Figure 3.1 is a simplified outline of a postulated direct route from speech to meaning. Speech reaches us as a stream of what physical analysis shows to be astonishingly indistinct sounds. These sounds, albeit imperfectly, approximate to phonemes. The auditory language analysis system in the brain analyses and guesstimates incoming speech into its constituent phonemes, then goes on to assemble these into words. At this point, the language is being stored as a form described by its sound, by phonemes. The words are being held in phonemic form, in phonemic code, in an auditory input lexicon. Each word in this auditory input lexicon, each unit of phonemically coded language, can be matched to its corresponding entry in the mental lexicon we envisaged before, where they are held according to their meaning, according to semantics. This mental lexicon we will call the semantic lexicon. When a phonemically coded entry, in the auditory lexicon, is successfully matched with its corresponding entry in the semantic lexicon we can say that the spoken word has been understood.

Figure 3.1 Understanding the spoken word.

I will now run another brief but important digression past you – on the evanescent phoneme. I have glibly pictured the identification and subsequent management of phonemes. There is a problem, however, as we have already noted. Spoken language is amazingly smeary. We mostly do not enunciate clearly, as newsreaders must. Analysis of everyday speech reveals it to be very messy stuff indeed. Phonemes are often barely delineated at all in conversational language; even word boundaries are usually very comprehensively blurred. Phonemes, in other words, are often not really there at all, except to a good speller, who already knows they are there (who is, a lot of the time, re-inserting them where they ought to have been in effect). Tiny examples: wossu’? is said, but you ‘hear’ what’s up?, eezaguboyinnee? says the man opposite and you ‘hear’ he’s a good boy, isn’t he? or someone shouts wazaca? but you ‘hear’ where’s the cat? You only ‘hear’ these phonemes if you already know how to spell the words of which you know they must be made. (Logically, this must be the case, even though you feel certain you hear them absolutely and precisely. This is a fine example of how your wonderful unconscious routinely puts together the best possible story for you from evidence which is usually thoroughly ragged and incomplete. See also notes to chapter 6.)

As Adams (1990 p.69) says ‘…however psychologically real a phoneme may be, it is acoustically evanescent… [phonemes] are not acoustically discrete.’ and ‘…it is not just that knowledge of a word’s spelling can influence the way in which people 'hear' its sounds but also the very likelihood that they will 'hear' a sound at all.’ (ibid. p.401). Speech does not deliver phonemes reliably; frequently it hardly delivers them at all. Adams again: ‘…there is no way to know that, say, the word cat is composed of three phonemes except by having, somehow, learned that it is.’ (ibid. p.69) We learn phonemic segmentation largely by learning to spell. (Read Adams 1990 pp. 65-81 for an elegant summary of the debate, as well as Goswami and Bryant 1990 and Scholes 1998.)