Original Paper

Adaptive Behavior 21(3) 199214 The Author(s) 2013 Reprints and permissions: sagepub.co.uk/journalsPermissions.nav DOI: 10.1177/1059712313483145 adb.sagepub.com

Turing instabilities in biology, culture, and consciousness? On the enactive origins of symbolic material culture
Tom Froese1,2,3, Alexander Woodward1 and Takashi Ikegami1

Abstract It has been argued that the worldwide prevalence of certain types of geometric visual patterns found in prehistoric art can be best explained by the common experience of these patterns as geometric hallucinations during altered states of consciousness induced by shamanic ritual practices. And in turn the worldwide prevalence of these types of hallucinations has been explained by appealing to humanitys shared neurobiological embodiment. Moreover, it has been proposed that neural network activity can exhibit similar types of spatiotemporal patterns, especially those caused by Turing instabilities under disinhibited, non-ordinary conditions. Altered states of consciousness thus provide a suitable pivot point from which to investigate the complex relationships between symbolic material culture, first-person experience, and neurobiology. We critique prominent theories of these relationships. Drawing inspiration from neurophenomenology, we sketch the beginnings of an alternative, enactive approach centered on the concepts of sense-making, value, and sensorimotor decoupling. Keywords Enaction, sense-making, representation, hallucination, Turing patterns, human cognition

1 Introduction
Forms of artistic expression, rituals, and language are universal features of all known human cultures. While animal behavior is mostly governed by immediate environmental demands and the meaning of animal communication is generally fixed by biological evolution, human symbolic practices and their meanings are the historical outcome of a seemingly open-ended social process of cultural evolution. Explaining the origin of such symbolic practices in terms of our biological embodiment and the social circumstances of our prehistoric past is one of the major outstanding challenges of science. Indeed, the origin of symbolic representation concerns the origin of the human condition as such, hence the idea of modern humans as the symbolic species (Deacon, 1997) or Homo symbolicus (Henshilwood & dErrico, 2011b). This is an issue that reaches across the great divide between the natural and social sciences, and has implications beyond the remit of science itself. The enactive approach to cognitive science is in a good position to advance the debate about the origins of the earliest symbolic practices. Whereas traditional cognitive science presupposes the existence of an internal symbol system as its basic starting point (Newell

& Simon, 1976), the enactive approach instead starts with the biologically embodied mind, and must therefore address the cognitive gap between adaptive behavior and abstract human cognition (De Jaegher & Froese, 2009; Froese & Di Paolo, 2009). Thus, rather than simply assuming the notion of representation as its most basic conceptual foundation for explaining the internal mechanisms of human cognition, it tries to understand the biological, social, and historical origins of the phenomenon of symbolic representation as an aspect of our cultural environment. This approach helps to avoid a lot of the unnecessary linguistic and theoretical confusion, which has plagued cognitive science for many years (Harvey, 2008). And it also helps
1

Ikegami Laboratory, Department of General Systems Studies, Graduate School of Arts and Sciences, University of Tokyo, Tokyo, Japan 2 ticas Aplicadas y en Sistemas Instituto de Investigaciones en Matema noma de Me xico (UNAM), Me xico (IIMAS), Universidad Nacional Auto 3 Centro de Ciencias de la Complejidad (C3), Universidad Nacional noma de Me xico (UNAM), Me xico Auto Corresponding author: n, Instituto de Departamento de Ciencias de la Computacio ticas Aplicadas y en Sistemas (IIMAS), Investigaciones en Matema noma de Me xico (UNAM), Apdo. 20-726, Universidad Nacional Auto 01000 Mexico D.F., Mexico. Email: t.froese@gmail.com

200 to bring to light some unresolved foundational issues. The question therefore becomes how primary adaptive processes of sense-making of the here and now could have been transformed into secondary forms of symbolic sense-making of the absent and the imaginary (Froese, 2012). Enactive accounts of this profound qualitative transition are still in their infancy, but there is broad agreement that there is no one single biological or social explanatory factor. We are dealing with a historical process emerging from the interaction between biological processes, social practices, and the cultural background (Froese, in press; Hutchins, 2010; McGann, 2007; Stewart, 2010). The development of an account that theoretically bridges the cognitive gap therefore promises to simultaneously provide an interdisciplinary bridge between the natural and social sciences. Furthermore, because the interaction between cultural processes and biological embodiment is mediated at the personal level through lived experience, this scientific integration is not limited to objective processes, but also includes phenomenology of the first-person perspective. Cognitive science thereby becomes an anthropologically informed cultural neurophenomenology (Laughlin & Throop, 2009). The aim of this paper is to contribute to this enactive approach to the various interplays between biology, culture, and consciousness, in particular by critically examining the role of altered states of consciousness in the generation of geometric prehistoric art.

Adaptive Behavior 21(3) for the cross-culturally shared value of these specific kinds of geometric patterns. We then evaluate current models of the neural basis of geometric hallucinations, because basic neural processes are prime candidates for explaining this crosscultural selective bias. These models typically propose to view the visual system as a potentially excitable medium whose autonomous dynamics are unleashed by disinhibiting altered states of consciousness. Essentially, it is suggested that the disinhibited visual system generates spatiotemporal patterns of neural activity due to Turing instabilities. Researchers also generally claim that the geometric hallucinations experienced by the subject are mental representations of these neural patterns. However, while these neural models are capable of reproducing some of the geometric patterns that are found in prehistoric art and non-ordinary visual experiences, their range remains severely limited. In addition, the models tend to trivialize the relationship between the structure of subpersonal neural processing and the content of personal visual experience by assuming that visual experience is a representation of inverse optics applied to neural activity in region V1 of the visual nervous system. We then propose an enactive approach to resolving these issues. By drawing inspiration from the method of neurophenomenology, we argue that the role of selfsustaining neural activity, which is closely associated with neural Turing instabilities, has been underappreciated. According to an enactive approach, selfsustaining neural dynamics can generate their own intrinsic value in relation to their conditions of selfmaintenance, and they can also serve as a neural mechanism by which to decouple autonomous brain activity from the influence of environmentally mediated sensorimotor dynamics. Both of these aspects can help to explain the aesthetic selective biases of the first artists, in particular their interest in inner experience as exemplified by abstract hallucinations and imaginary phenomena, which are not directly related to the demands of their physical environment. We speculate that the self-sustaining dynamics may account for why these geometric hallucinations were experienced as more significant than other phenomena, and that at the same time their underlying neural dynamics may have served to mediate and facilitate a form of imaginary sense-making that is not bound to immediate surroundings.

1.1 Overview of the paper

We first take a closer look at the one of the most prominent theories of the origin of prehistoric art and highlight its main shortcomings. The prevalence of certain geometric patterns in the symbolic material culture of many prehistoric cultures, starting shortly after the emergence of our biological species and continuing in some indigenous cultures until today, is explained in terms of the characteristic contents of biologically determined hallucinatory experiences. However, we argue that the correlation between the first artistic motifs and typical hallucinatory experiences is not sufficient to serve as a full explanation. In particular, there is a lack of consideration of the value associated with altered states of consciousness, both in terms of phenomenology and function. What is it about these nonordinary visual patterns that made them more attractive for artistic expression than most others of an almost infinite set of possible patterns, both physical and imaginary?1 Given that humans appear to be in principle capable of arbitrarily associating any kind of stimulus with any kind of meaning, as epitomized by language as an open-ended symbol system, there is a need to explain the shared selective biases that are in evidence across prehistoric cultures. In other words, we need to account

2 On the origins of the symbolic mind

Starting with Darwins (1871) evolutionary approach to the origins of human language in terms of natural and sexual selection, the date for the beginning of symbolic thought and creative imagination has been continually pushed back into prehistory. It was long assumed that

Froese et al. symbolic material culture originated during the last Ice Age in Europe. This standard view was supported by the available archaeological evidence, for instance by the famous Paleolithic cave paintings that were created by the first immigrating Homo sapiens starting from around 40,000 years ago. However, more recently the archaeological consensus has begun to be overturned by new evidence of an even older symbolic material culture (Henshilwood & dErrico, 2011a). These findings were uncovered in Africa, and indicate the presence of symbolic practices over 100,000 years ago, including the manufacture of paint and body ornamentation. In addition, and contrary to the idea that artistic expression began with primitive pictorial representations of the external environment, there was a tradition of engraving pieces of red ochre with a variety of abstract geometric patterns (Henshilwood, dErrico, & Watts, 2009). An example is shown in Figure 1. This discovery connects well with the ethnographical observation that geometric visual patterns have been playing an important role in prehistoric cultures from around the world (Froese, in press). Interestingly, these patterns represent only a small subset of the potentially infinite set of possible abstract visual patterns that could be created, and it seems that certain kinds of pattern are particularly frequent, such as dots, circles, cross-hatchings, parallel wavy lines, and especially various kinds of spirals (Lewis-Williams & Dowson, 1988). An important clue to the origin of this selective bias was found by psychologists investigating the experiential effects of various kinds of alterations of consciousness. They discovered that the patterns of many visual hallucinations could be categorized into a small number of so-called form constants (Klu ver, 1967). In the 1920s, Klu ver had systematically studied the effects of mescaline (the main psychoactive compound of the peyote cactus, Lophophora williamsii) on the experience of its users (including on himself). He observed that the non-ordinary visual experiences were often characterized by similar kinds of abstract geometric patterns, which he classified into four categories of form constants: (1) gratings, lattices,2 fretworks, filigrees, honeycombs, and checkerboards; (2) cobwebs; (3) tunnels and funnels, alleys, cones, vessels; and (4) spirals. Klu vers form constants have since been found to occur in a variety of other kinds of altered states (Bressloff, Cowan, Golubitsky, Thomas, & Wiener, 2001). Intriguingly, these form constants turned out to resemble many of the abstract motifs that are often associated with prehistoric art from around the world, including Paleolithic cave art in Europe. These insights led to the provocative hypothesis that much of the content of prehistoric art was inspired by patterns and visions seen during altered states of consciousness (e.g., Clottes & Lewis-Williams, 1998; Lewis-Williams, 2002; Lewis-Williams & Dowson, 1988; Winkelman, 2002).

201

Figure 1. Illustration of a piece of red ochre with abstract geometric incisions that were made on one side around 73,000 years ago. This piece was found during excavations in Blombos Cave, South Africa. This and similar pieces from the same location are currently the oldest known human expression of abstract geometric patterns. Reproduced with kind permission from Elsevier (Henshilwood et al., 2009).

Of course, it still remains to be explained why these particular motifs were highly regarded by the artists and how these people became artists capable of symbolic expression in the first place. According to LewisWilliams (2002), the symbolic capacity was already provided by the evolution of an appropriate mutation in the brain, and the value of the particular content was derived from the hierarchical power structures enforced by shamans who had exclusive access to the visionary experiences. However, this hypothesis has several shortcomings. While complex social stratification and strife was clearly in evidence during the Neolithic period in Europe (Lewis-Williams & Pearce, 2005), this kind of social class conflict may not be generalizable to earlier periods or other parts of the world. For instance, the Jomon culture of prehistoric Japan, one of the oldest pottery traditions in the world dating back to around 16,500 BC, was the first culture in the Japanese archipelago to produce pottery decorated with abstract geometric patterns. This symbolic material culture flourished for well over 10,000 years, producing large quantities of pieces with recognizable form constants (e.g., Figure 2). However, compared to later symbolic material cultures in this region, the Jomon period is remarkable for its long-term stability, and because of the absence of strong archaeological evidence for a complex social hierarchy, social strife, and intergroup warfare (Habu, 2004). The hypothesis that the first symbolic practices were an outcome of a prehistoric class struggle is lacking solid archaeological evidence. More importantly, even if we could find an early prehistoric society that fits Lewis-Williams hypothesis, the existence of a rigid social hierarchy and restricted accessibility of altered states of consciousness by an elite class does not explain why such altered states were highly valued by these people in the first place. If the ritualized alteration of consciousness and its symbolic expressions were to be a tool of empowerment restricted to the elite, then the general population must have already valued such

202

Adaptive Behavior 21(3) 2001). Regarding visually expressive phenomena, it has been shown that the skin patterns of some animals can be generated by underlying reactiondiffusion systems (Kondo & Miura, 2010). Given the similarity between Turing patterns and some of the motifs frequently found in prehistoric art and experienced by subjects in altered states of consciousness, it makes sense to investigate whether the biological mechanisms underlying the production of these visual phenomena is amenable to an analysis in terms of Turing instabilities. Indeed, this proposal is in line with Klu vers observation that the hallucination is [.] not a static process but a dynamic process, the instability of which reflects an instability in its conditions of origin (cited in Bressloff et al., 2001).

3.1 The neural Turing mechanism

One of the originally proposed mechanisms for geometric hallucinations is that of a neural Turing mechanism, embodied in the WilsonCowan equations (H. R. Wilson & Cowan, 1973) and first mathematically analyzed in terms of visual hallucinations by Ermentrout and Cowan (1979). This is the neural context for the famous Turing Instability, which describes a reaction diffusion process in which an ordered macroscopic structure can emerge from local interactions under the non-equilibrium state (Turing, 1952). Turings idea can be readily transposed to the functioning of the nervous system. For example, action potential propagation along a neurons axon can be directly described by reactiondiffusion equations, and reactiondiffusion equations are analogical to the WilsonCowan neural network equations (H. R. Wilson, 1999, pp. 267268). We can think of the reaction component as the interactions between neuronal cells, and of the diffusion component as the spread of neural activity through local synaptic connections. Similarly, the local structure of neural interconnectivity dictates the type of emergent phenomena that can be produced. Neural network models of geometric hallucinations have gradually incorporated these empirical insights from neural anatomy and physiology, including the spatial arrangement of different neuronal cell types. The first model of geometric hallucinations to follow this idea, by Ermentrout and Cowan (1979), developed an isotropically connected two-layer neural network of excitatory and inhibitory neurons to represent the primary visual cortex (V1). Increasing an excitability parameter destabilized the rest state, bifurcating the system, and led to macroscopic spatial patterns emerging due to the lateral interactions of negative feedback. Such a general description of a locally connected network shows up the reactiondiffusion type properties of such systems, but they require specificity in parameters in order to generate patterns. The main limitation of this early work is that they do not consider the neural

Figure 2. A deep pot covered with various spiral patterns; an early artistic example of the spiral type of Klu vers form constants. This particular style of vessel, known as Moroiso, was made around 4000 BC by people of the Jomon culture (Kobayashi, 2004, p. 31), a relatively peaceful and non-stratified society of prehistoric Japan (Habu, 2004). (Photo taken by Tom Froese at the Archaeological Museum of Kokugakuin University, Tokyo, Japan).

non-ordinary experiences; otherwise their restriction could hardly have served as an instrument of control. In addition, this hypothesis leaves any functional connection of altered states of consciousness to the origin of symbolic practices, including their underlying neural effects, unexplored. It seems plausible that, at least under some circumstances, a temporary alteration of normal sense-making can lead to heightened levels of imagination and creativity (Dobkin de Rios & Janiger, 2003).

3 Neural mechanisms
Turing patterns consist of various geometric forms, including spots, traveling waves, grids, and spirals, which emerge out of the distributed activity of nonlinear dynamical systems with local excitatory and sparse inhibitory connectivity. Some illustrative examples derived from a simple GrayScott reaction diffusion system are shown in Figure 3. Simple spatiotemporal pattern formation was first described by Turing (1952) for the case of chemical reactiondiffusion systems. Turing proposed that these kinds of systems could help to explain morphogenesis, i.e., pattern formation during biological development. Since then, Turing patterns have been found to be ubiquitous in a variety of biological systems (Goodwin,

Froese et al. architecture of V1: the arrangement of neurons in hypercolumns and their properties as low-level feature detectors. These are addressed in Bressloff et al.s (2002) work, to be discussed later in this section.

203 externally coupled state (Miyashita, 1995), although there is some variability in these results (Kosslyn & Thompson, 2003). Other results have pointed to the primary visual cortex interacting with the hippocampus during offline episodic memory consolidation; highorder replay of episodic memories occurs with quite specific patterns of visual activity in V1 (Ji & Wilson, 2007). The visual cortex is located in the posterior region of the brains occipital lobe. V1 takes information from the Lateral Geniculate Nucleus (LGN), a region of the thalamus, which in turn receives information from retinal ganglion cells that each receive information from around 5100 photoreceptors. Anatomically, V1 is an array of hypercolumns involved in the low-level processing of visual input, with feature detectors tuned to visual aspects such as color, edges, contours, motion, etc. (S. W. Wilson, 1983). Between regions, there is a progressive mapping to 2D cortical surfaces and all of the models to be discussed operate in such a dimension. After V1, higher-order processing of visual information occurs in roughly 30 identified regions within the visual cortex (Miyashita, 1995). But it is presumptive to speak of these brain processes happening in some sort of projective sequence. For example, in cat brains it has been identified that

3.2 The location of geometric hallucinations within the brain

Where exactly in the brain do such geometric hallucinations emanate from? Also, is it correct to say that there is a single region that constitutes the generation of a visual pattern, knowing that the brain is a highly interconnected dynamical system? Considering that geometric hallucinations are unaffected by the physical movement of the eyes means that such phenomena must emanate from the endogenous activity of the brain. Secondly, that we see such phenomena supports the assumption that they must be dependent on the brains visual sub-system. All models of geometric hallucinations propose that such patterns emerge from V1, the first region of the visual cortex to receive visual input. This assumption is supported by fMRI scans that have shown that V1 is sometimes activated during mental imagery, i.e., when one imagines something visual in either an awake or dreaming stateeven when they are not in an

Figure 3. Examples of dynamic patterns exhibited by the GrayScott reactiondiffusion system under various parameters. Patterns are chosen as exemplars of various phenomena; see Pearson (1993) for a more systematic classification. (a) A spiraling pattern; (b) a chaotic pattern of travelling waves; (c) a line pattern, whereby lines grow at the ends and then bend to fill space; (d) a labyrinth pattern; (e) a hole pattern; (f) a pattern of unstable spots, whose fluctuating population is maintained by a balance between reproduction and disintegration; (g) a stable spot pattern, whereby spots reproduce to fill empty space. Reproduced with kind permission from Nathaniel Virgo (Virgo, 2011).

204 retinal input accounts for about 7% of synaptic connections to relay cells in the LGN, therefore 93% of input is non-retinal, coming from other regions in the brain (Van Horn, Eris xir, & Sherman, 2000). It is certain that similar physiology exists within the human visual system and this shows that a great deal of dynamic feedback processes must be occurring. Accordingly, models that do not account for this massive feedback have questionable biological plausibility.

Adaptive Behavior 21(3) of the original retinal frame of reference. This reasoning has important implications for explaining geometric visual hallucinations. For example, taking a uniformly textured image (in V1) and going from rectangular (cortical) to polar (retinal) coordinates can generate a tunnel like warping of the image. This is an effect common in many geometric hallucinations and perhaps also related to the tunnel seen in near death experiences (Blackmore & Troscianko, 1989). However, the tunnel like warping effect of the inverse retinotopic mapping is not always fitting. For example, to accommodate lattice type phenomena that do not have a clear center of origin or convergence, Ermentrout and Cowan (1979) state that these must occur at the periphery of the visual field. On the other hand, a cobweb with a clearly defined center is said to appear nearer the fovea (see Figure 4 for illustrative examples of these distinct kinds of form constants). But this seems a rather odd prediction. If this model is indeed biologically accurate then a large number of geometric patterns that do not exhibit a tunnel-like appearance should be constrained to only appearing at the periphery of our visual experience. As far as we know, the extant psychological literature does not make any mention of a strict division between such focal and peripheral types of hallucinatory visual patterns. Thus, although this is a prediction that still deserves to be further investigated experimentally, it is more likely an indication that the relationship between the retinotopic mapping and geometric hallucinations is less direct than originally assumed by Ermentrout and Cowan. Another shortcoming of using inverse retinotopic mapping to derive an image of the subjects experience is that it lacks consideration of the temporal dimension of consciousness (Varela, 1999). Empirical research suggests that abnormal activity in the visual system

3.3 The retinotopic map

In general, in current models of geometric hallucinations, an inverse retinotopic map is applied to neural activity in V1 in order to virtually project back into what would have been the causally responsible retinal space under normal conditions. The resulting image is intended to enable us to see what the subject would visually experience. The retinotopic map describes how retinal input projects onto the visual cortex. This physiological mapping has been studied extensively and, in simplified terms, is a conformal projection from retinal polar coordinates into cortical rectangular coordinates. More space in V1 is assigned to the foveal region than to the periphery and this is why our spatial detail is highest at this focal point in our vision. The reasoning principle underlying the application of an inverse mapping to the neural network models is that under normal awake conditions, when a subject is observing the external world, an undistorted view of the world enters the retina and then, for physiological reasons, gets distortedly mapped to V1. However, we never consciously experience this distorted view of V1. Therefore, it is argued, there must be some subsequent inverting process such that, in some manner, phenomena generated in V1 are actually experienced in terms

Figure 4. Illustration of typical form constants. (a) Cobweb; (b) lattice consisting of parallelograms; and (c) lattice consisting out of hexagons. In contrast to the cobweb form, the lattice forms do not converge on a center point. Reproduced with kind permission from Springer (Ermentrout and Cowan, 1979).

Froese et al. only becomes consciously experienced in the form of geometric hallucinations after more extensive recurrent processing (e.g., P. C. J. Taylor, Walsh, & Eimer, 2010), thus again implying a more indirect mapping between activity in V1 and conscious visual experience. More generally, we can also take issue with the phenomenology of mental imagery that is implied by the idea of deriving the subjects experience from a neural image. Under closer phenomenological examination, the experience of mental imagery is qualitatively different from the experience of seeing a 2D picture (Thompson, 2007).

205 that can emerge from a realistic non-linear model of a cortical sheet of noisy spiking neurons. Their model included a drug parameter and they found that under a normal state of activity there was incoherent lowfrequency firing. As they increased the excitation of the network, beginning from random initial conditions, they categorized the following emergent properties: (1) stripes, (2) spiralsfound to be very stable, (3) rings, and (4) collective bursts. Collectively these were termed first stage imagery, corresponding to the idea of form constants. Second stage imagery, which goes beyond geometric hallucinations to include complex imagery, without any doubt involves several other areas of the brain, including memory systems (de Araujo et al., 2012), and was not modeled. Interestingly, once spirals exist within the modeled neural system they cannot be extinguished, even if the drug parameter is adjusted into the ranges to support categories 3 and 4. We note that this kind of hysteresis can be useful during normal brain functioning. Summarily, they consider form constants to be elementary excitations in the primary visual cortex and these spatiotemporal patterns involve the same kind of mechanism as in experimentally observed drug induced epilepsy (we will return to epilepsy below). Milton, Chu, and Cowan (1993) developed similar work for generating spirals from a fixed excitation center that mimic cardiac arrhythmia. This also demonstrates how changes in neural parameters generate quite varied macroscopic phenomena despite having a consistent local connectivity. In general, however, these models do not generate all four categories of form constant. And despite some authors claiming that the performance of a large network does not depend on details of the model once the neural essentials have been incorporated (Fohlmeister et al., 1995), other models indicate that more of the characteristic form constants could be produced if the exact structure of V1 was considered.

3.4 The ubiquity of spiral and wave dynamics in the brain

As stated earlier, most models assume that hallucinatory phenomena are produced in V1 and then pursue a description of their neural underpinnings. But spiral and wave (circular wavefront) dynamics seem to be a ubiquitous feature of brains, not just in V1, and also of biologically realistic neural models. Huang et al. (2010) state that spiral waves which emanate from a central point, as an emergent property, can organize and modulate cortical population activity on the mesoscopic scale and may contribute to both normal cortical processing and to pathological patterns of activity, e.g., epilepsy. Such wave activity is found to be robust when considered within local excitatory interactions such as within a cortical sheet (a locally connected 2D network). But the brain has modulatory long-range connections that can change the properties of local sub-networks that could serve as regulatory protective mechanisms. For example, Rulkov, Timofeev, and Bazhenov (2004) developed a set of computationally efficient yet highly realistic map-based neuron models to explore large-scale cortical networks and their dynamics. In particular, the spiking activity in different types of cortical pyramidal (PY) (fast spiking, regular spiking, intrinsically bursting) and interneurons (INs) was developed. Experiments showed that wave properties were dictated by coupling parameters and that self-sustained dynamics were a function dependent on synaptic interaction between neurons, and not spontaneously firing cells. These findings support the idea of neural architecture being an important factor for the emergence of geometric patterns that are potentially experienced as hallucinations. That spontaneous waves of activity could persist within the network indefinitely is also interesting with respect to the ideas of self-sustainability during externally decoupled mental states, maintaining network wide activity to prevent neuronal atrophy from disuse, or the brains default networkwhere selfsustained dynamics allows the brain to be in a primed, high-dimensional state. Fohlmeister, Gerstner, Ritz, and van Hemmen (1995) investigated the type of spontaneous excitations

3.5 Modeling geometric hallucinations as a function of V1 neural architecture

Bressloff et al. (2001, 2002) developed a model that aimed to connect geometric visual hallucinations with the structure of V1. It is well known that V1 consists of a number of neural circuits that are involved in lowlevel image processing, responding preferentially to features such as oriented edges and local contours, etc. Their work proposes that the specific spatial arrangement and local coupling of these feature detectors, along with the neural Turing mechanism combine to generate form constants. According to their model, under normal conditions the visual system is in a stable state, but the introduction of drugs destabilizes V1 and spontaneously generates cortical activity that reflects the underlying architecture of V1. The possible form

206 constants of this model are specific linear combinations of the eigenfunctions of the neural architecture, called Planforms, which possess both spatial and temporal properties. Since V1 must avoid falling into these states during normal visual processing, if the influences of the retinal input are to have any effect in driving the system, this places strong constraints on the evolution of the visual cortex architecture, for example in terms of the sparsity of long-range inhibitory connectivity (Butler et al., 2012). Bressloff et al.s work represents perhaps the most complete description of a neural mechanism for generating geometric hallucinations, however, it is not capable of generating all of Klu vers form constants and does not seem to account for the ubiquitous spiral wave dynamics of the brain, which are realized by the neural dynamics in a model such as Folhmeister et al.s. To address this, it would be interesting to incorporate a more biologically plausible, spiking neural model. Secondly, the authors only considered oriented edges and local contours in their architecture; what generative power would the modeling of other feature maps, such as spatial frequency, binocular disparity, motion, and color provide? Furthermore, none of the models seems to account for the entirety of geometric hallucinations that are commonly reported in the archeological and anthropological literature, for example as described by LewisWilliams and colleagues (Lewis-Williams & Dowson, 1988). Future work along the lines of Bressloff et al.s model may someday show that the intrinsic properties of V1 are enough to describe all geometric hallucinatory phenomena, but given the large variety of patterns, this is unlikely. It is interesting to consider what other brain mechanisms could contribute to their formation, such as the large number of feedback pathways that exist within the visual system.3

Adaptive Behavior 21(3) toward one of the stable Planforms as proposed in Bressloffs model. Secondly and more generally, we know that V1 and other regions of the visual pathway show neural activity during imagination (Ganis, Thompson, & Kosslyn, 2005), which shows that they can be integrated into higher-level processing in the absence of external input. Could a feedback loop between these brain areas generate geometric hallucinations? This is much like the idea of a video-feedback system, where a video camera is positioned to record the same video monitor to which it is connected (Crutchfield, 1984, 1988). Despite a video feedback system being physically different, a conceptual analogy can be formed by making firstly a connection between the neural Turing mechanism and modeling the video-feedback as a reactiondiffusion system (Crutchfield, 1988), and secondly with the known abundance of feedback pathways within the brain. The large number of geometric patterns that can be produced by such physical video feedback systems, such as logarithmic spirals and phyllotaxis, lends some credence to this novel hypothesis. However, what could constitute a neural visual feedback mechanism for geometric hallucinations is still unknown. It seems plausible that somehow the abundance of feedback loops could play this role, especially when considering that at some point, during an altered state of mind, geometric hallucinations may be combined, modulated by, and reinterpreted with hallucinations from memory and iconic imagery that must be generated elsewhere within the cerebral cortex (de Araujo et al., 2012). Another pathway that could be involved is through the superior colliculus (SC)a major region of the vertebrate midbrain that forms an important part of the visual system, and which is considered to play a critical role in the development of blind sight (Takaura, Yoshida, & Isa, 2011). The mutual interaction between the primary visual cortex and this subcortical neural structure, which is also involved in visual processing, could open new possibilities for explaining geometric hallucinations in the absence of external visual stimuli. In addition, the SCs role in multisensory integration (Hall & Moschovakis, 2004) might help to explain how visual hallucinations become integrated with hallucinations in other modalities, including synesthesia-like effects.

3.6 The role of neural feedback in visual processing

Complex and varied experimental results on the function of feedback pathways have been found, but it is generally considered that higher-level cortical regions are involved in large-scale perceptual integration which combine with the fine-scale resolution provided by the earlier visual regions (Sillito, Cudeiro, & Jones, 2006). Feedback is involved in modulating perception; such as figure ground separation, suppressing areas of low information to generate a sparse codification of the visual signal, coordination and tuning of visual feature detectors, and attention (Williams et al., 2008). Backprojecting pathways are therefore thought to facilitate, inhibit, and synchronize neural activity (Przybyszewski, 1998). Under the influence of drugs, or in an altered state of mind induced by other means, these pathways could all be involved in moving the visual system

3.7 Beyond structural isomorphism

This overview of the putative neural mechanisms underlying geometric hallucinations points to the key components being the architecture in V1 supporting a neural Turing mechanism along with the retinotopic mapping, the ubiquity of spiral and wave dynamics within the brain, and widespread and long-ranging feedback connections. But despite these insights, the

Froese et al. full range of reported geometric hallucinations has yet to be accommodated by a single mathematical neural network model. The incorporation of additional physiological detail may increase the range of geometric phenomena that can be modeled, especially long-range feedback connectivity. We must consider the neural projections into and out of V1, such as its direct connection back to the LGN, to the other 30 or so visual processing regions, and the rest of the brain. Visual hallucinations can be induced in a wide number of ways, but all models make use of a single drug parameter that affects the entire system. Yet how does the experienced pattern differ depending on the route taken to such a state of mind? Dietrich (2003) proposes that most altered states of mind (drug induced or otherwise) involve a transient decrease in prefrontal cortex activity, and that the difference between types of experience arises from how the induction method affects the various prefrontal neural circuits. Connecting this function to models of geometric hallucinations and to the types of patterns specific to certain altered states could make for some interesting predictions. We can also wonder at what type of widespread effects the prefrontal cortex can have on the whole visual cortex and the rest of the brain, not just in V1. It also remains to be seen whether Turing instabilities in the rest of the brain, including the peripheral nervous system, play a role in generating geometric hallucinations. Indeed, because of the generality of this dynamical phenomenon, its potential role may not be limited to electrical interactions between neurons, either. We know that Turing patterns are crucial for a number of other biological processes, including in the chemical domain (Goodwin, 2001). Therefore, more realistic models of geometric hallucinations might want to include details of the physical and chemical processes of the central and peripheral nervous system, e.g., reaction and diffusion of various neurotransmitters, vitreous liquids inside the eye, oily residues on the surface of the cornea, etc., which could give rise to Turing patterns under disinhibited conditions, and thus influence the outcome of visual processing. As a simple proof of concept, it has been shown that the dynamics of a GrayScott reactiondiffusion system (see patterns in Figure 3) can be used in order to control a mobile robot, giving rise to shape discrimination and memory (Dale & Husbands, 2010). But as we incorporate these additional features into the models, it may also turn out that the hypothesis of a strict structural isomorphism between patterns of neural activity in V1 and geometric hallucinations will become less tenable. As a case in point, the function of the inverse retinotopic map can be called into question when considering that it implies that many of the reported geometric hallucinations have to appear extrafoveally. Accordingly, rather than looking for a direct one-to-one structural mapping between patterns of

207 neural activity and patterns of hallucinatory experience, it may be more productive to consider other functionally relevant properties of such neural patterns than their precise spatiotemporal arrangement alone. For example, all modeled neural patterns show properties of robustness and self-sustainability. In the models of Ermentrout and Cowan (1979) and Bressloff et al. (2001), patterns appear as the stable configurations of the system after bifurcating from a low activity state. The neural model of Rulkov et al. (2004) displays self-sustained wave dynamics that can spontaneously emerge from low-level random neural activity. As already mentioned, in the model of Fohlmeister et al. (1995), once spiral waves emerge they cannot be destroyed, even when increasing a drug parameter into the range where other geometric phenomena usually appear. This case of apparent irreversibility is probably the strongest example of robustness and selfsustainability in the surveyed models, and it is suggestive as a biological explanation for the fact that spirals are one of the most prevalent motives in prehistoric art. How could this property of robust self-maintenance relate to the hallucinatory experiences? A striking feature that differentiates geometric hallucinations from other visual experiences is that they are generated intrinsically when the subject has been decoupled from its environment. The patterns form irrespective of our lifetime learned memories. Indeed, they could be considered internally directed perceptual experiences, since if the proposed models hold true, they are directly formed from the actual biological structure of the visual system. We are said to have a strange subjective experience of looking into oneself, where the patterns we see directly expose the underlying operation of our brains. At the same time, the robustness of the underlying neural patterns also reinforces this decoupling from the environment. In the next section, we argue that under some conditions such a neural mechanism of decoupling could be adaptive. Humans are adept at not only understanding patterns from external sensory input, but importantly, we have a powerful faculty for active generationbe it for purposes such as prediction, imagination or playing.

4 Steps toward a new approach to the origins of symbolic material culture

One promising approach for systematically evaluating the relationship between brain functioning and human experience is known as neurophenomenology, a method that was proposed by Varela (1996) as part of the enactive approach, and which has subsequently been elaborated by others (e.g., Petitmengin, Navarro, & Le Van Quyen, 2007; Thompson, Lutz, & Cosmelli, 2005). The basic idea of neurophenomenology is to relate the

208 empirical data of neuroscience and verbal descriptions of first-person experience in a mutually informing and enriching manner. In this way both third-person and first-person aspects of a phenomenon are taken into account, but without privileging one domain over the other. While the aim is to find a formal model at which the dynamics of the two domains can be integrated, this does not entail a necessity of structural isomorphism. Instead, there is a mutual braiding of structural constraints, which links the activity of the two domains into one unfolding process (Varela, 1999). In addition, because perception is conceived of as a relational process of sense-making that is extended over the entire brainbodyenvironment as a whole, the externalinternal distinction becomes relativized (Lenay & Steiner, 2010). That is, the perceptual experience of an environment as being external is itself an outcome of sense-making, and therefore becomes potentially reversible and dissolvable as the conditions of sense-making are modified under unusual conditions. In the case of geometric hallucinations, it could therefore be that we are in fact dealing with an abnormally oriented perceptual process, where the object of perception is an aspect of the nervous system itself. This is surely a strange hypothesis, but the consideration of perceptual mediation opens up a new perspective on the problem of the mismatch between the impoverished neural patterns and the rich geometric experiences. A similar mismatch has been observed to hold between the impoverished retinal input and our rich visual experience of the world (Noe , 2002). Accordingly, on this alternative view, the main shortcoming of the current neural network models is not an insufficiently detailed structural isomorphism, but rather a failure to take the enriching effects of sense-making into account.4

Adaptive Behavior 21(3) This consideration places a significant constraint on our explanation from the side of cultural practices, because the artists must have valued their altered visual experiences. Some researchers have tried to address this constraint by appealing to social practices. For example, Lewis-Williams (2002) argued that the value of the experiences derives from their value as a tool to enforce social hierarchies by restricting access to elite individuals via prohibition. But this only shifts the original problem without resolving it, because we must then explain why people would value these experiences as worthy of restriction to elites. One promising alternative is to assume that when these visual patterns are seen during altered states of consciousness they are directly experienced as highly charged with significance. In other words, the patterns are directly perceived as somehow meaningful and thereby offer themselves as salient motifs for use in rituals. This possibility seems to accord with verbal reports about the range of emotional effects that can be elicited by a variety of non-ordinary experiences (e.g., Dobkin de Rios & Janiger, 2003; Huxley, 1956; Shanon, 2002; Strassman, 2001). This appeal to an intrinsic value places constraints on the underlying neural dynamics. We therefore need to discuss how value could be realized in neurobiology, and then we determine whether this neurological mechanism is compatible with the neurological mechanism underlying the experience of the most common forms of geometric phenomena.

4.2 Sketches of an enactive neurobiological account of value

In the history of cognitive science, the question of value has been a profoundly puzzling one. The computational theory of mind holds that cognition is rule-based manipulation of abstract symbols, as epitomized by the digital computer program. However, this logical syntax leaves the concrete meaning of the symbols unaccounted for; they are simply empty placeholders. This problem of meaning has resulted in major stumbling blocks in theory of mind, as well as in the practice of AI and robotics, such as the symbol grounding problem and the frame problem. The enactive critique of these theoretical and practical problems is centered on the grounding of value in the self-maintenance of autonomous forms of biological organization (for a review, see Froese & Ziemke, 2009). Most importantly, the enactive approach tries to provide a viable alternative by defining value in operational terms. A good starting point is the work by Di Paolo, Rohde and De Jaegher (2010, p. 48): We propose to define value as the extent to which a situation affects the viability of a self-sustaining and precarious network of processes that generates an identity. There

4.1 Neurophenomenology of value

How might neurophenomenology address the correspondence between spatial patterns in the cultural, experiential and neurobiological domains? The first thing to note is that the artistic patterns under consideration form an extremely limited set compared to the total set of all possible patterns that could have been created. This universal selection bias is something that must be explained, but which has so far remained mysterious. It is not sufficient to note the correlation with visual experiences during altered states of consciousness. Even if we accept that internally oriented visual perception is possible during altered states of consciousness, this does not in itself explain why those perceptual experiences are more frequently expressed artistically. Why should these particular patterns be preferred over other kinds of ordinary and non-ordinary visual experiences? This selective bias must therefore be accounted for in terms of the value these patterns have had for the artists who made them.

Froese et al. are a number of entangled concepts implicated in this definition, which cannot be unpacked in detail here. We briefly summarize the main ideas. To start with, the networks identity is not a reified entity; it can be defined as operational closure, i.e., the mutual dependence of the processes of the network on each other for their existence. This co-dependence provides them with an internal relationship that integrates them with each other and that does not depend on a unifying distinction by an external observer. And this co-dependence also makes the networks identity precarious, since its existence is not externally guaranteed but must be selfsustained by its own ongoing activity. In other words, the network is only viable within certain boundary conditions. The value of an event is therefore defined in relation to its impact on these conditions, i.e., whether it contributes to the self-sustaining of the identity or not. The paradigmatic example of this kind of valuegenerating system is the minimal living system (Weber & Varela, 2002). But the same kind of organization can also be found in the nervous system, where we have the temporary emergence of cell assemblies (Varela, Lachaux, Rodriguez, & Martinerie, 2001). The firing patterns of some neurons (or of collections of neurons) can become entrained with each other, thereby constituting a precarious, self-sustaining network of processes, which in this case is a neurocognitive identity (Varela, 1997). This kind of neural pattern is therefore a suitable candidate for a process of value-generation. Now if we consider the models of the neural mechanisms underlying some of the common geometric hallucinations, we also find the same kind of self-sustaining cell assembly. Moreover, it is likely that the assemblys robustness is enhanced considerably, because activity in the visual brain regions during altered states is generally assumed to be disinhibited. In addition, under these conditions the sensory input from external events no longer has the power to modulate the cell assemblys boundary conditions, thus enhancing its viability. In the most extreme cases, this kind of autonomous, selfsustaining network can manifest itself in a complete loss of sensorimotor interaction. For example, it has long been known that seeing flickering light at certain frequencies can induce neural entrainment and geometrical hallucinations (for recent work in this area, see, e.g., Billock & Tsou, 2007), but when the brain as a whole becomes too entrained with those frequencies it can cause an epileptic fit. According to the enactive account of value, we would expect these kinds of geometric experiences to be experienced as significant, and the non-ordinary experiences encountered before or during epileptic fits as intensely meaningful.5 Verbal reports of people with epilepsy confirm this hypothesis. Sometimes the seizures leave people unconscious, but when they do have experiences during the prodromal phase and at the onset of the epileptic fit

209 they are complex in many ways (Le Van Quyen, 2010; Petitmengin et al., 2007). What is particularly remarkable is the reported vividness or intensity of these experiences. The content of the experiences depends on the person and the precise brain region of the epileptic discharge, but commonalities can be identified. Consider, for example the case of emotional experience:
Special emotional auras consist of an attack of anguish and terror that suddenly takes over the consciousness with such intensity that the subject has the impression she or he is losing control of the situation, which will have a terrible end, perhaps madness or even death. In other emotional auras, the experience consists of a sudden state of joy with no apparent cause, and it takes over the consciousness passively for a few short moments, filling it with awe and strangeness. (Le Van Quyen, 2010, pp. 247248)

Clearly, neurophenomenology is currently not advanced enough to explain the particular content of these experiences, but it does successfully explain why the experiences are characterized by such an intensely felt significance. Generalizing this insight to the experiences of geometric hallucinations, we can predict that these, too, are felt with an increased intensity of significance, as long as self-sustaining processes of neural activity generate them. In other words, intrinsic value is built into the experience of some altered states of consciousness from the start. The ambivalence of the nonordinary experiences reported by Le Van Quyen, i.e., whether they are experienced as good or bad, is not specific to altered states of consciousness induced by epilepsy, but is characteristic of altered states more generally (Huxley, 1956). Perhaps this inherent polarity of consciousness alteration, which Huxley famously contrasted as scenarios of heaven and hell, has to do with the fact that the organism is a whole meshwork of self-sustaining processes (Varela, 1991), and the intrinsic value of a specific self-sustaining neural process has to be balanced against the needs of the other selfsustaining processes of the organism as a whole.

4.3 Ideas for an enactive neurobiological account of imagination

The problem of explaining the origins of the human imagination is related to the transformation of basic sense-making of the here and now to the enaction of the absent and imaginary (Froese, 2012). We speculate that one manner of realizing this transformation is to mediate the organisms sensorimotor interactions in such a way that the autonomous dynamics of the nervous system temporarily become decoupled from the environment. Abstract forms of cognition are likely to be facilitated during such highly introverted states of consciousness, which would also help to explain their prehistoric origin.

210 As we have already indicated, a basic dynamic neural mechanism for realizing a temporary mode of environmental decoupling can be derived from the existence of self-sustaining processes of neural activity. These neural processes might not always be easily identifiable in terms of spatial patterns, especially when we are considering parts of the nervous system that do not exhibit such a regular spatial embedding as the retina or V1.6 Nevertheless, it is helpful to think of these neural processes in terms of an analogy with chemical reactiondiffusion systems, which are formally closely related to the activationinhibition models of the visual system (H. R. Wilson, 1999, pp. 267268). Chemical reactiondiffusion systems are a well-researched area, both formally and empirically, and therefore provide a useful source of inspiration. For example, it has been demonstrated that chemical Turing patterns, and especially spiral patterns, are highly robust against the influence of external perturbations such as attacks by parasitic chemical species (Boerlijst & Hogeweg, 1991). Accordingly, it seems plausible that the disinhibition of the visual cortex during altered states, perhaps via topdown projections from prefrontal cortex to the pathway from retina to V1 regions, could lead to Turing patterns of neural activity in V1 that are robust against perturbations from visual sensory input, thereby effectively shutting down external visual influence (Butler et al., 2012). The analogy with chemical reactiondiffusion systems also suggests that some of these neural Turing patterns may even exhibit their own self-individuated spatiotemporal identity, as well as the ability to engage in adaptive interactions with other emergent processes and structures. These kinds of emergent behaviors have been demonstrated to occur in chemical reaction diffusion systems, including self-moving droplets with tail structures (Froese, Virgo, & Ikegami, in press). It is therefore tempting to speculate that some neural Turing patterns can adapt to, and perhaps even make sense of, the brains own activity and structures during altered states of consciousness. Could this help to explain the commonly reported transition from purely geometric hallucinations to iconic hallucinations that appear to exhibit their own agency (Lewis-Williams, 2002)? Future work in systems neuroscience needs to verify if any of these possibilities are borne out by the empirical evidence. Nevertheless, an enactive approach is compatible with current models of the neural mechanisms underlying geometric hallucinations, and already has more explanatory power than merely assuming a direct structural isomorphism between visual experience and brain activity: (1) it allows for the possibility that the patterns of non-ordinary visual experience are more varied and richly detailed than the underlying neural patterns; (2) it can help to explain why these patterns were taken to be highly significant; and (3) it suggests that there may have been a functional link between the ritualistic alteration of consciousness and

Adaptive Behavior 21(3) the origin of abstract cognition and symbolic practices in terms of temporary environmental decoupling of neural processing.

5 Conclusion
We have criticized some of the current theories that have been proposed to explain the origins of symbolic practices, especially as exemplified by Lewis-Williams account. We have argued that Lewis-Williams insistence on social conflict and class struggle as the primary driving force behind changes in the first symbolic practices is not compelling for a variety of reasons. Nevertheless, his claim that the cultivation of altered states of consciousness was involved in the origins of the first symbolic material cultures may still turn out to be correct, although for alternative reasons which he fails to consider. On the basis of a review of current mathematical models of neural mechanisms underlying geometric hallucinations, we proposed a neurophenomenological approach that emphasizes the enactive account of sense-making and value. In particular, by considering some kinds of altered state of consciousness as largely internally mediated forms of perceptual sense-making, we can better account for the richness of non-ordinary experiences. Furthermore, the value generated by selfsustaining cell assemblies may help to explain the selective bias of the first artists for the kinds of geometric patterns that are typically experienced during these altered states of consciousness. The enactive approach also supports the idea that such altered states could have significantly influenced the operation of the nervous system, especially by temporarily decoupling the autonomous activity of the brain from the usual environmental influences. This switch from immediate sensorimotor sense-making, which is normally directed toward the external here and now, to a more internally mediated, decoupled sense-making of mental and bodily structures could thereby have facilitated the creation and diversification of abstract cognition and symbolic practices. Notes
1. Some researchers prefer the hypothesis that these prehistoric patterns are highly abstract representations of normal perceptual experience rather than direct expressions of abnormal hallucinatory experience. For example, spiral patterns could be derived from swirling water flows, swirling winds, winding stems of vines and winding snakes (Takaki & Ueda, 2007, p. 133). However, this hypothesis presupposes a process of geometric abstraction. Moreover, other common kinds of prehistoric patterns are not so readily found in nature. Most importantly, even if all of the patterns could be abstracted from environmental phenomena, the problem of explaining the cross-cultural significance of these patterns would remain the same.

Froese et al.
2. Note that the engraved pattern shown in Figure 1 can be interpreted as a lattice pattern made of triangles (i.e., one of Klu vers form constants). 3. In this respect it is suggestive that even the hallucinatory experience of small spots of light (phosphenes), induced by applying transcranial magnetic stimulation (TMS) over the human visual cortex, apparently cannot be reduced to a local neural mechanism, but arises only after more widespread recurrent processing (P. C. J. Taylor et al., 2010). A distribution across multiple neural substrates also helps to explain how such simple geometric hallucinations can be transformed into full-blown iconic hallucinations, for example under certain kinds of pathological conditions (J.-P. Taylor et al., 2011). 4. This shift in explanatory approach helps us to resolve one of the major outstanding puzzles of the current neuroscience of geometric hallucinations, namely the apparent mismatch between the resolution of neural Turing patterns and visually experienced patterns. However, it is not a solution to the explanatory gap of the mind-body problem as such. We thank Julien Hubert for helping us to clarify this important difference. 5. Strictly speaking, the enactive approach only predicts that there is value for the self-sustaining neural process itself. More remains to be said about how this value also becomes significant from the perspective of the selfsustaining organism as a whole. But considering that neural processes are constitutive of the whole organism (Varela, 1991), it seems reasonable to assume that their values are also constitutive of the values of the whole organism. Nevertheless, future work on an enactive theory of the personal self is needed in order to better connect these distinct levels of description. Thanks to Nathaniel Virgo for highlighting this issue. 6. Thanks to Chris Buckley for pressing this point during several discussions.

211
Bressloff, P. C., Cowan, J. D., Golubitsky, M., Thomas, P. J., & Wiener, M. C. (2001). Geometric visual hallucinations, Euclidean symmetry and the functional architecture of striate cortex. Philosophical Transactions of the Royal Society B: Biological Sciences, 356, 299330. Bressloff, P. C., Cowan, J. D., Golubitsky, M., Thomas, P. J., & Wiener, M. C. (2002). What geometric visual hallucinations tell us about the visual cortex. Neural Computation, 14, 473491. Butler, T. C., Benayoun, M., Wallace, E., von Drongelen, W., Goldenfeld, N., & Cowan, J. D. (2012). Evolutionary constraints on visual cortex architecture from the dynamics of hallucinations. Proceedings of the National Academy of Sciences of the USA, 109(2), 606609. Clottes, J., & Lewis-Williams, D. (1998). The shamans of prehistory: Trance and magic in the painted caves. New York, NY: Harry N. Abrams. Crutchfield, J. P. (1984). Space-time dynamics in video feedback. Physica D: Nonlinear Phenomena, 10(12), 229245. Crutchfield, J. P. (1988). Spatio-temporal complexity in nonlinear image processing. IEEE Transactions on Circuits and Systems, 35(7), 770780. Dale, K., & Husbands, P. (2010). The evolution of reaction diffusion controllers for minimally cognitive agents. Artificial Life, 16, 119. Darwin, C. (1871). The descent of man, and selection in relation to sex. London, UK: John Murray. de Araujo, D. B., Ribeiro, S., Cecchi, G. A., Carvalho, F. M., Sanchez, T. A., Pinto, J. P., de Martinis, B. S., Crippa, J. A., Hallak, J. E. C., & Santos, A. C. (2012). Seeing with the eyes shut: Neural basis of enhanced imagery following Ayahuasca ingestion. Human Brain Mapping, 33, 25502560. De Jaegher, H., & Froese, T. (2009). On the role of social interaction in individual agency. Adaptive Behavior, 17(5), 444460. Deacon, T. W. (1997). The symbolic species: The co-evolution of language and the brain. New York, NY: W. W. Norton & Company, Inc. Di Paolo, E. A., Rohde, M., & De Jaegher, H. (2010). Horizons for the enactive mind: Values, social interaction, and play. In J. Stewart, O. Gapenne & E. A. Di Paolo (Eds.), Enaction: Toward a new paradigm for cognitive science (pp. 3387). Cambridge, MA: The MIT Press. Dietrich, A. (2003). Functional neuroanatomy of altered states of consciousness: The transient hypofrontality hypothesis. Consciousness and Cognition, 12, 231156. Dobkin de Rios, M., & Janiger, O. (2003). LSD, Spirituality, and the creative process. Rochester, VT: Park Street Press. Ermentrout, G. B., & Cowan, J. D. (1979). A mathematical theory of visual hallucination patterns. Biological Cybernetics, 34, 137150. Fohlmeister, C., Gerstner, W., Ritz, R., & van Hemmen, J. L. (1995). Spontaneous excitations in the visual cortex: Stripes, spirals, rings, and collective bursts. Neural Computation, 7, 905914. Froese, T. (2012). From adaptive behavior to human cognition: A review of Enaction. Adaptive Behavior, 20(3), 209221. Froese, T. (in press). Altered states and the prehistoric ritualization of the modern human mind. In C. Adams, A. Waldstein, B. Sessa, D. Luke & D. King (Eds.), Breaking

Acknowledgments
We thank Chris Buckley, daboo, Julien Hubert, Guillaume Dumas, Etienne Roesch, Nathaniel Virgo and one anonymous reviewer for their thoughtful comments on an earlier draft of this paper. Many ideas of this paper took shape during several seminars of the Ikegami Laboratory, and they have benefited from generous audience feedback.

Funding
Tom Froese and Alexander Woodward were each financially supported by a Grant-in-Aid of the Japanese Society for the Promotion of Science (JSPS) during the initial phase of this work.

References
Billock, V. A., & Tsou, B. H. (2007). Neural interactions between flicker-induced self-organized visual hallucinations and physical stimuli. Proceedings of the National Academy of Sciences of the USA, 104(20), 84908495. Blackmore, S. J., & Troscianko, T. S. (1989). The physiology of the tunnel. Journal of Near-Death Studies, 8(1), 1528. Boerlijst, M. C., & Hogeweg, P. (1991). Spiral wave structure in pre-biotic evolution: Hypercycles stable against parasites. Physica D: Nonlinear Phenomena, 48, 1728.

212
convention: Essays on psychedelic consciousness. London, UK: Strange Attractor Press. Froese, T., & Di Paolo, E. A. (2009). Sociality and the lifemind continuity thesis. Phenomenology and the Cognitive Sciences, 8(4), 439463. Froese, T., Virgo, N., & Ikegami, T. (in press). Motility at the origin of life: Its characterization and a model. Artificial Life. Froese, T., & Ziemke, T. (2009). Enactive artificial intelligence: Investigating the systemic organization of life and mind. Artificial Intelligence, 173(34), 366500. Ganis, G., Thompson, W. L., & Kosslyn, S. M. (2005). Brain areas underlying visual mental imagery and visual perception: an fMRI study. Cognitive Brain Research, 20, 226241. Goodwin, B. (2001). How the leopard changed its spots: The evolution of complexity. Princeton, NJ: Princeton University Press. Habu, J. (2004). Ancient Jomon of Japan. Cambridge, UK: Cambridge University Press. Hall, W. C., & Moschovakis, A. (Eds.). (2004). The superior colliculus: New approaches for studying sensorimotor integration. New York, NY: CRC Press. Harvey, I. (2008). Misrepresentations. In S. Bullock, J. Noble, R. A. Watson & M. A. Bedau (Eds.), Artificial Life XI: Proceedings of the Eleventh International Conference on the Simulation and Synthesis of Living Systems (pp. 227233). Cambridge, MA: The MIT Press. Henshilwood, C. S., & dErrico, F. (2011a). Middle Stone Age engravings and their significance to the debate on the emergence of symbolic material culture. In C. S. Henshilwood & F. dErrico (Eds.), Homo symbolicus: The dawn of language, imagination and spirituality (pp. 7596). Amsterdam, Netherlands: John Benjamins Publishing Company. Henshilwood, C. S., & dErrico, F. (Eds.). (2011b). Homo symbolicus: The dawn of language, imagination and spirituality. Amsterdam, Netherlands: John Benjamins Publishing Company. Henshilwood, C. S., dErrico, F., & Watts, I. (2009). Engraved ochres from the Middle Stone Age levels at Blombos Cave, South Africa. Journal of Human Evolution, 57, 2747. Huang, X., Xu, W., Liang, J., Takagaki, K., Gao, X., & Wu, J.-y. (2010). Spiral wave dynamics in neocortex. Neuron, 68, 978990. Hutchins, E. (2010). Enaction, imagination, and insight. In J. Stewart, O. Gapenne & E. A. Di Paolo (Eds.), Enaction: Toward a new paradigm for cognitive science (pp. 425450). Cambridge, MA: The MIT Press. Huxley, A. (1956). The doors of perception and heaven and hell. New York, NY: Harper & Brothers. Ji, D., & Wilson, M. A. (2007). Coordinated memory replay in the visual cortex and hippocampus during sleep. Nature Neuroscience, 10(1), 100107. Klu ver, H. (1967). Mescal and the mechanisms of hallucination. Chicago, IL: The University of Chicago Press. Kobayashi, T. (2004). Jomon reflections: Forager life and culture in the prehistoric Japanese archipelago. Oxford, UK: Oxbow Books. Kondo, S., & Miura, T. (2010). Reactiondiffusion model as a framework for understanding biological pattern formation. Science, 329, 16161620. Kosslyn, S. M., & Thompson, W. L. (2003). When is early visual cortex activated during visual mental imager? Psychological Bulletin, 129(5), 723746.

Adaptive Behavior 21(3)

Laughlin, C. D., & Throop, C. J. (2009). Husserlian meditations and anthropological reflections: Toward a cultural neurophenomenology of experience and reality. Anthropology of Consciousness, 20(2), 130170. Le Van Quyen, M. (2010). Neurodynamics and phenomenology in mutual enlightenment: The example of the epileptic aura. In J. Stewart, O. Gapenne & E. A. Di Paolo (Eds.), Enaction: Toward a new paradigm for cognitive science (pp. 245266). Cambridge, MA: The MIT Press. Lenay, C., & Steiner, P. (2010). Beyond the internalism/ externalism debate: The constitution of the space of perception. Consciousness and Cognition, 19, 938952. Lewis-Williams, D. (2002). The mind in the cave: Consciousness and the origins of art. London, UK: Thames & Hudson. Lewis-Williams, D., & Dowson, T. A. (1988). The signs of all times: Entoptic phenomena in Upper Paleolithic art. Current Anthropology, 29(2), 201245. Lewis-Williams, D., & Pearce, D. (2005). Inside the Neolithic mind: Consciousness, cosmos and the realm of the gods. London, UK: Thames & Hudson. McGann, M. (2007). Enactive theorists do it on purpose: Toward an enactive account of goals and goal-directedness. Phenomenology and the Cognitive Sciences, 6(4), 463483. Milton, J. G., Chu, P. H., & Cowan, J. D. (1993). Spiral waves in integrate-and-fire neural networks. In S. J. Hanson, J. D. Cowan & C. L. Giles (Eds.), Advances in neural information processing systems 5 (pp. 10011006). San Mateo, CA: Morgan Kaufmann Publishers. Miyashita, Y. (1995). How the brain creates imagery: Projection to primary visual cortex. Science, 268(5218), 17191720. Newell, A., & Simon, H. A. (1976). Computer science as empirical enquiry: Symbols and search. Communications of the Association for Computing Machinery, 19(3), 113126. Noe , A. (2002). Is the visual world a grand illusion? Journal of Consciousness Studies, 9(56), 112. Pearson, J. E. (1993). Complex patterns in a simple system. Science, 261(5118), 189192. Petitmengin, C., Navarro, V., & Le Van Quyen, M. (2007). Anticipating seizure: Pre-reflective experience at the center of neuro-phenomenology. Consciousness and Cognition, 16(3), 746764. Przybyszewski, A. W. (1998). Vision: Does top-down processing help us to see? Current Biology, 8, R135E139. Rulkov, N. F., Timofeev, I., & Bazhenov, M. (2004). Oscillations in large-scale cortical networks: Map-based model. Journal of Computational Neuroscience, 17, 203223. Shanon, B. (2002). The antipodes of the mind: Charting the phenomenology of the Ayahuasca experience. Oxford, UK: Oxford University Press. Sillito, A. M., Cudeiro, J., & Jones, H. E. (2006). Always returning: feedback and sensory processing in visual cortex and thalamus. Trends in Neurosciences, 29(6), 307316. Stewart, J. (2010). Foundational issues in enaction as a paradigm for cognitive science: From the origin of life to consciousness and writing. In J. Stewart, O. Gapenne & E. A. Di Paolo (Eds.), Enaction: Toward a new paradigm for cognitive science (pp. 131). Cambridge, MA: The MIT Press. Strassman, R. J. (2001). DMT: The spirit molecule. Rochester, VT: Park Street Press. Takaki, R., & Ueda, N. (2007). Analysis of spiral curves in traditional cultures. Forma, 22, 133139.

Froese et al.
Takaura, K., Yoshida, M., & Isa, T. (2011). Neural substrate of spatial memory in the superior colliculus after damage to the primary visual cortex. The Journal of Neuroscience, 31(11), 42334241. Taylor, J.-P., Firbank, M., Barnett, N., Pearce, S., Livingstone, A., Mosimann, U., Eyre, J., McKeith, I. G., & OBrien, J. T. (2011). Visual hallucinations in dementia with Lewy bodies: Transcranial magnetic stimulation study. The British Journal of Psychiatry, 199, 492500. Taylor, P. C. J., Walsh, V., & Eimer, M. (2010). The neural signature of phosphene perception. Human Brain Mapping, 31, 14081417. Thompson, E. (2007). Look again: Phenomenology and mental imagery. Phenomenology and the Cognitive Sciences, 6, 137170. Thompson, E., Lutz, A., & Cosmelli, D. (2005). Neurophenomenology: An introduction for neurophilosophers. In A. Brook & K. Akins (Eds.), Cognition and the brain: The philosophy and neuroscience movement (pp. 4097). New York, NY: Cambridge University Press. Turing, A. M. (1952). The chemical basis of morphogenesis. Philosophical Transactions of the Royal Society B: Biological Sciences, 237(641), 3742. Van Horn, S. C., Eris xir, A., & Sherman, S. M. (2000). Relative distribution of synapses in the A-laminae of the lateral geniculate nucleus of the cat. The Journal of Comparative Neurology, 416, 509520. Varela, F. J. (1991). Organism: A meshwork of selfless selves. In A. I. Tauber (Ed.), Organism and the origins of self (pp. 79107). Dordrecht, Netherlands: Kluwer Academic Publishers. Varela, F. J. (1996). Neurophenomenology: A methodological remedy for the hard problem. Journal of Consciousness Studies, 3(4), 330349.

213
Varela, F. J. (1997). Patterns of life: Intertwining identity and cognition. Brain and Cognition, 34(1), 7287. Varela, F. J. (1999). The specious present: A neurophenomenology of time consciousness. In J. Petitot, F. J. Varela, B. Pachoud & J.-M. Roy (Eds.), Naturalizing phenomenology: Issues in contemporary phenomenology and cognitive science (pp. 266317). Stanford, CA: Stanford University Press. Varela, F. J., Lachaux, J.-P., Rodriguez, E., & Martinerie, J. (2001). The Brainweb: Phase synchronization and largescale integration. Nature Reviews Neuroscience, 2, 229239. Virgo, N. (2011). Thermodynamics and the structure of living systems. D.Phil. Dissertation, University of Sussex, Brighton, UK. Weber, A., & Varela, F. J. (2002). Life after Kant: Natural purposes and the autopoietic foundations of biological individuality. Phenomenology and the Cognitive Sciences, 1, 97125. Williams, M. A., Baker, C. I., Op de Beeck, H. P., Shim, W. M., Dang, S., Triantafyllou, C., & Kanwisher, N. (2008). Feedback of visual object information to foveal retinotopic cortex. Nature Neuroscience, 11(12), 14391445. Wilson, H. R. (1999). Spikes, decisions and actions: The dynamical foundations of neuroscience. New York, NY: Oxford University Press. Wilson, H. R., & Cowan, J. D. (1973). A mathematical theory of the functional dynamics of cortical and thalamic nervous tissue. Kybernetik, 13, 5580. Wilson, S. W. (1983). On the retino-cortical mapping. International Journal of Man-Machine Studies, 18, 361389. Winkelman, M. (2002). Shamanism and cognitive evolution. Cambridge Archaeological Journal, 12(1), 71101.

About the Authors Tom Froese received an MEng in computer science and cybernetics from the University of Reading, UK (2004). He then obtained a DPhil. in cognitive science from the University of Sussex, UK (2010). He was a postdoctoral research fellow at the neurodynamics and consciousness laboratory of the Sackler centre for consciousness science, University of Sussex, UK (2010). Froese then became a JSPS postdoctoral fellow at the Ikegami laboratory of the department of general systems studies, University of Tokyo, Japan (20102012). Currently, he is a postdoctoral researcher at the self-organizing systems laboratory of the Instituto de ticas Aplicadas y en Sistemas, Universidad Nacional Auto noma Investigaciones en Matema xico, Mexico (20122013). His research is focused on developing enactive approaches de Me to understanding the biology, phenomenology, and dynamics of life, mind, and sociality. Alexander Woodward received his PhD in computer science from the University of Auckland, New Zealand in 2009. He is currently a postdoctoral fellow at the Ikegami laboratory of the University of Tokyo (since 2010). His research into computer vision has led him to ask questions about the nature of visual perception in living systems. Current research focuses on computer vision, neural networks and reservoir computing, subjective time in the brain, scientific computing on the GPU, and a maximalist approach to artificial life.

214

Adaptive Behavior 21(3) Takashi Ikegami received his doctorate in physics from the University of Tokyo. His research interest is to build and study artificial life systems ranging from chemical droplets and evolutionary robotics to Web dynamics. Some of these results have been published in Life emerges in motion (Seido, 2007) and also The Sandwich theory of life (Kodansha, 2012). Takashi Ikegami gave the keynote address at the 20th anniversary of the Artificial Life conference in Winchester, UK. He is also a member of the editorial boards of Artificial Life, Adaptive Behavior and BioSystems.