Sensitivity to relevance 1 Sensitivity to Communicative Relevance Tells Young Children What to Imitate

How do children decide which elements of an action demonstration are important to reproduce in the context of an imitation game? We tested whether selective imitation of a demonstrator’s actions may be based on the same search for relevance that drives adult interpretation of ostensive communication. Three groups of 18-month-old infants were shown a toy animal either hopping or sliding (action style) into a toy house (action outcome), but the communicative relevance of the action style differed depending on the group. For the no prior information group, all the information in the demonstration was new and so equally relevant. However, for infants in the ostensive prior information group, the potential action outcome was already communicated to the infant prior to the main demonstration, rendering the action style more relevant. Infants in the ostensive prior information group imitated the action style significantly more than infants in the no prior information group, suggesting that the relevance manipulation modulated their interpretation of the action demonstration. Furthermore, infants in the the prior information made infants imitate the action style at the expense of the action outcome. A further condition (non-ostensive prior information) confirmed that this sensitivity to new information is only present when the ‘old’ information had been communicated, and not when infants discovered this information for themselves. These results indicate that, like adults, human infants expect communication to contain relevant content, and imitate action elements that, relative to their current knowledge state or to the common ground with the demonstrator, is identified as most relevant. Sensitivity to relevance 3 Introduction The ubiquity of human imitation gives the impression of an ability that is a trivial feat. We inadvertently imitate one another during social interactions (Chartrand & Bargh, 1999), newborn infants imitate the facial expressions of their caregivers (Meltzoff & Moore, 1983), and, by about 9 months of age, human infants spontaneously begin to imitate the actions of others. Recent findings from social neuroscience have led to the suggestion that a dedicated neural mechanism, which maps observed behaviours directly onto the observer's own motor system, may exist to sub serve this ability (Iacoboni et al., 1999). However, as several authors have noted, imitation requires not only the ability to map observed behaviours onto one’s own body, but also cognitive mechanisms to select which behaviours are necessary to be imitated (Gergely & Csibra, 2006; Brugger, Lariviere, Mumme & Bushnell, 2007; Csibra, 2007; Southgate & Hamilton, in press). The capacity to identify relevant behaviours for reproduction is essential for imitation to have evolved as an efficient tool for cultural transmission (Boyd & Richerson, 1995; Galef, 1992; Tomasello, 1999). As a number of studies have shown, children do not blindly imitate everything that they observe. For example, 12and 14-month-old infants take into account the action constraints of the demonstrator, and appear to modulate their imitation depending on whether their own situation is subject to the same constraints (Gergely, Bekkering & Kiraly, 2002; Schwier, van Maanen, Carpenter, & Tomasello, 2006). In another study, infants at 18 months did not imitate what an experimenter actually did when she failed to achieve a goal – they imitated what she had intended to do (Meltzoff, 1995). Sensitivity to relevance 4 In a recent paper, Carpenter, Call and Tomasello (2005) argued that infants imitate actions in terms of what they think the demonstrator's goal is. In their study, when 12and 18-month-olds were shown a toy mouse either hopping or sliding into a toy house, infants selectively imitated putting the animal in the house, but did not imitate the particular means (hopping or sliding) by which the animal went into the house. However, when there was no house present and they were shown the animal simply hopping or sliding around a mat, infants at both ages imitated the action style. Similar findings were reported in a different paradigm and in older children by Bekkering, Wohlschlager, and Gattis (2000). The authors concluded that infants copied actions in terms of goals: when there was a clearly visible goal (e.g. a house), infants interpreted the outcome as the goal (putting the mouse into the house), but when there was no visible goal, infants interpreted the action style (e.g., hopping or sliding) as the goal. Thus, since one could reproduce behaviours at a number of different levels (Byrne & Russon, 1998; Csibra, 2007), one can isolate relevant from irrelevant actions by identifying the goal of a particular action, and imitating at the goal level, disregarding any lower level components that do not appear to be causally related to the desired outcome. That infants as young as 14 months are capable of identifying causally relevant actions and imitating on this basis has recently been demonstrated by Brugger and colleagues (Brugger, Lariviere, Mumme, & Bushnell, 2007). However, some authors have argued that, unlike the many documented instances of other animal cultures (e.g. Laland & Hoppitt, 2003; Rendell & Whitehead, 2001; Whiten et al., 1999), the goals and the causal relations between performed actions and their outcomes are often not immediately obvious in human cultural practices (Gergely & Csibra, 2006). This cognitive opacity will often render selective Sensitivity to relevance 5 imitation, on the basis of goal identification, impossible. For example, humans engage in tool making for which, to a naive observer, there may appear no immediate and visible goal at the time of construction, and perform rituals that do not reveal how they are supposed to work. If much of human culture consists of such cognitively opaque practices, it would make little sense for infants' observational learning to be driven solely by the identification of goals. To cope with this problem of cognitive opacity, Gergely & Csibra (2005, 2006) have proposed that, as part of a suite of evolved adaptations, imitation has been selected to be sensitive to the communicative intent of the demonstrator. By this account, infants’ interpretation of action demonstrations directed to them is based on the same pragmatic assumptions that human adults employ when engaged in communicative interactions with others (Sperber & Wilson, 1986). One of these assumptions is that the communication is in some way relevant to the recipient, where relevance is determined in relation to the knowledge state of the individual. Thus, any information that is not already possessed by the recipient, or could not be inferred on the basis of her knowledge, will be identified as relevant, and as the intended content of the demonstration. That communication plays a role in imitation is suggested by a number of recent studies showing that selective imitation in the second year of life is influenced by the presence or absence of ostensive communication (Brugger et al., 2007; Kiraly, Csibra & Gergely, 2004; Nielsen, 2006), findings that are inconsistent with the proposal that infants simply imitate observed actions in terms of perceived goals. The present study aimed to test the hypothesis that the role played by communication in imitation is the expectation of relevance that it elicits in recipients. Specifically, we predicted that human infants seek relevance in others' communication Sensitivity to relevance 6 and selectively imitate what they infer to be the communicatively most relevant part of action demonstrations. We modelled our task after the paradigm used by Carpenter and colleagues (2005), described above. However, in our version, we varied the communicative relevance of some aspects of the information that 18-month-old infants received in each of three groups. Infants in the first group received a demonstration in which all of the information demonstrated to them, was new (no prior information condition). In this condition, infants watched as an experimenter either hopped or slid a toy animal into a toy house. In another condition (ostensive prior information), infants were first told and shown that the animal lives in the house, before seeing the same demonstration of the animal either hopping or sliding into the house. By showing infants that the animal lives in the house prior to the main demonstration, the placing of the animal into the house becomes ‘old’ information and should, if infants are sensitive to communicative relevance, receive less attention and processing resources than the ‘new’ information in the demonstration (the manner in which the animal moves). Our hypothesis was thus that infants who receive prior information would imitate the manner by which the experimenter moved the animal more than infants in the other condition in which all information is new. We expected that without prior information infants would perform as they did in Carpenter et al. (2005), and selectively imitate at the hierarchically highest level, putting the mouse into the house without reproducing movement style. Our predictions here are based on the assumption that it is the communicative context that generates the expectation that the demonstrator is going to manifest some relevant information for them (Gergely & Csibra, 2006). However, the expectation that communication is relevant is proposed to derive from a more general cognitive principle Sensitivity to relevance 7 of relevance (Sperber & Wilson, 1986), which describes the fact that human cognition is geared to the maximization of relevance (i.e. attend to information providing the most cognitive effects using the least processing resources). Thus, it is also possible that infants would simply attend more to the new information because of the greater cognitive effects it brings them, but that this relevance seeking is not related to the interpretation of the communicative intent of the demonstrator. To control for this possibility, we included a third condition (non-ostensive prior information), in which infants discovered for themselves that the animal could go in the house before the full demonstration. In this way, the action outcome component still constitutes cognitively ‘old’ information for the infant when it is subsequently ostensively demonstrated by the experimenter, but, crucially, it is not ‘old’ information in the communication. If infants’ imitation is driven by a sensitivity to communicative (rather than just cognitive) relevance, then even if they already know that the animal can go in the house, they should nonetheless treat this information as a relevant element of the experimenter’s communication. As such, we predicted that infants in this condition would behave as they do without prior information, imitating predominantly the action outcome, and ignoring the action style.