Fuster’s theory of cognits

J. Fuster has put forward a new model of thought based on the idea of memory networks he calls cognits. Here is the abstract from a recent paper, Cortex and memory: emergence of a new paradigm.

Converging evidence from humans and nonhuman primates is obliging us to abandon conventional models in favor of a radically different, distributed-network paradigm of cortical memory. Central to the new paradigm is the concept of memory network or cognit–that is, a memory or an item of knowledge defined by a pattern of connections between neuron populations associated by experience. Cognits are hierarchically organized in terms of semantic abstraction and complexity. Complex cognits link neurons in noncontiguous cortical areas of prefrontal and posterior association cortex. Cognits overlap and interconnect profusely, even across hierarchical levels (heterarchically), whereby a neuron can be part of many memory networks and thus many memories or items of knowledge.

And an abstract from a somewhat earlier paper, The cognit: a network model of cortical representation.

The prevalent concept in modular models is that there are discrete cortical domains dedicated more or less exclusively to such cognitive functions as visual discrimination, language, spatial attention, face recognition, motor programming, memory retrieval, and working memory. Most of these models have failed or languished for lack of conclusive evidence. In their stead, network models are emerging as more suitable and productive alternatives. Network models are predicated on the basic tenet that cognitive representations consist of widely distributed networks of cortical neurons. Cognitive functions, namely perception, attention, memory, language, and intelligence, consist of neural transactions within and between these networks. The present model postulates that memory and knowledge are represented by distributed, interactive, and overlapping networks of neurons in association cortex. Such networks, named cognits, constitute the basic units of memory or knowledge. The association cortex of posterior-post-rolandic-regions contains perceptual cognits: cognitive networks made of neurons associated by information acquired through the senses. Conversely, frontal association cortex contains executive cognits, made of neurons associated by information related to action. In both posterior and frontal cortex, cognits are hierarchically organized. At the bottom of that organization-that is, in parasensory and premotor cortex-cognits are small and relatively simple, representing simple percepts or motor acts. At the top of the organization-in temporo-parietal and prefrontal cortex-cognits are wider and represent complex and abstract information of perceptual or executive character. Posterior and frontal networks are associated by long reciprocal cortico-cortical connections. These connections support the dynamics of the perception-action cycle in sequential behavior, speech, and reasoning.

This work caught my attention in a posting to the blog The Quantum Lobe Chronicles by W. Lu. (here)

Although the modular modeling of the brain has utterly failed due to a lack of conclusive evidence, many neuroscientists continue to maintain this antiquated view… but why? Put quite simply, there was nothing better. However, thanks to Fuster, a new paradigm is emerging…
Introducing the cognit network model. It postulates that memory and knowledge are represented by interactive, distributed, and overlapping networks of neurons in association cortices.
The posterior-post-rolandic association cortex contains perceptual cognits and the frontal association cortex contains executive cognits. The prefrontal and posterior association cortices are linked by complex cognits in a hierarchical order. The parasensory and premotor cortex, found at the bottom of the hierarchy, contain relatively simple and small cognits which represent motor acts or simple percepts. At the top of the hierarchy is the temporo–parietal and prefrontal cortex containing larger cognits representing complex and abstract information of perception and executive control. The long reciprocal cortico–cortical connections between the posterior and frontal networks support sequential behavior, speech, and reasoning.

I and me

An article by N. Farb and others, ‘Attending to the present: mindfulness meditation reveals distinct neural modes of self-reference’, (here) looks at the difference between the present ‘I’ and the past ‘me’ in creating the conscious self.

Since William James’ early conceptualization, the ‘self ’ has been characterised as a source of permanence beneath the constantly shifting set of experiences that constitute conscious life. This permanence is often related to the construction of narratives that weave together the threads of temporally disparate experiences into a cohesive fabric. To account for this continuity, William James posited an explanatory ‘me’ to make sense of the ‘I’ acting in the present moment…Narrative self-reference stands in stark contrast to the immediate, agentic ‘I’ supporting the notion of momentary experience as an expression of selfhood. Most examinations of self-reference ignore mechanisms of momentary consciousness, which may represent core aspects of self-experience achieved earlier in development.

Here is the abstract:

It has long been theorised that there are two temporally distinct forms of self-reference: extended self-reference linking experiences across time, and momentary self-reference centred on the present. To characterise these two aspects of awareness, we used functional magnetic resonance imaging (fMRI) to examine monitoring of enduring traits (’narrative’ focus, NF) or momentary experience (’experiential’ focus, EF) in both novice participants and those having attended an 8 week course in mindfulness meditation, a program that trains individuals to develop focused attention on the present. In novices, EF yielded focal reductions in self-referential cortical midline regions (medial prefrontal cortex, mPFC) associated with NF. In trained participants, EF resulted in more marked and pervasive reductions in the mPFC, and increased engagement of a right lateralised network, comprising the lateral PFC and viscerosomatic areas such as the insula, secondary somatosensory cortex and inferior parietal lobule. Functional connectivity analyses further demonstrated a strong coupling between the right insula and the mPFC in novices that was uncoupled in the mindfulness group. These results suggest a fundamental neural dissociation between two distinct forms of self-awareness that are habitually integrated but can be dissociated through attentional training: the self across time and in the present moment.

Folk knowledge

A child growing up learns to predict – that is how the child becomes capable of moving without injury and obtaining the things it needs or wants. This is the first bit of folk knowledge the child has, ‘events have causes’. Whether the child gains this knowledge by experience or is born with it (or more likely both), it becomes a foundation of our relationship with the world and our movement in it. It is a piece of folk knowledge because we all tend to believe it, we use it as a tool of thought and action, but a modern theoretical physicist would find it a difficult idea to accept without caveats. Like most of folk knowledge it is a fairly good first approximation, it will usually do, it is good enough for most occasions. Other adjectives that often replace folk are naïve, vernacular, and commonsense.

Another example is ‘things fall down’ which is pretty good for most situations; we can think differently but still find this a useful idea. ‘The sun goes around the earth’, is one bit of knowledge that we no longer accept even as a first approximation. It has become just a idiom of our language.

But as well as the folk knowledge that we use to interact with the inanimate physical world, there is also a folk knowledge that we use in social contexts. This has been called folk psychology as opposed to folk physics. Again we have a spectrum from principles that we can hardly operate without, through useful approximations, to ideas that have completely failed us. One recent picture of folk psychology as most of us experience it is the Theory of Mind. We not not actually know which parts of this theory are with us to stay, which will not be believed but still used and which will be completely rejected.

The notion of intention is probably one that is with us for good and something we cannot operate without, even probably something we are born with. Our folk ideas about emotion might turn out to be useful for a long time if a bit simplistic. On the other hand, the notion of consciousness has been drifting about for centuries, changing in nature, importance and connections to other ideas. It continues to drift. What most would think of as commonsense about consciousness has become untenable scientifically. The folk version of consciousness is likely to go the way of the geocentric universe.

Working in the missing hierarchical level

The sort of research that may bridge the gap (see last posting on this blog, The missing hierarchical level) is illustrated by a ScienceDaily report on an article by D. Lutz. (here)

On the one hand, there are the individual nerve cells whose membrane depolarization is at the basis of everything and on the other hand, there’s lyric poetry, serial murder and the calculus. In between there are hundreds of billions of nerve cells, with hundreds of trillions of connections. Scientists understand the bottom end and can say useful things about the top end, but getting from one end to the other is another story…

…They devised a simple mathematical model that accurately represents a three-cell microcircuit in a chicken’s brain. They then found, to their surprise, that they could collapse their model mathematically to one equation with two parameters, which are derived from but not the same as the strength of the connections between neurons (the synaptic strength) that neural models usually emphasize…

one fruitful approach has been to analyze a microcircuit consisting of a few neurons all of whose interconnections can be traced.

Microcircuits that have been analyzed in this way include the network that controls the chewing rhythm of a lobster’s stomach (yes, it has teeth in its stomach), the reflex arc that allows a fruit fly to dodge a fly swatter and the timing network that controls the heartbeat of a medicinal leech.

Looking for a way to explain to the students in his Physics of the Brain class how delayed feedback produces complexity in a circuit, Ralf Wessel, Ph.D., associate professor of physics in Arts & Sciences, came up with a toy neural circuit simple enough to be stepped through time iterations at the blackboard. He used it to show his students that if there was feedback among the neurons, simple constant inputs could produce a long-period oscillation in their outputs.

Wessel then asked Matthew S. Caudill, Ph.D., graduate research assistant in physics, to create a computer model of the circuit so that it could be explored more thoroughly. As they worked with three-neuron microcircuit they realized it was very like one students in Wessel’s neurophysiology lab were studying.

That circuit, which consists of three neurons and their feedback projections, has a simple task: to detect motion in the chicken’s field of view. One neuron in the area called the optic tectum because it sits on the “roof” of the brain, sends axons to others in a knob of tissue called the nucleus isthmi. The neurons in the isthmi send projections back to the optic tectum, either directly to the neuron from which they got their input or back to the rest of the tectum (the crucial feedback loops). (This feedback resembles the loops between the cortex and thalamus in mammals, which are central components of consciousness.) There are similar microcircuits in the optical processing areas of reptilian and mammalian brains.

The microcircuit’s behavior could be captured mathematically by three equations, each of which describes one neuron’s output in terms of its inputs and parameters called synaptic weights, the standard way of expressing the strength of the connection between two neurons. Looking at the equations, Wessel and Caudill recognized that they could be reduced by algebraic substitution to one equation with two parameters (derived from the original five synaptic weights).

“It is as if,” says Caudill, “the system of three neurons was reduced to one abstract neuron that does the same thing, follows the same rule, as the more complicated circuit. “

…The finding also suggests why it might be difficult to achieve insight into neural circuits in the lab. Neurophysiologists probing neuronal circuits by inserting a glass electrode in one neuron, stimulating it, and recording the electrical activity of another neuron are likely to be baffled by physiological variations from animal to animal. The chicken microcircuit suggests that to understand a neural circuit they would have to measure combinations of synaptic weights. But the catch is they can’t know which ones to measure unless they already know the rule the circuit follows.

The missing hierarchical level

One of the problems with neural research is that there is a missing hierarchical level. To explain this I have to discuss the nature of science. The basic idea of science is to gather data (or facts, or observations) by investigations and then explain the the data with theories about the nature of the physical world. I know, I know; this is a very simplistic description, but it will do for now.

The larger the range of data explained by the theory, the more accurate the predictions that the theory makes and the more consistent the theory is with other knowledge (mathematics, logic, other established theories), than the better (or the more convincing and useful) the theory. Each large theory has its own model of how the physical world works; it has its own vocabulary of words and concepts; it has its own tools, methods and instruments; it has its own rules, laws and relationships. There is actually very little overlap between theories, but some. So for instance, the theory of plate tectonics and the theory of evolution by natural selection are very different. One can discuss either of them for days on end without needing to refer to the other one. But they do overlap when location or time of events in their narratives overlap or where their casual links entwine. These are sister theories, more or less at the same hierarchical level. However, not all theories are on the same hierarchical level. Theories can only overlap if they are on the same level or adjacent levels.

If we are looking at evolutionary biology and we want to explain the nature of fitness, we have to drop out of evolutionary biology into the level of functional physiology in order to see the functioning of the eye producing sight and how this helps an animal to survive. If we then want to talk about the changes that have occurred in eyes, we have to drop down another level to the cellular level with its overlapping theories of biochemistry, biophysics and genetics. For a deeper explanation we would drop down to macro chemistry and physics and then down to atomic physics and then sub-atomic physics. But you cannot jump over a level. You cannot get any reasonable explanation of physiological function from the concepts of sub-atomic physics – there is no overlap.

When we look at the brain/mind, we are missing an important level. On the one hand, there is behavior, ideas, emotion, cognition with theories in psychology, ethology, anthropology, philosophy and so on. On the other hand there is neurobiology, mostly at the cellular level. There is no overlap between the wet and squiggly neuron in brain tissue and the dry abstractions of a ‘black-box’ mind. There is experimental evidence for the very close connection between the brain’s activity and the mind but there is no accepted theory that bridges that gap.

What would such a theory look like? It would be about the functional physiology and anatomy of medium and large groups of neurons and in particular how they manipulate information. It would be about how the concepts of mind are manifested in neural activity.

One way to understand something is to try and make it. That is why creating artificial intelligence, work with neural networks, attempts to accurately simulate small areas of brain tissue and the like will contribute. Another approach is to study very simply ‘brains’ and completely understand how they function before climbing the stairs to ever more complex brains, trying to get a complete understanding at each level. Also, many are going to keep collecting data on connected events at the mind and brain levels, looking for generalizations and patterns. There is little doubt that the tools that exist today, primarily various kinds of scans and electrical recordings, will be improved and new types added. A few ‘eureka’ ideas will get investigators out of old habits of thinking.

Everyone is entitled to their opinion, but frankly it is not helpful to those looking for a theory that bridges the gap when someone stands to the side and says that such a theory is impossible. In effect they are saying that it can’t be done because no one has ever done it. Until recently, no one has tried very hard. Now quite a few are trying.

Holding something in mind

Edge has a speech by S. Dehaene, Signatures of Consciousness. Here is a small part of it on the possible usefulness of consciousness. The whole speech is worth reading (here).

In several experiments, we have contrasted directly what you can do subliminally and what you can only do consciously. Our results suggest that one very important difference is the time duration over which you can hold on to information. If information is subliminal, it enters the system, creates a temporary activation, but quickly dies out. It does so in the space of about one second, a little bit more perhaps depending on the experiments, but it dies out very fast anyway. …. When you are conscious of information, however, you can hold on to it essentially for as long as you wish. It is now in your working memory, and is now meta-stable. The claim is that conscious information is reverberating in your brain, and this reverberating state includes a self-stabilizing loop that keeps the information stable over a long duration. Think of repeating a telephone number. If you stop attending to it, you lose it. But as long as you attend to it, you can keep it in mind.

Our model proposes that this is really one of the main functions of consciousness: to provide an internal space where you can perform thought experiments, as it were, in an isolated way, detached from the external world. You can select a stimulus that comes from the outside world, and then lock it into this internal global workspace. You may stop other inputs from getting in, and play with this mental representation in your mind for as long as you wish.

In fact, what we need is a sort of gate mechanism that decides which stimulus may enter, and which stimuli are to be blocked, because they are not relevant to current thoughts. There may be additional complications in this architecture, but you get the idea: a network that begins to regulate itself, only occasionally letting inputs enter.

It would be interesting to know how Dehaene separates the contents of consciousness from focus of attention. I do think that one of the reasons for consciousness has to do with being able to remember in episodic memory things that are in consciousness.

Include the thalamus

ScienceDaily has a report of two studies on the thalamus from M. Sherman’s lab by B. Theyel and D. Llano and by C. Lee. (here).

Two new studies show that the thalamus–the small central brain structure often characterized as a mere pit-stop for sensory information on its way to the cortex–is heavily involved in sensory processing, and is an important conductor of the brain’s complex orchestra. …”The thalamus really hasn’t been a part of people’s thinking of how cortex functions,” said Sherman, “It’s viewed as a way to get information to cortex in the first place and then its role is done. But the hope is these kinds of demonstrations will start putting the thalamus on the map.”… information makes a stopover in the thalamus before being sent to the visual cortex of the brain to be processed. Similarly, auditory and somatosensory (touch) information is routed through the thalamus before traveling to cortex for more complex processing. …Once sensory information reaches the cortex, it is thought to remain segregated there as it moves from primary cortex to secondary cortex and higher-order areas. But when Theyel severed the direct connection between primary and secondary cortical regions, stimulating primary somatosensory cortex still activated secondary cortex as well as the thalamus, suggesting a robust pathway from cortex to thalamus and back. Only when the thalamus itself is interrupted does the activation of secondary cortex fail. The observation that at least a portion of sensory information passes back through the thalamus on its travels between cortical areas refutes the notion of the thalamus as a passive, one-time relay station, Theyel and Sherman said. “The ultimate reality is that without thalamus, the cortex is useless, it’s not receiving any information in the first place,” …. “But that may be because as a bottleneck, it provides a convenient way to control the flow of information. It is a very strategically organized structure.” … “These are two parallel streams serving different functions,” Lee said. “The thalamus is also the central hub for transferring information between cortical areas. Rather than carrying information, this second pathway winds up modulating information being sent between cortical areas.”… Both papers newly characterize the complexity of the thalamus and its role in shaping sensory information both before and after that information reaches higher cortical regions — not a crossroads, but a conductor. … “People who study how the cortex functions now have to take the thalamus into account. This can’t be ignored.”

I like to think of the neo-cortex as the thalamus’ on-line computer. The thalamus-cortex loop is certainly part of the neurological basis of consciousness.

Metaphor 2

A couple of months ago, Sing your own Lullaby posted a list of theories of meaning (here). I intended to comment on the list but never got my act together. There is more commonality then difference between various ways of looking at meaning. So we have constructs, frames, schema, conceptual networks, models – different but similar ways of explaining how we think and communicate. For years I have thought in terms of models and metaphors. These ideas are no better (or worse) then others but easier for me to write about.

Here is an old piece of mine about metaphor.

The problem with meaning is that we often think that something has meaning: a word has meaning, a symbol has, an event has. But this is wrong – a single thing cannot have meaning. Meaning is the relationship between things. We see this clearly in a dictionary. Every word is defined in terms of other words, which in turn are defined by others, and so on. When we have a large number of things that have mutual relationships so that each contributes meaning to the whole and each gets meaning from its place in the whole, then we have one of these structures that have been called constructs, schema, conceptual networks, maps of the territory, models and so on. They comprise our understanding. By blending or metaphor or elaboration, we can built more complex and larger structures – we can increase our understanding.

But this still leaves a problem. With the circular dictionary, we have to ground it by relating a few words to the world. We have to do some ‘pointing’ to shared, real experiences. Similarly, to make our meaningful structures we have to have some starting point. We have to be born with some primitive concepts with primitive relationships been them in order to ‘boot up’ our understanding. We need a foundation on which to build a structure. We have to have the first few schema or maps of the territory in order to get the process of creating metaphors started. Once started, a structure of metaphors or models or maps of reality can grow in number and size to form the sum total of a person’s concepts.

Some basic metaphors would be so natural that very little ‘hard wiring’ would be needed to make them close to universal. An article in Psychological Science, The Thermometer of Social Relations – Mapping Social Proximity on Temperature, by H. Ijerman and G.R. Semin, examines the two-way relationship between physical warmth and positive social feelings. (here)

In this view, abstract concepts and concrete experiences that are jointly expressed in a metaphor are coexperienced. In the case of ‘‘warmth is affection,’’ Lakoff and Johnson (1999, pp. 45–60) argued that this coexperience is primary: Babies experience the feeling of being held affectionately by their mothers, and being so held induces a warm sensation. This association is underlined by evidence that the insular cortex is involved in processing both psychological and physical warmth (see Williams & Bargh, 2008a). As a result, people express and share the abstract notion of affection in terms of the coexperienced sensation of warmth. Examples are abundant in mainstream culture: ‘‘The cold shoulder’’ and ‘‘a cold fish’’ are examples of metaphors relating lack of warmth to social distance, whereas ‘‘warm embrace’’ and ‘‘giving a warm welcome’’ are metaphors linking warmth to social proximity.

Interestingly, we can see here both a social learning path and a brain architecture path to this metaphor. I think it is likely that either alone could be a foundation for this particular complex of metaphors, but, in what is normal situations, they are both present and re-enforce each other. Lakoff and Johnson proposed that concrete experiences (e.g., temperature) ground abstract concepts (e.g., affection). This perspective is referred to as embodied realism. Some other pairings include: cleanliness and moral purity, physical dirtiness and self-disgust, physical and emotional pain (here), time and space, muscular movement and any effort or accomplishment.

We need metaphors because we have limited resources. Imagine the simplest nervous system – some sensory neurons with synapses with some motor neurons, like spinal cord reflexes. What allows sophisticated responses, memory, learning, thinking etc. is the complexity of networks of inter-neurons separating the sensory side from the motor side, in other words, a brain. But in the end the only thing that enters the brain is sensory signals and the only thing that leaves is motor (and glandular/chemical/emotional) signals. These primary signals must ground the metaphoric meanings we use to think with. The concepts and words we use have meaning by their relationships grounded in basic, primitive, hard-wired sensory, motor and chemical processes.


The papers are full of a story about a Belgian man who was thought to be in a vegetative state for 23 years and who has been found to be fully conscious but almost completely paralyzed. An expert called Steven Laureys has been working on ways to determine whether someone is in a coma or conscious. He uses many different scanning methods to, in effect, map the patterns within the brain. From these patterns he can categorize the brain’s degree of coma from permanently vegetative to fully conscious. Of 44 patients that he examined with the vegetative state diagnosis, he found 18 of them responded to communication under his methods. He is a respected expert and it seems that those in this field of medicine do not doubt his diagnoses.

The Belgian man, R. Houben, according to press reports, intends to write a book about his life using his one finger on a touch pad. This news, unlike the news that he is conscious, is not that reliable. Laureys himself is very noncommittal on this communication and does not want it associated with his work. The point is that Houben does not use the touch pad but instead his therapist uses something similar to the controversial ‘facilitated communication’ method popular with some therapists supposedly communicating with profoundly autistic children. This method has been discredited. So if anyone says that Houben said this, that or the other, you can bet that this was largely the words of the therapist and not Houben, without much fear of losing your money. Facilitated communication therapists tend to be well meaning but fool themselves more that they fool others.

It is unfortunate that a very effective new method to measure conscious activity has been put in a confusing relationship with a ineffective and probably unethical method of communication.

Working memory

JR Minkel has an interesting piece in the Scientific American site (here). He is commenting on the work of W. Zhang and S. Luck who have found that working memory does not fade away but disappears suddenly.

When you go from bed to bathroom on a dark night, a quick flick of the lights will leave a lingering impression on your mind’s eye. For decades evidence suggested that such visual working memories—which, even in daylight, connect the dots to create a complete scene as the eyes dart around rapidly—fade gradually over the span of several seconds. … (Zhang & Luck) tested subjects’ recall for the hues of colored squares flashed briefly on a screen up to 10 seconds earlier. Subjects marked their answer on a color wheel. If memories decay gradually, the guesses should have become increasingly imprecise as time wore on …. Instead subjects went straight from fairly accurate answers to random choices—no better than chance—indicating the memories were decaying all at once. According to Zhang and Luck’s mathematical analysis, most subjects’ memories went “poof” somewhere between four and 10 seconds after the stimulus. …Researchers say a sudden die-off is to be expected if working memories are stored in circuits that feed back on themselves.

I think that working memory and consciousness are related but that it is not clear what the detail of the relationship is. Clearly our conscious perceptions have more resolution, detail and accuracy then our later memories. Is this because of the detail in the working memory?