In asking where we get our idea of substances, Locke finds himself in one of the stickier sections of the Essay. He gives us the following picture of the origin of our ideas of substances: As we go through the world we carve up the dense sensory array into discrete objects, noticing which qualities regularly seem to cluster together. For instance, we see softness, blackness, a certain small size, a certain catlike shape moving all together throughout our experience, and we assume that all of these qualities make up a single object. However, he claims, this cluster of our ideas of observable qualities cannot in itself form the idea of a substance. We must also add to this an idea of whatever it is that these properties belong to; we do not simply believe that these properties exist out in the world, but rather that they are properties of something. That something, he argues, corresponds to our idea of substance in general or substratum. It is helpful to think of a substratum as an invisible pincushion, with all of the observable qualities that belong to it being the pins. The substratum itself is unobservable (and, hence, because of Locke's empiricism, unknowable) because it cannot itself have observable qualities; it is the thing in which observable qualities inhere. Anything we can observe or describe is a property rather than the substratum itself. Our idea of the substratum, therefore, is necessarily very obscure and confused. All we really know about the substratum is that it is supposed to support the observable properties of a substance. Beyond that, we have no hint and no hope of getting a hint. Locke is very eager to point out that the case is equally bleak for both mental and physical substances. Contrary to what most people believe, he argues, we do not know bodies any better than we know the mind. In both cases, we can only know the observable qualities. When it comes to what the properties belong to, we are completely in the dark in both cases. When he is being particularly careful. he remembers to point out that, really, since all we know are observable properties, there is no basis for even claiming that there are two different types of substance in the world. For the most part, however, he talks as if dualism were true (that is, as if mind and body were two distinct kinds of substances). In addition to treating the logico-linguistic problem of substances (i.e. What is metaphysically responsible for supporting properties? How can we make sense of the way we speak about them?), Locke also briefly touches on the scientific problem of substances: What is causally responsible for properties? The cause of properties, he claims, is the constitution of objects, their hidden microstructures. He treats this idea in greater depth in Book III.
Locke's discussion of substratum is probably one of the most confusing sections of the Essay, in large part because he himself is so obviously torn on the topic. In several instances, Locke uses language that would suggest he does not really believe substrata exist, that our idea of substratum refers to nothing and thus is meaningless. For instance, at I.iv.18 he says that we "signify nothing by the word "substance," but only an uncertain supposition of we know not what." At II.xxiii.18 he calls it a "promiscuous use of a doubtful term." Perhaps most provocatively, at II.xxiii.2 he compares the idea of a substratum to the explanatory tool of an Indian philosopher who, "saying that the world was supported by a great elephant, was asked what the elephant rested on, to which his answer was a great tortoise. Being again pressed to know what gave support to the broad-backed tortoise, he replied, something he knew not what." This mocking analogy seems to suggest that Locke considers "substratum" an entirely empty word, referring to nothing but our own limit of understanding. At the same time, Locke retains the idea in his picture. Given that one of his primary aims in the Essay is to encourage us to banish terms with no real meaning--terms that are supposed to refer to something in the world but do not or that have no clear ideas associated with them--his retention of this term is puzzling. Clearly, as suspicious as he was of the idea, he felt that it was necessary, though whether it is necessary only as a conceptual tool to make sense of our experience (as it would seem from the quotes above) or as something that must exist to make sense of the natural world itself (which he seems to suggest throughout the rest of the discussion) is really not clear. There are at least four reasons why Locke felt that it was crucial to include the notion of substratum in his account. First, he felt that the idea was needed in order to make sense of our language. If someone asks what a ladybug is, the answer would take the following form: "It is a thing that is black and red, with such and such a size and shape, that eats such and such..." There are qualities out in the world that correspond to the predicates in this sentence (even if the correspondence is not one of resemblance), so, Locke feels, there must also be something corresponding to the subject, the "thing." Not everyone in the history of philosophy felt this way. Some people, such as David Hume, felt that "thing" was just a peculiarity of how language works, a linguistic hanger on which we can hang qualities. In the world, however, there are only the qualities. When we say a thing that "is..." we do not really mean there is a thing that has these qualities, but simply that these qualities are the identity of the substance in question. This view is called the "bundle theory" of substances, because it regards substances as mere collections of observable properties. There are good reasons, though, why Locke did not want to go in this direction. This theory raises tremendous problems for itself. The biggest problem is the question of persistence through change. If a school bus just is a collection of yellow color, an oblong shape, powers of motion etc., what happens if I paint the school bus green, or if it breaks down and loses its powers of motion? If we have a new bundle of qualities, does that mean we have a new substance? The bundle theorist needs to come up with a good explanation of how the substance remains the same when the bundle changes. Persistence through change on Locke's view, however, is easy to account for, which is the second reason why he felt he needed to keep the notion of substratum. The substratum persists through any change. The substance, therefore, remains the same substance through changes in properties. A third reason Locke felt compelled to accept the notion of substratum was to explain what unifies co-occurring ideas, making them into a single thing, distinct from any other thing. The substratum, Locke claims at II.xxiii.1 and 37, helps elucidate this unity. It is not entirely clear, though, how the substratum is supposed to do this. Lastly, the substratum provides Locke with a way to account for the notion of support. The very idea of a quality involves dependence, being a quality of something. So what are qualities dependant on, what do they exist in? The answer, of course, is the substratum.
It is these considerations that push Locke to reluctantly embrace a notion that he himself admits may well be utterly meaningless.