The Jakobson‐Halle description of speech in terms of binary distinctive features is not valid. There are logical inconsistencies if features such as tone and compactness are defined in physical terms. It is impossible to give adequate definitions of features when they are required to be autonomous and relative, and when there are three contrasting linguistic units (such as tones in Yoruba) that differ only in the relative degrees in which they exhibit a particular property (rate of vibration of the vocal cords, or fundamental frequency, or nonlinguistically defined pitch). If the features are not required to be autonomous, the definitions lead to a disguised ternary (orn‐ary) system with a hierarchy of features. If recourse is made to definitions in terms of linguistically determined perceptual classes, then the system is trivial and cannot be said to permit phonetic specifiability of linguistic units is accordance with a universal phonetic alphabet. A better view is that phonological descriptions consist of an ordered set of necessary and sufficient statements linking grammatical statements with physical events. This view requires a new phonological theory.