The epistemological interests of analytic philosophers in the first half of the 20th century were largely focused on the relationship between knowledge and perception. The major figures in that period were Russell, Moore, H.H. Price (1899–1984), C.D. Broad (1887–1971), Ayer, and H. Paul Grice (1913–88). Although their views differed considerably, all of them were advocates of a general doctrine known as sense-data theory.

The technical term sense-data is sometimes explained by means of examples. If one is hallucinating and sees pink rats, one is having a certain visual sensation of rats of a certain colour, though there are no real rats present. The sensation is what is called a “sense-datum.” The image one sees with one’s eyes closed after looking fixedly at a bright light (an afterimage) is another example. Even in cases of normal vision, however, one can be said to be apprehending sense-data. For instance, when one looks at a round penny from a certain angle, the penny will seem to have an elliptical shape. In such a case, there is an elliptical sense-datum in one’s visual field, though the penny itself continues to be round. The last example was held by Broad, Price, and Moore to be particularly important, for it seems to make a strong case for holding that one always perceives sense-data, whether one’s perception is normal or abnormal.

In each of those examples, according to defenders of sense-data theory, there is something of which one is “directly” aware, meaning that one’s awareness of it is immediate and does not depend on any inference or judgment. A sense-datum is thus frequently defined as an object of direct perception. According to Broad, Price, and Ayer, sense-data differ from physical objects in that they always have the properties they appear to have; i.e., they cannot appear to have properties they do not really have. The problem for the philosopher who accepts sense-data is then to show how, on the basis of such private sensations, one can be justified in believing that there are physical objects that exist independently of one’s perceptions. Russell in particular tried to show, in such works as The Problems of Philosophy (1912) and Our Knowledge of the External World (1914), that knowledge of the external world could be logically constructed out of sense-data.

Sense-data theory was criticized by proponents of the so-called theory of appearing, who claimed that the arguments for the existence of sense-data are invalid. From the fact that a penny looks elliptical from a certain perspective, they objected, it does not follow that there must exist a separate entity, distinct from the penny itself, that has the property of being elliptical. To assume that it does is simply to misunderstand how common perceptual situations are described. The most powerful such attack on sense-data theory was presented by Austin in his posthumously published lectures Sense and Sensibilia (1962).

The theory of appearing was in turn rejected by many philosophers, who held that it failed to provide an adequate account of the epistemological status of illusions and other visual anomalies. The aim of those thinkers was to give a coherent account of how knowledge is possible given the existence of sense-data and the possibility of perceptual error. The two main types of theories they developed are realism and phenomenalism.

Realism

Realism is both an epistemological and a metaphysical doctrine. In its epistemological aspect, realism claims that at least some of the objects apprehended through perception are “public” rather than “private.” In its metaphysical aspect, realism holds that at least some objects of perception exist independently of the mind. It is especially the second of those principles that distinguishes realists from phenomenalists.

Realists believe that an intuitive, commonsense distinction can be made between two classes of entities perceived by human beings. One class, typically called “mental,” consists of things like headaches, thoughts, pains, and desires. The other class, typically called “physical,” consists of things such as tables, rocks, planets, human beings, and animals and certain physical phenomena such as rainbows, lightning, and shadows. According to realist epistemology, mental entities are private in the sense that each of them is apprehensible by one person only. Although more than one person can have a headache or feel pain, for example, no two people can have the very same headache or feel the very same pain. In contrast, physical objects are public: more than one person can see or touch the same chair.

Realists also believe that whereas physical objects are mind-independent, mental objects are not. To say that an object is mind-independent is just to say that its existence does not depend on its being perceived or experienced by anyone. Thus, whether or not a particular table is being seen or touched by someone has no effect upon its existence. Even if no one is perceiving it, it still exists (other things being equal). But this is not true of the mental. According to realists, if no one is having a headache, then it does not make sense to say that a headache exists. A headache is thus mind-dependent in a way in which tables, rocks, and shadows are not.

Traditional realist theories of knowledge thus begin by assuming the public-private distinction, and most realists also assume that one does not have to prove the existence of mental phenomena. Each person is directly aware of such things, and there is no special “problem” about their existence. But that is not true of physical phenomena. As the existence of visual aberrations, illusions, and other anomalies shows, one cannot be sure that in any perceptual situation one is apprehending physical objects. All that people can be sure of is that they are aware of something, an appearance of some sort—say, of a bent stick in water. Whether that appearance corresponds to anything actually existing in the external world is an open question.

In his work Foundations of Empirical Knowledge (1940), Ayer called the difficulty “the egocentric predicament.” When a person looks at what he thinks is a physical object, such as a chair, what he is directly apprehending is a sense-datum, a certain visual appearance. But such an appearance seems to be private to that person; it seems to be something mental and not publicly accessible. What, then, justifies the individual’s belief in the existence of supposedly external objects—i.e., physical entities that are public and exist independently of the mind? Realists developed two main responses to the challenge: direct (or “naive”) realism and representative realism, also called the “causal theory.”

In contrast to traditional realism, direct realism holds that physical objects themselves are perceived “directly.” That is, what one immediately perceives is the physical object itself (or a part of it). Thus, there is no problem about inferring the existence of such objects from the contents of one’s perception. Some direct realists, such as Moore and his followers, continued to accept the existence of sense-data, but, unlike traditional realists, they held that, rather than mental entities, sense-data might be physical parts of the surface of the perceived object itself. Other direct realists, such as the perceptual psychologist James J. Gibson (1904–79), rejected sense-data theory altogether, claiming that the surfaces of physical objects are normally directly observed. Thompson Clarke (1928–2012) went beyond Moore in arguing that normally the entire physical object, rather than only its surface, is perceived directly.

All such views have trouble explaining perceptual anomalies. Indeed, it was because of such difficulties that Moore, in his last published paper, “Visual Sense-Data” (1957), abandoned direct realism. He held that because the elliptical sense-datum one perceives when one looks at a round coin cannot be identical with the coin’s circular surface, one cannot be seeing the coin directly. Hence, one cannot have direct knowledge of physical objects.

Although developed in response to the failure of direct realism, the theory of representative realism is in essence an old view; its best-known exponent in modern philosophy was Locke. It is also sometimes called “the scientific theory” because it seems to be supported by findings in optics and physics. Like most forms of realism, representative realism holds that the direct objects of perception are sense-data (or their equivalents). What it adds is a scientifically grounded causal account of the origin of sense-data in the stimulation of sense organs and the operation of the central nervous system. Thus, the theory would explain visual sense-data as follows. Light is reflected from an opaque surface, traverses an intervening space, and, if certain standard conditions are met, strikes the retina, where it activates a series of nerve cells, including the rods and cones, the bipolar cells, and the ganglion cells of the optic nerve, eventually resulting in an event in the brain consisting of the experience of a visual sense-datum—i.e., “seeing.”

Given an appropriate normal causal connection between the original external object and the sense-datum, representative realists assert that the sense-datum will accurately represent the object as it really is. Visual illusion is explained in various ways but usually as the result of some anomaly in the causal chain that gives rise to distortions and other types of aberrant visual phenomena.

Representative realism is thus a theory of indirect perception, because it holds that human observers are directly aware of sense-data and only indirectly aware of the physical objects that cause those data in the brain. The difficulty with representative realism is that since people cannot compare the sense-datum that is directly perceived with the original object, they can never be sure that the former gives an accurate representation of the latter, and, therefore, they cannot know whether the real world corresponds to their perceptions. They are still confined within the circle of appearance after all. It thus seems that neither version of realism satisfactorily solves the problem with which it began.