What is it exactly that general terms signify, Locke asks in Chapter iii, section 12? It cannot be particulars, because then these terms would not be general. It also cannot be a plurality, because then there would be no difference between saying "cat" and saying "cats." What general terms signify, he concludes, are sorts.
With his analysis of sorts, Locke is heavily on the offensive. It is in trying to account for sorts that the Scholastics make their biggest mess, Locke feels. In order to define a sort they introduce the vague notion of an "essence," though none of them can say what essence really is. The definition and account of essences was one of the most hotly debated topics in natural philosophy for centuries, and still dominated the schools during Locke's lifetime. In sum, Locke is saying, "Look what a complex and incoherent story you told about a simple, simple thing." What makes something belong to a sort? There is no great mystery. Something belongs to a sort if it is classified under the abstract general idea that defines that sort. Essences, that great puzzle of the centuries, are simply abstract general ideas. In saying that it is abstract general ideas that determine sorts, Locke is making an important claim about natural kinds. We are the ones who create general ideas, so we are the ones who create sorts. Therefore, the world is not presorted into natural kinds that we discover. Rather, there is only a continuum of varying particulars, and we sort them ourselves into conventional kinds. As evidence for this claim, Locke points to "monsters," humans and animals that are born horribly deformed. These creatures do not fit into any one sort, which shows that nature itself is continuous and that it is only human beings who impose boundaries on it. Locke does not believe, however, that the sorts we create are arbitrary. We base our abstract general ideas on the observable properties we see, and these observable properties are really in the world. The differences really do exist, but we must decide which are relevant to our classifications. The final decision about how to carve things up is based on convention. Locke also introduces a further notion, the notion of real essences as opposed to nominal essences. By "nominal essence," Locke means that collection of observable properties that comprises our abstract general idea. In addition to observable properties, though, there is also another aspect to nature, the unobservable microstructure that gives rise to the observable properties. Locke denotes this internal constitution of objects as their "real essence." Unlike the nominal essence, the real essence has a basis in reality.
Given that Locke believes there is this real essence to things, it seems as if he should be able to give an account of sorts that allows for natural kinds. Why can he not say that sorts refer to these real essences, and thus have their basis in reality? Why does he insist on basing sorts on mind-created nominal essences, thus ensuring that there are no natural kinds? There are several levels on which to answer this question. First, Locke points out that we have had our classification of sorts long before we knew about the internal constitution of things. Our current sorts certainly cannot refer to these real essences, then. In addition, we still do not know the real essences of things. The real essences are in the unobservable microstructure. Our words, then, could not refer to them even if we wanted them to. Finally, even if we could discover the internal constitution of things (which Locke allows we might be able to do as soon as our microscopes become powerful enough), there is still a much larger problem, making it entirely impossible for us to use real essences to pick out natural kinds: Real essences are themselves determined by nominal essences. Each microstructure gives rise to a whole myriad of observable qualities, and it is on the basis of these observable qualities that we form our idea of each particular substance. In turn, on the basis of these ideas of particular substances, we form of our abstract general idea, or the nominal essence, by the method of abstraction. In order to determine which part of the microstructure counts as the real essence, we trace back the properties that make up the nominal essence to their causes in the microstructure. The real essence is just that subset of the real, internal constitution of things that determines the observable properties that go into the nominal essence. While microstructures occur naturally in the world, real essences do not. It is clear, then, that there is no chance of natural kinds on Locke's view. Despite this fact, Locke himself seems somewhat prone to talking as if there were. First of all, he often talks as if our ideas get better with time and science (as if, for instance, we get clearer on what gold really is the more we discover about its observable properties). Given Locke's view of sorts, though, it is impossible for ideas to get better over time (as Locke himself clearly states). Every time we discover something new about gold, we have a different idea, not a better one. We cannot get a richer understanding of gold, because there is no naturally occurring thing "gold" out in the world. There are only particular pieces of matter that we have chosen to classify together as a single kind of thing, called "gold." If we discover, for instance, that among the things that all look like what we call "gold," some of them have a different weight, we have not really discovered that these are not gold. We have only discovered that among pieces of matter that we happen to class together, there is a difference that we were not previously aware of. Whether we want to keep calling all these pieces of matter "gold" or call some of them "fool's gold" is entirely up to us. If we ignore the difference, they remain of the same sort or species. If we decide that the difference is salient and that we will attend to it in the future, then they become two separate sorts. Locke himself is the one who presents this example as an illustration of the consequences of his view. Still, he often talks as if it were not the case. It is not clear if he is just torn and confused or if he wants to find some way to reconcile his conventionalist theory of sorts with the possibility of deeper knowledge about what is in the world.
Another discussion that suggests that Locke is torn on the issue of natural kinds takes place in Book II and involves the relative adequacy of mixed modes as opposed to substances. There, he highlights the fact that our ideas of mixed modes cannot be inadequate because we create the archetypes that they are supposed to match. Substances, however, can be inadequate because the archetypes are out in the world. This makes it sound as if our ideas of substances were trying to match some actual essence out in the world. Locke might be able to get himself off the hook in this case, by objecting that in that discussion he was referring only to particular substances, not to sorts. Still, he does not draw this distinction in the discussion itself.