The phonemes of a language are categories defined for the native speaker by acoustic and functional attributes.' Workers at the Haskins and Bell Telephone laboratories have tried to specify the acoustic invariants of particular phonemes of American English; they have usually worked with single vowel allophones, and with the first three formants in the spectrograms of these allophones. Jakobson and his co-workers, drawing on these acoustic studies and on familiar articulational analyses, have proposed that the phonemes of English can be specified in terms of a small number of acoustic-articulational attributes having binary values.2 Each phoneme of English is said to be a unique bundle of such distinctive features. In all this, nothing is said about the functional attributes of phonemes. When studying the acoustic cues that affect phonemic identification of phones it is customary to present the different phones in the same phonetic context (e.g. /hid/, /haed/, /hud/). By using a common context the experimenter indicates to his subject that context will not be a useful cue to the identification of the phone. This is a reduced laboratory situation that permits experimental isolation of acoustic attributes. In ordinary speech perception, of course, the context functions as an important cue. In the utterance /sta( ) 'Oijf/ a native speaker of English expects a /p/ in the indicated position and is disposed to categorize whatever phone actually occurs in that position as a /p/. Bloomfield has pointed out3 that, in every language, forms are grouped into form classes by some recognizable phonetic or grammatical feature. In English, for instance, nouns are grammatically acceptable in certain sentence frames and unacceptable in other frames. Nouns are a form class by virtue of this limited functional equivalence. The sentence frame itself might be described as a functional attribute of nouns. A speaker of English will expect to find a noun in a particular frame. It is a functional characteristic of nouns that they are able to occur in such a frame. In similar fashion, we suggest that the phonetic contexts of a phoneme may be described as functional attributes of that phoneme. If the gap in the context is always filled by the same phoneme in English, then that context is a defining functional attribute of the phoneme. In written English the context q( )i defines the letter u as surely as the lines of the letter itself. To the degree that one phoneme is more probable than any other in a given context, that context is a distinctive attribute of the phoneme. Most of the categories of speech seem to be defined for the native speaker by