An Experimental Comparison Between Presuppositions and Indirect Scalar Implicatures

Abstract

We compare two aspects of meaning, namely the presupposition of stop in the scope of negation (John didn't stop going to the movies on Wednesday., → John used to go to the movies before Wednesday.) and scalar implicatures associated with strong scalar items under negation (John didn't always go to the movie last week. → John sometimes went to the movies last week.) ('Indirect Scalar Implicatures' (ISIs) Chierchia, 2004). Our results replicate the finding by Chemla and Bott (2013) that global presupposition interpretations are faster with a different methodology that avoids a potential confound of response bias. More surprisingly, ISIs are found to pattern with presuppositions in that responses reflecting an interpretation without an inference (corresponding to a 'literal' interpretation) are slower than ones based on the relevant inference, contrary to what has been found for direct scalar implicatures. These results are puzzling from the traditional perspective that ISIs are generated in the same way as direct implicatures. We explore two possible interpretations: first, strong scalar terms could receive a presuppositional analysis as well and presuppose that their domain is non-empty. Alternatively, we could group stop and ISIs together from another angle and see them as obligatory scalar implicatures, in contrast to the non-obligatory direct ones.

Presupposition projection from the scope of None: Universal, existential, or both?

“…The child data from Experiment 2 suggest that the UNIVERSAL reading arises from the basic projection mechanism, with the EXISTENTIAL reading being derived through some additional mechanism such as domain restriction. Reaction time data from adults reveal a delay associated with the PRESUPPOSI-TIONLESS reading, suggesting that it involves an additional costly mechanism (e.g., local accommodation), in line with previous results in the literature (Chemla & Bott 2013;Romoli & Schwarz 2014). In addition, the child data, combined with previous results from Bill et al (2016), suggest that in the case of negative quantifiers, there may be two routes for arriving at the PRESUPPOSITIONLESS reading.…”

“…One way to interpret this, which speaks to our third research question, is that the observed slow-down is indicative of an additional step that is required in order to access the PRESUPPOSITIONLESS reading. Such an interpretation is in line with previous results in the literature (Chemla & Bott 2013;Romoli & Schwarz 2014), and is compatible with all theories positing an extra mechanism responsible for suspension.…”

“…Second, if both EXISTENTIAL and UNIVERSAL readings are available, we will explore the relationship between them, in particular investigating whether one may be derived from the other. Finally, we will investigate the nature of the PRESUPPOSITIONLESS reading, in particular whether it corresponds to a literal meaning of sentences like (5) or rather involves further processes (Chemla & Bott 2013;Romoli & Schwarz 2014).…”

Intervention effects in NPI licensing: A quantitative assessment of the scalar implicature explanation

“…As we will see, the results of our Experiment 1 run counter to this expectation. Regarding the processing of implicatures, we will merely note here that the majority of previous experimental work suggests that interpreting a sentence without its implicatures is equivalent to not deriving the implicatures rather than cancelling them; for instance, Bott & Noveck (2004) and Bott et al (2012) provide evidence that participants who give implicature-less responses are faster to respond than those who give implicature-based responses, and Cremers & Chemla (2014) report similar results for indirect scalar implicatures, which are the kind that concern us here (though see also Romoli and Schwarz 2015).…”

Remarks on the Experimental Turn in the Study of Scalar Implicature, Part II

“… Cremers and Chemla () and Romoli and Schwarz () present evidence that indirect SIs are not costly, but Cremers and Chemla () argue that there are confounds in these findings, and they show that once these confounds are overcome, the computation of indirect SIs does come with a processing delay.…”