views updated


A distinction is often drawn in philosophy between two types of objects of awareness in perception. First, there are physical objects or substances (such as chairs, books, rocks, and water) and living organisms (animals, plants, and human beings insofar as they are perceptible, that is, their bodies). A common technical term for all these is material objects. Second, there are data of immediate awareness, which we shall refer to as "sensa" (singular, sensum), such as color patches or shapes, sounds, smells, and tactile feelings. This distinction is usually fourfold: (a ) in statusmaterial objects are external, located in physical space, and "public" (observable by different persons at once), while sensa are private and are usually held to have no external physical existence; (b ) in extentmaterial objects may at one time correspond to several sensa and normally persist throughout the occurrence of many sensa; (c ) in directnessthe perception of material objects is indirect, that is, it involves inference from or interpretation of sensa that are "given" directly to consciousness; (d ) in certaintyone is always certainly aware of sensa but not necessarily so of material objects.

There is no universally accepted term for sensa; sensations and sense data are commonest but indicate a further subdivision. Sensation is customarily used by scientists and psychologists and carries with it the suggestion that sensa are the immediate mental effects of brain activity resulting from the excitation of a sense organ by external stimuli. It and the less specialized term sense impression may be used interchangeably for the whole experience of awareness of sound, color, and the like, or for any sensum (such as a sound or a color patch) distinguished within it. The term sense datum (plural, sense data ) apparently originated with G. E. Moore but was introduced in print by Bertrand Russell in 1912. It later became particularly associated with the sense-datum theory of Moore, C. D. Broad, and H. H. Price, while Russell developed different views and came to use other terms.

Sense data are not meant to carry any implications of causal theory, and awareness of them is called sensing (the term sense datum is used for the sensum only, not for the whole experience). With the development of the sense-datum theory, controversy arose between those who regarded sense data as objects distinguishable from the act of awareness of them (act/object analysis) and those who denied this and claimed that sensing is really of "sense contents" (adverbial analysis). But the terminology is generally fluidfor instance, some modern neurologists use the term sense data instead of sensations in causal contexts. Similar concepts are found in earlier writers, though their language is different. John Locke's "ideas of sense," George Berkeley's "ideas" or "sensible qualities," and David Hume's "impressions" are all forms of sensa.


It has often been maintained, by philosophers as well as by psychologists, that perceiving consists in the synthesis and interpretation of sensations. But it must be realized that the occurrence of sensations in all perception is only a hypothesis and not an obvious feature of experience. In ordinary language, one may speak of having or feeling sensations of thirst, cold, or pressure and may refer to itches or pains as sensations. But the technical use of the word sensations involves a considerable extension of meaning, since one then speaks of visual or auditory sensations (that is, colors or sounds), while such locutions have no place in ordinary speech. We do not have green sensations in our eyes, nor do we normally feel or have sounds in our ears. Admittedly we do have afterimages, spots before the eyes, or ringing in the ears; but these are special cases because, unlike the objects or data of normal perception, the images, spots, or ringing "follows us around" and cannot be avoided by moving the head, closing the eyes, or stopping the ears. Indeed, in normal perception we are conscious not of colored shapes or of sounds as such but of material objects, or at least of ostensible material objects. Admittedly we may sometimes be aware of sounds, smells, tastes, or feelings of pressure, as distinct from objects or object properties, but it is doubtful how far these can be said to be sensations.

Sounds and smells seem public and external: Two or more people may hear the same sound or smell the same smell and agree on its source; sounds travel, and a smell may fill a room. Tastes are a borderline caseprivate and in the mouth, yet in a sense external to the skin and membraneswhile feelings of pressure or warmth are partly sensations proper and partly seem to be awarenesses of heavy or warm objects. However, colors and colored shapes normally seem quite external, public, and at a distance from us.

Sensations in this technical sense (private mental objects of immediate awareness) are thus mainly hypothetical occurrences. Their postulation can be justified only by its success in explaining the facts of perception, and it rests on two grounds. First, there is the causal argumentperception of objects depends on and is conditioned by a chain of causal processes; for example, light waves or sound waves stimulate the appropriate sense organ, causing impulses to travel along nerves to the brain and activate the appropriate receiving area. Perception cannot, therefore, be direct contact or confrontation with external objectsall immediate awareness must result from the causal process and be an awareness of mental sensations due to brain activity. Since they are thus separated from the external object in time and space, sensations cannot be identified with its properties, though they may resemble them.

Second, there is the psychological argumentmany characteristics of perception show that it is not a direct intuitive awareness but involves interpretation of sensations. Thus, error and illusion are really misinterpretations; perception of motion, depth, and distance involves the use of sensory "cues"; and perceptual identification and discrimination are interpretative, not immediate, since they can be improved by learning and experience. (Both these arguments are discussed at greater length under the Perception entry. Here we may simply note some relevant difficulties.)

the extent of sensations

Even if the causal argument forces us to distinguish between external material objects and the immediate objects of awareness caused by brain activity, it does not follow that the latter must be sensations, such as colors or sounds. They may be percepts, that is, mental contents that correspond to whole material objects, though here the psychological argument comes in, suggesting that percepts are the products of interpretation. Supporters of the theory of sensations, no doubt influenced by discoveries concerning the atomic structure of matter, at one time even claimed that the basic sensations are "atomic," that they are sensory point-elements, each corresponding to a different nerve cella patch of red color would thus be made up of many sensations of red. This view has now been completely abandoned, largely as the result of the experiments of the Gestalt psychologists, which show that our primary awareness is of organized wholes or figures (Gestalten in German), and not of elements into which these wholes might theoretically be analyzed. But even though sensations are not now thought of as minute elements that we synthesize, nonatomic sensations (colored patches of a larger size, or patterns of them, as well as sounds, smells, and so on) may still be regarded as data that we interpret in perception.

awareness and interpretation of sensations

The awareness of sensations or, for that matter, of percepts must itself be explained; the danger is that it will be construed as analogous to perceiving; for example, that seeing objects will be explained as seeing sensations caused by them, which is a circular explanation and can thus lead to an infinite regressseeing sensations must require seeing further sensations, and so on. (Compare the duplication objection to representative realism in the Realism entry). It is therefore necessary to maintain that the awareness of sensations or percepts ("having sensations") is a special kind of direct awareness different from perceiving, an amendment explicitly adopted by the sense-datum theory.

The problems of the psychological argument are (a ) that interpretation of anything would commonly be regarded as presupposing consciousness of what is interpreted, and we are normally conscious neither of having sensations (as opposed to perceiving objects) nor of interpreting them; and (b ) that the nature of the interpretation of sensations is controversiala range of theories is possible because it is not introspectable. The sensationalists (James Mill, J. S. Mill, and others who derived their inspiration from Hume) claimed that perceiving is the association of various sensations. Association is a vague term and was explained as the customary linking of ideas or sensations that are similar, contiguous in space and time, and so on. F. H. Bradley and other idealists successfully attacked the sensationalist view as inadequate to explain the facts of perception; instead, they claimed that the interpretation is an inference leading to a judgment, supposing that the possibility of error in perception required this. But this overintellectualized perceiving; inferences and judgments are not the only forms of mental activity liable to error.

Arguments for the Introduction of Sense Data

Since the start of the twentieth century, philosophers have made little use of the concept of sensation in their theories but have instead talked of sense data or sense contents. Though the same thingscolor patches, sounds, smells, and tasteshave been put forward as examples both of sensations and of sense data, the new terminology marks several changes. Recognition of the visual depth or stereoscopic qualities of sense data means that one visual sense datum or color patch is usually held to correspond to the whole of the visible part of an ostensible object (so that one may have striped or variegated sense data). Little detailed attention has been paid to psychological phenomena, except for discussion along traditional lines of error and illusion and their bearing on whether perceiving is a form of judgment. There has also been almost a revulsion from causal arguments, clearly influenced by their tendency to involve one in the notorious difficulties of representative realism. Instead, a fresh start has been made in the conviction that philosophy has its own distinct contribution to make in the logical and introspective analyses of perception and in the consideration of relevant epistemological issues, that is, of the extent to which perception provides knowledge of external reality. Nevertheless, with some adjustment the new arguments might be supplemented by and in turn supplement the causal and psychological arguments for sensations.

Sense data are defined as whatever is "given" or "directly present" in perceiving; they are the object of sensing, of "direct" or "immediate" or "actual" awareness in perception. The claim that this awareness occurs within perceiving is essential to the analysis. To most of its exponents it seems a clear fact of our experience as percipients, one revealed by reflective examination. "Direct" is explained by Price (in Perception ) as meaning intuitive or "not reached by inference, nor by any other intellectual process." This formal definition was often supplemented by a kind of ostensive one: Moore, J. R. Smythies, and others gave instructions for looking at an object or scene and picking out the sense datum, such as a colored shape. (Misleadingly, afterimages were sometimes offered as examples of sense data, but their difference from normal perception has already been noted; misleadingly also, some talked of seeing or hearing sense data.)

This definition of sense data naturally raises the question "Why not say that tables, chairs, and other material objects are given or directly seen?" In answering this, these philosophers produce various arguments for distinguishing sense data from material objects.

the certainty argument

The certainty argument was stressed by Price and by Russell in his search for "hard data," though it is also found in other sources. Directness or givenness implies certaintywhat is given must be limited to what we are absolutely certain of. But in any perceptual situation we cannot be sure that we are aware of any particular material object. For example, an object that seems to be a tomato may in fact be something quite differenta wax imitation, perhaps, or a reflected patch of light, or a hallucination (that is, not be a material object at all). Yet whatever the illusion may be, there can be no doubt, when we seem to see a tomato, that there is given a red, round, bulgy patch of color, a sense datum. Another version of this argument is the method of reduced claims; by confronting him with possible sources of error, you force the person concerned to reduce his claim from "I see a tomato" to what he actually and directly sees or, rather, senses: "I see a red, round color patch."

the partitive argument

When we observe a tomato or a bell, what we "actually see"the "objective constituent" of the situation, what is given or sensedis the colored shape that seems to be its front surface. This is a sense datum. We assume that the object has other surfaces and has other characteristics, such as causal properties, three-dimensionality, and persistence in time; and if we loosely say that we see a bell, we imply that we are perceiving an object possessing these properties, although we do not directly see or sense them. This argument, which stresses extent of sense experience rather than certainty, was preferred by Broad and Moore but seems inferior in suggesting that sense data are those parts of an object that we "actually see" on a given occasionwhich raises difficulties with respect to illusions.

the argument from the content of illusions

When a drunkard sees a hallucinatory pink elephant or sees two bottles when only one is present, what is the elephant or second bottle if it is nothing material? The sense-datum theorist answers, "A private object of awareness, a sense datum," and applies this also to cases of the relativity of perceiving: For example, when a round plate looks elliptical to a person standing at one side, the elliptical appearance cannot be the plate, which is round; it is an elliptical sense datum private to that person. Indeed, it is argued that at all times we are directly aware only of sense data, since there is no qualitative jump between the cases where one cannot be directly aware of an object, and so must be sensing sense data, and the normal cases where we think we are directly aware of an object. This gradation or lack of jump is particularly clear in the case of relativity, as when we gradually move from where the plate looks round to where it looks elliptical, but it also applies to many hallucinations where the illusory sensa are integrated with a genuine background. In short, perceiving a material object involves sensing sense data related or "belonging" to it; when the plate looks round to me and elliptical to you, I am sensing a round sense datum belonging to it and you are sensing an elliptical one.

The Full Sense-Datum Theory

The fundamental conception of sense data, as directly given elements of experience, spread far beyond epistemology. Both the atomic facts of the logical atomists and the supposedly incorrigible basic or protocol propositions of the logical positivists had as their prime examples simple statements about sense data (or sensa generally), such as "This is red." But the conception was also developed into a full theory of perception by consideration of the following topics, even though disagreements led to variant accounts.

the general nature of sense data

The arguments for the introduction of sense data, if valid, show that sense data are given and provide examples of them. Further alleged properties emerge from the discussion of illusions and relativity, namely, that sense data (1) are private, each sensed by only one percipient (see argument from the content of illusions); (2) are transitory existents, lasting only while they are sensed, so that they are usually claimed to be events rather than things or properties; (3) are distinct from the percipient and seem to be external (in contrast with sensations); (4) are without causal properties, for sounds (as opposed to sound waves) cannot act on other things, nor can colors or tastes, though the sensing of them may affect a person; (5) cannot be other than they appear to be, or the certainty argument is undermined.

Despite wide agreement on most of these points, a considerable divergence of view arose about (3) and (5). Point (3)that sense data are distinct from the percipient and seem to be externalinvolves what came to be called the act-object analysis of sensing. Largely on phenomenological groundson how direct experience of color patches, sounds, and such seem to the person concernedPrice and others claim that sense data have distinct existence, that they are objects distinguishable from the act of awareness of them. But some philosophers maintain that the data are only "sense contents" and do not exist apart from the sensing of them any more than does a pain or sensation. This view is formulated in the so-called adverbial analysis of sensing, namely, that "I sense a red color patch" is properly to be regarded as a statement of how I sense or, to put it in a different way, "red color patch" is an internal accusative of the verb sense, just as "waltz" is an internal accusative of dance in "I danced a waltz."

There is agreement on point (5)that sense data cannot appear to be what they are not, for example, sense data cannot appear elliptical when they are round. (Even this is dubiousan apparently pink expanse may, on examination, be found to consist of red dots on a white background.) But some say that sense data can fail to appear as they are (do not reveal their full properties at first sight); thus, one may see that a colored datum is striped without noting how many or how thick the stripes are. Others deny this, claiming that a closer look results in a fresh sense datum. In fact, the theory cannot deal satisfactorily with the phenomenon of attention. A thing may look quite different on careful examination from the way it looks at a casual glance, and the difference seems to be a matter of how attentively we look, a matter of changes in our mode of observation. In line with this evidence, one should say that sense data may reveal their full properties only on a closer examination, but then one is suggesting that sensing may at times be casual and inattentive and is thus undermining the fundamental claim that sensing is certain and incorrigible.

the relation of sensing to perceiving

The distinction between sensing and perceiving is threefold. First, perceiving is the awareness of some material object; except in certain kinds of illusion this awareness is the result of the object in question (or light or sound from it) acting on the percipient's sense organs. Sensing is the awareness of private sense data that differ from material objects and do not affect the sense organs. Second, sensing is claimed to be direct, immediate, and incorrigible, a form of knowing. Owing to illusions, perceiving cannot be this; it is fallible and indirect. Third, the indirectness of perceiving is said to consist in its being mediated by sensing; perceiving involves sensing, contains sensing within it.

Various views are possible about the nature of this mediation of perceiving by sensing, but they are best expressed as theories of perceptual consciousness. The same kind of consciousness of a tomato, for example, seems present in normal perception, when one sees a tomato as a tomato; in an illusion, when what one sees as a tomato is a piece of wax; and in a hallucination, when no corresponding material thing is there. The kind of consciousness present in these three cases may be called perceptual consciousness and is more conveniently discussed than perceiving, where the implication that there is an object acting on the sense organs complicates the issue.

Some, such as Brand Blanshard (The Nature of Thought, London, 1939, Ch. 2), claim that perceptual consciousness consists in sensing a datum and judging or inferring that it belongs to a material object. Price, however, argued that this is too intellectual and does not fit the facts. We unquestioningly accept or take for granted rather than infer or judge, and therefore he defined perceptual consciousness as sensing a sense datum (or data) and taking for granted that it (or they) belong to a material object. Others have said that we refer the sense datum to a material object, but refer is vague.

Two points of interest arise here. First, philosophers have most often said that we accept or judge that the sense datum belongs to a physical object. This seems obvious only about smells or tastes, and one would on first thought say we assume that the visual sense datum or color patch is the tomato. There is a reluctance on the part of sense-datum theorists to allow this, presumably because they are influenced by the partitive argument or by their knowledge that ex hypothesi the sense datum cannot possibly be the physical object. But there seems to be no reason why the ordinary person, whose mental processes are being described, may not mistakenly assume this; one would, for example, say "That patch of white over there on the hill is a sheep" (admittedly, the patch as "public" is hardly a sense datum, but it is the nearest one can get to a sense datum by ordinary examples).

Second, to say that we judge or infer that a sense datum belongs to (or is) a physical object is implausible, for it implies we are conscious of it first as a datum, which is not true to the facts: There is no passage of mind from datum to object, as in inference. Even to say we subconsciously judge or infer is unsatisfactory, for it seems extravagant to suppose that we constantly do subconsciously what we never do consciously. Price attempts to overcome this by maintaining that to take for granted that A is or belongs to B, one does not need to distinguish them at the timeindeed, the contrary is implied. Sensing thus comes to be regarded as a sort of sensory core within perceptual consciousness, surrounded, as it were, by the further activity of taking for granted. The two states of mind, sensing the red sense datum and consciousness of the tomato, arise together and simultaneously and can be distinguished only by subsequent analysis.

Even this account may be criticized on the grounds that it still does not do justice to the evidence of experience, namely, that perceptual consciousness is one unitary and unanalyzable state of mind, not two. No subsequent analysis of experience reveals sensing as an element within perceptual consciousness. Analysis or reflective examination can result in a "reduced" or critical phenomenological mode of observation in which one distinguishes sounds or colored shapes as such without attributing them to objects, but if this is sensingand it seems to be the nearest one can get to itthen it is a quite different state of mind from normal perceiving. There is no ground for supposing that this, achievable only by an effort of analysis, occurs as part of normal unconsidered perception. In general, therefore, the attempt to establish sensing sense data as an omnipresent basic element in perceiving faces the same difficulties that faced the claim that perceiving is the interpretation of sensations.

Another way of seeing the error is to consider the normal usage of "taking for granted." Price's analysis is at first sight closest to "Y saw the book and took for granted that it belonged to B," but then Y is referred to as conscious of the book, while the average percipient is not conscious of sense data as sense data; he is conscious only of the material object. This difficulty can be avoided by the formulation "X took for granted that A was B "; for example, that the piece of wax was a tomato, or that the visitor was the man he was expecting. In each case both A and B denote the same entity (the wax or the visitor). A describes this entity in a way that the speaker knows to be correct; B describes it as X saw it. Similarly, one might say, "He took for granted that the sense datum (A ) was a material object (B )." But this will not really save the analysis in which the datum and the physical object are alleged to be two quite different entities; to fit the analysis the first phrase (A ) must also be a description of the alleged object of awareness of X, not of the speaker. Price seems to be making the mistake of offering as a description of a percipient's actual mental content what is in fact a description of the situation that can be made only by someone correcting the percipient's error.

the relation of sense data to physical objects

One of the vaunted advantages of the sense-datum analysis of perception is its neutrality with respect to the traditional realist theories of knowledge. (Idealism was ruled out by the original claim that sense data are distinct from the sensing of them.) Indeed, sense data were even said to be neutral in that so far as the analysis is concerned, they can be mental or physical or neither. Consequently, it is possible to state the various theories of knowledge in terms of sense data. Naive realism reduces to the view that sense data are parts of the surface of material objects; representative realism would claim that sense data are mental existents caused or generated by cerebral activity ultimately due to material objects and that sense data resemble the properties of these objects. (The second view and, if not too naive, the first also, could admit "wild sense data"hallucinations that are not part of or caused by physical objects.)

Moore at times toyed with supposing that sense data are parts of the surface of objects (and even seriously discussed whether they might be identical with objects), though this must have been due to his affection for the partitive argument. The other arguments for sense data and general considerations about illusion do not allow this; for example, a round dish cannot have an elliptical sense datum as part of its surface. Representative realism is a more likely possibility: Neurologists such as Smythies advocate this theory in terms of sense data, and Broad proposed something not unlike it. Most of the philosophers have, however, rejected it in view of its traditional difficultyif our observation is limited to sense data while material objects are only assumed causes of sense data, then these objects are in fact never observed and therefore may, for all we know, not really exist.

A more common view is that sense data belong to material objects in the special sense that the latter are composed of "families" of sense data. This "family" relationship is not literally one of whole and part, as in naive realism; the material object is supposed to be a complex system or pattern of groups or sequences of sense data. But if a physical object is simply a family of sense data, then when no sense datum occurswhen the object is unobservedthe object must cease to exist. This is felt to be too paradoxical, and two main lines of development within this view have been put forward: (1) phenomenalism, in which the object is regarded as a family of actual and possible sense datawhen unobserved, it consists solely of possible sense data; (2) a compromise theory put forward by Price in which the material object, while mainly such a family, contains a physical occupant that persists, even while it is unobserved, as the source of all its causal properties. The notion of a physical occupant has some analogies to Immanuel Kant's notorious thing-in-itself, and this view has not obtained widespread acceptance.

This divergence of view reflects a central dilemma in the sense-datum theory. If the theory maintains that sense data belong to material objects or that the latter in some way consist of them, then it is difficult to explain (a ) the persistence of such objects when unobserved; (b ) the privacy that all versions attribute to sense datahow can a public object be a family of private sense data?; (c ) the conditioning or even generation of sense data by the sense organs and nervous system, which is required by the physiological facts, by the occurrence of hallucinations or color blindness, and by the effects of attention and learning on perception. (Most sense-datum theorists admit the generation as well as the conditioning.) But if one does not say that sense data belong to or constitute material objects, the distinctness and apparent depth of sense data (at least of visual ones) is difficult to explain; and, more important, sense data tend to become mental entities like sensations. This, together with the privacy and the generation by the brain, leads one into representative realism.

One attempt to escape this dilemma is to say that sense data are extended and located in their own private "sensible" space along the lines first suggested by Russell in his Mysticism and Logic and Our Knowledge of the External World. There is one such sensible space, with its own extension and dimensions, for every point in physical space, and the latter in fact becomes the system of points at which sensible space occurs. A physical object is thus, as it were, spread over physical space in a series of "perspectives" or "unperceived aspects," in the special sense that from different points in physical space, granted that sense organs and brain function properly, sense data may occur in sensible space but also belong to the object as appearances of it and reproduce its characteristics in a way modified both by the viewpoint and by the nature of the sensory apparatus.

This theory is very complex, which means that any summary of it is necessarily garbled. Two of the complexities are that a special interpretation is needed of what we normally call the volume occupied by a physical object and that account must be taken of the different senses, for sight, sound, and touch at least each have their own specific spaces. (Russell later spoke of sensible space as a construct of these spaces, but a construct cannot be the space in which immediately given sense data are located.) A further difficulty is that a given sensible space cannot really be at a point. Not only are the hands, say, at some distance from the eyes, but the brain and the sensory activity associated with perception of an object at one time and place are also really spread over an area. However, the major objection is once again the causation and conditioning of sense data by sense organs and nervous system. How do they influence or produce data in sensible space, or modify the appearance in sensible space of an object in physical space? As soon as one tries to fit in the causal processes, it is difficult to avoid straightforward representative realism, in which all this elaboration becomes unnecessary; perspectives become otiose, except as mere possibilities, or turn into light waves and sound waves. Hence, Russell's later views gradually approach representative realism (for example, in Human Knowledge, 1948).

There does, in fact, seem to be no satisfactory way out of this dilemma for the sense-datum theory. Upholders of it must embrace one horn or the otherthey must maintain pure phenomenalism or representative realism. Each has its well-known difficulties, but the second, though once thought hopeless, is now perhaps more easily made plausible than the first.

Difficulties concerning Sense Data

A number of difficulties have been noted already in the full theory, but others lie even in the arguments for sense data.

the certainty argument

Various objections may be made to the certainty argument. First, so far as introspective examination is concerned, our awareness is, as we have mentioned, of putative objects, not of color patchesone sees a tomato or something looking like one. Awareness of color patches as such is a different kind of observation from normal perceiving, not a sensory core within it. One may more readily be said to be directly aware of sounds or smells as such; but even then, as we saw concerning sensations, one is aware of them as public and external, not private.

Second, the assumed link between immediacy and certainty is questionable. If immediacy is put forward as an introspective characteristic of the awareness of sense data, nothing follows about its certainty because any awareness we point to as direct may be mistaken. However, if immediacy and certainty are linked conceptually, as the premise of the certainty argument suggestsif they are defined in terms of each otherthen it may be that what seems to be immediate, and hence certain, awareness is not immediate. This point may be illustrated in various ways. The certainty argument claims that sensing reveals existentsthat when we look at an (apparent) tomato, we cannot doubt that something red and round and bulgy exists. Strictly speaking, however, we are certain only of something red-looking; it may in fact be orange that looks red in this light. Indeed, as J. L. Austin pointed out, even statements about how a thing looks may have to be retracted. Further, the controversy over whether sense data can fail to appear as what they are throws further doubts on the incorrigibility claim, and the alternative adverbial analysis, that sense data are only sense contents, challenges the view that something exists distinct from the percipient's experience of it.

Third, the certainty argument is too ready to deny that we see physical objects in cases of illusion and distortion and to assume that we are aware of the same kind of existent in both perception and hallucination. Both these assumptions may plausibly be denied. When we look at the putative tomato, even if it is a piece of wax or a reflection of a tomato or an image on a screen, we are still seeing a material objectwax, or the tomato "in" (via) the mirror, or a screen illuminated in a certain way. There is no need to suppose that we are aware of something else, a sense datum. Contrastingly, the common explanation of hallucinations would be that they are unusually vivid mental images confused with perceptions. Such images, like afterimages, seem to be private, but one should not assume that they are identical with what we are aware of in normal perception. The sense-datum theory can, however, reply that hallucinations are normally quite indistinguishable by the victim from normal perception and may also be integrated with a perceived backgroundfor instance, the apparition may walk across the room and cast shadowsso if the hallucinatory images are private, so must be the data of the background. Although two entities are not necessarily identical because they are generally indistinguishable, identity may be the most plausible explanation of their indistinguishability, and the integration is very difficult to explain except on the sense-datum theory or on some form of representative realism. All the same, the sense-datum theory, if treated as an explanatory hypothesis, has the disadvantage of being very uneconomical in postulating so many distinct entities (the sense data).

the partitive argument

The partitive argument can be dismissed quite briefly, apart from its other troubles already mentioned. From the fact that we do not actually see the whole of an object at once, it does not follow that we do not then see the object, any more than the fact that we cannot visit all of New York at once means that we cannot visit it at all. Consequently, there is no ground for regarding what we actually see of an object as something different from it (a sense datum) or the actual seeing as some special direct awareness (sensing).

the argument from the content of illusions

The argument from the content of illusions presents problems similar to those of the certainty argument. The alternative to the sense-datum answer concerning what the drunkard sees in hallucinations is "a mental image," and in double vision "one bottle looking double." Neither answer is wholly satisfactory, since the first cannot explain the integration of the image with a real background, and the second has been accused of evading the issuelooking double is not like looking blue or looking elliptical, for it involves an extra apparent object, not a differing quality of the one object.

Ordinary cases of relativity are much more easily dealt with. When one sees a round dish that looks elliptical, one is simply seeing the dish and not some elliptical existent; the theory oddly assumes that things cannot look other than they are. This assumption is linked with the notion of immediacy: It is gratuitously supposed that in seeing the dish as elliptical, one is immediately aware of an elliptical existent. However, this begs the question by equating immediacy with incorrigibility, so that what looks elliptical is said to be elliptical. Furthermore, there is no cogent ground in experience or in the argument for supposing that nonhallucinatory sense data are private to a person: The elliptical shape of the plate or even the second bottle might also be sensed by others. The privacy is best supported by arguing that sense data are "generated" by brain processes (as in the causal argument for sensations).

other difficulties

Various other criticisms of sense data have been put forward, especially by Gilbert Ryle and J. L. Austin. First, sensing is either seeing under another namein which case there is the reduplication or regress noted concerning sensationsor else it is a myth. The notion of a mistake-proof awareness, Ryle claimed, arises from misunderstanding the character of perception words, which are achievement words or indicate the scoring of an investigational success. One cannot perceive unsuccessfully any more than one can win unsuccessfully, but that is a linguistic or conceptual matter; it does not mean that if one looks or plays, one is bound to see or to win.

Second, the theory, in speaking of sense data as existents, is simply reifying (treating as things) the sounds, smells, or looks of things. Ryle claimed a linguistic origin for this: By wrongly speaking of "seeing looks" or "smelling whiffs," which are pleonastic usages like "eating nibbles," the theory tends to treat looks and whiffs as the sort of things we can see or smellthat is, as objectsand fails to see that the point of such words is to show how we are perceiving objects. (He could hardly condemn hearing sounds, even if the other examples are correct.)

Third, Austin attacked the tendency of Moore, A. J. Ayer, and others to distinguish different senses of the word see : the normal sense (seeing objects) and the restricted "direct" or "actual" seeing (sensing, which is incorrigible). He claimed that the second sense is a myth: The basic fact is that one may describe the object one sees in various ways, depending on how advertent one is; for example, as a tomato or as a red object. But in these two cases it is the same thing described in different ways, not two different things; nor does it follow that there are two kinds of seeing or two senses of the word see.

Austin had other alleged linguistic grounds for the theory's mistakes, such as confusion of illusion with delusion, but it is doubtful whether the several different linguistic origins that he and Ryle claimed for the theory are really genuine and important. The reflective examination of experience seems a more likely origin for the theory, in view of the stress laid on it by the sense-datum philosophers. They have been so struck by the apparent immediacy of perceiving, by its apparently direct confrontation with a "given," that they have readily assumed that it does involve such an immediate awareness or confrontation; and because (on account of illusions) they cannot identify immediate awareness with the perception of a physical object, they have supposed it to be an inner awareness of special datathe sensing of sense data.

Further Developments

sense contents

As we have seen, the adverbial analysis of sensing claims that sensa no more exist as entities distinct from the sensing of them than do itches or pains; consequently, they are often referred to in this analysis as sense contents. Important advocates of this approach have been C. J. Ducasse and Ayer, and under its influence Moore modified and Russell abandoned an earlier faith in an actobject analysis (that sense data are separate entities distinct from the act of sensing). Russell's conversion to the adverbial analysis was brought about by his conclusion that the subject of awareness is a logical fiction; since the act-object analysis presupposes a subject of the act of awareness, it had to be dismissed (Analysis of Mind, p. 141). Probably few would follow him on this; it is, at any rate, not clear that the adverbial analysis can dispense with the subject, nor is it clear why one should wish to. Moore's Refutation of Idealism relied on the act-object analysis, but he later had doubts about this. He tended to see the problem as whether sense data have any existence when unperceived, or rather, unsensed; that is, whether their esse is percipi or sentiri. He regarded this as an open question, producing various arguments on either side at different times. Actually, the two questions are not quite identical: The adverbial analysis implies that sense data or sense contents cannot exist unsensed, while the act-object analysis is neutral on this.

It seems clear that whether sense data exist unsensed is not a question that can be settled by sensing them. Consequently, some would say it is a purely conceptual matter, one of how sense data are to be defined or how the general theory is to be framed. But factual issues are relevant and present a dilemma similar to the one of the relation of sense data to physical objects. If one accepts that sense data are generated by the brain, then it seems that they cannot exist unsensed. Even if they are only conditioned by the nervous system, they must appear different from what they really are in the unconditioned, unsensed state, thus undermining the certainty argument. At the same time, to say that a physical object is a family of sense data is scarcely meaningful if sense data do not exist unsensed; therefore, Russell at one period claimed that they do exist unsensed, calling them sensibilia in this state. More usually, however, phenomenalism is maintained; sense data do not exist unsensed, but possible ones or possibilities of sensation do.

So far as introspection is concerned, decision between the analyses depends on which sense is considered. Visual sense data, such as color shapes, would seem clearly to be distinct and to require an act-object analysis. (Afterimages are more doubtful, but anyhow are a special case.) Much the same applies to sounds and smells, which are normally experienced as external: By contrast, tactile and other bodily (somatic) sense data, such as pains or feelings of warmth or pressure, and the sensations of movement (kinesthetic data) seem clearly adverbial, as perhaps is taste; but there are marginal cases. Explanation of this variation is difficult for the theory, which would be more plausible if it could give one account of all sense data; it is also difficult to square the distinctness claimed in the act-object analysis with the privacy always claimed. Another possible line, which seems required for dreams and mental images and for hallucinations where no distinct objects are present, would be to say that while sense data seem to the person to be distinct, they are actually contents of adverbial experiences, as are sensations. However, this would undermine the claim of the theory to rely on introspective analysis.

sense-datum language

One suggestion that has been made is that the sense-datum philosophers have not, as they at first thought, produced a new theory of perception; they have simply introduced a new and more convenient terminology for discussing the facts of ordinary perceiving. This was accepted for a time by those who sought to see all philosophy as dealing with language and by those who, impressed by the difficulties the sense-datum theory encountered, sought to salvage something from the wreck. It is not popular now, for those with a linguistic bias have turned to the examination of ordinary language rather than to the advocacy of new terminologies, while the general decline of support for sense data has proceeded beyond this halfway house. Another reason for supposing that the sense-datum theory was only a terminology was the view that theories must be verifiable by observation of predicted consequences, which the sense-datum theory is not, but this seems to confuse a philosophical theory with a scientific theory.

Considered simply as a terminology, the language of sensing and sense data was claimed to have certain advantages; for example, that it is (a ) noncommittalone can describe the contents of one's experience independently of the physical objects they are thought to refer toand (b ) neater than ordinary language, for one can avoid periphrases like "there appears to be a red, bulgy tomato-like object" merely by listing the data sensed. But these are only slight advantages, and it seems that they are far outweighed by the fact that a sense-datum language cannot be truly neutral. It has been so long associated with the sense-datum theory that it must inevitably beg the question by suggesting that the data are private, transitory existents; that one is not "actually seeing" physical objects; or that in describing the scene in terms of visual and tactile data, one has described the experiences of normal perception and not of the different "reduced" phenomenological observation.

See also Perception; Phenomenalism; Realism.



Readers are advised to consult the historical surveys and then follow up any particular treatment that interests them with the aid of the references given. On the psychological side, see Edwin Garrigues Boring, Sensation and Perception in the History of Experimental Psychology (New York: Appleton-Century, 1942) and A History of Experimental Psychology, 2nd ed. (New York: Appleton-Century-Crofts, 1950); on the philosophical side, most useful is David Walter Hamlyn, Sensation and Perception (New York: Humanities Press, 1961). A stimulating but idiosyncratic discussion is given in Bertrand Russell, The Analysis of Mind (London: Allen and Unwin, 1921). Also, many of the books listed below under sense data discuss sensations. Gilbert Ryle's article "Sensations," in Contemporary British Philosophy, edited by Hywel David Lewis, 3rd series (London, 1956), attacks the whole concept; a careful study of a limited field is David Malet Armstrong, Bodily Sensations (London: Routledge and Paul, 1962).

sense data

The following contain statements of the sense-datum theory.

Pioneer work is found in George Edward Moore, Some Main Problems of Philosophy (introductory lectures of 19101911; London: Allen and Unwin, 1953) and his more difficult Philosophical Studies (London: Routledge, 1922). He kept returning to the problems of perception, not always consistently; for the advanced student, Alan Richard White, G. E. Moore (Oxford: Blackwell, 1958), is a useful guide with a full bibliography. Bertrand Russell was another pioneer in his lucid elementary Problems of Philosophy (London: Williams and Norgate, 1912) and his more significant Mysticism and Logic (articles of 1914; London: Allen and Unwin, 1918) and Our Knowledge of the External World (London: Allen and Unwin, 1914), but he tended to subordinate the topic of sense data to his special perspective theory. Clear and systematic is Henry Habberley Price, Perception (London: Methuen, 1932), which develops a full sense-datum theory. Another well-known account is Charlie Dunbar Broad, The Mind and Its Place in Nature (London: Kegan Paul, 1925); earlier and fuller statements of his views, with more attention to causal problems are in his Scientific Thought (London: Kegan Paul, 1923) and Perception, Physics and Reality (Cambridge, U.K.: Cambridge University Press, 1914). Alfred Jules Ayer gives several clear discussions, mainly from a phenomenalist point of view, in The Foundations of Empirical Knowledge (London: Macmillan, 1940), Philosophical Essays (London: Macmillan, 1954), and The Problem of Knowledge (London: Macmillan, 1956). For a modern version of representative realism stated in terms of sense data, see John Raymond Smythies, Analysis of Perception (London: Routledge and Paul, 1956).

There are many criticisms of the sense-datum theory. A general introductory survey, chiefly of Price's version, and extended criticisms are given in Rodney Julian Hirst, The Problems of Perception (London: Allen and Unwin, 1959). Gilbert Ryle's criticisms are found in his The Concept of Mind (London: Hutchinson, 1949) and Dilemmas (Cambridge, U.K.: Cambridge University Press, 1954); John Langshaw Austin's are in his lively defense of a commonsense approach, Sense and Sensibilia (London: Oxford University Press, 1962); Roderick M. Chisholm has some succinct criticisms in his more technical Perceiving: A Philosophical Study (Ithaca, NY: Cornell University Press, 1957); Harold Arthur Prichard, Knowledge and Perception, edited by W. D. Ross (Oxford: Clarendon Press, 1950), attacks Russell and the view that sensing is a form of knowing. Martin Lean, Sense Perception and Matter (London: Routledge and Paul, 1953), attacks Broad's version; and valuable is H. H. Price's "The Nature and Status of Sense-Data in Broad's Epistemology," in The Philosophy of C. D. Broad, edited by Paul Arthur Schilpp (New York: Tudor, 1959). On Moore, besides A. R. White's book, senior students should see P. A. Schilpp, ed., The Philosophy of G. E. Moore (Evanston, IL: Northwestern University Press, 1942), particularly C. J. Ducasse, "Moore's Refutation of Idealism"; O. K. Bouwsma, "Moore's Theory of Sense-Data"; and Paul Marhenke, "Moore's Analysis of Sense Perception"; and Moore's replies to them.

Among many critical articles, notable are Roderick Firth, "Sense-Data and the Percept Theory," in Mind 58 (232) (1949): 434465, and 59 (233) (1950): 3556, attacking the sense-datum analysis of perceptual consciousness; Winston H. F. Barnes, "The Myth of Sense-Data," in PAS 45 (19441945); and Anthony M. Quinton, "The Problem of Perception," in Mind 64 (253) (1955): 2851, criticizing the concept of sense data and the arguments for them; also a more advanced discussion of the earlier versions of the sense-datum theory in the symposia "The Status of Sense-Data," in PAS 14 (19131914): 355406, and "The Nature of Sensible Appearances," in PAS, Supp. 6 (1926): 142205.

sense-datum language

The linguistic interpretation of sense data is developed in George A. Paul, "Is There a Problem about Sense-Data?," in PAS, Supp. 15 (1936), and A. J. Ayer, "The Terminology of Sense-Data," in Mind 54 (216) (1945): 289312, reprinted in his Philosophical Essays (London: Macmillan, 1954); there is a useful discussion of it in the symposium "Seeming," in PAS, Supp. 26 (1952): 195252.

sense contents (adverbial analysis)

The most important advocate of this approach is C. J. Ducasse, Nature, Mind and Death (La Salle, IL: Open Court, 1951), largely identical with his contribution to The Philosophy of G. E. Moore. Brief statements may be found in A. J. Ayer, Language, Truth and Logic (London: Gollancz, 1946) and Philosophical Essays (London: Macmillan, 1954). A fuller discussion, with emphasis on the differences between the senses, is C. D. Broad, "Berkeley's Argument about Material Substances," in Proceedings of the British Academy 28 (1942): 127ff., and his Scientific Thought (London: Kegan Paul, 1923).

other recommended titles

Anscombe, G. E. M. "On Sensations of Position." Analysis 22 (1962): 5558.

Armstrong, David Malet. Perception and the Physical World. London: Routledge and Kegan Paul, 1961.

Casullo, Albert. "A Defense of Sense-Data." Philosophy and Phenomenological Research 48 (1987): 4561.

Chisholm, Roderick. The Foundations of Knowing. Minneapolis: University of Minnesota Press, 1982.

Chisholm, Roderick. "The Problem of the Speckled Hen." Mind 51 (1942): 368373.

Cornman, James, "Sellars, Scientific Realism, and Sensa." Review of Metaphysics 23 (1969): 417451.

Fales, Evan. A Defense of the Given. Lanham, MD: Rowman & Littlefield, 1996.

Fantl, Jeremy, and Robert Howell. "Sensations, Swatches, and Speckled Hens." Pacific Philosophical Quarterly 84 (2003): 371383.

Huemer, Michael. "Sense-Data." Stanford Encyclopedia of Philosophy. Available from

Jackson, Frank. Perception: A Representative Theory. Cambridge, U.K.: Cambridge University Press, 1977.

Lowe, E. J. "Indirect Perception and Sense Data." Philosophical Quarterly 31 (1981): 330342.

Martin, Michael G. F. "Beyond Dispute: Sense-Data, Intentionality, and the Mind Body Problem." In History of the Mind-Body Problem, edited by Tim Crane. New York: Routledge, 2000.

Perkins, M. Sensing the World. Indianapolis: Hackett, 1983.

Sellars, Wilfrid. "Phenomenalism." In his Science, Perception and Reality. London: Routledge, 1963.

Swartz, R. J., ed. Perceiving, Sensing, and Knowing. New York: Doubleday, 1965.

R. J. Hirst (1967)

Bibliography updated by Benjamin Fiedor (2005)