Humans can't help dividing up the world into types. After all, it would be incredibly inconvenient to be only able to talk about particular things and not the types to which they belong: why spend hours saying 'I like Fido, and Rex, and Spot ...' when I can say 'I like dogs' in a second or two? Regrouping particular things in conceptual 'boxes' allows us to make generalizations about such things, like 'dogs bark', which in turn enables us to predict future events ('the new neighbors have a dog, so we should expect to hear barking') and to draw probabilistic and explanatory inferences ('Spot keeps barking, he must be a dog', 'Rex barks because he is a dog').
This 'carving up' of the world has always been of great interest to philosophers. As we have seen, many philosophers take our categorizations to be somehow reflective of the world's real structure. The world, on this picture, is already divided up into real groups of things that resemble each other, and our categories simply correspond to such groups. Philosophers in this camp then propose explanations, like realism and trope nominalism, as to why the world is like that.
There are those, however, who dismiss the picture of a pre-divided reality. For them, types are useful ways of classifying particular things, and nothing else. The world has no preexisting 'architecture' that our concepts may or may not delineate. In fact, a completely different way of classifying things would be no more 'real' or 'accurate' than our way. A classificatory system that included the type 'shdog', where something is shdog if it is either a dog or a tree may be far less convenient than one that includes 'dog' and 'tree', but it wouldn't be any less real. On this view, types are very much like the constellations navigators used to find their way around the sea: constellations like Ursa Major are useful ways of mentally grouping individual stars, but at the end of the day, only the stars are real (Nanay 2007). Likewise, for philosophers in this second camp, types usefully regroup the individual things that come under them – their respective tokens – but only the tokens are mind-independently real. Rex, Fido and Spot are real enough, but ‘dog’ isn’t.
Philosophical positions that embrace a picture of this second sort abound: conceptualism, predicate nominalism, conceptual relativism (Putnam 1999), irrealism (Goodman 1975), singularist semirealism (Nanay 2007). For our purposes, I will call the rejection of types as mind-independently real ‘type-antirealism’. In my experience, it is popular among college students, particularly in the arts and humanities, who often talk of types as ‘social constructs’.
Undeniably, some types are mere constructs – ‘shdog’ is one of them, and I’m sure you can think of others. However, I’m going to briefly argue that dismissing all types as mere constructs may come at a high cost.
Take the type ‘dog’. The type-antirealist tells us that ‘dog’ is just one way of classifying individual things. We may as well have used the type ‘shdog’ which, once again, regroups all and only dogs and trees. ‘Dog’ is more useful than ‘shdog’, but there is no reason to consider it as any more real.
Notice that the reality of ‘dog’ has been denied on the basis that there are other types which, while less useful, are surely (we are told) no less real. But if this is an acceptable way of denying something’s mind-independent reality, then why stop at types?
Take, for instance, a token of ‘dog’, say, Spot. Spot seems about as real as anything. But ‘Spot’, like the other objects of everyday life, is just one way of mentally grouping bits of matter together. Consider instead ‘Spot*’, the composite object consisting of the dog Spot and my mobile phone. Sure, ‘Spot*’ is a much more inconvenient way of regrouping bits of matter than ‘Spot’. But why think that Spot* is any less real than Spot?
The natural reaction here is to say that Spot clearly is more real than Spot*. But it isn’t clear how we can argue for this without having to admit that, equally, ‘dog’ is clearly more real than ‘shdog’, on pain of inconsistency. For instance, we may say that Spot is real because the existence of Spot has plenty of predictive and explanatory power: Spot’s existence explains why I hear barking when I walk by the neighbour’s house, and allows me to predict that my hamster will die very quickly if he escapes to the neighbour’s garden. In contrast, the existence of Spot* doesn’t explain or predict anything that the existence of Spot and of my mobile phone don’t already explain or predict. We infer from this, using some version of Ockham’s razor, that Spot is real and Spot* isn’t. But if we infer this, we must also (again, on pain of inconsistency) infer the reality of ‘dog’ over ‘shdog’. We have already seen that ‘dog’ allows us to explain and predict quite a lot, and ‘shdog’ wouldn’t do nearly as well on that front.
At this point, the type-antirealist may bite the bullet and claim that Spot isn’t real either. All that is real, he could say, is mind-independent matter, and both types and tokens are just convenient ways of carving up matter.
Unfortunately, things may get worse for the type-antirealist. Take ‘matter’, the spatially extended stuff out of which the concrete things around us are supposed to be made. We generally think of our experiences as sometimes being of such external, mind-independent, material stuff, and other times as being illusory, as in dreams and hallucinations. This is very useful, because it allows us to predict and explain our experiences: the fact that my experience of getting my hand chopped off was only a nightmare explains why I don’t have an experience of lacking a hand the next day, and the fact that my experience of writing this article is not illusory allows me to predict a future experience of seeing the article published on my blog.
But there are other ways of interpreting experience. I could think that none of my experiences are experiences of mind-independent matter. Or I could think that those I would normally regard as illusory are not, and vice versa.
But as we have seen, the type-antirealist must hold that the explanatory and predictive usefulness of positing a thing’s reality does not warrant the belief that such thing is actually real – otherwise, he would have to admit that types like ‘dog’ are real since such types are explanatorily and predictively useful. Hence it seems that the consistent type-antirealist must also regard mind-independent matter as nothing more than a useful tool for organizing, interpreting and predicting experience.
To sum up, the type-antirealist, by disallowing the inference from explanatory and predictive usefulness to reality, ends up having to deny the reality, not only of types like ‘dog’, but also of individual things like Spot, and, finally, of the external world itself. What was intended as a relatively mild, sophisticated scepticism lands us in a scepticism of the most radical sort.
Perhaps the type-antirealist can find reasons to be sceptical about types that aren’t also reasons to be sceptical about individual things, or mind-independent matter. But I have yet to see such reasons.
So, what is the danger of believing that types aren’t real? It is that, if one is consistent, one may end up believing very little indeed.