We were very grateful for the thoughtful insights of both commentators. Both raised interesting theoretical points, which we are happy to address below. We suspect that our theoretical convictions are not actually as far away from each other as it might initially appear, as several primary remarks appear to be based on misunderstandings about the nature of our suggestions.
Hommel questions the need for introducing a new cognitive module (i.e., semantic resonance) to explain effects of action on language processing. He emphasizes that the strength of embodied theories of language lies in their assumption that language and action act on a set of common codes, and points out that introducing a new set of ''semantic codes'' unnecessarily complicates the embodied perspective, ultimately even disembodying cognition. We completely agree with Hommel. In fact, in our original paper, we do not suggest semantic resonance as a separate module or a set of new codes, but rather as a label to describe the direction of information flow between language and action systems. This can be seen quite clearly in Figure 1. Language and action are depicted as two cognitive systems, which influence one another because of their reliance on common codes (i.e., semantics). We point out that previous literature has focused on effects of language processing on sensorimotor systems (i.e., ''motor resonance''). We simply wish to expand the degree to which language and action are seen to interact, by suggesting that sensorimotor processing can also affect language processing. In other words, ''semantic resonance'' describes the other side of a well-documented coin.
Aarts and Veling question the role of intention in embodied language representations. They argue that embodied effects (i.e., motor and semantic resonance) should be elicited whenever action and language features overlap, irrespective of whether an action is intentional. This comment speaks to the final section of our original paper, in which we suggest that action intention must play a role in determining what semantic information is relevant in a given situation. We suggest that the meaning of cup in the action context of trapping wasps differs from the meaning of cup in the action context of drinking. In fact, we have no quarrel with the assumption that default embodied representations are activated in the absence of intentional actions. As Aarts and Veling point out, we review literature suggesting that automatic embodied representations exist. However, we do not think this can be the whole story. While some degree of semantic and motor resonance may indeed be automatic, the flexibility with which words (and objects) can be used in different contexts calls for additional explanation. We suggest that words are connected to multiple action representations, and that actions are connected to multiple semantic representations. Which of these representations becomes ultimately activated is modulated by the context in which the word/action is expressed. Thus, in our account embodied representations are not construed to be dependent on intention, but action intention can certainly modulate the weight given to one embodied representation over another.