The explanation of a child's discriminate responses to his environment turns on ascribing to the child a perceptual discrimination which counts certain things as more similar to one another than to some other thing. As Quine forcefully puts it:

Now for those determined to cleave to behaviourist canons, the problem is to use ‘perceptual similarity’ in explaining the subject's discriminating responses in a way which does not imply the existence of mental states and entities. What this really means is that the behaviourist must reconstruct the notion of ‘perceptual similarity', purifying it of its mentalistic dimension. So long as physicalism is a reasonable position, and while we are awaiting and abetting the neurophysiological millennium, the behaviourist's project is of significant moment. Now in Word and Object Quine does not seriously attempt to provide behavioural criteria for a subject's perceiving similarities, and he provisionally permits himself the mentalistic idiom he avows finally to eschew.