Why some children accept under-informative utterances

TL;DR: This article showed that children who accept under-informative utterances are in fact competent with implicature but do not consider pragmatic violations grave enough to reject the critical utterance.

Abstract: Binary judgement on under-informative utterances (e.g. Some horses jumped over the fence, when all horses did) is the most widely used methodology to test children’s ability to generate implicatures. Accepting under-informative utterances is considered a failure to generate implicatures. We present off-line and reaction time evidence for the Pragmatic Tolerance Hypothesis, according to which some children who accept under-informative utterances are in fact competent with implicature but do not consider pragmatic violations grave enough to reject the critical utterance. Seventy-five Dutch-speaking four to nine-year-olds completed a binary (Experiment A) and a ternary judgement task (Experiment B). Half of the children who accepted an utterance in Experiment A penalised it in Experiment B. Reaction times revealed that these children experienced a slow-down in the critical utterances in Experiment A, suggesting that they detected the pragmatic violation even though they did not reject it. We propose that binary judgement tasks systematically underestimate children’s competence with pragmatics.

Summary

1. Introduction

• In the case at hand, the hearer may reason that by only saying some apples the speaker also implies that ‘not all’ of the apples are in the basket.
• Noveck (2001) found that 8- and 10-year-old children were not rejecting a sentence with a weaker term on a scale in favour of the more informative term.
• Currently, there is ample evidence that when participants are rejecting under-informative sentences, they require more time to do so than when rejecting logically false ones.

2.2.2 Experiment B

• The ternary judgement task was created on the same principles as the binary judgement task.
• She informed the participants that she liked strawberries and that in this game, the participants should reward her with a small strawberry, a medium-sized strawberry, or a large one, in correspondence to how well they thought her sentence described the picture.
• After the interval, the participants continued with the graded judgement task, which took also about five minutes.
• Based on the results from the binary judgement task, one could conclude that there is a clear-cut distinction between children, with some being consistently competent with implicature (n = 22), some being in transition (n = 33), and some consistently lacking competence with implicature (n = 20).
• Such interpretations for this kind of data are consistently found in the literature (see, e.g., Guasti et al. 2005).

3.1.2 Experiment B

• In the graded judgement task, the same participants rewarded sentences with a large, medium, or small strawberry .
• No clear strategy is seen for the under-informative condition.
• The non-competent group only includes children who accepted all under- informative statements in the binary task and always awarded them the top reward, a large strawberry in the ternary task.
• Next, the authors compare the RTs for rejecting under-informative sentences with the RTs for rejecting false sentences (with one object) across pragmatically competent children.

4. Discussion

• The first objective of this study was to test whether evidence for Pragmatic Tolerance can be found with ad hoc scales within the same subjects.
• The authors therefore predicted that pragmatically tolerant children take longer to respond to under-informative sentences (for which an inference has to be made) compared to logically true sentences (for which no additional inference is needed).
• No such confusion was present for the true conditions, which were always completely true, or for the false condition with one object, which was completely false.
• In the present paper the authors reported response accuracy and, for the first time, reaction time evidence, showing that this is not the case.
• The authors should clarify that they do not suggest that all research that has used the binary judgement task to date is deficient.

Figures

Figure 1. The 3-point rating scale from the ternary judgement task

Figure 5. Distribution of under-informative statements that were rewarded with the top reward (large strawberry) by the twenty participants who accepted all statements in the binary task

Table 1. Predictions of behaviour towards under-informative statements.

Table 2. Conditions in the binary judgement task

Figure 4. The distribution of small, medium, and large strawberries awarded in each of the different conditions. The number in parentheses refers to the number of objects in the display.

Figure 3. Distribution of acceptances of under-informative statements

Table 3. Reaction times for the accepters of under-informative statements

Figure 2. Percentage of correct responses in the different conditions. The number in parentheses refers to the number of objects in the display. Note that in this paper, we label the rejection of an under-informative statement as a correct response.

Figure 7. Average age for each of the pragmatic competence categories

Figure 6. Average response times in the different conditions. The number in parentheses refers to the number of objects in the display.

Full text

1
Why some children accept under-informative utterances: Lack of competence or
Pragmatic Tolerance?
Alma Veenstra, University of Cambridge, UK
Bart Hollebrandse, University of Groningen, Netherlands
Napoleon Katsos, University of Cambridge, UK
Abstract
Binary judgement on under-informative utterances (e.g. Some horses jumped over the fence,
when all horses did) is the most widely used methodology to test children’s ability to generate
implicatures. Accepting under-informative utterances is considered a failure to generate
implicatures. We present off-line and reaction time evidence for the Pragmatic Tolerance
Hypothesis, according to which some children who accept under-informative utterances are in
fact competent with implicature but do not consider pragmatic violations grave enough to
reject the critical utterance. Seventy-five Dutch-speaking four to nine-year-olds completed a
binary (Experiment A) and a ternary judgement task (Experiment B). Half of the children who
accepted an utterance in Experiment A penalised it in Experiment B. Reaction times revealed
that these children experienced a slow-down in the critical utterances in Experiment A,
suggesting that they detected the pragmatic violation even though they did not reject it. We
propose that binary judgement tasks systematically underestimate children’s competence with
pragmatics.
Keywords: implicature, pragmatic development, under-informative utterances, reaction times

2
1. Introduction
According to Grice (1975 [1989]), interlocutors are expected to adhere to the Cooperative
Principle and a set of maxims that makes their conversation as efficient as possible.
Cooperative speakers will comply with the Maxims of Quantity, Quality, Relation, and
Manner, which enjoin them to say no more and no less than what is needed, to be truthful,
relevant, concise and unambiguous. In this paper we will focus on the development of
children’s competence with the Maxim of Quantity and especially its first sub-maxim, which
prescribes speakers to make their contribution as informative as is required.
Numerous studies in the field of pragmatics report that children are less competent using this
maxim to derive pragmatic inferences (known as implicatures) than adults. In a high-profile
paper that summarizes the emerging findings, Noveck and Reboul (2008) describe the
development of informativeness as moving from broad semantic interpretations to narrowed
pragmatic interpretations. The statement some apples are in the basket simply means that two
or more apples are in the basket. A narrowed interpretation based on the Maxim of Quantity
involves the evaluation of the speaker’s intentions and of other, more informative options:
what was said, and what else could have been said? If all apples were in the basket, then a
cooperative speaker would have stated the more informative proposition by saying all apples
are in the basket. Therefore, the fact that they did not use the more informative proposition
(assuming that they could have done so) suggests that this is not the case. In the case at hand,
the hearer may reason that by only saying some apples the speaker also implies that ‘not all’
of the apples are in the basket. This interpretation is typically known as a quantity
implicature. Moreover, because the implicature involves expressions that can form a scale of
informativeness (like ‘some and all), it is further known as a scalar implicature (see Geurts
2010).

3
In terms of child development, Noveck (2001) used a binary judgement paradigm, where
participants are presented with an under-informative statement and are asked to make a binary
judgement in terms of truth or falsity (or correctness, in other studies). Noveck (2001) found
that 8- and 10-year-old children were not rejecting a sentence with a weaker term on a scale in
favour of the more informative term. For example, the children judged sentences like some
giraffes have long necks as true, despite the fact that all giraffes have long necks. Importantly,
these children did not have problems rejecting sentences in general, as they were adult-like at
rejecting sentences that were logically false. Papafragou and Musolino (2003), who also used
a binary judgement paradigm, tested 5-year-old children’s performance on quantified,
numerical, and verbal scales, and found that children predominantly accepted under-
informative statements. Training and explicit instruction improved the children’s
performance, increasing the success on the numerical scales (but not the quantified or verbal
ones) to a near-ceiling 90%. Feeney and colleagues (2004) found that manipulation of the
relevance of the implicature also enhances rejections in 7-year-olds. Guasti and colleagues
(2005) studied children’s pragmatic competence further in line with Papafragou and
Musolino’s (2003) training and explicit instructions. They found that the performance of 7-
year-olds did improve, but that the effect did not persist over a longer period of time. Barner,
Brooks and Bale (2011) and Papafragou and Tantalou (2004) also used binary judgements
and further demonstrated that children’s performance does not differ when other types of
under-informative sentences are tested (e.g., when the sentence the cow and the dog are
sleeping describes a situation where a cat is sleeping as well). Foppolo, Guasti and Chierchia
(2012) demonstrated the role of different lexical choices (partitive vs. non-partitive
quantifiers) and the salience of the more informative expression that was not used (e.g. all),
while Skordos and Papafragou (2016) demonstrated the role of contextual relevance of the

4
more informative expression, in both cases by studying children’s rates of rejection of under-
informative utterances.
These and other papers have significantly advanced our understanding of the factors that are
relevant when children (and adults) decide how to judge the felicity of an under-informative
utterance. What is important for the present investigation is that these studies base their
conclusions on the assumption that the children who accept under-informative utterances
altogether lack (some or all of the) competence that is needed to generate implicatures. It is
beyond the scope of this paper to review exactly what kind of competence children who
accept under-informativeness lack (this is a theory-critical matter which has generated a lot of
attention and some controversy, see Katsos 2014). What we highlight is that they all assume
that these participants lack some kind of competence and that this prevents them from
generating implicatures.
However, Davies and Katsos (2010), Katsos and Smith (2010), and Katsos and Bishop (2008,
2011) have argued that children’s acceptance of under-informative sentences may have a
different origin. Katsos and Bishop (2011) used a ternary task in addition to a binary task, and
found that children were much more reluctant to endorse under-informative sentences in the
former compared to the latter. That is, while 5-year-old children accepted under-informative
utterances when given a binary choice (to decide if the utterance was right or wrong), they
penalised the utterances in a ternary scale (to decide whether to award a small, medium or
large strawberry), by giving the medium or the small rather than the top award. This finding
gave rise to the Pragmatic Tolerance Hypothesis: children are sensitive rather than blind to
violations of the Maxim of Quantity, but they do not always take the violation as severe
enough to warrant a downright rejection of the critical sentence. Once children are tested in a
paradigm that does not require them to demonstrate their competence by the categorical

5
rejection of pragmatically infelicitous sentences, then children’s true competence may be
evident at ages much younger than the ones tested up to now.
Katsos and Bishop (2011), and Katsos and Smith (2010), who reached similar conclusions,
tested two different groups of 5-year-old children: one that was given a binary judgement
task, and one that was given a ternary judgement task. Davies and Katsos (2010) tested
sensitivity to over- and under-informativeness with a binary and a graded judgement task
within the same groups of participants, and showed that children who seemed insensitive in
the binary task did display sensitivity in the graded task.
In this paper, we have two aims: First, we aim to replicate the evidence in support of
the Pragmatic Tolerance Hypothesis by using a binary and graded judgement task within the
same group of participants with a new set of stimuli, ad hoc scales. We propose the
developmental hypothesis that children are competent with implicature, and by extension,
with some crucial aspects of Gricean pragmatics, at a much younger age than previously
considered. This is anticipated by several studies in word-learning and mutual exclusivity,
which propose that children as young as two years old employ Gricean-like principles to infer
the likely referent of a novel word (see Clark 1990, among others; for an opposite view of
children’s competence with mutual exclusivity, see De Marchena et al. 2011).
Second, we wish to advance our understanding of the process through which pragmatically
competent children tolerate pragmatically under-informative sentences. Currently, there is
ample evidence that when participants are rejecting under-informative sentences, they require
more time to do so than when rejecting logically false ones. Relevant data were first presented
by Noveck and Posada (2003), who looked at reaction times to under-informative statements
in adults. Similar data are reported by Bott and Noveck (2004) and De Neys and Schaeken
(2007). The studies by Noveck and colleagues also report that rejecting under-informative
sentences tends to take more time than accepting them.