Private Language Argument

The private language argument is Wittgenstein's argument in the Philosophical Investigations (§§ 243–315) against the intelligibility of a language whose terms could be understood in principle only by a single person — most pressingly, a language for one's own inner sensations (pains, after-images, felt qualities). The argument: suppose I try to establish a name for a recurring inner sensation by concentrating on it and saying 'I call this S.' What criterion ensures that I apply 'S' correctly on subsequent occasions? The sensation is not publicly observable, so no external correction is possible. But without the possibility of correction, there is no distinction between correctly applying 'S' and merely seeming to apply it correctly — which means there is no rule being followed and therefore no genuine language. The argument is not a denial that inner states exist; it is a denial that inner ostension (mental pointing at a private object) can establish the meaning of a term. Meaning requires public practice, checkable use, the possibility of being wrong. The private language argument is the most technically demanding and debated section of the Investigations and underlies Wittgenstein's critique of Cartesian inner theater — the picture of the mind as a private arena to which only the owner has access, populated by objects that cannot in principle be publicly identified or verified.

The Empirical Challenge: Does Neuroscience Change the Conditions?

Wittgenstein's argument was framed against a specific background assumption: that inner states are, in principle, inaccessible to third-person observation. If the pain I report when touching a hot stove is a purely private Cartesian datum — immediately given to me, hidden from everyone else — then there is no public criterion for whether I apply 'pain' correctly across occasions. The argument's power depends on this privacy being effective.

Brain imaging research over the past thirty years has complicated this assumption in ways Wittgenstein did not anticipate. Multi-voxel pattern analysis can identify, with above-chance accuracy, which of several stimuli a subject is currently experiencing, or which of several mental images they are imagining, from activation patterns alone — without any behavioral report. This does not eliminate the epistemic asymmetry between first-person and third-person access, but it does introduce a criterion for inner states that is external, repeatable, and correctable. If two subjects who both report 'pain' show systematically different neural activation patterns, neuroimaging gives us grounds — independent of verbal report — to ask whether they are using the word to refer to the same kind of state.

The immediate objection is that neuroimaging tracks the neural correlates of inner states, not the states themselves — that there remains a gap between measuring the physical substrate and having access to the phenomenal quality. This objection is correct in one respect: neuroimaging does not give us direct access to what pain feels like. But Wittgenstein's argument does not require access to phenomenal quality — it requires a criterion for correct reapplication of a term. Neural activation patterns provide a criterion of that kind. Whether they provide the right kind of criterion is what is at stake.

The deeper challenge the empirical literature poses to the private language argument is not that it refutes Wittgenstein's conclusion, but that it relocates his question. The argument asks: what could establish the identity of a sensation across occasions? The neural stability of a sensation-type — its consistent causal signature in the brain — provides one answer. Not a philosophical answer (about meaning in the sense of semantic content) but a functional answer (about the reliable tracking of a stimulus category). Functionalism about mental states, the dominant view in cognitive science, identifies mental states with their functional roles — what causes them and what they cause — rather than with their intrinsic phenomenal character. Under functionalism, the private language argument's premise (that inner states are private in the relevant sense) becomes empirically contested rather than conceptually obvious.

What the argument retains, even under this pressure: the point that *meaning* cannot be constituted by inner ostension alone. Even if neuroimaging gives us a third-person criterion for tracking sensation-types, this criterion is not available to the subject in the moment of first learning the word 'pain'. The language acquisition process is still a social process, embedded in public behavioral cues — reaching for a burn, flinching, crying — that ground the word's application before any neural criterion is available. The public practice grounds the meaning; the neuroimaging study confirms the stability of what that practice refers to. These are compatible. But the compatibility needs to be argued, not assumed, and the article in its current form does not acknowledge that the argument is needed.