We congratulate Schilbach et al. for raising these theoretical issues concerning the nature of social engagement. We are also enthusiastic about the re-emergence of the concept of “affordance” to characterize what is properly perceived by an engaged observer during a social interaction. We share the view that observing social cues (i.e., features of the environment whose function is not to affect a recipient) is qualitatively very different from processing social signals. Social signals are, in contrast to cues, features of the environment whose function is to alter a recipient's behavior and which are recognized and processed as such by the engaged observer. This theoretical shift is fundamental and we see the attempt to free social cognition from the “inferential detour” as salutary and promising. We also consider with great interest the various methodological improvements brought by Schilbach et al., and see their practical suggestions as highly relevant. However, we regret that the authors do not specify in more detail what they understand by “social affordances,” and that they offer too few details about how they are perceived and processed by our cognitive apparatus. The clarification of these issues is crucial to considering the neural candidates, which might enable the perception of social affordances.

Contrary to what Schilbach and colleagues argue, we do not think that the mirror neuron system (MNS) is involved. First, this might be the case for purely anatomical constraints: Neurons with mirror properties (MNs) only account for 17% (Gallese et al. Reference Gallese, Fadiga, Fogassi and Rizzolatti1996) of the macaque premotor (PM) and parietal neurons. Among these 17%, only 5.5% code for a strictly congruent action in the motor and the visual domain, whereas 8.6% code for two or more actions in the visual domain, and 1.3% for non-congruent actions. In humans, Mukamel et al. (Reference Mukamel, Ekstrom, Kaplan, Iacoboni and Fried2010) revealed that 14% of the recorded neurons in the supplementary cortex area responded to congruent observed actions, but 10% responded to non-congruent observed actions. We do not think the latter should be considered MNs. Rather, they may be categorized as “social” canonical neurons, that is, neurons that are active when foreseeing a possible social interaction (vs. interaction with an object as for canonical neurons) and preparing oneself accordingly.

There is a second reason – a functional one – why MNs would not be good candidates for the underpinning of the processing of social affordances: What is involved in the perception of social affordances is, cognitively speaking, very different from what MNs are known to do, that is, to simulate an observed motor pattern (Rizzolatti et al. Reference Rizzolatti, Fogassi and Gallese2001). Instead, we assume that the affordance perspective of social signals calls for the reexamination of previous findings to identify specific neural networks.

Let us first consider the notion of “social affordance.” Schilbach et al. define social affordance as “the possibilities for interaction provided by others” (sect. 3.1.1, para. 5). These possibilities, according to the authors, are characterized by “an activation of motor programs that could allow for interpersonal coordination of behavior” (sect. 3.1.1, para. 5). It must be clear that social signals can be considered affordances in the sense that they trigger a wide range of opportunities for actions in the observer. We, as animals, are continuously facing opportunities for action in our environment and are constantly collecting information in order to select the most relevant action from among numerous potential actions (Cisek Reference Cisek2007; Cisek & Kalaska Reference Cisek and Kalaska2010). This entails that brain activity during social interactions reflects this parallel processing of multiple representations of potential actions (social affordances) and their evaluation through the use of external as well as internal sensory information. Activity in the parietal cortex and connected motor regions might therefore reflect the representation of the various affordances (Cisek & Kalaska Reference Cisek and Kalaska2010).

In a recent study conducted by our team (Conty et al. Reference Conty, Dezecache, Hugueville and Grèzes2012), participants faced dynamic stimuli depicting actors producing complex social signals involving gaze, a pointing gesture, and the expression of anger. We were able to show, using electroencephalography (EEG) coupled with functional magnetic resonance imaging (fMRI) that a binding of these social signals occurred in the PM as early as 200 msec after stimulus onset. Strict motor resonance (MNS) processing could not explain the activation in the PM, as anger expressions directed at an observer are perceived as a clear signals of non-affiliative intentions and are therefore less mimicked than averted anger expressions (Bourgeois & Hess Reference Bourgeois and Hess2008, Hess et al. Reference Hess, Adams and Kleck2007). We propose that activity in the PM is related to two complementary mechanisms: (1) the estimation of prior expectations about the perceived agent's immediate intent, and (2) the specification of currently available actions for the observer to deal with in the immediate situation. Indeed, only the combination of the two could explain the highest level of activity in the PM for the highest degree of potential social interaction, that is, an angry person pointing, facing, and looking towards oneself. This activity in the PM may therefore reflect the competition between several representations of action opportunities, triggered by the ongoing social interaction, as well as the selection of an adaptive behavioral response.

In sum, we think that it is crucial to first ask whether the kind of cognitive activity that is involved in the processing of social affordances is compatible with what is conventionally thought of as the MNS, before considering all activities in the PM as mirroring processes. Second, it is important to consider regions that do not display mirroring properties – such as the amygdala (Sander et al. Reference Sander, Grafman and Zalla2003) – as fundamental in the evaluation of social signals and in the triggering of one's own adaptive reaction. Consequently, it is reasonable to advance that the processing of social affordance involves a specific brain network, neuron populations, and mechanisms that differ (for the most part) from the MNS. Characterizing the neural specificities of the “social affordance network” beyond the MNS represents a challenging step in our understanding of the processing of social signals.