As a developmental scientist, I greatly appreciate this target article drawing attention to the importance of clarifying the origins of mirror neurons (MNs). This is no easy feat and I believe that this article is testament to the complexity of the problem. According to the authors, Cook et al., the standard view of MNs is that they represent an adaptation by the organism, and are thus genetically predetermined. By contrast, Cook et al. assert that mirror mechanisms are not predetermined or even probabilistically determined, but instead develop as a function of sensorimotor associative learning. My objection to this position is that associative learning may be necessary, but it is not sufficient for the development of MNs. In the remainder of this commentary, I will present three lines of evidence that question the sufficiency of associative learning for explaining the development of mirror mechanisms: (1) Genetic predispositions interact with associative learning; (2) infants show predispositions to imitate human as opposed to nonhuman actions, and (3) there are differences in learning during early and later development.
The first reason to question a strict associative account is because it cannot explain why some behaviors are learned more easily than others. Cook et al. claim that the observation–execution matching properties of mirror mechanisms are not a specific genetic predisposition, but rather a domain-general process of associative learning found in a wide range of vertebrate and invertebrate species. This position implies that all correlated sensory-motor experiences should be learned equally well by the organism, but this claim is not supported by the data. Beginning with the pioneering research of Garcia (e.g., Garcia et al. Reference Garcia, Kimeldorf and Koelling1955), it was shown that rats could not associate visual and auditory cues with food that made them ill, but could learn to associate olfactory cues with such foods (see Gould & Marler [Reference Gould and Marler1987] for other examples). These findings are not surprising from an adaptive standpoint (though I suspect eschewed by Cook et al.) because rats are biologically prepared to learn some things more readily than others. In the natural world, odor is a more reliable cue than color for rats because they are primarily nocturnal, so odor is better associated with dangerous foods than is color. The conclusion from these sorts of studies is that animals are biased to learn some associations more easily than others even though the contingencies are the same. Thus, it appears that there are genetic predispositions that interact with the success of associative learning.
The second reason to question the associative account derives from evidence showing that infants are biased to imitate human as opposed to nonhuman actions. In the target article, imitation is considered a behavioral index of the presence of mirror mechanisms. If all sensorimotor associative learning is sufficient to explain imitation, then infants should be as likely to imitate mechanical as opposed to human actions. However, the critical evidence is at best mixed. For example, Meltzoff (Reference Meltzoff1995) demonstrated that 18-month-old infants could imitate the actions of an adult pulling apart a barbell but not the actions of a mechanical pincer designed to match the movement, as well as the effect, produced by the adult.
Similarly, Longo and Bertenthal (Reference Longo and Bertenthal2006) tested 9-month-old infants in an observational version of the Piagetian A-not-B search task, and showed that infants still committed the search error on the B test trial because they covertly imitated the search behavior of the experimenter during the A trials. By contrast, Boyer et al. (Reference Boyer, Pan and Bertenthal2011) substituted a pair of mechanical claws for the human experimenter, and 9-month-old infants failed to commit the search error suggesting that they were less likely to covertly imitate the goal-directed behavior of the claws. It thus appears that infants, like the rats described above, are predisposed to learn some sensorimotor associations more readily than others. These findings are thus consistent with the human and nonhuman research literature suggesting that MNs are more likely to become activated to the observation of human actions than to mechanical devices or tools (Liepelt & Brass Reference Liepelt and Brass2010; Longo & Bertenthal Reference Longo and Bertenthal2009; Woodward Reference Woodward1998). If associative learning was sufficient for the development of mirror mechanisms, the response to human generated actions should not be privileged.
The last challenge concerns whether sensorimotor training studies with adults, considered an important source of support for the associative account, are necessarily relevant to the development of mirror mechanisms. This evidence is questionable because the training studies involve behavioral assessments which are inferred to reflect mirror mechanisms, but there is no direct evidence that the underlying mirror mechanisms are modified nor is there evidence that these observed short-term changes can translate into more permanent long-term effects. Also, at the neural level, there are two types of experiential learning (Greenough et al. Reference Greenough, Black and Wallace1987). One type is limited to sensitive periods during early development and is characterized by an overproduction of new synapses in anticipation of specific experiences that will contribute to the development of species typical behaviors, such as locomotion and language development (Bertenthal & Campos Reference Bertenthal and Campos1987). By contrast, experience-dependent processes are associated with the formation of new synapses that develop in response to unique experiences of the individual organism throughout development. If Cook et al. are correct that mirror mechanisms are exclusively a function of an inductive process involving sensorimotor learning, their development would correspond to an experience-dependent process. Currently, this claim is not defensible given that it is just as likely that the early behaviors associated with mirror mechanisms, such as imitation, are species-typical behaviors, and thus just as likely to be mediated by an experience-expectant process which predisposes infants to develop mirror mechanisms.
In conclusion, the target article raises legitimate reasons to question an extreme nativist position regarding the development of mirror mechanisms, but errs in the opposite direction by claiming a strong empiricist position. If a more probabilistic than predetermined view of epigenesis is considered (Gottlieb Reference Gottlieb2007), it is difficult to imagine how a genetic predisposition could not contribute to the development of mirror mechanisms.
As a developmental scientist, I greatly appreciate this target article drawing attention to the importance of clarifying the origins of mirror neurons (MNs). This is no easy feat and I believe that this article is testament to the complexity of the problem. According to the authors, Cook et al., the standard view of MNs is that they represent an adaptation by the organism, and are thus genetically predetermined. By contrast, Cook et al. assert that mirror mechanisms are not predetermined or even probabilistically determined, but instead develop as a function of sensorimotor associative learning. My objection to this position is that associative learning may be necessary, but it is not sufficient for the development of MNs. In the remainder of this commentary, I will present three lines of evidence that question the sufficiency of associative learning for explaining the development of mirror mechanisms: (1) Genetic predispositions interact with associative learning; (2) infants show predispositions to imitate human as opposed to nonhuman actions, and (3) there are differences in learning during early and later development.
The first reason to question a strict associative account is because it cannot explain why some behaviors are learned more easily than others. Cook et al. claim that the observation–execution matching properties of mirror mechanisms are not a specific genetic predisposition, but rather a domain-general process of associative learning found in a wide range of vertebrate and invertebrate species. This position implies that all correlated sensory-motor experiences should be learned equally well by the organism, but this claim is not supported by the data. Beginning with the pioneering research of Garcia (e.g., Garcia et al. Reference Garcia, Kimeldorf and Koelling1955), it was shown that rats could not associate visual and auditory cues with food that made them ill, but could learn to associate olfactory cues with such foods (see Gould & Marler [Reference Gould and Marler1987] for other examples). These findings are not surprising from an adaptive standpoint (though I suspect eschewed by Cook et al.) because rats are biologically prepared to learn some things more readily than others. In the natural world, odor is a more reliable cue than color for rats because they are primarily nocturnal, so odor is better associated with dangerous foods than is color. The conclusion from these sorts of studies is that animals are biased to learn some associations more easily than others even though the contingencies are the same. Thus, it appears that there are genetic predispositions that interact with the success of associative learning.
The second reason to question the associative account derives from evidence showing that infants are biased to imitate human as opposed to nonhuman actions. In the target article, imitation is considered a behavioral index of the presence of mirror mechanisms. If all sensorimotor associative learning is sufficient to explain imitation, then infants should be as likely to imitate mechanical as opposed to human actions. However, the critical evidence is at best mixed. For example, Meltzoff (Reference Meltzoff1995) demonstrated that 18-month-old infants could imitate the actions of an adult pulling apart a barbell but not the actions of a mechanical pincer designed to match the movement, as well as the effect, produced by the adult.
Similarly, Longo and Bertenthal (Reference Longo and Bertenthal2006) tested 9-month-old infants in an observational version of the Piagetian A-not-B search task, and showed that infants still committed the search error on the B test trial because they covertly imitated the search behavior of the experimenter during the A trials. By contrast, Boyer et al. (Reference Boyer, Pan and Bertenthal2011) substituted a pair of mechanical claws for the human experimenter, and 9-month-old infants failed to commit the search error suggesting that they were less likely to covertly imitate the goal-directed behavior of the claws. It thus appears that infants, like the rats described above, are predisposed to learn some sensorimotor associations more readily than others. These findings are thus consistent with the human and nonhuman research literature suggesting that MNs are more likely to become activated to the observation of human actions than to mechanical devices or tools (Liepelt & Brass Reference Liepelt and Brass2010; Longo & Bertenthal Reference Longo and Bertenthal2009; Woodward Reference Woodward1998). If associative learning was sufficient for the development of mirror mechanisms, the response to human generated actions should not be privileged.
The last challenge concerns whether sensorimotor training studies with adults, considered an important source of support for the associative account, are necessarily relevant to the development of mirror mechanisms. This evidence is questionable because the training studies involve behavioral assessments which are inferred to reflect mirror mechanisms, but there is no direct evidence that the underlying mirror mechanisms are modified nor is there evidence that these observed short-term changes can translate into more permanent long-term effects. Also, at the neural level, there are two types of experiential learning (Greenough et al. Reference Greenough, Black and Wallace1987). One type is limited to sensitive periods during early development and is characterized by an overproduction of new synapses in anticipation of specific experiences that will contribute to the development of species typical behaviors, such as locomotion and language development (Bertenthal & Campos Reference Bertenthal and Campos1987). By contrast, experience-dependent processes are associated with the formation of new synapses that develop in response to unique experiences of the individual organism throughout development. If Cook et al. are correct that mirror mechanisms are exclusively a function of an inductive process involving sensorimotor learning, their development would correspond to an experience-dependent process. Currently, this claim is not defensible given that it is just as likely that the early behaviors associated with mirror mechanisms, such as imitation, are species-typical behaviors, and thus just as likely to be mediated by an experience-expectant process which predisposes infants to develop mirror mechanisms.
In conclusion, the target article raises legitimate reasons to question an extreme nativist position regarding the development of mirror mechanisms, but errs in the opposite direction by claiming a strong empiricist position. If a more probabilistic than predetermined view of epigenesis is considered (Gottlieb Reference Gottlieb2007), it is difficult to imagine how a genetic predisposition could not contribute to the development of mirror mechanisms.