Though most psychologists and linguists have treated the phonemic segment as the basic unit for speech recognition in all of its aspects (perception, storage, retrieval, etc.), the supposed universal character of the phoneme has been questioned in the phonetic and psychological literature. This paper outlines (a) recent work in speech perception which suggests that the segment is the most appropriate perceptual unit, at least for English, (b) a new study that confirms prior work in modeling global sound similarity judgments for English and supports the all-purpose character of the segment, and (c) a recent extension to Arabic that points in the same direction.