Causative constructions come in lexical and periphrastic variants, exemplified in English by Sam killed Lee and Sam caused Lee to die. While use of the former, the lexical causative, entails the truth of the latter, an entailment in the other direction does not hold. The source of this asymmetry is commonly ascribed to the lexical causative having an additional prerequisite of “direct causation", such that the causative relation holds between a contiguous cause and effect (Fodor 1970, Katz 1970). However, this explanation encounters both empirical and theoretical problems (Nelleman & van der Koot 2012). To explain the source of the directness inferences (as well as other longstanding puzzles), we propose a formal analysis based on the framework of Structural Equation Models (SEMs) (Pearl 2000) which provides the necessary background for licensing causal inferences. Specifically, we provide a formalization of a 'sufficient set of conditions' within a model and demonstrate its role in the selectional parameters of causative descriptions. We argue that “causal sufficiency” is not a property of singular conditions, but rather sets of conditions, which are individually necessary but only sufficient when taken together (a view originally motivated in the philosophical literature by Mackie 1965). We further introduce the notion of a “completion event” of a sufficient set, which is critical to explain the particular inferential profile of lexical causatives.
Given that causative linguistic constructions are divisible into three parts: i) a cause (c); ii) an effect (e); and iii) the dependency (D) between (c) and (e), in studying the nature of (D), one should examine whether a one, all-encompassing, causative meaning component underlying the diverse linguistic phenomena is a justifiable position, or rather different ones should be distinguished for the various causative constructions. Only recently several philosophers argued in favor of theories of causal pluralism, allowing the co-existence of different notions of causation; some cognitive studies also indicate that people have a pluralistic conception of causation, similarly it has been proposed that the semantic content of (D) is different in various constructions, tracing whether the main verb encodes a necessary or a sufficient condition. This paper expands on this latter line of thought by focusing on the types of dependencies encoded within three verbal constructions in Hebrew, considering crucially whether these dependencies are asserted and/or presupposed. It argues, therefore, in favor of a non-unified semantic analysis for (D) denoted by the three verbal causative constructions to be passed under review here: overt causatives, verbs of change of state and caused activity verbs. According to the current proposal: Overt causatives assert necessary conditions; change of states causatives assert necessary conditions and presuppose potential sufficient conditions; and Caused activities only presuppose potential sufficient conditions.