The proposal that motor-visual neurons are formed through associative learning is an elegant way to explain a multitude of findings about motor-visual couplings that are otherwise more difficult to reconcile. Cook et al. argue that motor-visual neurons originate from associative learning and suggest the dissociation of function from origin of motor-visual neurons. We would like to stress that this view on motor-visual associations does not dispute the involvement and the importance of motor-visual couplings in social functioning, which has been demonstrated in numerous studies (e.g., Aglioti et al. Reference Aglioti, Cesari, Romani and Urgesi2008; Casile & Giese Reference Casile and Giese2006; Kilner et al. Reference Kilner, Neal, Weiskopf, Friston and Frith2009). Rather, Cook et al. encourage us to review the function of motor-visual couplings for social functioning in past and future studies. Here, we will argue that the perspective of motor-visual couplings being a result of associative learning is able to explain some findings of social functioning in physical social interactions (e.g., handshake) that are more difficult to explain with a genetic perspective of motor-visual associations. Specifically, we will discuss two aspects of this proposal that we deem particularly important for processing of information relevant for social cognition in physical social interactions.

The associative learning hypothesis accounts for context-specific learning of motor-visual associations. Although Cook et al. mainly discussed context specificity in terms of the emergence of motor-visual linkages, we would like to point to implications of context-sensitive motor-visual associations for processing socially relevant information in social interactions. Specifically, empirical evidence indicates that context-specific processing of socially relevant information is critical for social interactions. For example, Georgiou et al. (Reference Georgiou, Becchio, Glover and Castiello2007) showed that the kinematic patterns for the same block stacking actions are modulated by the nature of the social context (cooperative vs. competitive context). Likewise, Streuber and colleagues (Reference Streuber, Knoblich, Sebanz, Bülthoff and de la Rosa2011) demonstrated that sources of visual information about an interaction partner that allow better performance in a table-tennis task depend on the nature of the interactive context (cooperative vs. competitive play). In cooperative play, participants benefited from seeing the other person's racket, while in competitive play, participants showed performance improvements when seeing the other person's body. These and other studies (e.g., Hommel et al. Reference Hommel, Colzato and Van den Wildenberg2009) demonstrate that processing of information relevant for social cognition is influenced by social context. Since there is abundant evidence for the involvement of motor-visual units in processing of socially relevant information (Iacoboni et al. Reference Iacoboni, Molnar-Szakacs, Gallese, Buccino, Mazziotta and Rizzolatti2005; Kaplan & Iacoboni Reference Kaplan and Iacoboni2006), motor-visual units should exhibit some sensitivity to social context. Here, the associative learning hypothesis has the potential to provide a convenient and convincing way of explaining context-sensitive processing of socially relevant information by motor-visual units.

In a similar vein, contingency-based learning of motor-visual associations could be beneficial for processing of socially relevant information in physical social interactions. Physical social interactions often consist of socially agreed (and therefore probable) action sequences, which have been learned during development (e.g., the kissing of cheeks in a certain order as a greeting in some countries). Actions are learned, therefore, not as isolated physical events but as part of an action sequence (action context). Recent observations in our lab demonstrate that processing of socially relevant visual information depends on the preceding temporal action context (de la Rosa et al. Reference de la Rosa, Streuber, Giese, Bülthoff and Curio2014), which makes one action occurring more probable than another one. In particular, we used an adaptation paradigm to examine the sensitivity of action recognition processes to temporally precursory action contexts. Previous research has shown that adaptation to one of two actions biases the perception of a subsequently presented action towards the non-adapted action (Barraclough & Jellema Reference Barraclough and Jellema2011; Barraclough et al. Reference Barraclough, Keith, Xiao, Oram and Perrett2009). We investigated whether adaptation to action images is modulated by the presentation of an action context shown prior to one of the adaptors. We, therefore, created two experimental conditions that consisted of identical adaptors and test stimuli but differed in terms of a movie that was shown prior to one of the adaptors. The movies in the two conditions showed different action sequences and caused participants to differently interpret the action displayed by the adaptor. If action recognition is only based on the immediately available sensory information about the adaptors, then one would not expect any modulation of the action adaptation effect across the two conditions because the two conditions consisted of physically identical adaptor and test images. On the other hand, if action recognition is sensitive to the action context that preceded the adaptors, then the adaptation aftereffects between the two conditions should differ because both conditions were associated with different social contexts that induced a different action interpretation of the adaptor. Indeed, we found action adaptation effects to be different between the two conditions (de la Rosa et al. Reference de la Rosa, Streuber, Giese, Bülthoff and Curio2014). A control experiment showed that the presentation of the movie alone was not able to modulate the action adaptation effect between the two conditions. The results, therefore, demonstrate that action recognition does not only depend on the immediate sensory information, but also on the action context preceding an action.

We agree with the suggestion that a dissociation of function and origin of motor-visual associations is important for a better understanding of the involvement of these associations in social functioning. We think that the associative learning account provides a novel view that is able to stimulate research that improves our understanding about the ability of humans to interact, communicate, and socialize with others.