A Developmental Investigation of Processing Costs in Implicature Production

Abstract

Much developmental work has been devoted to scalar implicatures. These are implicitly communicated propositions linked to relatively weak terms (consider how Some pragmatically implies Not all) that are more likely to be carried out by adults than by children. Children tend to retain the linguistically encoded meaning of these terms (wherein Some is compatible with All). In three experiments, we gauge children's performance with scalars while investigating four factors that can have an effect on implicature production: (i) the role of (the presence or absence of) distractor items; (ii) the nature of the task (verbal judgments versus action-based judgments); (iii) the choice of scalar expression (the French quantifier quelques versus certains); and (iv) the type of scale that contextualizes the weak utterance (the affirmative All versus the negative None). Experiment 1 replicated earlier findings showing that 9-year-olds are more likely than adults to consider as true statements such as Some turtles are in the boxes (uttered when all turtles are in the boxes) while employing the quantifier certains in a truth evaluation task containing multiple distractor items. The task in Experiment 2 increased implicature production across all ages (4-, 5-, and 7-year-olds as well as adults) but maintained the developmental effect while using quelques in an action-based task containing no distractor items. Experiment 3 showed that 9-year-olds are more likely to produce implicatures with quelques than they are with certains in the action task while adults are not affected by the choice of term. Overall, these results identify seemingly harmless task features that can prevent even older children (9-year-olds) from carrying out implicatures (e.g., through the inclusion of distractors) while also showing how implicature production among even young children (4- to 5-year-olds) can be facilitated by task features (e.g., the use of an action task) and without the introduction of special training.

ACKNOWLEDGMENTS

The authors wish to thank Lewis Bott, Bart Geurts, Julien Musolino, Dan Sperber, Michael Tomasello, and two anonymous reviewers with respect to an earlier version of this article as well as to the RD&P team in Lyon for their helpful suggestions on all phases of the work. This work was supported by a grant from the CNRS (ATIPE) awarded to the second author (IN) and was conducted while the third author (GP) was affiliated with Cognition & Usages, Université de Paris-8.

Notes

¹ Papafragou and Musolino's (2003) study also investigated number terms (e.g., a puppet would use “three” to describe a situation where a larger number would have been more appropriate). They showed that 5-year-olds ultimately deal differently with number words than they do with other weak terms in that cardinals are more likely to get an adult treatment (even though the children are not completely adult-like). This finding confirms the prevailing view in recent linguistic theory, according to which number words do not have a scalar semantics (Horn (1992), Carston (1998), Geurts (1998; 2006), among others).

² Papafragou and Tantalou (2004) aim to show that 5-year-olds can be encouraged to produce scalar implicatures at adult levels. However, their data are based on a nonstandard paradigm (i.e., it gives participants no justifiable reason to accept the minimal interpretation of a term such as “some”). Moreover, much of the study's claims are based on children's self-reports and these lead to the conclusion that at most 56% of Papafragou and Tantalou's participants derived scalar implicatures.

³For further discussion see the Journal of Child Language (issue 2, 2004). For a detailed discussion of the limitations of the truth judgment evaluation task see also Musolino (2006) as well as Musolino and Lidz (2006).

⁴We point out straightaway, to avoid misunderstanding, that despite appearances the meaning of the French word “certains” is not equivalent to the English “certain.” For a semantic analysis of the French terms certains and quelques see Corblin (2001) and Gondret (1976).

⁵For a criticism of experimental designs using sentence judgments in investigating scalar implicatures in children, see also Guasti et al. (2005).

⁶Note that the familiarization technique used here, and in Experiments 2 and 3, merely confirms that the participants appreciate the details of the scenario. It is not comparable to the training given in Papafragou and Musolino (2003, Experiment 2) nor to the one used in Guasti et al. (2005, Experiment 2).

⁷All the responses in the tasks to be described in this article can be of one of two types; thus references to predictions based on chance are .5 here and elsewhere.

⁸One will notice that the boxes are quantified here, unlike in Experiment 1 where objects in and around the boxes were. Either formulation is technically possible (compare I would like all tokens to be in a box vs. I would like all boxes to contain a token). We adopted the latter because the one-to-one relation between a token and box is clearer with “boxes containing a token” than it is with “tokens being in a box.” We are assuming that participants' behavior is unaffected by this modification.

⁹ Certains is a partitive, but it isn't necessarily a specific indefinite. Besides, as mentioned earlier, it is important to bear in mind that it does not have the same function as the English certain.

¹⁰French is by no means the only language to have pairs of indefinite determiners where English only has some: Greek, Italian, Dutch (de Jong and Verkuyl (1985), de Hoop (1995)), Turkish (Enç (1991)), and Rumanian, among other languages, display the same pattern.

¹¹We are using the nomenclature of the American system; in French, these would be referred to as Maternelle moyenne section, Maternelle grande section, and Cours Elémentaire 1 (CE1).

∗This case is technically like the Some utterance in the All scenario, where no action can be taken to be a logical interpretation and an action (to add) would be considered pragmatic. However, this utterance involves a negatively quantified statement, making it prima facie difficult to process.

¹²An action here would be the result of an implicature because, technically speaking, no action is called for (two boxes do contain a token). As shown and argued elsewhere (Papafragou and Musolino (2003)), however, numbers are considered exceptional and prompt very high rates of implicature production in experimental tasks. Indeed, participants removed a token here without hesitation.

∗The values for the “keep as is” response in the subset condition are interpreted strictly; even harmless changes (e.g., adding a token to a box) are considered violations. However, the values in parentheses show the “keep as is” responses when taken to mean “keep the truth value the same.”

¹³We also conducted a second analysis to verify that the logical-to-pragmatic age transformation stands with a stricter analysis, i.e., we included only those participants (children and adults) who made no harmless changes upon hearing the Some utterance in the Subset scenario. This analysis also reveals that logical responses decrease significantly with age, z = 2.12, p = .017 (one-tailed) and in a near-monotonic order (36%, 17%, 11%, and 12% for 4-year-olds, 5-year-olds, 7-year-olds, and adults, respectively).

∗The values in parentheses are “keep as is” responses when responses are interpreted as “essentially, keep as is” because the changes made did not affect truth values.