Abstract
The Athabaskan and Eskimo approximants, particularly [j], [w], and [l], behave phonologically like obstruents. In both language families, [j], which alternates with a sibilant, is phonologically [+coronal], although its coronality has been questioned on phonetic grounds. These approximants best illustrate Sapir’s [1925] notion of ‘a true point in pattern’ which must be defined ‘over and above its natural [phonetic] classification on organic and acoustic grounds’. The fact that approximants are classified phonologically as sonorants or obstruents in different languages is incompatible with the universality of distinctive features and their phonetic properties. This problem is resolved by placing a phonological boundary for [+/––sonorant] somewhat differently in the continuum of the sonority hierarchy, depending on how approximants behave in individual languages.