The Role of Speaker Knowledge in Children's Pragmatic Inferences

Abstract

During communication, conversational partners should offer as much information as is required and relevant. For instance, the statement "Some Xs Y" is infelicitous if one knows that all Xs Y. Do children understand the link between speaker knowledge and utterance strength? In Experiment 1, 5-year-olds (N = 32) but not 4-year-olds (N = 32) reliably connected statements of different logical strength (e.g., "The girl colored all/some of the star") to observers who were fully or partially informed. Four-year-olds' performance improved when observer knowledge could be assessed more easily (Experiment 2a, N = 25) but remained the same in a nonlinguistic version of Experiment 1 that preserved the epistemic requirements of the original study (Experiment 2b, N = 26). These findings have implications for the development of early communicative abilities.

Four‐year‐olds incorporate speaker knowledge into pragmatic inferences

“…In fact, 5‐year‐olds performed at an adult‐like level in this new task. The findings of the present study lower prior estimates of the age at which children display the ability to take the epistemic step during scalar inference derivation (Hochstein et al, ; Papafragou et al, )—and are consistent with other studies with very simple paradigms that show an ability to compute scalar inferences (without special attention to the epistemic component) at young ages (Stiller et al, ).…”

“…Grice () proposed that pragmatic principles, including the maxim of quantity, extend to other types of cooperative communication (see also Sperber & Wilson, ), but relevant evidence so far in children and adults is limited (Bass, Bonawitz, & Gweon, ; Gweon & Asaba, ; Papafragou et al, ). In a previous study, 4‐year‐old children failed to incorporate speaker knowledge during scalar inference in a non‐linguistic paradigm adapted from a linguistic paradigm; however, they also failed in the linguistic version (Papafragou et al, ). In Experiment 3, we revisit the question of whether children (and adults) adapt to speaker knowledge during pragmatic inference in non‐linguistic exchanges using a simple task.…”

“…Available experimental evidence suggests that children are not able to incorporate information about a speaker's epistemic state into the derivation of scalar implicatures before the age of five, and even then, performance is highly fragile (Barner, Hochstein, Rubenson, & Bale, ; Hochstein, Bale, Fox, & Barner, ; Papafragou, Friedberg, & Cohen, ). In a study by Papafragou et al (), children were presented with videos of two observers, one with full knowledge of an event that occurred (e.g., a girl coloring a star) and one who fell asleep part way through. They then heard either a more‐informative or a less‐informative statement describing the event (e.g., ‘The girl colored some/all of the star’) and had to attribute the statement to either the limited‐knowledge or the full‐knowledge observer.…”

“…In the studies just reviewed, children had to watch lengthy scenarios while tracking the epistemic state of multiple speakers and the actions ofmultipleagents. These high task‐specific cognitive processing demands may have masked 4‐year‐olds’ ability to incorporate epistemic state into scalar implicature derivation (see Papafragou et al, for discussion of this possibility, also Pouscoulous et al, ). Additionally, Barner et al () argument that 4‐year‐olds compute scalar implicatures in the absence of epistemic reasoning hinges on the assumption that the ignorance implicature task in which children fail is minimally different (and only in terms of epistemic computations) from the ad hoc scalar implicature task that children pass.…”

“…The speaker had complete visual access to the contents of one box (and hence complete knowledge) but only partial visual access to the contents of the second box (hence partial knowledge); crucially, the boxes did not differ from the participant's perspective. We expected, following the logic of Papafragou et al (), that, if children and adults derive pragmatic inferences using a Gricean approach, they should use epistemic reasoning to link the more‐informative statement to the speaker under conditions of complete knowledge and the less‐informative statement to the speaker under conditions of partial knowledge. If children are inflexible or insensitive to the speaker's perspective, they would be expected to choose between the boxes at chance or rely on other potential biases (perhaps a preference for the fully knowledgeable speaker; see, e.g., Hochstein et al, ; Ozturk & Papafragou, ).…”