From the ALWS archives: A selection of papers from the International Wittgenstein Symposia in Kirchberg am Wechsel

Introduction

One answer to the question why the later Wittgenstein was so interested in the notion of expression is by now a familiar part of the philosophical landscape. The answer goes: “Through the notion of expression Wittgenstein wanted to break with a picture of language-use that focuses exclusively on assertion. That break provides an avenue to the solution of otherwise intractable philosophical problems such as the status of moral discourse and first person statements about mental states.” That this is indeed an important theme of PI won’t be contested in this paper; I think there is an important insight in this answer. I will, however, develop another answer to the same question, an answer that picks up a strand in PI that has not received due attention.

The answer that I would like to suggest in this essay is that the notion of expression helps to block a certain philosophically unfruitful way of thinking about the epistemology of other minds. Taking a well-known passage from PI as my starting point, I will show how Wittgenstein’s use of the notion of expression can be brought to bear on the idea that knowledge of other minds is inferential.

I. Wittgenstein on Expression

Here’s paragraph 244 of PI:

How do words refer to sensations?—There doesn’t seem to be any problem here; don’t we talk about sensations every day and give them names? But how is the connection between the name and the thing named set up? This question is the same as: how does a human being learn the meaning of the names of sensations?—of the word “pain” for example? Here’s one possibility: words are connected with the primitive, the natural, expressions of the sensation and used in their place. A child has hurt himself and cries; and then adults talk to him and teach him exclamations and, later, sentences. They teach the child new pain behaviour. (Wittgenstein, 2001, p. 75.)

In the paragraph immediately preceding this passage, one of Wittgenstein’s interlocutors asked us to consider a language with terms used only to refer to “what can only be known to the person speaking, to his immediate private sensations.” (Wittgenstein, 2001, p. 75.) The reason why the interlocutor asks us to consider this language is that he is captivated by the thought that our language somehow works in that way. He’s drawn to thinking that the words he uses to talk about his own inner states are such that they have (perhaps in addition to a publicly available meaning) a private meaning only accessible to him through his awareness of his own inner sensations.

In paragraph 244, Wittgenstein in effect refuses to play along with the way of talking about mental privacy that the interlocutor is trying to introduce. The interlocutor, presumably, would have his own way of giving an answer the question how words refer to sensations. His response would involve taking introspection to be the means by which a person attaches a meaning to a sensation term. But instead of letting the interlocutor formulate such answer, Wittgenstein suggests that learning the term ‘pain’ involves coming to use it to express pain in a new way.

Wittgenstein’s proposed quasi-empirical hypothesis about how children are taught the words for sensations is that the process is based on the fact that we express our inner states. The force of the notion of expression is that expression makes the inner knowable to other people. If we take the idea of expression seriously (and unless something more is said on behalf of the interlocutor’s conception, there is no reason why we shouldn’t) we can say that other people can perfectly well know when others, including children and animals, are in pain, or hungry, tired, happy, etc. They are states and emotions with characteristic expressions that make manifest what is going on with the person. By reminding us of the notion of expression Wittgenstein is drawing our attention to the fact the interlocutor’s conception of privacy is at odds with our ordinary supposition that we can know, and do know a lot of the time, the minds of others. Other people can know when the child is in pain, because the pain is expressed, and so there is nothing particularly problematic about teaching the child to use the word ‘pain’.

II. The Inferential Model and Skepticism About Expression

The moral of the preceding section, it seems, is that expression affords knowledge of other minds, and so blocks a conception of mental privacy according to which we cannot know the thoughts and feelings of others. Such was the interlocutor’s proposal: sensations are known only to the person having them. Since the interlocutor is denying that we know a range of facts we normally take ourselves to know, we can straightforwardly characterize his position as a form of skepticism. But what philosopher, nowadays, holds such a view about our knowledge of other minds? In this section I will argue that the notion of expression, when properly thought through, does more than prevent the interlocutor’s pure form of skepticism about knowledge of other minds. It also calls into question one particular and quite common view about what kind of knowledge we can have of other minds. This conception is one in which our knowledge of other minds is essentially inferential.

In arguing for the idea that the inferential model amounts to a form of skepticism about expression, I’m using the term skepticism as a term of criticism in a fairly non-standard way, although not in a sense without precedence.¹ I will in effect be claiming that even a philosopher who takes herself to be preoccupied precisely with giving an account of the kind of knowledge expression affords, can rightly be called a skeptic about expression. She is subject to that criticism on my view, if she is driven by certain philosophical considerations to give an account of expression that fails to reconstruct anything recognizable as the concept as we know and use it.

The central idea behind the inferential model is that even in the paradigmatic cases of knowing another mind, that knowledge is based on an inference from some publicly available fact to a mental state. Here’s Paul Churchland’s formulation of this thought:

It is of course by observing a creature’s behavior, including its verbal behavior, that we judge it to be a conscious, thinking creature—to be ’another mind’. From bodily damage and moaning, we infer pain. From smiles and laughter, we infer joy. From the dodging of a snowball, we infer perception. (Churchland, 1988, p. 67.)

The idea here is that inference is a must when it comes to other minds, there is no other way of accessing mental facts than inferring them from outer behaviour. Peter Singer, interestingly, provides a similar formulation of this idea in his plea for animal rights in Animal Liberation:

[Pain] is a state of consciousness, a ‘mental event’ and as such it can never be observed. Behaviour, like writhing, screaming, or drawing one’s hand away from the lighted cigarette is not pain itself; nor are the recordings a neurologist might make of activity within the brain observations of the pain itself. Pain is something we feel, and we can only infer that others are feeling it from various external indications. (Singer, 1990, p. 10.)

Now, neither Churchland nor Singer seems to be denying that there are expressions; indeed they talk about smiles and winces. Why not think that the inferential model precisely gives an account of how expression affords knowledge? In order to see how different the notion of expression that is deployed in the inferential model is from the ordinary notion of expression, let’s examine a straightforward attempt to define expression on the inferential model.

Alan Tormey, in The Concept of Expression, comes up with the following definition of expression:

If A’s behaviour B is an expression of X, then there is a warrantable inference from B to an intentional state of A, such that it would be true to say that A has (or is in state) S; and where S and X are identical. (Tormey, 1971, p. 43.)

This conditional, I will argue, states neither a necessary nor a sufficient condition for expression, and so it sheds no light on the concept what so ever.

The insufficiency of the condition is made apparent by the fact that other things than expression licence inferences to intentional states. If you see me going in to the library with jogging shoes sticking out of my back-pack, that might warrant the conclusion that I intend to go to the gym later. But the fact that I’m carrying around my gym-shoes doesn’t express my intention to go for a run on the treadmill. It can be a reliable indication of my intentions, and thus make my intentions knowable (for those who are familiar with my habits and ways) but that doesn’t mean that it makes any sense to call it an expression. So inference warrant is not a sufficient condition for expression. The more interesting point, however, is that inference warrant is not a necessary condition for expression.

The main problem with the idea that expression makes something knowable through warranting an inference is that it doesn’t capture the way in which expression, paradigmatically, makes what it expresses directly manifest. Charles Taylor has captured this point nicely,

When I know something or something is plain to me, through an inference, there is something else which I know or which is plain to me in a more direct way, and which I recognize as grounding my inference [.] It is characteristic of expression that it is not like this. I see the joy on your face, hear the sadness in the music. There is no set of properties that I notice from which I infer to your emotions or to the mood of the music. (Taylor, 1979, p. 74)

There are two arguments in this passage. One has to do with the phenomenology of recognizing expressions. The idea is that in paradigmatic cases, say, the expression of joy in a smiling face, the joy the joy seems immediately present to us. It doesn’t seem—as it does in the case of the intention gathered from the jogging-shoes—as a fact merely indicated, however reliably, by that which is immediately present. As Taylor points out, the words that come natural to us is that we see the joy in the face. If someone would say, about an apparently joyful face, that she inferred the joy from the way the face looks, that would probably strike us as an indicating a certain impaired ability to understand emotions. (If you read case studies of people diagnosed with autism, descriptions with this flavour are a commonplace.)

The second argument is logical. It hinges on the thought that for the idea of inference warrant to work, the expression must be thought of as a fact separable from the mental fact it provides a warrant for. This is, as Taylor rightly argues, not how it is with expressions. We are usually simply not able to describe the expression without specifying what the expression is an expression of. There is no independently available fact, “the expression itself,” from which the inferential step to the mental state is taken. When I recognize the joy in your smile I don’t do so by, say, noticing that your mouth is configured in such a way that the corners are pointing upwards and the upper row of your teeth are showing.

Taylor calls this phenomenon, discerning X in Y where there is not some other feature of Y which licences an inference to X, physiognomic reading. This captures a distinctive feature of the way in which expression allows something to be known.

In addition to the idea that expression offers a physiognomic reading, Taylor adds the important observation that expression is the most direct way of encountering the phenomenon expressed. This provides a further illumination of expression, since there are other phenomena, apart from expression, which allow a physiognomic reading. For instance, I can see the impending fall of a building, without being able to non-circularly specify what feature of the way the building looks makes me think it will fall. But the actual fall of the building can be observed on its own. Not so with expression. We can see the building fall, but it makes no sense to say that we can see the joy “in itself,” apart from the smile, the song, or the utterance. There could be and more adequate expressions, of course, but no such thing as observing what is expressed apart from its expression. (Cf. Taylor, 1979, p. 74)

At this point a proponent of the inferential model will think that I am begging an important philosophical question. Can’t we observe, or at least encounter, that which is expressed in a more direct way, namely in the first person? One of the main reasons why the inferential model has seemed so attractive is because of the undeniable difference between our own relation to our pain and other inner states, and the relations other can bear to them. However, on my view (which I can’t argue for here) first/other person asymmetry doesn’t directly lead to the inferential model, it only does so given a certain philosophical gloss. I conclude with the suggestion that one lesson from the previous considerations should be that we consider it to be a constraint on an adequate conception of first/other person asymmetry that it acknowledges expression and so avoids the implication that our knowledge of other minds is inferential.²