A prominent theory in the face perception literature--the parallel-route hypothesis (Bruce & Young, 1986)--assumes a dedicated channel for the processing of identity that is separate and independent from the channel(s) in which nonidentity information is processed (e.g., expression, eye gaze). The current work subjected this assumption to experimental tests from the perspective of three characterizations of independence and separability: Garner's speeded classification task (Garner, 1974), general recognition theory (Ashby & Townsend, 1986), and systems factorial technology (Townsend & Nozawa, 1995). The goals of the study were to (a) provide converging evidence on validity of the parallel-route hypothesis, and (b) understand the relationships between various definitions of independence and separability. The results of the current work demonstrated that identity and expression are not completely independent, using data from all three perspectives. Consequently, a modification of the central assumption of the parallel-route hypothesis is required. In addition, the intricate but orderly relationships between various types of separability and independence have implications for theories of perceptual independence. (Contains 15 tables, 6 figures and 7 footnotes.)