Scalar Implicatures in Child Language: Give Children a Chance

Abstract

Children's pragmatic competence in deriving conversational implicatures (and scalar implicatures in particular) offers an intriguing standpoint to explore how developmental, methodological, and purely theoretical perspectives interact and feed each other. In this paper, we focus mainly on developmental and methodological issues, showing that children from age 6 on are adult-like in deriving the scalar implicature related to the scalar quantifier some (i.e. they interpret some as some but not all), while children at age 4 and 5 only sometimes reject underinformative-some in a classical Truth Value Judgment Task (Experiment 1). They do so despite their excellent performance in pragmatic tasks that evaluate their competence with the rules of talk exchange, such as the Conversational Violations Test (Experiment 4) and the Felicity Judgment Task (Experiment 5). To give children a better chance to reject underinformative-some when all is at stake, we manipulated the experimental design and materials in three different ways: 1) in Experiment 2, we tested the partitive alcuni dei (some of) instead of the bare quantifier qualche (some); 2) in Experiment 3, we attempted to prime the scale <some, all> by asking children to judge a correct statement with all before the critical underinformative statement with some; 3) in Experiment 6, we aimed at making children more aware of the ambiguity of some, between its basic meaning (at least some, possibly all) and its strengthened meaning (some but not all). A surprising improvement is recorded in the last experiment, in which the rejection of underinformative-some by 5-year-old children rose to 72.5% (it was 42% in Experiment 1). We suggest that the children's low performance with scalar inference might be linked to the interplay of different factors as in the development of other general cognitive abilities, such as the ability to change one's strategy (Shallice, 1982) or to shift one's perspective (Gopnik & Rosati, 2001), the maturation of the lexicon (Barner & Bachrach, 2010), and especially their great sensitivity to the task, methodology and materials used to test their pragmatic abilities.

ACKNOWLEDGMENTS

We are grateful to Luca Surian for allowing us to use the Conversational Violations Test. We also thank Joke De Lange, Chiara Iogha and Marco Bernardi for their generous help in testing children and Claudia Caprin for her help with the statistical analysis on preliminary data. Participation in the Bicocca RRReading-Group was also fruitful (as well as enjoyable); special thanks go to Marco Marelli and Davide Crepaldi for sharing their expertise. We are also indebted to the three anonymous reviewers and the editors of this journal, for their sharp and fruitful suggestions that have contributed greatly to the final shape of this paper. Lastly, we cannot forget all the children, parents, directors and staff of the schools that participated in these studies; we send them our utmost gratitude for their kind cooperation.

Notes

¹ Although we think that this manipulation interestingly shows that children are capable of some sort of early understanding of pragmatic implicature in this task, we believe it unclear whether we can interpret children's refusal in this task as a definite indication that they have really derived the SI related to some or whether they are simply contrasting alternative descriptions. Given that the puppet is explicitly asked if she colored the stars, a simple “yes” answer would suffice to get the prize. Answers other than “yes” may suggest that the task was not completed, independently from the calculation of the SI related to the scalar term used in the response (see also Papafragou & Tantalou, 2004, p. 76, footnote 6).

² This possibility was suggested by one of the reviewers.

³ Qualche cannot be used in its partitive form; one possibility is to use qualcuno dei, but its use is controversial.

⁴ Statistical analysis, by means of logistic regression, revealed that children performed differently from adults in the control condition in which a piece of was used (χ²(1) = 6.89, p < .0087): contrast estimate results show that errors are about 9.75 times more common in children than in adults in this condition. Taking a closer look at how children responded, we see that in fact some children rejected true statements containing un pezzo (a piece of) because they interpreted this item in a “numerical” sense. For example, in a situation in which a character had built a piece of of the puzzle (and, crucially, not the whole puzzle), some children said that the puppet was wrong because the character had used three pezzi (pieces) of the puzzle and not just one. In Italian, in fact, there's no phonological distinction between un pezzo (lit. a part of) in which un is used as an indefinite article, and un pezzo (lit. one piece of) in which un is used as the numeral “one”. The ambiguity between un pezzo (part) and un pezzo (piece) might have led some children to take the expression un pezzo in its numerical sense, thus causing them to reject the otherwise true proposition. For this reason, leaving aside the data on a piece of, no difference is observed between children and adults in the incidence of correct responses in control trials.