Complications in diagnosing lexical meaning: A rejoinder to Horvath and Siloni (2013)

We respond to Horvath and Siloni’s (2013) continued arguments against the reflexivization analysis of anticausatives, which we show suffer many of the same problems Beavers and Koontz-Garboden (2013) identified with the arguments in Horvath and Siloni (2011). In Beavers and Koontz-Garboden (2013) (BKG) we point to methodological, empirical, and conceptual inadequacies in Horvath and Siloni’s (2011) arguments against the analysis of anticausativization as a reflexivization operation, as argued by Chierchia (2004), Beavers and Zubair (2013), and particularly Koontz-Garboden (2009). One of BKG’s main points is a methodological one — a key task in evaluating claims of semantic unacceptability is ensuring that context is taken into account in evaluating the relevant data. As BKG demonstrated repeatedly, examples that sound odd absent a context are instead judged acceptable when appropriate context is supplied. Such observations have ramifications for whether certain acceptability patterns are due to pragmatics vs. lexical meaning qua the set of lexical entailments encoded by a given word (following e.g. Dowty 1991). BKG conclude that Horvath and Siloni’s (2011) claim that their data indicate an absence of reflexive causal meaning in anticausatives is not supported, and that the data instead support such an analysis. Despite this, Horvath and Siloni (2013) (HS) suffers from essentially the same problems. As such, our overarching point in this rejoinder is the same as in BKG. In what follows we largely just point out where these problems arise again, and close by responding to their assessment of advantages of the reflexivization analysis over alternatives such as causer deletion, showing that there are key empirical and formal misunderstandings that lead HS again to unsupported conclusions. 1 Context, acceptability judgements, by itself, and negation The problems regarding absence of context in HS are seen most clearly in relation to by itself modification. BKG (following both Chierchia 2004 and Koontz-Garboden 2009) claim that by itself only modifies predicates that lexically encode causal meaning, including causatives and, we argued, anticausatives, but not statives. Underived inchoatives vary depending on whether a given inchoative lexicalizes causal meaning. The crux of HS’s (p. 219ff.) counterargument is that this predicts by itself modifiers to be acceptable with all causatives. However, they note that while such modifiers are acceptable with causatives with animate subjects, they are less acceptable if the subject is inanimate. Consider (1) from Spanish (their (10a,b)) and (2) from Hebrew (their (9a,b)). (1) a. El the viento wind rompió broke el the vidrio glass (??por by sı́ REFL solo). only ‘The wind broke the glass by itself.’ b. El the calor heat fundió melted el the hielo ice (??por by sı́ REFL solo). only ‘The heat melted the ice by itself.’ (2) a. Ha-ruax the-wind šavra