In this paper we report on the analyses of alternative approaches to semantic role annotation (FrameNet (FrameNet, 2005), PropBank (Palmer et al., 2005) and VerbNet (Kipper, 2006)) with respect to their models of description; granularity of semantic role sets; definitions of semantic roles concepts; and consistency and reliability of annotations, and we propose a methodological basis for identifying and analysing semantic roles, including a data-driven account of defining semantic role concepts. We present evaluation results of the defined concepts for semantic role annotation concerning the redundancy and completeness of the tagset, and concerning the reliability of annotations in terms of inter-annotator agreement.