The role of social interaction and pedagogical cues for eliciting and reducing overimitation in preschoolers

https://doi.org/10.1016/j.jecp.2013.12.012Get rights and content

Abstract

The tendency to imitate causally irrelevant actions is termed overimitation. Here we investigated (a) whether communication of a model performing irrelevant actions is necessary to elicit overimitation in preschoolers and (b) whether communication of another model performing an efficient action modulates the subsequent reduction of overimitation. In the study, 5-year-olds imitated irrelevant actions both when they were modeled by a communicative and pedagogical experimenter and when they were modeled by a non-communicative and non-pedagogical experimenter. However, children stopped using the previously learned irrelevant actions only when they were subsequently shown the more efficient way to achieve the goal by a pedagogical experimenter. Thus, communication leads preschoolers to adapt their imitative behavior but does not seem to affect overimitation in the first place. Results are discussed with regard to the importance of communication for the transmission of cultural knowledge during development.

Introduction

Imitation is a powerful mechanism that allows humans to learn novel actions from others (Meltzoff, 1988). In contrast to emulation, which is accomplished by copying the end state of an action without performing the observed action steps, imitation entails copying the action sequence itself (Whiten, McGuigan, Marshall-Pescini, & Hopper, 2009). Although in many situations imitation is a quick and efficient learning tool, in other situations copying the exact actions observed in others is quite inefficient. For instance, in a study by Horner and Whiten (2005), the experimenter performed relevant actions as well as irrelevant actions while demonstrating how to retrieve a reward from a puzzle box to wild-born chimpanzees. When the box was opaque, chimpanzees imitated both kinds of actions. When the box was transparent, thereby revealing that irrelevant actions had no effect, chimpanzees employed a more efficient strategy of emulation and omitted the irrelevant actions.

In contrast to chimpanzees, human children and adults tend to faithfully imitate actions that are not the most efficient way to accomplish a certain aim (Flynn and Smith, 2012, Horner and Whiten, 2005, McGuigan et al., 2011). The imitation of causally goal-irrelevant actions has been termed overimitation (Lyons, Young, & Keil, 2007). The phenomenon is usually studied by showing participants, most often preschoolers, how to retrieve a reward from a novel, causally transparent container by using one or more irrelevant actions and one relevant action. After observing the model, participants typically reproduce both the causally relevant and irrelevant actions, thereby adopting an inefficient strategy. Crucially, this strategy is not spontaneously performed when participants operate the container without observing a model first (Lyons et al., 2011, Lyons et al., 2007).

There is currently much debate about why overimitation occurs. Lyons et al. (2007) argued that children automatically encode observed actions as causally relevant and, therefore, reproduce them. This process has been dubbed automatic causal encoding (ACE). The ACE claim is based on the observation that children overimitate even if they are explicitly encouraged to omit any unnecessary actions and even when performing the irrelevant actions ultimately endangers receiving a reward (Lyons et al., 2011). Others have argued that social norm learning and/or the desire to affiliate with the experimenter underlie the phenomenon of overimitation (Kenward et al., 2011, Keupp et al., 2013, Nielsen and Blank, 2011). Kenward (2012) had 3- and 5-year-olds observe an experimenter perform relevant actions as well as unnecessary actions in the presence of a puppet. Most children protested, some of them using normative language, when the puppet subsequently performed the task but omitted the unnecessary actions.

Neither norm learning, nor social affiliation, nor the ACE hypothesis can be ruled out at the moment. Regardless of which of these accounts holds true, some have suggested that overimitation results from children expecting others to teach them how something is done (Gergely & Csibra, 2006). Because the primary goal of the current study was not to distinguish among norm learning, social affiliation, and the ACE hypothesis, “how something is done” may henceforth refer to social norms as well as causal necessities and functional properties of artifacts.

According to the theory of natural pedagogy, humans have evolved mechanisms to transmit generic knowledge through communication (Csibra & Gergely, 2011). This generic knowledge is supposed to be shared by all members of a social group and may entail, for instance, knowledge about the functions of tools as well as cultural norms and rituals that are often cognitively opaque (Kiraly, Csibra, & Gergely, 2013). According to this theory, the (usually adult) teacher addresses the child in pedagogical interactions using certain ostensive cues such as eye contact, calling the learner’s name, and speaking in a child-directed manner. These ostensive signals prompt the expectation in learners that they are about to be taught relevant and generic information that can be generalized across situations and other individuals. Several studies have demonstrated infants’ and children’s sensitivity to these ostensive signals and the effects of communication on early learning (e.g., Gergely et al., 2007, Topal et al., 2008).

Corroborating the theory of natural pedagogy, some empirical evidence shows that social interaction and communicative cues increase imitative behavior in infants (Brugger et al., 2007, Kiraly et al., 2013, Nielsen, 2006). In a study by Kiraly et al. (2013), 14-month-olds performed a head touch to turn on a light more frequently after observing a communicative model demonstrating this novel and relatively inefficient action than after incidentally observing a non-communicative model. The authors proposed that direct communication and ostensive signals may support overimitation in older children as well. However, there is evidence that toddlers (i.e., 24-month-olds) may actually rely less on communicative cues than slightly younger infants (18-month-olds) when copying specific actions as opposed to action outcomes (Nielsen, 2006). In that study, 24-month-olds, but not younger infants, tended to copy specific actions irrespective of whether the model had interacted with them or not. Shimpi, Akhtar, and Moore (2013) reported that when the model is unfamiliar, direct interaction can even suppress the imitation of arbitrary object-directed actions in 18- and 24-month-olds when compared with the observation of a third-party interaction.

It has been suggested that the importance of communicative cues directed at the participant may decline from infancy to preschool age (Lyons et al., 2011, McGuigan et al., 2011). Yet, the role of the model’s communicative behavior in overimitation studies with preschoolers is currently unclear because, to our knowledge, no study so far has directly compared children’s imitation of obviously irrelevant actions performed by a pedagogical model compared with a completely non-communicative model. In a study by Nielsen, Moore, and Mohamedally (2012), the model did not demonstrate the actions to the child directly but rather demonstrated the actions to another adult (explicitly expressing his intention to “show [someone] how to use this”). Children imitated irrelevant actions even though some of them had already discovered a more efficient way of achieving the goal. In that study, the knowledgeable model communicated with the child prior to the demonstration and ostensive signals were transmitted, although they were directed at another individual. In another study on overimitation in children and adults, participants watched a video-recorded presentation of relevant and irrelevant actions, but an experimenter instructed them to watch closely because they were going to “have a go in a minute,” thereby also establishing a pedagogical context in which participants were supposed to learn from others (McGuigan et al., 2011). The question remains open whether preschoolers imitate causally irrelevant actions demonstrated by a completely unfamiliar and non-communicative experimenter in the absence of any instruction to learn how to perform an action or how to use a novel object.

Furthermore, it is currently unclear whether children’s omission of previously learned irrelevant actions and their adoption of more efficient strategies depend on the communicative context. According to the natural pedagogy account, children should expect pedagogically transmitted knowledge to be generalizable and shared among members of a social group. The subsequent presentation of an efficient strategy by a non-communicative model, therefore, should not lead to a switch in strategies. A communicative and pedagogical second model may, in contrast, be able to teach children the efficient action as a second strategy. The latter assumption is based on the previous finding that preschoolers are able to flexibly shift between different strategies of retrieving a reward after social demonstration (Wood, Kendal, & Flynn, 2013).

In the current study we tested (a) whether communication of a model performing irrelevant actions is necessary to elicit overimitation in preschoolers and (b) whether communication of another model performing an efficient action modulates the subsequent reduction of overimitation. In Phase 1 of the current experiment, 5-year-olds observed either a communicative experimenter showing them causally relevant actions as well as clearly irrelevant actions to retrieve a reward from a transparent container (pedagogical) or an unfamiliar experimenter who never engaged with them at all (no contact). We then observed to what extent children reenacted the irrelevant actions in comparison with a baseline condition in which another group of same-aged children operated the container without a prior demonstration. In Phase 2 of the experiment, the same children were shown the efficient way to retrieve a reward from the container either by a communicative and pedagogical experimenter (no-contact-then-pedagogical and pedagogical-then-pedagogical conditions) or by an unfamiliar experimenter who did not communicate with them at all (pedagogical-then-no-contact condition). Hence, we ran three different conditions (see Table 1).

We predicted that children would imitate irrelevant actions in Phase 1 of the pedagogical-then-no-contact and pedagogical-then-pedagogical conditions, thereby replicating previous findings of overimitation following the demonstration of irrelevant actions by a communicative and pedagogical model. In Phase 1 of the no-contact-then-pedagogical condition, less or no overimitation was expected if direct communication indeed supports learning of causally irrelevant actions from others in children.

In addition, communicative cues may affect whether children continue to use irrelevant actions after seeing the efficient way to achieve a goal. Therefore, we predicted that children would continue to perform the irrelevant actions they were taught by a pedagogical experimenter in Phase 2 of the pedagogical-then-no-contact condition even after seeing a non-communicative experimenter perform the more efficient action. This would speak to the robustness of pedagogically transmitted information. It would also be in accord with the norm learning and social affiliation hypotheses because children should be less motivated to conform to a non-communicative model than to the pedagogical experimenter because they should feel less affiliated with a person who does not establish contact with them. The ACE account, in contrast, would be more compatible with a switch to the efficient strategy regardless of the communicative context because any presentation of the efficient strategy demonstrates the expendability of the irrelevant actions and, thus, should be able to correct distorted causal beliefs. In Phase 2 of the pedagogical-then-pedagogical condition, however, the second model was also communicative. We hypothesized that this communicative model would be able to actively teach children the efficient way to retrieve the reward after they had learned the irrelevant actions from another communicative and pedagogical model because children have been shown to switch flexibly between different socially demonstrated strategies (Wood et al., 2013). In Phase 2 of the no-contact-then-pedagogical condition, children were expected to continue to use the efficient action.

Section snippets

Participants

The study was conducted in a medium-sized German town with participants recruited from a middle-class socioeconomic background. A total of 99 5-year-old children (M = 62.5 months, SD = 1.69, 49 boys and 50 girls) participated. Participants were recruited from a pool of children who had taken part in previous studies. Children were assigned to one of four conditions: three experimental conditions (in each condition: n = 28, 14 boys and 14 girls) and one baseline condition (n = 15, 7 boys and 8 girls).

Results

The number of children showing each of the four irrelevant actions in each condition is presented in Table 2. As expected based on previous findings (Lyons et al., 2007), the most frequently imitated actions involved direct contact with the container and the least frequently imitated action involved no contact with either the rod or the container (i.e., clapping hands). This was the case in all of the experimental conditions.

Preliminary analyses revealed that children’s sex had no significant

Discussion

In the current study, 5-year-olds were first shown an inefficient method, involving several irrelevant actions, to retrieve tokens from a container, demonstrated either by a pedagogical experimenter or by a non-communicative experimenter. Then, children observed the efficient method to retrieve the tokens presented by another experimenter who either acted in a pedagogical manner or did not. The irrelevant actions were rarely performed spontaneously by a group of same-aged children in a baseline

References (23)

  • E. Flynn et al.

    Investigating the mechanisms of cultural acquisition: How pervasive is overimitation in adults?

    Social Psychology

    (2012)
  • Cited by (43)

    • A dual-process perspective on over-imitation

      2020, Developmental Review
      Citation Excerpt :

      When observing a person performing such an action, one needs to check the causal functionality of the test object to understand that this action is irrelevant. Noncontact actions are actions which do not involve any physical contact with the test object, such as noncontact actions directed at a reward container (for instance circling with the index finger 5 cm above the reward container; e.g., Clay et al., 2018), disconnected object-directed actions (for instance tapping on the lid of a separate disconnected container, e.g., Lyons et al., 2007; Taniguchi & Sanefuji, 2017), and body-directed noncontact actions (for instance clapping or tapping with a tool on the palm of one’s hand, e.g., Hoehl et al., 2014). When one sees a person performing such an action, one does not need to check the causal functionality of the test object to understand that this action is causally irrelevant; it is seen at first sight.

    View all citing articles on Scopus
    View full text