Insight and the Sciences

Franscico Varela was, among other neuro-scientific things, a neurophenomenologist. Neurophenomenologists are interested in how the brain (neuro-) gives rise to consciousness (phenomenology, or the study of the lived experience). Varela was particularly interested in how contemplative eastern spiritual traditions seemed to have implicitly captured the workings of the brain in their teachings and went on to found the Mind and Life Institute which seeks to unite contemplative traditions with the sciences.

Alongside Natalie Depraz and Pierre Vermersch, Varela sketched out three curiously compelling gestures. These ‘gestures’ are the direct product of Varela’s attempts to model the structural dynamics of a common and powerful contemplative practice: ‘becoming aware’ of new ways of thinking, being, and seeing the world. The gestures of awareness: suspension, redirection, and letting go.

Varela’s gestures, as noted in the article linked there, provide a platform for understanding the hard questions of the human experience. In this case, the question of transcending the self. And in particular, Varela’s gestures describe the fundamental attributes of insight, demonstrating that the same insights that dawn upon us in the philosophy class are the ghosts of profound spiritual insight that opens us to new ways of being-in-the-world.

One way of recognising the value of Varela’s gestures in this regard is comparing his almost trivial-sounding descriptions to the total output of the sciences of the mind. It makes you wonder just what, exactly, it is we have been wasting our time doing (though I admit, I’ve wondered myself if brain science was perhaps only ever destined to illuminate the small things).

¶A quick summary of the gestures

Varela’s gestures are worth exploring in detail. But a brief summary here better illustrates the informational overlap. Varela describes:

• suspension: the suspension of habitual patterns of thinking and behaving, a prerequisite for any kind of change;
• redirection: the willful act of turning one’s attention from the external world to the internal world.
• letting go: the releasing of that effortful act of holding internal attention, and instead a passive observation.

For Varela, both redirection and letting go allow us to notice new patterns of thinking and behaving that might be possible—new ways to be that aren’t trapped in the web we have weaved for ourselves.

Again, almost trite, described superficially thus. Well worth further exploration to get a sense of just how well it helps us comprehend insight from the mundane to the profound.

But, perhaps we don’t even have to explore Varela’s gestures to see that his description might just be all that’s worth describing. Let’s journey across the sciences of the mind and see what we come up with.

¶The sciences and insight

¶Information Processing and Cognitivist Heuristics

One key approach to the study of insight comes from the Information Processing literature. Here, humans are viewed as serial processors of information, with a small short term storage and an infinite long term memory. This means that new information comes in and can be held and manipulated for some period of time, but is vulnerable to being pushed out by new information coming in. Unless that information is placed into long-term storage it is lost, and must be relearned.

The short term storage permits us to represent the problem space in which a task is solved. We can hold and manipulate information related to the issue at hand, configuring it in different ways until some satisfactory solution emerges.

On this account, it is the interaction of the problem space (internal representation) and task environment that produces (or does not produce) insight. As such, the major constraints on human problem solving, therefore, is the structure of the task environment (or our knowledge of it) and the structure of the problem space (i.e. what kind of information we hold, and how we are manipulating it). The constraints of either can be explored for their utility or their role as obstacle in the generation of insight. For more detail on this, it is worth reading the seminal work of @simonHuman1971, in particular page 155 on Constructing Problem Spaces as well as the book by the same authors and the same name which emerged the following year.

This forms the core of the cognitivist perspective on problem-solving and insight. Two approaches have developed around this core.

One is inspired by Gestalt psychology, who believed that humans interpret the world as some combination of patterns, configurations, or gestalts, rather than percieving each individual thing. Here, prior knowledge is combined with perceptual ‘chunks’ of information to create a representation of some kind of goal state. To attain insight, one must be good at both decomposing chunks (assign new meanings to perceptual elements) and relaxing constraints (changing the goal representation—overcoming self-imposed constraints). So, for example, one might represent a person’s name in as many ‘chunks’ as there are letters. This would be an ideal strategy for completing an anagram puzzle. But, for remembering the name, perhaps it would be better to represent the name as one ‘chunk’ for the forename and one for the surname. Similarly, if the anagram puzzle asks for as many words to be created as possible from a name, perhaps it would be more efficient to simply write down all the three and four letter words, rather than aspire to produce the most elaborate words. This is known as representational change theory, and insight comes when individuals discover novel ways of ‘chunking’ and novel goal states.

This theory, though heavily computational in recent years, is the closest to earlier gestalt research which focused on the phenomenological experience of the sudden emergence of knowing (insight) as opposed to some more gradual experience of analytic comprehension. Subsequent research in all fields have attempted to address this feature directly or indirectly.

The other approach in the cognitivist domain focuses more on the nature of the internal problem space, noting that most problems have problem spaces that are extremely large. There are an enormous number of ways one could use a cup. Deciding among all of them to determine how to drink water would be a paralysing affair. One must therefore employ heuristics that guide and restrict the search space. Often, these heuristics are relatively stereotypical, learned through experience or taught, and used because ‘they work.’ However, stereotypical heuristics are not always the most efficient. On this account, insight is the product of an heuristic that is particularly accurate or appropriate to the problem. The more flexible one is in deploying heuristics, or the more heuristics one has access to, the more likely the experience of insight. This is broadly known as progress monitoring or satisfactory progress theory.

An excellent overview of these approaches can be found in the reviews of Ollinger and colleagues [e.g. @ollingerPsychological2009]. These reviews are not particularly recent, but do cover the history and ideas in depth.

It is well worth noting that cognitivist perspectives are frequently critiqued for their distance from our understanding of neural function. That is to say, it is difficult to imagine how these cognitive experiences are discharged in the context of actual neurons. Yet, to borrow from progress monitoring theory, they remain excellent heuristics to understanding insight.

¶Attentional episodes

A tangentially related field, that is often agnostic on cognitive representations but quite gnostic on ‘chunks’ is hierarchical reinforcement learning theory [@botvinickHierarchical2012] or the theory of attentional episodes [@duncanStructure2013]. Both are conceptually very similar, and suggest that tasks can be naturally decomposed into ‘chunks.’ Here, complex problems can be described in terms of any number of sub-problems. First, we much reach with the arm, then grasp with the hand, then put the hand to the mouth, then we can eat. This ‘episodic’ nature of tasks is a core concern of modern cognitive science. The role of the mind is thought to be that of a ‘combiner’—putting together these attentional episodes into appropriate sequences to achieve our various goals. This is particularly important when many sequences are possible but conflicting. The classic example might be that of texting and driving, both possible but not at once.

These theories have two unique contributions when thinking about insight.

The first is the concept of abstraction. Some tasks require us to go beyond the obvious structual relations between different episodes. We can note the similarities between different episodes or sequences and thus reduce the dimensionality of the task. For example, the episodes required to mount a bike are similar in many ways to those required to mount a horse. With experience of either, learning the sequence we do not know becomes much easier. Insight emerges when we recognise new abstractions or are able to use old abstractions in new ways.

The second contribution is a way of thinking why insight is uncommon, or difficult. On this account, difficulty to gain insight can come about when a task is not easily differentiated into discrete episodes. If we are unable to recognise the different sub-components of a task, we are much less likely to recognise how it might be similar to more familiar sequences. Similarly, if the episodes of a task are not maintained in the correct sequence (i.e. maintaining appropriate structural relations) we are unlikely to recognise its similarity to other sequences, or recognise what sub-components might be used in new ways.

¶Active Inference (the Bayesian brain or predictive coding)

The active inference approach [@linsonActive2018], and those like it, attempt to explain action and perception in terms of the minimisation of ‘variational free energy’. That is, minimising the difference between some internal generative state and the external world. Typically, this is done by bringing the coupling of the two into a more preferable relationship. When we are hungry, we manipulate the world such that we are no longer hungry (i.e. we eat).

The details of the theory are immensely difficult to understand, even for seasoned cognitive scientists. There is an entire Twitter account dedicated to the struggle of understanding Karl Friston’s free energy principle, and here is an amusing article grappling with the ideas.

However, the core concept is that we live our lives in an internal world. Our past (our priors) and our future (our desired states) guide far more of our behaviour than our present. We must therefore be the ‘author’ of our sensations—guiding our internal world to align with the external world and the external world to align with our internal world. We must minimise the difference.

Here, there are imagined to be lawful links between sensorimotor systems and the environment, and we act to avoid surprise and maximise evidence. When we cross the road, we look both ways, not merely because the paint on the crosswalk tells us to but because there is a degree of uncertainty that surrounds the endeavour. We look to reduce that uncertainty, and to bring our model of the road (a thing on which there may or may not be fast moving cars) closer to the road in front of us (a thing on which we can determine the exact number of cars by examination). The internal generative model cannot exist without our acting upon the environment, and to us the only environment that exists is that which we act upon. In this way, as I mentioned, we are the ‘author’ of our sensations.

As implied by the title, the core of this theory is the inference—a prior, or a belief about what future states are possible, and the constant updating of our priors by unfolding events. Thus, humans are curious and inclined to create structural models of the world.

On this account, ‘aha’ moments come about when the possible model states are reduced to a minimum, and an accurate model can be selected. This kind of model-state-reduction can only be achieved by effective evidence gathering—knowing where or how to find information that will reduce uncertainty, and then a qualitative transition from uncertainty to knowing. A fabulous article, though dense, explains the broad principle and touches on the ‘aha’ moment [@fristonActive2017].

¶Neural bases for insight

While each of these theoretic approaches to insight have merit, an examination of the actual neural activity of these ‘aha’-type moments leads us also to an interesting conclusion. In short, the neural correlates of insightful thought involve a relaxing of constraining patterns of activity and the adoption of more domain-general forms of processing.

Early research concentrated on hemispheric differences—differences between the sides of our brains. Analytic thought was more often associated with the left side of the brain, and insight with the right. This is most likely because analytic thought is more likely to employ linguistic techniques to solve problems. The language centres of the brain are highly lateralised, concentrated about dead centre of the left side of your head. Language is very helpful for problem solving, but is necessarily more constraining in relation to what kinds of things we can think—calling something a chair, for example, makes it very easy to collect all the things chairs are useful for but it make thinking of a chair as a projectile, or a ceiling decoration more difficult, though things things could be equally true.

This became known as the semantic coding hypothesis. The heavy emphasis on hemispheric bias was probably an artefact somewhat of methodology. If you test people in a way that uses words, you’re likely to get the part of the brain that does words excited. Lateralised hypotheses are out of favour in recent years, but the hypothesis has lived on in updated form. The anterior temporal lobe (bilateral—both sides of the brain) is now thought to be the ‘semantic hub’ and similar reasoning is thought to play a role in the difference between analytic and insightful thought—the more constraining the network of semantic objects (schemas/models/priors including linguistic concepts) used to solve a problem the less likely one is to have a breakthrough ‘aha’ moment. It does seem that more activity appears in the right hemisphere during moments of insight however, and once again, one can imagine this is related to the constraints placed on information coding related to the language centres of the brain—perhaps we release ourselves from our linguistic tendencies during these moments.

Pre-understanding neural activity is also distinct between analytic and insightful thought. Analytic thought is associated with occipital cortex. This part of the brain is at the back of the head and is responsible for coding what we see—vision. As such, it’s thought that analytic thought has a tight relationship with outward attention. Insightful thought is associated with anterior cingulate cortex and temporal regions (at the sides) and also presents as more diffuse—spreading across more of the brain. This possibly is indicating less focused attention. The anterior cingulate cortex is often thought to be involved in conflict monitoring, and it has been hypothesised that the ACC is detecting conflicting and possibly less obvious solutions to the problem at hand, associations or information related to a problem that might be more abstract or novel than those solutions we might normally jump to.

A good starting point, and rough summary of most of the above is available in a somewhat old Annual Review paper [@kouniosCognitive2014]. The semantic hub is also worth exploring, particularly the ‘hub and spoke’ model [@chiouControlled2018]. This model is an excellent segue into another line of enquiry, the ethological action maps of Graziano [@grazianoEthological2016].

One key feature of the neocortex (the wrinkly part on the outside of the brain is the predominance of topological maps—the visual cortex, the auditory cortex, the motor cortex, and the somatosensory cortex. Each of these regions appear to code for very specific aspects of perception in a way that could be drawn out in a map.

For example, the auditory cortex has neurons that code for specific frequencies of sound, but as you move away from the auditory cortex and towards the motor cortex, you find parts of the brain that are active for language. This is sensible indeed, because language is both a motor task (producing sounds) and an auditory task (words necessarily sound like things).

The semantic hub and spoke model appears to be a similar map. It describes the anterior temporal lobe (at the sides of the brain) as a place that codes for semantic objects (concepts of things). Different patches of brain code for specific objects—the fusiform gyrus codes faces, and another nearby region houses for example. As you move away from these regions, it codes for more abstract ideas like the semantic concept of home. That might activate the ‘house’ patch, but also other regions because your home is not simply your house.

In essence, certain regions of the brain appear to represent aspects of the world in detail, and as one moves away from these regions, it represents more general features of the world. If one follows that logic to it’s conclusion, one would end up with some regions of the brain that were quite general indeed—coding for the most abstract ideas and concepts. These domain-general regions would likely be involved in more-or-less all activities that were not heavily practiced or heavily tied to one specific aspect of the world.

Of course, these regions do exist. Once thought to be the purview of the prefrontal cortex, but now often thought to be distributed frontal parietal networks. These networks are thought to be involved in complex problem solving and higher-order processes. It may not, therefore, come as a surprise that they feature heavily in research on personalised spiritual experiences [e.g.@mcclintockSpiritual2019]—something that is often associated with a dissolution of the self and unity with ‘god’ or nature.

In sum, the neural literature indicates that insightful thought is the product of diffuse attention, a relaxing of constraining patterns of activity linked to specific concepts or behaviours and the adoption of more domain-general forms of processing.

¶Clinical Insight

An interesting counter-perspective to one of the themes here comes from clinical psychology. While cognitive insight into illness has long been known to be a crucial aspect of successful therapeutic intervention (particularly in disorders with positive psychotic symptoms), it also demonstrates the possible risks.

Cognitive insight, awareness of one’s illness and attribution of symptoms to the illness, is most readily associated with features of self-reflection, with an emphasis on metacognitive skills. It also heavily emphasises a flexibility of self-certainty. One must be able to evaluate from different perspectives, and not invest too heavily in one’s self-beliefs.

However, self-reflection can also contribute to mood instability, with particular reference to depression. Indeed, a theoretical argument could be made that self-absorption is a key feature of a constellation of disordered symptoms. Certainly it is a feature of more ‘neurotic’ personality types. More self-reflection has an unfortunate habit of becoming negative. It is thus well worth noting that self-reflective thought is not always the solution, but rather a tool. Like anything, too much can become problematic.

An excellent recent review was published recently in the Clinical Psychology Review [@vancampCognitive2017].

¶Summary

While the content here is dense, what we wanted to do was summarise the insights about insight from across the sciences of the mind. See how helpful Varela’s gestures are in understanding one of the crowning achievements of human thought. So, here we go:

1. Insight is crucially reliant on an inward, not an outward focus;
2. Insight is determined by processes that are more diffuse—less associated with clear or typical patterns of behaviour or neural activity, or more open to different possibilities and patterns than those more readily available to us; and
3. Insight can be distinguished from analytic thought. The former is spontaneous and the latter is gradual, both with distinct patterns of activation in the brain.

To remind you, Varela’s gestures are:

• suspension: the suspension of habitual patterns of thinking and behaving, a prerequisite for any kind of change;
• redirection: the willful act of turning one’s attention from the external world to the internal world.
• letting go: the releasing of that effortful act of holding internal attention, and instead a passive observation.

The overlap is uncanny, no? That these findings of the sciences of the mind so neatly map to these gestures, and the these gestures similarly map just as straightforwardly to 500-year-old theological scholarship in the same domain suggests that Varela’s gestures are well worth paying attention to.

¶References

Botvinick, Matthew Michael. 2012. “Hierarchical Reinforcement Learning and Decision Making.” Current Opinion in Neurobiology 22 (6): 956–62. https://doi.org/10.1016/j.conb.2012.05.008.

Chiou, Rocco, Gina F. Humphreys, JeYoung Jung, and Matthew A. Lambon Ralph. 2018. “Controlled Semantic Cognition Relies Upon Dynamic and Flexible Interactions Between the Executive ‘Semantic Control’ and Hub-and-Spoke ‘Semantic Representation’ Systems.” Cortex 103 (June): 100–116. https://doi.org/10.1016/j.cortex.2018.02.018.

Duncan, John. 2013. “The Structure of Cognition: Attentional Episodes in Mind and Brain.” Neuron 80 (1): 35–50. https://doi.org/10.1016/j.neuron.2013.09.015.

Friston, Karl J., Marco Lin, Christopher D. Frith, Giovanni Pezzulo, J. Allan Hobson, and Sasha Ondobaka. 2017. “Active Inference, Curiosity and Insight.” Neural Computation 29 (10): 2633–83. https://doi.org/10.1162/neco_a_00999.

Graziano, Michael S. A. 2016. “Ethological Action Maps: A Paradigm Shift for the Motor Cortex.” Trends in Cognitive Sciences 20 (2): 121–32. https://doi.org/10.1016/j.tics.2015.10.008.

Kounios, John, and Mark Beeman. 2014. “The Cognitive Neuroscience of Insight.” Annual Review of Psychology 65 (1): 71–93. https://doi.org/10.1146/annurev-psych-010213-115154.

Linson, Adam, Andy Clark, Subramanian Ramamoorthy, and Karl Friston. 2018. “The Active Inference Approach to Ecological Perception: General Information Dynamics for Natural and Artificial Embodied Cognition.” Frontiers in Robotics and AI 5 (March): 21. https://doi.org/10.3389/frobt.2018.00021.

Öllinger, Michael, and G. Knoblich. 2009. “Psychological Research on Insight Problem Solving.” Journal of Neuroscience - J NEUROSCI, January. https://doi.org/10.1007/978-3-540-85198-1_14.

Simon, Herbert A., and Allen Newell. 1971. “Human Problem Solving: The State of the Theory in 1970.” American Psychologist 26 (2): 145–59. https://doi.org/10.1037/h0030806.

Van Camp, L. S. C., B. G. C. Sabbe, and J. F. E. Oldenburg. 2017. “Cognitive Insight: A Systematic Review.” Clinical Psychology Review 55 (July): 12–24. https://doi.org/10.1016/j.cpr.2017.04.011.