An Analogy Between the Neurophysiology of Thought and the Polypedal Locomotion of an Octopus

Abstract

This article presents theoretical interpretations of the neurological events that underlie two fundamental brain processes: 1) the selection and activation of neocortical assemblies; and, 2) the generation and subsequent embellishment of mental imagery as it passes between sensory and association areas. These two processes together are taken to be responsible for the transitions between mammalian brain states where mental representations change gradually due to cycles of activation, deactivation and coactivation of cortical assemblies. Gradual changes in a pool of simultaneously coactivated neurons occur as cortical assemblies that continue to receive sufficient activation energy are maintained; assemblies that receive reduced energy are released from activation and new assemblies that are tuned so as to receive sufficient energy from the current constellation of coactivates are converged upon, recruited and incorporated into the remaining amalgam of active assemblies from the previous cycle. This neurophysiological process is presented here as analogous to the locomotive behavior of a many-armed octopus that grabs and releases footholds as it pulls itself from place to place. Specifically, the process whereby neural assemblies fluctuate spatio-temporally is taken to be analogous to the nonlinear stride of an octopus that plants the majority of its arms temporarily, while actively repositioning arms that have let go of their footholds. This analogy uniquely describes a system where certain nodes are conserved through time as others are actively repositioned.

Because individual cortical assemblies are tuned to code for a discrete element of long-term memory, when one is coactivated with other assemblies, the individual elements can be united into composite, mental representations. These representations fluctuate back and forth between early, bottom-up sensory cortex (where they are metric and topographic) and late, top-down association cortex (where they are abstract and conceptual) on the order of brain oscillations. Sensory areas and association areas continually stimulate each other into building interpretations of the others outputs resulting in a conversational interchange with minimized informational redundancy. The fact that some assemblies within association areas are conserved (i.e. the octopus arms remain planted), during these reciprocal oscillations, is taken to account for the continuity found between successive brain states. The longer assemblies in association areas can be continuously activated - over a series of states - the longer they can influence sequences of bottom-up imagery in a sustained and consistent way allowing modeling, planning and working memory in general. The result is a stream of consciousness where each thought is slightly different from the ones preceding, as newly relevant assemblies are added and the least relevant ones are removed. Highly intelligent mammals have a larger selection of available assemblies, can coactivate a larger number of assemblies together simultaneously and have the capacity to prolong activation in relevant assemblies for extended periods.

Keywords: association cortex, cortical assembly, perception, primary sensory cortex, systems neuroscience, top-down

Bullet Pointed Introduction:

1) How can the thought process, the experience of consciousness and the functionality of working memory be described in terms of brain events? 2) How do features from long-term memory combine together to create original thoughts? 3) What brain events take place when we move from thought to thought?

- Models such as Baars global workspace theory, Baddeleys theory of working memory, Damasios convergence-divergence paradigm and Edelmans theories of reentrance and neural Darwinism have done much to lend perspective and insight into the mechanics of thought progression.

The Octopus Arms Represent Cortical Modules

- Unlike subcortical areas, strictly one-to-one, linear activation is probably rare in the cortex. Rather, cortical modules coactivate together to spread the activation energy necessary to recruit the next set of modules that will be coactivated with the remaining modules from the previous cycle. This is highly analogous to the seafloor walking behavior seen in the octopus because these animals plant the majority of their arms on the ocean floor while repositioning arms in the back towards the front.

- The longer modules in association areas can be continuously activated, over a series of thoughts, the longer they can influence sequences of bottom-up imagery in a sustained and consistent way allowing modeling, planning and working memory in general.

- It may be correct to say that someone with a working memory deficit has fewer modules to select from, fewer modules bound during coactivation and modules that cannot maintain their activation for as long as they do in other people. Because modules work cooperatively, having fewer modules of less duration will reduce network searching power and specificity.

- Fluid intelligence derives from the number and duration of modules whereas crystallized intelligence derives from the connections between modules and their tuning properties.

- Some modules are retained as coactivates even after a number of thought cycles. This happens when ones thoughts transition and change but hold a common element constant and is often due to potentiation by the PFC.

- Sometimes modules are not conserved from thought to thought and the octopus drops most of them all at once. This happens when one abandons a train of thought and quickly reorients to a new, salient, perhaps emotionally laden stimulus.

- Another component of this analogy is the idea that the octopus will fall if it loses its grip on a sufficient number of branches. Since the body of the octopus is analogous to consciousness this is appropriate because brains become unconscious once coactivation and the accompanying binding (especially in the frontal and parietal fields) is sufficiently diminished.

- The number of octopus arms is set and this represents our fixed, innate capacity for working memory. Even though the number of chunks that can be held in working memory, 7 plus or minus 2, coincidentally coincides with the number of arms an octopus has (8), this is not a reliable indication of the number of modules that can be coactivated in the cortex. It is not clear: 1) how many modules are normally coactivated at once, 2) what organization of neurons or their assemblies constitutes a module or 3) exactly how rhythmic binding plays a role in module coactivation.

- The hippocampus has an ability to detect a set of cortical coactivations and reactivate the rest of the assemblies that were coincident with these in the past in a process called pattern completion. The hippocampus then, has the ability to guide the legs of the octopus toward historically coactivated patterns.

- The PFC helps the octopus control the spatio-temporal layout of coactivations by prolonging activation in modules that correspond to salient concerns in the environment in order to allow the uninterrupted persistence of such features in the imagery, enabling forethought, planning and modeling.

Interactions Between Association Modules and Sensory Imagery

- The process of thinking involves cycles between internally generated imagery and ones higher-order perception of it- reciprocal priming between bottom-up sensory and top-down association areas. This process is similar to what it would be like to watch a television program that one could control with their ideas, conceptions and conceptualizations.

- The modules in association areas cooperatively converge on sensory modules which combine this unique set of higher-order coactivations into a composite, lower-order, feature-based image. The ability to do this is fine-tuned during early development, makes use of the vast architecture of recurrent (back-propagating) pathways and is accomplished rapidly, based on prior probabilities.

- Importantly, things that follow from our conceptualizations, but that we did not expect to see are routinely rendered in imagery. For instance our sensory areas might pull up the imagery specified by association areas, but elaborate on it with closely associated but unforeseeable embellishments. Thus, the cyclical oscillations of information between sensory and association areas allow them to learn from each other and allow them to integrate their knowledge like two people in a conversation.

- This suggests that one can only perceive the relationship between two abstract ideas if one already has implicit information in the sensory cortex about how to co-represent them in an image.

- As an association node is converged on from nodes upstream, it becomes active and in turn divergently activates the downstream sensory nodes that ordinarily converge upon it. The resultant sensory imagery is then either superimposed over objects perceived in the environment or combined with other features in the minds eye. Thus contemplative thought takes place on the same Cartesian stage that sensory experience takes place on. Sense, remembered or reactivated is the substrate of thought.

- The octopus arms (modules) in posterior sensory areas move faster, from module to module, but are can hold a larger number of simultaneous representations. This accounts for the transience of sensory memory but also for its greater capacity (echoic and iconic memory decay faster but hold a larger number of chunks).

- A module becomes implicit, and its features become unconscious, when it is no longer needed - during coactivation with its normal coactivates - to recruit another particular module.

- Some modules within association areas remain activated because they are restimulated by the imagery that they contributed to. Other association modules that are not restimulated by this imagery deactivate as the projection neurons associated with them stop firing as rapidly. In other words, we create imagery in our minds, but we dont pay attention to every aspect of the imagery just like we dont notice every aspect of the perceptions that we create of our environment. Thus this analogy of the TV you control with your mind is closely related to the octopus analogy because the elements of the imagery that are attended to drive the placements of the octopus free arms.

Introduction

Some of the important questions in cognitive neuroscience today include: 1) How can the thought process, the feeling of consciousness and the functionality of working memory be described in terms of brain events? 2) How do elemental features (fragments) of long-term memories combine together to create original thoughts? 3) What brain events take place when we move from thought to thought? 4) What is the nature of communication between association and sensory areas? 5) How does the human brain permit such sophisticated working memory relative to other animals? Without being able to tie together all of the neurological, psychological and philosophical loose ends necessary to answer these questions directly and comprehensively, this paper will attempt to address them using novel approaches based on simple analogy.

There are currently many theories of consciousness and working memory. Some do a fine job of tying together a large number of relevant phenomena into a cohesive picture. Models such as Baars global workspace theory (Baars, 1997; 2002), Baddeleys theory of working memory (Baddeley, 2000; 2007), Damasios convergence-divergence paradigm (Damasio, 1989; Meyer & Damasio, 2009), Edelmans theories of reentrance and neural Darwinism (Edelman, 1987; 2006), Edelman and Tononis conceptualization of a functional cluster or dynamic core (2001) and Grossberg and Carpenters adaptive resonance theory (Carpenter and Grossberg, 2003) have done much to lend perspective and insight into the mechanics of attention, consciousness, perception and working memory. Despite much progress, most scientists report that current theory is unsatisfying because it cannot yet bridge the gaps between epiphenomenal consciousness, brain processes and neural connectionism (Chalmers, 1995; Chalmers, 2010; Shear, 1997). The present work intends to present a narrative that bridges the biological/psychological divide of neurocognitive processing while remaining consistent with most other theory in this literature. It approaches the questions posed above from what is perhaps a much neglected perspective, from the transitions between brain states. By looking closer at how the brain transitions between individual brain states it is hoped that light will be shed on the progression of thoughts. Thoughts are taken here to be units of cognition where an individual thought corresponds to a full cycle between bottom-up and top-down processing. In my opinion, this means that a representation must go from association cortex to sensory cortex and back (and/or perhaps vice versa, from sensory, to association and back). After elaborating on this perspective it subsequently attempts to integrate the customary approaches of consciousness, attention and working memory.

Like several other models of mind-brain processes, this model views cognition as a system responsible for using representations in long-term memory to guide goal-directed processing (Moscovich, 1992). The model is consistent with connectionism and parallel distributed processing in that it conceptualizes mental representations as being built from decentralized, interconnected networks of modular clusters of neurons that have multiple inputs and outputs (Gurney, 2009). Like other models, it envisions these clusters as neural assemblies in the cerebral cortex and assumes that individual assemblies represent elementary features of long-term memories. These assemblies are capable of being activated and spreading their activation energy to closely associated assemblies. The assemblies, like the neurons that compose them function as coincidence detectors (Fujji et al., 1998). When these assemblies are activated at the same time, the features that they code for are amalgamated into composite mental imagery in whatever way prior probability and previous experience dictate.

Here, an assembly is taken to be comprised of a number of similarly-tuned neurons that are Hebbian-bound to create a discrete, functional unit (Lansner, 2009). Assemblies are often thought to correspond to cortical minicolumns of cells with similar receptive fields that map onto a specific, elementary, perceptual feature (or fragment of such a feature). How these discrete fields of cells interact has been mysterious since the columnar organization of the cerebral cortex was first delineated by Vernon Mountcastle (1978). Groups of these vertical or columnar cortical assemblies can be Hebbian-bound to create ensembles that stretch laterally through the cortex and that are comprised of cells with disparate but related receptive fields. It is thought that these assemblies and ensembles coactivate and bind together through reentrant connections in the corticocortial and thalamocortical system creating stable constellations of activity on the order of tens or hundreds of milliseconds (Crick and Koch, 2003). Moreover, information passing through these connections tends to flow hierarchically between cortical sensory areas and cortical associations areas on the order of brain oscillations (Klimesch, Freunberger, Sauseng, 2010).

This model is also consistent with the consolidation hypothesis which states that memory is stored in the same areas that allow active, real-time function (Moscovitch et al., 2007). Furthermore, it assumes that imagining a particular sensory image largely activates the same neural networks that are involved in actually perceiving the imagery in the environment (Crick and Koch, 2003). Thus, sensory cortex acts as an active canvas for either the environment, or the imagination (association cortices) to paint on. As stated, the model addresses aspects of attention, working memory and consciousness but is largely about how neural networks, built of assemblies, combine their features to create cycles of mental imagery and thus thought. Again, the model proposed will attempt to accomplish two main objectives: 1) offer an analogy between the locomotive behavior of an octopus, and the coactivation behavior or neural assemblies; and, 2) offer explanations for how imagery is generated in sensory cortex, subsequently interpreted by association cortex, and then is sent back to sensory cortex, resulting in the creation of updated imagery.

The Hypothesis

The general pattern of activation in cortical assemblies is addressed by the present analogy, which involves a many-armed octopus grabbing and releasing footholds (cortical assemblies) as it pulls itself from place to place. This is meant to illustrate that the thought process involves the simultaneous coactivation of several cortical assemblies at a time (multiple footholds held by an octopus) as well as the activation of new assemblies and the deactivation of others. Unlike subcortical areas, strictly one-to-one, linear activation is probably rare in the cortex. Also, unlike subcortical areas, information processing in the cortex is not compartmentalized into individual nuclei that are relatively isolated from processing occuring elsewhere. Rather, cortical assemblies coactivate together to spread the activation energy necessary to recruit, or converge upon, the next set of assemblies that will be coactivated with the remaining assemblies from the previous cycle. This pooling of activity might be called multiassociative. To differentiate from multiassociative neural networks we can abandon the prefix multi and call this brand of associativity among neurons, polyassociative. The concept of polyassociativity makes it so that Hebbian bound neural circuits can no longer be thought of as indivisible units, instead the unit must be a small group of cells with the same or nearly the same receptive field and we will refer to these as assemblies. An assembly is released from coactivation when it no longer receives sufficient activation energy from its inputs i.e. its relevance to the processing demands diminishes. When an assembly is deactivated, the perceptual or conceptual element corresponding to it becomes deactivated and can no longer impact present experience. Whatever new assembly is introduced will inform the present sum of coactivates in a unique and informative way.

The assemblies that are coactivated, at any one time, sum their component features together to create imagery and this occurs in both sensory and association areas. Imagery changes plastically as assemblies that continue to be useful are maintained, assemblies that are rendered less useful are released from activation, and assemblies that are newly recognized as useful are activated and incorporated into the remaining amalgamation of useful coactivations. This is analogous to the seafloor walking behavior of an octopus that plants the majority of its arms temporarily and actively repositions arms that lie behind it, toward the front, in the direction of its movement. The fact that some assemblies are conserved, during these transitions, provides the physical basis for the continuity of thought. The assemblies that are conserved reside mostly in association areas whereas assemblies in early sensory areas are activated much more transiently. This is as if the arms in back of the octopus (corresponding to posterior sensory cortices) move much more quickly than the forelimbs (which find firm footholds in anterior association cortices).

It is not clear if individual assemblies represent consciously perceptible constructs. In order to escape this uncertainty, we will refer to groups of assemblies that do this as ensembles. This is a helpful distinction that will lead us to assume that when a psychologically perceptible construct is displaced from working memory (and the ensemble associated with it loses activation), each of the individual assemblies that constitute the ensemble also stop firing. In other words, assemblies (which are neural units) can be bound by learning to constitute an ensemble (which is a psychological unit). In the present analogy, each octopus arm corresponds to an active ensemble, the suction cups on one arm can be taken to correspond to the assemblies that make up the ensemble, and the grains of sand under each suction cup on an arm represent cortical neurons. This analogy is apt because, like the grains of sand on the sea floor, cortical neurons do not move, only the pattern of activation the octopus - moves. The same neuron can probably contribute its representational content to different assemblies just as assemblies can be combined in different ways to create different ensembles and ensembles combined differentially to create thoughts. For the most part, humans have some ability to guide thought on the level of the selection of ensembles but have almost no control over the selection of assemblies or neurons because the brain is

Feedforward activation from bottom-up sensory areas selects among potential assemblies in association cortex. Conversely, feedback activation from top-down association areas modulates and drives the selection of assemblies in early sensory cortex. Because the imagery in early visual areas is tied retinotopically to the visual field, this interaction is like watching a TV screen that one can influence with ideas and conceptualizations. Similarly, the early auditory area can be equated with a tape recorder that can be recorded upon and played back. The association cortices, including the prefrontal cortex, do not equate to a homunculus, but rather an integrated extension of sensory areas that allow multimodality, temporal continuity, motor instruction and advanced multiassociative guidance for the creation of early sensory imagery. Thus, the oscillation of information between abstract association areas and veridical sensory areas allows these two types of cortex to converse, learning from the others unique brand of content, like two people in a conversation. The following sections will elaborate on this interchange pointing to the consequences of the octopedal patterning and relating other models to this one. Before this is done, the next section will consider how the present model has changed over time.

The Old Analogy: An Ape Swinging From Cortical Branches

I have played with a few different schemas for representing the workings of the mind, realizing that a good model would have to satisfy certain criteria. The original allegory that I used was of an ape swinging from branch to branch (hand over hand) where each branch represented a group of neural assemblies in the cortex that coded for a new thought. This early analogy tried to convey the idea that the branches, or concepts, from the immediate past determine what branches will be held in the future. It was meant to convey that we move from one thought to the neurologically nearest, most appropriate thought in a deterministic manner. The next chosen branch in the cortical canopy represented to me, the probabilistically most likely association given the persons current thought, given their past and given the structure of their memory. This model of the thinking process is limited because it is linear. I came to understand that localized groups of neural assemblies cannot code for complete thoughts, images, or memories and that assemblies are not activated and then deactivated, one set at a time, in sequence. This caricature of memory was limited, vague and failed to capture the polyassociative and oscillatory nature of thought.

Instead, mental activities must involve the simultaneous coactivation of numerous assemblies from multiple locations each that code for different features of long-term memories. I have since replaced this branch-swinging ape with a walking octopus. I changed the animal because the octopus has more arms and can simultaneously possess more footholds. The many arms introduced important and divergent features to the locomotive behavior that I think creates instructive analogies when superimposed on the neural processes of thought. For example, the octopus analogy has the advantage of demonstrating how more intelligent animals are able to have more complicated thoughts they are able to coactivate a larger number of assemblies that each have the potential to be active for a longer period.

This routine of assembly activation and deactivation is very similar to polypedal locomotion or movement in animals with many legs. It is not much like the locomotion of an insect such as a millipede or a centipede though because these animals move their legs in stereotypical, repetitive ways where the placement of each leg is not actively influenced by the placements of other legs or of the qualities of the footholds. The pattern of activations in the brain is more like the polypedal locomotion of an octopus that is seafloor walking because it is asymmetrical, dynamic and the placement of the next legs are influenced by the octopus stance, posture and the characteristics of the footholds themselves. Most importantly, this model can accommodate nonlinear aspects of thinking. One neural assembly does not activate the next in sequence. Several assemblies are coactivated together and they pool their activation energy to determine which assemblies will be activated next.

The New Analogy: An Octopus Walking on a Cortical Sea Floor

Cortical assemblies are used only when they are needed. However, many can be found working together, in concert, at any one time. In the analogy, the octopus footholds represent simultaneously activated assemblies. But as thoughts do not hold still, the octopus is constantly repositioning its arms. The octopus is continually letting go of some of these so that it has the resources to grab new ones

Even though assemblies are constantly being deactivated, we take many assemblies with us through time. If we did not do this, we would continually forget what we were just thinking, and we could not have a progressive train of thought. Because some assemblies and their associates can remain active for a number of seconds we are able to transition between closely-related thoughts. Thus, mental continuity has a neural basis: When an assembly receives sufficient activation energy from its inputs it will fire at its targets (its projective field), often firing recurrently at the sources that targeted it (its receptive field), until the configuration of assemblies changes to the point where it no longer receives sufficient activation from either the bottom-up or top-down assemblies that converge on it.

Some assemblies can probably be retained even after the transitions between a number of thoughts. This happens when your thoughts cycle and change but hold a common element or theme constant. When we attempt to solve a novel and complex problem we try to keep the majority of our octopus arms firmly planted so that we can keep the problem set in mind. Some aspects of creative thinking or free association, on the other hand, might involve strategically pivoting around a smaller set of continually active assemblies and using these to determine the next set of coactivates.

It is not always the case that the majority of assemblies are conserved from one thought to another. Most assemblies can be dropped or abandoned at the same time, i.e. when they become a lower priority. This readily happens when we are exposed to a new, salient, perhaps emotionally laden, stimulus. When this occurs, the octopus jumps, taking all of its arms with it, and reorients to the new stimulus and its accompanying set of features. Such a jump would constitute a disruption of mental continuity. So clearly mental continuity can be viewed on a continuum where a high proportion of assemblies are conserved between brain states on one end of the continuum and a low proportion are conserved on the other end. Such a disruption might occur due to a stimulus in the environment, or from an internally generated stimulus. Evolution has probably programmed the octopus to jump and reposition its arms quickly in order to respond to important sensory stimuli, so that mammals react to them with all of their cognitive resources. Mental continuity is less easily disrupted in humans than it is in other mammals, although perhaps more easily disrupted in people with habituation deficits. Attention and distraction must be intimately related to the temporal conservation of cortical assemblies. In fact, the extent of attention deficit and distractibility should be inversely related to the neurological capacity to conserve assemblies in association areas from second to second. Creating an operational definition for this proportion and ways to measure it (on a scale of neurons per millisecond) may prove informative.

Another component of this analogy is the idea that the octopus will topple if it loses its grip on a sufficient number of assemblies. This makes the body of the octopus analogous to consciousness because brains become unconscious once coactivation (especially in the frontal and parietal fields) is sufficiently diminished. Assemblies in early sensory areas are often active during unconscious states, but assemblies in association areas are less active and out of sync with those in sensory areas. Thus anterior-posterior balance and coordination are important for our allegorical octopus. Table 1 below summarizes some of the important concepts related thus far.

Definitions of Terms

Working Memory: The Coordination of 7 Plus or Minus 2 Arms

The number of available octopus arms in this analogy is relatively stable and this represents our fixed, innate capacity for working memory. Even though the number of chunks (psychologically perceptible units of perception and meaning) that can be held in working memory, 7 plus or minus 2, coincidentally coincides with the number of arms an octopus has (8), this is not necessarily a reliable indication of the number of ensembles that can be coactivated in association cortex. Also, the relationship between chunks and ensembles is unclear. It is clear though that the octopus has a relatively set number of arms and that generally, in order to bring a new ensemble into the train of thought it must first let go of some other ensemble. Surely the number of primable assemblies/ensembles differs from area to area and from task to task but it probably remains relatively constant within tasks. Relative to Baddeleys model of working memory, active assemblies in visual areas can be equated with the visuospatial sketchpad, assemblies in auditory and language areas equated with the phonological loop and assemblies in the prefrontal cortex (PFC) equated with the central executive.

It may be correct to say that someone with a working memory deficit (because of mental retardation, intoxication, psychosis or dementia) has fewer of these allegorical octopus arms. Someone with a general mental deficit, temporary or chronic, cannot bring as many assemblies with them through time and cannot coactivate as many of them together simultaneously. Because assemblies work cooperatively, having fewer assemblies of less duration will reduce network searching power and specificity. In other words, the larger the number of active assemblies, the more vivid and precise the mental imagery created in the minds eye; whereas, fewer means less accurate, less precise perceptions. Similarly, the longer certain assemblies are activated, the more new thoughts are informed by recent thinking. Smart people, endowed with large working memories, can activate many assemblies that can remain active for a long time and this allows the priming of meaningful nodes that allows the person to be perceptive and keen.

A deficient working memory (or one lower on the phylogenetic scale) may have the following characteristics: 1) fewer assemblies to select from (based on reduced cortical surface and the resulting smaller number of unique receptive fields), 2) fewer assemblies bound during coactivation, and 3) individual assemblies cannot maintain their activation for very long. The extended activation of assemblies changes the learning process as well. Prolonged coactivation causes synaptic changes to reflect higher-order, temporally-structured representations; altering the weights of receptive fields, tuning ensembles and their assemblies to be able to respond to even more complex features in the future. Thus, fluid intelligence derives from the number and duration of assemblies whereas crystallized intelligence derives from the connections between assemblies and their tuning properties.

The Selection of New Assemblies: Where The Octopus Sends Its Free Arm

The way that new assemblies are primed in this model is consistent with connectionism and spreading activation theory. In spreading activation theory associative networks can be searched by labeling a set of source nodes which spread their activation energy to closely associated nodes. The propagation of activation follows weighted links to other nodes. Several alternate paths through these links can reach a specific end node. When enough of these alternate links reach the same node this node is likely to be activated. In the brain, these links are thought to represent connections between neurons and the weights are found in the synapses. In other words, when a particular number of cortical areas are active they will converge on particular other areas according to the weights found in the network. Many areas will be converged upon weakly, but a few areas will be converged upon enough to increase the frequency of action potential firing which in turn increases metabolic activity.

A cortical cell has many inputs (in the form of synapses) and a large number of these inputs must be actively sending it neurotransmitters (creating excitatory post synaptic potentials) in order for the cells firing rate to increase appreciably. Cells in the cortex are open to being activated maximally, but they remain at a resting level until just the right complement of inputs takes place. Once the cell becomes activated sufficiently, the cell will send outputs to other cells within its projective field. On an assembly level, this happens when the pyramidal projection neurons associated with the assembly fire out rapidly to other assemblies and interneurons in the cortex. Inhibitory interneurons will determine what neurons will be inhibited from contributing to their assemblies and pyramidal neurons determine what assemblies will become active. I believe the outputs of pyramidal projection neurons in sensory areas broadcast mental imagery by converging on downstream assemblies. Projection neurons in association areas, far downstream, recurrently direct and modulates this imagery through top-down, backpropagating retroactivation.

Because only a small minority of cells become highly active at any one time, many memories and much imagery remains dormant. Only the precisely appropriate cells are chosen and this creates the specificity of thought. It takes just the right combination of inputs from other primed areas (anywhere in the cortex and even subcortical areas) to pull these cells into the octopus embrace. Input from four associated assemblies may not activate a new assembly without the contribution of a fifth assembliess EPSPs. From a psychological viewpoint, we may not be able to recall a special memory unless just the right combination of related memories are coactivated. By the same account, sometimes arbitrary assemblies are activated (and superfluous memories are recalled) because a large, but not large enough, group of assemblies were coactivated to activate it. In other words, the precise combination of active assemblies determine together (by spreading activation) which assemblies will be activated next. This reasoning is consistent with the conclusion of psychologists that the thinking process works associatively. On millisecond time scales, we do not pick and choose our thoughts, they are chosen for us based on how the currently active assemblies interact with the associative network. In this way thinking appears haphazard because ultimately the way new assemblies are selected is not overseen by any rational process. There is no hidden logic or computation aside from that found in the epigenetic structure of memory due to past learning. Thus, the fewer the number of assemblies coactivated to choose the next assemblies, the more random, mercurial, deterministic and unguided this process seems.

This model causes cortical thought to appear as a closed system due to the fact that we can only select preexisting assemblies for coactivation. We are working with a limited number of ways to represent things given that we have a limited number of unique receptive fields. However, even though our thoughts are forced to choose from currently existing assemblies, these recycled assemblies can be combined in unique ways to create imagery that has never been created before. Does recombination of old parts truly create something new? Well, at least it seems like it does when the brain throws in multimodal assemblies and temporally persistent assemblies, when there are a large number of assemblies being combined at once and when the imagery composed of the combined parts actively retunes the existing assemblies.

We may not be able to remember someones name, or face, until just the right combination of other assemblies are primed. Their face might correspond to a constellation of neural assemblies (facial features) that have fired together in the past, creating an ensemble. Once assemblies that are key in recognition are activated they converge on and activate the network of assemblies responsible for holding the name (in an auditory area) or a face (in the fusiform face area) to bring back the memory. Again, such localized, well-connected assemblies are analagous to the suction cups on an octopus arm. Thus, the octopus coactivates certain localized neural assemblies using the suckers on its arm. The sum of these assemblies (which code for elementary features that alone do not capture attention) produces an ensemble (a trait which can capture psychological attention) that is then summed with other active ensembles to create the complete thought.

The allegory of an interactive TV: How the octopus guides mental imagery

This section will attempt to explain how coactivated assemblies combine to create mental imagery, how features of images come together and how subsequent images are chosen. To do this we will consider hierarchical brain processing. To a certain extent, the cortex is organized hierarchically. The primary visual area (V1) (along with the thalamus) processes the stream of information sent to it from the retina allowing it to distinguish dots. The secondary visual area (V2) puts these dots together to form lines- the edges and curves that make up the visual scenery. Even higher-order, downstream areas put these lines and curves together to discern more complex visual features amounting to the recognition of movement, color and even objects, scenes and faces. Young children probably use a larger proportion of their visual areas to decompose visual scenery into its component parts in an attempt to ensure that they are not misidentifying visual objects. Older individuals probably retune many of these neurons to work more closely with higher-order concerns due to the fact that they are experienced at identifying the objects in their environment. Now we turn to how the cycling of information between lower-order, bottom-up areas and higher-order, top-down areas is accomplished and worked into our conceptual schema of the octopus. In other words, this section will address how imagery is created, interpreted and then modified.

We will now describe the brain events involved in a cycle between internally generated imagery and our higher-order perception of it- reciprocal priming between sensory and association areas. This process is similar to what it would be like to watch a television program that we could control with our ideas, conceptions and conceptualizations. The early sensory areas constitute the TV in this analogy because, unlike association areas, they map imagery that is spatially or retinotopically bound to the visual field. Early visual areas take inputs from higher association areas and, given these specifications, paint metric imagery. Importantly, things that follow from our conceptualizations, but that we did not expect to see are routinely rendered in imagery. For instance, our sensory areas might pull up the imagery specified by association areas, but elaborate on it with closely associated but unforeseeable embellishments. Thus, the cyclical oscillations of information between sensory and association areas allow them to learn from each other and allow them to integrate their knowledge like two people in a conversation.

This analogy of the TV you control with your mind, represents the process whereby higher-order (top-down) association areas interpret and then control sensory imagery in early (bottom-up) sensory areas. The higher-order association areas influence this imagery through their outputs to sensory areas and then receive feedback from the sensory areas (almost as if they actively watch the imagery that is created). At this time some of the association areas remain activated because they are restimulated by the imagery. Other association areas that are not restimulated or those that are stimulated by interneurons might deactivate. We create imagery in our minds, but we dont pay attention to every aspect of the imagery just like we dont notice every aspect of the perceptions that we create of our environment. When an association assembly contributes to mental sensory imagery in a way that it not noticed by association areas in the next cycle, that assembly is not reactivated and does not contribute to subsequent imagery until reactivated. The sensory imagery that is generated is not seen as a whole as we might like to think, in fact, many features of the mental imagery that is created probably remains preattentive. We might like to think that our mental footwork is smooth and streamlined but instead we respond in scattered, discrete ways that feel continuous because of the fluidity of our movements and language and because of the vividness of our sensory systems. Thus this analogy of the TV you control with your mind can be combined with the octopus analogy because the elements of the TV that are noticed drive the placements of the octopus free arms. Technically, the imagery in early association areas also involves assemblies and thus octopus arms as well.

This analogy of a TV that you control with your mind is open to harsh criticism from philosophers of mind because it evokes a homunculus (another thinking entity inside of the first entity). Of course, incorporating an irreducibly conscious entity into a model of consciousness is not helpful in explaining the nature of consciousness. Thus it is important to stress that the entity watching the TV set is actually many entities (individual assemblies) that have no ability to render geometrical, contour or line-based representations. Also, the TV and its observer can be thought of as continuous with each other where the observer can be thought of as an extreme form of the TV itself and similarly, the TV can be thought of an extreme form of the observer. Also, not every oscillation goes from the posterior pole to the anterior, meaning that occipital areas can probably give feedback without having to activate V1.

Antonio Damasio has proposed that early sensory cortices construct image space and that association cortices construct dispositional space that do not hold any imagery themselves. In my opinion, association areas do hold imagery. They hold imagery of higher-order concepts that are disoriented from spatial mapping or retinotopic coordinates. In other words, visual sensory areas hold spatially-oriented optical imagery, auditory sensory areas hold temporally-oriented sound imagery and association areas hold abstracted, multimodal conceptually-oriented imagery that is relatively free of reality-imposed, unimodal, spatio-temporal constraints. Contrary to Damasios notion that association areas only guide the construction of imagery, I think that association areas hold true imagery in the sense that they can invoke high-level perceptions of things that the person can become conscious of. However, consistent with Damasio, this model agrees that association areas do not hold all of the information held in the early sensory cortices that converge upon them. The firing of a grandmother neuron in the anterior temporal cortex alone does not produce a conscious visual depiction of a grandmother in the minds eye. In other words, you cannot visualize a line-bound image of your grandmother without early visual cortex. However, without early visual cortex, you can still hold associative, conceptual imagery about her as long as your anterior temporal cortex is intact. In this sense, the imagery (its construction, and manipulation) is truly in the process of bottom-up to top-down reciprocal oscillations. The thing that is the most unclear about this is how the dynamic pathways between assemblies interact with one another and the extent to which historically unrelated assemblies cooperate to drive new images. It is also unclear if these assemblies

Even though the brain is a very dark place, early visual areas have been tuned by experience to represent variations in brightness, color and form so that when they become active, from either retinal or top-down inputs, they display vibrant imagery. It may be correct to say that only sensory areas constitute this TV, and that association areas hold higher-order concepts that drive image construction within the TV. On the contrary, the association areas may just be extentions of the TV because they, like the sensory areas, are turned on every time a specific visual concept is invoked so there is no reason to assume that they dont hold their own form of imagery. Arguing that association areas do not hold imagery is tenuous and is akin to saying that only V1 holds imagery and that V2 simply modulates the imagery. Thus cortical areas responsible for visual processing - from the posterior occipital pole to the dorsolateral prefrontal cortex lie together on a continuum with retinotopic imagery on one side and abstract, conceptual imagery on the other.

The progression of thought involves oscillatory messaging between bottom-up sensory imagery and top-down interpretations of that imagery where none of the brain areas know exactly what they are going to invoke in the areas they are communicating with until they receive feedback. This is to say that the brain probably does not make any plans about how sensory areas will integrate the various association inputs, it simply does so reflexively based on prior probabilities. At first it seems that early sensory areas would have a difficult time integrating multiple concepts from association areas into a single, meaningful image. However, the ability to take incongruous elements and integrate them has become the sensory cortexs speciality as, over developmental time, it has been trained to do this with environmental perceptions.

The process of reconstituting diverse association specifications into sensory imagery is probably identical in many ways to the way that sensory areas combine features of the sensory environment to create early sensory perceptions. When sensory areas create perceptions based on inputs from the retina they construct scenes by conjoining dissimilar elements into a cohesive interpretation based on what they have been rewarded for creating in the past. Sensory areas must do the same thing with inputs, not from the retina, but from the association areas to create imagined imagery. This suggests that one can only perceive the relationship between two abstract ideas if one already has implicit information in the sensory cortex about how to co-represent them in an image. If the person is missing instrumental conceptual knowledge in their sensory areas then these areas will not be able to create the image (although it is possible that association areas could manipulate a series of images to lead up to a final image where the important elements are depicted retrospectively). In turn, things that we hadnt considered are brought up on the screen, important things that help to guide our train of thought, things that only our unconscious visual memory system can conjure up. Again, our higher-order, abstract, more highly evolved associative memory systems inject their own unique take on things, that are still very much mechanical and deterministic, but more reflective of what we have done and learned, instead of what we have been passively exposed to.

The reciprocal activity allows sensory and association areas to learn from each other. Association areas might as well be saying to themselves: Well, it will be interesting to see how the visual system will combine this unique set of higher-order coactivations into a composite, lower-order, feature-based image. This sensory image does what association areas cannot do on their own take various components and integrate them into a visage that has ecological utility. The way that sensory areas integrate when they construct images is informed directly by reality as they have been tuned directly by real environmental inputs unlike associative areas which are tuned indirectly by reality due to the intervening effects of motivation, temporal delay and inference. The whole reason that these coactivations make a perception that is good is this early visual system has been exposed to so much sensorily, it has a huge repertoire, accumulated over time, of tricks and insights into how physical and psychological things work. The sensory activations potentiated by the sum of coactivated association areas determine the next group of association assemblies that will be activated- the placements of the octopus free arms.

What happens psychologically when multiple sensory assemblies are coactivated together to form an imagined perception? I believe that, when this happens, mental imagery is created of the sort that we are all accustomed to seeing in our minds eye. The cortical assemblies that we have been talking about, that can be primed by internal inputs, also correspond to brain areas that are primed by external inputs during sensory perception. In fact, early sensory assemblies were first tuned by external inputs- sensory experiences that we have been having since before birth. When these assemblies are activated by the environment we have a sensory perception - a rich and vivid experience that can involve any sense but the most vivid and highly processed are probably visual and auditory. As an association assembly is converged on from sensory assemblies upstream, it becomes active and in turn divergently reactivates the upstream sensory nodes that just converged on it and additionally may also activate the other nodes that ordinarily converge upon it. The resultant sensory imagery is then either superimposed over objects perceived in the environment (during perception) or combined with other features in the minds eye (during imagination).

When we think and imagine, we activate early perceptual networks. The PFC and other associative areas do not appreciably influence processing in V1 or the LGN of the thalamus (the earliest of visual processing areas) via recurrent projections, but can profoundly influence the activity in extrastriate visual areas. Thus contemplative thought takes place on the same Cartesian stage that sensory experience takes place on. To me there is far less mystery left about the origin of conscious imagery after this is taken into account. Sense, remembered or reactivated is the substrate of thought. The unique pattern of coactivations that are primed in the cortex by a particular external visual scene is perceived as a sensory image. An internally driven sensory image is a pattern of coactivations in sensory cortex that is not primed by information from the retina but information from higher-order association areas. The internally driven imagery does not activate sensory areas as much as true sensory experience does, and this probably accounts for why imaginary imagery is not as vivid as actual imagery. Internally driven imagery, on the other hand, probably activates higher-order, top-down, association areas more than does true sensory experience. Both externally and internally generated imagery have the capacity to activate higher order sensory areas. In other words, we can alternate sequentially from externally generated imagery to our higher-order perception of it; we can alternate reciprocally between internally generated imagery and our higher-order perception of it, or combinations of both.

Thinking about perceptions and reperceiving our thoughts about them

First we take a look at the TV. This idea can easily be generalized to other sensory areas, but for the moment, let us use vision as the example. The TV screen is our visual sensory cortex, composed of primary and secondary sensory areas- known to have their own very short-term memory called sensory memory. Sensory memory has been shown to hold more than working memories 7 plus or minus 2 chunks, although it does so very fleetingly (2.5 seconds for auditory sensory memory and 250 milliseconds for visual sensory memory). This means that seeing a vista activates many early visual assemblies but only for about a quarter of one second. Once the vista is no longer seen by the eyes the viewer has 250 milliseconds to choose which aspects of that sensory scene to keep in mental imagery because after this time the only aspects left will be those consciously attended to - those that progressed downstream to activate association assemblies. Importantly though, sensory input activates only a subset of the association assemblies that the octopus has access to, it primes assemblies in the early sensory areas - creating rich tapestries of lower order information. The sum of coactivations here determine which receptive fields of higher order (secondary sensory and association cortex) neurons will be activated. Mental imagery probably works in much the same way in the sense that the imagery holds more information than we can consciously attend to and that it fades within a quarter of a second if it is not bound to, or reactivated, by higher-order associations. In other words, there are many association assemblies watching the fast-paced TV but these will only speak up if they are adequately activated.

Importantly, this cycle demonstrates how Baddeleys visuospatial sketchpad and phonological loop operate. Psychologically this feels as if new concepts evoke apposite imagery which we then analyze and modify. Our sensory areas conjure up their best sensory representation of the concepts the higher areas conceived of and, given the specifications handed down to them, use receptive fields and prior probabilities to present this visage. This ability is probably fine-tuned during early visual development and probably makes use of the vast architecture of recurrent (back-propagating) pathways and is accomplished rapidly and automatically. The nervous system is wired in a way that almost all cortical to cortical connections can go both ways. The sensory imagery generated causes association areas to think of something else, something that could be a slight modification on what we saw last, or seemingly a paradigm shift away from it. The next configuration of octopus arms may seem wildly different, because it evokes a different sensory image, but unless emotions were evoked, it is likely that many of the high-order arms are still in place even when the early sensory imagery changes profoundly. So sequential images created on the TV may look very different but they are likely to be highly interrelated.

It is hard to say whether the bottom-up or top-down components are more conscious. It is easy to think that the conscious part of the octopus is the higher-order areas, but these areas are not where the most vivid imagery is. We could say that the unconscious part of the octopus is the lower-order areas - but if we are going to say these are unconscious then we have to specify that we nonetheless are conscious of them. Both the activity in lower-order and higher-order areas is due to coactivated assemblies so it is probably appropriate to think of the octopus of consciousness as the sum of assemblies regardless of whether they are abstract (higher-order) or literal (lower-order).

One might ask then, where and when does consciousness happen? Most neuroscientists agree that, when considering sensory information from the environment, lower-order, primary areas are activated, then it takes a quarter of a second to become aware of the information. This time is thought to correspond to the time it takes for these lower areas to activate the most important high-order area, the PFC. Once the PFC becomes aware we are thought to be conscious of the sensory stimulus. This makes some sense that associative, convergence areas such as the PFC, hippocampus and the angular gyrus must be activated for us to become aware of something. But these areas do not contain vivid sensory imagery, at least not the kind we associate with sight for instance. Because of this, we probably have to wait for the PFC and other high-order areas to contact sensory areas in order to experience our response to a stimulus. What does it mean to be conscious of something if we have not yet responded with imagery? Maybe consciousness is the ability to form new imagery from previous imagery. Then there is the question of how many times we have to reexperience our response to lower-order imagery for us to be aware of it. It seems that this happens as soon as assemblies associated with imagery related to self-awareness are activated.

We are conscious of the early sensory imagery, that is the experience that we remain in through time. But the association cortex is the buffer of coactivating concepts that actively chooses what mental imagery we see next. Only our lifes insight allow us to see this.

The octopus momentum keeps it moving inexorably, mechanistically and deterministically to the nearest associated concepts. The PFC and hippocampus though can intervene and place its arms on footholds in order to reflect a past experience (the hippocampus) or a past motivation (the PFC). Hippocampal pattern completion helps the octopus place a number of its arms on a constellation of assemblies that were activated at some point in the past. The PFC helps the octopus maintain its posture in the sense that it trains and controls the spatio-temporal layout of coactivations, it also pins a few arms down at a time to allow planning and modeling. They work together too. The PFC keeps several things online long enough so that when the hippocampus takes a snap shot, the snap shot contains several different activates.

The Role of the Hippocampus

Not all activation derives from intracortical priming. The hippocampus plays an integral role in activating multiple assemblies that correspond to past (contextual and episodic) patterns of coactivation. The hippocampus has an ability to detect a subset of nodes that were coactivated together in the past and reactivate the rest of the assemblies that were previously coincident with these in a process called pattern completion. The hippocampus then, has the ability to guide the legs of the octopus toward historically coactivated patterns. It can do this for recalled events from 30 seconds to two years in the past. This means that even though the hippocampus can help to guide the selection of features that combine to create working memory and long-term memory, it does not hold them itself. In this sense, the hippocampus can be thought of as the memory that the octopus has for where its arms have been in the past. The hippocampus repositions the octopus arms and in so doing allows long-term memory to effect the progression of thoughts. This process allows a different kind of continuity. The pattern of coactivations among assemblies in the cortex allows continuity on the order of seconds but the long-term memory of the hippocampus allows mental continuity on the order of minutes to years.

Research on individuals with anterograde amnesia due to hippocampal damage evince that without the hippocampus thoughts tend to lose long-term continuity because past patterns of assemblies cannot be recalled. The anterograde amnesia caused by bilateral hippocampal lesion makes it so that conceptions tend to remain in working memory only very briefly, usually between 7 and 30 seconds. This suggests that the individual assemblies in the association cortices, that we have been considering, may have the ability to remain active for about as long without intervention from the hippocampus. However, many assemblies are probably much more fleeting as they may be inhibited or may lose their activation inputs from other areas on quicker time scales, probably on the order of rythmic brain oscillations (4 to 100 per second). It is important to note that, as discussed earlier, assemblies in sensory cortex are probably much more fleeting than those in association cortex, corresponding to the differences between sensory memory and working memory. The octopus arms (assemblies) in posterior sensory areas move faster, from assembly to assembly, but can hold a larger number of simultaneous representations. This accounts for the transience of sensory memory but also for its greater capacity (echoic and iconic memory decay faster but hold a larger number of chunks).

The octopus without the hippocampus can only model semantic sensory knowledge, but with the hippocampus can consider contextual and episodic knowledge. The formation of long-term memories in the cortex necessitates a number of learning cycles whereas the hippocampus seems to bind discrete assemblies much more quickly. Most researchers assume that memories in the hippocampus migrate out towards the cortex in the span of a couple of years. Perhaps this happens when the hippocampus repeatedly pulls up the same constellation of assemblies causing them to become associated within cortical maps. Perhaps after two years, the hippocampus will attempt to complete a pattern and find that the cortical assemblies that it activates no longer code for the same elementary features that they once did. The weights of cortical maps are changing everyday with experience, and this must make hippocampus-driven coactivations fuzzier with the progression of time.

The Role of the PFC

Cortical assemblies that are activated are deactivated rather quickly unless another area extends its primed state. The PFC can extend activation in many cortical areas for several seconds. The PFC helps the octopus control the spatio-temporal layout of coactivations by prolonging activation in assemblies that correspond to salient concerns in the environment in order to allow the uninterrupted persistence of such features in the imagery, enabling forethought, planning and modeling. In general terms, the PFC is tuned to remember what elements of a situation should be preserved through time in order to properly inform subsequent coactivations.

There is not a clear distinction between assemblies that are in working memory and nodes that recently fell out of working memory and are held in nonhippocampal-dependent short term memory. It is known that primed nodes recently used in working memory transfer to short term memory where they are much more likely to be reprimed in the near future. These nodes are still metabolically active, slightly above baseline; however, they might not contribute much to imagery. Primes that lay in the octopus trail are easily reprimed (even if subliminally so) and the real question is do these recently primed nodes contribute to imagery? If they do then they are also part of the octopus. There is a lot of grey area between currently contributing primes and recent primes that might contribute a little. I think that recent primes do contribute in subtle ways to mental imagery because they often remain metabolically active for some time, contributing their activation energy to the network. This implies that our octopus is heavily influenced by the footholds that it let go of recently.

The brain cells corresponding to recently primed nodes are more metabolically active and have increased firing rates. They may be only slightly more active than if they hadnt recently been primed, this might make them hard to consciously recall, but it might make them recognizable or at least more likely to be pulled into awareness again later. But even if they do not come all the way back and never end up contributing the feature that they code for to the global composite image, they probably have the capacity to alter the weights of nodes to be pulled up in the near future. This may happen in the case where a columns neurons continue to fire to other nearby columns laterally within cortical layers without causing the column to fire its projection neurons. Such neurons may alter the weights of different nodes in different ways. If recently primed node A excites primed node B, B may be more likely to come into the picture later on if Bs other associates are primed. Conversely, A may eventually inhibit B if their general relationship is inhibitory. In other words, recent primes blur a line between, consciousness and past conscious states, between the current orientation of our octopus and its past orientations. Different areas lose their reprimability at different rates, neuronal networks that subserve the phonological or articulatory loop lose prime in a few seconds without rehearsal (intentional, PFC mediated repriming) whereas areas primed by the PFC stay primed for longer and are more easily reprimable.

Something that has not been mentioned yet, but is very important is behavior. When assemblies in the motor and premotor cortex are activated we move, act, speak and behave. Early in life, certain lower-order sensory activations are directly connected to movement and these often correspond to innate instincts. As association areas, including the PFC, become metabolic and continue to mylinate they has the ability to inhibit these impulses in certain scenarios, scenarios where it has learned that it is better not to act, or better to do something different. The PFC has to be able to visualize the advanced payoff and must use mental imagery to see this. The person must use imagery to conceptualize how and why it is better to inhibit in order to trick dopamine systems into being driven to repress a response. In order to see the payoff and organize itself to achieve it, it must make certain primes last longer. This allows certain primes to have priority and to interact with future thoughts in order to guide behavior that is informed about the past.

The PFC is just another area, that doesnt know what it is doing at all at first, this is why it doesnt come online for so long, why it is slowly programmed. It has some functional organization but is really an empty vessel (with subdivisions). Existing behavioral reactions are filtered out by inhibitory neurons (this is my alternative to the brain performing computational operations) until a behavioral output that is not inhibited runs to completion. A stimulus activates some sensory primes, and also some PFC inhibitory neurons to make sure that a wrong sensory prime is left out of the picture that is created. The PFC talks to top down areas more so than bottom up, this is why it deals with ideas, more than imagery. It pulls up diverse and variable imagery based on the unique pattern of top down concepts it primes. The PFC evolved to guide behavior though and if it is not fine-tuning behavior it is guiding the imagery created in lower-order areas. One way the PFC informs behavior is by planning. We realize that we may not be adequately prepared to deal with something and so we model it in our mind. This use of imagination, makes it so that when we actually find ourselves in the situation that we modeled we have the right associations to help guide us.

In order for the frontal cortex to do its job and make us do more work for the better, but delayed payoff, it must first be able to see that the new behavior will pay off. They have to be able to actually predict or see the benefit in this behavior. Mice, cats, monkeys, humans are on a continuum for being able to not only choose the harder thing, but see it first. The frontal cortex is activated more when someone is choosing the delayed but greater payoff. Especially when this decision in a novel one, it is more taxing on the PFC to do the harder thing. Almost like we have to use lots of frontal resources to identify the causal mechanisms that make two in the future is better than one now, or else we never are able to justify the harder thing. Without frontal insight, delaying gratification appears as an unjustified or superstitious conception.

We try to do a certain amount of planning before the next big event. This gives us more behavioral flexibility when we get there. Some people dont do this planning they just act how they want and never really make time to plan for this or that. It is too difficult for them to really model the situation because they forget important elements (of their last thoughts) before they can really understand the time-deferred construct. This limits their ability to act in this situation when it comes, and instead of acting with a plan they just react. Planning is learning and thinking in the future. You need flexible and persistant PFC coactivations to do this. It means that you show more instances of inhibition of random responses and will cause your ultimate association to be more likely to be correct. When you have a thought, everything else is inhibited. In the same way that when a reptile moves, all other movements were inhibited because this was deemed to be the best one to solve its problem based on what it knows.

This model has suggested that association cortices, most notably the PFC, conduct information processing in sensory areas. The processing in early sensory areas occurs very fast, in an automatic and deterministic way. These circuits of perception are created by evolution to build appropriate perceptions while minimizing processing time. The PFC and other association cortices say: ok, you found this out in a very fast and automatic way, now I want you to find out this, ok good, now I want you to find out that So the direct feedback from the sensory assemblies of imagery explains how the brain chooses its next set of associations. Without working memory this process allows simpler animals to respond quickly to available perceptions. STM and continued activation help the association areas build a continuous narrative.

The ontology of coactivation is very important. What you have learned in the past helps you to ensure that the next octopus branch that is picked when a certain set of premises are coactivated, is the appropriate one. You might have such a set coactivated and not really get it. You might have someone coactivate all of the right premises for you and still not reach the conclusion by yourself, further, if the person tells you the punch line you might not even appreciate it. The more you know about a set of coactivates, the more meaningful such a punch line is, especially when it revolutionizes your outlook and effects coactivates in the future. Invoking the importance of life experience seems vague and nebulous, but these are exactly the kinds of things that parents teach their children and watch their children slowly come to perceive, understand and put into their own words.

Working memory is the reactivation of multiple long term memories in an organized, unique and dynamic way. Working memory and long term memories are completely dependent on each other because one cannot do mental work without long-term memories and one cannot retrieve appropriate long-term memories without working memory.

Valid Hallucinations

Neurons in early sensory processing areas have very specific jobs. Such a neuron might become active when a certain portion of your visual field contains a bright dot in it. This neuron and the neurons that immediately surround it, which code for the same or nearly the same thing, will send messages to all of the other neurons that it is connected to. Often its efforts go unnoticed. However, if a number of neurons are activated because a series of dots in the visual field are lined-up consecutively, they will all send messages to the neurons they are connected to, inevitably converging (maximally activating) on a neuron (or group of neurons) responsible for coding information about a line. This is how brains use neurons to connect the dots into lines.

Different neurons in visual areas, code for all kinds of visual features and generally these are arranged hierarchically from dots to lines, to contours, to shapes, to objects to concepts. Individual features, once activated at the same time (coactivated) are thought to work cooperatively to activate, or converge upon, a higher (or lower) order area that corresponds to their coactivation. This higher order area holds information about both representations and in essence binds them together to form a new representation that is equivalent to their sum. For instance, if one low-order visual area that holds information about two dots, and another that holds information about the presence of an arc are coactivated (and oriented in a very particular way) this configuration could activate a higher-order visual area involved in object recognition that corresponds to the presence of two eyes and a smile. If the activation of this area is sent to the PFC, it will be attended to consciously and result in the conscious perception of a smiley-face.

The PFC and object recognition area though, can send information back to the lower-order areas responsible for holding the visualization, altering the visualization, how it is perceived, or what aspects of it are focused on. If the higher-order convergence area fires back, this activation branches toward its many inputs (diverging backwards toward the early sensory areas coding for dots and lines), keeping the visual perception active. The COOL thing is that these higher-order convergence areas send backward activations (recurrent retroactivations) that reactivate the sensory inputs that were activated by the initial stimulus, BUT ALSO activate the other inputs that normally feed into it, embellishing the smile. In other words, top-down processing centers activate the very neurons that hold visual imagery, superimposing their expectations over our perceptions. This causes the smile to look more like a prototypical smile or an average smile than it objectively should. This same kind of top-down manipulation of early visual areas is responsible for imagined imagery seen in the minds eye. I think that a good term for this phenomenon is reperception. Mistaken reperception is certainly at the heart of hallucination.

The term valid hallucination seems self-contradictory but I think it is a interesting phrase that helps to communicate how the visual system creates perceptions. I am proposing that when we perceive something we naturally embellish the mental representation of it with top down meaning.

Let me offer an example. The other night I was walking in the dark and I saw a rectangular patch of light on the carpet. I took this patch of light to be a cream colored pillow, and after looking twice I swore I could see folds and convolutions in the medium-thread count fabric. I actually bent down to pick up the pillow and scraped my fingers across the carpeting. This was a true hallucination. Last week I thought I saw my cat in the corner of my eye but it was just his water bowl out of its place. I actually thought to myself, oh he looks hungry, even though I was not seeing him at all. My faucet makes a fluctuating, high-pitched sound when it pours water and twice in the last week I thought that this was the perimeter alarm. I ran out to the alarm panel to punch in the code only to see that the alarm was not sounding at all. Our sensory systems deduce events from prior experience, and the modulated squeaking that my faucet makes bore a close enough resemblance to the alarm to allow it to activate the alarm identification assembly in my head. More often than not though, I correctly identify things in my environment. But it is like mistaking a stranger for a friend, the further away they are, the darker it is and the more they look like your friend the more likely you are to misidentify them.

When you misidentify something you activate narratives and schemas about how to interact with this thing. I tried to grab the carpeting, I ran out to Clearly, it must be possible to identify something, given very little information, and still to do so correctly. When this is done

Later when I thought I saw my cat in the corner of my eye (and got a good picture of him before I actually confirmed it was him) I committed a valid hallucination. I think that people are constantly having valid hallucinations. I think that usually the activation sent from higher-order convergence areas often do not progress all the way back to the earliest sensory areas (like area V1 for dots and V2 for lines). However, I think that we are constantly superimposing what we assume we are seeing over the early, perceptual mental imagery.

For instance, when we see a bright yellow car our visual system only sees the lines and contours that make up the car until they send this information to higher-order systems that

The neuronal assemblies in early sensory areas are building structure by taking their inputs and converging them on other assemblies with a higher-order meaning. This, in the terms of Antonio Damasio, convergent-divergent zone becomes activated and sends info to higher stages of processing. That will rise hierarchically until it activates the PFC and either informs movement or inhibits it. If this zone is turned on maximally it will oscillate back, to its inputs, reactivating them and also activating the other lower-order nodes that are ordinarily associated with them. In other words, if activated sufficiently the higher-order node will reactivate, not only the nodes that activated it, but also the other nodes that ordinarily converge upon it. This is usually right and it embellishes (colors them in and draws around them) perceptions with imagined imagery of prototypical examples. When these prototypes are superimposed over the real image (like a bistable image) the image does not lose its orientation or spatial layout, it is simply embellished, not uprooted or replaced. When this goes wrong though, we see something that we are not supposed to see like an illusion. When this system is working fine, but top-down processing from higher levels fails to veto a perceptual determination it leads to a hallucination. Like when you accept a perception (embellished by a seemingly valid hallucination) as if you were not taking into account where you are and what you are doing. I thought a shoe box was Niko even though I was away from home. I thought I heard Niko even though it was my wheezing. I thought a game was a banging on the gate because I stepped away from the game. I saw a patch of light on the ground and reached down to pick up a pillow. Higher order areas do not properly inhibit perceptions. The perception that is chosen is either the one that sounds the best to the sensory expert due to prior probability or the one that is validated by other areas of different modalities. This tells us that probably a number of nodes are activated by each perception, but only some make it all the way, the ones accepted by higher-up areas and they become attention. These different pathways compete for consciousness.

In other words, perception, and higher-order cognitive processing are inextricably tied together. Thought is the process whereby the PFC, hippocampus and other association areas use different forms of pattern completion to coactivate different sensory areas which leads these sensory areas to automatically run their inputs to completion until new imagery is generated. Once this imagery reaches association areas, the PFC and consciousness, the process starts over.

These LTM representations are generally considered to be memory traces held in cortical networks. The individual networks are groups of neurons that respond to the same inputs and share the same receptive fields. Such neurons are usually found within cortical minicolumns. There are thought to be around 20,000,000 minicolumns in the human cortex each of which is about 30 to 40 micrometers in diameter. The neurons code for specific features that can be activated above baseline so that the cortical networks responsible for them necessitate an increased supply of oxygen and glucose. Bound to positions in a cognitive coordinate system. Individual assemblies can fix chunks against a temporal or a spatial context.

Conclusion

This is a low-level model that has attempted to describe some general patterns of mental activity in terms of neuroanatomical space and the spreading of activation between processing units over time. Mental activity was likened to the polypedal locomotion of an octopus that is seafloor walking. This octopus leaves the majority of its arms where they are and only moves an arm when the foothold it is placed on is distant. This is meant to show that we drop neural assemblies when their relevance to the processing demands diminishes. When one arm is taken off of a branch, the concept corresponding to that branch becomes deactivated and no longer informs present thought.

So how does the interactive television work? It is a continuous cycle between coactivations in higher order areas and those in lower-order (primary and secondary sensory) areas. The lower order imagery automatically activates a new set of higher order nodes that are tuned to the different features of the lower order imagery. These newly activated higher-order nodes are added to the remaining previous nodes (or octopus arms that remained on the past branches) to create a unique set of coactivations that again (via recurrent connections) activates lower order sensory neurons that are responsible for pulling up a new scene of mental imagery. The important point is that the priming lasts longer in the higher order areas and the lower order sensory areas can be wiped clean quickly to accommodate a new picture. The lower-order sensory areas probably last just as long as sensory memory whether they are activated by the environment or by higher order areas. The octopus is always straddling the line between bottom up and top down and the arms (or assemblies) on the bottom-up side move faster.

The vast majority of mammals have small prefrontal cortices so they rarely have assemblies that remain active over a number of brain oscillations. It is probably maladaptive for animals to prolong the influence of a particular feature that is not found in the environment. The persistence of assemblies probably causes animals to react slowly to their environment because their imagery is influenced by past concerns instead of very present concerns in real-time. Mental continuity slows the octopus down. The continuity provided by the PFC allows systemization of the environment and higher-order learning, things that our species adapted to necessitate. Most other vertebrates though dont need this kind of continuity to create adaptive behavior, they probably find the persistence of working memory distracting, noisy and task irrelevant. Assembly activation in most animals is usually due to external stimuli in the environment sensory areas are activated by environmental inputs, these then activate association areas, which in turn activate motor areas then the animal waits for the next group of stimuli. Humans, on the other hand, can activate their sensory areas using only internal, associative stimuli, allowing a reciprocal loop between association and sensory areas. This ability is very limited in most animals and this is why they must learn from trial and error rather than mental modeling.

Maybe all of these bottom-up to top-down reciprocations are organized into oscillations that propagate in regularly timed intervals, across the brain so that they do not interfere with each other. The oscillations reciprocate back and forth at just the right speed so that each area has the time to process its inputs before reprojecting so that they have time to finish processing before they get the next complement of inputs. Messaging would be muddled if areas were to get information while they are processing, or if they didnt receive all of their inputs at the same time. Perhaps these bottom-up to top-down cycles of imagery map on neatly to the synchronized oscillations of neural populations known to give rise to macroscopic oscillatory electric fields, which can be observed in the electroencephalogram.

Some of these assemblies we are aware of (we can report knowledge about the imagery that they invoked) and some influence us without our awareness of them, further, some primes can be attentively activated, some can only be activated by the environment and others can only be activated by a special combination of the two. It is as if the octopus realizes where some of its arms are, but cannot tell in the hurry, where other arms are positioned.

A longstanding debate in this field has been between connectionism and computationalism. The present hypotheses have been largely connectionist as they emphasize the importance of interconnected networks of simple and often uniform units rather than modeling the computational manipulation of explicit symbols. Scientists comparing human brains to computers sometimes assume that the brain accomplishes what it does by performing vast numbers of calculations or computations, using hidden logic and special algorithms to process inputs into outputs. It seems clear to me that this is not going on. Life experience gives the functional structure to these processes that illusorily appears to be computation. Surely much of the function can be quantified and turned into math, but I believe that the connectionistic function must be understood before insights into the processing will be gained from the mathematics.

Gerald Edelman has argued that it is reentry that provides the continuity of thought, but I argue here that it is prolonged activation. Consciousness is the feeling or the sensations associated with the activity of taking information with you through time. And it is not only active assemblies but also assemblies potentiated through priming and potentiated by the hippocampus.

You have to ask yourself, does this systematization sound right? Does this seem like the kind of process that happens in my own thinking on the order of 8-40 times per second?

If we needed size and orientation in our mental imagery, then we would do the work to generate it until primary sensory areas were better linked up to secondary ones in the generation of mental representations. Most people do not do this and so their mental imagery is lacking physical constraints making it seem even more difficult to explain.

Here, the significant neural correlate of consciousness (NCC) is taken to be open and rapid communication between sensory and association areas where sensory areas are creating imagery, and association areas are attending to aspects of the imagery by activating the higher order assemblies that best correspond to these aspects, given the association assemblies already active. Binding then does not occur due to coactivation alone, but via the convergence of coactivated features in both association and early sensory areas.

Step-by-step procedures for visual thinking:

1. A retinotopic image corresponding to the visual field is represented in early visual cortex.
2. Active assemblies in primary visual cortex converge on higher-order representations, activating their assemblies, turning dots into lines, lines into contours and contours into objects. Once these objects have the opportunity to interact with multimodal assemblies in association cortex they can be construed into conceptual constructs.
3. Only aspects of the early sensory imagery that are interesting are attended to. In other words, not-so-early assemblies compete for activation energy and will gain it if they have been properly tuned or sensitized in the past.
4. Individual conceptual constructs are coded for by ensembles made of association area assemblies. Certain assemblies in association areas become newly active and these are added to those that were already activated from the previous cycle(s).
5. The sum of active association areas, previously activated and newly activated, feedback to early visual areas with their new, altered set of specifications, changing the imagery that is automatically pulled up in visual areas.
6. This leads to: a) a new mental image in the minds eye or b) a top-down image is subtly superimposed over the perceptual experience.
7. This new image that is made, that has different features than the image made before it, will be sent back to association areas and will activate new assemblies that were not previously active.

Evolutionarily Important Procedures in Thought:

1. Behave in ways that ensure that our sensory organs are receptive to important stimuli.
2. Notice the things that are salient by sensitizing to them and not habituating against them.
3. Activating associative assemblies that correspond to early sensory imagery of important stimuli.
4. Orienting to the stimulus.
5. Recurrently reactivate the early sensory imagery of the important stimulus, attempting to verify it if it is deemed salient.
6. Activate other association assemblies that hold information about how to react.
7. Continue to pass information about the stimulus back and forth hoping that the right association area assemblies become active so that you can gain coactivated insight about how to deal with the stimulus and then create visual and motor imagery about it that will facilitate the adaptive behavior.