§64. The Sense-Data Theory

In the early twentieth century Bertrand Russell made the sense-data theory prominent in British empiricism. He adapts the theory from G. E. Moore, who lectured on it in 1910. Moore’s argument was that our most primitive knowledge of material objects is by means of the senses.

The color, the shape—these are things that he sees, qualities given to vi­sion. He explains that sense data exist only in their apprehension, are pri­vate, different for each perceiving subject, and never occur in the same space with sense data apprehended by others. We are cautioned to distinguish sense data from sensation. The whitish patch is one thing and Moore’s sensa­tion is another. Sense data are qualitative entities, a quasi-Epicurean eidola; their apprehension, when they strike us, is a sensation. This is the abstrac­tion Berkeley, Hodgson, Bradley, and Oakeshott denounce as a disguised contradiction, as if sensory qualities might exist without any sensing. Moore says that sense-data apprehension is direct, which he explains by contrast to memory.

How could we know this material reality? Russell supplies the explana­tion Moore withheld: sense data are signs. We come to an indirect apprehen­sion of material objects by the direct apprehension of indicative sense-data signs, “a sign of something existing independently of us, something differing, perhaps completely, from our sense-data, and yet to be regarded as causing those sense-data.” What Russell seems to take from Moore's theory is that knowledge of bodies is inferred from immediate knowledge of sense data, which he regards as physical because they depend on the body and are not located in the mind, that having been Berkeley's mistake. Like Plato he has the perceiving take place in the soul but locates the datum itself in the body, explaining that what we are directly aware of when, for instance, we see color is an event in the nervous system.¹³⁴

Writing on the relation of sense data to physics in 1914, Russell formulates “the supreme maxim in scientific philosophizing,” which is that “wherever possible, logical constructions are to be substituted for inferred entities.” That explains his interest in Moore's sense data—they fit his substitution pro­ject, a project Moore did not share. Sense data are perfect for logical con­struction because they exist if and only if we sense them, so there is never a problem of empty reference. Physical objects are equated with sets of sense data, bringing these objects closer to the empirical foundation of knowledge, a boon for scientific rationality, and sense data are the foundation of lin­guistic meaning no less than knowledge. Every name and predicate has to analyze into a logical expression entirely defined in terms of sense data.

Sense data also made a good fit with Russell's acquaintance principle.

The sense-data theory implies that in cases of illusion, when, say, an oar in water appears broken, something really is just as we perceive it—not the oar but the qualitative mosaic of sense data. Even when things are not as we see them, what we directly, immediately see is just as we see it to be. Russell's proposal to reduce material bodies to sets of sense data ran into all kinds of technical difficulties, but even if they were overcome and the theory proved consistent, one might ask, so what? “Admittedly, physical objects have been dispensed with,” a critic observes, “but each physical object has been replaced by a whole gamut of inferred entities, so the replacement looks, if anything, harder to know than what it replaces. The overall effect of [logical] construc­tion is to increase, rather than diminish, the number of inferred entities.”¹³⁶

The semantic relations Russell promised between sense data, bodies, and their names also proved impossible to define. It was unexpectedly difficult to spell out the content of even simple claims (“The cat is on the mat”) in terms of sense data. The problem seems to lie with the principle of acquaintance, which may explain why Russell gave it up. He assumes that it takes affec­tion by entities directly apprehended in sense to establish the link between language and thought. Wittgenstein demolished the idea in his Philosophical Investigations, and it now seems acknowledged that “experience does not have the capacity to establish, as if by some mental magic, a connection be­tween a name and an object.”¹³⁷

Sense data and the idea of constructing objects from them enjoyed what another unimpressed critic describes as “surprising respectability among eminent philosophers in the first half of the twentieth century. The role of sense-data in philosophy was so central that the ontological status of physical bodies was constantly in question.” Then postwar anglophone philosophy almost unanimously repudiated sense data. They are subjective, invoking private episodes; and atomistic, supposedly supplying representation with simple starting points, overlooking contextual stage-s etting and supple­mentary psychological capacities. Thus Wittgenstein, Strawson, Sellars, and Quine.¹³⁸