Acta Neuropsychiatrica 2007: 19: 139–148 All rights reserved DOI: 10.1111/j.1601-5215.2007.00204.

x

# 2007 The Authors Journal compilation # 2007 Blackwell Munksgaard

ACTA NEUROPSYCHIATRICA

Review article

When brains expand: mind and the evolution of cortex
Kirkcaldie MTK, Kitchener PD. When brains expand: mind and the evolution of cortex. Objective: To critically examine the relationship between evolutionary and developmental influences on human neocortex and the properties of the conscious mind it creates. Methods: Using PubMed searches and the bibliographies of several monographs, we selected 50 key works, which offer empirical support for a novel understanding of the organization of the neocortex. Results: The cognitive gulf between humans and our closest primate relatives has usually been taken as evidence that our brains evolved crucial new mechanisms somehow conferring advanced capacities, particularly in association areas of the neocortex. In this overview of neocortical development and comparative brain morphometry, we propose an alternative view: that an increase in neocortical size, alone, could account for novel and powerful cognitive capabilities. Other than humans’ very large brain in relation to the body weight, the morphometric relations between neocortex and all other brain regions show remarkably consistent exponential ratios across the range of primate species, including humans. For an increase in neocortical size to produce new abilities, the developmental mechanisms of neocortex would need to be able to generate an interarchy of functionally diverse cortical domains in the absence of explicit specification, and in this respect, the mammalian neocortex is unique: its relationship to the rest of the nervous system is unusually plastic, allowing great changes in cortical organization to occur in relatively short periods of evolution. The fact that even advanced abilities like self-recognition have arisen in species from different mammalian orders suggests that expansion of the neocortex quite naturally generates new levels of cognitive sophistication. Our cognitive and behavioural sophistication may, therefore, be attributable to these intrinsic mechanisms’ ability to generate complex interarchies when the neocortex reaches a sufficient size. Conclusion: Our analysis offers a parsimonious explanation for key properties of the human mind based on evolutionary influences and developmental processes. This view is perhaps surprising in its simplicity, but offers a fresh perspective on the evolutionary basis of mental complexity.

Matthew T. K. Kirkcaldie1, Peter D. Kitchener2
1 Department of Physiology, School of Medical Sciences, The University of New South Wales, Randwick, New South Wales, Australia; and 2 Department of Anatomy and Cell Biology, The University of Melbourne, Melbourne Victoria, Australia

Keywords: cerebral cortex; cognition; prefrontal cortex; neuroanatomy Dr Matthew T. K. Kirkcaldie, Department of Physiology, School of Medical Sciences, Rm 308a1, Wallace Wurth Building, The University of New South Wales, Randwick, NSW 2052, Australia. Tel: 161 2 9385 2560; Fax: 161 2 9385 1059; E-mail: m.kirkcaldie@unsw.edu.au Both the authors contributed equally.

Consciousness presents a special challenge for neuroscientists; despite increasingly detailed confirmation that it is produced by the action of the brain, it has been considered an inappropriate, or at least unwise, area of study. Omitting the brain’s most important function has placed neuroscientists in a strange situation. If we maintain that consciousness is, in principle, not amenable to investigation, then we are effectively dualists – or

Ôcloset dualists’ (1), if we also adopt the fashion that dualism is untenable. Otherwise, we may be tempted to endorse the claim that consciousness arises from some complexity of neuronal interaction, as an Ôemergent feature’ of the highly recursive information processing transformations afforded by neuronal interactions. But as Kevin Kelly points out in his postscript to Out of Control (2), Ôemerges’ can be just another way of saying 139

Kirkcaldie and Kitchener Ôhappens’, in which case an Ôemergent’ explanation for consciousness might be little more than an appeal to magic. More cautiously, we could take the position that the problem of consciousness is best avoided; that the gulf separating genes, membranes, synapses and circuits from embodiment, unity, agency and continuity (3) is unbridgeable. There has, nevertheless, been a growing interest in consciousness paralleled by an increasing sophistication in describing brain activity during specific conscious experiences. Some correlates, such as electroencephalograph (EEG) activity in different states of consciousnesses, are well established, whereas others such as functional magnetic resonance imaging (fMRI) or exploiting definable conscious percepts in binocular rivalry (4) have only recently enhanced our knowledge. Despite better empirical data, many fundamental questions about the forms and distribution of consciousness in animals remain unanswered. Our own consciousness is irrefutably self-evident, and we are comfortable in ascribing this consciousness to other people, but we have little insight into other species: are they conscious in a rudimentary or prototypic form? Do they have affective experiences, and if so, are they comparable to human emotions? Some insight has been gleaned by studying conscious behaviour in animals, for example, the mirror test (recognizing a reflection as the self, instead of another individual) is passed by some species: humans, some of our close relatives (5), elephants (6) and dolphins (7). This capacity seems closely aligned with Ôtheory of mind’: the ability to conceptualize the viewpoint of, and attribute mental states to, other individuals. Such a test is relatively specific but leaves unanswered the difficulty of explaining what consciousness is like in animals exhibiting empathetic behaviours to other members of their species (8). Even if we discount the apparent self-consciousness of such behaviour, how do we conceptualize what these animals experienced in themselves? Many mammals exhibit sophisticated perceptual and social behaviour, including recognition of conspecifics and their actions. As Griffin (9) has asked, would there really be within the conscious perception of the activity of other animals a Ôperceptual black hole’ where its own actions would be represented? It is similarly difficult for us to conceive of forms of consciousnesses that solely elaborate the experience of the present, with little conception of how the future may be shaped, and a memory embodied only as behavioural tendencies shaped by the unrecalled past. The biggest difficulty in understanding consciousness is understanding how there can be any sort of awareness or phenomenality. This has been termed 140 Ôthe hard problem’ (10) in recognition that consciousness is fundamentally unlike everything else we know of: there seems no way to relate it to any property or process, or any member of the set of all things that are not consciousness. The study of the brain in relation to consciousness is therefore necessarily confined to the Ôeasy’ problem – the neural events associated with the generation of phenomenality, the physiological correlates of consciousness. A contrasting view is that consciousness only seems unique but is in fact nothing more than the result of complex brain activity, as much explicable in terms of the physiology of biological entities as any other complex and superficially mysterious processes. Dennett (11) points out that while our intuition that consciousness is special may be powerfully compelling, intuition is a poor guide to the underlying nature of complex processes and suggests that we let go of this fondly held, unsupported belief, and set about better understanding the biological processes that generate consciousnesses: the easy, and only, problem. Calling the problem Ôeasy’ implies only that it comprises instances or combinations of known phenomena; a separate issue is whether practical considerations might nonetheless prevent a full explanation (12). Fortunately, progress in understanding how the brain generates consciousness may not require a choice between philosophical stances. The easy problems of consciousness are practical and useful pursuits, regardless of whether we believe that their solution offers a complete explanation of conscious experience. We might suspect that there are crucial organizing principles or fundamental phenomena yet to be apprehended and included in the analysis, but even if a detailed account of the brain’s activity cannot fully explain the associated phenomenality, such an account will almost certainly be an essential part of the comprehensive explanation. This view stems from Hubel and Wiesel’s seminal insight (13) about the visual system: that the experienced features of visual percepts coincide with the features extracted by the computational activity of neurons. There is still considerable uncertainty as to what analyses neurons in the visual system are actually performing (14), but it seems extraordinarily unlikely that a later understanding of how perceptual consciousness arises will show that this activity is arbitrary and unrelated to the generation of the associated percepts.
Neuronal activity and consciousness

It has been noted that one of the difficulties in studying consciousness is the lack of a fixed point from which to theorize (15). The lack of consensus

Mind and the evolution of the neocortex on the form that a theory of consciousness might take may also obstruct the interpretation of less overtly theoretical approaches, like attempting to characterize the neurophysiological activity related to aspects of consciousness. The obvious, and demonstrable, characterization of neurons as Ôinformation processing automata’ connotes that the neurophysiological correlates will need to be resolved to this computational level for any insight into the mechanism of phenomenality. Supporting such a conception are two of the most significant insights into neural function: the ionic basis of the action potential (16) and the extraction of higher order visual features by connections of neurons in the visual pathway (13). These insights rank among the greatest achievements in science, but they may not be the only conceptual tools needed to extend neurophysiological studies to the correlates of consciousness. It could be, for example, that there are higher order arrangements of these neural properties that the current computational view may not reveal (17) (analogously, the primary amino acid sequence of an enzyme does not explain its catalytic properties). Even at the level of single neurons, the characterization of information processing by neurons may be open to different interpretations (18): the discovery of retrograde dendritic action potentials in cortical pyramidal neurons (19) has supported the alternative concept of the action potential as a reset mechanism, acting to synchronize activity rather than express the sum of synaptic inputs. It might also be conjectured that features at other levels could be critical – can we, for example, overlook complex electrodynamic phenomena generated by concerted neuronal activity [see, e.g., Freeman (20)] in our eagerness to focus on the action potential? When relating neuronal activity to the more global functions of the brain, we soon run into interpretational difficulties in analyzing neurons beyond those which initially process sensory inputs. An example is the representation of perceptual objects by neurons that are not involved in the initial grouping and segmentation of the retinal image – itself a task of breathtaking complexity. Understanding the way in which visual percepts could correspond to the concerted activity of vast numbers of individual neural responses is conceptualized as an encoding problem; sparse and distributed coding offer contrasting models of how percepts or their antecedents might be encoded in such activity. Sparse encoding is an extension of the hierarchical feature extraction posited for lower levels of visual processing, such that particular combinations of low-level features trigger responses in a few specific neurons. In the extreme case, there are single neurons that respond to a single distinguishable object of perception. Alternatively, discernible percepts may arise from unique spatial and temporal distributions of neuronal activity. The difficulty of deciding between such explanations is highlighted by a recent investigation (21) of the medial temporal lobe in surgical patients, in which experimenters recorded single-unit responses to visual stimuli of well-known landmarks and people. The authors interpreted their results as supportive of sparse representation, but a most surprising aspect of the data was that an extraordinarily high proportion of recorded units (14% of the total and about one third of those with invariant responses to repeated stimuli) had markedly selective responses. Several neurons showed responses to multiple stimuli that were different representations of a single individual (such as photographs showing different views, a hand-drawn portrait and the written name of the actress Halle Berry) but not to other individuals. Such specificity could be consistent with a sparse code for an abstract representation of individuals but does not explain why so many of the neurons encountered were responsive to the relatively few objects in the stimulus set. The set of stimuli used was limited (extremely small, given the range of all possible visual stimuli) and although Berry would, by virtue of film and publicity, likely be familiar to the American subjects tested, it would be expected that neurons at the top of a sparse hierarchy representing the actress would be encountered only rarely. Part of the difficulty in revealing the Ôneuronal code’ of abstract visual percepts is that the role of these neurons within the entire neural response to the stimuli is unknown: the highly selective response exhibited could relate to perceptual representations and to top-down visual processes, attentional mechanisms, memory and motivational factors (21). The authors of the study also suggest that what is revealed by these responses may be something like place cells of the hippocampus, which generate a dynamic reference system for position in the environment (22). Such ÔHalle Berry neurons’ would analogously participate in dynamic bindings within a conceptual space, leaving questions of sparse and distributed representation unresolved, and offering little insight to their relationship to conscious percepts.

Does quality equal quantity?

An important consideration in comparing human minds to those of other animals (especially other primates) is the extent to which phenomena such as Halle Berry neurons, demonstrations of theory of 141

Kirkcaldie and Kitchener mind, self-recognition in reflections or the recently embraced Ômirror neuron system’ (23,24) necessarily indicate recent evolutionary changes in the human lineage. Other than the demonstrated differences in cognitive capacity and cultural sophistication, what evidence is there that human brains are different to those of other primates? The most obvious measure in which humans clearly deviate from the pattern exhibited by other primates is that their brain size to body size ratio is much greater; interestingly, other brain metrics are surprisingly uninformative in revealing where humans might be unique in relation to other primates. For example, correlations of the sizes of all brain subregions (including neocortex) to total brain size across a wide variety of primates fit rather tightly to a simple exponential function (25). Because all primates have a neocortex whose size is a simple exponential ratio of overall brain size (or, perhaps more meaningfully, subcortical brain size), we can assume that the underlying mechanisms controlling the size of the neocortex are the same in all primates. As summarized by Striedter (26), the human neocortex is roughly as large as we would expect, given the size of the brain. The same argument applies to the evident complexity of prefrontal connectivity in humans: although the proportion of cortex that is prefrontal in humans is not vastly different to that in other primates, the human prefrontal cortex (PFC) has recently been shown to possess a greater ratio of white matter to grey matter (27). But here again there is a simple exponential fit for all primates examined in this study, suggesting that the human PFC has exactly the degree of white matter connectivity that the size of the neocortex would predict. We view such consistent relationships across species as evidence that the human brain is just a proportionally larger version of a typical primate brain and that the increase in human cognitive sophistication occurred not by virtue of new genetic programs, specifying new types of neocortical properties, but simply as a result of human brains evolving to be much larger than those of other primates. An implication of this proposal is that few genetic differences between human neocortex and that of other primates are likely to emerge from exhaustive analysis (in developing and mature brains) of the neocortical regions that subserve higher cognitive functions. Although this is a parsimonious interpretation of the morphological data, it is contrary to the more usual assumption that those regions of the brain responsible for the capacities in which humans excel (such as abstract problem solving, extended temporal organization, language and symbolic reasoning) are areas of the 142 human brain that have undergone most evolutionary change. However, it is also possible that few, if any, of the known genetic differences between humans and chimpanzees (28,29) are involved in neocortical function. There is also no compelling reason to suspect that new kinds of neocortical architecture have evolved in humans: for example, there is no extant evidence for new cytoarchitectonic parcellations of the PFC, and what were initially thought to be a neuronal phenotype unique to human neocortex (Ôspindle’ cells) have also been found in other great apes (30) and cetaceans (31). Instead of positing mutations to somehow produce novel cerebral architectures or modes of connectivity, we propose an alternative hypothesis that the phenomena represented by Halle Berry neurons, sensory processing, mirror neurons and self-consciousness are latent in all primate brains, but are variably manifested depending on the relative amount of neocortex. To justify the view that simply generating larger sheets of neocortex could produce new types of abilities, we need to examine the way that functions are assigned to cortical regions during development and to understand how the relationship between neocortex and the rest of the nervous system is uniquely tolerant of radical changes in their relative proportion.

Evolutionary and developmental plasticity of the neocortex

At first glance, the idea that the human neocortex is not qualitatively special is counterintuitive. When we look for the substrate of mental experiences, it is clear from neuropsychological evidence that the most closely related structure is the neocortex. Strokes and traumatic damage have shown that specific cortical regions are sometimes closely matched to specific abilities and qualities of visual experience, such as recognition and naming, the experience of color, or emotional responses to faces; not only can these abilities or facets be selectively damaged by circumscribed lesions but such damage can even abolish them entirely from all mental states – for instance, bilateral damage resulting in achromatopsia not only removes the ability to perceive color but also destroys the memory of color and all understanding of its subjective qualities [e.g., Ogden (32)]. In such cases, a single ability or subjective experience has been excised from the conscious mind, although the conscious mind itself is still experienced, and the individual who has lost the ability is aware of the loss and retains much of the Ôself’. Complementarily,

Mind and the evolution of the neocortex damage to prefrontal regions often changes the self as perceived by others: planning, evaluation and propriety of behaviour are severely compromised, without the individual being aware that his or her behaviour is observably different. The neocortex is therefore clearly associated with specific abilities and qualities of phenomenal experience, as well as with the regulation and integration of behaviour and the self. What is less clear is that such abilities and qualities may be continuously variable across mammal species. Whereas the nervous system below the diencephalon is fundamentally similar across distantly related vertebrate orders (mammalia, aves, reptilia) (33,34), the most obvious evolutionary changes in mammalian nervous systems have involved the neocortex and the diencephalon. Understanding how humans have been able to develop, and effectively utilize, much larger expanses of neocortex requires an understanding of the evolutionary uniqueness of the neocortex, and why such radical reconstruction is even possible in the brain, whose functions are so critical to survival. Allman’s lucid book Evolving Brains (35) provides a worthwhile analogy: while visiting a power plant, the author saw several generations of mechanical, electromechanical, vacuum- and solid-state electronic devices, all simultaneously regulating the plant’s operation.
When I asked why the older control systems were still in use, I was told that the demand for the continuous generation of power was too great to allow the plant to be shut down for [renovation] . thus there had been a progressive overlay of control technologies, the pneumatic, vacuum tube, and computer systems integrated into one functional system. (pp. 40–41)

The point here is that the regulatory and survival functions of the brain are so critical that it is simply not possible to substitute novel mechanisms (or to introduce new types of developmental principles) when progressing through the slow, stepwise refinement of evolution. There may be a far more effective brain that could be built out of the biological materials we have available, but starting from scratch on a more efficient approach is not compatible with intergenerational survival. Given the blind nature of mutation, any radical change in the nervous system that breaks existing functions will be fatal, or at least deleterious. Thus, to radically change the brain, evolution can only focus on regions that are not intimately engaged in physiological regulation and the survival functions of the organism.

One way past this bind is to make changes in a structure that interacts with the rest of the nervous system in a less direct manner: processing, elaborating and modulating the functions of the rest of the nervous system. In fact this is the evolutionary trump card of the mammals: the development of the neocortex, whose properties permit rapid and radical changes in function without fatally compromising the rest of the nervous system. With its generalized, six-layer structure, the relatively unspecialized anatomy of the neocortex (suggested by its other name isocortex) indicates the remove at which it operates: no cortical regions receive direct inputs from sensory fibers, and very few cortical neurons make direct synapses with effector neurons [with the exception of humans’ unusually overdeveloped corticospinal projection (36)]. Early in mammalian evolution [prior to the split between monotremes, marsupials and eutherian mammals (34)], this telencephalic structure consisting of multiple, generalized radial units emerged. Its adaptable structure proved useful both for processing sensory inputs and for elaborating and sequencing behaviour on large scales. The uniformity of the neocortex allows representations of inputs and outputs to directly interact, despite great differences in organizational structure in lower regions dealing directly with receptors and effectors. Cortex extracts abstract properties from the senses, applying its intrinsic processing abilities (lateral inhibition, tuning, extraction of temporal and spatial structure, pattern learning) to inputs derived from very different sources. An interesting phylogenetic hypothesis advanced by Marcello Rosa and Rowan Tweedale (37) rests on another key property of the mammalian neocortex: its self-connectivity at characteristic scales, which implies that neocortical regions developing beside each other will interconnect extensively and in a structured manner. When one of those cortical regions is driven by sensory activity relayed by the thalamus, the effect of this interconnection is that the region more removed from the input will form a derived representation of the activity in the Ôprimary’ region and will reciprocally project to modulate it. In effect, sensory inputs to a region of neocortex not only cause that region to reflect the input patterns but also cause surrounding cortical regions to mirror secondary features derived from activity in the primary region. If evolution adds more cortex, the characteristic length of corticocortical interconnection will ensure that regions further out will instead derive their activity from the secondary region, extracting third-order properties and reciprocally modulating the source. As more neocortex is added, the degree of remove 143

Kirkcaldie and Kitchener increases such that the animal ends up with primary, secondary, tertiary, quaternary and yet more abstracted regions, all interacting in a sequential, polysynaptic stream radiating out from and projecting back toward the source of the sensory inputs. It is not much of a conceptual stretch to propose similar arrangements in the motor sequencing and pattern making regions of cortex: a primary region surrounded by regions operating at greater and greater levels of abstraction. If we consider a developing cortical sheet with several of these organizing centers influencing surrounding regions, the patches of cortex at several removes from multiple organizing centers are likely to derive their activity from a combination of these influences. Such a combination would allow cross-modal influences to be felt across large cortical distances, spreading outward through feedforward paths through one system and feedback paths in another. This is the arrangement we observe in the human brain: parietal cortex bordered by visual, auditory and somatosensory processing regions ends up representing polymodal associations and multimodal abstractions like spatial awareness. Recent studies show that such flexible functional assignation is evident in the mechanisms of neocortical development. During the generation of the cortical sheet, diffusible factors spreading from the rostral, caudal, lateral and medial margins cause differential expression of control genes, depending on their specific combination at various locations. Similar gradients in the thalamus – derived from simple genetic duplication – create gradients of expression there (38). As development proceeds, these control genes in turn specify the transcription of a range of cell surface adhesion molecules that allow outgrowing thalamocortical fibers, groping blindly along gradients of chemoattractants and shepherded by barriers of repellent molecules, to recognize corticothalamic axons growing the other way along the same pathways [e.g., Molnar & Blakemore (39), Hevner et al. (40)]. Axons that find each other divert to follow their counterparts back to their origin, reciprocally interconnecting thalamic and cortical regions that express matching markers [e.g., the visual region specifier Ten_m3 (41)] and structurally patterning each by means of expression cascades triggered by the same control genes and their derivatives (42,43). As these islands of specific connectivity lock primary sensory and motor centers into the developing cortical sheet, it appears that the surrounding cortex functionally coalesces around them, forming as many derivative regions as there is available cortex around the organizing centers. The cortical sheet begins as an 144 almost isotropic expanse patterned by a few gene gradients into regions of connectional preference; simple rules and properties ensure that the initial connections are then able to produce a diverse cortical map with multiple regions involved in various sensory and motor systems. The unique flexibility of this system is highlighted in observations such as those of Bishop et al. (44): all it takes to drastically change the relative importance of visual, somatosensory and motor regions of the mouse brain, is to modify or abolish one or two gene gradients acting across the cortical sheet. If the Emx2 gene, normally expressed strongly at the caudal margin of the developing cortex, is mutated, the visual regions that normally form where it is strongly expressed will shrink down to tiny vestiges. Crucially, this reduction does not leave a region of useless cortex: instead, the rostral functional areas extend over more of the cortical sheet. Even quite remarkable changes in function at the periphery can be accommodated by smooth reconfiguration of cortical processing without requiring extragenetic specification, shown recently by a study in which mice augmented with a human visual cone pigment gene were able to utilize the extrareceptor information to achieve trichromatic vision (45). Although this picture is greatly simplified, it captures the essential qualities of the neocortex in the context of the evolution of complex brains. That is, the neocortex’s sophisticated sensory processing and behavioural modulation is achieved through indirect connectivity, which allows significant changes in organization to occur while leaving the regulatory and survival functions of the nervous system intact. Second, when changes are made to the large-scale functional groupings of the cortex, they typically result in a smooth reconfiguration of cortical territory; furthermore, if larger areas are given to particular functions, the characteristic radius of corticocortical connection ensures that islands of derived activity will form in the surrounding uncommitted cortex. Taken together, these properties imply that in environments that favor more sophisticated sensory processing and more elaborate motor planning, animals with a neocortex will be able to evolve these functions quite readily: the mutational changes need only include an increase in cortical progenitor division (46) and changes in the relative expression of master control genes. In this way, the brain can grow and place different priorities on vision, audition or motor control because the neocortex will self-configure to make the best use of these changes. The relative ease of such changes is evident if we compare the vast range of mammal

Mind and the evolution of the neocortex cortex sizes and layouts, to the tiny changes that have occurred in the rest of their nervous systems over the same time period: features such as the types of peripheral receptors and the reflex control of muscles are near-identical across all mammals (and most vertebrates). Even a Ônew sense’ such as echolocation in bats is achieved by a neocortical reconfiguration, rather than a fundamental change in the sensory structures of the ear (47). A further implication of these mechanisms is that the human brain could have achieved its distinction from those of other primates the same way: by an expansion of the neocortex, and a shifting of its functional priorities. Given the roughly equal size of primary sensory regions [e.g., Rosa & Tweedale, fig. 3 (37)], it is apparent that the expansion seen in human neocortex is principally of Ôuncommitted’ regions, whose inputs and activity derive from surrounding neocortex. In neocortex close to the primary sensory regions, it is easy to experimentally investigate the large proportion of activity derived from incoming stimuli, and accordingly these regions have been identified as secondary, tertiary and quaternary visual regions, for example. However, the further removed a cortical region is from a primary sensory region, the more difficult it is to characterize its activity in terms of experimentally manipulable stimuli. There is no reason to assume that the developmental process of these so-called association areas is significantly different from those of the well-defined secondary and tertiary sensory representations, and we should therefore conclude that their activity consists of similar cortical processing, albeit at more distant removes from the flux of sensation and the requirements of motor sequencing. Their activity could be characterized as derivatives of sensory and motor patterns – perhaps not so much high-level abstractions as multiply-processed facets of the activity of other brain regions, abstracted in proportion to the degree of their physical separation on the sheet of the cortex. Less a hierarchy than an interarchy of evoked and generated patterns, this view of the neocortex implies that a rich repertoire of behavioural responses and evoked conscious states would be available to individuals with sufficient cortex to generate a diversity of cross-modal interactions (48). These self-organizing qualities of developing cortex imply that more sophisticated individual behaviour might be produced simply by mutations that yield a larger neocortex. In conditions that permit the individual to support its greater metabolic demands (e.g., using a broader behavioural repertoire to obtain more food), a large brain with this capability could quickly become advantageous, rather than requiring multiple complementary mutations to slowly accumulate. This view suggests that an advanced so-called higher order consciousness like ours might be a natural consequence of growing enough neocortex to produce areas whose functions are sufficiently abstracted from primary sensory and motor activity. Even such advanced notions such as awareness of the self appear to be shared by mammals as diverse as primates, dolphins and elephants (6,7); intriguingly, primates with brains of intermediate size have been shown to exhibit behaviours with some – but not all – of the characteristics of self-recognition when confronted with a mirror (49). This cognitive property common to diverse mammals with large brains suggests that recognizing the self as an entity is an ability latent in the representational abilities of mammalian neocortex.

Choosing the simplest explanation

In examining the relationship between consciousness and neurophysiology, we have touched on several behaviours and observations of neural activity: the celebrated Ômirror neuron system’, activated by planning actions but also by similar actions in others; behaviour indicating an awareness of self, in several mammalian orders; and distributed representation of complex real-world entities (ÔHalle Berry’ neurons). The complexities of these phenomena can be reduced by seeing them as instances of the same type of neocortical activity – reflecting specific features of sensory data, depending on the available inputs, and thereby capturing properties of real-world entities to guide behaviour: if we can predict an entityÔs actions or future movement, we can interact with it more effectively (e.g., move out of its way, or catch it). Such explanations do not require explicit, abstract modeling in the neocortex: rather, by representing and analyzing the world so many different ways, behaviour is guided as if the brain had made abstract models. We know that there is no distinct representational modeling process because cortical damage breaks aspects of the understanding of objects (50) but never completely breaks interaction with the world, unless consciousness itself is abolished. Thus, Ôrepresentations’ are distributed and dynamic – even a stimulus as specific as a particular person will evoke dynamically bound activity in neurons scattered throughout all the cortical regions activated by the sight/sound/name of that person, 145

Kirkcaldie and Kitchener and their learned associations. Even in the vanishingly small proportion of neocortical neurons sampled by Quiroga et al. (21), it was possible to identify neurons involved in representing Halle Berry, among all possible humans – neurons that were obviously specific to Berry but not exclusive to her. Similarly, in investigating the premotor regions, Rizzolatti and Craighero (23) have repeatedly encountered neurons resonating with the notion of particular movements, whether in the process of being generated by the animal whose brain they are in or visible in the actions of other individuals. Of course, the sensory consequences of the animal’s own movement are part of the way that movement is originally learned, so there should be no surprise that visual and auditory sensory activity forms part of the inputs available to premotor neurons. Rather than a specific Ômirror neuron system’, it is helpful to view these neurons as exemplifying the way that derivatives of sensory inputs spread throughout the cortex, resonating with and activating neurons whose other associations have a similar quality (48). If a neuron is involved in guiding an arm, sensory experiences of seeing another animal move its arm are likely to match and resonate. Thus, we would argue that the Ôrepresentations’ of the neocortex are an epiphenomenon of its tendency to analyze and derive new information from its inputs – objects having identifiably different features and qualities will trigger different meshes of cortical activity, which we could choose to interpret as feature representations in primary cortex (14), as distributed representations of cognitive entities (21) or as mirrors of observed behaviour (23). The interpretational difference is prompted by the neurons’ location on the neocortical sheet and hence their specific mix of sensory inputs and motor response planning; however, they are all reflective of the same types of evoked and elaborated activity. Singling out a Ômirror neuron’ is akin to identifying a Ôfeature neuron’ in V1 – it is much simpler to assume that all cortical neurons exhibit facets of feature representation and mirroring of external events. Mirror neurons have attracted attention because their efferent connectivity is uniquely well understood because they help to produce observable movements. The simplest conclusion to draw from the available evidence is that the rudiments of a notion of self arise when an animal has sufficient neocortex for representations of sufficient complexity, rather than being reliant on some distinct ability that arose separately in three or more mammalian orders. This naturally leads to the idea that the self might be a structured recognition of the states of 146 other parts of the neocortex, a notion remarkably close to the ideas of Hughlings Jackson (51). Just as we might understand, explain and predict the behaviour of an external entity by observing it, we may understand, explain and predict our own behaviour by applying exactly the same analysis to the responses of our own nervous system. That is to say, when cortex achieves sufficient size, the representational complexity that naturally ensues might effectively Ômodel’ the animal’s own behaviour, giving it a notion of its own preferences, and how it might usually react in particular circumstances. Such self-knowledge, mediated in humans by the PFC, would be of direct survival advantage to animals whose lives are dominated by social context (48). As we have argued, the evidence of phylogeny (35,37) and developmental genetics (38,44) suggest that the neocortex generates increasingly complex fields of sensorimotor interaction naturally – as long as evolutionary selection favors the growth of a larger cortex. If the neocortex becomes large enough, we suggest that behaviours related to the notion of the self emerge from the greater diversity of cortical representation and interaction available. We are not suggesting that the considerable [and possibly accelerated (52)] evolution of genes expressed in the central nervous system is irrelevant to brain development and function – it is, of course, highly likely that mutations along the human lineage have affected the brain in ways related to our cognitive capabilities. A small number of genes (53,54) and regulatory regions (55) are known to have changed since humans diverged from other apes, but how these changes relate to any aspect of brain function is hard to evaluate. For example, the human FOXP2 gene was initially hailed as a Ôlanguage gene’ because of the profound language deficits seen in members of a family with a mutation in this gene; however, this description was not supported by subsequent recognition of broader cognitive deficits caused by the mutation, and the failure to implicate FOXP2 in any specific language impairments (53). While mutations that disrupt FOXP2 protein function have markedly deleterious effects on brain function, it remains to be shown whether our variant of FOXP2 confers specifically human brain properties: we do not know, for example, if substituting the chimpanzee form of the gene would affect language or other functions. Of possible significance to our proposal is the recent interest in microcephaly-related genes, mutations of which cause a marked overall reduction in brain size accompanied by mental retardation (52). These genes, including ASPM and microcephalin,

Mind and the evolution of the neocortex might therefore be developmental regulators of brain size and thus critical in determining our large brain to body ratio. There is evidence that both of these genes have undergone, and continue to undergo, recent evolution [see Ponting & Jackson (54) for review]: in this context, our proposal is that any resultant brain size increase could provide new functional capacity without requiring a new suite of genetic mechanisms to specify the functions of the expanded neocortical territory. If the future of human evolution includes an expansion of neocortical size, the implication of our hypothesis is that new abilities would arise at levels of abstraction beyond even our sophisticated awareness. Although consciousness itself may be frowned upon as a proper subject for neuroscience, it is remarkable to find that its fundamental problems can at least be framed in the terms of phylogeny, developmental biology and neurophysiology. By tackling the neural basis of mind as another instance of neocortical processing and elaboration, it may end up being far more tractable than many neuroscientists appear to believe. Whether the Ôhard problem’ reduces to easy problems, or whether crucial mechanistic insights lie in wait, the ultimate appreciation of consciousness will require neuroscience to call on all of its subdisciplines for analyses of brain function spanning all levels of organization, from descriptions of disease (56) to the flux of ions through neuronal membranes.
Acknowledgements
Our thanks to the participants in the INA Workshop on Neurophilosophy for broadening our outlook, as well as to Richard Vickery, Charles Watson, Tony Hannan, Tristan Hardy, Colin Hales and Peter Snow for conversations and critiques that greatly shaped our views on these subjects. 8. PRESTON SD, DE WAAL FBM. Empathy: its ultimate and proximate bases. Behav Brain Sci 2002;25:1–72. 9. GRIFFIN DR. Animal minds. Chicago: University of Chicago Press, 1992. 10. CHALMERS DJ. Facing up to the problem of consciousness. J Consciousness Studies 1995;2:200–219. 11. DENNETT DC. Sweet dreams: philosophical obstacles to a science of consciousness. Cambridge: MIT Press, 2005. 12. MILLER SM. On the correlation/constitution distinction (and other hard problems) in the scientific study of consciousness. Acta Neuropsychiatr 2007;19:159–176. 13. HUBEL DH, WIESEL TN. Brain and visual perception: the story of a 25-year collaboration. Oxford: Oxford University Press, 2004. 14. OLSHAUSEN BA, FIELDS DJ. How close are we to understanding V1? Neural Comput 2005;17:1665–1699. 15. ARCISZEWSKI MB, AKINS K. ‘‘The Race for Consciousness’’ by John Taylor (book review). Trends Neurosci 2000; 23:648–649. 16. HODGKIN AL, HUXLEY AF. A quantitative description of membrane current and its application to conduction and excitation in nerve. J Physiol 1952;117:500–544. 17. VICKERY RM. Mind the neuron. Acta Neuropsychiatr 2007;19:177–182. ¨ 18. KONIG P, ENGEL AK, SINGER W. Integrator or coincidence detector? The role of the cortical neuron revisited. Trends Neurosci 1996;19:130–137. ¨ 19. STUART G, SPRUSTON N, SAKMANN B, HAUSSER M. Action potential initiation and backpropagation in neurons of the mammalian CNS. Trends Neurosci 1997;20:125–131. 20. FREEMAN WJ. The wave packet: an action potential for the 21st century’. J Integ Neurosci 2003;2:3–30. 21. QUIROGA QR, REDDY L, KREIMAN G, KOCH C, FRIED I. Invariant visual representation by single neurons in the human brain. Nature 2005;435:1102–1107. 22. WILSON MA, MCNAUGHTON BL. Dynamics of the hippocampal ensemble code for space. Science 1993;261:1055–1058. 23. RIZZOLATTI G, CRAIGHERO L. The mirror-neuron system. Ann Rev Neurosci 2004;27:169–192. 24. OBERMAN LM, HUBBARD EM, MCCLEERY JP, ALTSCHULER EL, RAMACHANDRAN VS, PINEDA JA. EEG evidence for mirror neuron dysfunction in autism spectrum disorders. Brain Res Cogn Brain Res 2005;24:190–198. 25. STEPHAN H, FRAHM H, BARON G. New and revised data on volumes of brain structures in insectivores and primates. Folia Primatol 1981;5:1–29. 26. STRIEDTER GF. Principles of brain evolution. Massachusetts: Sinauer Press, 2005. 27. SCHOENEMANN PT, SHEEHAN MJ, GLOTZER LD. Prefrontal white matter volume is disproportionately larger in humans than in other primates. Nature Neurosci 2005; 8:242–252. 28. SIKELA JM. The jewels of our genome: the search for the genomic changes underlying the evolutionarily unique capacities of the human brain. PLoS Genet 2006;2:e80. 29. UDDIN M, WILDMAN DE, LIU G et al. Sister grouping of chimpanzees and humans as revealed by genome-wide phylogenetic analysis of brain gene expression profiles. Proc Natl Acad Sci USA 2004;101:2957–2962. 30. NIMCHINSKY EA, GILISSEN E, ALLMAN JM, PERL DP, ERWIN JM, HOF PR. A neuronal morphologic type unique to humans and great apes. Proc Natl Acad Sci USA 1999;96:5268–5273. 31. HOF PR, VAN DER GUCHT E. Structure of the cerebral cortex of the humpback whale, Megaptera novaeangliae (Cetacea, Mysticeti, Balaenopteridae). Anat Rec 2007;290: 1–31.

References
1. ARSHAVSKY YI. ‘‘Scientific roots’’ of dualism in neuroscience. Prog Neurobiol 2006;79:190–204. 2. KELLY K. Out of control: the new biology of machines. London: 4th Estate, 1994. 3. CAMPBELL K. Body and mind. London: Macmillan Press, 1971. 4. MILLER, SM, NGO, T. Studies of caloric vestibular stimulation: implications for the cognitive neurosciences, the clinical neurosciences and neurophilosophy. Acta Neuropsychiatr 2007;19:183–203. 5. KEENAN JP. The face in the mirror: the search for the origins of consciousness. New York: HarperCollins Press, 2003. 6. PLOTNIK JM, DE WAAL FB, REISS D. Self-recognition in an Asian elephant. Proc Natl Acad Sci USA 2006;103: 17053–17057. 7. REISS D, MARINO L. Mirror self-recognition in the bottlenose dolphin: a case of cognitive convergence. Proc Natl Acad Sci USA 2001;98:5937–5942.

147

Kirkcaldie and Kitchener
32. OGDEN JA. Visual object agnosia, prosopagnosia, achromatopsia, loss of visual imagery, and autobiographical amnesia following recovery from cortical blindness: case M.H. Neuropsychologia 1993;31:571–589. ´ 33. ABOITIZ F, MONTIEL J, LOPEZ J. Critical steps in the early evolution of the isocortex: insights from developmental biology. Braz J Med Biol Res 2002;35:1455–1472. 34. REINER AJ. A hypothesis as to the organization of cerebral cortex in the common amniote ancestor of modern reptiles and mammals. Novartis Found Symp 2000;228:83–102. 35. ALLMAN JM. Evolving brains. New York: W. H. Freeman & Company, 1999. 36. IWANIUK AN, PELLIS SM, WHISHAW IQ. Is digital dexterity really related to corticospinal projections?: a re-analysis of the Heffner and Masterton data set using modern comparative statistics. Behav Brain Res 1999;101:173–187. 37. ROSA MGP, TWEEDALE R. Brain maps, great and small: lessons from comparative studies of primate visual cortical organization. Philos Trans R Soc Lond B Biol Sci 2005; 360:665–691. 38. O’LEARY DD, NAKAGAWA Y. Patterning centers, regulatory genes and extrinsic mechanisms controlling arealization of the neocortex. Curr Opin Neurobiol 2002;12:14–25. 39. MOLNAR Z, BLAKEMORE C. How do thalamic axons find their way to the cortex? Trends Neurosci 1995;8:389–397. 40. HEVNER RF, MIYASHITA-LIN E, RUBENSTEIN JL. Cortical and thalamic axon pathfinding defects in Tbr1, Gbx2, and Pax6 mutant mice: evidence that cortical and thalamic axons interact and guide each other. J Comp Neurol 2002 May;447:8–17. 41. LEAMEY CA, KANG N, CROISIER E et al. Ten_m3 is expressed in the developing visual pathway. Program No. 567.19. Washington D.C.: Society for Neuroscience, 2003. 42. INCE-DUNN G, HALL BJ, HU SC et al. Regulation of thalamocortical patterning and synaptic maturation by NeuroD2. Neuron 2006;49:683–95. 43. SUR M, LEAMEY CA. Development and plasticity of cortical areas and networks. Nat Rev Neurosci 2001; 2:251–62. 44. BISHOP KM, GOUDREAU G, O’LEARY DDM. Regulation of area identity in the mammalian neocortex by Emx2 and Pax6. Science 2000;288:344–349. 45. JACOBS GH, WILLIAMS GA, CAHILL H, NATHANS J. Emergence of novel color vision in mice engineered to express a human cone photopigment. Science 2007;315: 1723–1725. 46. RAKIC P. Less is more: progenitor death and cortical size. Nat Neurosci 2005;8:981–982. 47. RAZAK KA, SHEN W, ZUMSTEG T, FUZESSERY ZM. Parallel thalamocortical pathways for echolocation and passive sound localization in a gleaning bat, Antrozous pallidus. J Comp Neurol 2007;500:322–338. 48. BARRETT L, HENZI P. The social nature of primate cognition. Proc Biol Sci 2005;272:1865–1875. 49. DE WAAL FB, DINDO M, FREEMAN CA, HALL MJ. The monkey in the mirror: hardly a stranger. Proc Natl Acad Sci USA 2005;102:11140–11147. 50. DE RENZI E. Disorders of visual recognition. Semin Neurol 2000;20:479–485. 51. MEARES R. The contribution of Hughlings Jackson to an understanding of dissociation. Am J Psychiatry 1999; 156:1850–1855. 52. DORUS S, VALLENDER EJ, EVANS PD et al. Accelerated evolution of nervous system genes in the origin of Homo sapiens. Cell 2004;119:1027–1040. 53. FISHER SE, LAI CSL, MONACO AP. Deciphering the genetic basis of speech and language disorders. Ann Rev Neurosci 2003;26:57–80. 54. PONTING C, JACKSON AP. Evolution of primary microcephaly genes and the enlargement of primate brains. Curr Opin Genet Dev 2005;15:241–248. 55. POLLARD KS, SALAMA SR, LAMBERT N et al. An RNA gene expressed during cortical development evolved rapidly in humans. Nature 2006;443:167–172. 56. HANNAN AJ. Brain phylogeny, ontogeny and dysfunction: integrating evolutionary, developmental and clinical perspectives in cognitive neuroscience. Acta Neuropsychiatr 2007;19:149–158.

148