Locally Accommodated Presuppositions Are Also Backgrounded

Two generally agreed upon characteristics of presuppositions are projection and backgroundedness. Yet, presuppositions sometimes fail to project. To derive the necessary local interpretations, standard semantic local accommodation accounts posit an operation that, inside the scope of an embedding operator, turns content lexically marked as presupposed into non-backgrounded content and conjoins it with the clause’s entailed content. Such accounts, as well as syntactic operator accounts descended from them, suggest that locally accommodated presuppositions might differ from projecting presuppositions in lacking not just projectivity, but also the second basic presuppositional property of backgroundedness. Recent pragmatic accounts arrive at a parallel, even stronger prediction of non-backgroundedness for locally accommodated content via their claim that all and only backgrounded material projects. To date, though, this idea, that non-projecting presuppositional content is also not backgrounded, has never been empirically tested, perhaps due to challenges in testing embedded material for backgroundedness directly. Using reduced cognitive salience as a measure of presuppositional backgroundedness, we present data from picture-matching and novel information-seeking tasks to indirectly evaluate the backgroundedness of locally accommodated presuppositions of also and continue. Baselines are provided by their explicit, non-backgrounded conjunction paraphrases and equivalent non-presuppositional elisions. Existing standard local accommodation accounts lead us to expect equivalence across all three, since they represent all three as non-backgrounded. However, our outcome measures indicate greater backgroundedness for locally interpreted presuppositions than equivalent explicit proffered content and elision controls (though to a lesser degree for the latter). We find similar patterns for global accommodation and for occurrences of also and continue that require no accommodation at all. These findings indicate that the interpretation of presuppositional content associated with lexical triggers should be modelled in a way that marks it as having a discourse status more backgrounded than that of main content in the same clause. Standard models accomplish this for globally accommodated and non-accommodated presuppositions by ensuring that the dynamic semantics includes their content in the context of interpretation of the clause in which the trigger occurs. However, such a direct solution is not available for locally accommodated presuppositions, as it fails to distinguish between the significantly backgrounded status we find for locally accommodated content and the non-backgrounding of identical content presented non-presuppositionally. We suggest augmenting standard representations of backgroundedness with variable givenness features in order to represent accurately the conventional backgroundedness marking of presuppositions.