Whenever the discussion turns to neuroscience, I tell people if they want to read the best thing written about the brain in this century, they need to read After Phrenology (Anderson Reference Anderson2014). I say this up front to put any differences between Anderson and me in perspective. I view these differences as relatively minor, mostly a matter of emphasis. Herein I will focus on this two-part question: If Anderson is right about: (a) neural reuse, redeployment, and multiuse at multiple spatial and temporal scales in the brain and (b) the implications for extended and embodied cognition and ecological psychology (I believe he is right on both counts), then (1) what ought the future of cognitive neuroscience look like, and (2) what does the future hold for the autonomy of folk psychology? He says he is advocating explicitly for giving the brain its scientific voice and hence the likelihood of a significant revision to the vocabulary of cognition (p. xxii). I will argue that if Anderson is right about the nature of the brain, then he needs to move even further away from a focus on mapping structure to function and even closer to a more radically extended and embodied account of mind and behavior. Second, I will argue that while the vocabulary of scientific psychology – that is, RTM and CTM – must change, Anderson's vision of the brain and embodied cognition is a good argument for the autonomy of folk psychology and hence a good argument against reduction.
I agree completely with Anderson when he says “individual pieces of the brain, from cells to regions to networks, are used and reused in a variety of circumstances, as determined by social, environmental, neurochemical, and genetic contexts” and that “what is reuse at one level of organization can be multiuse at another” (p. 36). He says the brain achieves its functions by assembling the right functional coalitions neural and extra-neural, including body, environment, and external artifacts such as symbolic ones, all in the service of adaptive behavior (pp. 242, 302). Hence, intelligence is not so much about local processing as it is about “cooperative connectivity” (240). This implies not just the failure of massive modularity, but also a complex many-many relationship between structure and function at multiple scales. In Silberstein and Chemero (Reference Silberstein and Chemero2013), which focuses on networks and graph theory in systems neuroscience, we note that very different neurochemical mechanisms and wiring diagrams can instantiate the same networks and hence perform the same cognitive functions. Indeed, in these models, it is primarily the topological features of various types of small-world networks that explain essential organizational features of brains, as opposed to lower-level, local purely structural features. As Sporns puts it, “a reentrant system operates less as a hierarchy and more as a heterarchy, where super- and subordinate levels are indistinct, most interactions are circular, and control is decentralized” (Reference Sporns2011, p. 193). Hence, topological features such as the properties of small-world networks exhibit a kind of universality with respect to lower-level structural details. Structural and topological processes occur at radically different and hard to relate timescales. The behavior and distribution of various nodes such as local networks are determined by their nonlocal or global connections. That is, such global organizational principles or features of brains are not explicable in principle via localization and decomposition.
Anderson therefore is absolutely right that this picture of the brain tells against componentiality and mechanistic explanation as standardly conceived. As he says, “Global function is not built from componential local function, but rather the reverse” (p. 93) and “local neural assemblies are polymorphic and ‘multifunctional’” (p. 104). He says then that we must use current and invent new increasingly sophisticated tools for mapping structure to function, such as measures of effective connectivity between various interacting parts of the brain. He gives the example of Granger causality for exploring effective connectivity between large-scale systems of the brain. He describes the bottom-up approach that creates mappings between network motifs (common configurations of functional relations between neural elements) and causal/functional effects. To do this properly, he says, we must first learn to identify small-scale network configurations in living animals (p. 307).
I certainly agree with all this as far as it goes, but my worry is that it is still too brain-centric, still too focused on the “where strategy,” to use Anderson's language. My worry is based on the following: (1) brains and other complex biological systems exhibit multiple realizability at all scales and levels of description defined both structurally and functionally, and (2) as Anderson stresses, brains are embedded in larger physical and social networks that play a huge role in determining their activity and interactions at all scales. He argues for example that “actual social interaction appears necessary for language acquisition” (p. 258). Given 1 and 2, the first main point I want to make is that going beyond the brain, we need a neuroscience that is inherently multi- and inter-disciplinary, one that focuses on various complex relationships at multiple scales and levels of organization such as between gene networks, RNA networks, epigenetics, various levels of scale in the brain from neurons to networks, behavior, cultural and social features, phenomenology, and so forth. These relations need to be studied both synchronically and diachronically and at various timescales including the developmental and evolutionary.
The point is that a neuroscience-in-full is inescapably interlevel and deeply historical – a point Anderson makes as well, to be sure. However, I think the point deserves greater emphasis because the focus on mapping structure to function, however sophisticated, comes from the perspective of intervention and manipulation versus what I will call “seeing the big picture.” Of course, there is nothing wrong with intervention and manipulation, especially in the biological sciences, but in the long run, with respect to both control and the big picture, the full potential of systems neuroscience will be realized only if we move further away from such mechanistic thinking. Of course, this violates some conceptions of what biology is and potentially starts to make biology look a lot more like mathematical physics in general or the study of many-bodied systems such as condensed matter theory. However, this does not imply a reduction of biology to physics, just an increasing use of the physics tool kit, which – as illustrated by graph theoretical neuroscience – is of course already happening.
Anderson characterizes the brain as a primarily action oriented control system as opposed to computationally driven. The brain is an action controller responsible for managing the values of salient organism-environment relationships, that is, affordances (p. 307). So the brain controls or manages biased affordance competition. Of course, the brain for Anderson is decidedly not a CPU, but why does he say the brain is the controller, the brain assembles the right functional coalition, and so forth, and not the extended organism itself. The worry is that given his Gibsonian picture, this way of talking about the brain makes it sound too autonomous from the body, the environment, and the affordances of the organism, too CPU like. He talks about the “multidimensional functional biases of individual brain regions as indices of some set of underlying causal dispositions” (p. 307). He talks about this in terms of neural dispositions, differential propensities, and causal powers (pp. 114, 308), and in terms of “personalities of brain regions.” He calls this the dispositional vector account of brain activity. My worry is that this sort of language suggests not merely that different networks or regions have a tendency to contribute to certain tasks, but that there is something intrinsic and mechanism-like in virtue of which they so contribute. I am not angling for full-blooded holism, and I do not doubt that there are important differences, but I am suggesting these differences, rather than being intrinsic, are primarily fixed by evolutionary, developmental, functional constraints, and historical contingencies. Furthermore, there may be no optimal brain architecture, and even if we discover such TALoNS (p. 94), we may be learning about said constraints and historical contingencies only with respect to humans or a subset thereof: There is probably multiple realizability with respect to TALoNS as well. And, as Anderson acknowledges, these constraints often get trumped by cultural and social constraints that also get preserved and conserved across time. So I want to push back a little against the picture of brain regions evolving particular dispositions that get combined in various ways like primary colors to collaborate on different tasks.
Rather than speak of brains as control systems with intrinsic dispositions, I want to say that brains are hubs in the graph theoretic sense, but they are living hubs in multilevel dynamical processes that are extended in space and time. The networks in question are physio-bio-cultural, and brains are truly historical artifacts that bear the marks of their origins and development. In this model, extended organisms or agents spanning brain, body, and environment are the primary locus of control, and the structural elements that instantiate such networks are often secondary to the networks themselves (Silberstein & Chemero Reference Silberstein, Chemero, Manetti and Caiani2011).
This brings me directly to my second main point about the autonomy of folk psychology. Rather than reduction or elimination, neural reuse and extended cognition so conceived actually support functionalism in the sense that we individuate processes with regard to their role and effects as opposed to their structural constituents. Again, looked at in this way, we find a great deal of multiple realizability within and across human and non-human brains. Anderson places a lot of emphasis on the importance of selection in understanding brain function (p. 296), and I agree, but I think there are others aspects of evolutionary and developmental biology that are equally important and that do not fully reduce to selection. Evolutionary theory has its own version of functionalism in the form of convergent or parallel evolution such as mimicry and flying (McGhee Reference McGhee2011). Focusing on cognitive convergence, creatures with very different brains and selective histories seemingly converge on similar behavioral and cognitive strategies for dealing with “socioecological” problems. For example, corvids do not even possess a prefrontal cortex, yet they exhibit very intelligent behavioral and cognitive strategies similar to primates. There is growing evidence across the board that creatures with very different brains have in many respects converged on relatively similar minds (McGhee Reference McGhee2011). Well-known examples of this from insects to mammals include: tool use, architectural behavior, agricultural behavior, social or collective behavior, mathematical behavior, and language use (McGhee Reference McGhee2011, Ch. 6). There is also a growing consensus that a wide variety of different species with very different brain structures and nervous systems possess some form of not only sophisticated convergent cognition, but also consciousness, self-awareness, and metacognition. Examples abound, such as the mourning behavior of gorillas and dolphins (McGhee Reference McGhee2011, p. 240). I think the best explanation for cognitive convergence is that the affordances, environmental and social features often trump structural neural constraints whether imposed by physics or selection.
So, although I agree with Anderson that the brain must have its voice, what I think it is telling us is that neural reuse is best viewed as a subset of what developmental and evolutionary biologists call plasticity and robustness (Bateson & Gluckman Reference Bateson and Gluckman2011). Although there are many different kinds of both robustness and plasticity, in general robustness refers to relative stability or invariance across environmental, genetic, or cellular transformations, and plasticity refers to cases wherein features of the organism are held constant such as genotype, and yet because of environmental transformations the organism nonetheless manifests very different or unique adaptive traits or behaviors (Bateson & Gluckman Reference Bateson and Gluckman2011, p. 8). Different forms of plasticity include: phenotypic, molecular, variety of neural types, immunological, and behavioral (Bateson & Gluckman Reference Bateson and Gluckman2011, p. 46). Both of these features of complex biological systems are of course at the heart of the epigenetic revolution in biology. As many people have pointed out, robustness and plasticity are two sides of the same coin: “Plasticity is often regulated by robust mechanisms and robustness is often generated by plastic mechanisms” (Bateson & Gluckman Reference Bateson and Gluckman2011, p. 46). It is important to note that, although many biological mechanisms possess the properties of robustness and plasticity, these features cannot in principle be explained mechanistically in terms of localization and decomposition. These are global/systemic and scale-invariant features of such biological systems. Convergent evolution, robustness, and plasticity all go hand-in-hand, and they all point to the strongly extended nature of phenotype, behavior, and cognition. For one take on what such a Gibsonian cognitive neuroscience might look like, see Silberstein and Chemero (Reference Silberstein, Chemero, Orden and Stephen2012).
Whenever the discussion turns to neuroscience, I tell people if they want to read the best thing written about the brain in this century, they need to read After Phrenology (Anderson Reference Anderson2014). I say this up front to put any differences between Anderson and me in perspective. I view these differences as relatively minor, mostly a matter of emphasis. Herein I will focus on this two-part question: If Anderson is right about: (a) neural reuse, redeployment, and multiuse at multiple spatial and temporal scales in the brain and (b) the implications for extended and embodied cognition and ecological psychology (I believe he is right on both counts), then (1) what ought the future of cognitive neuroscience look like, and (2) what does the future hold for the autonomy of folk psychology? He says he is advocating explicitly for giving the brain its scientific voice and hence the likelihood of a significant revision to the vocabulary of cognition (p. xxii). I will argue that if Anderson is right about the nature of the brain, then he needs to move even further away from a focus on mapping structure to function and even closer to a more radically extended and embodied account of mind and behavior. Second, I will argue that while the vocabulary of scientific psychology – that is, RTM and CTM – must change, Anderson's vision of the brain and embodied cognition is a good argument for the autonomy of folk psychology and hence a good argument against reduction.
I agree completely with Anderson when he says “individual pieces of the brain, from cells to regions to networks, are used and reused in a variety of circumstances, as determined by social, environmental, neurochemical, and genetic contexts” and that “what is reuse at one level of organization can be multiuse at another” (p. 36). He says the brain achieves its functions by assembling the right functional coalitions neural and extra-neural, including body, environment, and external artifacts such as symbolic ones, all in the service of adaptive behavior (pp. 242, 302). Hence, intelligence is not so much about local processing as it is about “cooperative connectivity” (240). This implies not just the failure of massive modularity, but also a complex many-many relationship between structure and function at multiple scales. In Silberstein and Chemero (Reference Silberstein and Chemero2013), which focuses on networks and graph theory in systems neuroscience, we note that very different neurochemical mechanisms and wiring diagrams can instantiate the same networks and hence perform the same cognitive functions. Indeed, in these models, it is primarily the topological features of various types of small-world networks that explain essential organizational features of brains, as opposed to lower-level, local purely structural features. As Sporns puts it, “a reentrant system operates less as a hierarchy and more as a heterarchy, where super- and subordinate levels are indistinct, most interactions are circular, and control is decentralized” (Reference Sporns2011, p. 193). Hence, topological features such as the properties of small-world networks exhibit a kind of universality with respect to lower-level structural details. Structural and topological processes occur at radically different and hard to relate timescales. The behavior and distribution of various nodes such as local networks are determined by their nonlocal or global connections. That is, such global organizational principles or features of brains are not explicable in principle via localization and decomposition.
Anderson therefore is absolutely right that this picture of the brain tells against componentiality and mechanistic explanation as standardly conceived. As he says, “Global function is not built from componential local function, but rather the reverse” (p. 93) and “local neural assemblies are polymorphic and ‘multifunctional’” (p. 104). He says then that we must use current and invent new increasingly sophisticated tools for mapping structure to function, such as measures of effective connectivity between various interacting parts of the brain. He gives the example of Granger causality for exploring effective connectivity between large-scale systems of the brain. He describes the bottom-up approach that creates mappings between network motifs (common configurations of functional relations between neural elements) and causal/functional effects. To do this properly, he says, we must first learn to identify small-scale network configurations in living animals (p. 307).
I certainly agree with all this as far as it goes, but my worry is that it is still too brain-centric, still too focused on the “where strategy,” to use Anderson's language. My worry is based on the following: (1) brains and other complex biological systems exhibit multiple realizability at all scales and levels of description defined both structurally and functionally, and (2) as Anderson stresses, brains are embedded in larger physical and social networks that play a huge role in determining their activity and interactions at all scales. He argues for example that “actual social interaction appears necessary for language acquisition” (p. 258). Given 1 and 2, the first main point I want to make is that going beyond the brain, we need a neuroscience that is inherently multi- and inter-disciplinary, one that focuses on various complex relationships at multiple scales and levels of organization such as between gene networks, RNA networks, epigenetics, various levels of scale in the brain from neurons to networks, behavior, cultural and social features, phenomenology, and so forth. These relations need to be studied both synchronically and diachronically and at various timescales including the developmental and evolutionary.
The point is that a neuroscience-in-full is inescapably interlevel and deeply historical – a point Anderson makes as well, to be sure. However, I think the point deserves greater emphasis because the focus on mapping structure to function, however sophisticated, comes from the perspective of intervention and manipulation versus what I will call “seeing the big picture.” Of course, there is nothing wrong with intervention and manipulation, especially in the biological sciences, but in the long run, with respect to both control and the big picture, the full potential of systems neuroscience will be realized only if we move further away from such mechanistic thinking. Of course, this violates some conceptions of what biology is and potentially starts to make biology look a lot more like mathematical physics in general or the study of many-bodied systems such as condensed matter theory. However, this does not imply a reduction of biology to physics, just an increasing use of the physics tool kit, which – as illustrated by graph theoretical neuroscience – is of course already happening.
Anderson characterizes the brain as a primarily action oriented control system as opposed to computationally driven. The brain is an action controller responsible for managing the values of salient organism-environment relationships, that is, affordances (p. 307). So the brain controls or manages biased affordance competition. Of course, the brain for Anderson is decidedly not a CPU, but why does he say the brain is the controller, the brain assembles the right functional coalition, and so forth, and not the extended organism itself. The worry is that given his Gibsonian picture, this way of talking about the brain makes it sound too autonomous from the body, the environment, and the affordances of the organism, too CPU like. He talks about the “multidimensional functional biases of individual brain regions as indices of some set of underlying causal dispositions” (p. 307). He talks about this in terms of neural dispositions, differential propensities, and causal powers (pp. 114, 308), and in terms of “personalities of brain regions.” He calls this the dispositional vector account of brain activity. My worry is that this sort of language suggests not merely that different networks or regions have a tendency to contribute to certain tasks, but that there is something intrinsic and mechanism-like in virtue of which they so contribute. I am not angling for full-blooded holism, and I do not doubt that there are important differences, but I am suggesting these differences, rather than being intrinsic, are primarily fixed by evolutionary, developmental, functional constraints, and historical contingencies. Furthermore, there may be no optimal brain architecture, and even if we discover such TALoNS (p. 94), we may be learning about said constraints and historical contingencies only with respect to humans or a subset thereof: There is probably multiple realizability with respect to TALoNS as well. And, as Anderson acknowledges, these constraints often get trumped by cultural and social constraints that also get preserved and conserved across time. So I want to push back a little against the picture of brain regions evolving particular dispositions that get combined in various ways like primary colors to collaborate on different tasks.
Rather than speak of brains as control systems with intrinsic dispositions, I want to say that brains are hubs in the graph theoretic sense, but they are living hubs in multilevel dynamical processes that are extended in space and time. The networks in question are physio-bio-cultural, and brains are truly historical artifacts that bear the marks of their origins and development. In this model, extended organisms or agents spanning brain, body, and environment are the primary locus of control, and the structural elements that instantiate such networks are often secondary to the networks themselves (Silberstein & Chemero Reference Silberstein, Chemero, Manetti and Caiani2011).
This brings me directly to my second main point about the autonomy of folk psychology. Rather than reduction or elimination, neural reuse and extended cognition so conceived actually support functionalism in the sense that we individuate processes with regard to their role and effects as opposed to their structural constituents. Again, looked at in this way, we find a great deal of multiple realizability within and across human and non-human brains. Anderson places a lot of emphasis on the importance of selection in understanding brain function (p. 296), and I agree, but I think there are others aspects of evolutionary and developmental biology that are equally important and that do not fully reduce to selection. Evolutionary theory has its own version of functionalism in the form of convergent or parallel evolution such as mimicry and flying (McGhee Reference McGhee2011). Focusing on cognitive convergence, creatures with very different brains and selective histories seemingly converge on similar behavioral and cognitive strategies for dealing with “socioecological” problems. For example, corvids do not even possess a prefrontal cortex, yet they exhibit very intelligent behavioral and cognitive strategies similar to primates. There is growing evidence across the board that creatures with very different brains have in many respects converged on relatively similar minds (McGhee Reference McGhee2011). Well-known examples of this from insects to mammals include: tool use, architectural behavior, agricultural behavior, social or collective behavior, mathematical behavior, and language use (McGhee Reference McGhee2011, Ch. 6). There is also a growing consensus that a wide variety of different species with very different brain structures and nervous systems possess some form of not only sophisticated convergent cognition, but also consciousness, self-awareness, and metacognition. Examples abound, such as the mourning behavior of gorillas and dolphins (McGhee Reference McGhee2011, p. 240). I think the best explanation for cognitive convergence is that the affordances, environmental and social features often trump structural neural constraints whether imposed by physics or selection.
So, although I agree with Anderson that the brain must have its voice, what I think it is telling us is that neural reuse is best viewed as a subset of what developmental and evolutionary biologists call plasticity and robustness (Bateson & Gluckman Reference Bateson and Gluckman2011). Although there are many different kinds of both robustness and plasticity, in general robustness refers to relative stability or invariance across environmental, genetic, or cellular transformations, and plasticity refers to cases wherein features of the organism are held constant such as genotype, and yet because of environmental transformations the organism nonetheless manifests very different or unique adaptive traits or behaviors (Bateson & Gluckman Reference Bateson and Gluckman2011, p. 8). Different forms of plasticity include: phenotypic, molecular, variety of neural types, immunological, and behavioral (Bateson & Gluckman Reference Bateson and Gluckman2011, p. 46). Both of these features of complex biological systems are of course at the heart of the epigenetic revolution in biology. As many people have pointed out, robustness and plasticity are two sides of the same coin: “Plasticity is often regulated by robust mechanisms and robustness is often generated by plastic mechanisms” (Bateson & Gluckman Reference Bateson and Gluckman2011, p. 46). It is important to note that, although many biological mechanisms possess the properties of robustness and plasticity, these features cannot in principle be explained mechanistically in terms of localization and decomposition. These are global/systemic and scale-invariant features of such biological systems. Convergent evolution, robustness, and plasticity all go hand-in-hand, and they all point to the strongly extended nature of phenotype, behavior, and cognition. For one take on what such a Gibsonian cognitive neuroscience might look like, see Silberstein and Chemero (Reference Silberstein, Chemero, Orden and Stephen2012).