Behaviorism is a bit like a zombie: just when you think you’ve killed it the thing comes lumbering back into the room. So it may be worth reminding ourselves quite how bad the reasoning was that led to it. The doctrine may be defined thus: mental states are dispositions to observable behavior. There are three concepts at work here: disposition, observation, and behavior. None of these is redundant: each could be satisfied without the others being satisfied. Mental states could be dispositions to produce certain effects without those effects being observable or behavioral, since they could be entirely mental. They could be known by observation without being dispositional or behavioral, since they could be known by observing brain states or possibly by something like telepathy (assuming this to be logically possible). And they could be behavioral without being either dispositional or observable, since the behavior might be invisible—it might only occur at a microscopic level or possibly wear a cloak of invisibility (assuming this logically possible). Presumably no one would advocate behaviorism if the behavior was as unobservable as the mental state within, but this is not to be ruled out on logical grounds. The whole point of introducing behavior is to provide something that will function as observable evidence, so unobservable behavior is not going to cut it: hence the need to specify observable behavior (sometimes “gross behavior”).
The reasoning, then, goes from observation to behavior to disposition (the last introduced in order to allow for mental states on which the organism is not currently acting). We want something observable if psychology is to be a real science; behavior is observable; therefore mind is behavior. There is already a non sequitur here: it needs to be the case that only behavior is observable. That may seem de facto true, but it’s worth noting how contingent it is. What if brain waves detectable by an EEG machine were more accessible and revelatory than they actually are? Then scientists could read the mind off the recordings taken by such a machine without recourse to behavior.Psychology could thus be a science based on electrical observations, not unlike parts of physics. It is merely a contingent fact that this is not the case. Or suppose there were super-scientists equipped with a special sense capable of detecting other people’s mental states directly—again no need to invoke behavior. Or suppose brain chemistry has reached such a point that chemical analysis can reveal what is going on mentally. None of these is practically possible as things stand, but they are not ruled out as a matter of principle. So the behaviorist is in effect saying that as a practical matter the only evidence we have is behavioral, not that no other evidence is logically conceivable. Consider how psychology might be conducted by blind psychologists: no one can see anyone’s behavior. Would the behaviorist approach seem so attractive then? What if EEG recordings delivered by Braille, though crude, actually outperformed attempts to hear or feel or smell other people’s behavior? Then the appeal of behaviorism as an evidence-based approach to psychology would presumably lapse. Behaviorism only seems appealing because we sighted people in fact generally judge other people’s mental condition by looking at what they do, but that is hardly a logical necessity. It amounts to the claim that as a matter of contingent fact we judge other people’s states of mind by visually observing their behavior, while allowing that there is nothing necessary about that.
But this gives the game away: for how could the mind be dispositions to behavior, simply because behavior is the only evidence we actually have, given our senses and the state of our technology, for ascribing mental states to others? Isn’t this a blatant case of trying to derive an ontological conclusion from an epistemological premise? It may be that the only evidence we practically have for knowing about distant stars is the light that reaches us from those stars, but that hardly implies that stars are patterns of light traveling across space. Evidence is one thing, fact another. A star is not a collection of dispositions to send light into our eyes, even though we have no evidence apart from this; and it is perfectly possible that other forms of evidence might emerge as technology develops—for example, we might travel to the stars and have an up-close look. So the reasoning in support of behaviorism commits a glaring non sequitur: it moves from a claim about the evidence actually available to us to a claim about the nature of the thing for which this is evidence. This is an attempt to move from a highly contingent fact about evidence to a constitutive truth about the nature of mind. If we made such a move in the possible scenarios I sketched, we would end up saying that mental states are dispositions to cause EEG recordings or chemical changes in the brain or even telepathic intuitions in a certain class of observers. None of these are what the mind is; they are merely possible sources of evidence regarding the mind. Behaviorism is no different: it is the elevation of one source of evidence, itself quite shaky, to the status of constitutive truth. There is nothing privileged about behavior in providing possible evidence about the mind, so converting it into a constitutive claim is bizarre and unwarranted. The mind cannot be what just happens to provide evidence for it to us now, given our senses and state of technology. Behavior is really just one effect of mental activity to be set beside others (electrical fields around the brain, chemical changes within it, remote stimulations of human senses at some distance from brains). Choosing another type of effect by which to define the mind would strike us as bizarre, but then why is behavior regarded as constitutive? And add to this the point that gross observable behavior is actually a very unreliable and crude guide to what is going on in someone’s mind, being just a displacement of the body caused by an internal state. Your larynx moves thus and so when you vocally express your pain by groaning, but the pain itself may have a complex internal reality that exceeds this relatively coarse mode of expression. Also, behavior can be deceptive and misleading: it is hardly a certain guide to the other’s state of mind. EEG recordings would be much more reliable and attuned to what is really going on inwardly. Behaviorism is appealing to a rather impoverished source of evidence as well as committing a logical fallacy (trying to deduce an ontological conclusion from an epistemological premise). In ordinary life we have nothing better to go on most of the time, but as a basis for solid science behavior is far from ideal—so why try to convert it into a definition of the mental? That’s like taking unaided observations of the stars from planet earth as constituting the very nature of the astronomical world. Looked at like this, it is hard to see how anyone could have taken behaviorism seriously.
 A follower of Wittgenstein might insist that behavior supplies a criterion of the mental not merely a symptom; it is woven into the concept of mind. But why should the same thing not be true of EEG recordings if they formed part of our language game with mental words? What if children were brought up this way and took it for granted?
 There is also the familiar (and good) point that it is wrong to reduce a postulated theoretical entity to the evidence for it: but this concedes too much to the behaviorist in allowing that behavior is the unique and necessary form of evidence for mental attributions. What we need to appreciate is its thoroughly contingent status as evidence—whatever necessity it has is purely practical. Presumably if brain scanning were more realistic at the time behaviorism gained a foothold it would not have had the appeal it did have. The introspectionist school in psychology would be opposed by the brain scan school not by the behaviorist school. Maybe in the future behavioral evidence will be completely replaced by brain scan evidence; then someone will no doubt proclaim that the mind is a set of dispositions to produce images on scanning machines.