In this work two linguists from different theoretical paradigms develop a new general theory of natural language predicates. This theory is capable of addressing a broad range of issues concerning (complex) predicates, many of which remain unresolved in previous theoretical proposals. Grounded in empirical evidence from a wide variety of genetically and geographically unrelated languages (German, Hungarian, Fox, Nenets, Tzotzil, Malayalam, among others), this new theory synthesizes conceptual and representational assumptions from several different theoretical traditions. The authors focus on cross-linguistically recurring patterns of predicate formation where identical contentive notions (i.e., lexical semantic, grammatical function, and morphosyntactic information) are expressed by predicates consisting of a single morphological word or by combinations of independent words that need not form a single syntactic unit. They provide a detailed implementation of their theory for German tense-aspect, passive, causative, and verb-particle predicates. In addition, the authors discuss extensions of these representative analyses to the same predicate constructions in other languages. Beyond providing a formalism for the analysis of language-particular predicates, they demonstrate how the basic theoretical mechanisms they develop can be employed to explain universal tendencies of predicate formation. For this purpose, Ackerman and Webelhuth introduce the construct `grammatical archetype' into linguistic theory, relating universal patterns of predicate formation to language-particular patterns in a principled fashion.