Two theories of speech—one quite conventional, the other much less so—account very differently for the biological advantage of speech over writing/reading. The guiding assumption of the more conventional theory is that the elements of speech are sounds, and that these are served by processes of motor control and auditory perception that are in no way specialized for language. Accordingly, there must be a cognitive stage, beyond action and perception, where the motor and auditory representations are somehow invested with linguistic significance. On the conventional view, then, the sounds of speech are just like the letters of the alphabet. Neither has more than an arbitrary relation to language, hence the difference between them is trivially a matter of which of the equally large gaps between signal and message needs to be bridged. On the less conventional theory, the ultimate constituents of speech are not sounds, but articulatory gestures. Having evolved exclusively in the service of language, they form a natural class, a phonetic modality. Being phonetic to begin with, they do not require to be made so by cognitive translation. And that, very simply, is the advantage of speech over writing/reading. Speech has the corollary advantage that it is managed by a module biologically adapted to circumvent limitations of tongue and ear by automatically coarticulating the constituent gestures and coping with the complex acoustic consequences. But a result is that awareness of phonetic structure is not normally a product of having learned to speak: The module “spells”—that is, sequences phonetic segments—for the speaker and recovers the segments for the listener, leaving both in the dark about the way that is done; the gestural representations are immediately phonetic in nature, precluding the cognitive translation that would bring them to notice; and coarticulation destroys all correspondence in segmentation between acoustic and phonetic structures, making it that much harder to demonstrate the alphabetic nature of speech at the acoustic surface. Accordingly, special difficulty in becoming literate might be caused by a weakness of the phonetic module, for that would produce primary representations of a fragile sort, with the consequence that they would be that much harder to bring to awareness—as is required if they are to serve writers and readers as the units of an alphabetic script—and also that much less able to bear the weight of working memory.