1 Intentional States and Intentional Attributions

There can be no doubt that our intentional speech is extraordinarily rich. Natural languages, such as English, distinguish a great number of intentional verbs. One can say of Caesar that he believed that Rome is situated on the Tiber, that he desired Gaul to be conquered, that he hoped that his actions will result in the conquest of Gaul, that he regretted that he allowed Pompey to be murdered, that he was afraid of the Ides of March, that he understood well the Roman raison d’État, that he doubted that he might be assassinated, that he expected that Cato will never support him, that he remembered that he was an enemy of Sulla, that he predicted the end of the First Triumvirate, that he intended Pompey to marry his daughter, and that he knew that all men are mortal. All such statements (together with numerous others) attempt to classify Caesar’s mental states both in terms of the content and in terms of the psychological mode (to borrow Searle’s terminology).Footnote 1 The former aspect attempts to capture the representational or semantic dimension of the states of Caesar (as representing the world being made in a certain way). The latter places Caesar’s states in a complex web of relations with other states of Caesar’s, his potential actions and stimuli. Both aspects contribute to what Akeel Bilgrami called”the point and rationale” of intentional states, i.e. their role in “common sense psychological explanation of human behavior” (cf. Bilgrami 1992, p. 15).

That said, one must clearly note that probably nobody has ever claimed that every distinction between non-synonymous attitude verbs (and corresponding attitude attributions Footnote 2) entails the distinction between different types of intentional states. Footnote 3 In other words, probably nobody has ever claimed that all linguistically encoded kinds of constraints that can be put on states of cognitive agents are relevant to the individuation of their states qua intentional states (in other words: there is clearly no one-to-one correspondence between intentional verbs and intentional states). Quite the contrary: knowledge attributions, for example, differ in general from belief attributions (due to their factiveness, among other things). Nevertheless knowledge states are belief states of a certain sort.Footnote 4 Intentional language is obviously an imperfect indicator of the intentional ontology.

Given this much, when it comes to the linguistically guided intentional ontology it may be useful to distinguish two kinds of candidates for real intentional states: ones that may be called basic and others that may be called impure (a similar distinction applies to intentional attributions). In the case of the former kind we assume that simple intentional sentences neither logically entail other simple intentional sentences nor entail sentences about non-mental events or states of affairs. In such a case an intentional verb of the sentence may be said to be an indicator of a basic intentional state. In the case of the latter kind the opposite holds: either the sentence entails other simple intentional sentences or sentences about non-mental events or states of affairs. For the present purpose it is sufficient to bear in mind that intentional impurity is something that distinguishes belief and knowledge states or belief states and fancying states. Those of the latter kind are impure, those of the former are not. The use of the phrase “candidate for a real intentional state” is deliberate. I do not consider the basic/impure distinction as having a direct existential impact. At the end of the day it may happen that basic intentional attributions do not correspond to a real kind of intentional states or that impure intentional attributions do correspond to a real kind of intentional state. Nothing in the discussion below presupposes a particular decision on that complicated matter.Footnote 5

It may be worth noting that early attempts to provide criteria for singling out the class of intentional states were, among other things (and, probably, contrary to the intention of their proponents), the criteria of distinguishing basic kinds of intentional states. Chisholm, for example, provides the following condition:

(…) any compound sentence which contains a propositional clause (…) is intentional provided that neither the sentence nor its contradictory implies either that the propositional clause is true or that it is false” (Chisholm 1957, p. 171–172)

As one of three criteria for singling out the class of intentional sentences (and states). Unless very benevolently interpreted, this criterion excludes, inter alia, knowledge, memory and regret from the class of intentional states. Benevolently interpreted, it allows us to draw the following conclusion: if a compound sentence which contains a propositional clause does imply that the propositional clause is true (or false), then it is either a non-intentional sentence or an impure intentional sentence.

Due to the variety of distinct intentional attributions and the ontological doubts regarding the existence of intentional statesFootnote 6 it might be useful to think about sundry intentional states as varying with respect to their location in a complex pragmatically determined quality space that we might call an intentional space. The concept of “quality space” is used here in a manner similar to Gärdenfors (Gärdenfors 2000) and it follows on from Gärdenfors in being a consistently conceptualistic approach to properties.Footnote 7 Within such an account one may investigate conceptual (“what is the concept of belief?”) and ontological (“what is a belief state?”) questions independently.

Now, within the intentional space one can distinguish an inferential dimension that, for instance, connects beliefs and knowledge states or knowledge states and the truth of their propositional contents. There is also a causal dimension that requires states with particular psychological modes to have appropriate causal connections with other states, actions and stimuli (connections of this sort should be somehow reflected in intentional laws and law-like generalizations). Yet another dimension that might be called formal, and which was comprehensively described by Searle, characterizes states in terms such as direction of fit and causal self-reference.

It was probably Sellars who first forcefully argued for including a normative aspect in the list of dimensions of the intentional space. He noticed in the famous passage from Empiricism and the Philosophy of Mind that:”(…) in characterizing an episode or a state as that of knowing, we are not giving an empirical description of that episode or state; we are placing it in the logical space of reasons, of justifying and being able to justify what it says.” (Sellars 1956, p. 169). Sellars thesis might be generalized to attitudes other than knowledge. Roughly speaking, one might propose for each intentional state (of a type V that somehow corresponds to an intentional verb V) a version of Sellars hypothesis:

(Generalized Sellars Hypothesis) In characterizing an episode or a state as that of V-ing (believing, doubting, thinking, hoping etc.) we are not giving an empirical description of that episode or state; we are placing it in the normative space of obligations, of being able to do O or being able to end up in a state S.

An intuitive assessment of the Generalized Sellars Hypothesis (GSH henceforth) depends, of course, on the choice of V, O and S. In one interpretation, intentional normativists presuppose that GSH is basically applicable to one particular kind of intentional state, i.e. beliefs. It is important to note that GSH is connected with normativity in a dual manner. Firstly, it imposes on the intentional subject a requirement of having a capability of doing something (or that of being in a particular state). For instance, one might argue that believing that p imposes on a cognitive agent an obligation of being able to defend every logical consequence of p that they grasp, it also imposes on them a slightly more elusive obligation of being able to justify the belief. Secondly, the fact that they are expected to meet the obligations in question results in the fact that they are allowed to assess whether (and to what extent) they meet the appropriate obligations. Thus, if they are unable to justify their belief, we will assess it as unjustified or poorly justified and, if they are unable to defend (or simply: do not defend when required) some simple logical consequence of a particular belief, we might suspect them of being irrational, inconsequential or insincere (and the belief of being haphazard).

2 Meanings of “Intentionality is Normative”

Now I would like to suggest that we can classify intentional normativists as either claiming that an interesting type of the normativity of the intentional has its source in one or other version of GSH, or as claiming that it has its source somewhere else. In other words, that the normativity of the intentional is either connected with the fact that being in an intentional state of a particular type imposes some obligations on the agent of the state, or that it is connected with the fact that some or all intentional states have direct or indirect normative aspects. The latter option might take various forms. Basically it might mean that intentional states can be externally classified as correct or incorrect simpliciter. It might also mean that commonsense intentional explanations have normative presuppositions. This interpretation of the IN slogan has been popularized by Davidson and McDowell (cf. Davidson 1970; McDowell 1985). Although I feel slightly uncomfortable to dub this interpretation a version of the IN thesis,Footnote 8 it is quite popular among various authors. Hence, I will take it on par with others.

The diagram below schematically represents the possible interpretations of the IN slogan. Let me just note that I do not presuppose the options labeled as [2A] and [2B] to be mutually exclusive. I treat the options as separate because it is logically possible to be committed to one of them without being committed to othersFootnote 9 (Fig. 1).

Fig. 1
figure 1

Possible interpretations of the normativity of the intentional

At one end of the spectrum [1A] are those philosophers who accept some version of GSE and who conceive the normativity of the intentional in terms of obligations put on agents of the states by the very fact that they are in the intentional state in question. It is important to keep in mind that the term “obligation” as used here is open to several interpretations. Although one may be tempted to think that it hardly makes sense to speak of states as imposing normative requirements on agents unless they involve a social or an institutional aspect, this in fact depends on what one thinks of the (supposed) norms that appeal to someone’s being in a given intentional state. Intentional states in themselves (without relevant norms) cannot have a normative impact. Available alternatives here range from taking such norms as purely social or institutional, to taking them as mind and society independent, possibly appealing to non-natural properties.Footnote 10 Of course, the latter options do not look very attractive (as Geach noted once: “(…) nobody has ever given a coherent and understandable account of what it is for an attribute to be non-natural” (Geach 1956, p. 35)). However, the option just described is not committed to any particular view on norms.

We might refer to such theories as imposition theories (and to the normativism as Imposition Normativism). As we have seen, Sellars is a philosopher who defended such a view for knowledge states (or: belief states meeting certain additional conditions). It also applies to those philosophers who accept the inferences similar to the following case of Searle’s is-ought derivation (cf. Searle 1964):

  • Monica believes that Bombay is the most populous city in India.

  • Monica places herself under an obligation to deny that Delhi is the most populous city in India.

  • Monica is under an obligation to deny that Delhi is the most populous city in India.

  • Therefore: Monica ought to deny that Delhi is the most populous city in India.

One possible worry regarding this interpretation of the IN thesis is that it seems to, at least prima facie, contradict Hume’s Law and, as such, is guilty of naturalistic fallacy.Footnote 11 But this is not the interpretation that is intended by philosophers who adhere to this version of IN. When asked, they will say that the argument shows that there are no such things as brute intentional facts. This reaction is consistent both with intentional irrealism and with the belief that the term “obligation”, as used in the argument, has a constitutive sense (and that there are institutional facts that make the appropriate obligations, to use Searle’s phrase, institutionalized). This constitutive interpretation of IN neatly corresponds to other claims to the effect that certain normative statements count as conceptual truths or meaning postulates for some intentional attributions. For instance, some authors claim that it is a conceptual truth about beliefs that we ought to believe truths.Footnote 12 Similar things can be probably said about the principle stating that if x believes that p, p entails q and x grasps that p entails q, then x ought to believe q.Footnote 13

There exists a version of the imposition approach that says that intentional states might put certain normative requirements on actions that are causally connected with the intentional states in question. For instance, my desires might be causally responsible for some actions aimed at achieving the goals specified in those desires. However, the actions in question will count as incorrect or inappropriate if they, in fact, lead to results that exclude the possibility of achieving the goal of the desire or, at least, if they bring about some unexpected effects. For example, it is widely argued that the American policy (that can be understood in terms of desires of individuals in power) aimed at eradicating Saddam Hussein’s regime was causally responsible for sending troops to Iraq. This resulted in a serious destabilization of the region. This unforeseen consequence might result in evaluating the action in question as irrational, unwise or incorrect.

To sum up: it seems that [1A] has at least two versions: State Normativism that locates the agent’s obligations in the domain of other intentional states of the agent (for example, by claiming that belief p puts on the agent an obligation to believe that non–non-p) and the Action Normativism that locates the agent’s obligations in the realm of actions that are caused by their intentional states. The Normative Essentialism of Zangwill (Zangwill 2005) that postulates obligation only in the realm of “horizontal relations” (i.e. relations between mental states) is an example of the former theory. Sellars’s approach to knowledge states is an example of the latter. As explained above, both interpretations presuppose the existence of norms that appeal to the relevant states or actions.

Let us consider the option [2A] now (we may dub it Assessment Normativism). There can be no doubt that we actually classify some intentional states as correct, incorrect, rational, irrational etc. However, without further analysis, this simple fact remains oblivious with respect to various possible uses of terms like “correct” (in particular, we have no self-evident reason to believe that such terms, when applied to intentional states like beliefs, are used in a homogenous manner). Thus, the proponents of [2A] attempt to unpack the notion of correctness. One way of doing this is to attempt to distinguish the descriptive and the explanatory aspect of that concept. The former is an attempt to describe in normative terms the essential features of intentional states. The latter attempts to explain, by “mentioning” normative properties, the fact that one is capable of being in certain intentional states.

When describing the nature of correct intentional states one might refer to concepts such as goal oriented standards that are supposed to essentially regulate and guide certain types of mental states. Wedgwood, for instance, describes this option as follows:

(…) the property of being a correct mental state is based on two claims. The first of these claims is that it is essential to the relevant type of mental state that it tends to be regulated by certain standards of rationality or justification. The second claim is that these standards have some ultimate point or purpose; for example, the ultimate point or purpose of conforming to the standards of justified or rational belief is to ensure that one believes the proposition that one is considering if and only if that proposition is true. (Wedgwood 2007, p. 154)

Still, what is the meaning of phrases like “regulate” and “guide”? “Regulating” is usually understood as having a regular and systematic influence on forming, maintaining and discarding beliefs (thus, “normative theories of belief formation” (cf. Stich 1983, p. 2) can be seen as variants of [2A]). According to this understanding of “regulation” the following picture emerges: there are certain norms (laws of logic, for example), which are regularly used to assess (and self-assess) certain intentional states; this assessment results in regular and predictable “behavior” of the intentional structure of the agent whose states are being assessed, i.e. it results in contractions, revisions and expansions of that structure.

With that in mind, what does it mean that a tendency for being regulated by a standard is essential to an intentional state? It cannot clearly mean that the states in question are in fact assessed and intentional structure of the agent is in fact updated as a result of this assessment (the adjective “essential” would be redundant in that case). It also cannot mean that this is necessarily the case since this is plainly false: it is perfectly conceivable that intentional states are either not assessed according to relevant norms or that intentional structure is insensitive to that assessment (this happens often in the actual world). Finally, it also cannot mean that certain mental states are necessarily capable of being assessed and updated according to some standards. The reason is that this extremely weak claim holds for every arbitrary standard one can formulate (for every possible world where a state is capable of being assessed and updated according to a standard S there is a possible world where an intentional state is capable of being assessed and updated according to a standard N, incompatible with S). Consider this example: Josh believes falsely that Chicago is bigger than Los Angeles, he also learns that Los Angeles is the second biggest city in the USA. Based on that he forms a false belief that Chicago is the biggest city in the USA. According to Wedgwood (ibid. p. 162) there are two standards for beliefs:

  • (Correctness) A belief is correct iff it is true

  • (Rationality) A belief is rational with respect to some body of information BI iff it is sufficiently likely that BI makes the belief true

In that case, Josh’s belief that Chicago is the biggest city in the USA does not meet the Correctness standard. On the other hand it meets the Rationality standard with respect to the body of information that Josh has. There can be no doubt that Josh’s system of beliefs might be updated if assessed according to the Correctness norm. But it also might be preserved if assessed according to the Rationality norm and it also might be preserved if assessed according to the following Pascalsque “norm”:

(Stupefying) A belief is correct iff it is uncritically accepted.

And again, it might be updated if assessed according to the following “norm”:

(Papal norm) A belief is correct iff its content is identical with one of the Pope’s beliefs

All these “norms” might be used to assess and regulate the beliefs of Josh.

The source of the problem ought to be clear. Wedgwood attempts to provide essential properties of intentional states in dispositional and relational terms, i.e. he attempts to say that it is essential for intentional states to be disposed to being possibly related in a particular way to certain norms. However, it is, in general, hard to single out essential properties in that manner. In particular, the state in question is disposed to being possibly related in a regulative manner to a great number of other external norms. Consider a parallel case: someone proposes that it is an essential property of dogs to have a disposition to having an owner (or: to have a disposition to being possibly related in an “ownership manner” to some human beings). We will say that this property is a very bad candidate for an essential property of dogs partially because dogs also have a disposition to not having an owner, as well as several others uninteresting relational dispositions (for instance, dogs have dispositions to being smaller/bigger than some cats, they have a disposition to having rabies/to not having rabies, they have a disposition to running faster/slower than rats etc.). Roughly speaking, dispositions might be constituents of the essence of something but—if one ignores properties that in “no way limit the logical potential of an individual” (cf. Stalnaker 1979)—they can hardly be dispositions to being possibly externally related to something. Thus, it is difficult to say what it means that “it is essential to the relevant type of mental state that it tends to be regulated by certain standards of rationality or justification.”.

What about the explanatory aspect of the concept of correctness? Wedgwood describes the issue as follows:

(…) the core of this normative theory of intentional mental states (…) will need to be surrounded by a shell that explains what it is for a thinker to possess those concepts, or to be capable of those attitudes. If this is the correct account of the nature of the concepts that are expressed by ‘ought’ (or of the concept ‘yellow’, or of logical concepts like ‘or’ and ‘not’), and of the nature of such attitudes as belief or choice, then what has to be true of a thinker if she is to possess these concepts, or if she is to be capable of having these attitudes? (Wedgwood 2007, p. 163)

and gives the following answer:

(…) that everyone who is capable of having a certain type of mental state must have a disposition to a certain sort of rational thinking involving mental states of that type (…) (Wedgwood 2007, p. 165)

He also defends the claim that this is consistent with saying that the disposition is not always manifesting. As he puts it:

(…) the claim that a thinker has a disposition does not imply this disposition will be manifested in every possible case. (…) It is to say that ceteris paribus, or in any normal case, when the object is in one of these stimulus conditions, it will also go into the corresponding response condition. But the case can fail to be normal, and cetera can fail to be paria, in various ways. (Wedgwood 2007, p. 164–165)

The last point is essential for Wedgwood’s normative dispositionalism because he thinks that any specification of the relevant dispositions (which are dispositions to being engaged in certain forms of rational thinking or reasoning) must involve the notion of defeasibility. More precisely, that appropriate forms of thinking or reasoning presuppose that certain defeating conditions are absent. It goes without saying that for Wedgwood the notion of defeating condition is normative.

This way of introducing normativity into intentionality, however, comes at a cost: if we agree that this is where the normativity of the intentional can be located, we must consequently say that normativity is involved in most dispositional properties. There can be no doubt that the counterfactuals that capture the essence of particular dispositions presuppose particular similarity standards for our closest possible worlds (or sets of premises, if we prefer premise semantics). For example, hemlock is disposed to harm x when drank by x iff it would harm x if it were the case that x drank it in normal conditions. Thus, for instance, it is not disposed to harm x if x would die before digesting it. Possible worlds where x dies before digesting hemlock count as distant and irrelevant when it comes to assessing whether hemlock is disposed to harming anybody. The absence of the relevant defeating conditions amounts precisely to this, and it seems that this sort of “normativity” is ubiquitous and trivial.

I believe that one can look for a weak or low-profile normativity of the intentional elsewhere. As Akeel Bilgrami noted with respect to the notion of the normativity of meaning:

The ≪oughts≫ I am admitting are hypothetical and extrinsic in the sense that if one has a certain end and a certain view of the means to achieve that end, then one ought to pursue those means. There is nothing special about the normativity which attaches itself to the meanings of words. Meanings are normative in the sense that a lot of other things might be: such as that I ought to play with a straight bat if I do not want to get caught in the slips, or that I ought to listen to songs by Hugo Wolf or The Cure if I want to get depressed. Similarly, there are certain things I ought to do if I want to be easily understood: I ought to speak as I have in the past or I ought to speak like others do (…) The point of these hypothetical, extrinsic norms is that by neither intentionally conforming to the others usage nor by even intentionally sticking to the past ways of one’s own usage, is a sign that meaning has an essential or primary aspect of correct and incorrect usage. The entire notion of correctness is entirely secondary to the desire and intention to communicate without causing strain, which underlies the notion of meaning. (Bilgrami 1992, p. 111)

Analogous things happen in the case of the supposed truthfulness and rationality of beliefs. One ought to believe truths if one wants to act efficiently,Footnote 14 and one ought to hold consistent beliefs if one wants to believe true propositions (since having inconsistent beliefs entails believing some false propositions). The normativity of beliefs (and other attitudes) consists in the fact that attitudes are sensitive to pragmatic factors and practical goals.

According to [2B] (we may appeal to [2B] as Explanatory Normativism) the normativity of the intentional consists in the fact that commonsense intentional explanations have normative presuppositions. One possible interpretation of this thesis is described in the manner presented in the following quotes:

[P]ropositional attitudes have their proper home in explanations of a special sort: explanations in which things are made intelligible by being revealed to be, or to approximate to being, as they rationally ought to be. This is to be contrasted with a style of explanation in which one makes things intelligible by representing their coming into being as a particular instance of how things generally tend to happen. (McDowell 1985, p. 389)

From the viewpoint of commonsense psychology (understood in the way which the autonomy theory understands it), in contrast, we view agents as rational beings. We attempt to make sense of their behavior, to understand what they will do and why they did what they did, on the assumption that they are rational agents. Without that assumption we will have no way of getting from what we know of their preferences and beliefs to their behavior. But, we can only make predictive and explanatory use of that assumption if we abstract away from the descriptive details of how they might actually be reasoning on a particular occasion and attend instead to the prescriptive or normative dimension of how they ought to reason. (Bermudez 2005, p. 43–44)

It has been argued (cf. Knowles 2002) that the main problem with this interpretation of the IN thesis is to state precisely what it means for an explanation to be of “a special sort”. Imagine, for instance, that someone believes that Castor and Pollux are siblings and that she believes that all male siblings have identical surnames. We would clearly expect and predict them to address both brothers with the same last name and to form a belief on which this action is based. A regular (non-normative) explanation might follow the following pattern:

A believes that Castor and Pollux are male siblings.

A believes that all male siblings bear identical surnames.

A believes that Pollux’s last name is “Dioscurus”

If A believes that Castor and Pollux are male siblings and A believes that all male siblings bear identical surnames, and A believes that Pollux’s last name is “Dioscurus”, then A believes that Castor’s last name is “Dioscurus”


A believes that Castor’s last name is “Dioscurus”

or, more generallyFootnote 15:

  1. I.

    A believes that aRb

  2. II.

    A believes that if xRy, then ln(x) = ln(y) (for each x and y)

  3. III.

    A believes that ln(a) = n

  4. IV.

    If A believes that xRy and A believes that if xRy, then ln(x) = ln(y), and A believes that ln(x) = z, then A believes that ln(y) = z (for each x, y, z).


  5. V.

    A believes that ln(b) = n

    Such an explanation seems to be legitimate and intuitive. Now, how can one introduce a normative element into it? As Knowles correctly observes replacing the generalization (IV) with:

  6. IV′.

    If A believes that xRy and A believes that if xRy, then ln(x) = ln(y), and A believes that ln(x) = z, then A ought to (or should rationally) believe that ln(y) = z (for each x, y, z).

    does not help: an obligation to do something cannot explain the fact that someone actually does something. If one intends to preserve the alleged normative character of the explanation, (IV) cannot also be replaced with:

  7. IV″

    If A believes that: if A believes that xRy and A believes that if xRy, then ln(x) = ln(y), and A believes that ln(x) = z, then A ought to (or should rationally) believe that ln(y) = z (for each x, y, z)

    because this premise is not normative and the explanation has nothing special in it.Footnote 16

A more promising interpretation of [2A] can be found in the work of authors like Daniel Dennett who defended the view that the rationality assumption constitutes a precondition for all intentional explanations.Footnote 17 In his reply to Stich he describes the issue as follows:

This is not to say that we are always rational, but that when we are not, the cases defy descriptions in ordinary terms of belief and desire (Dennett 1989b, p. 87)

In Dennett’s intentional system theory this means that in such cases we retreat to a description in whatever terms are offered by design or physical stances. There can be no doubt that Dennett is right that the strategy of saving (for example) the consistency of a description of someone’s belief system is common. For instance, in the case of regular mistakes (like that of a boy at a lemonade stand who gives incorrect change) we tend to describe the mistake in terms of acting on false beliefs rather than acting in a particular manner while having an inconsistent belief system. However, one cannot generalize Dennett’s observation to all cases. It is simply untrue that inconsistency resists description in regular propositional attitude terms. In fact, there can be no doubt that, on some occasions, explaining the action requires the assumption that the agent’s belief system is incoherent. Consider, as an example, the following quote from Nicholas of Cusa:

God is completely everywhere and not at all anywhere; He is of every form and of no form, alike; He is completely ineffable; in all things He is all things, in nothing He is nothing, and in Him all things and nothing are Himself; He is wholly and indivisibly present in any given thing (no matter how small) and, at the same time, is present in no thing at all.

The fact that Nicholas of Cusa said all these things can be explained by the fact that he said what he believed. Footnote 18 So, among other things, it can be explained by the fact that he believed that God is of no form and that God is of every form, that he believed that God is present in everything and in nothing etc.Footnote 19

Or consider a patient who is suffering from Cotard’s Syndrome (Nishio and Mori 2012), a rare disease that results in a genuine belief that the ill person is dead or even nonexistent. In 2012, doctors in Japan reported a man who came in for an appointment claiming “I guess I am dead. I’d like to ask for your opinion.” The doctor argued that dead men could not speak, but the patient was unable to get over his conviction that he was dead. “(…) he understood that his words defied logic, but he could not change his thinking.” (Nishio and Mori 2012, p. 218). In light of the explicit statement that he believes what he believes and that his words defy logic it is hard to see how one can describe a belief state of the patient as consistent. In particular the belief system of the patient contains the following elements:

I am dead.

Dead people cannot speak.

I am speaking.

My belief that I am dead is inconsistent with my belief that dead people cannot speak and that I am speaking.

And, in fact, inconsistent parts of the patient’s belief system can be used to explain some of the patient’s utterances (and non-verbal actions as well). It is important to note that in this case the patient’s belief system is inconsistent in the strongest sense: he does not only believe p and q for some p and q that are mutually inconsistent but he also knows that p and q are mutually inconsistent. This case can be contrasted, for instance, with the case of weak inconsistency such as a belief the someone is in two distant places at a given time (see: Bayne and Pachery 2005 where such a clinical case is described). If you find Nicholas of Cusa or clinical examples unconvincing, you might consider cases like the lottery paradox and ask yourself this: does the fact that an agent holds inconsistent beliefs precludes predicting that they will assent to “Some ticket will win”? I am not denying here that authors like Dennett are basically correct when they say that we have a natural propensity towards re-interpreting inconsistencies in someone’s belief system as cases of apparent irrationality (for instance, one might claim that Nicholas of Cusa did not mean what he said literally). However, our initial thesis was not that there is such a tendency among homo sapiens, but rather that such cases “defy descriptions in ordinary terms of belief and desire”. The latter claim is patently false.

Proponents of Dennett’s interpretation of the normativity thesis might respond here by pointing out that examples quoted above make use of a single norm of consistency and, due to that fact, show only that this particular norm is not essential for intentional explanations, i.e. that there are other rationality norms that are constitutive presuppositions of all intentional explanations. This, of course, might be correct. However, one ought to be more precise about the nature of those norms. Saying, for instance, that “(…) norms of good reasoning cannot be fully codified in a formal theory” (Bermudez 2001, p. 466) sounds, at best, as a vague declaration or, at worst, like an expression of the unconditional philosophical surrender.

There are, then, two possible interpretations of [2B]. Unfortunately, both face serious difficulties: the first refers to a rather mysterious notion of explanation, the other presupposes falsely that intentional explanations always fail in cases of inconsistency.

At best, therefore, of all the interpretations of the IN thesis we have distinguished only imposition theories seem to be, prima facie, free of serious drawbacks. The other two interpretations are either unclear ([2B]) or presuppose an extremely weak and ubiquitous understanding of normativity ([2A]). Below I will argue that a closer look at imposition theories reveals their problematic nature as well.

3 Belief Attributions, Normativism and Relativism

Here is the sketch of the argument presented in the pages to follow:

[1] Imposition theories might be interpreted either in a perspectivist or constitutivist manner.

[2] Perspectivism entails either conventionalism or relativism about obligations.

[3] Constitutivism entails deflationism regarding normative terms.

[4] All three options (conventionalism, relativism and deflationism) exclude robust normative versions of imposition theories.

Therefore: Imposition theories do not support the IN thesis, if the claim that there is normativity in intentionality is assumed to have a substantive (non-trivial) meaning.

Imposition theories attribute themselves to agents of propositional attitudes with complex conditional properties in the form:

If (…) V-s that p, then (…) ought to A

where V represents a particular psychological mode and A, in case of Action Normativism, represents dispositions to act and, in the case of State Normativism, other intentional states of the agent. Thus, for instance, a possible Action Normativism attribution might make use of the following property:

(*) If (…) knows that p, then (…) ought to be able to justify that p.

while a possible State Normativism attribution has the following:

(**) If (…) believes that p, then (…) ought to believe that p or q.

Both types of attributions are possible in virtue of the supposition that appropriate conditionals are true for all values of (…), i.e. that they follow on from appropriate true conditional generalizations. I believe that at this point one can identify a certain problematic aspect of the imposition theory. The aspect I have in mind is connected with two possible approaches to belief and propositional attitude attributions.Footnote 21

According to the first approach (let us call it “perspectivism”) belief attributions are perspective dependent. This account has been defended by several authors, of which Stephen Stich (e.g. Stich 1983) and Daniel Dennett (e.g. Dennett 1989b) might be used as examples. For Stich, for instance, belief attributions presuppose complex context-dependent similarity judgments. Roughly speaking, in the case of beliefs this might be the result of replacing the content identity theories of belief attributionsFootnote 22 with a theory that assumes that utterances of “S believes that p” mean each time they are used that S is in a belief state that is content-similar to the one that plays a typical causal role in the sincere assertion of ‘p’ by the belief’s attributor. The notion of content-similarity is then unpacked as involving several dimensions called by Stich causal-pattern, ideological and referential. Dennett’s approach is motivated mainly by the indeterminacy of functional attributions (for him, the indeterminacy of content is just a special case of this general problem):

We cannot begin to make sense of functional attributions until we abandon the idea that there has to be one, determinate, right answer to the question: What is it for? And if there is no deeper fact that could settle that question, there can be no deeper fact to settle its twin: What does it mean? (Dennett 1989a, p. 319).

Stich and Dennett’s accounts differ in important details but they share a common feature: they deny that there are attributor independent intentional facts. An alternative approach, inspired by the work of authors such as Carnap (1955), Sellars (1956) and Lewis (1972), presents propositional attitude terms and concepts as theoretical. It presupposes that the meaning of propositional attitude terms is given by the theory or theories in which they figure. The meaning of propositional attitude attributions is, then, not perspective dependent but theory dependent. I consider the two options as exhausting the logical space of adequate theories of belief attributions.

Now, assume for the sake of argument that the former approach is correct. Then, it seems, no matter what particular generalizations entailing particular attributions like (*) or (**) you take for granted, it is hard to see how one can infer something about particular obligations of a cognitive agent. Consider, for example, the case of John who is unable to distinguish Castor and Pollux. Assume that he saw Castor yesterday and he is disposed to utter affirmatively “I saw Castor yesterday”. Now, if you believe that (**) is true of John, you shall also believe:

If John believes that he himself saw Castor yesterday, then he ought to believe that he himself saw Castor yesterday or that today is Monday.

However, due to indeterminacy of content, we will also have:

If John believes that he himself saw Pollux yesterday, then he ought to believe that he himself saw Pollux yesterday or that today is Monday.

As well as:

If John believes that he himself saw Pollux-or-Castor yesterday, then he ought to believe that he himself saw Pollux-or-Castor yesterday or that today is Monday.

If you follow Dennett’s suggestions, you will be disposed to say that there is absolutely nothing that could decide about the obligations of John. Note that the problem here is not that John’s state is not determined by empirical factors (such as his acts and dispositions to act, his causal history etc.) but rather that, according to the theory we are considering, there is nothing to be determined. If you follow Stich’s suggestions the situation becomes even more awkward. If context-dependency and similarity standards are involved in choosing the correct antecedents of the three conditionals, then what John is obliged to do depends on the similarity standard relevant in a given context. This entails relativism with respect to John’s obligations. Now, relativism and robust normativism are in the sharp conflict. Robust normativism claims, inter alia, that normative properties (like correctness, rationality, goodness, fairness, moral appropriateness) are inhered absolutely. In other words (using the language of attributions and predicates) robust normativism is the claim that normative predicates, if truly applied to something, are absolutely true of that thing. Hence there seems to be no way to reconcile the normativity presumption and the perspective dependency.

Assume, therefore, that the second approach is correct, i.e. that the meaning of propositional attitude predicates and concepts is given by the theory in which they figure. The theory in question being probably a set of platitudes collected in the manner suggested by Lewis. Assuming this, how can one interpret generalizations on which attributions like (*) or (**) are based? The reply is that such generalizations must somehow follow on from the platitudes of our folk psychological theory. In other words, that they are logical consequences of what is constitutive for the meaning of folk psychological predicates or concepts. As such, they are also constitutive for the meaning of appropriate folk psychological terms.

There are, of course, worries regarding the whole notion of constitutive normativity.Footnote 23 However, I do not think that we need to address the whole complicated issue of constitutive normativity in order to see why we should not take for granted that a folk psychological theory will give as a robust normativity of actions or intentional states. Consider a simple analogy. One might say that the meaning of a given propositional logic connective is given by a set of valid inference rules that make use of that connective (the set of such rules is a counterpart of our folk theory that gives meaning to propositional attitude predicates). Thus, for instance, the conjunction elimination rule might be said to be constitutive for the meaning of “∧” in propositional logic. It should be clear that one of the many ways of stating this very simple fact is to say that if one accepts “AB”, then one ought to accept A. The “ought” in this formulation has clearly a redundant or deflationary sense: we learn nothing special about the meaning of “∧” when we compare it with the non-normative ways of stating it (like: “if you have “AB” in the proof, A follows”)—it plays no informative role at all. The situation with the folk psychological theory and propositional attitude predicates is analogous if we assume that the former gives meaning to the latter. Assume now that the following conditional is an instantiation of some folk psychological platitude:

If John believes that he himself saw Castor yesterday and that he did not meet Pollux on the very same day, then he ought to believe that he himself saw Castor yesterday.

What additional meaning does it have when it is compared to:

If John believes that he himself saw Castor yesterday and that he did not meet Pollux on the very same day, then he believes that he himself saw Castor yesterday.

I think that the answer is: none. Assuming that the first conditional follows from a folk psychological platitude that endows folk psychological predicates with meaning, there is a very clear reading of the second conditional that ascribes to “ought” a purely decorative or pragmatic role. By the same token, the appropriate normative predicates are redundant and the conditionals have no special normative sense.


I do not consider this argument and other analyses presented above to be decisive. It might happen that there is a third way between perspectivism and constitutivism or that the functional-theoretical interpretation of folk psychological concepts contains a genuine normative element. It might also happen that other proposals I have considered in the paper have interpretations that were unintentionally ignored. If this is true, then the paper contains only a list of challenges that should be addressed by the proponents of the IN thesis . I can only hope that considering some of them will enable us to better understand the nature of attitudes and intentionality.