It is often claimed that we can’t really describe smells, or that our ability to do so is severely limited. This is taken to show a limitation of our language, a limitation that can’t be overcome. What is the source of this idea?
Clearly, we do describe smells sometimes. A client may tell a carpenter about the way his house smells. Say, there is a putrid smell, or a sharp, stinging smell in the basement, or it smells like old wet rags, etc, and the carpenter may tell him that there’s nothing to worry about, or maybe that the house suffers from dry rot or mildew or what have you. Or a wine connoisseur might describe a wine to a colleague, and his colleague might be able to tell what type of wine it was – or they might disagree about the aroma of some particular wine, etc.
“But here we are not really describing the smell itself, simply saying what the smell is like. We’re not bringing the smell itself into words.” It will also be said that what we miss is a smell vocabulary. Now strictly speaking this isn’t true, there’s a whole range of words for describing smells (as illustrated here, for instance). It is true, though, that our smell vocabulary is very generic, and that we often try to be more specific either by referring to the source of the smell (it smells of fish) or by comparing it to a smell identified by its source (it smells like fish).
But why should we think that in referring to the source or in invoking a similarity to a smell coming from some source we are not describing the smell itself? What would it be like to “describe the smell itself”? I should like to suggest that our thinking here is under the spell of colour vocabulary. Colour words, we think, are a paradigm of what it means to bring our very sensations into words. Colour words mention only the visible property itself, independently of the object carrying the property. The colour red is an abstract property compared to the smell of fish, for instance.
Colour words, however, are really a special case. Quite apart from the fact that the application of colour words is not in reality wholly object independent (consider the difference between establishing that a book is red, that an apple is red, that the sky is red, and that someone is red in the face), colour, it seems, has no direct counterpart when it comes to other perceptual properties. Our vocabulary for sound or texture or even shape (felt or seen) is quite limited, and most of the time we describe the perceptual qualities of an object by referring to the kind of object it is (I felt that it was silk) or is like (it felt like silk). Where shapes are concerned for instance, we have recourse to a few schematic concepts, which do not apply to most natural objects, and when they do apply, it is usually just by approximation. (We use the alphabet for recording spoken sounds, and we have musical notation for musical sounds, but there is no general abstract vocabulary for sounds.)
The relatively abstract nature of colour vocabulary, I would imagine, is ultimately grounded in our practices of painting or colouring objects; for instance in the fact that we can use the same paint for a great many different kinds of object.
Colours, in fact, provide a misleading paradigm of what it means to describe perceptual qualities, leading us to say that smells “can’t really be described”.