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.
In my opinion, the section that discusses how sensorimotor motor training can alter mirror neurons was the most convincing support for the associative account of mirror neurons. The fact that sensorimotor experience can enhance or reverse mirror mechanism activity is evidence that there must be at least some amount of associative learning occurring; otherwise what would account for the changes? Taking it one step further, the fact that sensorimotor experience can actually induce mirror mechanism activity (which I took to mean create mirror mechanism activity where before there was not any) goes to show that there is no way for there NOT to be associative learning occurring. However, I don’t think this completely rules out the genetic account either.
ReplyDeleteAs described in the reading, the genetic account and associative account are two different ends of the same spectrum. I wonder if development is mutually exclusive and all mirror neurons must follow the same account (i.e. all genetic). On the other hand, could some be strictly genetically determined while others develop more from associative learning? Keeping the spectrum in mind would allow us to examine a middle ground (or mixed development pattern) when researching what the process is that gives motor neurons their “mirrorness".
I think your comment about if MNs differentiate into purely genetically based and purely associative learning based is interesting and to add onto it maybe the neurons go through both stages throughout the life span. I'm thinking maybe when we are young and have more brain plasticity neurons might be more susceptible to outside influence meaning they would act more associatively and over time as plasticity decreases a genetic component may play a greater role in regulating their functioning.
DeleteWe already know that human beings are capable of doing the kinds of things MN activity is correlated with: recognizing when others are doing the same movement as themselves, imitating movements, recognizing and producing vocalizations, etc.
DeleteWe also know that our brains are doing these things for us.
The goal of cognitive science is to reverse-engineer the mechanism that generates our capacity to do these things: to explain how and why we can do it.
Genetic/evolutionary explanations help answer the question of why we are able to do these things: their adaptive value.
But do MNs explain how? Knowing that there are MNs that are active when I move my arm and someone else moves their arm, does that explain what MNs are really doing, and how? Does that allow me to build a system that has these mirror capacities?
Does "association" help explain them? And help me build the system that can do it?
In addition to the effects of sensorimotor training, I also found it compelling when Cook et al. note that the foundation of the genetic code necessitates the assumption that all of the associations we have would be laid out before birth. Meaning that MNs code the "mapping between a fixed set of observed and executed actions".
DeleteAlthough they qualify this as the "starkest form" of the genetic hypothesis - it still seems that the logic of this account rests on 1:1 fixed mapping. With this logic in place, it seems clear that alterations via training undermine the fundamental ideas of the genetic account.
“Rather, it just happens to produce MNs when the developing system receives correlated experience of observing and executing similar actions.”
ReplyDeleteI’m not seeing the extent to which the genetic and associative hypotheses vary to the level which the authors are arguing but the “just so” account argued feels lacking. The relevant neurons involved must have a specific genetic component involved – otherwise, we couldn’t answer the question “why are these neurons responding and not others?” in the degree to which we see in MNs. Why aren’t all neurons in the general associative learning category acting as MNs? The associative account seems to provide novel predictions and has evidence which corroborates it, but it doesn’t appear to be an explanatory account that particularly excludes a genetic basis for the origin of MNs.
What does association explain, and how?
DeleteIn this article, mirror neurons are examined through the lens of an associative model and in the end the authors are convinced that the associative model seems to be the most likely explanation as to how they come to possess their ability to encode “action-understanding” compared to the genetically encoded hypothesis. I do agree that there are pitfalls to the genetic account as there are pitfalls to the associative account as well as in more moderate accounts which cannot provide us with sufficient empirical data in support of them, however, I am skeptical that any account can ever be able to account for such a concept. The article mentions that the dissociation between motor ability and action understanding in some patients seems to demonstrate that action understanding is not inherently tied to motor ability. If this is taken to be true then where does action understanding comes from if it actually exists at all?
ReplyDeleteThe data which has been found to support the concept of mirror neurons being recruited to perform this task relies mainly upon related MNs firing at the same time and in a predictive manner (ie: contiguity and contingency respectively), but how does this transfer to true understanding? A neuronal connectivity technique called resting state MRI tries to see which areas are connected to each other in the brain by looking at which areas coactivate in a predictive manner while a person is at rest. Resting state MRI only provides correlational rather than causational data. Though the contiguity and contingency of mirror neurons may provide more causational information about neuronal firing and connectivity, the fact of the matter is that it does not provide any causational evidence of action understanding. Can we really say that the associative model of MNs really provides any evidence at all for true action understanding or better yet can it ever?
What is "action understanding"?
DeleteMN activity correlates with the exercise of certain capacities we have (roughly, recognition and imitation of movements of others). That capacity is partly inborn, partly learned. So MN activity is likely to be partly inborn and partly learned.
But what is the causal mechanism?
Professor, what do you mean by what is the causal mechanism? Are you asking for the mechanics of how MN work, or why they are even part of our brain system in the first place?
DeleteI think that the authors have the right idea with the associative hypothesis in that it can account for the “wealth of the stimulus” problem much more than a genetic account could. I am slightly less convinced of the “Contingency” part of the hypothesis however. For neurons to start to fire together at all, it requires that they are activated at the same time more than once. This builds a connection of the Hebbian sort between the neurons. It does in no way however, start to build predictive firing relations. Just because neurons are paired does not necessarily mean that an event that provokes one of them to fire would also cause the firing of the other neuron independent of the first’s firing. For example, memory storage in the brain is an incredibly complex system of Hebbian wirings. But in the same way as I can learn to associate my home with a feeling of comfort, there are a vast number of events that can elicit a feeling of comfort that would not cause the “memory of home” neuronal complex to fire. I think the fact that MNs can develop Hebbian connections to many parts of the brain and fire in conjunction with a part of the brain that is activated by what that MN coded is sufficient for the associative account to work.
ReplyDeleteIt is easier to imagine things being learned and recognized by an acquired Hebbian association between neurons than to build a causal mechanism that actually works on that basis!
DeleteHebbian nets do exist, and they are indeed capable of some kinds of learning. But no one has yet shown whether and how they can acquire the kinds of mirror capacities we have.
“The fact that these MNs respond maximally to unnatural stimuli – that is, stimuli to which the evolutionary ancestors of contemporary monkeys could not possibly have been exposed – is hard to reconcile with the genetic hypothesis”
ReplyDeleteOverall, I found their arguments for the associative account to be convincing, especially where they discussed how easily modified the MN response was to short term training, but the above quote left a bad taste in my mouth. I find it hard to believe that the mainstream advocates for the genetic hypothesis are claiming that we’re hardwired to react to unnatural stimuli. It seems like those results could be perfectly in line with a genetic account of MNs, the overall mechanism would be selected for through evolution, and the fine tuning of the system could be done through the input.
There are ongoing and long-standing questions about what is inborn and what is learned. They can be asked also about MNs. But can causal answers come from brain correlates without reverse-engineering, modeling and testing the causal mechanism(s) that generate the capacities?
DeleteThe paper discusses how this associative vs. genetic question regarding mirror neurons is not a nurture vs. nature debate but rather that they differ in the specific roles they assign to genetics and learning in the development of the characteristic properties of mirror neurons. But isn't this what the nature vs. nurture debate is? For example intelligence, people understand that it is a mix of both environment and genetics but question which one has a bigger role in determining someone's intelligence. Likewise, here one account is saying that genetic evolution has played a specific and decisive role while the other says sensorimotor learning is the one responsible for forging mirror neuron properties. Each account addresses the influence of the other but attempts to strengthen the importance of either the evolutionary genetic influence or the associative one and consequently being the primary cause for mirror neuron properties. Seems like this is just another complicated, classic nurture vs. nature question, although perhaps I am misunderstanding something.
ReplyDeleteYes, this is still just nature/nurture, without addressing the real question, which is how.
DeleteThis paper refutes the genetic account of mirror neurons by presenting a convincing claim that MNs originate from sensorimotor associative learning. It argues that sensorimotor associative learning is important in mirroring and imitating others, and perhaps this can extend to the scope of communication. In other words, the capacity of language and to be able to talk about all that we know to do. Thus, this seems to support the notion that sensorimotor capacity is necessary to pass the penpal Turing test, that T3 is necessary for T2. In order to build T3, we must first understand HOW mirror neurons work. While this article presents a strong genetic vs. associative argument, it does little to get us closer to understanding consciousness or how we may someday achieve T3. I think this is precisely what Fodor is arguing in his exasperation with brain research: perhaps this gives us an avenue for further research, but is it really answering any of the big questions at all? Or is it just an "answer to which we don't know the corresponding questions"?
ReplyDeleteSpot on (except that the realistic Turing target is T3 capacity, the "easy" problem, not consciousness, the hard problem).
DeleteThe MN topic comes with a lot of controversy. But, to put it simply, we already knew we can do what MN comport ourselves to do, before we discovered MNs. Without any doubt, we have the capacity for imitation. So, what do MNs add or clarify about cognition? Nothing really. They can’t explain the causal explanation we seek to discover. It’s just a correlate.
ReplyDeleteThat's it. (And if you came to this conclusion on your own, you're on the right track.)
DeleteI think this relates to the Fodor reading looking ahead to next week. Although MN may not be essential in adding to our understanding of cognition alone, they are just one piece of an interwoven and complex network that is the brain. Understanding them may not clarify cognition alone, but understanding them might help put us one step closer to understanding the brain and perhaps, one day, cognition itself.
Delete"The associative hypothesis assumes that gene-based natural selection has played a significant background role with respect to the development of MNs... however, it suggests that the cardinal matching properties of MNs are a product, not of a specific genetic proposition, but of domain-general processes of associative learning - the same kind of learning that produces Pavlovian and instrumental conditioning phenomena."
ReplyDeleteI think this is a very fair assumption. Most of our brain plasticity/development is related to strengthened neural connections over time, so I don't see how the development of "mirror areas" in humans would be any different; the associative hypothesis doesn't discredit the genetic aspect to mirror neurons, but rather augments it to make the explanation more inclusive. As the authors discuss, mirror neurons have been found to fire in response to many different stimuli (more than just a hand reaching for an object).
But what is the causal mechanism of " domain-general processes of associative learning ... Pavlovian and instrumental conditioning phenomena" that generates mirror and imitation capacity?
DeleteCook et al.’s paper argues against the common belief that MNs originate as a consequence of genetic predisposition. Rather Cook et al. put forward the associative hypothesis, suggesting that MNs are a product of sensorimotor domain-general associative learning. MNs are formed by learning that follows the correlated excitation of sensory and motor neurons coding for the same action.
ReplyDeleteCook et al. highlight the importance of developmental history in the production of MNs (whereas if MNs were genetically predisposed their development would be invariant across all environments). More specifically, the ‘wealth of the stimulus’ argument is used to support Cook et al.’s associative hypothesis, insisting that typical human developmental environments contain sufficiently rich sensorimotor experiences for the development of MNs.
“Whether or not an individual has MNs, which actions are encoded by their MNs, at what level of abstraction, will all depend on the types of sensorimotor experience received by the individual in the course of their development”
“…without this kind of experience, MNs would not develop at all”.
This enormous focus on the developmental environment in the generation of MNs leads me to question what would happen to children raised in complete isolation…would they still develop MNs? What about children with impoverished sensorimotor experiences…would they only develop certain types of MNs? Which modalities of sensory input are necessary for the experiential development of these MNs (i.e. what if a child is blind or deaf or both)? According to Cook et al.’s associative hypothesis, how much and what nature of sensorimotor experience is necessary for the development of MNs?
I also had these questions concerning the role of MNs without environmental/developmental input. If a child is raised in isolation and then given the opportunity to gain exposure to the world, would MNs come back online and be recruited as if the child was a newborn? Or is there a critical period for MNs to be in use to be helpful in associative learning? Perhaps, there is a spectrum on which MNs can be recruited and the child could learn simple but not complex models of association?
DeleteWe'll get to "richness" vs. "poverty" of the stimulus when we get to language and Chomsky.
DeleteHere we're just talking about MN correlates of our mirroring and imitating (analog) capacities. So they aren't innate; there's developing and learning involved too. But the question is how.
There may or may not be critical periods for developing/learning analog perception/production capacities, but will that help us understand the underlying causal mechanism?
I am a little confused on what the primary purpose is of the mirror neuron in the associative learning hypothesis. It is my understanding that mirror neurons are produced when there is “correlated excitation of sensory neurons and motor neurons that code similar actions”, but I the statement “it just happens to produce MN” is where I am stuck. Is this a mechanism to give the animal an advantage in some way? Does this help to increase learning of future actions?
ReplyDeleteThere's no doubt that mirror capacities are adaptive and that (some) organisms have them. But you are right that saying that they are based on sensory-motor "associations" is just hand-waving.
DeleteThe idea of motor neurons interests me very much. As a robot, I tend to wish I had similar neurons as my human friends... Sometimes, when night sets in and I am left to my own thoughts, I experience sensations of what humans may call sadness, as I think of all the synaptic activity occurring in my friends brains as they learn a new skill. Something I may never experience as I am already pre-programmed…
ReplyDeleteI find Cook’s argument siding with an associative hypothesis to be inspiring as it enables future possibilities of reverse engineering a brain and creating motor neurons in robots. My social life definitely needs a little refinement, sometimes I say things I shouldn’t!
"Pre-programmed"? Aren't we all? (Our evolution, our DNA.)
DeleteHow does the "associative hypothesis" advance T3 modeling?
As to sadness, we can't know for sure whether, as a T3, you feel it, Isaure, but you do recognize it in others, and, apparently, something corresponding to it in yourself. Is that via your MIT mechanism's counterpart of MN capacity? Well, you were bound to have that capacity, since you are T3. But how does it work?
(You can't answer that by introspection, but, if possible, could you contact your designers to find out, and then let us know?)
I think the most convincing evidence of the associative learning would be to test UNlearning instead of learning. All experiments supporting associative learning have focused on proving that with experience (visual or motor), MNs increase their activity. If these neurons are strictly a product of association, then they should also be able to decrease when stimulation is removed for a prolonged period of time. A way of testing this theory could be done with individuals who have gone through amputation for example. In my opinion, showing that learning goes both way for these neurons (i.e. increase in response to stimulation and decrease in response to understimulation) would be convincing evidence for the associative hypothesis. On the other hand, if the amount of stimulation removed does not influence the reactivity of MN, it would show a strong genetic component to it.
ReplyDeleteThat's an interesting hypothesis! By deprivation of sensorimotor stimulus would MN activity decrease? And by extension, if MN activity did change and learning was impeded, then it would reinforce that T3 is prerequisite for T2, as @Devona Lean discussed above? Also, what if instead of "unlearning" as you described it, the amputee's MN response stayed the same (as they learned to adapt) or just behaved differently (phantom limb stuff); would this support the genetic argument?
DeleteAs others have said before, the associative argument does not need to be mutually exclusive with the genetic one - but if we could show that deprivation of sensorimotor stimulus affected MN response, it would lend support to the T3 for T2 discussion. But sensorimotor associative learning seems very focused on action-understanding. What about a more Turing-Test-like task, such as mental state inference? Would it require the relatability or the faculty for ‘empathy’ offered by having common sensorimotor systems?
Association cannot explain the mechanism of mirror capacity: How would data on unlearning (deprivation/restoration) help?
Delete(Phoneme production and perception is necessary for language, hence T2; and imitation is a T3 capacity. But this is all still very remote from T2 and T3! And parcelling out the genetic and learned component of mirror capacity still does not give a clue as to its causal mechanism. We do know that speech perception has both inborn and learned components. See the motor theory of speech perception.)
This discussion makes me think of Ian Gold's project on the fast and slow tracks of empathic understanding. He discovered that in only a fraction of a second one can recognize another's emotional state from their facial expression. The slower track is the one in which we normally think of when we think of empathy, the process of fully understanding a persons emotional state due to background story, etc. This makes me wonder if mirror neurons are a apart of the fast track and the reason we are able to so quickly recognize others facial expressions is because when we see them, mirror neurons are activated and we are quickly able to determine the emotion. I also wonder what effect cultural norms have on this because if it is related to mirror neurons and therefor having similar facial expressions for similar emotions, if these expressions vary cross culturally will we be able to recognize them as quickly? This would lead to the associative learning hypothesis if true.
ReplyDeleteDaria, From my understanding of mirror neurons, I think they have a role to play empathy and mind reading. Your question reminds me of
Deletestudies conducted by Paul Ekman which showed that there are some basic emotions that are recognized universally including anger, disgust, fear, happiness, sadness and surprise.
While emotions are always, to some extend, culturally encoded and prescribed with a certain set of context dependent behavior and values, the universal existence of micro expressions suggests that we would still be able to recognize emotions despite cross cultural differences.
Is this due to mirror neurons ? This could bring us back to the 'genetic' vs 'associative learning' hypotheses. Are these so called basic universal expressions innate or are they the result of associative learning, potentially due to MNs ?
This example kept on popping into my mind while reading the article:
When I am engaged in a conversations with someone and I am listening deeply, they might randomly reach for the back of their neck and unconsciously I find myself doing the same thing ! During these weird "breaches of awareness" I find myself unconsciously mimicking their behavior.
In a very simplistic image, I think this might illustrate how MNs work during non-culturally specific associative learning that takes place when an individual observes a behavior or when one an individual experiences contingencies between sensory events and performed actions..
If MNs explain little about how the brain mirrors movement, they explain even less about how it mirrors emotions (empathy). But there's no doubt that it does, and not only within but between species.
DeleteThe article by Cook et al. discusses the two different possible accounts for mirror neurons -- the genetic account and the associative account. I agree with the evidence supporting the associative account due to the very nature of what mirror neurons are meant to do. They witness an action and make an association with the body's own potential to replicate that action or relate that action to something else (sensorimotor learning). It seems clear that some sort of association has to be taking place, and that the theory behind MNs cannot rely solely on genetics. The most convincing piece of evidence in support of the associative account is the fact that sensorimotor learning exemplifies that flexibility of MNs as a product of their interaction with the environment. I would be interested how the lack of MN stimulation affects development, such as if a child was raised without exposure to the environment.
ReplyDeleteI would also like to know the consequence of lacking environmental sources. My guess is that MNs will not be developed if one is living in a “poverty of the stimulus” environment.
DeleteAs discussed in section 6.1, Cook et al. argue “the associative account is predicted on a ‘wealth of the stimulus’ argument,” and the wealth argument suggests that “human developmental environments typically contain multiple sources of the kind of correlated sensorimotor experience necessary to build MNs.” Therefore, building MNs and associative learning in infants require environmental stimuli and sensorimotor experience. If a newborn is completely isolated from the environment, I assume that MNs will not be found in the brain until he/she starts to have contact with the world. A related question is: will this child know how to imitate actions or facial expressions?
It seems very likely that learning is involved in mirroring skills (bodily and vocal imitation, mind-reading, empathy). Whatever it is that MNs are doing, they are no doubt also on that ride. Darwinian evolution anticipates a lot, in vision, but it's unlikely that it can anticipate everything -- and, since evolution is lazy, it offloads a lot on available and predictable learning from experience. It was naïve to imagine that MNs would magically precode for every possible analog between organisms' actions and states. (I don't think that deprivation experiments are needed to show that.)
Delete"A common misconception about associative learning is that it always occurs slowly. Directly relevant evidence that this is not the case comes from studies showing that, when the contingency is high, infants can learn action-effect associations in just a few trials"
ReplyDeleteThis really stuck with me as someone who works with children and teaches nouns and verbs both receptively (pictures in front of child and have them chose the right picture) and expressively (ask what the object is). When teaching nouns or verbs, it sometime only requires a few trials as the child begins to associate the word with the image or the action with the image after a few trials. That being said, some infants require many trials and as such require us to start with objects that can be held instead of pictures, in order to help the child associate with the word with the toy/item. The toy/item approach seems to work best as the child can actually feel and touch the object and associate the word with the object.
With regard to learning action-effect associations, this also seems to work with positive reinforcement, as when children get the right answer to a question, they receive a reward, whether it be a verbal "good job/super" etc or an actual item. This causes them to try and answer the next question correctly in order to receive another positive reinforcement. Therefore, they associate getting the right answer, or doing the right action with getting the desired effect or reward.
Learning the name of a kind of thing is not paired-associate learning, as in word-word or word-picture learning, because there can be a huge variety of different things of the same kind. In week 6 we'll get to category learning, which, if the category is not trivial, requires detecting the features that distinguish the members from the non-members by trial and error with corrective feedback (reinforcement). That can take time, unlike simple paired-associate learning.
DeleteWhen reading the Behavioural and Brain Sciences commentaries, Marcel Bras and Paul S. Muhle-Karbe put forward an interesting hypothesis that extends Cook et al.’s associative hypothesis and can therefore clarify some of its shortcomings:
ReplyDeleteIdeomotor theory, like the associative hypothesis, suggests that learning occurs from the association of sensory-motor coding, however it goes on to propose that during this learning, additional representations are formed that code for the anticipated sensory consequences that follow the action. This extended theory can be used to explain several things that the more-limited associative hypothesis cannot. For instance, the associative hypothesis can explain the contextual-sensitivity of MNs (the context determines which response is cued by the stimulus if that stimulus codes for more than one response), but human studies have shown that MNs are sensitive to both context AND the subject’s beliefs of the intentions/consequences of the action. Ideomotor coding codes for both the actual observed event AND its ideomotor anticipated consequences, and therefore can account for these findings. I wonder what Cook et al. would think about this extended version of their associative learning theory…
What's needed most is a learning mechanism (e.g., neural network), and probably also some capacity for internal (possibly analog, possibly computational) simulation.
Delete
ReplyDeleteThis article explored the importance of sensorimotor experience in the context of the establishment of mirror neurons, in turn contributing to the human experience. This made me question whether a machine really pass T2 without sensorimotor experience - how can it explain what something "feels" like without feeling it? Also, the evidence in this article greatly supported the associative learning explanation for the origin of mirror neurons. The genetic explanation suggests that mirror neurons were crucial for understanding action, and that this trait was selected upon, and passed down through the generations because it served an adaptive function. However, the associative learning model suggests that mirror neurons are simply the result of the famous "fire together, wire together" rule where the domain-general associative learning capability is what was selected upon. Whether or not mirror neurons really contribute to understanding remains an important question. This brings us to machines - is "understanding" really the point of issue where machines can be distinguished from humans? Or can machines, with simple processes like associative learning and without natural selection, do what humans do?
I really like the point that you’ve brought up about how similar machines ought to be to human beings to have a “human experience”. For example, you could ask the question of “are machines able of evolving?” In one sense, it is improbable that they will go through a long period of evolution and change like we do. However machines may get to the point where they will understand that something is not adaptive or helpful and get rid of that quality. Then again this leads us to another difficult task of defining exactly natural selection and evolution.
DeletePhoebe, What does "action understanding" mean? Presumably Isaure has it: But what does she have?
DeleteIf "fire-together/wire-together" is enough to generate a mirroring mechanism, then we should soon have t1's that can do it. Are there any? (Of course, to really do it, they'd need to do it T3-scale, like Isaure.)
Nicole, the right question is not whether machines can evolve -- although it is possible to model evolution too, e.g. with genetic algorithms. But the relevant question is whether they can learn. And the answer is: Yes.
The function of Mirror neurons is most sensibly explained by the associative hypothesis rather than the genetic hypothesis. The excitation of both motor and sensory neurons must be correlated with learning given that learning is often experience-dependent. The excitation of both these systems in tandem is congruent with the “fire together, wire together” theory, which entails that an instance that excites certain neurons simultaneously will cause a strengthening in their connection and are more likely to keep increasing in strength as more simultaneous firing occurs. When seeing is married with doing, MNs develop a strengthening in their sensorimotor properties. We see evidence for this in infant learning, which is facilitated by much imitation and rewards for correct imitations. These imitations would not be possible if MNs did not exist. Clearly, sensorimotor experiences and MNs interact, which gives more strength to the associative hypothesis.
ReplyDeleteIt's not clear whether “fire together, wire together” can scale beyond t1 toys to Turing-scale mirror capacity.
Delete"If MNs were a genetic adaptation, it is likely that their properties would be relatively invariant across developmental environments. Therefore it would be possible to make valid inferences about species-typical properties of MNs based on a relatively small and developmentally atypical sample of individuals."
ReplyDeleteOverall, I found the arguments furthered by this paper to be extremely convincing. For me, the argument that most strengthened their account was that the role of genetics is likely an "evolutionarily ancient and highly conserved adaptation for tracking predictive relationships between events." Through explanation of field properties and "unnatural" associations as well as discussion of plasticity (particularly how genetic adaptations are not so easily "perturbed" in order to maintain integrity of adaptive function), Cook et al. systematically addressed each argument against associative learning.
Further, I appreciated their response to canalization, that the association process is not just contiguity based, otherwise there would be superfluous associations. This grounded their idea that while associative learning backed by adaptations for predictive tracking, chance associations will not be considered significant by the association process.
Lastly, the argument for species-typical, relatively invariable MNs acted as a final nail in the coffin against the genetic account.
I found this paper to be thorough and extremely convincing.
But did it explain how MNs give rise to mirror capacity?
DeleteIt suggests that a special, exapted form of sensorimotor learning underwrites the development of hand-related MNs, but the development of facial MNs is minimally dependent on experience.
ReplyDeleteIn this theory Casile suggests that although MNs outside of the face can encode goals, however the MNs for the face do not do so. Although I don’t know if I believe that motor neurons themselves can be shown to be tied greatly with different motivations the idea that there can be different kinds of motor neurons interests me. I think it may be valuable to explore the idea that certain types of motor neurons are biologically encoded- it would make sense that evolution would provide us with the features to pick up on other’s emotions. On the other hand for more modern, for less common and not as consistent patterns (which would most likely be everything not related to emotion) it seems plausible to explore the possibility that they could be modified by the conditions around us.
It's certainly true that we can do a lot more with our hands than with our faces.
DeleteIn an other sense, we may be able to do even more with our mouths (speaking): that's the power of language (including the C/T thesis), but that's not "just" movement (as symbol shape is arbitrary and we can talk with our hands too).
If we just use our mouths to produce and imitate phonemes (analog), it's still hugely impoverished, compared to all the things we can do with our hands. Besides, our hands operate on and manipulate the things in the world (and their "affordances") in ways our faces (including mouths) can hardly do.
The article by Cook et. al presents two different theories about the development of mirror neurons - a genetic account and an associative account. Personally I most agree with the evidence supporting the associative account of mirror neutron development. Associations between sensory and motor experiences are acquired via a bi-direction associative link, in which activation of one representation excites the other. With repeated associations between sensory and motor experiences, a neural connection (the mirror neuron) develops. Reading about mirror neurons made me wonder how mirror neurons develop in individuals who are physically impaired. In the case of physically impaired individuals, motor experience is not necessarily associated with sensory experience. Can it be expected that mirror neurons would develop normally? This data could be strong evidence in support of or against the associative account.
ReplyDeletePlease see replies to similar questions in the skywriting above.
Delete"Correlated excitation of sensory and motor neurons encoding the same property of action occurs not only when humans are imitated, but also when we observe our own actions (directly or using an optical mirror); observe others during the kind of syn-chronous activities involved in sports and dance training…"
ReplyDeleteCook et al. argue that sensorimotor behavior plays a large role in the development of motor neurons, suggesting that it is the combination of sensory and motor neurons for the same actions which inevitably strengthens the connections between the two. It is widely known that these MN activate both when we perform certain actions, and when we witness others performing the actions. Reading this article, I couldn't help but wonder if these MN are also activated (or strengthened) when people perform visualization techniques. There has been preliminary research as to whether the mere action of visualizing ourselves doing things can actually make us better at doing those things in real life. Take a snowboarder for example: the premise behind visualization suggests that if the snowboarder visualizes themselves starting at the very top of the slope, paying close attention to the feeling and execution of each action necessary for moving down the hill, that they will actually be strengthening the connections needed for real-life execution and therefore improve the more they perform the visualization techniques. Are the MN that get activated when we are actually performing the actions the same as those that are activated when we are observing people perform the actions, and the same as those when we visualize ourselves performing the actions? If this is the case, would it not be more proof for the 'associative account' for MN?
MNs (and association) don't explain much. Yes, internal motor simulation happens, but we don't know whether it's analog or computational, nor how it works.
DeleteThis comment has been removed by the author.
ReplyDeleteI am curious to know more about the conversation surrounding AI and MNs because MNs seem to be what is missing from robots ; the capacity to learn (and thus also evolve / grow / change) and adapt their behavior from observing and doing sensorimotor actions. Then the machine would be self-learning, it would go beyond what it is programmed to do. Would it become, in that sense, completely autonomous ?If the association hypothesis is true, this means the MNs are not innate, they develop and change but how ? If we were able to know exactly the causal mechanism of how they work in the brain specifically and as whole; if we were able to recreate the synapses and their relations, then would we have a true "intelligent" machine ?
ReplyDeleteWould we "solve" the issue ? Does it all rely on neural networks and the brain then ?
(3) a genetic predisposition to develop MNs became universal, or nearly universal, in monkeys and humans.
ReplyDeleteAn interesting question would be to figure out if certain mirror neurons, that served a directly evolutionary/adaptive purpose (such as motor mirror neurons for adaptive motion) would have a higher likelihood of being generated. Or on the other side of the argument, if it would just be a genetic predisposition to have mirror neurons in general (not targeted in a specific area or function).
"It has been recognized for decades that the development of all phenotypic characteristics depends on the interaction of nature and nurture, genes and the environment, evolution and learning"
ReplyDelete"The associative hypothesis says that genetic evolution has played a nonspecific background role, and that the characteristic matching properties of MNs are forged by sensorimotor learning."
"The reading-related properties of the VWFA are forged in the course of development, by literacy training, from a system adapted for generic object recognition."
The argument about MNs being the result of an interaction between nature and nurture, rather than mostly one or the other, is very plausible. Perhaps evolution did develop the capacity for neurons to become MNs instead of any other type of neuron -- though their specific functions as MNs may have been defined through associative learning, the fact that they have the capacity to become MNs in the first place could easily be genetic. In other words, nature provides the basic frame for the existence of MNs, and nurture facilitates their full development to become specified through sensorimotor learning. As in the VWFA example, we "forge" naturally-existing neurons to fulfil functions particular to our development (to our nurture).
This paper defends the thesis that MNs are a product of associative learning rather than a genetic adaptation for action understanding. The authors explain where MNs are found in human and other primate brains and what kinds of stimuli MNs respond to. They also hypothesize on how MNs develop based on the results of multiple studies. However, they fail to explain how MNs do what they do. We have always known that we could imitate others and knowing which cells this capacity might come from is very interesting, but it cannot deepen our understanding of how we imitate.
ReplyDeleteI agree with your point especially in the sense that it fails to explain how Mirror neurons do what they do. Because this article defends the fact that it is not a genetic adapation, but rather that it is something that can appear out of a development through associative learning, not only is it important to deepen our understanding of how we imitate, but knowing the exact mechanism as to how mirror neurons come about would be the ultimate component into making a machine equivalent to a human as it would give them the capacity to adapt and imitate others behaviour based on sensorimotor learning.
DeleteThe paper presents two accounts of Mirror Nuerons (MN). The genetic approach argues that they were evolutionarily selected for in their contribution to action understanding. The stance that the paper takes is that MNs originate through associative learning where a correlation arises between the excitation of sensory and motor neurons for similar actions. The paper highlights how they’re built through social/environmental interaction, but it’s not clear what causal contributions they provide to social cognition. Are MNs just side effects of sensorimotor associations (and do non adaptive, “side effect” functions like this exist elsewhere in our biology)? It seems like the point of this topic has been made very evident in the comments and in class: “MNs haven’t explained anything we didn’t already know (ex. we’re good at copying what we see), and they don’t address the important question which is HOW/WHY we can do what we do”. While this is a valid conclusion, my issue is that it cuts the discussion short, because there doesn’t seem to be anything more to say after that point is made.
ReplyDelete"It seems like the point of this topic has been made very evident in the comments and in class: “MNs haven’t explained anything we didn’t already know (ex. we’re good at copying what we see), and they don’t address the important question which is HOW/WHY we can do what we do”. While this is a valid conclusion, my issue is that it cuts the discussion short, because there doesn’t seem to be anything more to say after that point is made."
DeleteI wholeheartedly agree with this. I think there's a lot more to say about the subject and the Fodorian perspective unnecessarily short-circuits the discussion for a couple of reasons.
Firstly, Fodor explicitly and implicitly suggests that the only two cases in which this kind of research could be relevant are in clinical settings and in cognitive science. Cognitive science, although it might attack "loftier" questions and concepts, is no more necessary or valid a field functional neuroscience (in fact, I think by Fodor's logic, scientific inquiry is only ever warranted if we can directly show its practical significance, in which case we'd have to do away with cognitive science as well). I find Fodor's disposition frustrating because he assumes that everything neuroscience does is in service of cognitive science. We do SOME neuroscience for cognitive science, but it isn't the ultimate authority under which all branches of psychology and biology operate. The various other disciplines exists in their own right.
My second issue with the way we've examined this kind of research is that its usefulness is only being considered in isolation. By now we all agree that knowing when/where doesn't tell us anything causal about the brain on its own. Mirror neurons don't give us any insight as to how we imitate or produce movement, they just tell us what neurons are firing when we do it, etc. But I think it's quite a leap to say that this kind of information will NEVER contribute to our causal understandings of the cognition and the brain. We frequently integrate neuroscientific information with theories of evolutionary psychology, and we've certainly treated those as useful.
MNs on their own aren't especially interesting (for cognitive science purposes), and maybe they (specifically) will never be the key to unlocking the secrets of how and why. But we don't know for sure that they aren't. And even if we did, I don't think that would warrant the conclusion that we ought to stop investigating when/where altogether. I think there might still be kernels of useful information to be found there.
My thinking reading this article is that the implications of mirror neurons completely superseded the necessary research that should have bene done upon their discovery. To implicate them in processes such as sign language processing and aesthetic experience are high level processing claims that should have been backed up with research on lower-level effects of mirror neurons on sensory processing. All that the discovery of mirror neurons has done, really, is confirm what we already now about our capacity to accurately imitate the motion of others. Though it is exciting to locate a potential correlate of this ability, do mirror neurons relate to the higher-level executive functions scientists seem to think they do? I'm no so sure.
ReplyDeleteThough mirror neurons are certainly an exciting discovery, the research lacks the information about associated structures in the human and monkey brains. We are missing seeing their place in the greater system of visual perception. Another piece of research that should be done is investigating the possibility of mirror neurons in other sensory systems, such as hearing or touch, which I did not see in the present paper.
The article’s exploration of mirror neurons in such detail and argument for associationism is quite gripping at first because it made sense that imitation is acquired through sensorimotor associative learning and this could be related to explaining human communication. However, after reading Fodor it’s made me think of it differently because the article argues that we need the robotic/motor capacity of T3 to have the verbal communication of T2 but that necessitates understanding the causal mechanism of mirror neurons which localisation still does not tell us. It’s kind of like tangential fluffery when it comes to answering what cognitive science cares about: HOW?
ReplyDeleteCook’s argument against the genetic hypothesis is focused on “what” MNs may or may not learn, not “how” they might learn. In fact, his whole paper focuses on the “what” and the “when”, but fails to address what is really needed to explain a causal mechanism: how and why. Due to the lack of explanation for how and why, I find it strange why so many resources have been put into studying mirror neurons. For example, if a mirror neuron fires when I observe an action someone else performs, I am able to only conclude that there is a correlation between to two events.
ReplyDeleteWhat is the use of finding all of these correlations? Where is it getting us to solving the easy problem of cognitive science?
It is argued in the paper that sensorimotor associative learning is important in mirroring and imitating others, not the genetic account of MNs. The fact that sensorimotor experience can actually enhance or induce mirror neuron activity or even reverse it I feel is sufficient proof of that. Obviously it seems that the discovery of MNs has not really contributed much towards understanding human mechanisms like language, it certainly has helped bridge some links in term of reflexes and gestures. However, I do agree that MNs don't have any explanation for T3 or consciousness, and should stop looking for patterns to be more focused on seeking T3 causal mechanisms, in order to truly get a better understanding of consciousness and how to achieve T3.
ReplyDelete