Speech is a fleeting phenomenon. In order to study it, we must capture it - keep it 'present' to us, re-present it to ourselves - by letting something more permanent stand in for it, or represent it. Representation is thus a necessary precursor to any analysis of speech, whether practical or theoretical. However, representation faces us with several kinds of problems. First, we have to choose appropriately among many ways of representing speech - any of several kinds of writing systems, any of several kinds of transcription systems, output from any of several kinds of phonetic analysis equipment (spectograms, electropalatograms, etc.), abstract diagrams in any of several specialist theories. Second, whatever choice we make inevitably brings with it the danger that we might confuse characteristics of our representation with characteristics of speech itself.