Daniele Panizza (GSGG, University of Gottingen): Entailment patterns in processing and development

This talk consists into two separate presentations (after each presentation there will be some time for questions, as well as a final discussion at the end of the talk). While these presentations illustrate two experimental works employing different techniques and investigating two different populations, there is one concept that is central to both works, which is the critical role played by the entailment patterns of the sentence (i.e. downward vs. upward entailment). Entailment patterns are logical properties displayed by natural languages, which are coded in their grammar and systematically affect the interpretation of scalar quantifiers (some) and numerals, and the licensing of Negative Polarity Items (NPIs). The first part of this talk is about the former issue whereas the second part will focus on the latter one.
First part.
A sentence such as 'two boys came' may be assigned an interpretation compatible with a situation in which more than two boys came ('at least' reading) or an interpretation that is not compatible with such a situation ('exact' reading). The proportion of 'exact' vs. 'at least' interpretations adopted by adults (Panizza, Chierchia, Clifton, 2011) is higher when numerals occur in declarative sentences, in which the 'exact' reading is logically stronger ('exactly two' entails 'at least two'), than in conditional antecedents, in which the 'at least' reading is stronger ('at least two' entails 'exactly two'). We argue that the mechanism that underlies this alternation of readings is a pragmatic Gricean-like inference that is sensitive to the entailment patterns of the proposition.
The aim of this work is to investigate whether 3-5-year-old children’s interpretation of numerals is affected by entailment patterns in the same way as adults' interpretation.
To test this, we designed an experiment in which a blindfolded puppet (Kermit the Frog) wanted to organize a race, and needed to find an adequate number of race contestants. Kermit asked some friends to find the contestants for him. In the critical scenario one of Kermit’s friends brought three animals (e.g. butterflies) when only two were actually required for the race. Kermit then stated "I remember what I told this boy I needed: two butterflies", stressing the minimum number of contestants that was required. Following this pragmatic boost, 16/32 children heard declarative sentences like (1), whereas the other 16 participants heard conditional sentences like (2).
(1) I think this boy brought two butterflies. Give him a coin if I'm right.
(2) If this boy brought two butterflies, give him a coin.
Children who rewarded the boy who brought three animals with a coin would show they had interpreted the numeral two as meaning 'at least two'. Our results show that young children’ interpretation of numerals is sensitive to entailment patterns in the same way as adults' interpretation.
Second Part.
The Italian word mai is an NPI that can mean either ever or never. When mai occurs in post-verbal position its distribution and interpretation is identical to that of ever. If mai occurs under negation or downward entailing contexts it is interpreted like ever, as in (3a). If it occurs in non-downward entailing contexts such as affirmative sentences like (3b) it generates the ungrammaticality of the sentence.

(3) a. Gianni non ha mai bevuto una birra.
Gianni hasn't ever drunk a beer.
b. *Gianni ha mai bevuto una birra.
Gianni has ever drunk a beer.

By contrast, when mai occurs in pre-verbal position, as in (4), it is capable of introducing a negation in the sentence, that is, it is interpreted as never.

(4) Gianni mai ha bevuto una birra.
Gianni never has drunk a beer.

However if it occurs in pre-verbal position but in a downward entailing context, as in (5) it is still interpreted with the existential meaning (i.e. ever)

(5) Se mai dovessi andare a Cuba, ci andrei da solo
If I should ever go to Cuba, I will go there alone

This tells us that the information about the entailment properties of the context in which it occurs is critical to determine how mai is interpreted. We conducted an Event Related Potential (ERP) experiment presenting participants sentences such as (4) to (3a-b). The ERP pattern that was elicited by these conditions provides us some insight as to how the processor accesses the information about the entailment pattern of the sentence in which mai occurs in order to assign the positive/negative interpretation to mai. In particular, in line with other works we conclude that an NPI violation Furthermore we also found that the condition in (4), in which mai is grammaticaly interpreted as never, elicits a pattern composed by a negativity (N400) followed by a frontal positivity (FP600), which is very similar to that elicited by (3b), as compared with (3a). We discuss these results and present a model to account for the interpretation of mai in real time and how it is affected by the entailment patterns of the sentence.