Language is a strikingly social behavior. While it is possible to have social behavior that is not linguistic, the converse is not true. Language is learned, perceived, and produced within the fabric of social interaction. Using developmental and comparative literature, we have previously contended that the presence of structural and functional linkages between subcortical motivation systems, and conventional language and social circuits in the brain, are critical determinants of the evolution and development of language in a given species (Syal & Finlay Reference Syal and Finlay2011). A research program that aims to study language-learning needs to attend to language as embedded in its ecological context, acquired in the early development of an obligatorily social, gregarious, and often-altruistic species, where the motivation to learn to communicate with conspecifics drives both its ontogeny and phylogeny. To that end, researchers have thus far focused on the role of socially derived motivation in language learning through studying infant-caregiver interactions in the development of vocal communication (Goldstein & Schwade Reference Goldstein and Schwade2008; Kuhl Reference Kuhl2007b). However, this corpus of knowledge has been constrained by the inability to study brain-behavior linkages through the acquisition of functional neuroimaging data that model ecologically valid social interactions. The novel methodologies discussed by Schilbach et al. in the target article use eye-tracking in conjunction with functional magnetic resonance imaging (fMRI) to create interactive paradigms that allow human participants to experience the effects of their gaze on that of a social counterpart, simulating a naturalistic social interaction, while allowing researchers to gather MRI data that elucidates underlying neural networks. This approach provides an immensely pliable platform on which social motivation in vocal learning can now be placed and probed, in both adult and infant language learners. For instance, the role of joint attention in guiding the learning of artificial object-labels in (a) adults, using both eye-tracking and fMRI, or (b) infants, using interactive eye-tracking setups, are immediate examples of experimental questions that can be addressed using these paradigms.

From early life, human children attend to social cues, share information, join games, and generally cooperate, serving a form of social learning limited largely to humans (Moll & Tomasello Reference Moll and Tomasello2007). These prosocial tendencies also sustain vocal learning behavior during development, wherein numerous structural aids to language learning are presented to the infant in a characteristically social environment replete with positive feedback. For example, mothers reliably use predictable prosodic contours to modulate infant affect and attention (Fernald Reference Fernald, Papousek, Jurgens and Papousek1992; Fernald & Simon Reference Fernald and Simon1984). They engage in contingent turn-taking vocal interactions with their infants that facilitate vocal development (Goldstein & Schwade Reference Goldstein and Schwade2008). They use variation sets – sentences with partially overlapping syntactical structure – to aid word learning (Onnis et al. Reference Onnis, Waterfall and Edelman2008). Recent work has shown how learning of the structural regularities of language can emerge from the richness of social interactivity embedded in the human ecological niche. Specific forms of contingently delivered vocal reinforcement from a social counterpart cause infants to change correspondingly specific features of their own vocalizations towards developmental advancement (Goldstein & Schwade Reference Goldstein and Schwade2008). Contingency of social interaction remains a core requirement in these learning processes – infants do not display vocal learning when the same amount of stimulation is provided in a non-contingent social interaction (Goldstein & Schwade Reference Goldstein and Schwade2008; Goldstein et al. Reference Goldstein, King and West2003; Goldstein et al. Reference Goldstein, Schwade, Briesch and Syal2010a), or through audio-visual media (Kuhl et al. Reference Kuhl, Tsao and Liu2003). Indeed, the extent of learning is in fact determined by the amount of social engagement – greater shared visual attention between infants and interactive social agents facilitates greater language learning (Conboy et al. Reference Conboy, Brooks, Taylor, Meltzoff and Kuhl2008).

An idea essential to this approach is that social contingency in dyadic or triadic interactions is inherently rewarding, and promotes learning through the recruitment of motivational neurocircuitry, and the facilitation of shared attention. In support of this, initial data using the experimental paradigm outlined in the target article suggest a role for reward-related circuitry in initiating joint attention on both neural and behavioral levels (Schilbach et al. Reference Schilbach, Wilms, Eickhoff, Romanzetti, Tepest, Bente, Shah, Fink and Vogeley2010b). Numerous studies have highlighted the role of contingency in specific forms of reward-based learning. In adults, the caudate nucleus within the dorsal striatum is sensitive to reinforcement of action and shows a robust response when subjects perceive a contingency between their actions and task-outcomes (Tricomi et al. Reference Tricomi, Delgado and Fiez2004). The caudate is also involved in encoding stimulus salience (Zink et al. Reference Zink, Pagnoni, Chappelow, Martin-Skurski and Berns2006) and, in language learning, reward-related caudate activation in response to contingent feedback facilitates the learning of non-native phonetic contrasts in adults (Tricomi et al. Reference Tricomi, Delgado, McCandliss, McClelland and Fiez2006). This social hypothesis of language learning suggests that reward-based instrumental learning and positive affect systems may be critical to language development, not only in terms of acquisition, but also in the flexible integration of newly learned information within the existing lexicon. Positive affect embedded in social interaction could facilitate language-learning through salience-tagging information and/or shifting focus towards a broader information encoding context. Data from adults indicates that positive affect can lead to fundamental shifts in information processing through the facilitation of flexible modes of cognition (Isen Reference Isen, Snyder and Lopez2002), which increase the breadth of attentional allocation in both perceptual and conceptual domains (Rowe et al. Reference Rowe, Hirsh and Anderson2007; Schmitz et al. Reference Schmitz, De Rosa and Anderson2009). Additionally, as a social act, language learning and use involves not only the ability to make narrow associations between words and their referents, but also the broader capacity of reading another's mental states, possibly invoking a network of shared activation between minds. The affiliative role of positive emotions is likely critical to this interpersonal network resonance.

Past work on language learning has accorded limited significance to emergent properties of social interactions. The second-person approach to neuroscience posits social interaction and emotional engagement with social counterparts as fundamental features of social living that enable us to understand and learn from the minds of others, highlighting the importance of studying real-time interaction dynamics between individuals in an ecologically valid manner. Within the ecological framework of vocal learning, the parent-infant dyad constitutes a socially distributed system of learning, best viewed as a consolidated network that incorporates the learner, the social interactor, the interaction between the two, and the effect of each actor on its own and the other's nervous system.