for example, Davies’ (1989) appeal to the generality constraint. I think the objections I raise will carry over to alternative formulations. See note 30 below for further discussion.
- Gendler (2008b) pp. 565-566. In a précis of her work on belief (2012, p 763) she says: ‘beliefs are, roughly speaking, evidentially sensitive commitments to content that are quickly revisable in the face of novel information’ – so this appears to be her considered position. See also Brownstein and Madva (2012) p71 who make a similar argument concerning cases of implicit bias.
- Gendler (2008b) 565-566. She repeats this claim in her (2012) – see note 11.
- This line of thought is endorsed by Schwitzgebel (2010).
- This example comes from Stalnaker (1991)
- Variations on this example are possible where what decides whether Sam is sensitive to linguistic evidence is not his social relation to the speaker but whether the speaker talks in his teenage vernacular, is sufficiently rhetorically adept, charismatic etc. This further illustrates the messiness of our relation to linguistic evidence.
- One might object that it could be that Sam’s intellectual states are sensitive to the testimony of his parents; it’s just that he has a standing belief that they are unreliable testifiers, which defeats such evidence. However, we can imagine that Sam is disposed to sincerely assert that his parents are reliable sources of information, and yet ignore their advice regardless – and the intellectualist is committed to saying that he believe what he says. I’m sure this is in fact the case with implicitly biased subjects – they say, e.g., that the testimony of black people is equally reliable and yet they completely ignore it.
- See Anderson (2007).
- As variant on this case we can consider auditory rather than visual imagination. For example, a jazz musician without formal training might be able to imagine how various motifs would sound, and which would work in a given circumstance – this might guide her playing though she would be unable to verbally articulate what she was imagining. Moreover, the imaginative abilities might be updated as she heard new performances.
- Schwitzgebel (2010) pp. 539-541 makes a similar point
19. Stich (1978) pp. 42-43.
20. ibid pp. 40-41.
- See also Levy (2014b)
- I think that Stich himself is working with a slightly different dialectic. He takes it to be intuitively obvious that all beliefs are intellectual states (mistakenly, in my view) and wants to argue that inferential integration is a property distinctive of them – and thus that our concept of belief is a category of theoretical interest.
- A related argument is Fodor’s (1983) influential claim that beliefs are states within the central system – as opposed to informationally encapsulated modules. I stick with Stich’s formulation since it requires less technical machinery.
- See ibid p 44
- Levy appeals to empirical data to argue that implicit biases are not inferentially integrated.
- See, e.g., Tye (2000).
- So in the argument above, if we work with strong integration then premise 1 is false, while if we work with weak integration premise 2 is false – either way the argument is unsound.
- See Schooler and Engstler-Schooler (1990).
- The original trick regarding ‘deny’ is found in Powers (1978).
- This is a paraphrase of the commentary given by David Foster Wallace (2006).
- One might argue that Federer possesses ‘mere practical knowledge’. However, the nature of practical knowledge – in particular whether it involves beliefs – is very much an open question. Indeed I take it to turn on the kinds of issues being discussed in this paper. Therefore, it would be begging the question to assume practical knowledge is not belief.
- See Mandelbaum (2012) for further argument that implicit bias must involve inferential reasoning.
- As I mentioned above, Davies (1989) – among others – has argued that one better captures the sophistication of belief by appeal to the generality constraint, rather than inferential integration. I think if one were to try to run the argument in these terms, my objection would still apply: intellectual states only satisfy the generality constraint partially and non-intellectual states do so too. Spelling out this argument, however, would requiring setting everything up in terms of the contentious theory of concepts the generality constraint presupposes, so I won’t pursue the matter here.