ABSTRACT: The terms S (intransitive), A, P (transitive), as well as T and R (ditransitive) have been used since the 1970s to allow linguists to characterize the differences between major alignment patterns such as accusative vs. ergative. They are often taken for granted, but a closer examination of the literature reveals that they can have three rather different meanings: In the Dixonian approach, they are used as universal syntactic functions based on transitivity; in the Comrian approach, they are seen as comparative concepts for the two arguments of a typical action clause, and in the Bickelian approach, they are taken as generalized semantic roles (Dowtyan proto-roles). In this paper, I explain the three approaches in some detail, and I argue that the Comrian approach is superior to the other two approaches. It is better than the Dixonian approach, because it does not take an undefined transitivity notion for granted but defines transitivity in terms of A and P. It is also better than the Bickelian approach, because the Dowtyan proto-roles were designed for the description of a particular class of English verbs; for cross-linguistic comparison, we need to limit ourselves to restricted (and semantically coherent) classes of verbs if we want to arrive at general statements.