Why children and adults sometimes (but not always) compute implicatures

Abstract

Noveck (2001) argued that children even as old as 11 do not reliably endorse a scalar interpretation of weak scalar terms (some, might, or) (cf. Braine & Rumain, 1981; Smith, 1980). More recent studies suggest, however, that children's apparent failures may depend on the experimental demands (Papafragou & Musolino, 2003). Although previous studies involved children of different ages as well as different tasks, and are thus not directly comparable, nevertheless a common finding is that children do not seem to derive scalar implicatures to the same extent as adults do. The present article describes a series of experiments that were conducted with Italian speaking subjects (children and adults), focusing mainly on the scalar term some. Our goal was to carefully examine the specific conditions that allow the computation of implicatures by children. In so doing, we demonstrate that children as young as 7 (the youngest age of the children who participated in the Noveck study) are able to compute implicatures in experimental conditions that properly satisfy certain contextual prerequisites for deriving such implicatures. We also present further results that have general consequences for the research methodology employed in this area of study. Our research indicates that certain tasks mask children's understanding of scalar terms, not only including the task used by Noveck, but also tasks that employ certain explicit instructions, such as the training task used by Papafragou & Musolino (2003). Our findings indicate further that, although explicit training apparently improves children's ability to draw implicatures, children nevertheless fail to achieve adult levels of performance for most scalar terms even in such tasks, and that the effects of instruction do not last beyond the training session itself for most children. Another relevant finding of the present study is that some of the manipulations of the experimental context have an effect on all subjects, whereas others produce effects on just a subset of children. Individual differences of this kind may have been concealed in previous research because performance by individual subjects was not reported. Our general conclusions are that even young children (7-year olds) have the prerequisites for deriving scalar implicatures, although these abilities are revealed only when the conversational background is natural.

Notes

¹Scalar implicatures are cancellable. In fact, we can continue (1a) with denial of the implicature without incurring in a contradiction, as seen in (i):

 (i) Some students passed the exam. In fact, they all did. In this respect, scalar implicatures differ from entailments, which give rise to contradiction if they are negated.

 (ii) Some students passed the exam. In fact, none of them did.

²Adults’ responses gave rise to a bimodal distribution also in Noveck's study, but no attention was devoted to this fact.

³In this case, there was also a significant difference between absurd statement and appropriate statements introduced by some; between absurd statements and false statements introduced by all and between absurd statements introduced by all and appropriate statements introduced by some (Scheffé test, p<.05). In all these cases, the difference was due to the fact that the absurd statements were answered correctly 100% of the time and the other statements were answered correctly between 93 and 95% of the time.

⁴A less controversial training would have been one in which the terms used formed a scale under any model of implicatures (e.g., the connectives and/or).

⁵Based on these considerations, one might raise doubts on the view that training enhances subjects’ readiness to derive implicatures. It is possible that training merely serves to instruct subjects to search for statements that describe the observable facts more perspicuously, without necessarily leading them to really derive implicatures, that is, to recognise that by uttering some giraffes have a long neck the speaker intended to communicate that some, but not all giraffes have a long neck. Although it is hard to distinguish between the search of a more adequate description and the derivation of implicatures, these concerns should not be dismissed.

⁶Unfortunately, these authors only provide means and result from the ANOVA and do not tell us how subjects are distributed depending on the number of times they accepted under-informative statements. But we guess that subjects’ responses did not have a normal distribution.