Elsevier

Cognition

Volume 107, Issue 3, June 2008, Pages 1018-1034
Cognition

Three- and four-year-olds spontaneously use others’ past performance to guide their learning

https://doi.org/10.1016/j.cognition.2007.12.008Get rights and content

Abstract

A wealth of human knowledge is acquired by attending to information provided by other people – but some people are more credible sources than others. In two experiments, we explored whether young children spontaneously keep track of an individual’s history of being accurate or inaccurate and use this information to facilitate subsequent learning. We found that 3- and 4-year-olds favor a previously accurate individual when learning new words and learning new object functions and applied the principle of mutual exclusivity to the newly learned words but not the newly learned functions. These findings expand upon previous research in a number of ways, most importantly by showing that (a) children spontaneously keep track of an individual’s history and use it to guide subsequent learning without any prompting, and (b) children’s sensitivity to others’ prior accuracy is not specific to the domain of language.

Introduction

A tremendous amount of what humans learn is learned from other people. Indeed, social transmission, either implicit or explicit, is the main route through which humans learn about language, culture, science, religion, and human relations. Typically-developing children avidly seek out information provided by others, starting with their expressions and actions, and then, as they get older, learning from information conveyed through speech (for discussion, see Baldwin, 2000, Bloom, 2000, Tomasello et al., 2005).

Such reliance on other people poses a unique set of challenges, however. Although a baby is prudent to rely on the cues provided by his or her parent – being more willing to crawl over a visual cliff if the parent makes supportive expressions and sounds, for instance (Vaish & Striano, 2004) – it would be maladaptive for children and adults to indiscriminately trust all information that others provide. Humans make mistakes; they trick; they lie; they have different levels of knowledge and different areas of expertise; and they offer information even when they are uncertain (see Callanan, Sabbagh, Perez, & Cervantes, 1995). Hence, learners are faced with the task of determining when, and if, someone is a credible information source.

Sabbagh and Baldwin (2001) found that 3- and 4-year-olds are sensitive to explicit verbal cues about the credibility of a source. When taught novel words by a speaker who verbally conveyed his knowledge or ignorance about the word’s referent, (e.g., the speaker said, “You know, I’d like to help my friend Birdie and I know just which one’s her blicket.” versus “You know, I’d like to help my friend Birdie, but I don’t know what a blicket is. Hmmm. Maybe, it’s this one.”), they were more likely to learn the word from the speaker who claimed certainty. Relatedly, Chris Moore and his colleagues (Moore et al., 1989, Moore and Davidge, 1989) have found that 4-year-olds, but not 3-year-olds, are sensitive to explicit linguistic markers of epistemic states (e.g., ‘know’ versus ‘think’).

Can children use less explicit cues to credibility – such as the speaker’s prior history of being correct or incorrect – to determine who is the most credible? All else being equal, it is a useful heuristic to assume that someone who has been correct in the past is more likely to be correct in the future than someone who has been incorrect in the past. Do young children appreciate this and can they use this information to facilitate the learning process? Recent research suggests that they can.

Koenig, Clement, and Harris (2004) showed 3- and 4-year-old children video clips in which an actor asked questions of two informants, one who consistently responded correctly (e.g., calling a ball “a ball”) the other who consistently responded incorrectly (e.g., calling a ball, “a shoe”). Children were asked to identity the person who said something right and the person who said something wrong. The children then observed the actors label a single object with different names, and were asked which was correct. For instance, the previously correct speaker would call the object a “mido”, the previously incorrect one would call it a “toma”, and children would be asked: “Can you tell me what this is called, a mido or a toma?”. Koenig et al. (2004) found that 3-year-olds and 4-year-olds who passed the explicit questions regarding which informant was right/wrong were significantly more likely to choose what the accurate informant said as the label for the novel object, suggesting they understand that a person’s prior credibility is a useful cue to a person’s future credibility (see also Clément et al., 2004, Pasquini et al., 2007).

In follow-up studies, Koenig and Harris (2005) explored whether, in addition to endorsing information from the most credible source, children could also use a person’s past performance to make related decisions (such as choosing to seek out new information from the previously accurate person). In this procedure 3- and 4-year-olds were able to distinguish the accurate and inaccurate speakers when asked, ‘Which one was (not) very good at answering questions?’, but only the 4-year-olds used that judgment to: (a) predict their future assertions, (b) seek information from the more accurate speaker, and (c) endorse the more accurate speaker’s claims. Koenig and Harris (2005) also noted an important distinction between someone with a history of being inaccurate (e.g., calling a ball ‘a book’) and someone with a history of being ignorant (e.g., saying ‘I don’t know’ when asked what a ball is called). They found that when the alternative speaker was inaccurate (Experiment 1) 3-year-olds did not favor the previously accurate speaker, but when the alternative speaker was ignorant (Experiment 2), 3-year-olds directed their questions to, and preferred to learn from, the accurate speaker (see Koenig & Harris, 2005 for a discussion).

More recently, Jaswal and Neely (2007) provided another demonstration of children’s ability to use a person’s past performance when choosing the best source of information. They found that when given a choice to learn new words from either a child or an adult – both of whom had a history of being accurate – children tended to choose the words provided by the adult. However, when given a choice to learn new words from either a child who had a history of being accurate or an adult who had a history of being inaccurate, children preferred to learn the words provided by the child. In other words, prior accuracy trumped age.

For this sensitivity to previous accuracy to play an important role in real-world learning, however, at least four prerequisites must be met. First, children must be sensitive to others’ mistakes. That is, they will need to detect when others offer inaccurate information. Second, children need to possess at least an implicit appreciation that a person’s prior competence is a useful indicator of that person’s future competence. Third, they must be able to keep person-specific information about prior accuracy separate from accuracy information about other individuals and be able to use it to evaluate new information from those same individuals. Importantly, for these abilities to have much influence on real-world learning a fourth condition is essential: Children must spontaneously keep this track record and spontaneously use it to facilitate subsequent learning. It is not sufficient, after all, that they can capitalize on the fact that someone is ‘good’ or ‘not very good’ at answering questions when prompted by an experimenter; for this capacity to impact learning outside of the laboratory, children must spontaneously track and store this information and choose to use it in the normal course of social interaction.

The first prerequisite – a sensitivity to others’ mistakes – emerges relatively early (at least in the domain of language). Pea (1982) demonstrated that 24- to 30-month-old children will say “no” to speakers who mislabeled visible objects, and Koenig and Echols (2003) demonstrated that children as young as 16-months of age looked longer at speakers who mislabeled a series of common objects.

The second and third prerequisites – an implicit understanding of the relationship between past behavior and future behavior, and an ability to use person-specific information about prior accuracy to aid the learning of new information from that person – were shown in Koenig et al. (2004) study described above. They found that 3- and 4-year-olds who passed the explicit questions regarding which informant was right/wrong were significantly more likely to choose what the accurate informant said as the label for a novel object. These results suggest that young children appreciate, at some level, that a person’s prior competence can be a useful cue to a person’s future competence, and can use person-specific information to guide future learning (at least when they are prompted to explicitly note the person’s prior accuracy).

The fourth prerequisite has not yet been explored. None of the research to date has addressed whether children naturally hold person-specific information about previous accuracy in mind and spontaneously use it in the course of subsequent learning. In the studies described above, the children were always explicitly asked about the speaker’s performance prior to being tested on new items (e.g., the experimenter said, ‘One of these people was not very good at answering these questions. Which one was not very good at answering questions?). It is unclear whether children would utilize the speaker’s previous history to guide their learning if this priming was not provided. Indeed, Dunfield and Fitneva (2007) found that even after a single trial in which one person was correct and the other incorrect, 4-year-olds subsequently sought information from the previously correct speaker, but, they only did so when given a trait-like prompt such as ‘Which one was (not) very good at answering the question?’.

Moreover, it is unclear from the studies that have explored children’s sensitivity to a person’s prior accuracy whether the children are actually learning the new information. It is possible that in these studies, the child is simply siding with the individual who was correct in the past. For example, they might have adopted a heuristic such as “say whatever the person who is good at answering questions says”, or “ignore whatever the person who is not good at answering questions says” without actually learning the new word. For instance, children who were asked, “Can you tell me what this is called, a mido or a toma?”, usually responded with the name given by the previously accurate informant (“mido”), but they might have done so without storing “mido” in long-term memory; that is, they might have done so without learning the new word.

To address this concern, the experiments reported here include two sorts of test trials. There is a ‘Preference Condition’, similar to the studies described above, where the correct answer is the same answer given by the previously accurate speaker. But there is also a more difficult ‘Contrast Condition’. In this condition, two objects are placed in front of the child and the previously accurate speaker names one of them “a ferber” (for example), and the previously inaccurate speaker names the other object with the same name: “a ferber”. Children are then asked “Where’s the cheena?”, cheena being a word that they have not previously heard. We know from other research that children of this age are biased to assume that different words refer to different objects (i.e., the constraint of ‘mutual exclusivity’; see Markman & Wachtel, 1988; see also Byers-Heinlein and Werker, 2006, Golinkoff et al., 1994, Hall and Graham, 1999, Merriman and Bowman, 1989, Mervis et al., 1994). Hence, if children have fully encoded that “ferber” refers to the object labeled by the previously accurate speaker then they should make the inference that this new word “cheena” refers to the other object. This experiment addresses the robustness of the learning of this socially-transmitted information by capitalizing on children’s tendency to treat word labels as mutually exclusive.

In sum, the following experiments address whether 3- and 4-year-old children spontaneously keep track of an individual’s history of providing accurate or inaccurate information and use this person-specific information to guide their subsequent learning.

Section snippets

Participants

Forty children from middle-class families participated in the current study: Twenty 3-year-olds (8 Males; 12 Females; M = 41 months; range = 34 months to 47 months) and 20 4-year-olds (12 Males; 8 Females; M = 54 months; range = 49 months to 59 months). Sixty-three percent were White, 5% were Black, and 33% were Asian.

Materials

The materials consisted of four common objects, eight novel objects and two hand puppets (one child-like girl “Jenny”; one child-like boy “Ben”) with moveable mouths who served as the

Experiment 2

In Experiment 2, we were interested in whether children would spontaneously keep track of a person’s history of being accurate or inaccurate in a domain other than word learning – in this case, the typical functions of objects. The design here parallels that of Experiment 1, including the presence of the Contrast Condition. Here, however, there was no clear prediction about how children would respond in the Contrast Condition given that this experiment involved functions instead of words. The

General discussion

Our results revealed that 3- and 4-year-olds spontaneously keep track of others’ history of being accurate versus inaccurate. They are more likely to learn new words (Experiment 1) and new object functions (Experiment 2) from someone who has been accurate in the past than from someone who showed signs of incompetence (in this case by repeatedly being inaccurate about the words for, or typical functions of, common objects).

Experiment 1 replicated Koenig et al. (2004) previous finding that

Acknowledgements

This work was supported by a Social Sciences and Humanities Research Council of Canada (SSHRC) grant, start-up funds from the Human Early Learning Partnership (HELP) of British Columbia and a Peter Wall Institute for Advanced Studies Early Career Award to Susan A.J. Birch. We are grateful to Smaranda Luca and Kristen Frampton for their assistance with data collection and recruiting, and to the editor and 3 anonymous reviewers for their helpful comments.

References (47)

  • Brosseau-Liard, P., & Birch, S. A. J. (2007). How children’s evaluations are influenced by an individual’s past...
  • Byers-Heinlein, K., & Werker, J. F. (2006). Word-learning at 18-months: A comparison of monolingual and bilingual...
  • Callanan, M. A., Sabbagh, M. A., Perez, D. R., & Cervantes, C. A. (1995). Names for objects in conversations between...
  • K. Casler et al.

    Young children’s rapid learning about artifacts

    Developmental Science

    (2005)
  • F. Clément et al.

    The ontogenesis of trust

    Mind & Language

    (2004)
  • J.H. Danovitch et al.

    Should you ask a fisherman or a biologist?: Developmental shifts in ways of clustering knowledge

    Child Development

    (2004)
  • G. Diesendruck et al.

    Children’s avoidance of lexical overlap: A pragmatic account

    Developmental psychology

    (2001)
  • G. Diesendruck et al.

    Two-year-olds’ sensitivity to speakers’ intent: An alternative account of Samuelson and Smith

    Developmental Science

    (2004)
  • K. Duncker

    On problem solving

    Psychological Monographs

    (1945)
  • Dunfield, K. A., & Fitneva, S. A. (2007). Selective information seeking: From informant accuracy to informant...
  • G. Gergely et al.

    The social construction of the cultural mind: Mitative learning as a mechanism of human pedagogy

    Interaction Studies: Social Behavior and Communication in Biological and Artificial Systems

    (2004)
  • T.P. German et al.

    Functional fixedness in a technologically sparse culture

    Psychological Science

    (2005)
  • T.P. German et al.

    Immunity to functional fixedness in young children

    Psychonomic Bulletin & Review

    (2000)
  • Cited by (315)

    View all citing articles on Scopus
    View full text