If you trained someone to emit a particular sound at the sight of something red, another at the sight of something yellow, and so on for other colors, still he would not yet be describing objects by their colors. Though he might be a help to us in giving a description. A description is a representation of a distribution in a space (in that of time, for instance).

Suppose you and I belong to the same linguistic community and express in our public behaviour just the same colour discriminations and comparisons. Could you and I still have inverted colour experiences? Some philosophers think so. 1 In recent years, those who have doubted this have often appealed to the various "asymmetries " between colours as we perceive them. The argument has been roughly this: The relevant inversion requires that our colour experiences could differ with respect to their intrinsic qualities while having the same structure or form. But a close consideration of colours as we see them shows that this is not possible. For example, orange is perceived as a mix of yellow and red but red is not perceived as a mix of any two colours. A significantly darkened blue still looks blue while a significantly darkened yellow does not look yellow. Again, red is generally perceived as a warm colour while green is perceived as a cool colour. Because of these and many other differences, the intrinsic qualities of our experiences could not differ systematically while having the same structure. 2 However, the existence of such asymmetries is not the only difficulty for the view that you and I could be spectrum inverted while expressing the same colour discriminations and comparisons. Nor, I think, is it the main difficulty. I shall argue that the main difficulty is instead to give a satisfactory account of what you and I would believe and say on such a scenario. This problem is independent of any consideration regarding asymmetries. And as far as I am aware, nobody has to this day made plausible that it can be solved. In section 1, I specify the scenario I will consider. In section 2, I present an argument for its impossibility. In section 3, I consider some recent accounts of spectrum inversion. I try to illustrate, here, that contemporary defenders of the

Wittgenstein (1968) endorsed one kind of inverted spectrum hypothesis and rejected another. This paper argues that the kind of inverted spectrum hypothesis that Wittgenstein endorsed (the “innocuous ” inverted spectrum hypothesis) is the thin end of the wedge that precludes a Wittgensteinian critique of the kind of inverted spectrum hypothesis he rejected (the “dangerous ” kind). The danger of the dangerous kind is that it provides an argument for qualia, where qualia are (for the purposes of this paper) contents of experiential states which cannot be fully captured in public language. I will pinpoint the difference between the innocuous and dangerous scenarios that matters for the argument for qualia, give arguments in favor of the coherence and possibility of the dangerous scenario, and try to show that some standard arguments against qualia are ineffective against the version of the dangerous scenario I will be advocating. One of the two arguments for qualia I will give is a “shifted spectrum ” argument that is much less controversial than the version I gave in Block (1999), and the other argument for qualia is an “inverted spectrum” argument that is much less controversial than the one I gave in Block (1990). The inverted spectrum argument is much less controversial because it does not require a behaviorally indistinguishable spectrum inversion. Wittgenstein’s views provide a convenient starting point for a paper that is much more about qualia than about Wittgenstein.