Cook et al. present an elucidative perspective on how mirror neurons (MNs) could emerge in the human brain. By clarifying the conceptual distinction between questions about the function of MNs and questions about the origin of the matching properties of these neurons, their article makes an important contribution to the field of MN research. With respect to the origin of MNs, we think that the associative account put forward by Cook et al. provides a strong theoretical framework by which experimental findings can be assessed. In this commentary, we would like to elaborate on the potential function of MNs in action understanding, assuming that the associative account put forward by Cook et al. is correct.
Cook et al. note that if MN activity can be understood as arising from associative learning, then its function, if any, in action understanding remains an open question. We would like to take their proposal one step further and argue that if MN activity indeed is a purely associative phenomenon, then there are strong theoretical reasons to believe that MN activity itself cannot be constitutive of genuine action understanding, in the sense of understanding the “why” of actions (e.g., the goals and intentions underlying actions). Our argument consists of two parts. First, we argue that, at least in humans, higher-level cognitive processes can guide the formation of appropriate associations, and that such guidance seems necessary for forming associations that code for the “why” of actions. Second, we argue that even if such appropriate associations have been formed successfully, higher-cognitive processes seem required for selecting which of many possible associations codes the “why” in a particular situation.
Cook et al. review evidence showing that in order to build an associative connection between two events, it is sufficient that the two events occur around the same time and that one reliably predicts the other. This suggests that associative learning is a low-level process where no reasoning is involved. The fact that even basic organisms that are unlikely to have higher-order reasoning ability (e.g., invertebrate sea slugs; Walters et al. Reference Walters, Carew and Kandel1979) are sensitive to classical conditioning suggests that cognitive processing is indeed not necessary for the formation of associations. However, there is evidence that suggests that the degree to which any two events are associated can be guided by beliefs about the causal structure of the world and other prior knowledge. For instance, in a trace conditioning paradigm, Clark and Squire (Reference Clark and Squire1998) found that participants could be conditioned to blink their eyes when hearing a tone, but only if they were aware of the relation between that tone and a puff of air to the eye that caused them to blink. A similar influence of causal beliefs on association seems operational in the well-known “blocking” effect in conditioning. This is the effect that the association of an event A with event Y is prevented if A is presented together with another event B that has previously been associated with event Y. Notably, Waldmann (Reference Waldmann2000) found that this effect is modulated by whether the participants were led to believe that A and B were either possible causes or possible effects of Y. Findings such as these provide strong evidence that, at least in humans, which types and degrees of associations are formed is partly shaped by higher-level cognitive processes. If MNs indeed code for associations, as Cook et al. suggest, then the degree to which they can be supposed to support action understanding, if at all, will depend on the degree to which the associative processes that guide the MN matching properties are in turn guided by relevant higher-level processes involving beliefs about the causal structure of actions. In other words, if associations are formed without consideration of the causes of actions, including mental causes such as goals and intentions, then they cannot come to code for the “why” of actions.
Admittedly, it is conceivable that the associations coded by the MN system are shaped in part by higher-level processes and thereby could form associations in a way that is sensitive to the causal structure of actions. But even then, for the interpretation of newly observed actions, the MN system can probably not interpret these actions without the involvement of higher-level cognitive processes. This leads us to our second point. There are many possible intentions that may explain an observed action, as well as many possible actions that result from an intention. This many-to-many mapping implies that any given observed action can activate – for example, in the MN system – many possible intentions that have been previously associated with that action. To select which of these associations applies to the current situation, we need some other, context-sensitive process. For reasons that have been detailed elsewhere (Uithol et al. Reference Uithol, van Rooij, Bekkering and Haselager2011), this process most plausibly involves some form of (possibly implicit, unconscious) reasoning that takes into account some form of knowledge of the world and how actions interact with intentions and contexts. This idea is supported by studies showing that when people consider why an action was performed, for example, when they observe novel actions (Brass et al. Reference Brass, Schmitt, Spengler and Gergely2007) or when they are instructed to attend to the intentionality of an action (de Lange et al. Reference de Lange, Spronk, Willems, Toni and Bekkering2008), areas other than MN areas are recruited. Hence, it seems that, if MN associations play a functional role in action understanding at all, these associations need to be integrated with prior knowledge and beliefs in areas outside the MN system for a full appreciation of the intentionality of actions.
In summary, we agree with Cook et al. that the associative account of MN activity is viable. The properties of MNs seem indeed explainable by associative learning. We go one step further than Cook et al. by proposing that a purely associative account implies that MNs cannot explain genuine action understanding, including understanding of the “why” of actions. Higher-level processes are important in the formation of appropriate associations that may code the intentions of actions, as well as in the application of learned associations during intentional action understanding.
Cook et al. present an elucidative perspective on how mirror neurons (MNs) could emerge in the human brain. By clarifying the conceptual distinction between questions about the function of MNs and questions about the origin of the matching properties of these neurons, their article makes an important contribution to the field of MN research. With respect to the origin of MNs, we think that the associative account put forward by Cook et al. provides a strong theoretical framework by which experimental findings can be assessed. In this commentary, we would like to elaborate on the potential function of MNs in action understanding, assuming that the associative account put forward by Cook et al. is correct.
Cook et al. note that if MN activity can be understood as arising from associative learning, then its function, if any, in action understanding remains an open question. We would like to take their proposal one step further and argue that if MN activity indeed is a purely associative phenomenon, then there are strong theoretical reasons to believe that MN activity itself cannot be constitutive of genuine action understanding, in the sense of understanding the “why” of actions (e.g., the goals and intentions underlying actions). Our argument consists of two parts. First, we argue that, at least in humans, higher-level cognitive processes can guide the formation of appropriate associations, and that such guidance seems necessary for forming associations that code for the “why” of actions. Second, we argue that even if such appropriate associations have been formed successfully, higher-cognitive processes seem required for selecting which of many possible associations codes the “why” in a particular situation.
Cook et al. review evidence showing that in order to build an associative connection between two events, it is sufficient that the two events occur around the same time and that one reliably predicts the other. This suggests that associative learning is a low-level process where no reasoning is involved. The fact that even basic organisms that are unlikely to have higher-order reasoning ability (e.g., invertebrate sea slugs; Walters et al. Reference Walters, Carew and Kandel1979) are sensitive to classical conditioning suggests that cognitive processing is indeed not necessary for the formation of associations. However, there is evidence that suggests that the degree to which any two events are associated can be guided by beliefs about the causal structure of the world and other prior knowledge. For instance, in a trace conditioning paradigm, Clark and Squire (Reference Clark and Squire1998) found that participants could be conditioned to blink their eyes when hearing a tone, but only if they were aware of the relation between that tone and a puff of air to the eye that caused them to blink. A similar influence of causal beliefs on association seems operational in the well-known “blocking” effect in conditioning. This is the effect that the association of an event A with event Y is prevented if A is presented together with another event B that has previously been associated with event Y. Notably, Waldmann (Reference Waldmann2000) found that this effect is modulated by whether the participants were led to believe that A and B were either possible causes or possible effects of Y. Findings such as these provide strong evidence that, at least in humans, which types and degrees of associations are formed is partly shaped by higher-level cognitive processes. If MNs indeed code for associations, as Cook et al. suggest, then the degree to which they can be supposed to support action understanding, if at all, will depend on the degree to which the associative processes that guide the MN matching properties are in turn guided by relevant higher-level processes involving beliefs about the causal structure of actions. In other words, if associations are formed without consideration of the causes of actions, including mental causes such as goals and intentions, then they cannot come to code for the “why” of actions.
Admittedly, it is conceivable that the associations coded by the MN system are shaped in part by higher-level processes and thereby could form associations in a way that is sensitive to the causal structure of actions. But even then, for the interpretation of newly observed actions, the MN system can probably not interpret these actions without the involvement of higher-level cognitive processes. This leads us to our second point. There are many possible intentions that may explain an observed action, as well as many possible actions that result from an intention. This many-to-many mapping implies that any given observed action can activate – for example, in the MN system – many possible intentions that have been previously associated with that action. To select which of these associations applies to the current situation, we need some other, context-sensitive process. For reasons that have been detailed elsewhere (Uithol et al. Reference Uithol, van Rooij, Bekkering and Haselager2011), this process most plausibly involves some form of (possibly implicit, unconscious) reasoning that takes into account some form of knowledge of the world and how actions interact with intentions and contexts. This idea is supported by studies showing that when people consider why an action was performed, for example, when they observe novel actions (Brass et al. Reference Brass, Schmitt, Spengler and Gergely2007) or when they are instructed to attend to the intentionality of an action (de Lange et al. Reference de Lange, Spronk, Willems, Toni and Bekkering2008), areas other than MN areas are recruited. Hence, it seems that, if MN associations play a functional role in action understanding at all, these associations need to be integrated with prior knowledge and beliefs in areas outside the MN system for a full appreciation of the intentionality of actions.
In summary, we agree with Cook et al. that the associative account of MN activity is viable. The properties of MNs seem indeed explainable by associative learning. We go one step further than Cook et al. by proposing that a purely associative account implies that MNs cannot explain genuine action understanding, including understanding of the “why” of actions. Higher-level processes are important in the formation of appropriate associations that may code the intentions of actions, as well as in the application of learned associations during intentional action understanding.