§34. But suppose someone said: "I always do the same thing when I attend to the shape: my eye follows the outline and I feel....". And suppose this person to give someone else the ostensive definition "That is called a 'circle'", pointing to a circular object and having all these experiences—cannot his hearer still interpret the definition differently, even though he sees the other's eyes following the outline, and even though he feels what the other feels? [Yes!] That is to say: this 'interpretation' may [will] also consist in how he now makes use of the word; in what he points to, for example, when told: "Point to a circle".—For neither the expression "to intend the definition in such-and-such a way" nor the expression "to interpret the definition in such-and-such a way" stands for a [inner, private - or at least decisive] process which accompanies the giving and hearing of the definition.

ELUCIDATION

  1. 'Intending a definition' and 'interpreting a definition' could seem, if not in OL then certainly to philosophers, to refer - to uniquely refer - to a private, inner, process. W. proposes that there is no such unique, private, process. Something goes on privately, no doubt, and various things go on publicly. The sum total of these events, notably including the public events such as B pointing at something and saying "That's a circle" are what we mean by 'correctly interpreting a definition'. In particular, such events are not (merely) confirming that B now is interpreting the definition correctly. The agreement of the human community, the nodding at B's use, is what 'interpreting the definition correctly' means. There is no 'correct definition' of 'circle', or 'sepia', which lurks somewhere, in a book, or in someone's mind - some arrow or dotted line that links the words to abstract ideas in Plato's Third World, or to some objects in the world. Circle' does not have a meaningx aside from its community use.
  2. Compare his later discussion of 'understanding', of  'reading', of 'rules', of 'S', and of 'pain'.  The constant theme of these discussions is that words that appear to refer uniquely to private stuff do not do so. They refer to a more holistic set of events, partly private, and partly public. B learns from A how to use the words, by various public events (pointing, repeating, speaking...). A checks (collects evidence), as far as possible, that B is an accredited user of the public language by public testing public demonstrations - no other kind of testing is possible. It is not that 'defining' necessarily does not somewhat involve private stuff. It is that the public language, by its public nature, cannot work by uniquely targeting this private stuff. We may delude ourselves that it does, because the private stuff is what is closestx to us. But it is a self-aggrandising delusion; our language is a public tool, which works with public material - because there is no other possibility.