Categorization and features
Phonological features allow for formal expression of sound patterns used by speakers of a language. To understand where features come from, it is worth exploring where the patterns themselves come from. In this paper, we argue that the retroflex (tongue tip up) or bunched (tongue tip down) articulation of American English /ɹ/ is speaker- and context-dependent. We provide arguments against two overt sources for these patterns, phonological patterns and perception, as well as against their being purely the result of physiology. The conclusion we come to is that these patterns are spontaneously created by the speakers in order to provide order to their articulations of the sound /ɹ/. We conjecture that if patterns arise spontaneously, so too might features.