Standard semantics for counterfactuals (Lewis 1973, Kratzer 1981) predicts that logically equivalent antecedents should be substitutable salva veritate. A recent study by Ciardelli et al. (2018) challenges this picture by finding different endorsement rates for counterfactuals of the form in (1a) and (1b) in the same scenario.

1.    a. (not A or not B) > C  
      b.  not (A and B) > C

Ciardelli et al. propose an explanation based on inquisitive semantics, which allows for differences in meaning between de Morgan-equivalent sentences by encoding key lexical differences in the denotation of disjunction and conjunction. Bar-Lev (2018, 2019) proposes an implicature account, based on assumptions about the alternatives of negation. The two accounts make different predictions with respect to the source of the effect, linking it respectively to disjunction and negation. We report a series of experiments testing these predictions. We use simpler examples than Ciardelli et al. and control more systematically the role of negation. We find an effect of negation but no difference between connectives. These results are challenging for the proposal by Ciardelli et al, and are instead in line with the negation-based approach.