What makes a house a home? Mechanisms of lexical alignment in preschoolers’ referential communication

When conversing, people must choose between alternative names to refer to an object (e.g. cup vs mug). An important determinant of referential choice is lexical alignment: the tendency of speakers to converge on the same name for a particular object (Brennan & Clark, 1996). Lexical alignment (or entrainment) has been found in school-aged children as well as adults (Branigan, Tosi, & Gillespie-Smith, 2016;). More recent research has shown that in a highly structured picture-naming ‘snap’ task, even 3-4 year old children display lexical alignment (Lindsay, Hopkins, & Branigan, in prep). However, the nature of the ’snap’ task promoted the possible influence of lexical priming mechanisms on children’s referential choices, and made it less likely that perspective-taking and social-affiliative mechanisms – which mediate alignment in adults (Branigan, Pickering, Pearson, & McLean, 2010) – would play any role. In the present study, we examined whether 3-4 year olds show lexical alignment in a less structured context involving a more demanding task. We also ask whether social-affective mechanisms influence alignment when affiliation goals are salient. Participants (N=70) played a novel referential communication game – the “Moving House” game – with an experimenter. In the game, the participant and experimenter each moved items from a moving truck into different rooms of a house. Experimental items had two alternative names. Pre-tests established that children knew and understood both names, but had a strong preference for one alternative. The experimenter told the participant where to place items from the truck into their house (prime rounds); the participant and experimenter then switched roles, and the participant directed the experimenter where to place her items (different exemplars of the same category, e.g., a different cup; target rounds). During prime rounds, we manipulated the name that the experimenter used to name target items (preferred vs. dispreferred), and examined children’s likelihood of producing the dispreferred name for a different exemplar of the same category in target rounds. Before the game, we also manipulated children’s affiliative motivation, by showing half the participants a video depicting third-party ostracism, and the other half a control video (Over & Carpenter, 2009). Children were more likely to use a dispreferred name (mug) if the experimenter had previously used a dispreferred name than a preferred alternative (cup; .33 vs. .12, p < .001). Their tendency to lexically align was not affected by having watched the ostracism vs. control video (.35 vs .31; p = .46). Our findings show that 3-4 year old children spontaneously lexically align with a conversational partner, even when this means using a normally dispreferred name. Importantly, such alignment is not restricted to simple and highly structured tasks where referential communication was not necessary for task success: in the Moving House task, children had to determine the correct location for each object, and communicate both the identity of the relevant object and its location to the experimenter. Moreover, their alignment occurred over a substantial number of intervening turns and associated time delay. However, such alignment was not influenced by exposure to third party ostracism. Together these results suggest that preschoolers’ strong tendency to spontaneously lexically align is not restricted to highly structured and cognitively undemanding tasks, and that it is not primarily driven by social-affective mechanisms whereby children imitate in order to promote affiliation, though we cannot rule out that such mechanisms may be operative under some conditions. The fact that alignment occurred for different exemplars of a category also suggests that it does not arise from a simple memory association between a referent and an episode of language use. Our results are compatible with lexical priming mechanisms, but the durability of alignment effects – in the presence of greater task demands – also suggests the possibility of perspective-taking. Children may have used shared linguistic context (i.e., the experimenter’s previous name use to infer what she was likely to understand), and aligned to enhance communicative success.