This paper examines the validity of the conception of the sick role which underlies much research into illness behavior. The method by which this role conception is tested is the Multitrait-Multimethod matrix. The essentially negative findings of the matrix analysis support and expand recent findings of Twaddle (1969) and Gordon (1966), and they lend credence to the suggestion that the sick role as it is presently conceived is an inadequate unit of analysis for the study of illness behavior. The comments of the respondents to this study are analyzed, and they are used to support the view that behavioral expectations of the ill are specific to person-illness units. An inductive strategy of research is suggested in which commonalities of expectations among person-illness units across various subgroups in the population are explored as a means of pursuing the role theoretic implications of illness behavior.