States and changes of state: A crosslinguistic study of the roots of verbal meaning

What are the basic building blocks of verb meanings, how are they composed into more complex meanings, and how does this explain the grammatical properties of verbs and their relationships to other words with related meanings? These questions are fundamental to the study of verb meaning, and some of the most fruitful attempts to answer them have come from event-structural theories. wherein verb meanings are assumed to be decomposed into an event template, which captures the verb's broad temporal and causal contours, and an idiosyncratic root shared across templates, which describes specific actions and states for a given verb. An open question is what the division of labor is between the template and the root in a given verb's event template, and whether their meanings are bifurcated: are broad eventive lexical entailments introduced only by the templates, never the idiosyncratic roots? Since event templates and not roots are the primary semantic correlates of a verb's grammatical properties, bifurcation would make strong predictions about the correlation of a verb's broad temporal and causal semantics and its syntax and morphology. We argue against this bifurcation by comparing translation equivalents of Levin's (1993) non-deadjectival vs. deadjectival change-of-state verb roots in English (e.g. result vs. property concept roots) across languages. A broad-scale typological study reveals that property concept roots tend to have unmarked stative forms and marked verbal forms, while result roots have the opposite pattern. Semantic studies of several languages confirm that terms built on result roots always entail change, while terms based on property concept roots do not. This supports a theory wherein result roots entail change independent of the template, contra bifurcation. This supports a more complex, albeit still principled, theory of possible event-structural meaning and its grammatical correlates, one that takes subclasses of roots into account, while showing the value of this type of cross-linguistic methodology for testing the predictions of event-structural approaches.*