What is it for a thinker to be entitled to form a given belief? The answer to this question is central in the project of understanding rationality and knowledge. It is equally important in understanding the relations between rationality and meaning or content. A good answer to the question will bear on many matters, including the issues of rationality that give various forms of skepticism their bite.
I begin to address the question of the nature of entitlement by distinguishing three levels at which we can characterize the relation of entitlement. There is an increase in generality and explanatory power as one proceeds through these three levels of characterization. My plan is then to propose hypotheses at two of these levels. The hypotheses I will be developing apply theses in the philosophy of mind to issues in the theory of entitlement, and thereby to issues in epistemology. If correct, the hypotheses suggest a general treatment of one species of entitlement. They also bear upon skepticism about perceptual knowledge.
I will be concerned primarily with a thinker’s entitlement to rely on certain contents of her own perceptual experience. The hypotheses I develop do have some application beyond the perceptual case. The main reason, however, for concentrating on the perceptual case (which is certainly only one kind of case) is its fundamental and central role. If we cannot give a good account of entitlement in these cases, we will not have an adequate account of empirical knowledge in general.
1. Entitlement: The Three Levels
The three levels at which we may characterize the entitlement relation are levels which we can distinguish for any property or relation.
Level (1). The first of the three levels I distinguish is simply the level of instances or examples of the entitlement relation. So characterizations at this level comprises true statements of the form ‘a thinker in such-and-such circumstances with so-and-so background information is entitled to judge that p’. These examples involve a specification of types of circumstances in which an entitlement exists. The types may concern the thinker’s environment, his other conscious states characterized in terms of content and his general capacities. If a thinker is entitled to make the judgement, of a seen object, that it is curved, when he visually experiences it as curved, and when there is no reason for doubting his senses, then that would be a statement included at this first level of instances. So would statements about the entitlement to rely on apparent personal memory in making judgements about one’s own past.
These entitlements are defeasible, in the sense that someone’s entitlement to rely on perceptual experience or apparent memories may be defeated by further information that does give reason for doubting one’s senses or apparent memories. Here, as in other areas, we need to distinguish two kinds of defeasibility. Something one thought to be a mathematical proof may turn out not to be so; or reasonable doubt may be cast on it by the most distinguished experts. That is, no doubt, some kind of defeasibility, but it is quite different from the perceptual and memory cases. If something really is a proof, no additional information can establish that its conclusion is not true. If something really is not a proof, it is not a proof all along, whatever we think about it. We cannot say that something is not really an experience as of something’s being the case, if we come to discover that it is not a genuine perception. Experience itself is never conclusive, in the way that a proof is. We distinguish defeating conditions that show that something isn’t embedded in world in right way to be genuine perception from those which show something isn’t of the right kind to be conclusive.
There is a factive notion of being a perception that p which behaves more like the notion of being a proof that p. The question of the relation between the factive notion and the non-factive property of being an experience which represents it as being the case that p is important, and I will return to it after developing some positive theses. At present, I am taking it that the non-factive state is entitling in the absence of genuine reasons for doubt.
Level (2). The second level is the level of generalizations about the entitlement relation. This level consists of true generalizations which, in the presence of additional information determined by the generalizations, have statements at level (1) as instances. It may be helpful to think of the relation of this level to others as the same as that exhibited by grammatical generalizations to other levels in the theory of grammars. In his 1965 theory, Chomsky wrote that a grammar “is descriptively adequate to the extent that it correctly describes the intrinsic competence of the idealized native speaker”. In the same spirit as Chomsky’s use of the term, we could call this second level ‘the level of descriptive generalization’. A descriptively adequate grammar for a language will have as consequences instances or examples of the property of grammaticality for the language.
We equally operate at this second level of characterization when, in developing a logic for some particular expression in natural language, we move from particular valid transitions containing the expression but not containing schematic letters to the stage of formulating general schemata that are valid. That is a move to a level of descriptive adequacy. As in the other cases, it can be an important step towards theoretical understandings. It is, for example, illuminating, and a step towards an explanatory theory, to note that though transitivity does not hold for the counterfactual conditional, the schema ‘If A were the case, then B would be the case; if A & B were the case, then C would be the case; so if A were the case, C would be the case’ is generally valid.
The generalizations about the relation of entitlement at this second level may be more or less extensive, and correspondingly more or less illuminating. The generalizations may use theoretical notions in classifying circumstances, contents, and capacities. It is information about the extension of these classifying notions that one will need if one is to use the generalizations to derive truths at level (1) about instances of the entitlement relation.
Level (3). The third level is the level of explanation. This third level consists of philosophical explanations of the true generalizations at level (2). If certain theoretical notions seem to capture the correct generalizations at that second level, then one of the philosophical tasks at this third level is also to explain why they do so. In linguistic theory, Chomsky distinguished a level of explanatory adequacy, a level of theory at which one aims to explain why the child selects a particular descriptively adequate grammar. Explanations at that level would be empirical explanations of acquisition in the linguistic case. In the case of the development of a logic, this third level would be the development of a semantical theory that explains the patterns of validity and invalidity that would be captured in schemata at the second level.
In the case of the explanatory level for the entitlement relation, we are concerned rather with philosophical explanations. In contrast at least with the linguistic case, the task of characterizing this third level for the relation of entitlement is that of explaining timeless generalizations rather than historical (extended) events of acquisition of a grammar. In the particular case of explaining the entitlements provided by perception and memory, the task is to explain the true generalizations about defeasible entitlement, and to say why those generalizations (and nothing weaker or stronger) captures the extension of the entitlement relation.
How would we describe these three levels in more detail for the case of perceptual entitlement? At the level of examples, I wrote, like James Pryor, of an entitlement to take the perceptual content of experience at face value, in the absence of reasons for doubting it. For a wide range of perceptual contents, there is such an entitlement. It exists for many spatial contents, temporal contents, contents relating to surface texture, colour and illumination, and a range of material properties.
There are, however, also cases where it is much less plausible that perceptual experience alone can supply the entitlement. Consider furniture that looks Swedish; appliances that look like Mac computers; or the properties of looking sad or looking delighted. Are we entitled, in the absence of reasons for doubt, to judge on the basis of such experiences that some furniture we see is Swedish, that some presented object is a Mac computer, or that someone is sad or delighted? In some of these cases, one can follow a strategy of divide-and-rule. One can explain the apparently perceptual phenomenon thus. There is some kind such that the thing or person appears to be of that kind, and the person judges that things of that kind are (say) Swedish people, or Mac computers. The perceptual entitlement holds only for the kind which is given in the content of perception, as opposed to the content of the judgement. But such a division is not plausibly available in all cases. It does not, for example, fit the case of perception of the expression of an emotion. To describe, when seeing the face of a person, the experience in which they look sad in non-emotional terms is not to capture its distinctive representational content. There is no kind, described without reference to the emotions, of which one can say that the facial expression appears to be of that kind, and it is merely an additional judgment on the part of the person that people looking that way are sad.
It is tempting to say that the purely perceptual entitlement holds only for observational concepts. That may well be true; but it is hardly an illuminating generalization at level (2), unless we have some independent account of observationality. We are in danger of moving in a circle, for it is only too plausible to say that observational concepts are those that can be applied with entitlement simply on the basis of perceptual experience, and without further information. Unless we have some characterization of observational concepts distinct from that, then to say that the perceptual entitlement holds only for observational concepts will become the vacuous claim that it holds for those contents for which it holds.
I will suggest later on that a by-product of a proper characterization of the second level will be a starting point for characterizing the relevant notion of observationality without circularity. Suppose we can formulate a sufficiently wide-reaching true generalized conditional about the conditions under which perceptual entitlement holds. It will have the form ‘If such-and-such conditions hold for the content p and for the thinker’s circumstances, then the thinker is entitled to take the content p of his perceptual experience at face value’. One way to characterize the non-observational will be as contents not meeting the antecedent of that conditional.
The task of formulating such a generalization about perceptual entitlement lies at the second of the three levels. So I aim to characterize, without using the notion of an observational concept, a relation which holds between a perceptual experience and a particular content p which it represents as being the case, a relation with the following property: the holding of that relation is sufficient for a subject who enjoys the experience to have a perceptual entitlement to judge that p, in the absence of reasons for doubt. Various other perceptual entitlements, I will later argue, have their status as such in virtue of the relations in which they stand to this sufficient condition.
In some cases, and to a first approximation, what is constitutive of an experience’s having a certain representational content is that when the thinker’s perceptual apparatus is functioning properly, in a normal environment, experiences with that content are caused by the holding of the condition which is in fact the correctness condition for that content. This is plausible for the spatial representational contents of perception: the representational contents concerning such matters as distance, direction, shape and size. In Being Known, I argued that the same is true for the temporal contents of perception. When all is working properly, in a normal environment, and in the most fundamental cases, a subject’s perception of temporal order and magnitude is caused by instances of those order-relations and temporal magnitudes. (The perception of temporal magnitudes, like the perception of spatial magnitudes, is commonly unit-free.)
Suppose we agree that it is constitutive of a particular kind of experience’s having a spatial or temporal content that such experiences have certain causes in specified conditions. It does not follow that it is constitutive of that content that it feature in experiences of that kind in any perceiver capable of having states with that content. Experience of different kinds, in more than one sense modality, may have the same spatial representational content. A given subject may be capable of having experiences in only one of those modalities. Furthermore, the given content may also feature in proprioception. The feeling of moving one’s arm in a straight line may involve the same content straight line as also features in visual or tactile experience. The spatial content straight line can also feature in the non-proprioceptive awareness of acting that can be present even when one’s limbs are anesthetized. Indeed, the very fact that perceptual experiences with these contents are individuated in part by facts about their causes in certain circumstances opens up the possibility of the occurrence of such contents in other conscious states, both in perception and in action. The cause that is involved in the individuation may cause other experiences too. Equally, the state of affairs that is the cause may be mentioned in constitutive accounts in which it features as an effect, as in the non-proprioceptive awareness of action. In short, we must be careful not to overstate the constitutive principle which links the individuation of some perceptual contents with the holding of those contents in the perceiver’s environment.
What is true, however, is that the spatial and temporal contents of experience are in a certain sense constitutively basic with respect to these experiences. That is, these experiences do not have these elements of their content in virtue of the experiences’ having certain other relations to other states with the same contents. One can contrast this with contents of perceptual experience which seem to use such concepts as soldier or judge, as when one says that it looks as if there is a soldier guarding the building, or a judge speaking from the bench in a courtroom. Such experiences, if that is their literal content, have those contents by virtue of their having contents which also feature in the ability to come to believe that someone is a soldier or a judge; and these capacities in turn have to do with some knowledge, perhaps rudimentary, of what it is to be a soldier or a judge. Unlike the case of contents concerning the properties of being a soldier, or of being a judge, perceptual experience can provide a thinker’s fundamental fix on spatial and temporal properties and relations.
The experiences of which I am writing are sometimes called externally or anti-individualistically individuated. For present purposes, this not an ideal label. For the essential characteristics of the phenomenon are present in, for instance, proprioception of limb position, and the disposition of one’s own body in space. One has an awareness that represents one’s limbs and body as being a certain way spatially. It is highly plausible that what gives this awareness the content it has is that, when all is functioning properly, an awareness as of one’s arm being straight is caused by one’s arm being straight. This is not a relation between conditions external to the perceiver’s body and the perceiver’s own mental states. So I prefer to speak of perceptions which are instance-individuated with respect to certain of their contents. What makes these perceptions have the content they do is the fact that when the subject is properly related to the world, the holding of these contents causally explains the subject’s experience as of their holding.
Even when the subject is properly connected to the world, and the environment is normal, still some spatial, temporal and bodily contents of these sorts do misrepresent. There are some reliable illusions – such as the Müller-Lyer – which occur even in ordinary circumstances when the embedding of subject is as proper as it is ever going to be. In these cases, the experiences have the contents they do because of their relations to those experiences that are directly instance-individuated with respect to perception. These cases of illusion can be described as derivatively instance-individuated with respect to perception.
We c an then formulate this generalization at the second level about the relation of entitlement:
A perceptual experience which represents a content as correct, and which is instance-individuated with respect to that content, is also one which entitles a thinker to judge that content, in the absence of reasons for doubting that he is perceiving properly.
This needs some adaptation if perception has nonconceptual content. Suppose it does. Then for conceptual contents for which there is a perceptual entitlement to judgment, there will be a range of nonconceptual contents of experience which generate an entitlement to judge such a conceptual content. Call these ‘the range of nonconceptual contents which canonically correspond to the conceptual content’. The generalization at the second level would then be formulated by saying that:
A perceptual experience which represents a nonconceptual content as correct, and which is instance-individuated with respect to that nonconceptual content, is one which entitles a thinker to judge a conceptual content as correct, in the absence of reasons for doubt, when the nonconceptual content is in the range which canonically corresponds to the conceptual content.
Under either variant, the generalization at the second level is equivalent to something simple and intuitive. The generalization is in effect saying that when making perceptual judgments, one is entitled to take it, in default of evidence to the contrary, that one is in the circumstances with respect to which one’s perceptions are instance-individuated with respect to the contents in question. So I will call this thesis about the second level ‘the Individuation Thesis about Perceptual Entitlement’.
This Individuation Thesis suggests an approach to the issue of what makes something an observational concept. The intuitive idea is that a non-observational concept will not be instance-individuated, because it has commitments which go beyond what is involved in instance-individuation. An experience of something as a Mac computer, or as a PET-scanner, cannot be purely instance-individuated, because that would not capture the commitments of these concepts, the commitment that objects falling under them are capable of carrying out certain functions.
Such a development of a criterion for observationality has to be carried out with some care, because instance-individuation is not to be taken as meaning that nothing more than causal interaction is involved in an experience’s having a certain content. There is causal interaction in ordinary visual experience with patterns of light reaching the eye; with retinal stimulation patterns; and with the state of the optic nerve. None of these matters enter the representational content of ordinary visual experience. The representational contents of visual experience also serve as input to the subject’s construction of a conception of the layout of the objective spatial world around him. This is a feature of perceptual experiences with instance-individuated contents. So the criterion for a concept to be non-observational might be better formulated thus. Non-observational concepts have commitments going beyond the minimal conditions for objective content which are met by the contents of experiences which are instance-individuated. I simply note the possibility of this approach to observationality, as one by-product of the Individuation Thesis about Perceptual Entitlement. It would take us too far off our main path to pursue here the further elaboration which would be necessary to develop the criterion in detail, and with a rationale.
Our actual entitlement to perceptual judgements rests on far more than the Individuation Thesis alone, applied atomistically experience-by-experience. Consider a stream of experiences, each of which entitles a thinker to believe some corresponding content. If these objective contents cohere, each being a spatial content of a perception reasonably expected to follow its predecessors, then the resulting entitlement to judge each content is massively greater than if each experience had occurred in isolation. Holism of confirmation is as pervasive in the sphere of perceptual judgements as it is in other areas. Even the proposition that objects have rears is something that requires perception from more than one angle, and cannot be confirmed by a single view.
All the same, it seems to me that this holism of confirmation serves to increase a prior level of entitlement that can already exist in the individual case before additional perceptions or evidence are brought in. The additional perceptions or evidence are important because they can give further reasons for thinking things are as an initial perception represents them as being. (They can also serve to show that certain kind of defeating conditions do not hold.) In a sequence of coherent experiences, the later experiences themselves give defeasible reasons for making certain judgements, independently of the occurrence of the earlier experiences. It is only because this is so that the later experiences can then give further confirmation of the judgements supported by the earlier experiences.
There is an abstract, structural argument that if rational, entitled thought is to be possible at all, some concepts must be such that one is default-entitled to presume that one is in the circumstances in which they are individuated. Maybe there could be a concept whose possession-condition makes reference to applications in circumstances one is not default-entitled to presume are one’s own. Perhaps there could be a concept which, as a matter of its nature, is to be applied to objects which look a certain way, but only under a certain kind of abnormal illumination. Any entitled application of this concept on the basis of experience will require inference, or some other entitled transition, to the conclusion that the illumination is of the special abnormal kind. Now could it always be that inference, or some other entitled transition, has to be made before we are entitled to apply a concept? It seems that this could not be so, if entitled application is ever to get started.
This abstract, structural argument seems to me to be sound. But absract arguments by themselves have only abstract conclusions. It is one thing to know that default entitlements must exist. It is another to explain how they are possible in the first place, and to explain why they have the particular character and contents they do. The abstract argument does not give us an understanding, of any particular generalization about the entitlement relation, of why it, rather than some other principle, holds. To move towards such understanding in the perceptual case is the purpose of my next question, which is at the third of the three levels I distinguished. The question is: if the Individuation Thesis about Perceptual Entitlement is a true generalization about entitlement, what explains its truth?
There would be no further task of answering this explanatory question if the Individuation Thesis about Perceptual Entitlement were derivable simply from the truths about the individuation of perceptual content together with principles about the nature of entitlement in general. But I cannot see what such a derivation from those premises alone would be like. The generalization which is the Individuation Thesis about Perceptual Entitlement does not itself explain why one is, in the default case, entitled to accept what would hold in the circumstances with respect to which perceptual content is individuated. Those circumstances may be special from the standpoint of the theory of the individuation of content; but what is so special about them for epistemology and the theory of entitlement? Why is one entitled, in the default case, to form perceptual beliefs as if one were in the circumstances with respect to which the content of the perception is individuated? To achieve philosophical understanding of these issues, we have to undertake the further substantive philosophical task of explaining the epistemic significance of facts about the individuation of perceptual content. It is part of the task of connecting the epistemology of the theory of content with its metaphysics.
2. An Explanatory Hypothesis
I propose the following hypothesis at the third level of characterization of the entitlement relation, the explanatory level. Suppose a state or kind S of event is individuated by the relations in which its instances stand to other events and objects. Suppose also that we can draw a distinction between genuine instances of the state S, instances which stand in the required relations, and ‘merely as if’ states or events which are in a quite specific sense qualitatively similar to and parasitic upon to those genuine instances. The ‘merely as if’ states do not stand in those relations as things actually are. They are also parasitic in the sense that they are given as states which, although they do not in fact stand in the required relations, it is as if they do. For illustration, we can use a helpful example of Crispin Wright’s, to which we will return. Suppose, remarkably, a group of people ran around kicking a ball for 90 minutes, without any idea or intention of playing soccer, but engaging in the same bodily movements that would be involved in a game of soccer. We can say that their movements are as if they are playing soccer, even though those movements do not have the right relations to their own and to others’ mental states for it to be a game of soccer. Similarly, a perceptual hallucination does not stand in the right relations to things in the environment to be a genuine perception; but it is, subjectively, as if it were so related. On the way I will use the terms here, we will say that genuine soccer games and genuine perceptions are as-if states, though of course they are not merely as-if states.
The general hypothesis at the third level that I propose then states this: in a significant range of cases, given just the information that an as-if state qualitatively similar to an instance of S occurs, the easiest way for this to be the case is for it to be a genuine instance of S, and not a mere as-if state.
I shall also be arguing for a further claim: the fact that it is easy for something to be the case if it has a certain explanation is itself to be elucidated in terms of the explanation’s explaining condition involving a reduction in complexity from that of the explained state of affairs. So I label this general hypothesis ‘the Complexity-Reduction Principle’. I will be arguing that the Complexity-Reduction Principle, in combination with the other principles I have endorsed, implies that the easiest way for an as-if state to occur is for it not to be a mere as-if state, but for it to be veridical, to stand in the required external relations.
The Complexity-Reduction Principle may seem very strong, so I will start with some illustrations to suggest how it operates. Let us take Wright's soccer example first. (We will in fact end up with a very different position on the issues from Wright’s, and I will eventually be drawing different conclusions from his own examples.)
For the movements of a set of 22 people to replicate those of a soccer match without their having any idea of soccer is not metaphysically impossible. It would, however, involve a massive, extraordinary series of coincidences. The hypothesis of accidental replication is hugely more complex than the simple hypothesis that they are playing soccer, and that their bodily movements are controlled by the intentions that are made understandable by their meaning to play the soccer.
It would not be a coincidence that the agents’ movements matched those of a game in the other case Wright mentions, that in which they are under the control of a movie director who wants his movie to represent a game taking place. But the hypothesis, given only that people are moving as if playing soccer, that they are under the control of a movie director, seems to me more complex, to demand more of the world, than that they are simply playing soccer. It demands not just that the agents have the notion of soccer, but that they all be influenced by some further individual.
The Complexity-Reduction Principle seems, then, to receive some confirmation from the soccer examples. In those examples, given only that the movements of the agents are as if they are playing soccer, the easiest way for this to be true, the explanatory hypothesis which best contributes to the explanation of the complexity of an event which involves the movements of a soccer match, is that they are really playing soccer.
Now consider an apparently perceptual experience with the representational content that p, and suppose this experience, in respect of this content, is instance-individuated. What is the easiest way for it to come about that such an experience occurs? It is certainly metaphysically possible that the experience occurs, and that its occurrence is not explained by its being the case that p, or indeed by anything in the environment of the subject to whom it occurs. But the occurrence of a perceptual experience with the representational content that p is a highly complex thing. To have the content that p, the experience must be of a kind which, when the subject is properly connected to the world, has its instances caused by the fact that p (or else it is derivatively instance-individuated).
For an event to have that property is already somewhat complex. But in fact the complexity of the property of having an experience with a certain representational content goes further. An experience with spatial representational content, for instance, must be one whose content is capable of contributing to its subject’s conception of the spatial layout of the world around him. Without this, the experience would not amount to having a spatial content at all. The content must also be capable of integrating with other spatial representations in confirming or disconfirming the subject’s conception of the layout of the world. These holistic elements in the possession of spatial perceptual content all contribute further to the complexity of the property of having an experience with a given spatial representational content that p. I will be arguing that, in the absence of any information to the contrary, it is not a good explanation, one reducing complexity, to suppose that an experience with this complex property occurs without being explained by the condition that p.
The occurrence of an experience with the representational content that p would hardly be a coincidence if its representational content held of the world, and the subject had a properly functioning perceptual system whose holistic complexities were adapted to its spatial embedding in the world. For a subject with such a perceptual system, its being the case that p would then explain the occurrence of an experience with whatever complex relational property is involved in representing it as being the case that p. But how easily does it come about that the subject has a properly functioning perceptual system?
It is a relatively a priori truth that since subjects rely substantially on their perceptual systems in the formation of belief, there will be selection for roughly accurate perceptual systems. As always in evolution by natural selection, there are trade-offs, and some, perhaps considerable, inaccuracy may be traded for speed or range of representations. But the explanation by natural selection of the existence of roughly accurate perceptual systems does reduce complexity. The explanation succeeds by citing states of affairs of lesser complexity than that which is to be explained.
By contrast, the hypothesis that an experience is produced by an intentional manipulator is much more complex. There is no reason to think it easy for such manipulators to exist; and the attribution to a manipulator of any particular purposes is not a matter of choosing the easiest way the experience can come about. The hypothesis that there is a manipulator would explain the occurrence of the experience, but would do so by citing mental states of at least as great complexity as the perceptual state to be explained.
So I suggest that the supposition that an apparently perceptual experience is veridical is a consequence of an explanation that is better, because it reduces complexity. The explanation is better than the hypothesis that the experience is not a perception of the world at all, and better than the hypothesis that it is produced by some intentional manipulator. Both of these latter hypotheses involve unexplained complexity.
It is plausible that if one has the information that a certain state is instantiated, and, as an a priori matter, the easiest way it can be so is for p to hold, then it is rational to judge that p. If that is so, then the Complexity-Reduction Principle, together with the claims I have made about the individuation of content, jointly imply the preceding section’s Individuation Thesis about Perceptual Entitlement. Thinkers are entitled to take the contents of perceptual states which are instance-individuated at face value, because the easiest way for such experiences to occur is for them to be genuinely perceptual, that is for their representational content to be true. Perceptual knowledge remains, of course, not a matter of inference from the occurrence of a perceptual experience. The instances of the entitlement relation at level (1) remain entitlements to take perceptual experience at face value, in the absence of reasons for doubt, rather than inference from the occurrence of an experience. The suggestion is just that the Complexity-Reduction Principle contributes to an explanation of the Individuation Thesis about Perceptual Entitlement. That Individuation Thesis holds because of truths about what are and what are not easy ways in which complexity may arise.
Not just any property coextensive with the cases in which the entitlement relation holds can amount to an explanation of those instances. Since judgement aims at truth, an explanation of the instances must cite a property that gives good reason for thinking that the content to which one is entitled is true. The easiness of the way in which it comes about that one has the experience is a property which is connected, defeasibly, with the truth of the representational content of the entitling experience.
The Complexity-Reduction Principle seems to me to need restriction to the case in which the experience in question is instance-individuated. This corresponds to, and if correct helps to explain, the same restriction in the Individuation Thesis about Perceptual Entitlement. When an experience has a content that is not instance-individuated, such as a content involving the concepts soldier, or judge, or photocopying machine, what makes it have that content is not a matter of the relations of experiences of that type to instances of those concepts. We have an explanation of how the experience comes to stand in the relations necessary to have these non-observational concepts if we can explain how the experience comes to have a content which, when taken at face value, leads the thinker to judge that the presented items fall under the relevant non-observational concept. This need not involve any relation, however indirect, to encounters with instances of the concepts.
This point highlights a difference in the direction of individuation of the perceptual content as between instance-individuated cases and other cases. For an instance-individuated content, the experience has such a content in part because of its complex relations to encounters in which that content holds, and which causally produce experiences of that kind. For other contents, the direction is reversed: the experience has such a content because of its connection with concepts the thinker has, and whose possession does not constitutively involve such encounters. That still involves a kind of complexity, certainly, but the easiest way for it to arise need not at all involve the supposition that such experiences are veridical. A good explanation has only to say how the antecedently-individuated concept comes to enter the representational content of the perceptual experience, and this need not involve anything which implies the correctness of that content. In the instance-individuated case, the content is not antecedently-individuated, and (if my argument is sound) the best explanation does generally involve correctness of the content.
The Complexity-Reduction Principle as applied to instance-individuated experiences is an instance of a form of reasoning used repeatedly in the theory of evolution by natural selection. Functional complexity needs an explanation. Such complexity is found in an organism that is able to fly, and more generally in any organism that is able to survive, gaining energy by substances obtained from its environment. Such complexity will have come about by a step-by-step process, the ‘easiest’ way. Both principles are summarized in this passage from Richard Dawkins:
“A complicated thing is one whose existence we do not feel inclined to take for granted, because it is too ‘improbable’. It could not have come into existence in a single act of chance. We shall explain its coming into existence as a consequence of gradual, cumulative, step-by-step transformations from simpler things...”
The ‘could not’ in this passage does not contradict what we earlier said about metaphysical possibility. Dawkins’ ‘could not’ is that of empirical implausibility.
There is not just a parallel with this style of reasoning in the theory of evolution. Some of the considerations of the present argument are actually an instance of this reasoning in the theory of natural selection. If we are to explain how organisms come to enjoy perceptual states with their complex panoply of environmental relations, a process of step-by-step evolution by natural selection provides the only account which does not rely on coincidences, miracles, or the implausibly complicated.
This does not mean that the concept of entitlement contains the concept of natural selection. We have been concerned with the philosophical explanation of why the entitlement relation holds in the cases in which it does. The concept of the easiest way something can come about enters that explanation. It is then a substantive claim that the easiest way is the path provided by evolution by natural selection. So this is a position that makes evolutionary considerations relevant to epistemic status, without changing the subject to something non-normative.
The principle that good explanations of complexity cite states with reduced complexity is not, however, confined to cases of natural selection and its applications. A different example concerns the subpersonal psychological explanation of intelligent action by states that do not involve the same level of intelligence, or the same psychological capacities that are to be explained. The principle that a good explanation of complexity must not, unless it is required by other truths, cite explaining states that display the same level of complexity as what was to be explained seems to me to apply to arbitrary empirical subject-matters.
I note the following features of this approach.
(1) The approach does not say that it is a priori that hallucinations are rare, nor that an evil-demon world is impossible. Hallucinations may be frequent, and there are genuinely possible worlds in which there is a deceiving evil demon. The present position is only that such worlds are more complex than those in which there is a predominance of genuine perception as a result of the perceptual system’s coming into existence in an easy way.
(2) The argument goes far beyond the very modest position which states that if we’re going to commit ourselves to anything about the relations of experience to nonmental world, the perceptual hypothesis is best, but that it is preferable outright just to remain neutral on whether the experience stands in any such relations at all. This very modest position is offering no explanation at all of the complexity involved in the occurrence of an experience with a representational content. The conclusion of our argument is not just that if we say anything at all about the environmental relations of the experience, then the hypothesis that it is a genuine perception is a consequence of the most simple hypothesis. What needs explanation is that a perceptual experience with a certain representational content occurs at all, with the complex of relations to the nonmental world this requires in the case in which its subject is properly connected to the world (and the complex of relations to other mental states whether or not the subject is so connected). Remaining neutral on the experience’s relations to the environment is no explanation of these relations at all. Correct application of the Complexity-Reduction Principle takes us all the way from the mental world to the nonmental, unconditionally - though, as always, defeasibly.
(3) Complexity-Reduction does not need to be thought about by the subject who forms an observational belief to which he is entitled because he has a certain perceptual experience. To be sensitive, in the formation of one's judgments, to the boundary between those states which are entitling and those which are not entitling it is not necessary that one know the philosophical basis of the distinction.
(4) It would be a topic for a different essay to relate the Complexity-Reduction Principle to the vast literature on Simplicity. A first point to note is that I am concerned with a certain feature of explanations, not of hypotheses considered in isolation. The Complexity-Reduction Principle does not classify hypotheses outright as simple or more complex. It rather places a constraint on the relative complexity of explaining condition and explained condition if a proposed explanation is to be a good one. Moreover, some notions of simplicity used in the literature do not match up at all well with the comparative notion relied upon in the Complexity-Reduction Principle. Conditions which would be described in some of the literature on simplicity as extremely simple nevertheless seem to demand special explanation. Suppose the whole of Earth’s northern hemisphere were covered with land, and the whole of its southern hemisphere were covered with water. Some accounts of simplicity would count that as more simple than a random distribution of bodies of water across the globe: but it would certain seem to demand explanation. If one wanted to trace connections between the notions involved in the Complexity-Reduction Principle and those used elsewhere, I conjecture that the notion of entropy as used in thermodynamics would seem to be a better starting-point. We need explanations of complex states of affairs which are consistent with the fact that in the system as a whole, entropy is increasing (even if in local subsystems it is not). I also note that the whole thrust of the present argument is in tension with the widely-canvassed view that simplicity is merely some kind of pragmatic virtue, perhaps even a human value, having nothing to do with truth or the likelihood of truth. No one could reasonably suggest that the explanations by natural selection that explain complexity in accordance with a Complexity Reduction Principle have merely pragmatic virtues, and have nothing to do with truth. I think the suggestion would be equally unreasonable as applied to complexity-reducing level-(3) explanations of the conditions under which we are entitled to make perceptual judgements.
(5) There is obviously some convergence between the present account of perceptual entitlement and the much more general epistemological theory that a state’s providing entitlement to belief in a given content is on occasion a matter of the best explanation of that state’s occurrence being the correctness of that very content. To keep this paper within acceptable bounds, I will not discuss these relations in detail here, but I will note that the crucial difference between the present approach and those theories is the appeal I have been making to the individuation of the content of an entitling state in the case of perception. I conjecture that if one tries to develop a ‘best-explanation’ theory of perceptual entitlement without appeal to the individuation of content, one will end up with something that is either too wide, or not sufficiently explanatory. Consider, for instance, an unrestricted general claim that the best explanation of the occurrence of an experience with representational content p, is that p is true, regardless of whether p is observational or not. This claim is not plausible for non-observational contents. One is entitled to take perceptions of things as PET scanners, or laser-pointers, or glaciated valleys, only in the context of background beliefs about these kinds of things and their characteristic spatial properties. Without such background beliefs, there does not seem to be even a defeasible entitlement: the transition from experiences with these contents to judgement would not be rational. On the other hand, if the entitlement is restricted to the observational cases, it is unexplained, on the general ‘best-explanation’ theory, why this restriction should be there. Its presence is wholly intelligible, and a consequence of the theory, on the level-(3) Complexity-Reduction explanation of generalizations about perceptual entitlement.
Many questions arise about this argument. I take two of them, before going on to compare this approach with a very different attitude to the possibility of explanatory epistemology.
(a) Is the argument I have given too limited, in that it applies only to perceptual experiences that are instance-individuated? Won’t this exclude representational colour contents, if we hold a theory stating that colour properties are to be explained in part in terms of their production of certain experiences in humans? Theories of this kind include the account I gave in Sense and Content. But aren’t we equally entitled to take the representational content of experience at face value in the case of colour, as we are for its spatial and its temporal content?
We do have such an equal entitlement. I think it can be captured consequentially in this theory. It is a consequence of what I have been arguing that the thinker is entitled to accept, in the absence of reasons for doubt, that he is perceiving properly. But if he is so perceiving, and has an experience in which something is presented in what in Sense and Content I called a red' region of his visual field, then by such accounts it will be true that the presented object is red. So the thinker will in these circumstances be entitled to the judgement that the presented object is red. (This also presumes upon constancy of categorical basis of the power to produce the experiences in question.) Some adaptation of this reasoning is available to those who do not agree that there are sensational properties of experience, but who still hold that the colour content of experience is subjectively determined in a way in which, for example, its spatial content is not.
(b) How can the Complexity-Reduction Principle be the basis of perceptual entitlement when, for instance, it seems to be so different from the notion of validity which underlies logical transitions to which one is entitled? I offer three remarks in reply. First, when we consider other transitions to which a thinker is entitled but which are not conclusive, moving to the least complex hypothesis seems to play a role. It would be hard to deny that in the entitlement to inductive inference by enumeration, supposing that not all F’s are G when all the encountered F’s are G is more complex that supposing that all F’s are G. If we think such cases are to be explained as tacit inference to the best explanation, avoiding complexity still plays a part in the choice of hypothesis, and of what one takes as needing explanation. Second, I will be arguing below that a range of nonperceptual entitlements that involve relying on psychological states are also ones in which the Complexity-Reduction Principle is implicated. If this is right, the perceptual case is not unique. Third, if we see a spectrum of cases ranging from conclusive entitlement through strong but nonconclusive, to weaker nonconclusive cases, we can see conclusive entitlement as the special case in which the entitling grounds give a sufficient condition of truth without needing to appeal to complexity-reduction in our philosophical explanations. With the nonconclusive, complexity-reduction needs to be brought into the philosophical account, but it is still serving a purpose which is uniform across the conclusive and the nonconclusive cases - contributing to the determination of which grounds are really reasons for thinking something to be true.
There are various approaches to these issues that would declare the explanatory task – that of providing characterizations at level (3) - to be either illusory or impossible (if not both). The explanatory task is that of explaining why certain methods, characterized in the generalizations at level (2), really are entitling, and the task is conceived as one which involves connecting use of those methods with the truth of the beliefs to which use of those methods leads. It follows that there are at least two ways in which the charge of impossibility may be supported. One way is to hold that there is no such notion of truth, that is, no notion of truth for which it is a genuine question whether those methods lead to true belief. The other way grants that there is such a notion of truth, but says that these generally characterized methods cannot be shown to be rational means of reaching the truth.
Some neo-Wittgensteinian views tale the first tack. These views hold that experiences provide ‘criteria’ for observational statements. This thesis is held in combination with a criterial theory of meaning, together with a purely minimalist, ‘redundancy’ theory of truth. Such a position is outlined in some of Crispin Wright’s early writings. For such a neo-Wittgensteinian treatment, there can be no real task of explaining why the fulfilment of a criterion is reason for believing the truth of the content for which it is a criterion. There is no notion of truth, on such a theory, for which this could be a real task On the theory of meaning endorsed by the criterial approach, the only conception we have of the truth of the observational sentence is one under which certain experiential conditions are criteria for its truth. Though the matter is complex and needs further argument, it seems to me that it is a genuine question, in need of an answer, “Why do these experiences give reasons to think that the content in question is true?” It is a question of a sort one can legitimately ask about any other entitling state and content that the state entitles one to judge. Even if the answer to this question about the perceptual case is one in which experiences of the kind in question play a role in individuating the content, that role must be linked to the truth-conditions of the content if the answer to the question is to be satisfying.
A different kind of theorist holds that we do indeed have a notion of truth which permits intelligible framing of questions at the third level, but holds also that we cannot give philosophical explanations at the third level of why generalizations at the second level hold. One prominent position of this sort is that of Humean naturalism as characterized by Peter Strawson in Scepticism and Naturalism: Some Varieties. Strawson endorses a view he attributes to Hume, and to Wittgenstein, “the view that our “beliefs’ in the existence of body and, to speak roughly, in the reliability of induction are not grounded beliefs and at the same time are not open to serious doubt” (19). “To attempt to confront the professional skeptical doubt with argument in support of these beliefs, with rational justifications, is simply to show a total misunderstanding of the role they actually play in our belief systems” (19). “…there is no such thing as the reasons for which we hold these beliefs” (20). In short, “one must refuse the challenge” (25). He also observes that a theory commonly offered in justification of these beliefs, that their content provides the best explanation of our experiences, is not the ordinary person’s reason for holding these beliefs.
It is surely true that we do naturally believe in the existence of material bodies. That we naturally believe, without philosophical reasoning, what we are entitled to believe is not something the theorist of level (3)-tasks is committed to denying.
On Strawson’s other grounds for his skeptical naturalism, it seems to me we need to start by distinguishing the entitling states from the principles which make them entitling. I suggest that some of what Strawson says about justifications is not true when taken to be about the entitling states; while if it is taken to concern the principles which make states entitling, it is not an objection to the present approach.
The entitling states – the experiences of our senses – do give us reasons (non-inferential reasons) for believing in bodies. The answer of the ordinary person to the question “Why do you believe there are material bodies?” is likely to be “I see and feel them”, and this is a pre-theoretical answer that accords with the theory of entitlement outlined here. The answer cites states that are entitling. The ordinary person will not or need not come up with any particular theory of why they are entitling. That could be an objection only if thinkers who are entitled to make certain transitions in thought also have to have some explicit account of why they are so entitled. This is not something we would accept when considering other topics. The idea that the thinker is entitled to make certain transitions, but can be without a theory he can articulate of why they are transitions to which he is entitled, seems to me a combination that we find when the non-philosophical thinker is making logical transitions, mathematical transitions, transitions involving the notion of probability, and indeed almost any transition whose legitimacy is grounded in the nature of notions of which he need not have a fully explicit grasp.
The theorist of level (3) characterizations need not at all be denying that the framework of material objects underlies “all questions and all thinking”. I myself doubt that one can give an account of thought about experiences themselves without treating it as parasitic on the ability to think of material things. This is just one of many ways in which the framework of material things underlies a domain of thought. Such framework theses seem to me to be entirely consistent with the existence of explanations of why certain states are entitling. They may not be consistent with certain bad attempts to carry through the level-3 task, attempts which suppose we have a conception of entitling experience entirely independent of our capacities to think about the external world. But we should not rule out a type of approach only because it has some bad versions, versions which contain extraneous elements that are in no way essential features of other answers to level-3 questions.
Other widely-discussed approaches in epistemology do not explicitly refuse the skeptic’s challenge, in the style of Strawsonian naturalism, but they still characterize normative notions in ways that seem to leave the challenge unaddressed. One such treatment is Goldman’s, in his Epistemology and Cognition. He requires of a legitimate method only that it reliably produce truth in ‘normal’ worlds. He writes:
“Imagine, the objection goes, that our actual world turns out to be an evil demon world. (Or imagine that we are actually brains in a vat being deceived by scheming scientists.) Intuitively, our beliefs would still be justified; yet the belief-forming processes being deployed are not reliable. Again the case is easy to handle. Its apparent strength rests on the assumption that the justificational status of the beliefs is determined by the reliability of their causal processes in the actual world. But this does not accord with our theory. Reliability is measured in normal worlds; and in this case, the actual world is an abnormal world!” (p.113).
So Goldman is classifying worlds in which the thinker is a brain in a vat as abnormal. Now I would not want to say that they are normal. But there seems to be a legitimate question: ‘Why should I rely on a method which yields true belief only in worlds which I haven’t, on Goldman’s theory, been given reason to think are actual?’ A more satisfying treatment must give the thinker some reason, if only a defeasible one, for thinking that he is not the brain in a brain-in-a-vat world. Otherwise, we will be back with skeptical conclusions. The Complexity-Reduction Principle aims to meet this need.
4. Links and Applications
(a) The Complexity-Reduction Principle is pertinent to wider issues about the relations between rationality and truth. It can be deployed in arguments over the issue of whether such notions as the default entitlement or default reasonableness of a method or rule can be elucidated in terms of its tendency to yield true beliefs. Hartry Field regards such an elucidation as “thoroughly implausible, on numerous grounds”. One of his grounds is
“The standard ‘internalist’ criticism: it is implausible to hold that our methods (assuming them reliable in the actual world) would be straightforwardly unreasonable in a ‘demon world’ (a world designed to make those methods unreliable, but undetectably so).” (Op. cit., p.125).
The treatment I have been proposing still permits a truth-based elucidation of default reasonableness for the practice of taking certain experiential contents at face value. Although this method is certainly not productive of truths in a demon-world, we argued that that world provides a more complex explanation of why there are experiences than does a world in which there is no such demon. The default-reasonableness of taking certain perceptual experiences at face value can be elucidated in terms of the tendency of doing so to produce true beliefs in the worlds which have the least complex explanation of why such an experience occurs, the ones in which it comes about in an easy way that there is such a perceptual experience. This is only the first step in accounting for default reasonableness in terms of a more elaborate relation to the production of true beliefs: but it is an essential first step.
(b) On the present approach, a thinker is entitled, in the absence of reasons for doubt, to take certain perceptual experiences which represent it as being the case that p at face value, and to judge that p. In the right circumstances, this judgment can be knowledge. Under this account, the transition is from perceptual experience to knowledge. There is no reliance on a premise to the effect that this experience, or experiences of some kind under which it falls, is perceptual. If there were such reliance, it is not clear that perceptual knowledge would ever be possible at all. If the transition to perceptual knowledge were even partly inferential, it could yield knowledge only if the premises of the inference were also known. But how is the premise that this experience (or experiences of such-and-such a kind) are perceptions to be known? It is not known a priori. If it is known a posteriori, it must rest on other cases of perceptual knowledge. But how are these other cases of perceptual knowledge to be attained, if they themselves rely on some premise to the effect that the experiences they involve are perceptual? This way lies infinite regress. Entitlement will never be attained unless some perceptual entitlement is non-inferential.
If this is correct, it bears on the correct diagnosis of what is wrong with Moore’s ‘Proof of an External World’. Moore’s ‘Proof’ is sometimes criticized on the ground that Moore’s own perceptual experience entitles him to judge ‘This is a hand’ only in the presence of the additional premise that his experience is produced (in the right way) by the external world. Under this diagnosis, since this is what the skeptic is questioning, Moore’s ‘Proof’ fails because its conclusion is already taken for granted in one of the argument’s (suppressed) premises. If what I have said is right, this cannot be the correct diagnosis. Perceptual entitlement does not rely on such additional premises.
Does this mean that the present approach is committed to accepting Moore’s ‘Proof’ as successful? Here we must distinguish between the existence of an entitlement, and having a dialectically effective reply to the skeptic. If there are not in fact any reasons for Moore to doubt his perceptual experience, he is entitled to judge that he has two hands, and to move from this to the conclusion that material objects exist. Entitlement is preserved throughout Moore’s line of thought. (One suspects he would have had other routes to this conclusion that material objects exist.)
If, however, the skeptic is challenging whether there really is an entitlement to rely on perceptual experience, then to offer Moore’s reasoning and nothing more is to beg the question. One needs at the very least to say more about why there is an entitlement to rely on perceptual experience. If the skeptic has more specific grounds for doubt, those too must be addressed. The important point is that it is entirely consistent to acknowledge that Moore’s argument should not by itself rationally convince the skeptic, whilst also holding that an entitlement to perceptual judgment is not a matter of inference.
For any given application we make of the earlier claims about entitlement, we can ask: is that application dependent only upon the first level of characterization, or does it additionally dependent upon the second level, or upon the third as well? If an application depends only upon a given level, that application is neutral on theses about the deeper levels. Theories that disagree on correct characterizations at the later levels may still agree on the application in question. In these initial remarks about the diagnosis of Moore’s ‘Proof’, the application depends only on the idea that there is defeasible perceptual entitlement of a non-inferential nature. This particular application is not dependent upon any one theory of the second and third levels. Theorists who disagree about those levels may nevertheless agree on a diagnosis of Moore’s ‘Proof’ which does not construe perceptual knowledge as inferential (and does not attribute that construction to Moore either).
(c) The generalization I formulated about the conditions under which a thinker is entitled to take for granted the representational content of his perceptual experience bears on the relationship between entitlement and factive states. I suggest that the generalization I formulated at level (2) supports the view that we can in some cases formulate the conditions under which a thinker is entitled to make a judgment in terms of his sensitivity to factive states, such as genuinely perceiving something to be thus-and-so, rather than formulating them in terms of a sensitivity to experiences whose content may or may not be correct.
Let us take an observational content Fa, and consider the conditions under which a thinker has a perceptual entitlement to accept it. We have so far been considering a defeasible rule, concerning the nonfactive state of perceptual experience (D):
A thinker is entitled to judge Fa if perceptual experience represents it as being so, where Fa is observational, in the absence of reasons for doubting he is perceiving properly.
A factive/outright rule for perceptual entitlement could be given the formulation (O):
A thinker is entitled tout court to judge Fa, where Fa is observational, if he perceives a (so given) to be F.
There is an argument that, under the level 2 generalization I formulated, the Nonfactive/defeasible rule (D) and the Factive/outright rule (O) are in a certain sense equivalent. A thinker is entitled to judge the observational content Fa in exactly the same circumstances whether he is following the defeasible rule (D) or the factive/outright rule (O).
We consider two cases, according as (1) the thinker does not, or (2) does, have reasons for doubting his perceptual states. In case (1), where the thinker has no reasons for such doubts, consider the case in which he judges that Fa by rule (D). Then, by our earlier arguments, he is entitled to assume he is in the circumstances with respect to which the observational content of his experience is individuated, and these, I argued earlier, are circumstances in which his experience is a genuine perception. So in the absence of reasons for doubt, our thinker is entitled to treat his experience as perceptual. Hence he will equally judge that Fa if he is following rule (O). Conversely, and trivially, if our thinker is entitled to judge Fa when following rule (O), since any genuine perception is a perceptual experience, he will equally be entitled to judge Fa if following rule (D). If there were any reasons for doubting that he is perceiving, then he would not be entitled to judge Fa under rule (O) either.
In case (2), where the thinker has reasons for doubting that he is perceiving properly, he will not be entitled to judge Fa under either rule.
The argument could be refined, without essential alteration, to treat the case in which the thinker has reasons for doubting just certain of his perceptual states. We would just consider cases according as states with the perceptual content Fa are in the doubtful category.
The factive/outright rule (O) seems to me to be more fundamental than the defeasible rule (D). The complexity of (D) is precisely what one would expect if the thinker were aiming to make his judgments sensitive to his genuinely perceptual states. It is not as if one would find intelligible a statement of perceptual entitlement that had defeating clauses relating not only to reasons for thinking that one is not perceiving properly, but rather to some other arbitrary condition not having to do with one’s perceptual mechanisms. The practice of taking one’s perceptual states at face value is the practice of taking it that they are delivering factual information about the world. Anything which makes it rational not to take them at face value must be something which undermines the proposition that one’s senses are delivering factual information about the world.
This is confirmed by the point that the qualification in (D), that the thinker is entitled only if he has no reason for doubting he is perceiving properly, could not be replaced by something more general, to the effect that there is no reason for doubting that Fa. If that were a correct formulation of a principle about entitlement, one could never be entitled to set to rest one’s doubts about whether it is the case that Fa by coming to perceive that Fa. One is frequently entitled to do just that. Doing so seems to be a paradigm of rationality. The fact that the qualification in the correct formulation of (D) concerns reasons for doubting that one’s experiences are genuinely perceptual further highlights the fact that this defeasible condition is simply aiming to make judgments to which it counts one as entitled sensitive to whether one’s experiences are genuine perceptions of the way the world is.
The claim of equivalence for (D) and (O) will have analogues for other informational states in cases in which there are contents which stand to those states as observational content stands to perceptual states.
It is a plausible principle that what a thinker is entitled to judge, and what is justified and what is rational, depends only what seems to the thinker to be the case, and not on which factive states he stands in. The truth of this principle (if it is true) does not imply that principles of entitlement, justification and rationality cannot mention factive states. It does not follow, because the thinker who responds to seeming-, non-factive states, may be doing so because he thinks or takes for granted that they are perceptual. The rules he is trying to follow may still mention factive states.
If the claim of equivalence of (D) and (O) is correct, then at least some entitlement rules mentioning only non-factive states are equivalent to rules mentioning factive states. If the equivalence holds, it must be false that there are entitlements which are captured by rules mentioning non-factive states, but not by rules mentioning factive states.
(d) The above explanation of perceptual entitlement bears upon the theses of my book Being Known. There I argued that certain concepts can be individuated by the conditions under which certain contents containing them are not merely rationally judged, but are known. This was part of the ‘Linking Thesis’ of Chapter 2 of that work. The idea was, for instance, that the concept a Babylonian expressed by ‘Hesperus’ is distinct from the concept he expressed by ‘Phosphorus’ because there are certain circumstances in which he can come to know that Hesperus is F without thereby being in a position to know that Phosphorus is F. Now if a skeptic questions whether a thinker is entitled to take perceptual experience at face value, his skepticism will extend to this Linking Thesis too. If perceptual knowledge is not possible, it follows that it is not possible to know by perception that Hesperus has some property without knowing that Phosphorus has it. Much of the discussion of Being Known would then collapse. One would be left only with skeptical responses to the challenge of integrating metaphysics and epistemology, rather than the ones attempted in that book which aimed to show that we really do know much of what we think we know, and without weakening our conception of truth for the propositions in question.
Being Known thus presupposed that some answer to skepticism exists, without actually supplying that answer. I take the theses about perceptual entitlement at the third level of explanation to be the start of such an answer. They can be seen as a contribution to the task of explaining why, in the perceptual cases, the Linking Thesis is true. In a discussion essay written after Being Known, I spoke of a ‘Second Linking Thesis’, linking instance-individuation with entitlement and, thereby, with knowledge. The theses at the second and third levels in the present paper certainly say more than the Linking Thesis of Being Known. They are, however, contributions to the tasks of explaining the conditions under which the Linking Thesis is true, and of explaining why it is true, rather than being autonomous, additional theses.
The claims of the present paper also bear on the somewhat creaky discussion of the ‘rationally nondiscretionary’ in Being Known. The rationality of judging an observational content on the basis of perceptual experience requires the rationality of two things: the rationality of moving from the content of the perceptual experience to the content of the observational judgment (in theories under which these are distinct contents); and the rationality of taking perceptual experience at face value in the first place. What I have been offering is an explanation of why this second thing is rational; the approach of Being Known is incomplete without it.
(e) The existence of defeasible but non-inferential entitlement structures goes far beyond such cases as perception and the various forms of memory, and possibly testimony. The existence of defeasible, non-inferential entitlement relations can also provide more room for maneuver in the philosophical account of some areas of moral thought. One example is provided by the discussion in an important recent paper by Allan Gibbard, ‘Normative and Recognitional Concepts’. In the part of his paper concerned with ‘Thick Recognition’, Gibbard observes that one’s understanding of a situation may be ‘heavy with demands for action’ (MS 11). One may, for instance, perceive an unjust act as demanding rectification. Such cases pose a problem for views that sharply separate how things are and what to do. Gibbard observes, acutely, that it would be completely unacceptable to think that the thing to do is to act on every impression that the situation demands a certain action. We are all subject to prejudices, and there can be illusions of demands. He says that “the principle we’d need to accept” in order to take the apparent demands as face value is “appalling” if it means we should act on any impression of any demand. His conclusion is that we should just take it as part of our situation that we have this sense, and this is “a psychological aspect, not plan-laden in itself” (MS 11).
These cannot be the only two possibilities, if defeasible but non-inferential entitlement structures exist. Taking the seeming-demands of a situation at face value in deciding what to do may be something to which one is prima face entitled, an entitlement which can be defeated if the seeming-demands are promoting a course of action one has reason to think is morally wrong. The “appalling” principle Gibbard cites is the analogue of the epistemically wholly unacceptable principle that one should always take perceptual experience at face value – even an experience one knows to be of a perfect trompe l’oeil, or knows to have an inconsistent content. Putting the fact that there are certain seeming-demands into the specification of one’s situation is the analogue, for the practical case, of the treatment of perception that says one has a premise to the effect that one is having an experience of a certain kind. That approach has made it impossible to see how perceptual knowledge could be attained. The believer in the importance of thick concepts may insist that the apprehension of demands for action should not be assimilated to a model that has proved unworkable in the perceptual case. The defender of thick concepts and their significance should invoke the structure of a defeasible, non-inferential entitlement relation. I am not necessarily endorsing this position: my point is just that defeasible, non-inferential entitlement relations makes available this position in logical space.
Some of the features of this account of perceptual entitlement generalize beyond perception to cases in which the direction of the relation of causation between world and mind is the opposite of that in perception. It is not crucial to the general form of the account of entitlement I outlined in the perceptual case that the entitling mental state be caused by the conditions to which one is entitled. The general structure of the account can still get a grip provided that the mental state is individuated by certain of its relations to the conditions mentioned in a statement of the entitlement. In my view, this applies in the case of action - as one might well expect from the many symmetries, now widely recognized, between perception and action.
For basic bodily action-types φ, the mental event-kind of trying to φ is individuated by the fact that events of that kind tend to produce φ-ings, when the subject’s central control system is properly connected to his body. Now thinkers normally know what they are doing. In fact, they have a distinctive phenomenology of action. It can seem to the subject that he is φ-ing, and this apparent awareness can be present even in the subject who is acting with an anaesthetized or damaged limb from which there is no proprioceptive feedback (nor any illusion of such perceptual states). In fact, some striking experiments by Tony Marcel suggest that apparent awareness of one’s φ-ing can be produced by one’s trying to φ, even if one is not in fact φ-ing, and even when the limb employed is neither anaesthetized nor damaged. In such cases, the distinctive apparent awareness a subject has of his own actions seems to be a result of his tryings.
The question then arises: how can this distinctive awareness yield knowledge, on the part of the subject, that he is acting a certain way? A reliabilist would say that in circumstances in which the agent does know, trying to φ is reliably correlated with φ-ing. But there are strong objections in other cases to pure reliabilism. Is there some explanation of how we have knowledge of what we are doing which is not dependent upon perception of ourselves, or upon proprioceptive feedback, but which does not involve a reversion to reliabilism? I suggest that there is, and that it relies on a generalized version of the principle on which we have relied in the account of perceptual entitlement. The event-type of trying to φ is individuated by its relation to φ-ings in the case in which the agent’s control center is properly connected with his body. The thinker is entitled to take it that he is in the circumstances with respect to which these event-types are individuated. (Again, an explanation in terms of complexity-reduction could be given of why he is so entitled; though the present argument for knowledge of action requires only that such an entitlement exists.) But in these circumstances, tryings to φ do produce φ-ings. The awareness that is produced by the trying to φ can then, with entitlement, be taken at face value. In suitable circumstances, this can yield knowledge on the subject’s part that he is φ-ing. If we reject this approach, it would be a real task to explain philosophically how we have knowledge of actions we are performing, without reverting to reliabilism. One cannot simply apply the perceptual model straight, since as we saw, it is not the bodily action itself that causes the apparent awareness of action.
It is tempting to apply the same generalization to other cases too, outside the realm of perception and action. Consider, for instance, the entitlement to self-ascribe beliefs on the basis of one’s own judgments. Judgment is individuated as an event of a kind which, when all is working properly, leads to belief. So one can explain how one can know what beliefs one has by making self-ascriptions that are sensitive to one’s one judgments – even though this is certainly a fallible method. This explanation does not involve a reversion to pure reliabilism. The method is a rational one. Again too, one could develop an appeal to complexity-reduction to explain why the entitlement exists, given the complex relations an event must stand in if it is to be a judgment with a given intentional content.
At this point, there are many tasks for further work. Here are some of the questions which arise. Can every case of entitlement by a state with intentional representational content be assimilated to the present model, or to some natural extension thereof? If so, how? If not, why not? And if not, what is the correct explanation of entitlement for cases which no extension of the present model can capture? There is evidently massive further work to be done. But I do conjecture that entitlements that are rather different from that of the perceptual case, or any extension thereof, are possible only because perceptual entitlement is also possible. The cases in which the Complexity-Reduction Principle applies are fundamental.
* * * * * * * * *
 The position outlined here was foreshadowed in my remarks as a commentator at the Rutgers Epistemology Conference in April 2001, and developed in writing the following summer. I thank Tyler Burge and Stephen Schiffer for valuable advice and comments.
 This distinction is further discussed as that between defeasibility of identification and defeasibility of gournds, in my paper ‘The A Priori’, forthcoming in The Oxford Handbook of Analytical Philosophy, ed. F. Jackson and M. Smith (Oxford: Oxford University Press, expected 2003).
 Aspects of the Theory of Syntax (Cambridge, Mass: MIT Press, 1965), p.24.
 R. Stalnaker, ‘A Theory of Conditionals’, Studies in Logical Theory ed. N. Rescher (Oxford: Blackwell, 1968); D. Lewis, Counterfactuals (Oxford: Blackwell, 1973), esp. pp. 32-5.
 Aspects, pp.25-7.
 J. Pryor, ‘The Skeptic and the Dogmatist’, Noûs 34 (2000) 517-49, at pp. 536ff.. This position is in the same spirit for entitlement in perception as Burge’s on the entitlement to accept the utterances of interlocutors: see T. Burge, ‘Content Preservation’, Philosophical Review 102 (1993), 457-88..
 Pryor uses the notion of propositons which our experience ‘basically represent’: these are propositions we seem to perceive to be so, but not in virtue of seeming to perceive other propositions to be so (p.539). This will give a wider class of entitling states than results from application of the criterion of observationality outlined below.
 Being Known (Oxford: Oxford University Press, 1999), Chapter 3.
 C. Wright, ‘(Anti-)Sceptics Simple and Subtle: G.E.Moore and John McDowell’, forthcoming in Philosophy and Phenomenological Research.
 From Chapter 1, 'Explaining the very improbable', of R. Dawkins’ The Blind Watchmaker (Longman: Harlow, Essex, 1986) p.14.
 Cp. Dennett, ‘Artificial Intelligence as Philosophy and as Psychology’, in his Brainstorms (Montgomery, Vt.: Bradford Books, 1978). Given the theses of the present paper, the following quote from Dennett also seems prescient. It concerns what he calls ‘skyhooks’, procedures, capacities or information that does not result from earlier selection processes and testing of the sort envisaged in the theory of natural selection. Dennet writes, “The renunciation of skyhooks is, I think, the deepest and most important legacy of Darwin in philosophy, and it has a huge domain of influence, extending far beyond the skirmishes of evolutionary epistemology and evolutionary ethics.” (‘In Darwin’s Wake, Where am I?’, Proceedings and Addresses of the American Philosophical Association 75 (2001), p.23.) If the present paper is right, this Darwinian legacy is of significance even in the relatively a priori domain of theories about the normative concept of entitlement.
 Oxford: Oxford University Press, 1983.
 See for instance his ‘Strawson on Anti-Realism’, repr. in his Realism, Meaning and Truth (Oxford: Blackwell, 1987), esp. at p.75ff..
 London: Methuen, 1985.
 Harvard University Press: Cambridge, Mass., 1986.
 ‘Apriority as an Evaluative Notion’, in New Essays on the A Priori, at p.124.
 In Moore’s Philosophical Papers (London: George Allen and Unwin, 1959).
 ‘(Anti-)Sceptics’, op.cit..
 Contrast R. Wedgwood, ‘Internalism Explained’, forthcoming in Philosophy and Phenomenological Research.
‘The Past, Necessity, Externalism and Entitlement’, Contribution to a Symposium on Being Known, Philosophical Books 42 (2001) 106-17.
 On testimony, see Burge, ‘Content Preservation’.
 Forthcoming in Philosophy and PhenomenologicalResearch.
 See the description of his vibro-tactile experiments in his paper ‘The Sense of Agency – Ownership and Awareness of Action’ forthcoming in Agency and Self-Awareness, ed. J. Roessler and N. Eilan (Oxford University Press, publication expected in 2002). For further philosophical discussion of the issues, see also my paper ‘Action: Awareness, Ownership and Knowledge’ in the same volume.