Cook, R., Bird, G., Catmur, C., Press, C., & Heyes, C. (2014). Mirror neurons: from origin to function. Behavioral and Brain Sciences, 37(02), 177-192.
This article argues that mirror neurons originate in sensorimotor associative learning and therefore a new approach is needed to investigate their functions. Mirror neurons were discovered about 20 years ago in the monkey brain, and there is now evidence that they are also present in the human brain. The intriguing feature of many mirror neurons is that they fire not only when the animal is performing an action, such as grasping an object using a power grip, but also when the animal passively observes a similar action performed by another agent. It is widely believed that mirror neurons are a genetic adaptation for action understanding; that they were designed by evolution to fulfill a specific socio-cognitive function. In contrast, we argue that mirror neurons are forged by domain-general processes of associative learning in the course of individual development, and, although they may have psychological functions, they do not necessarily have a specific evolutionary purpose or adaptive function. The evidence supporting this view shows that (1) mirror neurons do not consistently encode action “goals”; (2) the contingency- and context-sensitive nature of associative learning explains the full range of mirror neuron properties; (3) human infants receive enough sensorimotor experience to support associative learning of mirror neurons (“wealth of the stimulus”); and (4) mirror neurons can be changed in radical ways by sensorimotor training. The associative account implies that reliable information about the function of mirror neurons can be obtained only by research based on developmental history, system-level theory, and careful experimentation.
I found this article extremely enriching. When I learned about mirror neurons in my neuro-anatomy class, I didn’t suspect there would be objections to their origin or function. I think the evidence made a strong case for the associative model as opposed to the genetic/evolutionary one. If I am interpreting this correctly, mirror neurons were originally believed to facilitate understanding (i.e. something neuroscientific was relevant to cognition). The evidence for the associative learning model however suggests that this is not necessarily the case. Mirror neurons are simply the results of sensory and motor neurons firing together in a context dependent and contingent fashion. Whether they contribute to understanding remains elusive. I also think the fact that something as complex as mirror neurons can be explained by simple associations, suggests the possibility that complex processes can also be established in machines without the elaborate mechanisms of natural selection (although the case of MN in particular does suggest some need for genetically pre-determined “capacity for associative learning in neural tissue”). Lastly I thought that on a broader level this article highlighted the importance of sensorimotor experience. This made me think about the need for T3 to pass T2.
ReplyDeleteI definitely agree with your last point on T3. While the evidence for action understanding is unclear as to whether mirror neurons are involved in “action understanding” or a more perceptual process, the evidence for mirror neurons’ role in action perception and imitation is somewhat clearer. Both the latter and the former are involved in communication. This evidence suggests that sensorimotor associative learning, as the origin of mirror neurons, sets the stage for communication, which is required for passing T2.
DeleteI also think this highlights the fact that we lack an operational definition of "action understanding". As long as we are unable to properly define "understanding" this aspect of cognition will always remain elusive as it cannot be tested empirically. But, as you both note, this paper suggests sensorimotor experience and associative learning rather than "action understanding" may be the key to passing T2. In that sense, attempting to precisely define "understanding" would be unnecessary (and perhaps futile).
DeleteJessica, do you think the discovery of mirror neurons explains how I know when you are making the same movement I am? Having discovered them, do we know how they do whatever they do?
DeleteAustin, Do "associations" explain it (or anything)? And if mirror capacity is required for language, does the existence of motor neurons (where they are, when they fire and what action or capacity their firing is correlated with) explain either mirror capacity or language capacity?
Chloe, "action understanding" sounds somewhat homuncular, don't you think? I know when you are making the same movement as I am: but how do I know it? In fact, what does it even mean that I "know" it. That I can make the same movement as you? Fine, but how I able to do that? Do mirror neurons explain that to me?
This comment has been removed by the author.
DeleteRe: Professor Harnad’s questions. For the first question, I would think that the sensorimotor associative learning explains at least some part of language capacity. From the reading, I would take the stance that it’s not the “action understanding” or the “mind-reading” that matters when we speak of motor neurons (as we discussed in the lecture). Rather, the evidence that Cook et al. present suggests stronger support that mirror neurons matter to the extent that they help us imitate movements. So, I think that motor imitations explain language. To answer how this might be and to answer the second question, I would say that the existence of motor neurons explains mirror capacity, which in turn, explains language capacity. But I would argue that the mirror capacity does not need to be “mind-reading”; it only needs to involve imitations of movements. The crucial point, in my view, for mirror neurons in explaining language is the capacity to imitate others’ bodily movements/gestures. This then leads to language capacity.
DeleteProf Harnad, I think that your comment about "how" makes more sense after doing the readings on categorical perception. To "know" that someone is doing the same movement requires categorization (ie you must "recognize" that movement using its invariant features from a set of sensorimotor inputs).
DeleteIn that case, I do not think mirror neurons can fully explain how that categorization process occurs. It would be necessary to understand the full neural circuitry connected with mirror neurons not only once this association is established but throughout the learning process (assuming that this "understanding" of the action in question is not innate). To answer this question we would need to know how those invariant features are identified (e.g. how do you recognize waving from similar non-waving hand and arm movements).
Can someone please explain to me what they understood by "field properties of MNs"? Here's what I understood, please correct me if I am wrong: motor neurons are mapped in the brain so that all the motor neurons that are linked to movement of the right hand are bundelled up in the same area of the motor cortex on the brain. Simmilarly, MNs related to the visual features representing happiness are all represented in a bundle of MNs spatially together somewhere in the brain, so there are fields of MNs in the brain with groups of MNs related to one action being localized together in the brain. Is that correct?
ReplyDeleteMy answer may not be entirely right, but here is what I understood. Field properties can be translated into “goal” or “purpose” of these mirror neurons. The fact that these are found in the pre-motor area, and not in the motor area might infer that they are involved in the learning, understanding, and planning of movement. Once these mirror neurons have achieved their field properties, they may better inform the neurons found in the motor area on how to execute a movement based on the observed model. Hope this helps?
DeleteJulie, Josiane, the "receptive field" of a visual neuron (or set of neurons) is the part of visual space to which it responds (e.g., upper left, lower right, etc.). The idea of spatial field can be extended to the kind of stimulus or object the neuron is responding to (e.g., 45 degree lines of a certain left, moving rightward in a central field). Notice that once the input becomes a kind or an object, the idea of a "field" is already becoming theoretical, rather than just a matter of observing and describing where in space the stimulus is coming from. In the case of Hubel & Wiesel's dot/line receptors there is still uncertainty over whether what is being reponded to is a line or or stripes of different thickness).
DeleteThe idea of receptive field can be extended to other senses, such as touch, hearing and taste. It can also be extended to movement -- both movement direction and position in space and more details about the kind of movement. But again, once the "movement" becomes a more cognitive "action," with a pattern and a goal, the notion has again become theoretical. That doesn't mean the theory is wrong; but it means that the reference to "field" is more abstract or even metaphorical. (Something similar happens with the even vaguer (and somewhat homuncular) notion of "representation.")
The important thing to note is that once an explanatory notion becomes theoretical (i.e., hypothetical, an inference, a guess) rather than just a straightforward observation, it can be wrong, or, more often, empty: If we rename or interpret neurons that are active whenever I or someone else is making the same movement, we have pointed out an observable correlation, but we have not explained what the neurons are actually, doing, how, or why.
Although I found this reading very interesting, I first failed to see a direct link with what had been covered in the previous readings/lectures. After further thinking, my interpretation is that the refutation of the genetic account as the explanation of the origin of mirror neurons reflects quite well the failure of computationalism. Genetic information, just as a computer program, can't be completely responsible for cognition. The example of MNs show that experience not only shape our thinking, but in some cases can originate it (I find the fact that MNs can encode actions that are not in the evolution context particularly convincing).
ReplyDeleteThe only limit to this interpretation is that it only seems to show that computation isn't sufficient to explain neural behavior, not cognition. It says that T2 isn't sufficient to explain T4, but doesn't say "T2 is cognition" or "T4 only is cognition".
I think that "the human computer program" (if such a thing could exist) is more than just the genetic component of our brains. Genetics are part of our biology which comprise a significant part of course, but I think that MN's bring up a bit more subtle point about how cognition arises. Perhaps genetic makeup can be thought of as the original code onto which the "programmer" originally runs, and thus MNs demonstrate that significant learning must occur beyond the programmed code? But I think that this article better brings into question the role of experience that must exist to form cognition.
DeleteThe association theory of MN's implicates a biological account of critical cognitive development occurring as a result of a human/monkey interacting with the environment and with eachother. Without direct experience, sensory information cannot be cognitively linked to motor information, thus the various proposed functions of MNs - which appear to be quite critical to what a human is able to do - could not have developed.
Therefore is it enough to simulate sensory motor and perceptual information in a T2 robot? Without actual experience of the world how would this link be made in AI's? I see the relevance of this paper as a further justification for why T3 is the correct AI to pass the Turing Test. In order to do what humans do sensory-motor functionality must be there.
I actually disagree that the genetic account and associated learning accounts provide a computationalism versus non-computationalism analogy. Our human genetic code may provide the syntactic programme for functioning but hebbian associative learning is exactly a syntactic function, because it operates on a rule-based: if neuron A and neuron B fires simultaneously or contingently, some operation is applied to them for them to be more likely to fire simultaneously subsequently" This type of associative learning is exactly made possible by a syntactic programme.
DeleteI believe that we could construct a hebbian associative learning-capable machine that is entirely computational that would not have understanding.
From this, it becomes very unclear at what level of symbolic representation is meaning encoded in. The meaning of an action is evidently not encoded at the neuronal level, nor is it encoded at the whole brain level. So it must be at some sort of intermediate circuitry but how extensive that circuit must be is unclear.
Also, I'm still not convinced that the symbol grounding problem can be explained by sensorimotor experience since sensorimotor systems are neural and so, syntactical, and experience is just exactly semantical. It seems to reiterate the question.
Also from this section:
"However, there is still no consensus about exactly what is meant by “action understanding,” or how it differs from cognate functions such as “action perception,” “action recognition,” and “action selection”
(Gallese et al. 2011). Attempts to clarify have emphasized that, in comparison with purely visual processing of action, MN activity relates to the “meaning” of an action and yields a “richer understanding,” “real understanding,” or “understanding from within” (Gallese et al. 2011; Rizzolatti & Sinigaglia 2010)."
I think this attempt at relying on action understanding is misguided because the observability of action understanding is the same as action recognition and action selection. This points again to the other mind's problem.
Mael, the purpose of this reading was two-fold: (1) to get you to reflect on the explanatory power of data on the correlation between brain function and behavior (is Fodor right that where/when does not explain how/why?) and to prepare the road for later weeks in which we will discuss the evolutionary origins of language, both gestural and verbal.
DeleteCassie, even before we learned about MNs we already knew that people could recognize and imitate when another person was making a particular movement. The question to ask yourself is whether the discovery that there are neurons (or sets of neurons) whose activity is correlated with doing that explains how the brain does that? Does it help us generate part of T3 ability? Or does the correlation leave us just as unable to explain how to generate T3 mirror capacity as when we did not yet know the T4 detail that there were particular neurons whose activity was correlated with the behavior?
Yi Yang, is neuronal activity computational (syntactic, formal) if connections get stronger with repetition? Is an oven computational if temperature increases when you turn the dial? The correlation can be formulated as a rule, but is that not just a computational simulation of the neurons' or oven's input/output function rather than evidence that the brain or the oven is computing (i.e., rulefully manipulating symbols based only on their arbitrary shapes)? (It is true, though, that it is possible to simulate a parallel, distributed neural network with a serial, computational algorithm, so that whatever the net can do with input/output, the algorithm can do too. But are real neurons just the hardware implementing either a parallel neural net or a serial algorithm for one?) I agree that the leap from "action" to "understanding" is too quick in this article.
This comment has been removed by the author.
DeleteRE: “To reflect on the explanatory power of data on the correlation between brain function and behavior”
DeleteAfter reading this article, I now know 3 different studies that contributed their findings to the ‘associated learning” mechanism- MNs, Dopamine regions of midbrain structures, and the purkinje and climbing fibers of the cerebellum. Three different classes, three different studies, same explanation. We are explaining random noise in hindsight, and this is leading to problems. I think this problem can be explained by this question: "how can you have an objective measure of a cognitive processes?"(Girl in class, feb 3) . We are looking for mechanisms to a machine that doesn’t have a definitive function. In other words, the function of a heart is to pump, the function of a kidney is to filter, the function of cognition is to ______. If this article was meant to have us reflect on the explanatory power of data on the correlation between brain function and behavior, it seems as though reverse engineering is fruitless.
(3) The genetic account assigns a relatively minor, facilitative role to sensory and/or motor experience in the development of MNs, whereas the associative account suggests that sensorimotor experience plays a major, instructive role in their development.
ReplyDeleteThis article states that the focus of the mirror neurons (MNs) have largely been directed towards the function, at which point the article redirects the focus to the origin of MNs. The third point of the article could be misunderstood as the instance of the “Nature versus Nurture” debate; however, the article clarifies the distinction to be made in detail, by describing the plural nature of processes. It appears that the article defines itself as laying a foundation for the phenomenological flaw of the current research investigating the functions of MNs—the presuppositions that cause the way in which the question is posed.
I found that the argument that mirror neurons originate in sensorimotor associative learning, relating back to previous material, supports why we need the T3 model and why computationalism doesn’t stand. The article shows how important sensorimotor associative learning, to interact with things in the world, is for mirroring and imitation; with mirroring and imitation, we’re now getting into communication. Therefore, it looks like sensorimotor capacity leads into communication, including language, which is needed to pass T2. But before language, correlated sensorimotor experience generates both strictly-congruent and broadly congruent mirror neurons. This correlated sensorimotor experience depends on experience, is built through social interaction, and the sensory and motor neurons fire in a predictive relationship. Relating to T3, these aspects of sensorimotor associative learning mean that any model of cognition has to be able to respond to the “wealth of stimulus” in its environment. Thus, it will have a developmental history and thereby, its environment will be part and parcel of its cognitive ability.
ReplyDeleteIt would be very interesting to explore if other animals have MNs, as a way to establish consciousness. I personally do not doubt that animals such as fish and reptiles possess self awareness, but if they could encode their movements in a similar way to the way humans encode their motor movements in MNs, I believe that would be proof of 'feeling'. In order to recognise your personal walking pattern more than another's, you must know what it feels like for you to personally walk. In the same way, if fish were to encode their swimming patterns in MNs, they would know what it felt like to swim, therefore establishing a conscious mental state.
ReplyDeleteAdrian, I completely agree that other animals like fish and reptiles have some sort of mirror neurons and about your further point about their “feeling”. A question further elaborating on that is: At what point phylogenetically are there no longer mirror neurons? My guess would be that they stop appearing in organisms that no longer perceive and subsequently interact with their environment (ex: immobile jellyfish that rely on currents to move). Additionally, what about animals that live all/most of their lives alone (https://en.wikipedia.org/wiki/List_of_solitary_animals)? A lot of the animals on this list are undoubtably conscious, but when we use MNs to determine “feeling”, doesn’t that leave out at least a few of these?
DeleteKarl, this is a good point. I do agree that many solitary animals are concious but may lack mirror neurons. To add a complication to your point, what about humans who have lived in isolation? We have already established that if humans are isolated from a young age they do not develop language in the same capacity as humans who are exposed to language in the critical period do. Could there be a critical period for MNs, and if so, how would missing the critical period affect humans?
DeleteAdrian, I think it would make sense for there to be a critical window if we believe in the associative account. If there is no evolutionary history, and mirror neurons develop by associative learning, then surely there must be a sensitive period like most other things that require context and contingency for learning to occur. A person/animal living in isolation will not have this window, will not develop mirror neurons, and they will be fine because they won't need it, if actually the purpose of mirror neurons are for understanding.
DeleteOther animals have indeed been studied for mirror neurons. Since the first experiments that found mirror neurons in monkeys, scientists have also discovered mirror neurons in songbirds, which fire when they hear other birds singing their songs. This is obviously another example of very social species, so I agree that it would be interesting to look further at isolated individuals and species.
Deletehttps://www.scientificamerican.com/article/do-dogs-have-mirror-neurons/
Adrian, in response to animals having self-awareness: there are a bunch of cool studies they did using mirrors to test this. I think in the study they found 'higher-level' mammals and some birds definitely do have it, but it would be interesting to see if fish or reptiles could have it but need to be tested in a different way!
DeleteLooking into this idea could also help to understand the 'building blocks' of cognition. I think it's believed that what reptiles/fish/insects can experience are the absolute fundamentals of thought and awareness. Mirror neurons might be clues to direct researchers in the right direction for this.
Adrian, I myself have wondered over the years about studying humans and animals who have been raised in isolation – ethics aside, I think we would be able to learn a lot about the interaction between biology and sociology through case studies like these. A tentative hypothesis (operating off merely instinct through a few years of studying psychology) might be that we all possess the ability to process social cues, much like our brains have developed the ability to process language. While we don't obviously come out of the womb knowing language, we do emerge primed and ready to learn language(s). Mirror neurons may be one of the earlier biological pieces of evidence that our brain development is primed similarly for social interaction – not particularly surprising, as we are incredibly social creatures.
DeleteDominique, I find that very interesting that mirror neurons have been found in songbirds while they listen to other birds singing. In all the articles I’ve read about mirror neurons for classes, they just touch on physical actions. I wonder if mirror neurons would also fire in humans in response to other senses like auditory senses, or if mirror neurons only fire in songbirds while listening to other birds because the songs are important for their species?
Delete
DeleteI am really unsure if many other animals have MNs, but I think it is a problem with the article and the research that we cannot properly define them. If MNs are functionally connected to socio-cognitive behaviour then we should only expect to see them in more-or less social animals (Or basically any animals that benefit from learning behaviour observationally). That said if we define MNs so functionally we may fail to notice them in animals using them in a different manner. Do only animals that learn behaviour observationally have MNs, or do all animals have them because they serve some essential underlying function, but they are only observable to scientists when they are evidently firing in observation?
Are mirror neurones mirror neurones ontogenetically special. Are they built differently, or are they just like any other neutron in a network but gain their mirrorness from their position in the network? Because the declared cardinal feature of mirror neurons is that they fire both under conditions of observation and execution it is likely impossible to observe their firing when the scientists cannot empirically confirm the animal is observing.
Mirror neurones seem to have a fascinating connection to socio-cognitive function. Thus the logical place to look for mirror neurones (if you don’t see it as ethically wrong (V)) is in social mammals such as monkeys, dogs, etc. But from an evolutionary standpoint (the standpoint these authors are backgrounding) one cannot make a good determination of MNs’ function if they are not observed in other animals. Since absence of evidence is not evidence of absence it is entirely possible that MNs (or even proto-MNs) exist in any subset of neurologically complicated animals. Finding MNs in other animals, especially neurologically or socially more “primitive” animals would give point strongly to the evolutionary story these authors are ignoring. However, since the scientists will naturally struggle to find a non-social animal that exhibits imitative behaviour it would be incredibly difficult to identify neutrons that fire both during observation and execution of the same action. However, this does not mean that they are not present.
Regarding the possibility of a critical period for mirror neurones, perhaps there is, but I have also heard of professional dancers who exhibit significantly more MN activity than a non-dancer while watching another dancer (sorry can’t recall my source). Since these were adult dancers who may have begun dancing a different ages, I suspect that any critical period may not be that significant.
I am curious as to other possible emergent functions whose development could be enhanced by the same process found in these mirror neurons. For example, certain physical aspects of language could undoubtedly benefit from this type of “understanding” based on observation. The movement of the mouth when producing sounds is one of the more complex movements an infant must learn, given the range of the phonetic alphabet. The fact that speech is learned so rapidly may hint at a large contribution of mirror neurons in learning the basic (observable) movements of the mouth and face, while the more unobservable motions (ex. movement of the larynx) must be learned independently. Hence why the first sound a baby makes (crying) involves much more movement of the vocal cords than the rest of the vocal tract: they are starting their independent learning (which would take longer) as soon as possible.
ReplyDeleteI am curious to see what function this process might play in emotion – we know that there are some innate reactions (the universal emotions) that are consistent across populations, but I wonder how much they could be modulated by observation at the time of birth. From seeing someone over-reacting, or just not reacting at all, what sort of effect would this leave on the developing child, and the ability to move facial muscles for expression?
DeleteThe argument made in this paper that MNs allows for domain-general associative learning seems fairly convincing. I’m guessing that prof. Harnad will use it to argue that T3 is a necessary for T2. However, before we can make T3 we’ll need to understand how mirror neurones do what they do. Presumably this is where Fodor comes in: how much does all this research really tell us? It might give us an avenue for further research, but the big ‘how’ question is still unanswered.
ReplyDeleteI think you may have gotten it a bit backwards; Harnad argues that T2 is necessary for T3, not the other way around. For a robot to talk about everything we know and talk about, it needs to be able to do those activities first, and understand what they are. Any roboticist who aspires to create a robot that passes T3 will have to create one that passes T2 first. In this way, T3 (and T4 by extension) becomes redundant. Sure, it would be awesome to have a robot that could physically do all human actions, but it would be entirely unnecessary if we already had a robot that could tell us convincingly that it could perform all of those actions.
DeleteIt was mentioned in the paper that monkey mirror neurons have been found corresponding to hand and mouth actions. Likewise, humans have mirror neurons corresponding to hand, foot, arm, and mouth movements. My question is - why is it that mirror neurons have only been found in these areas? Do we think this is just an instance where with time we will find mirror neurons corresponding to every body part? Or, is it thought that mirror neurons are only found to correspond to places where imitation is most likely to occur. Perhaps this would fit into the associative hypothesis because the places where imitation is most likely to occur are the same places where contingent learning will take place?
ReplyDeleteI think that the fact that these mirror neurons are found in the premotor area explain the fact that their expressions can only be displayed in motor movements. Therefore, it makes sense that the hand, foot, etc. are involved since these are the limbs we use for all types of movements. Otherwise, what body part is then left for us to mirror? Or maybe were you making a more abstract reference at mirroring emotional response?
DeleteThe article presents a very convincing claim that mirror neurons are likely the by-product of associative learning, primarily due to the fact that mirror neurons require an association between sensory perception (seeing another do a task) and motor action (doing a task) – an association that is achieved, strengthened or modulated via social interaction. The genetic account’s claim that MNs evolved for “action understanding” raises the question of “what is meant by ‘action understanding,’ or how does it differs from cognate functions such as ‘action perception,’ ‘action recognition,’ and ‘action selection’.” From my understanding, this “action understanding” refers to the process of attaching (or associating) meaning to motor actions. If that is the case, what exactly differentiates “understanding” from “association”? Moreover, what is meant by “meaning”? This clarification is crucial in differentiating action understanding from action perception. Ultimately, it is scientifically proven that mirror neurons link sensory stimuli to motor responses, so why the need to introduce “understanding” into the equation? It seems that the entire argument for “action understanding” is merely an attempt to find the metaphysical within scientific evidence.
ReplyDeleteIn response to p. 178: “The primary focus of this article is the origin of MNs. Our principal questions are not “What do MNs do?” or “What are they for?”, but “What is the process that gives MNs their ‘mirrorness’; their fascinating, cardinal capacity to match observed with executed actions?”
ReplyDeleteIn this article, Cook and colleagues lay out some very convincing evidence in favor of the associative account of mirror neurons (MNs). The more standard genetic account holds that our genetic predisposition to develop MNs evolved because of MNs’ usefulness in action understanding. The authors, however, argue that MNs are able to do what they do via sensorimotor associative learning – independent of their possible function in action understanding. I thought that the central question of the article itself was the best evidence against the genetic account: the genetic account seems to mix up a possible function of MNs with their origin. The authors’ explanation of the genetic account made me think about our class discussions on the reverse engineering of cognition: it's hard to dissociate ultimate function with origin when working backwards.
I found it interesting that the paper included an entire subsection titled "not nature vs. nurture" yet the whole genetic vs. associate view of mirror neuron's simply argue this problem. Did the motor neuron's come first via. evolutionary advantage and develop with associative learning? Or did associative learning come first through evolution which produced these mirror neuron's? I feel like the two view's distinction was somewhat aimless.
ReplyDeleteHowever, I feel like it is more important to figure out HOW mirror neurons do what they do than worrying about genetic vs. associative. Based on the evidence that Cook has provided, it does not seem like the discovery of mirror neurons get us closer to understanding consciousness but simply an example of an act of computation. It seems to reflect a input/output situation where if you were to artificially place an input into a mirror neuron network and sent an electric signal to them that is identical to one that a brain would receive, wouldn't the output be the same? Would we be able to "put in" mirror neurons into a robot and have another robot recognise and imitate one another as a same movement as itself? I think these are some interesting aspects that Cook had not really gone into.
By the associative learning hypothesis, mirror neurons rely strongly on experience and social interactions in order to develop, whereas according to the genetic hypothesis they don’t. The paper mentions that because of these differences, the genetic hypothesis would say that MNs would be uniform across individuals and the associative hypothesis predicts that they could vary among individuals depending on their developmental history and experience they’ve had. I’m wondering if this has been looked into, comparing the MN development and activity among a large group with varying developmental histories, amounts of social exposure and perhaps different ages. By doing this, wouldn’t it provide strong evidence for one account or another?
ReplyDeleteWhat I found connected most in this article to what we have been covering in class, is MNs role in empathy. The article briefly discusses MNs role in the empathy and more so in its role in action understanding. Looking to learn more about its role in empathy, I came across this article by Lamm and Majdandzic, which actually refutes the importance of MNs in involvement in both. Focusing on empathy, the authors argue that “there are many situations in which empathic responses are evoked, without an action or even a person being perceived...seems more likely that the automaticity and apparent predisposition for empathic resonance can be attributed to (early) learning experiences, culture and socialization rather than to hard-wired mirroring.” It is also foolish to assume that the neural make-up of monkeys will be the same as that in humans (this has been shown more recently in clinical trials where treatments on mice have often been failing on humans). Even if we have similar looking systems, how can we know for sure that the same thing is happening in the brain. I think that the multiple ways empathy can be expressed in the brain shows that it isn’t something hardwired into us from birth, and it is a sign of consciousness.
ReplyDeletehttp://www.sciencedirect.com/science/article/pii/S0168010214002314
Hi Eugenia,
DeleteI see your point, but I’m not sure I agree. I don’t think that the fact that the brain is able to express empathy in multiple ways necessarily means that it’s not something hardwired into us from birth. The brain is very plastic, especially in early years. It could be that through early learning experiences, culture, and socialization, our brain alters the way it processes empathy in order to better adapt to our environment. I feel like humans probably have some kind of genetic predisposition to developing a sense of empathy. I think that the ability to understand other people’s feelings is necessary to be able to communicate effectively and develop relationships with others, so there is definitely a survival benefit to being able to feel empathy.
I was also very intrigued with how the mirror neurons might be connected to empathy. Experiments have shown that when people see someone they love/care about experience pain they themselves also feel some sort of pain – is it possible that mirror neurons could be the underlying neurobiological cause? Perhaps just as the audiovisual neurons described in the article connect the visual system to the motor system, there exists some sort of connection between an emotional centre in the brain and the visual system, so that when you observe an event occurring to another person you understand how it feels to them?
DeleteI also think that if this were the case it would be interesting to explore how it all fits in with the evolution perspective. From what I’ve read, there is not as obvious of a definitive survival benefit to being able to feel empathy as you seem to suggest Kristina. I was under the impression that the purpose of empathy and altruism were highly contentious topics; why would it be beneficial to understand how someone else was feeling? Would it make you a more attractive mate? If it was determined that this was the case, and that empathy was a trait inherited through natural selection, I think it would lend a lot of support to the evolutionary origins of mirror neurons.
However, if the associative hypothesis is correct and mirror neurons arise from associative learning this might suggest that empathy is a learned skill.
Finally, it’s interesting to consider how all this would connect to T3. If empathy is a trait that (most) humans posses it would be essential that T3 posses this skill as well. And whether or not we were successful in giving T3 this skill through an algorithm (genetic makeup) or learning (associative) might reveal information about the human origins of empathy
"Many of the field properties cited as evidence of goal (intention) coding by MNs can also be explained by contextual modulation within an associative framework."
ReplyDeleteIf I understand correctly, Cook is arguing that perhaps MNs firing specifically and selectively in different contexts is not due to the genetic account but rather due to the associative hypothesis. He compares MN response to conditioning rats to respond differently to the same tone in different contexts. Is he proposing that a neural circuit responding to a movement is not actually "understanding" the movement but rather responding to the context in which it is occurring? What is he implying by this? Does he assume we have a neural circuit wired for every salient context?
ReplyDeleteRE: ”The associative account is predicated on a “wealth of the stimulus” argument and therefore anticipates MN activity in young infants"
Cook assumes here that all children have MN activity which is developed via ”self-observation, being imitated by adults, being rewarded by adults for imitation, and the kind of experience in which, for example, lip movements make the same smacking or popping sound when observed and executed.”
He later states that whether or not an individual has MNs will depend on “the types of sensorimotor experience received by the individual in the course of their development.” If this were true, you'd assume that children who have grown up in isolation (without social interaction) would not have MN activity. I'm wondering if they have tested "wild" children like Genie for MN activity?
When I read this part of the paper I had the exact same thought as you, Elise. The paper also mentioned that without any sensorimotor experience, the motor neurons wouldn't develop altogether. I thought that if this were true and they were able to test individuals who grew up in isolation, like Genie, then it would be strong evidence for the associative-learning hypothesis.
DeleteFrom what I understood of this article, MNs occur partially from experiences of seeing other people's movements and partially from seeing your own movements (self-observation). Testing "wild" children or isolated animals on MNs could test these theories and give more evidence towards whether seeing one's own movements is enough to develop MNs and whether these are the exact same as typically developed mirror neurons.
DeleteIt would also be interesting to investigate people who were born blind to see if they develop mirror neurons, whether these are in the same area as typical mirror neurons or whether they are more associated with brain areas of other sensory modalities.
Rebecca, your comment about mirror neuron development in blind individuals made me curious. I found this article (http://www.jneurosci.org/content/29/31/9719) that suggests mirror neurons develop in the absence of sight. They demonstrated that familiar sounds (such as clapping) elicited the same activation patterns in (born) blind and not blind individuals. They also showed that unfamiliar sounds did not elicit these responses. This suggests that the auditory experience of the individuals has an effect on the development of the mirror system.
DeleteI wonder if a blind person would elicit these same mirror system activation patterns if the experiment consisted of them touching another human as they carried out the action. If the individual had never seen (or “felt”) someone shrug, would their own experience of shrugging elicit the same firing patterns as someone who had the visual experience of shrugging?
The common belief that the "body image" (in humans) exists generally within the parietal lobe colours the discussion on MN localization with a certain restrictive bias.
ReplyDeleteThe authors suggest fMRI repetition suppression and TMS adaptation as possible methods to help navigate the localization problem by isolating behavioral effects to specific populations of neurons.
Assuming that mirror neuron processes *could* be shown to be task-dependent, firing in cell assemblies with varying overlap and localization depending on the context, the natural next step, as noted by the authors, would then be lesion-studies to observe behavioural competence.
The problem as I see it however, is that the proposed associative-learning mechanisms for mirror neurons makes training (for both humans, monkeys etc.) kind of a self-fulfilling prophecy.
It is hard to expect natural results in a laboratory setting, especially when manual animal lesioning is added, the propensity for MN learning will be inextricably biased towards the conditions of the experiment, and thus not hugely useful in making wider claims about MN development.
“It suggests that MNs acquire their capacity to match observed with executed actions through domain-general processes of sensorimotor associative learning, and that the role of MNs in action understanding, or any other social cognitive function, is an open empirical question.”
ReplyDeleteHow far does the domain-general claim that MNs are a by-product of associative learning go? Do MNs underlie associative learning between other sensory modalities and relationships besides the association between the visual perception of seeing another perform a task and physically performing the task oneself, in the motor modality. Can a claim be made that there exist MNs (in another part of the brain) that fire when we smell a freshly baked bread and also fire when someone tells us they are experiencing that smell or we see others walk into a bakery?
RE: Thus, the associative hypothesis implies that the characteristic, matching properties of MNs result from a genetically evolved process, associative learning, but this process was not “designed” by genetic evolution to produce MNs. Rather, it just happens to produce MNs when the developing system receives correlated experience of observing and executing similar actions.
ReplyDeleteI'm wondering what people think is the primary purpose of the mirror neuron in this hypothesis. The associated hypothesis explained that MNs came from a "associated learning" this suggest that these neurons were designed to give the organism some sort of social advantage. These neurons fire when "observing and executing similar actions" what is the advantage of these neurons doing so? If a monkey sees the action performed again and again does it speed the learning process so that the same neurons are firing which allows them to learn faster? Can learning faster relate to social cognition?
I have the same wonder too! I wonder if mirroring/ repeating an action (i.e. MNs associative learning in the neural level) is how social interaction first formed? The act of imitating reminds me of a social psychological phenomenon where people on the street just started to line up when they see a few people are in line. Somehow I wonder the first ever learning process is made by imitation too.
DeleteI think that social interaction is definitely a possibility as to why these neurons formed! That’s an interesting example about imitation, do you mean that people will line up if others are even if they don’t know what’s going on?
Delete(1) Sensorimotor learning plays a crucial, inductive role in the development of MNs, and, because of this, (2) we will get reliable information about the function of MNs only by applying an approach based on developmental history, system-level theory, and rigorous experimentation.
ReplyDeleteAll together this paper is considering the ideas of nature v nurture and monkey see monkey do associative learning. Cook et al seem to fall into a trap that places too much priority on rigid genetic encoding. Christina Behme points out, appropriately, that the acquisition of knowledge is highly flexible and that “mirror neurons for words encode recognition of the articulatory form… must be linked to other neural networks for the encoding of meaning.” While mirror neurons are a great resource to explain associative learning, I feel they are only one step to the learning process in the brain.
One of the evidence supporting the argument in this article is that mirror neurons do not consistently encode action “goals”. It seems to me that most of the artificial intelligent models are designed to have a “goal” and what happens is that they run in order to reach the goal. However, it is not (necessarily) the case for MNs and humans. I wonder if it is one of the reasons why reverse-engineering / AI models cannot tell much about cognition. Yes, cognition is not JUST computation, but computation could be part of cognition. I wonder, could the discrepancy between brain (not goal-oriented) and goal-oriented artificial neural network/AI be a step that we have to come across, in order to get closer to knowing cognition (or at least the part on computation)? What happens to a computation model if it is developed to be not consistently encode action goals?
ReplyDeleteI’m not sure I understand the refutation of evolution for mirror neutrons in section 4.1, is the author arguing that because mirror neutrons respond to sensory information that was not present in evolutionary circumstances that it cannot have evolved? Unless I’ve misunderstood, this would also mean that for humans to have minor neutron activation watching someone dance the Tango there would have needed to be a pre-historic Tango for them to evolve to. This seems like a very bold claim.
ReplyDeleteIn addition, section 8.1 describes ‘Hebbian learning’ as a process which relies only on contiguity, and ‘associative learning’ as relying on both contiguity and contingency. This seems like a misunderstanding of Hebbian learning, which to my understanding relies on repeated and causal activation of a neuron (as the author claims associative leaning does). Is the author’s distinction here a contemporary one, or have I misunderstood Hebbian learning?
Edit: *Neuron, not neutron
DeleteIgnore my second pert, the ‘Open Peer Commentary’ section cleared it up for me (Christian Keysers, David I. Perrett, and Valeria Gazzola) It seems like they are different, although all the neuroscience jargon is going a bit over my head
I think the argument concerns more so mirror response to information which is not goal-oriented. Since mirror firing to sensory information (what they're calling object-free movement) is not-goal oriented, they don't respond to survival necessity alone. This leaves a weak proof that mirror neurons evolved out of evolutionary necessity if they are not serving the object-directed function.
DeleteFollowing this logic it seems like mirror activation when watching another dance the Tango would be an acquired reaction (experience with dancing it, or having seen others dance it). This would make an argument for the associative hypothesis they're rooting for.
Although this is not one of the most vital parts of the article, something that I found interesting was the fundamentals of how MNs are created. It was noted that as the same visual and motor neurons fire together in a dependent way, these connections become strengthened, causing motor neurons to "become" mirror neurons. This raises concerns as to what exactly are mirror neurons if they are other neurons that slightly change function of firing patterns. It also brings up questions as to why these particular neurons are the ones firing? Why do they get mapped onto motor neurons and not visual neurons? Do all visuomotor actions eventually lead to mirror neurons? Why does this hierarchy that they suggest occur?
ReplyDeleteThe genetic account still agrees that "motor and sensory experience plays a facilitative role in development of MNs,” while the associative learning account assumes that “gene-based natural selection” has played a background role in the process of developing MNs. It seems that the two camps are in agreement of one another’s main argument, but are in disagreement of which should play a more important role.
ReplyDeleteWhile the evolutionary argument for mirror neurons and other brain functions like face processing is often debated, it is important to consider its validity. Although we may not be born with complete ability to recognize faces nor to recognize and mimic certain motor abilities over others, our attention is still somehow driven towards these things at a young age. Therefore we must have some innate mechanism that tells us it is important to encode particular experiences and visual or sensorimotor stimuli. Because we are aware of the brain's plastic abilities, then it would make sense that as a human undergoes certain sensorimotor experiences, certain neurons will encode those experiences more strongly than others. Although this supports the associative based learning hypothesis, it doesn’t discredit the genetic account because maybe this type of plasticity is indeed adaptive. Isn’t it also possible that while some circumscribed mirror neurons may be more specialized for sensorimotor learning, others may be independently designed for genetic and evolutionary purposes? These neurons could be grouped together as “mirror neurons" simply for having similar functions and properties, but they may have completely different origins.
In the paper I found the genetic account very interesting because it stresses an innate and evolutionary component. The strict view suggests that certain brain areas are made for MNs. However, this is a problematic view because it entails that what MNs respond to is largely predetermined. Then specializing MNs for novel actions seems like it would be difficult.
ReplyDeleteI also found the associative theory of MNs interesting in how it could be related to AI. The associative theory says that MNs needs sensorimotor experience. This is in support of the T3 level for AI. Maybe the associative view could explain why the T3 level is the strongest level.
I’m having some trouble separating the genetics account and the associative learning account because they seem to be pretty intertwined (maybe that’s the whole point of the latter, since they are hoping to separate the origin from the function of MNs). I’m not entirely sure how you can argue against natural selection for MNs – or at least the processes that favour the growth of MNs – since (a) there are so many ways I can imagine these helping human beings to survive and (b) there are so many mirroring actions that on the surface seem to separate humans from other animal (language being the big one).
ReplyDeleteI don’t know if I am convinced that there can be any associative learning at all if there weren’t even a few preliminary MNs present. The trajectory in development of these MNs, I can see that being affected greatly by individuals’ sensorimotor experiences; but I’m not sure how there could be any initial mirror learning if there wasn’t already some mirror mechanism in place…(have I missed the point? Was this addressed in the article? It felt unanswered to me after I completed the reading…)
I second your question regarding whether associative learning can occur without some genetic predisposition. The article following Cook's original paper sheds some light on this as the author points out that Cook's domain-general hypothesis is not supported by the existing evidence.
DeleteAlso, from a computational standpoint, there have been no models developed thus far that can fully support the associative perspective; the existing ones all have some existing connectivity and input representations (Bonaiuto, 2014). Of course, as we've discussed repeatedly in class, cognition is not just computation. However, reading Bonaiuto's article made me increasingly aware of the gaps in Cook's push for the associative model as it fails to provide a satisfactory explanation for how mirror neurons may function in a dynamic fashion.
My understanding is that Cook seems to argue that there is a line drawn in the sand between the associative vs. genetic view. It seems arbitrary to decide that the only driving force behind these MNs is their function. This form follows function argument is convincing in this article but also (I think) difficult to justify in a biological model. Darwin's theory of evolution suggests that form precedes function (while Lamarck's long discredited stance was that form follows function), thus Cook's stance entirely discredits the biological nature of this behaviour. I'm inclined to agree that the mechanism by which these emerged is indeed this coincident firing in a specific context. However, ignoring the biological/neurological context of the form of these MNs lends itself to the multiple realization theory of cognitive activity. This theory often remains disinterested in neuroscience to preserve the connectionist conception that rules or symbols have to be embodied into networks and often fails to take into account the insight that can be gained from looking at the biological or genetic behaviour of their models. Thus, Cook can easily claim to have explained how MNs function when he entirely dismisses the question as to why did they fire together, and why was this useful to the model, what did the model learn from this?
ReplyDeleteProbably MNs are not enough for the understanding for consciousness or cognition. For example, how does the audiovisual property of MNs explain for blind people who can also learn from the sound in the world? In previous research, it was discovered that there are indeed “visual maps” which are activated in blind people’s cortex when they try to perceive the world around them. As such, if MN projections are genetically predisposed, how are them activated by those “to-become-audible” stimuli instead of visual cues? Moreover, how does a learning hypothesis explain for those who become blind only later in their life (e.g. because of an accident) while still holding the capability of learning from perceiving, e.g. the sound of, others in the world? Does the activation of MN overlap with some visual areas that are active in both normal and blind people? Thus, is there a complementary mechanism to the mirror neurons?
ReplyDeleteIt is interesting to look at the contending arguments regarding learning/environment and genetic hypothesis trying to explain the functions of MNs. Indeed, it is hard to separate the environmental influence from the genetic predisposition. However, some questions remain undiscovered after reading this paper. For instance, how does a learning hypothesis account for behaviors facilitated by the MNs? Can it be explained by a near transfer or far transfer mechanism in the brain since sometimes “different object-directed actions with different intentions” are seen in observe and execute conditions? As such, is the associative learning formed by a confirmation of the observer’s prediction of certain action from experience, or are functions of MNs more likely to be genetically predisposed thus one is just imitating and exploring the consequences of different behaviors?
ReplyDelete”The associative account is predicated on a “wealth of the stimulus” argument and therefore anticipates MN activity in young infants".
ReplyDeleteThis quote brings up a Chomskian argument. The rate at which babies and children learn language does not occur at the rate at which they are taught therefore some part of language, its rules, must be innate. Following the point Victoria brought up, not all of the actions encapsulated by MNs can be learnt from the get go. And not all actions arise from mirroring others. Mirror neurons must have some biologic or genetic base or rule system or mechanism for these infants to activate causing "MN activity".
Since MN's are so entrenched in motor and visual activity as they are the most, I wonder if there are MN's associated with motor and hearing or tactile information but perhaps are not as common? If someone became blind in their twenties would their motor neurons undergo neural degeneration from lack of input and are they replaced by anything if that's the case?
RE: Adaptive Function
ReplyDeleteI found the argument referencing the visual word form area (VWFA) of the brain to be particularly compelling, as the VWFA area evolutionarily predates its supposed adaptive function (reading), much in the same way that mirror neurons (such as those in the gorilla) predate language. Another comparison might be able to be drawn to the parts of the brain relevant to the appreciation of music – they are utilized for functions that they were not necessarily “designed by nature” to carry out.
What do these comparisons implicate about theories of language such as Universal Grammar? If we assume that natural language developed from the imitation capacities of mirror neurons (which is what we are calling into question here), and if the functions of mirror neurons are learned rather than innate, would this imply that language is learned rather than innate?
RE: “MN activity relates to the “meaning” of an action and yields a “richer understanding,” “real understanding” or “understanding from within”
ReplyDeleteAlthough the definition of “action understanding” is somewhat ambiguous, if a T4 were built with this MN capacity, it seems as though one would have a causal, Searle-proof, explanation for cognition.
Searle’s objection to strong AI in part targeted the fact that the symbols in a T2 were just shape manipulation (without any meaning/understanding). The MN mechanism, however, provides an explanation for how symbols may be grounded in meaning. The mechanism does so by mediating/integrating sensorimotor information into our understanding.
For example, “MNs acquire sensorimotor properties whenever individuals experience a contingency between ‘seeing’ and ‘doing’.” The fact that MNs are altered depending on sensorimotor information shows that there is a bi-directional relationship between MN activity and what we can do.
Oh, this paper. I remember I was somewhat frustrated getting through this paper, because I felt that there is little to no relevance in the MN’s discovery to progress our understanding of cognition. But, I will attempt my stab now: the paper aims to answer the question of the process that gives MN’s their capacity to match observed with executed action, or, as they said, their “mirrorness”.
ReplyDeleteCook went on about the genetic and associate alternatives, which seems to be vague and unclear as to how it contributes to both their question and the questions of cognition. If we put aside the fact that MN does not explain anything, the path from MN to imitation is mostly straightforward, and even MN to empathy as well. Such that: when we feel that emotion, we would do something, such as frowning, that others could recognize the action and correlate it to what we are feeling when we perform the action (I believe similar to what we discussed of “mind-reading”).
However, bringing back to our point, that the discovery of MN is just another brain correlation. And, as we all know, that correlation does not imply causality. Perhaps we know that the standard scientific method attempts to infer causality from a well designed randomized controlled experiment. So perhaps this MN correlation can sheds light on further research, but even with just working off of the neurons, will we really be able to answer the big question?
To this end, we still don’t know HOW the brain or these newly discovered MNs allow us to detect that others is making the same movement. It seems that the existence of MNs can’t really explain much on how we can do all these…
Hello, I really like your comment, as it constitutes a very nice summary of the content and "future directions" in this article. I agree that the "how" still remains to be investigated, since correlation will not explain causation. I'd say that this paper is useful though because it criticizes the vagueness of the concept of "goal" as it is sometimes used in psychology (for example in evolutionary psychology, i often hear people talking about selection as if it had some sort of volition by itself).
DeleteI find the associative hypothesis quite compelling in that it allows for a more flexible view as to the purpose of mirror neurons, whereas the genetic hypothesis with its emphasis on goal- orientation is a more deterministic view. Yet the idea that mirror neurons are formed when the brain links two actions or an expression and response together, is not an intuitive to me either.
ReplyDeleteRelating this to the idea of an AI which is functionally indistinguishable from a human, I wonder when designing or attempting to create this hypothetical being, which view of MNs would be more compelling. Would we want to hard-wire certain capabilities into it (genetic code) which allows it to empathize and relate itself to others? Or would we want these neural indicators to arise and coalesce in unison with learned responses?
Learning that mirror neurons have been found in brain regions such as the inferior parietal lobule, which is concerned with language and body image, definitely indicates an association between these neurons and the communicative abilities of humans. This made me wonder whether mirror neurons are implicated in mimicking other people’s mannerisms. When you spend a lot of time with the same person, you have a tendency to “catch” their mannerisms and start producing them yourself. My assumption would be that by continuously watching another person perform the same action, a particular group of mirror neurons gets strengthened, making the behaviour more likely to be produced. I think this ties into the idea of being able to learn actively from your environment, and points to the necessity of the T3 model. Sensorimotor associative learning allows us to interact with our environment, and thus to adapt and learn from it. I’m not sure if this ability is present in T2. It seems to be that any machine programmed to operate at T2 would be limited in the sense that it would only be able to operate with the “knowledge” or the capacities that it had been pre-programmed with. Thus, even if we were able to successfully reverse engineer cognition using a T2 model, I feel like it would be a static representation of cognition. In T3, it’s possible to focus on reverse engineering a more dynamic representation of cognition, which, in my opinion, seems more valid as our mental capacities are consistently being altered and added to on a daily basis.
ReplyDeleteThis article uncovered that the discovery of mirror neurons do not EXPLAIN how we do what we do; how we empathize, mirror a movement, or “think”. It explains a neurological mechanism that probably has a part to play, but the discover of MNS does not provide us with the whole picture. Like was said in class, through FMRIs, the study of MNs can (vaguely) help us get at the WHERE and the WHEN of what is happening in the brain, but that is about it. Intellectually honest people cannot be satisfied by how much the discovery tells us about the HOW. Correlation is not causation.
ReplyDeleteDespite the cynical tone of this article, our fMRI studies have indeed gotten us somewhere. They allow us to reason that mirror neurons link sensory stimuli to motor responses. Now, the response to this discovery shouldn’t be, wow cool! What a leap forward! We understand the mechanism of empathy and learning via association! But instead, okay, we have a tiny little piece, now we need to formulate some good research questions to learn more about this mechanism and its role in learning. We should also make AI models as we go, to check our understanding.
I agree with the author that we have been overly zealous in our satisfaction with the explanatory power of MNS in our behaviour, and especially in ascribing the evolutionary WHY to this mechanism. However, I don’t share the same cynical attitude. It’s true, if scientists are overly zealous in how they present their discoveries it generate misconceptions. The public should be invited to think critically about how MNs work, their place in the larger brain and in the complexities in their functioning to the public. I think this is a really important article. But, I don’t think that theorizing about the larger role of the MN is a bad thing per se. We need to work on all fronts in order to solve the easy problem, and some people need to theorize based on the latest evidence, even if there is limited empirical data. However, these theories should never be presented in a way that makes out that we know more than we know, that's intellectually dishonest and turns research into MEsearch.
The discovery of mirror neurons and illumination of their origin and function is not, I think, a worthless task. Gaining a deeper understanding of the types of neurons and neural circuity within the brain can have many clinical implications and can illuminate the overall computational structure - that is, not the specific computations performed, but the function of the circuitry - of parts of the brain. However, there's an absolute limit to how much knowledge of the activity, location, and origin of specific neurons can contribute to cognitive science (as opposed to brain science), in the sense that this knowledge on its own does little to causally explain cognition and performance. That said, I don't think it's wise to dismiss neuroscience in general in solving the easy problem, given that the brain is as yet the only system capable of producing all the behaviour we are interested in explaining.
ReplyDeleteThe assumptions of the associative hypothesis on Mirror Neurons contain much stronger evidence and likelihood than those of the genetic account. For one, strong experience-dependence suggests that correlated sensorimotor experience plays an inductive or instructive role. Learning must certainly be correlated through excitation of both sensory and motor neurons coding similar actions. Hebbian theory states that such synapses that fire together will wire together. With enough repetition, these mirror neurons can form associations with a subject's sensorimotor control. The idea of social construction also seems to be quite accurate with its multiple examples. When an individual experiences a contingency between seeing and doing, MNs acquire stronger sensorimotor properties. Infants learn through correlated sensorimotor experiences such as self-observation, being imitated by adults, and being rewarded by adults for doing the imitations themselves. All these seem much more intuitive than the idea that natural selection is responsible for the development of mirror neurons. Sensorimotor experience plays an active role in shaping such neurons and to say they are just passive would reduce the existence of MNs to chance. This paper also discusses the importance of context effects in the function of MNs and it seems that learned responses are often subject to contextual control.
ReplyDeleteHow can we further test for the viability of either account (genetic/associative)?
Should we look for the origins of MNs before we try to figure out what exactly it is that they do?
Are there other pathways besides sensory, visual, and motor, that can activate MNs?
What effects would we predict to arise if these MNs are lesioned or strongly inhibited?
RE: Relating the “mirrorness” of mirror neurons to their origins
ReplyDeleteI found one of the major arguments towards the associative aspect of MN, relying on sensorimotor information very interesting. The original view, more in line with the genetic argument, purported that mirror neurons encode the end goal of a certain action. Although we know many cognitive processes rely on goal directed behavior, neurophysiological findings did not support this encoding mechanism. I think the most interesting suggestion is that MNs fire not directly due to visual input during an action, but rather as a sensory predictor. In this manor they act as prediction error signals. As we know the dopamine rewards system facilitates information in a very similar way, DopamineResponse = RewardOccurred – RewardPredicted. Could this signal underlie more cognitive processes? And if so, wouldn’t this lend more information as to the computational processes that our brains undergo?
It’s interesting how you compare the function of MNs to the dopamine prediction error signal. If you consider how associative learning is often mediated by dopamine signaling and how this dopamine circuit also plays a role in promoting social interactions, it’s intriguing to think about how closely tied to social cognition MNs potentially are as well. If the MNs’ function is somehow involved in the circuit implicated with associative learning and social reward, what kind of potential could this have for future experimentation? Would it be possible to show MN’s involvement in associative learning by seeing if MN function is affected if the dopamine signaling that encodes the brain’s value of a certain reward or goal-directed behavior is experimentally manipulated?
DeleteRE: Evolution to have MN
ReplyDeleteSo if MN were evolved to help us facilitate understanding, I'm curious as to what this could mean for the future of the brain. In hundreds of thousands of years (assuming life is still here!) could neurons actually simulate things when watched? Communication could also become much easier if MN evolved further. I think it's very interesting how it could relate to empathy and how we can experience what others are feeling. As farfetched as it may seem, I wonder if MNs are or could be a subtle form of 'telepathy'.
I guess MNs are not like “telepathy”, they are more like imitating and thus understanding and learning from your own similar experience?
DeleteWhat I was trying to say was that this could be where the evolution of neurons could be headed. Right now they imitate what we see in other things, but I was just trying to think of/imagine how they could develop in the future. It's very powerful to be able to stimulate another being's neurons just by watching and I was wondering how this could be applied to communication in the future.
DeleteI would be interested in their future clinical relevance. Neurorehabilitation may be an interesting domain for further investigation. I read a bit about the system's capacity for cortical reorganization driving new therapeutic approaches for stroke patients.
DeleteBased on the authors’ evidence, I agree with their argument that mirror neurons were not produced by evolution for the purpose of aiding with action understanding, but were likely part of a more general pattern related to visuomotor control, and produced through the association of sensory inputs and motor outputs throughout a person’s experience. I find the most convincing evidence to be the fact that mirror neuron properties can be changed, in a relatively short amount of time, through sensorimotor training. One example was the experiment in which people watched a pincer opening and closing while performing the movement with their own hand, and eventually the pincer elicited as much mirror neuron activity as a human hand performing the same action. Similarly, when people watched a movement of the index and pinky fingers while moving the opposite finger, they eventually developed reversed mirror neuron activation such that the areas that originally responded to the movement of the pinky finger responded to the index finger, and vice versa. This, along with the evidence that mirroring does not take place in infancy for the most part, demonstrates that young children would have had enough stimulus to facilitate the development of mirror neurons by the time they are able to mirror actions.
ReplyDeleteAdditionally, the evidence that mirror neurons are not only activated by goal-oriented actions, but also by pantomimed (object-directed actions without an object) and intransitive (e.g. tongue protrusion) actions was interesting as this demonstrates that the sole function of mirror neurons is not to facilitate the understanding of goal-related actions. However, I think that, as the authors mentioned, it is important to have a clear definition of action understanding, because it is unclear to me why a person couldn’t “understand” an action that was unrelated to a goal, for instance understanding that a pantomimed action represented a certain goal, or that a transitive action was merely a movement without a specific goal. Finally, the fact that mirror neurons are only activated when the sensory input and motor output are linked both by contiguity and contingency is not something I knew previously. It would be difficult to image how mirror neurons might have been evolved genetically to facilitate action understanding, since a person could not be born with the knowledge of which inputs and actions are contingent on one another, and thus when mirror neurons should fire.
I think that the arguments of the genetic account are part of a larger problem of assuming the evolutionary purpose of something based on its current function. I think that action understanding is too vague of a construct to properly test, and the assertion that mirror neurons facilitate it is still an inadequate explanation of how “action understanding” takes place— for instance, if the firing of mirror neurons is what allows a person to understand the goal of an action, this does not explain how the sight of an action causes these specific neurons to fire. It would seem that there would already need to be some kind of understanding for this to take place. I think the associative account is right to look to more general processes such as visuomotor functioning as evidence of how mirror neurons were developed to function as they do.
I believe the analysis of Mirror Neurons could be more effective than analysing the problem of symbol grounding. Given what the paper says, the implications are that a person who lived alone in the world (first of their kind) would be predisposed with MNs, and that if the person were to procreate somehow then the MNs would develop and enhance – evolution. But where did the initial setup come from? Would it be based off animals in the environment?
ReplyDeleteThis leads me to what the potential definition of consciousness could be, because it appears as if the way we define consciousness now, in terms of I-language, computation, symbol grounding, relies on multiple interactions with other people. But sure, there still would be some sort of I-language with just one person in the world, and it would probably be image based and involve some computation, especially if the person wished to survive. The person is both conscious because they are predisposed with these traits, but not conscious because they cannot prove that they conscious by TM.
The associative hypothesis says that we rely on the visual system and the motor system, and experience plays an important part and claims that MBs are forged by sensorimotor learning, which makes sense in the scenario above. Associative learning appears to be what lead to the development of MNs. How much then, of what we can do, is actually predisposed?
It seems like this article provides evidence that computationalism alone is not enough to explain cognition. This research indicates that, for something to cognize, it must interact with its environment. Mirroring and imitating things in the world seem to play an important role in a creature’s ability to convincingly pass a Turing Test, which would require social interaction, knowledge of language, etc, and perhaps a creature’s ability to “think”. This seems to indicate that environment and social interaction, and having the sensory-motor capacity to interact engage with its environment, is crucial for cognitive development, and, thus, human cognition. Would machine that could not interact with its environment (ie. cognition alone) lack something in its cognitive capacity? The evidence from mirror neurons seems to indicate that that may be the case.
ReplyDeleteThe article made me question whether this correlational data on mirror neurons (a set of neurons firing when an action is performed or observed), has any weight in terms of explaining the ‘hows’ of cognition. Unfortunately mirror neurons do not tell us how we are able to know that a person is making the same movement as we are. Having a specific set of neurons fire is not sufficient in explaining how primates are able to make the same movement as whomever they are imitating. Correlational data is not causation.
DeleteI had previously learned about mirror neurons as discoveries that served as a possible explanation for the development of language in humans. At first I could not see how mirror neurons has anything to do with cognition. The associative theory discussed in this article shows that mirror neurons need sensorimotor experience. Kara questions whether or not a machine that could not interact with its environment lacks something in its cognitive capacity. I believe this machine would be lacking a crucial feat of cognition. Like we have discussed so far in class, a T3 robot has to have ability to use sensory modalities to interact with its environment and only then will it be able to fully pass the TT and appear to cognize as humans do.
I found this article very interesting. I thought that the authors made a strong argument against the genetic hypothesis. It did make me wonder about the associative learning account and how our mirror neurons would differ from others. I think that’s a more sound explanation than the genetic hypothesis. It does make sense, to me, that your mirror neurons would be more active in response to certain actions that you are more attuned to. For example, I would definitely expect a child who is learning to play the piano to have mirror neurons that are more sensitive to watching another pianist play. They’d be more aware of certain actions that the pianist is doing and how that will affect the consequent sound that is being produced, whereas if you took someone with zero experience in piano, I can’t imagine that their mirror neurons would fire as much. What use would mirroring a pianist do for them?
ReplyDeleteThis reading brought to mind the argument of social constructivism vs biological determinism. Cook et al support the associative hypothesis over the genetic hypothesis for the development of mirror neurons. The way I interpreted it, is that the genetic hypothesis is based on the philosophy of biological determinism. The genes for MN predisposition were naturally selected until they became universal. The associative hypothesis is based on the theory of social constructivism. Through associative learning, the neurons that fire when we observe an action are conditioned to fire when we perform the action, by strengthening the connection over repeated experiences. I tend to more strongly come down on the side of social constrtuctivism in general, so I found cook et al’s argument for domain-general associative learning to be strong. I like how they stated that mirror neurons are “built through social interaction” which in a way outlines how social experience can have measurable, tangible effects on physiology, displaying the powerful interaction between biology and environment. Nothing can be exclusively biological or exclusively social, although it may be difficult to tease out the differences between the two.
ReplyDeleteIn the system-level theoretical approach to the debate of whether “MN activity causes or constitutes ‘action understanding’”, I found it a little difficult to distinguish between the interpretations of MN activity that would lead to answering the hard problem of cognition and those contributing to solving the easy problem. I am under the impression that choosing the alternative explanation that “firing of MNs during action observation is, in itself, a form of ‘action understanding’” would lead us to the dead end of hard problem of cognition for which we don’t have any empirical ways of finding an answer. But if we go with the first option in interpreting MNs’ function stating that they only cause action understanding by being a contributing factor in generating the competence for action understanding, then all we have available to measure is behavioral manifestations of such competence. I can’t see how this approach would not encounter the same problems that behaviorism did. Is there any connection between accepting the second alternative for the role of MNs in action understanding and finding the answer to the easy problem?
ReplyDelete