Cognition and the Brain

What cognition means

Cognition is a collective noun. It covers every process by which the brain takes in information, keeps hold of it, changes it, and puts it to use. Perceiving a face, remembering where you parked, working out a tip, choosing not to reply to a rude email, following an argument in a lecture: all of these are cognition. The word does not name a single faculty but a family of operations that share a common feature, namely that they work on information rather than directly on the world.

Psychologists usually break cognition into components for study: perception, attention, memory, language, reasoning, decision-making, and the executive control that coordinates the rest. These divisions are useful, but they are analytical conveniences rather than natural joints in the brain. Real tasks recruit several at once. Reading this sentence requires visual perception, attention to the page, retrieval of word meanings from long-term memory, a working memory buffer to hold the start of the sentence while you reach the end, and executive control to keep you on task rather than checking your phone.

This library treats cognition as the level of description that sits above cells and signals but below whole-person behaviour. If you want to know how a neuron fires or how a synapse passes a message, those pages cover the machinery. Here we are concerned with what that machinery, assembled into networks, actually computes.

Why there is no thinking centre

The most persistent popular idea about the thinking brain is that somewhere inside the skull there is a region responsible for intelligence, a control room where thought happens. It is an appealing picture, and it is wrong. More than a century of lesion studies, and three decades of functional imaging, point consistently in the opposite direction: cognition is produced by distributed networks whose members are widely separated, and almost no cognitive function depends on a single region alone.

Two lines of evidence make the case. The first is lesion data. When a stroke or injury damages one region, the resulting deficit is usually specific and partial rather than a wholesale loss of thought. Damage to the left inferior frontal gyrus disrupts fluent speech production but often leaves comprehension and reasoning comparatively intact. Damage to the hippocampus devastates the formation of new long-term memories while leaving working memory, language, and intelligence broadly preserved. If there were a single seat of thought, damaging it would abolish thinking; nothing of the kind is observed.

The second is functional imaging. When people perform demanding cognitive tasks in a scanner, activity does not concentrate at one spot. It appears across a reliable constellation of regions: lateral prefrontal cortex, posterior parietal cortex, the anterior cingulate, the insula, and, depending on the task, sensory and motor areas. Change the task and the pattern shifts, but the same core regions keep reappearing, engaged by almost any task that is difficult. That pattern, a shared network recruited by many different demands, is precisely what a distributed account predicts and a single-centre account does not.

The frontoparietal control network

Among the networks that support cognition, one has a special claim on attention: the frontoparietal control network, sometimes called the multiple-demand system or the central executive network. It comprises regions of the lateral prefrontal cortex, particularly the dorsolateral surface, together with the posterior parietal cortex around the intraparietal sulcus, plus the anterior insula and dorsal anterior cingulate. These are connected by long white matter tracts, most notably the superior longitudinal fasciculus, that let them exchange information rapidly.

What makes this network distinctive is its generality. Ask people to hold digits in mind, switch between rules, inhibit a prepotent response, solve a matrix reasoning problem, or sustain attention against distraction, and this same network lights up. It is not tuned to a particular kind of content. Instead it appears to implement the control operations that all demanding tasks require: representing the current goal, keeping it active in the face of interference, selecting which information to prioritise, and updating when circumstances change.

It is worth noting what the frontoparietal network is not. It is not the whole of cognition. Perception, long-term memory, emotion, and habitual action all rely on other systems, and the control network typically works by modulating those systems rather than replacing them. Nor is it a homunculus, a little person inside the head who does the deciding. It is a set of regions whose activity implements control functions; the control is distributed, not delegated to an inner agent.

The four building blocks

Cognition is easier to understand if you separate the operations it performs. Four of them are foundational, and each has a page in this section. A convenient way to remember their relationship is a simple division of labour: attention selects, working memory holds, executive function controls, and language encodes.

1. Attention selects

   The brain receives far more input than it can process. Attention is the mechanism that decides what gets through: a dorsal frontoparietal system biases processing towards what your goals make relevant, while a ventral system interrupts when something unexpected demands a look.

2. Working memory holds

   Once selected, information must be kept available for a few seconds while you work on it. Working memory is that temporary, capacity-limited workspace, and its capacity is one of the strongest single correlates of reasoning ability that psychology has found.

3. Executive function controls

   Executive function is the family of control processes that keep behaviour aligned with a goal: inhibiting an unwanted response, updating the contents of working memory, and shifting flexibly between tasks or rules.

4. Language encodes

   Language lets the brain compress experience into symbols, hold it, share it, and reason with it. Its neural basis is a pair of distributed streams across the left hemisphere in most people, not the two spots of the textbook diagram.

These four are not independent modules stacked in a pipeline. They overlap heavily, share neural substrate, and constrain one another. Attention and working memory in particular are so intertwined that some theorists treat the contents of working memory as simply the items currently in the focus of attention. Treat the fourfold division as a useful scaffold for learning, not as a claim about how the brain partitions itself.

The brain and IQ: the honest evidence

Because this site is about intelligence testing, it matters to be straight about what neuroscience has actually established regarding the brain and IQ. The honest summary is that the associations are real, replicable, and modest, and that they are group-level statistical findings that do not license conclusions about individuals.

The Parieto-Frontal Integration Theory

The most influential attempt to synthesise the imaging literature is the Parieto-Frontal Integration Theory, or P-FIT, proposed by Rex Jung and Richard Haier in 2007. They reviewed dozens of structural and functional imaging studies of intelligence and found that the regions implicated most consistently were not scattered at random. They clustered in the dorsolateral prefrontal cortex, the inferior and superior parietal lobules, the anterior cingulate, and several temporal and occipital regions, along with the white matter tracts connecting them, especially the arcuate fasciculus.

P-FIT's central claim is therefore about a network. On this view, individual differences in reasoning ability arise from how efficiently information flows from posterior regions, where sensory input is elaborated, into parietal regions, where it is abstracted and integrated, and then into frontal regions, where hypotheses are tested and a response is selected. Intelligence, in this framing, is a property of a distributed circuit, not of any single structure. Notably, the P-FIT regions overlap substantially with the frontoparietal control network described above, which is exactly what you would expect if reasoning tests measure control capacity.

Brain volume and IQ

The most frequently cited structural finding is that total brain volume correlates positively with IQ. It does. But the size of the correlation is what matters, and it is small: meta-analytic estimates of the association between in vivo brain volume and IQ typically land in the region of 0.2 to 0.3, and more recent large-sample analyses tend towards the lower end of that range rather than the higher.

A correlation of 0.25 means brain volume accounts for somewhere around 6 per cent of the variation in test scores across a population. Ninety-four per cent lies elsewhere. In practical terms, if you took two people at random and knew only which had the larger brain, you would do barely better than chance at guessing which scored higher. Any claim that a scan can tell you how clever someone is misrepresents the evidence badly.

White matter and processing speed

A second strand of evidence looks not at the size of regions but at the quality of the connections between them. Diffusion imaging studies find that the microstructural integrity of white matter tracts, particularly the long association fibres linking frontal and parietal cortex, is associated with better performance on reasoning tests, and that much of this association appears to run through processing speed: better-organised white matter is linked to faster, more reliable information transfer, and speed of processing is itself a robust correlate of reasoning ability.

This fits P-FIT neatly. If intelligence depends on efficient integration across a network, then the wiring should matter at least as much as the nodes, and it appears to. It also fits the well-documented fact that processing speed and reasoning both decline together in later life, a pattern consistent with age-related degradation of white matter.

If you want to see what an intelligence test actually measures and how the resulting numbers work, our guide to IQ scores and our comparison of crystallised and fluid intelligence cover the psychometric side, and you can take a test to see the format for yourself.

Common misconceptions

The cognition reference pages

Each of the pages below takes one part of the cognitive system and treats it in depth: the anatomy, the classic experiments, the current models, and the honest state of the evidence linking it to measured intelligence.

Related topics