Lost ALL of your marbles? The impact of the Question Under Discussion on embedded implicatures

Team: Nicole Gotzner (SIGames, ZAS Berlin), Judith Degen (ProComPrag, Stanford) and Noah Goodman (Stanford)

A sentence like “Ann found some of the marbles” may give rise to the inference that Ann did not find all marbles, referred to as an implicature (Grice, 1975). Whether or not listeners draw implicatures depends on several contextual factors such as the Question Under Discussion the context evokes (QUD, Roberts, 1996; Zondervan, 2010; Degen & Goodman, 2014). Degen & Goodman (2014) found that implicature rates increase if the context triggers the implicit QUD Did Ann find all marbles? compared to the QUD Did Ann find any marbles? (see also Zondervan, 2010). The likely reason for such an effect is that the former QUD makes the stronger alternative all more relevant. In fact, most theories of implicature assume that alternatives need to be relevant/active as a precondition for an implicature to arise.
Here, we apply the paradigm proposed by Degen & Goodman (2014) to embedded implicatures. It is a controversial theoretical debate whether implicatures can arise in embedded positions. To give an example, researchers disagree whether the sentence “Every person found some of the marbles” implicates that every person found some but not all marbles (e.g., Sauerland, 2004; van Rooij & Schulz, 2004; Chierchia, 2004; 2006).
Experimental evidence on whether listeners draw embedded implicatures is mixed (Geurts & Pouscoulous, 2009; Clifton & Dube, 2010; Chemla & Spector, 2011; Benz & Gotzner, 2014), suggesting that the experimental paradigm and the contexts evoked by these paradigms affect the probability of drawing embedded implicatures.
With the proposed project, we try to bring the discussion on embedded implicatures to a different level, addressing the following questions: (1) What are the contextual factors that facilitate the computation of embedded implicatures? (2) What role does the relevance and activation of alternatives play? (3) Which models can capture these contextual effects (e.g., Goodman & Stuhlmuller, 2013)? We will compare different manipulations of contextual factors such as the QUD manipulation and a manipulation of prosody which has been found to encourage implicature computation of simple sentences (e.g., Gotzner & Spalek, 2014). On a methodological level, we address the question which dependent measures and experimental designs are most appropriate to investigate (embedded) implicatures (see Degen & Goodman, 2014).