David Krassnig

We have conducted an acceptability judgement experiment to examine two hypotheses related to the felicity of reverse Sobel sequences (rSS). The first hypothesis was based upon Lewis’ (2018) relevance-based variably-strict semantics: If two rSS are identical except for their respective difference in dissimilarity between their antecedent worlds, then the rSS with a higher degree of dissimilarity should be, on average, more acceptable. Our results, however, seem to support this hypothesis only weakly and appear more contradictory than supportive to the model behind it: Any kind of clear-cut dissimilarity will render an rSS felicitous, so long as its conditionals are counterfactual by nature. The second hypothesis, that rSS whose domains of quantification are entirely disjoint should be just as acceptable as regular utterances, was quasi-confirmed. Whilst a significant difference to the control items was found, these rSS differ only minimally in average acceptability and are the highest rated rSS so far. We therefore explain the slight, statistically significant decrease in acceptability as a result of the markedness of rSS structures. Finally, we provide two analyses to account for the data gathered. First, one modification of Lewis’ (2018) account, where we argue that relevance may not increase closeness beyond the levels set forth by worlds similarity. Then, with the second account, we attempt to motivate the need for and the pragmatic contribution of contrastive stress with regards to rSS, using Ebert et al.’s (2008) assumption that antecedents represent their conditional’s aboutness topic – thereby deriving contrastive topic with regard to the two sets of antecedent worlds.

In this paper, we show that an account centred around modal subordination and contrastive topic may account for most if not all of the known data on reverse Sobel sequence felicity. We show that contrastive stress in rSSs often targets the auxiliary verb and that, in doing so, we actually target the tense-aspect-mood information that is encoded by it. By treating tense-aspect-mood as a type of bound pro-form, adopting the pro-form semantics of Jacobson (2000; 2004), we show that, in doing so, we are actually trying to contrast the two conditionals’ domains of quantification. As such, the contrastive topic is successful iff the two domains are disjoint, preventing a modally subordinate reading, allowing for the possibility of reverse Sobel sequence felicity.

In this thesis, I investigate whether conditionals can be more accurately modeled with a (semi-)dynamic strict semantics or a variably-strict semantics. Two linguistic phenomena related to conditionals are examined: negative polarity items and (reverse) Sobel sequences. Negative polarity items are words with a variable distribution, often occurring in negative contexts. I reject the traditional account of licensing these items in downward monotone environments and propose an alternative operator-based approach, suggesting they are licensed by a covert even-like operator with a scalar presupposition. Regarding (reverse) Sobel sequences, which have been debated in conditional modeling, the initial favoring of (semi-)dynamic strict semantics is challenged by recent findings that some reverse Sobel sequences can be felicitous. This supports a return to the variably-strict approach, incorporating selective pragmatic mechanisms to explain their felicity. Through a reverse Sobel sequence felicity experiment, I identify contrastive stress in the antecedent as a crucial factor for their felicity. Additionally, counterfactuality, a lack of causal link between φ and ψ, and the use of overt or covert exclusion of ψ as an epistemic possibility influence felicity. I propose a pragmatic model centered on contrastive stress, which effectively accounts for all empirical data on reverse Sobel sequences. This pragmatic mechanism can be integrated with either the variably-strict or (semi-)dynamic strict model without altering the predicted felicity distribution. Ultimately, I conclude that the (semi-)dynamic strict model has a slight explanatory advantage over the variably-strict approach, but further modifications to the latter could render both approaches equally viable for explaining (reverse) Sobel sequences and negative polarity items.