Abstract

The prefrontal cortex is the anterior region of the frontal lobe, a mosaic of association areas that supports the control of thought and action in the service of goals. Its definition is still argued over, because architectonic, connectional and functional criteria do not line up neatly across species. This article reviews its subdivisions and comparative size, the structure of executive functions, the shift from a single theory of control to frontoparietal networks, the working-memory debate, population codes, the control hierarchy, orbitofrontal and ventromedial valuation, neuromodulation, development, lesion evidence and psychiatric conditions. Domain-general frontoparietal recruitment is well supported, and the unity-and-diversity structure of executive functions is supported in adult samples, although no single factor model fits every sample. The substrate of working memory, the role of somatic markers, the form of the control hierarchy and the orbitofrontal value-versus-map question remain open.

Keywords: prefrontal cortex, executive function, cognitive control

The prefrontal cortex is the part of the brain most closely tied to organising behaviour around objectives. A review of focal damage in human patients and nonhuman primates concluded that the region is necessary, not merely active, when thought and action must be steered toward a goal (Szczepanski & Knight, 2014). Fuster's synthesis cast its overarching contribution as the temporal organisation of behaviour and reasoning, and declined to tie the region to any single function or single locus (Fuster, 2001).

What counts as prefrontal is nonetheless unsettled. A review that asked what constitutes the prefrontal cortex judged that its structure and function have not been resolved across species, so that proposals of uniquely human prefrontal regions remain arguments rather than findings (Carlén, 2017). Functional accounts have meanwhile shifted from cataloguing abilities to describing a family of related control processes: a 2022 review organised the evidence on cognitive control and executive function around one shared component and two separable ones, and distinguished its common component from general intelligence while relating control to a general factor of psychopathology and to impulsivity (Friedman & Robbins, 2022). The sections below trace these debates and give each contested account its current standing.

Key Takeaways

  • Lesion evidence from patients and monkeys shows that the prefrontal cortex is necessary for goal-directed control, the steering of behaviour toward goals, yet its boundaries and its definition across species remain unresolved.
  • Whether humans have a disproportionately large prefrontal cortex is contested: one MRI comparison found the human frontal cortex was not disproportionately large relative to great apes, while allometric analyses, which judge a structure's size against other brain measures, and multimodal analyses, which combine several kinds of brain data, found the prefrontal cortex itself larger than expected, each with definitional caveats.
  • In adult samples, executive functions, the mental skills that manage thinking, are correlated but separable: current models pair a common factor, one ability shared by every task, with updating-specific and shifting-specific factors, although results for inhibition have been inconsistent and no single model fits every sample.
  • Many kinds of demanding task recruit a domain-general frontoparietal multiple-demand system, a network of frontal and parietal areas that very different hard tasks share, and accounts of cognitive control have moved from a single prefrontal theory toward interacting networks.
  • Whether working memory rests on persistent prefrontal firing, sparse bursts or activity-silent synaptic states, in which information is held in temporary changes at synapses rather than in continuous firing, is an open debate, and one review draws on human and monkey evidence that storage is distributed beyond the prefrontal cortex.
  • The somatic marker hypothesis, a strict rostro-caudal (front-to-back) hierarchy of control and the orbitofrontal value-versus-map question are all contested, and cortical change continues through adolescence along non-linear trajectories rather than ending at a fixed age.
  • Groups of people with ADHD, schizophrenia or addiction show prefrontal differences in brain structure or activity, but the effects are often small, partly entangled with medication and found only in group comparisons, so they cannot diagnose any individual.

Anatomy and Definition

Locating the prefrontal cortex is easier than defining it. A 2017 review concluded that the structure and function of the region have not been resolved across species, a gap that matters more as mouse models are used increasingly to study it (Carlén, 2017). The difficulty is therefore not only technical but conceptual: a homology claim requires agreement on which criteria make a patch of cortex prefrontal in the first place.

In primates the classical answer is architectonic. Fuster counted Brodmann areas 8 through 13, 24, 32, 46 and 47 as prefrontal, grouped them into three major regions, and estimated that together they account for nearly one-third of the neocortex (Fuster, 2001). Petrides's comparative review found a broadly similar architectonic plan in the lateral prefrontal cortex of macaques and humans, with areas arranged from top to bottom as well as from front to back, though the human areas are more developed (Petrides, 2005). That scheme separates mid-dorsolateral from mid-ventrolateral zones and relates lateral prefrontal regions to monitoring within working memory and to cognitive control (Petrides, 2005).

Imaging has since refined the map. Drawing on several kinds of Human Connectome Project imaging data, a semi-automated parcellation, paired with a machine-learning classifier for locating areas in individuals, divided each human cortical hemisphere into 180 areas, 97 of them newly described, using 210 healthy young adults and a further 210 for validation (Glasser et al., 2016). The authors treated 180 as a probable lower bound, suspected that some parcels are complexes of several areas, and noted that the sample contained only healthy young adults (Glasser et al., 2016).

At the anterior tip lies the frontopolar cortex, also termed the anterior prefrontal cortex, which corresponds to Brodmann area 10; a review of its role in managing competing goals linked it to keeping several goals under watch at once and to switching between them, while describing the human evidence as speculative and the monkey evidence as coming from early studies (Mansouri et al., 2017). The orbitofrontal cortex and the ventromedial prefrontal cortex occupy the underside and the medial wall, as Figure 1 sketches. Drawing on meta-analytic data, one review tied the ventromedial prefrontal cortex to social cognition, to the regulation of negative emotion and to value-based choice, while noting that different subregions of the area serve distinct functions (Hiser & Koenigs, 2018).

Figure 1

The Major Functional Subdivisions of the Prefrontal Cortex

A schematic left-lateral view of the prefrontal cortex with its dorsolateral, ventrolateral, orbital, and frontopolar subdivisions marked A simplified left-facing brain outline. The front third is shaded and divided into an upper dorsolateral region, a lower ventrolateral region, an orbital region on the underside, and a frontopolar region at the tip. A separate inset marks the medial anterior cingulate, which lies on the inner surface and is not visible from the side. Dorsolateral working memory, reasoning Ventrolateral retrieval, selection Orbital / ventromedial Fronto- polar Medial surface Anterior cingulate conflict, error monitoring
Note. A schematic left-lateral view. Dorsolateral and ventrolateral regions occupy the outer surface; the orbital and ventromedial cortex lies on the underside; the frontopolar cortex sits at the anterior tip. The anterior cingulate lies on the medial wall, shown separately because it is hidden in a side view. Boundaries are approximate and functional, not exact cytoarchitectonic borders. Original schematic.

Types of Prefrontal Cortex

MeSH files Prefrontal Cortex under Frontal Lobe and lists two narrower descriptors beneath it. The subtypes need not be mutually exclusive or exhaustive, and MeSH is an indexing classification for organising the biomedical literature, not a theory of how the cortex is carved into functional units.

Table 1. Direct subtypes of Prefrontal Cortex in the MeSH classification (tree A08.186.211.200.885.287.500.270.700).

SubtypeIn brief
Broca AreaAn inferior frontal territory conventionally identified with BA44 and BA45; analysis of individual participants shows that it holds a language-selective region and a separate multiple-demand region, so it is not a single functional unit (Fedorenko & Blank, 2020).
Dorsolateral Prefrontal CortexThe lateral region whose left-hemisphere activity tracked the implementation of control in a task-switching Stroop study (MacDonald et al., 2000) and whose neurons in monkeys code a remembered location across a delay (Funahashi et al., 1989).

Note. Neither subtype has its own article on this site. Both regions recur in the body sections on cognitive control and working memory.

Is the Human Prefrontal Cortex Special?

Whether the human prefrontal cortex is unusually large has been argued for decades, and the answer depends heavily on what is measured. An MRI study of living primates that included every hominoid species found that the human frontal cortex was not disproportionately large compared with that of the great apes, and suggested that differences among individual cortical areas and in their interconnections may matter more for human cognition (Semendeferi et al., 2002). That result concerned the frontal cortex as a whole rather than its prefrontal portion.

Analyses aimed at the prefrontal cortex itself have reached a different conclusion. Applying regression to cytoarchitectonic measurements across primates, Passingham and Smaers found the human prefrontal cortex larger than allometry predicts even when apes were included in the comparison, although only two datasets were available and one of them relied on the frontal lobe as a stand-in for prefrontal cortex (Passingham & Smaers, 2014). A multimodal structural MRI study of 60 humans, 29 chimpanzees and 19 macaques likewise found the human prefrontal cortex disproportionately large, examining both cortical gray matter and subcortical white matter under conservative, liberal and genu-based definitions of the region (Donahue et al., 2018). Its authors were candid about the limits: three species give little leverage for estimating scaling slopes, drawing prefrontal boundaries involved judgment, homologies in the chimpanzee are not conclusive, orbitofrontal signal dropped out in the nonhuman scans, and most of the nonhuman animals scanned were female (Donahue et al., 2018).

Rodent research adds a separate problem. Combining a word-frequency analysis of PubMed, a review of rodent brain atlases and a survey of 38 prefrontal researchers, Laubach et al. found that the naming of rodent prefrontal areas is inconsistent and that researchers agree on it only weakly (Laubach et al., 2018). The survey sample was small and self-selected, and the homologies the authors proposed were offered tentatively (Laubach et al., 2018). Even so, a review of the mouse prefrontal cortex argued that the mouse experimental toolbox can inform theories that were founded on primate studies, while conceding that the mouse data are still accumulating (Le Merre et al., 2021). The question of what, if anything, is distinctive about the human prefrontal cortex therefore remains unresolved (Carlén, 2017).

Executive Functions: Unity and Diversity

Much of the psychology of the prefrontal cortex runs through the construct of executive function. Its modern structure comes from a latent-variable study of one hundred thirty-seven college students in which shifting, updating and inhibition emerged as moderately correlated abilities that were nonetheless clearly separable (Miyake et al., 2000). The same study also asked whether these latent abilities predicted performance on more complex tasks: they related to operation span, random number generation, the Tower of Hanoi and the Wisconsin Card Sorting Test, whereas dual tasking bore no relation to any of the three functions (Miyake et al., 2000).

Individual-differences work built on simple laboratory tasks was later summed up in four general conclusions, including that executive functions stay fairly stable over time and carry a genetic influence (Miyake & Friedman, 2012). The resulting unity and diversity model of executive functions decomposes performance into a common executive factor shared by all tasks plus factors specific to updating and to shifting, estimated with bifactor and twin models (Friedman & Miyake, 2017). Its current standing is supported, with stated limits: the three-function model is not meant to be comprehensive, results for an inhibition factor have been inconsistent, factor structures vary between studies, and executive tasks suffer from task impurity and low reliability (Friedman & Miyake, 2017). An independent systematic review with bootstrap reanalysis, covering 46 samples (N = 9,756), found that no single model of executive functions fit every sample and that the structure differed with age (Karr et al., 2018). Acceptance rates for the models were low, which the authors suggested may reflect a bias toward publishing models that fit well but may not replicate, drawn from underpowered samples (Karr et al., 2018).

A 2022 review carried this framework into the neuroscience of cognitive control, describing a common control component alongside components specific to set shifting and working memory updating, with the common component closely related to response inhibition and distinct from general intelligence, distinguishing cool from hot forms of control, and relating control to the general factor of psychopathology (Friedman & Robbins, 2022). Research on working memory converged on the same territory. Kane and Engle argued that working-memory capacity, and its association with general fluid intelligence, reflects executive attention for which the dorsolateral prefrontal cortex is critical, while treating that region as one node in a wider network and disclaiming any new theory of the prefrontal cortex (Kane & Engle, 2002).

Clinical cases show why such control matters outside the laboratory. Shallice and Burgess described three patients with traumatic prefrontal injuries who failed at open-ended tasks that required them to organise several subgoals and to reactivate intentions at the right moment, despite intact intelligence; two of the three passed other standard frontal tests (Shallice & Burgess, 1991). With only three patients, the series shows that the dissociation can occur, not how common it is.

Cognitive Control: From a Single Theory to Networks

The most influential single account is the integrative theory of prefrontal cortex function. It holds that the prefrontal cortex actively maintains representations of goals and that these maintained patterns send bias signals to other brain regions, steering processing along the pathways a task requires (Miller & Cohen, 2001). Its current standing is revised rather than overturned: recent reviews locate flexible control in the dynamic interaction of several prefrontal networks within large-scale brain organisation (Menon & D'Esposito, 2022) and in a structured set of shared and separable control processes (Friedman & Robbins, 2022).

Imaging soon pulled control apart into components. Event-related fMRI with a task-switching Stroop task showed the left dorsolateral prefrontal cortex, in Brodmann's area 9, engaged when control had to be put into effect, while monitoring of performance was associated with activity in areas 24 and 32 of the anterior cingulate cortex (MacDonald et al., 2000). At the cellular level, recordings in behaving monkeys showed that single prefrontal neurons encode abstract rules, in this case the two rules of a same/different picture task; the published summary gives no neuron counts (Wallis et al., 2001).

The medial frontal contribution received a normative treatment of its own. The expected value of control proposes that the dorsal anterior cingulate cortex estimates whether investing control is worthwhile, setting the expected payoff of a given control signal against the cost of cognitive effort and so determining how much control to allocate (Shenhav et al., 2013). The account concerns the anterior cingulate on the medial wall rather than the lateral prefrontal convexity, and its standing is supported as a normative proposal, a theory of what control allocation ought to compute (Shenhav et al., 2013). An earlier account of the medial frontal role, conflict monitoring, holds that the anterior cingulate cortex registers conflict in processing and signals that more cognitive control is needed, so that conflict and control form a feedback loop (Botvinick et al., 2001). The proposal drew on anterior cingulate responses to conflict and was worked out in two computational modeling studies (Botvinick et al., 2001). Conflict monitoring now stands as a revised account, since the expected-value-of-control framework later sought to explain the dorsal anterior cingulate's many proposed functions, performance monitoring among them, as expressions of a single underlying function (Shenhav et al., 2013). A separate answer to what the region does treats the medial prefrontal cortex, anterior cingulate cortex included, as a learner that forecasts where an action is likely to lead, whether the result is good or bad; in this view cognitive control follows from appraising those forecasts, and one simple learning-rule model reproduced in simulation a broad range of reported medial frontal effects, among them effects of conflict and of reward (Alexander & Brown, 2011). That support comes from modelling rather than new data, and the authors described the region's basic function as disputed among rival accounts (Alexander & Brown, 2011).

A separate line of work asked whether particular regions are recruited by many kinds of demand. Reviewing neuroimaging work in 2000, Duncan and Owen found that tasks as unlike one another as problem solving, choosing a response and holding information in working memory recruited a similar set of frontal regions for many demands: the dorsal anterior cingulate cortex and the middle portions of the ventrolateral and dorsolateral prefrontal cortex (Duncan & Owen, 2000). Cortex on the orbital and medial faces of the frontal lobe responded little to these demands (Duncan & Owen, 2000). The same review observed how little agreement then existed on prefrontal specialisation (Duncan & Owen, 2000). Duncan later proposed the multiple-demand system, a network of frontal and parietal cortex engaged across diverse demands and closely tied to fluid intelligence, which he cast as the machinery for assembling structured mental programs, drawing on single-cell recording and fMRI (Duncan, 2010). That link has been challenged by work combining factor models with neuroimaging-based simulations, which argues that intelligence is better described as several components, each tied to a distinct functional brain network, than as a single higher-order factor g (Hampshire et al., 2012). Analyses at the level of individual participants then showed that focal frontal and parietal regions in each person responded across seven varied and demanding tasks, overlapping in a way that implies general-purpose machinery alongside the brain's specialised components (Fedorenko et al., 2013).

The most detailed current map comes from multimodal parcellation of 449 Human Connectome Project participants. Across three cognitive contrasts, a core of 10 multiple-demand parcels per hemisphere emerged, together with 17 additional areas, forming a distributed and strongly interconnected domain-general network surrounded by a weaker penumbra (Assem et al., 2020). The standing of the multiple-demand system is supported and refined: its regions are recruited together across tasks yet show only modest relative preferences among them, a pattern that older imaging, with its coarser localisation, could not resolve (Assem et al., 2020).

Network approaches recast control as a matter of connectivity. In a within-subject fMRI study of 15 adults performing 64 tasks built from 12 rules, the frontoparietal control network showed globally variable connectivity that shifted with the task, and its connectivity patterns carried enough information to decode which task was under way, consistent with flexible hubs (Cole et al., 2013). The evidence is modest in size: the sample was small, 64-way decoding reached 5% against a chance level of 1.56% (p = .013), and decoding of sensory rules was not significant (p = .077) (Cole et al., 2013). A framework proposed in 2020 extends the idea, arguing that network coding models, in which activity flows over measured connections, can predict the cognitive representations a region will carry from the organisation of the network (Ito et al., 2020). Menon and D'Esposito's synthesis identifies six key prefrontal networks whose changing interplay supports flexible control, and lists unresolved questions about how those networks operate (Menon & D'Esposito, 2022).

Working Memory: Persistent Activity and Its Rivals

Working memory, the brief holding and use of information that is no longer present, is the function most often attributed to the lateral prefrontal cortex, and how the region implements it is disputed. The historical anchor is single-neuron recording in monkeys performing an oculomotor delayed-response task. Of 288 neurons recorded around the principal sulcus, 170 showed activity related to the task and 87 were active during the delay, with delay activity that coded the direction of the cue the animal had to remember (Funahashi et al., 1989). A 1995 review by Goldman-Rakic addressed the cellular basis of working memory (Goldman-Rakic, 1995). Human imaging offered a different reading of such delay activity. A critical review of fMRI studies argued that sustained dorsolateral prefrontal activity during the retention interval may direct attention toward representations held in posterior regions rather than store the content itself, and conceded that what this persistent activity codes is still unclear (Curtis & D'Esposito, 2003).

Recordings like these became the basis of the persistent-activity model of working memory, in which remembered content is held by sustained spiking across the delay. Taking the affirmative side of a 2018 dual-perspectives exchange, Constantinidis and colleagues held that working memory rests on prefrontal neurons that keep firing through the delay, and they rejected rival accounts that rely instead on rhythmic discharges or on short-term synaptic plasticity alone (Constantinidis et al., 2018). Its companion paper took the opposite side. Lundqvist, Herman and Miller argued that, examined trial by trial, working-memory spiking occurs in brief and sparse episodes whose code changes over time, and that averaging across trials hides this structure, while agreeing that spiking remains important (Lundqvist et al., 2018). The standing of the persistent-activity model is therefore contested (Constantinidis et al., 2018; Lundqvist et al., 2018).

The burst account rests on recordings of local field potentials and spikes in monkey prefrontal cortex. In those recordings, brief gamma bursts (45-100 Hz) accompanied the encoding and reactivation of the remembered items, while beta bursts (20-35 Hz) reflected a default state that encoding and decoding interrupted; gamma appeared at only a minority of recording sites, and a gating role for the bursts was put forward as a hypothesis (Lundqvist et al., 2016).

A more radical alternative is activity-silent working memory. Stokes proposed that information can be carried through periods without elevated firing by hidden states, chiefly short-term synaptic plasticity, within a dynamic coding framework in which population patterns evolve over the delay (Stokes, 2015). He conceded several limits. The synaptic proposal is difficult to test, most of its support comes from monkeys, and human evidence is still needed (Stokes, 2015). Forms of sustained firing that use less energy had not been ruled out, and the functional role of short-term plasticity had not yet been fully established (Stokes, 2015). Human evidence has since begun to test the idea. In a study combining transcranial magnetic stimulation with multivariate pattern decoding, a single pulse brought an item that was held but unattended back into decodable activity (Rose et al., 2016). This happened only when the item might still be needed later, and the authors treated the result as suggestive rather than conclusive support for a synaptic account of working memory (Rose et al., 2016). Across two EEG experiments with 30 and 19 healthy adults, a visual impulse presented during the delay of a retro-cue task revealed hidden states whose decodable patterns tracked the cued memory contents, and a third, behavioural experiment tested 20 more (Wolff et al., 2017). An item cued to be forgotten could no longer be decoded, although temporarily unattended items remained coded, and the impulse brought no behavioural benefit (Wolff et al., 2017). In three monkeys, large single-trial recordings from many lateral prefrontal neurons at once showed that the remembered information was not carried by spiking sustained across the delay; instead, the population switched back and forth, in concert, between 'On' and 'Off' states (Panichello et al., 2024). During the Off states, information about the item was still reflected in which ensembles of neurons were functionally coupled, a coupling estimated from spike timing across the trials of each cue condition rather than observed directly at synapses (Panichello et al., 2024). From this the authors infer that brief, item-specific bursts of spiking work alongside synaptic mechanisms, an interpretation the recordings suggest but do not establish (Panichello et al., 2024). Its standing is contested: it remains a proposal whose central mechanism is unconfirmed, and other burst-based work emphasises active, spatially organised dynamics in controlling what working memory holds (Stokes, 2015; Lundqvist et al., 2023).

A further strand questions whether the prefrontal cortex is the store at all. Drawing on studies in humans and monkeys, one review argued that working-memory content is held in a distributed way across sensory, parietal and prefrontal cortex rather than in prefrontal cortex alone, with persistent activity a general property of cortex (Christophel et al., 2017). The open question it set for future work is how these areas together turn sensory input into a delayed response (Christophel et al., 2017). On this view prefrontal neurons may matter most for controlling what is held. In two rhesus monkeys, multi-area recordings that included 682 lateral prefrontal neurons showed that the prefrontal cortex used a common code when selecting an item already in working memory and when directing attention to a perceived one, consistent with a domain-general controller (Panichello & Buschman, 2021). The equivalent generalisation was weaker and not significant in the frontal eye fields, and it was also not significant in parietal cortex and V4; with two animals, the result awaits replication (Panichello & Buschman, 2021).

The newest synthesis in the burst line is a reanalysis of prefrontal recordings from five rhesus monkeys across four working-memory tasks. It proposes spatial computing: beta and gamma bursts, organised in spatial patterns across the recording arrays, carry low-dimensional control signals over working memory that coexist with mixed selectivity in individual neurons, and the authors discuss the scheme's relevance to zero-shot learning (Lundqvist et al., 2023). The work mostly reanalysed earlier datasets, the mapping of locations to functions was task-specific, and the data are available only on request (Lundqvist et al., 2023). The latest evidence thus strengthens the case for burst-based control without settling where content is stored: persistent firing, sparse bursts, silent synaptic traces and distributed codes each retain defenders.

The interactive demonstration below sets the two main accounts of maintenance side by side. Readers choose an account, lengthen the delay or strengthen a distractor, and see whether the stored cue still reaches the recall threshold; the model is illustrative rather than fitted to data. The recordings described above fit the second account more closely than the first, since they suggest that intermittent, item-specific spiking works alongside synaptic mechanisms rather than spiking sustained across the delay.

Interactive demonstration · Holding A Cue Across Time

Two Accounts of Working-Memory Maintenance

Choose an account, then set the length of the delay and the strength of an intervening distractor. In the persistent account the visible firing itself carries the memory. In the bursts-plus-trace account firing is sparse, and the cue is held in a slowly fading hidden trace that brief bursts refresh, so spiking and synaptic storage work together. Watch whether the stored quantity still clears the recall threshold when the delay ends.

Delay length (steps)10
Distractor strength20
recall thresholdmaintenance delaystored (%)050100
Cue retained
With sustained firing, a delay of 10 steps and distractor strength 20 leave the stored quantity at about 50.5%. That is above threshold, so the cue can still guide the response. Here the memory is visible as continuous firing throughout the delay.
Interactive demonstration. An illustrative model contrasting two accounts of how prefrontal cortex holds information across a delay: sustained firing (after Constantinidis et al., 2018) and sparse bursts over a hidden synaptic trace (after Stokes, 2015; Lundqvist et al., 2018). No parameter is fitted to data, and the burst schedule is arbitrary. The point is qualitative: both schemes can retain the cue, yet they predict very different moment-to-moment activity, which is why the debate turns on single-trial recordings rather than trial-averaged firing. Large-scale single-trial recordings in monkeys found that cue information did not persist in spiking across the delay, and suggest that intermittent, item-specific spiking and synaptic mechanisms may work together (after Panichello et al., 2024), a combination the second option sketches.

Population Codes and Representational Geometry

The most active front in prefrontal research treats the population, not the single cell, as the unit of explanation. A reanalysis of 237 lateral prefrontal neurons from two monkeys showed that many cells respond to nonlinear combinations of task variables, a property termed mixed selectivity, and that this endows the population with a high-dimensional code (Rigotti et al., 2013). The populations were assembled pseudo-simultaneously rather than recorded together, only linear readouts were considered, and the authors speculated that the collapse in dimensionality on error trials reflects nonlinear mixed selectivity being the most useful but also the most fragile form of selectivity (Rigotti et al., 2013).

Population dynamics also explain how context changes a computation. Two macaques judged either the motion or the colour of random dots depending on a cue; recordings from prefrontal cortex in the region of the frontal eye field, comprising 388 single units and 1014 multi-units and interpreted with a trained recurrent network, indicated that irrelevant input was not filtered out early but selected late, within the population dynamics, through a selection vector and a line attractor (Mante et al., 2013). Most recordings were not simultaneous, and the authors allowed that other selection mechanisms may operate alongside this one (Mante et al., 2013).

Later work examined the shape of these codes. Bernardi et al. recorded prefrontal and hippocampal ensembles in monkeys and found that task variables were laid out in an abstract geometry, measured by how well a linear decoder trained on some conditions generalised to the rest, a format that supports generalisation; the geometry changed with task events (Bernardi et al., 2020). Flesch et al. combined neural network simulations, fMRI in 32 human participants and a reanalysis of macaque recordings, and reported that dorsolateral prefrontal and parietal cortex represent the two contexts of a context-dependent decision task on low-dimensional manifolds orthogonal to each other, resembling the rich rather than the lazy learning regime of networks (Flesch et al., 2022). Only one monkey supported the orthogonality result, the BOLD signal is an indirect window on neural coding, the network was simpler than the fMRI task, and the finding conflicts with reports of high-dimensional prefrontal codes (Flesch et al., 2022).

The broader tension between high- and low-dimensional codes is the subject of an earlier review by Badre et al., which argues that the dimensionality of control representations embodies a trade-off between keeping task conditions separable and allowing what is learned to generalise, and which leaves several questions open (Badre et al., 2021). Studies of this kind illustrate a stance Ebitz and Hayden name the population doctrine, which they see rising beside the longer-established single-neuron doctrine; their review rests on a small set of core concepts, manifolds and subspaces among them, and observes that motor neuroscience is where the doctrine has so far made its largest mark (Ebitz & Hayden, 2021).

The meta-reinforcement learning account of prefrontal cortex extends population thinking to learning itself. In six simulation experiments with a recurrent network of LSTM units, including two-armed bandit and two-step tasks, Wang et al. proposed that reward prediction errors carried by dopamine train prefrontal circuitry, whose activity dynamics then implement a second learning process, a form of learning to learn (Wang et al., 2018). The account's standing is contested: it is a simulation-based proposal that reproduces effects qualitatively rather than by fitting parameters, and it abstracts away from neuroanatomical detail (Wang et al., 2018).

The Hierarchy of Control

The rostro-caudal hierarchy of cognitive control holds that the lateral frontal cortex is ordered along its front-to-back axis, with more anterior regions governing more abstract or temporally extended control. An influential early statement came from an imaging study in which Koechlin et al. described cognitive control in the lateral prefrontal cortex as a cascade running from premotor regions toward anterior prefrontal regions, distinguishing sensory, contextual and episodic levels of control (Koechlin et al., 2003).

A 2009 review generalised the cascade into a gradient of abstraction along the rostro-caudal axis, from area 6 caudally through areas 46 and 9/46 to area 10 at the pole, and judged a true hierarchy plausible but not yet demonstrated (Badre & D'Esposito, 2009). It set out the open problems candidly: a gradient does not by itself imply a hierarchy, evidence that rostral regions dominate caudal ones was preliminary, studies operationalised abstraction in different ways, the effects of perturbing caudal regions had not been tested, and the difference between dorsal and ventral patterns was unexplained (Badre & D'Esposito, 2009).

Nee and D'Esposito tested the scheme with fMRI and dynamic causal modelling in healthy adults. They placed the apex of the hierarchy not at the frontal pole but in mid-lateral prefrontal cortex, with abstraction graded both rostrally (temporal and contextual control) and caudally (feature control), and related the connectivity estimates to higher-level cognitive ability measured with Raven's Advanced Progressive Matrices and span tasks (Nee & D'Esposito, 2016). Dynamic causal modelling supports only relative claims, the circuits modelled were simplified, one key correlation was marginal (r = 0.39, p = 0.07), and partial correlations were inconclusive about the mid-lateral region (Nee & D'Esposito, 2016).

A 2018 review drew the consequence. A strict rostro-caudal gradient of abstraction is not supported, and neither is a unitary model of frontal function; hierarchical control is better described in terms of interacting frontal networks, although frontal organisation remains controversial (Badre & Nee, 2018). The hierarchy's standing is therefore revised: hierarchical control survives as an idea, but the simple principle that more anterior means more abstract does not (Nee & D'Esposito, 2016; Badre & Nee, 2018).

The frontal pole retains a distinctive role in this revised picture. In the monkey work that Mansouri et al. review, frontopolar cortex is tied to keeping track of goals, and the authors speculate that the human region may additionally follow several goals at once and move between them (Mansouri et al., 2017).

The interactive demonstration below illustrates the cascade proposal described above. Readers change how many response options, contexts and temporal episodes a task involves and see which part of the front-to-back lateral prefrontal axis the proposal would engage for each kind of demand.

Interactive demonstration · Front To Back, Abstract To Concrete

The Rostro-Caudal Control Cascade

In the cascade proposal, control is graded along the front-to-back axis of the lateral prefrontal cortex: posterior regions select a response from the immediate stimulus, more anterior regions add the current context, and the most anterior regions add the episode, the events that came before. Raise each demand and see how many bits of control each level must resolve and how far forward the cortex is engaged.

Stimulus-response alternatives (S)4
Contexts (C)2
Temporal episodes (E)1
Premotor cortex — Sensory control2 bits
Caudal lateral PFC — Contextual control1 bits
Rostral lateral PFC — Episodic control0 bits
Engaged to: Caudal lateral PFC
Sensory (premotor)Contextual (caudal)Episodic (rostral)
The task carries 3 bits of control in all: 2 sensory, 1 contextual, and 0 episodic. Because the response depends on context, control reaches the caudal lateral prefrontal cortex.
Interactive demonstration. An illustrative model of the cascade account of prefrontal control (after Koechlin, Ody, & Kouneiher, 2003; Badre & D'Esposito, 2009), in which sensory, contextual and episodic demands recruit progressively more anterior regions of the lateral prefrontal axis, each resolving log-base-2 of its factor in bits. The model follows the original proposal; later work found that a strict rostro-caudal gradient is not supported and placed the apex of the hierarchy in mid-lateral prefrontal cortex (Badre & Nee, 2018), so the mapping of levels to regions here is illustrative.

Value, Decision Making and the Orbitofrontal Cortex

Ventral prefrontal damage has a striking effect on decision making. On a novel laboratory task that simulated real-life choices involving reward and punishment, patients with prefrontal damage behaved as though blind to the future consequences of their choices, favouring immediate prospects; the report's abstract gives no statistics and left the cause of the impairment unexplained (Bechara et al., 1994).

Damasio's somatic marker hypothesis offered an explanation (Damasio, 1996). It proposed that bodily signals tied to emotion, which may act covertly or overtly, mark the anticipated outcomes of options and thereby guide reasoning and choice, with the ventromedial prefrontal cortex as a key structure (Damasio, 1996). Damasio did not present it as an account of the whole prefrontal cortex or as a single unifying frontal mechanism, and he rejected explanations based purely on conditioning or purely on cognition (Damasio, 1996).

The hypothesis's standing is contested, and three critiques are considered here. Maia and McClelland gave 40 undergraduates, 20 per condition, a structured awareness questionnaire during the Iowa Gambling Task and found that participants held conscious knowledge of which decks paid off, which removes the need to invoke nonconscious somatic markers to explain their choices (Maia & McClelland, 2004). They were explicit about the limits: the sample was small, nonconscious knowledge was not ruled out, the causal role of awareness was unresolved, and performance did not differ significantly between the questionnaire and replication conditions (two-sided P = 0.28) (Maia & McClelland, 2004). Dunn et al.'s critical review judged that the somatic marker hypothesis still lacks adequate support, pointing among other problems to thin causal evidence that bodily feedback steers choice and to simpler rival accounts of Iowa Gambling Task performance (Dunn et al., 2006). Steingroever et al. combined two extensive literature reviews with a reanalysis of eight data sets from healthy participants and found three key assumptions about healthy performance questionable: healthy choice was heterogeneous, which undercuts clinical readings of poor performance (Steingroever et al., 2013).

Imaging has located valuation more generally. Pooling coordinates from 206 published fMRI studies, Bartra and colleagues located subjective-value activity in the anterior ventral striatum as well as the ventromedial prefrontal cortex, while responses in the dorsomedial prefrontal cortex and anterior insula went in both directions, a mixed pattern that may reflect a U-shaped salience or arousal signal (Bartra et al., 2013). Single-neuron recordings in monkeys making economic choices found that orbitofrontal neurons encode both the value of the goods on offer and the value of the good finally chosen (Padoa-Schioppa & Assad, 2006). On that evidence the orbitofrontal cortex is a candidate network for economic choice, not an established one (Padoa-Schioppa & Assad, 2006). Drawing mainly on neurophysiology, Padoa-Schioppa and Conen argued that economic choices between goods are made within a circuit in the orbitofrontal cortex, contrasting this good-based view with rival models while acknowledging open questions (Padoa-Schioppa & Conen, 2017).

The orbitofrontal cognitive map offers a different reading of the same region. Wilson et al. proposed that the orbitofrontal cortex tracks which state of a task the animal currently occupies, giving reinforcement learning a map of task space (Wilson et al., 2014). They related the idea to several classic learning findings, devaluation and reversal effects among them (Wilson et al., 2014). They also noted that testing its predictions requires tasks whose states cannot be observed directly, and the paper itself did not test them (Wilson et al., 2014). Human evidence followed: fMRI pattern classification showed that the orbitofrontal cortex represented the hidden states of a decision task with 16 states (Schuck et al., 2016). A 2022 review set the two traditions side by side, found the value-only account challenged by the map view, and examined how the two might be reconciled; the orbitofrontal cognitive map is therefore contested rather than established (Knudsen & Wallis, 2022).

The medial and orbital frontal cortex is not one region. Without presenting new data, a review of work in rodents, macaques and humans gave several medial and orbital frontal subregions, from the orbitofrontal cortex to the dorsal anterior cingulate cortex, their own distinct parts in behavioural flexibility and decision making (Klein-Flügge et al., 2022). The ventromedial prefrontal cortex itself serves several roles, from social cognition, theory of mind included, to value-based choice, and some of its functions cut across psychiatric diagnoses (Hiser & Koenigs, 2018).

The interactive demonstration below is a simplified, illustrative version of the gambling task. Readers draw cards from four decks, two that pay more per card but lose over time and two that pay less but gain, and follow the running total.

Interactive demonstration · Reward Now, Penalty Later

The Iowa Gambling Task

Draw cards from any deck. Two decks pay a large reward on every card but hide larger penalties, and two pay a small reward but only small penalties. Nothing here is random: each deck runs a fixed schedule that repeats every ten cards. In the classic studies, healthy participants came to favour the paler, winning decks while patients with ventromedial damage kept choosing the richer, losing ones; later work found many healthy participants never settle on the winning decks.

Net: +0
No cards drawn yet. Deck A returns about negative 25 per card over its cycle and deck C about positive 25, though the reward on each single card points the other way.
Interactive demonstration. An illustrative version of the gambling task used to probe ventromedial prefrontal function (after Bechara, Damasio, Damasio, & Anderson, 1994). The fixed schedules are illustrative and cycle every ten cards: decks A and B pay more per card but lose 250 per ten cards, and decks C and D pay less but gain 250, so play is fully deterministic. What the task measures is contested: participants can report knowledge of the decks (Maia & McClelland, 2004), and healthy participants' choices are far more varied than the classic summary implies (Steingroever et al., 2013).

Neuromodulation: Dopamine and Stress

Prefrontal control depends on neuromodulatory input. Reviewing primate studies, Ott and Nieder assigned dopamine three major computational roles in prefrontal cognitive control, among them contributions to working memory and to the gating of sensory information; the microcircuits they describe are proposed and corroborated by models rather than directly demonstrated (Ott & Nieder, 2019).

Stress acts through related chemistry. Arnsten's review described how uncontrollable stress rapidly impairs prefrontal function through catecholamines, whose influence on prefrontal cognition follows an inverted U, and traced the effect through intracellular pathways including cAMP-HCN channel signalling and protein kinase C signalling, with the amygdala also implicated (Arnsten, 2009). Much of the evidence comes from spatial working memory tasks such as the oculomotor delayed response task, some of the mechanisms are proposed rather than established, and ethical limits restrict how far uncontrollability can be manipulated in human studies, where procedures such as the Trier social stress test are used instead (Arnsten, 2009).

Development Across the Lifespan

A widely cited longitudinal study scanned 13 healthy children every two years between the ages of 4 and 21, producing 52 scans, and found that gray matter matured from the back of the cortex to the front, with the dorsolateral prefrontal cortex among the last regions to mature (Gogtay et al., 2004). The sample was small and unrepresentative, with an average IQ of 125; sex differences were not analysed, and the link to synaptic pruning was speculative (Gogtay et al., 2004). Behaviour shows a related timetable. On oculomotor tasks given to 245 healthy 8- to 30-year-olds, adult levels arrived at about 14 years for voluntary response suppression, about 15 for processing speed and about 19 for spatial working memory (Luna et al., 2004). Although each ability followed a course that was partly its own, gains in processing speed fed into working-memory development, and inhibition and working memory matured in mutual dependence (Luna et al., 2004).

Larger samples have refined the picture. A multisample study pooling 388 participants and 854 scans from four longitudinal samples, aged 7 to 29 years, found that cortical thinning accounted for most of the loss of cortical volume during adolescence, including in prefrontal cortex (Tamnes et al., 2017). Annual rates of change differed significantly between samples, scanners and acquisition protocols varied, and morphometry offers only indirect clues to the underlying biology (Tamnes et al., 2017).

These patterns fit a broader proposal that cortical maturation proceeds along a sensorimotor-to-association axis, with association cortices retaining plasticity for longest; and the review cites accumulating evidence that this protracted plasticity also confers risk for developmental psychopathology (Sydnor et al., 2021). Larsen and Luna went further, proposing that adolescence is a critical period for the prefrontal development that underlies higher-order cognition, governed by mechanisms of developmental plasticity; the review presents this as a hypothesis that the correspondence it found supports (Larsen & Luna, 2018). For behaviour, Casey's review explained adolescent risk taking through diminished self-control arising from developing brain circuitry under evolutionary and biological constraints, while noting that diminished self-control accounts for only part of adolescent mortality (Casey, 2015).

At the largest scale, brain charts built from 123,984 MRI scans of 101,457 participants in more than 100 primary studies, spanning 115 days post-conception to 100 years, describe non-linear trajectories of brain morphology and supply centile scores for individuals (Bethlehem et al., 2022). The charts concern brain morphology in general rather than the prefrontal cortex specifically, and the MRI studies they draw on are biased relative to the global population (Bethlehem et al., 2022).

Lesions and Clinical Evidence

Lesions remain the most direct test of whether a region is necessary. Evidence from focal damage in humans and nonhuman primates indicates that the prefrontal cortex is required for goal-directed control (Szczepanski & Knight, 2014). A 2019 review argued that permanent lesions provide evidence distinct from that of reversible manipulations and remain essential in contemporary neuroscience, while discussing their inferential limitations (Vaidya et al., 2019).

Large lesion samples can dissociate functions that imaging finds intertwined. Voxel-based lesion-symptom mapping in 344 patients with focal damage, 165 of whom had lesions involving the PFC, revealed two separable networks, one supporting cognitive control and the other value-based decision making (Gläscher et al., 2012). Control was measured through set shifting, monitoring of conflict and the suppression of responses (Gläscher et al., 2012). The regions implicated included frontopolar, ventromedial prefrontal and orbitofrontal cortex as well as the anterior cingulate and dorsolateral prefrontal cortex (Gläscher et al., 2012). The patients were drawn from a dataset assembled over several decades (Gläscher et al., 2012).

The most famous prefrontal case has been re-read. Combining older evidence with new material, Macmillan and Lena argued that Phineas Gage recovered and achieved good psychosocial adaptation, and they discussed his recovery in terms of rehabilitation supported by external structure; their sources strongly imply this rather than prove it (Macmillan & Lena, 2010).

Clinically, the prefrontal cortex is a target for transcranial magnetic stimulation in depression. In a double-blind, sham-controlled trial of Stanford neuromodulation therapy, 29 patients with treatment-resistant depression received active (N = 14) or sham (N = 15) intermittent theta-burst stimulation of the left dorsolateral prefrontal cortex, targeted with resting-state functional MRI on the basis of its relation to the subgenual anterior cingulate cortex, and active treatment reduced Montgomery-Åsberg Depression Rating Scale scores more than sham (Cole et al., 2022). The result comes from a planned interim analysis of a small sample, the durability of the effect is unknown, and the therapy has not been compared with other treatments (Cole et al., 2022). A larger double-blind, sham-controlled trial by the same group took up the small-sample limitation, randomizing 48 participants with treatment-resistant depression to active (N=24) or sham (N=24) treatment (Kratter et al., 2026). At one month, the primary outcome, 50.0% of the active group were in remission against 20.8% of the sham group, and response rates likewise favoured active treatment, at 54.2% versus 25.0% (Kratter et al., 2026). Electroencephalography showed frontal beta power falling after active stimulation but not after sham, and beta power in the left anterior cingulate cortex at baseline is a candidate, not an established, predictor of who benefits (Kratter et al., 2026).

Table 2 summarises the major accounts discussed above and the standing each currently holds.

Table 2. Major Accounts of Prefrontal Function and Their Current Standing

AccountCore claimCurrent standing
Integrative theory of prefrontal cortex functionMaintained goal representations in prefrontal cortex bias processing in other regions (Miller & Cohen, 2001).Revised: flexible control is now placed in interacting prefrontal networks and in shared plus separable control processes (Menon & D'Esposito, 2022; Friedman & Robbins, 2022).
Persistent-activity model of working memorySustained delay-period spiking of prefrontal neurons holds remembered content (Constantinidis et al., 2018).Contested: single-trial spiking appears brief, sparse and changing, though still important (Lundqvist et al., 2018).
Activity-silent working memoryContent can be carried by hidden synaptic states without elevated firing (Stokes, 2015).Contested: hard to test and mostly monkey evidence; recent work stresses active burst dynamics in control (Lundqvist et al., 2023).
Somatic marker hypothesisEmotion-related bodily signals guide choice, with ventromedial prefrontal cortex central (Damasio, 1996).Contested: conscious deck knowledge, ambiguous physiology and heterogeneous healthy performance (Maia & McClelland, 2004; Dunn et al., 2006; Steingroever et al., 2013).
Rostro-caudal hierarchy of cognitive controlLateral prefrontal control forms a cascade from premotor toward anterior regions (Koechlin et al., 2003).Revised: apex in mid-lateral cortex; strict abstraction gradient not supported (Nee & D'Esposito, 2016; Badre & Nee, 2018).
Multiple-demand systemFrontal and parietal regions are engaged across diverse demands and tied to fluid intelligence (Duncan, 2010).Supported and refined: a parcellated domain-general core with only modest task preferences (Assem et al., 2020).
Unity and diversity model of executive functionsA common executive factor plus updating-specific and shifting-specific factors (Friedman & Miyake, 2017).Supported in adult samples, though no single factor model fits every sample (Karr et al., 2018).
Conflict monitoringThe anterior cingulate registers processing conflict and signals a need for more control (Botvinick et al., 2001).Revised: performance monitoring became one of the functions a later single-function account of the dorsal anterior cingulate sets out to explain (Shenhav et al., 2013).
Expected value of controlDorsal anterior cingulate weighs the expected payoff of control against its effort cost (Shenhav et al., 2013).Supported as a normative proposal (Shenhav et al., 2013).
Orbitofrontal cognitive mapOrbitofrontal cortex represents the current task state, including hidden states (Wilson et al., 2014; Schuck et al., 2016).Contested against value coding; reconciliation open (Knudsen & Wallis, 2022).
Meta-reinforcement learning account of prefrontal cortexDopamine-trained prefrontal circuitry comes to learn how to learn (Wang et al., 2018).Contested: a simulation-based proposal with qualitative, unfitted effects (Wang et al., 2018).

Note. Standings summarise the most recent appraisal cited in the body sections; contested accounts retain active defenders.

The Prefrontal Cortex in Psychiatric and Neurodevelopmental Conditions

A neuroimaging-centred review ties abnormalities in prefrontal circuits to the functions compromised in several childhood neurodevelopmental and psychiatric disorders, among them ADHD, obsessive-compulsive disorder and major depressive disorder, drawing on patient studies as well as on work in animals and in typically developing children (Arnsten & Rubia, 2012). The stimulation trials in treatment-resistant depression described in the lesion section, which targeted the left dorsolateral prefrontal cortex, are attempts to act on such a circuit directly (Cole et al., 2022).

ADHD

Functional imaging does not confine ADHD to the loop linking the prefrontal cortex and the striatum. An activation likelihood estimation meta-analysis pooled 55 fMRI studies, with approximate totals of 741 participants with ADHD and 801 comparison participants, and found altered activity across several large-scale networks, the frontoparietal and default networks among them, rather than in prefrontal-striatal circuits alone (Cortese et al., 2012). That analysis had too few studies for some finer comparisons, such as single tasks within each age group, and it left out studies that examined only preselected regions (Cortese et al., 2012). On the structural side, a cross-sectional MRI analysis by the multi-site ENIGMA-ADHD consortium found lower cortical surface area in children with ADHD, but no differences in adolescents or adults, and the authors described the differences as subtle (Hoogman et al., 2019). They took the pattern as confirming that the frontal cortex is involved in the disorder (Hoogman et al., 2019).

Schizophrenia

In schizophrenia, a meta-analysis comparing adult patients with healthy controls on executive tasks such as the N-back and the Stroop found reduced activation in the thalamus, the anterior cingulate cortex and the left dorsolateral prefrontal cortex (Minzenberg et al., 2009). Patients also showed greater activity in some midline regions, which may be compensatory, and the effects of task type were modest (Minzenberg et al., 2009). Structural data point the same way: pooling 4474 people with schizophrenia and 5098 healthy volunteers from 39 centres, the ENIGMA Schizophrenia Working Group found that patients had thinner cortex and smaller cortical surface area (van Erp et al., 2018). Antipsychotic medication may account for part of the thinning, and the reduction in area appeared global, although effects on both measures were largest in frontal and temporal regions (van Erp et al., 2018).

Addiction

For addiction, a review of a decade of functional neuroimaging argued that the disorder involves the prefrontal cortex and not just subcortical limbic reward regions (Goldstein & Volkow, 2011). The authors grouped the resulting deficits under the label iRISA, for impaired response inhibition and salience attribution: a weakened capacity to stop unwanted responses, combined with a distorted sense of which things matter, which together undermine self-control (Goldstein & Volkow, 2011). A later review of human and non-human primate studies, drawing on structural as well as functional imaging, suggests that drug use may itself contribute to prefrontal damage, a causal reading its authors base on converging human and non-human primate evidence (Ceceli et al., 2022). The same review treats prefrontal plasticity during abstinence as evidence that recovery may be possible (Ceceli et al., 2022).

Recovery During Abstinence

Structural imaging offers one test of whether that recovery occurs. When 19 treatment-seeking individuals with cocaine use disorder and 12 healthy controls were rescanned after six months, during which the patients had markedly reduced or stopped their drug use, gray matter volume had increased in the patients' left inferior frontal gyrus and ventromedial prefrontal cortex but had not changed in the controls, and larger volume gains in the inferior frontal gyrus went with better cognitive flexibility (Parvaz et al., 2017). A voxel based morphometry comparison of 43 abstinent cocaine users with 43 controls pointed the same way, with more gray matter among those abstinent for longer and less among those with more years of use; because the design was cross-sectional, however, it cannot show that abstinence caused the difference, and the effects may have predated abstinence (Connolly et al., 2013). A comparable pattern appears in alcohol dependence, where repeated MRI of 85 alcohol-dependent adults showed volumes in several regions, the dorsolateral prefrontal cortex among them, increasing over abstinence at a rate that was not constant from one interval to the next (Zou et al., 2018). That sample was mostly male, and attrition meant that only 22 of the alcohol-dependent participants contributed scans at every time point (Zou et al., 2018).

An early dual-tracer PET study suggests that some deficits outlast early abstinence: cocaine abusers showed lower dopamine D2 receptor availability than controls, the reduction tracked lower metabolism in frontal regions, and the deficit was still present 3-4 months after detoxification (Volkow et al., 1993). Repeat PET scans of five methamphetamine abusers found that the loss of dopamine transporters had partly reversed after 12-17 months of abstinence, yet the accompanying cognitive improvement was not significant, and so small a sample limits any conclusion (Volkow et al., 2001). Reactivity to drug cues follows a time course of its own: in an animal model, rats trained to self-administer cocaine sought the drug more strongly in response to its cues over the first 2 months of withdrawal, a progressive strengthening termed incubation of craving (Grimm et al., 2001). In humans, a cross-sectional EEG study of 76 cocaine-addicted adults measured the late positive potential, a brain response that indexes motivated attention to salient pictures, and found it larger for cocaine-related images at 1 and 6 months of abstinence than at 2 days, 1 week or 1 year, even as self-reported craving fell steadily with longer abstinence; the groups were separate and self-selected rather than the same people followed over time (Parvaz et al., 2016). Taken together, these studies suggest that reactivity to drug cues may peak weeks to months into abstinence rather than at its start (Parvaz et al., 2016; Grimm et al., 2001).

Interpreting the Clinical Evidence

These results describe group averages, not individuals: the ADHD surface-area differences were subtle, limited to children and drawn from a cross-sectional design whose authors call for longitudinal follow-up (Hoogman et al., 2019). Medication complicates the picture, both because it may confound the cortical thickness findings in schizophrenia, which also come from a case-control comparison (van Erp et al., 2018), and because it alters prefrontal chemistry in treated patients (Arnsten & Rubia, 2012). Abnormalities in prefrontal circuits are reported in childhood conditions as different as ADHD, obsessive-compulsive disorder, conduct disorder and major depressive disorder, although the review links each to a distinctive circuit (Arnsten & Rubia, 2012).

Worked Example

The arithmetic below is illustrative. It uses round numbers. In this version, two high-paying decks give 100 per card but carry penalties averaging 125 per card, so each draw from them is worth 100 - 125 = -25 on average. Two low-paying decks give 50 per card with penalties averaging 25, so each draw is worth 50 - 25 = +25.

Over 40 draws, a player who chose only from the high-paying decks would expect 40 × (-25) = -1,000, and a player who chose only from the low-paying decks would expect 40 × (+25) = +1,000. The 2,000-point gap between these strategies arises entirely from the penalties, because the immediate reward on every card favours the losing decks by 100 to 50.

That asymmetry explains why a learner who tracks immediate gains drifts toward the bad decks. Each card from a high-paying deck delivers twice the visible reward, and penalties arrive intermittently, so their cumulative cost becomes apparent only after enough losses have accrued. The bad decks are bad in expectation, not on every draw.

The arithmetic cannot, however, decide between the somatic marker account and its critics. Learning to avoid the high-paying decks is compatible with an implicit bodily warning, but it is equally compatible with explicit reasoning about payoffs, and participants questioned during the task could report which decks were advantageous (Maia & McClelland, 2004). The calculation also treats convergence on the good decks as the norm, whereas reanalyses of healthy participants found their choices heterogeneous, so failure to converge is not by itself a sign of impairment (Steingroever et al., 2013). The interactive gambling task in the decision-making section above offers an illustrative version that readers can play.

Discussion

Research on the prefrontal cortex has moved through three framings. Early work localised functions to areas and cells, exemplified by delay-period neurons that coded remembered locations (Funahashi et al., 1989) and neurons that coded abstract rules (Wallis et al., 2001). A second framing treated the region as the source of top-down bias that makes control possible (Miller & Cohen, 2001). The current framing is distributed: control is attributed to interacting networks (Menon & D'Esposito, 2022) and to the geometry of population activity (Ebitz & Hayden, 2021). Each framing absorbed rather than discarded the previous one, which is why single-cell findings remain the anchor of debates now conducted in population terms.

Two claims are well supported. Demanding tasks of many kinds recruit a common frontoparietal core, now mapped at the level of individual parcels (Assem et al., 2020), and individual differences in executive functions show a common factor alongside specific ones in adults (Friedman & Miyake, 2017), although no single factor model fits every sample (Karr et al., 2018). Even these are refinements in progress: the core's regions show modest task preferences, and the inhibition factor has behaved inconsistently. Four questions remain open. The substrate of working memory is disputed, with persistent firing set against sparse, transient bursts (Lundqvist et al., 2018). The somatic marker hypothesis lacks the causal and psychophysiological support its critics require (Dunn et al., 2006). The hierarchy of control no longer rests on a single abstraction gradient; in the most recent appraisal, interacting frontal networks have taken its place (Badre & Nee, 2018). And the orbitofrontal cortex is read by some as a value region and by others as a map of task states (Knudsen & Wallis, 2022).

The designs that would settle these questions are identifiable even where they have not been run. Large-scale simultaneous recordings analysed trial by trial would address the objection that averaging manufactures persistence (Lundqvist et al., 2018); causal perturbation of caudal frontal regions would test the dominance relations a hierarchy predicts (Badre & D'Esposito, 2009); and tasks whose states cannot be observed directly are needed to separate map coding from value coding (Wilson et al., 2014).

Beneath all of these lies the definitional problem. Comparative claims change with the boundaries drawn (Donahue et al., 2018), rodent nomenclature remains inconsistent (Laubach et al., 2018), and the cross-species definition of the region is unresolved (Carlén, 2017). Answers to functional questions therefore carry an implicit qualifier about which cortex, in which species, is meant.

Current Directions

The representational geometry of prefrontal codes is the most active front. Human fMRI and network modelling point to low-dimensional, orthogonal context codes in dorsolateral prefrontal and parietal cortex, in tension with earlier evidence for high-dimensional mixed selectivity (Flesch et al., 2022), and the proposal that dimensionality trades separability against generalisation frames that tension as a design choice rather than a contradiction, with the relevant experiments still to come (Badre et al., 2021). Abstract geometries shared by prefrontal cortex and hippocampus suggest how such codes might support generalisation (Bernardi et al., 2020).

Network-level models are moving from description to prediction, with activity-flow approaches proposed as a way to derive a region's representations from its connections (Ito et al., 2020). Meta-learning accounts recast prefrontal cortex as a learned learning system, but they remain simulation-based (Wang et al., 2018). In working memory, spatial computing proposes that bursts organised across the cortical surface implement control, on evidence that is largely reanalysed (Lundqvist et al., 2023), and a shared prefrontal code for selection in memory and attention was found in two monkeys (Panichello & Buschman, 2021).

Mouse studies are expanding the toolkit for testing primate-derived theories (Le Merre et al., 2021). Developmental work increasingly situates prefrontal maturation on a sensorimotor-to-association axis (Sydnor et al., 2021) and benchmarks individuals against lifespan brain charts (Bethlehem et al., 2022). Clinically, connectivity-guided stimulation of left dorsolateral prefrontal cortex outperformed sham in a small interim analysis whose durability is unknown (Cole et al., 2022), and a larger sham-controlled replication found higher remission at one month with active treatment (Kratter et al., 2026).

Common Misconceptions

The prefrontal cortex finishes maturing at a fixed age.
Cortical change continues across a broad age range: a multisample longitudinal study tracked thinning-driven volume loss across participants aged 7 to 29 (Tamnes et al., 2017), and lifespan brain charts describe non-linear trajectories rather than a single endpoint (Bethlehem et al., 2022). The belief draws on a real finding, that prefrontal gray matter matured late in an early longitudinal study, but that study followed only 13 children of high average IQ (Gogtay et al., 2004).
Humans are known to have a uniquely, disproportionately large prefrontal cortex.
The question is contested. An MRI comparison of hominoids found the human frontal cortex was not disproportionately large relative to great apes (Semendeferi et al., 2002), whereas a multimodal comparison with chimpanzees and macaques found the human prefrontal cortex disproportionately large, while noting that its boundaries involved judgment and that only three species were compared (Donahue et al., 2018).
Phineas Gage never recovered from his injury.
A historical re-examination concluded that Gage recovered and achieved good psychosocial adaptation, although the evidence strongly implies this rather than proving it (Macmillan & Lena, 2010).
Working memory is simply sustained firing in prefrontal neurons.
Trial-by-trial analyses suggest that working-memory spiking is sparse and transient, with averaging across trials creating an appearance of persistence (Lundqvist et al., 2018), and one review cites human and monkey evidence that content is distributed across sensory, parietal and prefrontal cortex (Christophel et al., 2017). The belief rests on genuine delay-period neurons recorded in monkeys, which remain part of the evidence.

Glossary

Activity-silent working memory.
A proposed form of working-memory storage in which information is held in hidden synaptic states, such as short-term synaptic plasticity, between periods of elevated firing; the mechanism is hard to test and its functional role has not yet been fully established.
Allometric scaling.
The relation across species between the size of a brain structure and that of a common reference, such as another cortical area, used to ask whether a structure is larger than expected for a given species.
Cognitive control.
The processes that configure perception, memory and action in line with current goals, including maintaining task rules, shifting between them and inhibiting inappropriate responses.
Domain-general.
Engaged across many kinds of task rather than specialised for one kind of content, such as language or faces; the term is applied to the frontoparietal multiple-demand system.
Dorsolateral prefrontal cortex.
The upper lateral part of the prefrontal cortex, associated with implementing control, executive attention and delay-period coding in working memory; it is one node in wider networks rather than an isolated controller.
Executive functions.
Control abilities that are correlated but separable in adult samples, usually modelled as a common factor plus specific factors for updating and shifting; results for a separate inhibition factor have been inconsistent.
Frontopolar cortex.
The anterior tip of the prefrontal cortex, also called the anterior prefrontal cortex or Brodmann area 10, implicated mainly by monkey studies in tracking alternative goals; its role in humans is speculative.
Hypofrontality.
Lower-than-typical activation of frontal cortex during a task, reported for example in imaging studies of executive tasks in schizophrenia, where it can occur alongside increased activity in other regions.
Iowa Gambling Task.
A card-choice task in which some decks yield larger immediate rewards but net losses, used to study decision making after ventromedial damage; its assumptions about healthy performance are disputed.
iRISA.
Short for impaired response inhibition and salience attribution, a framework that describes addiction as a weakened ability to stop unwanted responses combined with excessive weight placed on drug-related cues over other rewards.
Mixed selectivity.
The tendency of single neurons to respond to nonlinear combinations of several task variables, which raises the dimensionality of the population code; its original authors suggested it is the most useful but also the most fragile form of selectivity.
Multimodal parcellation.
Dividing the cortex into distinct areas by combining several kinds of brain imaging data rather than relying on a single measure.
Multiple-demand system.
A set of frontal and parietal regions recruited across many kinds of demanding task and linked to fluid intelligence, whose regions show only modest relative preferences among tasks.
Orbitofrontal cortex.
Prefrontal cortex on the ventral surface above the orbits, studied as a site of economic choice and value coding and, on a contested alternative view, as a map of task states.
Persistent activity.
Sustained neuronal firing across a delay, long taken as the mechanism of working memory; whether it reflects steady firing on single trials or an average over brief bursts is contested.
Population doctrine.
The view, gaining ground beside the single-neuron doctrine, that neural computation is best described through the joint activity of populations; its impact has been greatest in motor neuroscience.
Representational geometry.
The arrangement of task conditions in the space of population activity, including its dimensionality and the angles between coding directions, which determines what a downstream readout can separate or generalise.
Rostro-caudal axis.
The front-to-back axis of the brain; in prefrontal research, the proposal that more anterior regions handle more abstract or temporally extended control, an ordering that later work found weaker than first proposed.
Somatic marker hypothesis.
Damasio's contested proposal that emotion-related bodily signals, acting covertly or overtly, guide decision making, with the ventromedial prefrontal cortex as a key structure.
Ventromedial prefrontal cortex.
The lower medial prefrontal region, implicated in valuation among several other functions, whose subregions appear to serve distinct roles.
Working memory.
The brief maintenance and manipulation of information that is no longer present; its neural substrate, whether persistent firing, bursts, synaptic states or distributed codes, is disputed.

Key Researchers

Amy F. T. Arnsten

(b. 1954). Yale School of Medicine (Albert E. Kent Professor of Neuroscience and Professor of Psychology); neuroscientist of the stress-signalling pathways that impair prefrontal structure and function. Wikipedia - Wikidata - Faculty Page

David Badre

Brown University (Department of Cognitive, Linguistic and Psychological Sciences); cognitive neuroscientist of hierarchical control along the rostro-caudal axis of the frontal lobe. Google Scholar - Faculty Page

Jonathan D. Cohen

Princeton University (Princeton Neuroscience Institute); computational neuroscientist; co-author of the integrative theory of prefrontal function and of the expected-value-of-control account. ORCID - Wikipedia - Wikidata - Google Scholar - Faculty Page

Mark D'Esposito

Helen Wills Neuroscience Institute and Department of Psychology, University of California, Berkeley; cognitive neuroscientist of lateral prefrontal hierarchy and of prefrontal networks in executive function. ORCID - Google Scholar - Faculty Page

John Duncan

MRC Cognition and Brain Sciences Unit, University of Cambridge; cognitive neuroscientist who defined the multiple-demand system of frontal and parietal cortex. Wikipedia - Wikidata - Google Scholar - Faculty Page

Naomi P. Friedman

University of Colorado Boulder (Institute for Behavioral Genetics); behavioural geneticist and psychologist; co-developer of the unity and diversity model of executive functions. ORCID - Google Scholar - Faculty Page

Patricia Goldman-Rakic

(1937-2003). Department of Neurobiology, Yale University School of Medicine; primate neurophysiologist whose delay-task recordings made persistent prefrontal activity the classic account of working memory. Wikipedia

Etienne Koechlin

Ecole Normale Superieure, Paris (Laboratoire de Neurosciences Cognitives); cognitive neuroscientist who proposed a cascade architecture of cognitive control in prefrontal cortex. ORCID - Wikidata - Google Scholar - Faculty Page

Beatriz Luna

University of Pittsburgh (Staunton Professor of Psychiatry and Pediatrics; Laboratory of Neurocognitive Development); developmental cognitive neuroscientist of adolescence as a critical period for higher-order cognition. ORCID - Wikipedia - Wikidata - Google Scholar - Faculty Page

Earl K. Miller

Massachusetts Institute of Technology (Picower Institute for Learning and Memory); systems neuroscientist; co-author of the integrative theory of prefrontal function and of the burst-based account of working memory. ORCID - Wikipedia - Wikidata - Google Scholar - Faculty Page

Yael Niv

Princeton University (Professor, Princeton Neuroscience Institute and Department of Psychology); computational neuroscientist; co-author of the account of orbitofrontal cortex as a cognitive map of task space. ORCID - Wikipedia - Wikidata - Google Scholar - Faculty Page

Camillo Padoa-Schioppa

Washington University in St. Louis; neurophysiologist of economic choice and the orbitofrontal circuit for value-based decisions. ORCID - Google Scholar - Faculty Page

Frequently Asked Questions

What does the prefrontal cortex do?

The prefrontal cortex supports the control of thought and action in pursuit of goals. Lesion studies in human patients and nonhuman primates indicate that the region is necessary for this kind of goal-directed control (Szczepanski & Knight, 2014).

Is the prefrontal cortex the seat of working memory?

Not exclusively. One review cites studies in humans and monkeys that found working memory content stored across sensory, parietal and prefrontal cortex, and asks future studies to examine how these regions work together (Christophel et al., 2017).

At what age does the prefrontal cortex finish developing?

No single age is supported. A multisample longitudinal study of participants aged 7 to 29 found cortical thinning and volume loss continuing across adolescence, with rates differing between samples (Tamnes et al., 2017).

Do humans have an unusually large prefrontal cortex?

The question is contested. One multimodal MRI comparison of humans, chimpanzees and macaques found the human prefrontal cortex disproportionately large, while noting that only three species were compared and that drawing its boundaries required judgment (Donahue et al., 2018).

What is the multiple-demand system?

It is a set of frontal and parietal regions engaged across many kinds of demanding task. A parcellation of 449 Human Connectome Project participants identified a domain-general core of 10 parcels per hemisphere whose regions show only modest task preferences (Assem et al., 2020).

Is the somatic marker hypothesis accepted?

It remains contested: a critical review judged that the hypothesis still lacks adequate support, pointing to scant causal evidence for peripheral bodily feedback, to psychophysiological data open to more than one reading, and to an Iowa Gambling Task whose schedule participants can grasp cognitively (Dunn et al., 2006).

Did Phineas Gage recover after his injury?

A historical re-examination concluded that Gage recovered and achieved good psychosocial adaptation, although the evidence strongly implies rather than proves this (Macmillan & Lena, 2010).

Can stimulating the prefrontal cortex treat depression?

In a small double-blind trial of 29 patients with treatment-resistant depression, active stimulation of the left dorsolateral prefrontal cortex reduced depression ratings more than sham. The result came from an interim analysis, and its durability is unknown (Cole et al., 2022). A larger sham-controlled replication by the same group, with 48 participants, found higher remission at one month with active treatment (Kratter et al., 2026).

References

Alexander, W. H., & Brown, J. W. (2011). Medial prefrontal cortex as an action-outcome predictor. Nature Neuroscience, 14(10), 1338-1344. https://doi.org/10.1038/nn.2921

Arnsten, A. F. T. (2009). Stress signalling pathways that impair prefrontal cortex structure and function. Nature Reviews Neuroscience, 10(6), 410-422. https://doi.org/10.1038/nrn2648

Arnsten, A. F. T., & Rubia, K. (2012). Neurobiological circuits regulating attention, cognitive control, motivation, and emotion: Disruptions in neurodevelopmental psychiatric disorders. Journal of the American Academy of Child & Adolescent Psychiatry, 51(4), 356-367. https://doi.org/10.1016/j.jaac.2012.01.008

Assem, M., Glasser, M. F., Van Essen, D. C., & Duncan, J. (2020). A domain-general cognitive core defined in multimodally parcellated human cortex. Cerebral Cortex, 30(8), 4361-4380. https://doi.org/10.1093/cercor/bhaa023

Badre, D., & D'Esposito, M. (2009). Is the rostro-caudal axis of the frontal lobe hierarchical? Nature Reviews Neuroscience, 10(9), 659-669. https://doi.org/10.1038/nrn2667

Badre, D., & Nee, D. E. (2018). Frontal cortex and the hierarchical control of behavior. Trends in Cognitive Sciences, 22(2), 170-188. https://doi.org/10.1016/j.tics.2017.11.005

Badre, D., Bhandari, A., Keglovits, H., & Kikumoto, A. (2021). The dimensionality of neural representations for control. Current Opinion in Behavioral Sciences, 38, 20-28. https://doi.org/10.1016/j.cobeha.2020.07.002

Bartra, O., McGuire, J. T., & Kable, J. W. (2013). The valuation system: A coordinate-based meta-analysis of BOLD fMRI experiments examining neural correlates of subjective value. NeuroImage, 76, 412-427. https://doi.org/10.1016/j.neuroimage.2013.02.063

Bechara, A., Damasio, A. R., Damasio, H., & Anderson, S. W. (1994). Insensitivity to future consequences following damage to human prefrontal cortex. Cognition, 50(1-3), 7-15. https://doi.org/10.1016/0010-0277(94)90018-3

Bernardi, S., Benna, M. K., Rigotti, M., Munuera, J., Fusi, S., & Salzman, C. D. (2020). The geometry of abstraction in the hippocampus and prefrontal cortex. Cell, 183(4), 954-967.e21. https://doi.org/10.1016/j.cell.2020.09.031

Bethlehem, R. A. I., Seidlitz, J., White, S. R., Vogel, J. W., Anderson, K. M., Adamson, C., Adler, S., Alexopoulos, G. S., Anagnostou, E., Areces-Gonzalez, A., et al. (2022). Brain charts for the human lifespan. Nature, 604(7906), 525-533. https://doi.org/10.1038/s41586-022-04554-y

Botvinick, M. M., Braver, T. S., Barch, D. M., Carter, C. S., & Cohen, J. D. (2001). Conflict monitoring and cognitive control. Psychological Review, 108(3), 624-652. https://doi.org/10.1037/0033-295X.108.3.624

Carlén, M. (2017). What constitutes the prefrontal cortex? Science, 358(6362), 478-482. https://doi.org/10.1126/science.aan8868

Casey, B. J. (2015). Beyond simple models of self-control to circuit-based accounts of adolescent behavior. Annual Review of Psychology, 66, 295-319. https://doi.org/10.1146/annurev-psych-010814-015156

Ceceli, A. O., Bradberry, C. W., & Goldstein, R. Z. (2022). The neurobiology of drug addiction: Cross-species insights into the dysfunction and recovery of the prefrontal cortex. Neuropsychopharmacology, 47(1), 276-291. https://doi.org/10.1038/s41386-021-01153-9

Christophel, T. B., Klink, P. C., Spitzer, B., Roelfsema, P. R., & Haynes, J.-D. (2017). The distributed nature of working memory. Trends in Cognitive Sciences, 21(2), 111-124. https://doi.org/10.1016/j.tics.2016.12.007

Cole, E. J., Phillips, A. L., Bentzley, B. S., Stimpson, K. H., Nejad, R., Barmak, F., Veerapal, C., Khan, N., Cherian, K., Felber, E., Brown, R., Choi, E., King, S., Pankow, H., Bishop, J. H., Azeez, A., Coetzee, J., Rapier, R., Odenwald, N., ... Williams, N. R. (2022). Stanford Neuromodulation Therapy (SNT): A double-blind randomized controlled trial. American Journal of Psychiatry, 179(2), 132-141. https://doi.org/10.1176/appi.ajp.2021.20101429

Cole, M. W., Reynolds, J. R., Power, J. D., Repovs, G., Anticevic, A., & Braver, T. S. (2013). Multi-task connectivity reveals flexible hubs for adaptive task control. Nature Neuroscience, 16(9), 1348-1355. https://doi.org/10.1038/nn.3470

Connolly, C. G., Bell, R. P., Foxe, J. J., & Garavan, H. (2013). Dissociated grey matter changes with prolonged addiction and extended abstinence in cocaine users. PLoS ONE, 8(3), Article e59645. https://doi.org/10.1371/journal.pone.0059645

Constantinidis, C., Funahashi, S., Lee, D., Murray, J. D., Qi, X. L., Wang, M., & Arnsten, A. F. T. (2018). Persistent spiking activity underlies working memory. Journal of Neuroscience, 38(32), 7020-7028. https://doi.org/10.1523/jneurosci.2486-17.2018

Cortese, S., Kelly, C., Chabernaud, C., Proal, E., Di Martino, A., Milham, M. P., & Castellanos, F. X. (2012). Toward systems neuroscience of ADHD: A meta-analysis of 55 fMRI studies. American Journal of Psychiatry, 169(10), 1038-1055. https://doi.org/10.1176/appi.ajp.2012.11101521

Curtis, C. E., & D'Esposito, M. (2003). Persistent activity in the prefrontal cortex during working memory. Trends in Cognitive Sciences, 7(9), 415-423. https://doi.org/10.1016/S1364-6613(03)00197-9

Damasio, A. R. (1996). The somatic marker hypothesis and the possible functions of the prefrontal cortex. Philosophical Transactions of the Royal Society of London. Series B: Biological Sciences, 351(1346), 1413-1420. https://doi.org/10.1098/rstb.1996.0125

Donahue, C. J., Glasser, M. F., Preuss, T. M., Rilling, J. K., & Van Essen, D. C. (2018). Quantitative assessment of prefrontal cortex in humans relative to nonhuman primates. Proceedings of the National Academy of Sciences, 115(22), E5183-E5192. https://doi.org/10.1073/pnas.1721653115

Duncan, J. (2010). The multiple-demand (MD) system of the primate brain: Mental programs for intelligent behaviour. Trends in Cognitive Sciences, 14(4), 172-179. https://doi.org/10.1016/j.tics.2010.01.004

Duncan, J., & Owen, A. M. (2000). Common regions of the human frontal lobe recruited by diverse cognitive demands. Trends in Neurosciences, 23(10), 475-483. https://doi.org/10.1016/s0166-2236(00)01633-7

Dunn, B. D., Dalgleish, T., & Lawrence, A. D. (2006). The somatic marker hypothesis: A critical evaluation. Neuroscience & Biobehavioral Reviews, 30(2), 239-271. https://doi.org/10.1016/j.neubiorev.2005.07.001

Ebitz, R. B., & Hayden, B. Y. (2021). The population doctrine in cognitive neuroscience. Neuron, 109(19), 3055-3068. https://doi.org/10.1016/j.neuron.2021.07.011

Fedorenko, E., & Blank, I. A. (2020). Broca's area is not a natural kind. Trends in Cognitive Sciences, 24(4), 270-284. https://doi.org/10.1016/j.tics.2020.01.001

Fedorenko, E., Duncan, J., & Kanwisher, N. (2013). Broad domain generality in focal regions of frontal and parietal cortex. Proceedings of the National Academy of Sciences, 110(41), 16616-16621. https://doi.org/10.1073/pnas.1315235110

Flesch, T., Juechems, K., Dumbalska, T., Saxe, A., & Summerfield, C. (2022). Orthogonal representations for robust context-dependent task performance in brains and neural networks. Neuron, 110(7), 1258-1270.e11. https://doi.org/10.1016/j.neuron.2022.01.005

Friedman, N. P., & Miyake, A. (2017). Unity and diversity of executive functions: Individual differences as a window on cognitive structure. Cortex, 86, 186-204. https://doi.org/10.1016/j.cortex.2016.04.023

Friedman, N. P., & Robbins, T. W. (2022). The role of prefrontal cortex in cognitive control and executive function. Neuropsychopharmacology, 47(1), 72-89. https://doi.org/10.1038/s41386-021-01132-0

Funahashi, S., Bruce, C. J., & Goldman-Rakic, P. S. (1989). Mnemonic coding of visual space in the monkey's dorsolateral prefrontal cortex. Journal of Neurophysiology, 61(2), 331-349. https://doi.org/10.1152/jn.1989.61.2.331

Fuster, J. M. (2001). The prefrontal cortex—an update: Time is of the essence. Neuron, 30(2), 319-333. https://doi.org/10.1016/s0896-6273(01)00285-9

Glasser, M. F., Coalson, T. S., Robinson, E. C., Hacker, C. D., Harwell, J., Yacoub, E., Ugurbil, K., Andersson, J., Beckmann, C. F., Jenkinson, M., Smith, S. M., & Van Essen, D. C. (2016). A multi-modal parcellation of human cerebral cortex. Nature, 536(7615), 171-178. https://doi.org/10.1038/nature18933

Gläscher, J., Adolphs, R., Damasio, H., Bechara, A., Rudrauf, D., Calamia, M., Paul, L. K., & Tranel, D. (2012). Lesion mapping of cognitive control and value-based decision making in the prefrontal cortex. Proceedings of the National Academy of Sciences, 109(36), 14681-14686. https://doi.org/10.1073/pnas.1206608109

Gogtay, N., Giedd, J. N., Lusk, L., Hayashi, K. M., Greenstein, D., Vaituzis, A. C., Nugent, T. F., Herman, D. H., Clasen, L. S., Toga, A. W., Rapoport, J. L., & Thompson, P. M. (2004). Dynamic mapping of human cortical development during childhood through early adulthood. Proceedings of the National Academy of Sciences, 101(21), 8174-8179. https://doi.org/10.1073/pnas.0402680101

Goldman-Rakic, P. S. (1995). Cellular basis of working memory. Neuron, 14(3), 477-485. https://doi.org/10.1016/0896-6273(95)90304-6

Goldstein, R. Z., & Volkow, N. D. (2011). Dysfunction of the prefrontal cortex in addiction: Neuroimaging findings and clinical implications. Nature Reviews Neuroscience, 12(11), 652-669. https://doi.org/10.1038/nrn3119

Grimm, J. W., Hope, B. T., Wise, R. A., & Shaham, Y. (2001). Incubation of cocaine craving after withdrawal. Nature, 412(6843), 141-142. https://doi.org/10.1038/35084134

Hampshire, A., Highfield, R. R., Parkin, B. L., & Owen, A. M. (2012). Fractionating human intelligence. Neuron, 76(6), 1225-1237. https://doi.org/10.1016/j.neuron.2012.06.022

Hiser, J., & Koenigs, M. (2018). The multifaceted role of the ventromedial prefrontal cortex in emotion, decision making, social cognition, and psychopathology. Biological Psychiatry, 83(8), 638-647. https://doi.org/10.1016/j.biopsych.2017.10.030

Hoogman, M., Muetzel, R., Guimaraes, J. P., Shumskaya, E., Mennes, M., Zwiers, M. P., Jahanshad, N., Sudre, G., Wolfers, T., Earl, E. A., Soliva Vila, J. C., Vives-Gilabert, Y., Khadka, S., Novotny, S. E., Hartman, C. A., Heslenfeld, D. J., Schweren, L. J. S., Ambrosino, S., Oranje, B., ... Franke, B. (2019). Brain imaging of the cortex in ADHD: A coordinated analysis of large-scale clinical and population-based samples. American Journal of Psychiatry, 176(7), 531-542. https://doi.org/10.1176/appi.ajp.2019.18091033

Ito, T., Hearne, L., Mill, R., Cocuzza, C., & Cole, M. W. (2020). Discovering the computational relevance of brain network organization. Trends in Cognitive Sciences, 24(1), 25-38. https://doi.org/10.1016/j.tics.2019.10.005

Kane, M. J., & Engle, R. W. (2002). The role of prefrontal cortex in working-memory capacity, executive attention, and general fluid intelligence: An individual-differences perspective. Psychonomic Bulletin & Review, 9(4), 637-671. https://doi.org/10.3758/BF03196323

Karr, J. E., Areshenkoff, C. N., Rast, P., Hofer, S. M., Iverson, G. L., & Garcia-Barrera, M. A. (2018). The unity and diversity of executive functions: A systematic review and re-analysis of latent variable studies. Psychological Bulletin, 144(11), 1147-1185. https://doi.org/10.1037/bul0000160

Klein-Flügge, M. C., Bongioanni, A., & Rushworth, M. F. S. (2022). Medial and orbital frontal cortex in decision-making and flexible behavior. Neuron, 110(17), 2743-2770. https://doi.org/10.1016/j.neuron.2022.05.022

Knudsen, E. B., & Wallis, J. D. (2022). Taking stock of value in the orbitofrontal cortex. Nature Reviews Neuroscience, 23(7), 428-438. https://doi.org/10.1038/s41583-022-00589-2

Koechlin, E., Ody, C., & Kouneiher, F. (2003). The architecture of cognitive control in the human prefrontal cortex. Science, 302(5648), 1181-1185. https://doi.org/10.1126/science.1088545

Kratter, I. H., Austelle, C. W., Lissemore, J. I., Wada, M., Geoly, A., Chaiken, A., Kaloiani, I., Johnson, N., Wan, S., Kozyr, L., Makarewycz, E., Wong, B., Sridhar, M., Espil, F. M., Bassano, N., Kim, B., Ehrie, J., Maron-Katz, A., Tischler, C., ... Williams, N. R. (2026). Stanford neuromodulation therapy for treatment-resistant depression: A randomized controlled trial confirming efficacy, and an EEG study providing insight into mechanism of action and a potentially predictive biomarker of efficacy. World Psychiatry, 25(1), 105-116. https://doi.org/10.1002/wps.70032

Larsen, B., & Luna, B. (2018). Adolescence as a neurobiological critical period for the development of higher-order cognition. Neuroscience & Biobehavioral Reviews, 94, 179-195. https://doi.org/10.1016/j.neubiorev.2018.09.005

Laubach, M., Amarante, L. M., Swanson, K., & White, S. R. (2018). What, if anything, is rodent prefrontal cortex? eNeuro, 5(5), ENEURO.0315-18.2018. https://doi.org/10.1523/ENEURO.0315-18.2018

Le Merre, P., Ährlund-Richter, S., & Carlén, M. (2021). The mouse prefrontal cortex: Unity in diversity. Neuron, 109(12), 1925-1944. https://doi.org/10.1016/j.neuron.2021.03.035

Luna, B., Garver, K. E., Urban, T. A., Lazar, N. A., & Sweeney, J. A. (2004). Maturation of cognitive processes from late childhood to adulthood. Child Development, 75(5), 1357-1372. https://doi.org/10.1111/j.1467-8624.2004.00745.x

Lundqvist, M., Brincat, S. L., Rose, J., Warden, M. R., Buschman, T. J., Miller, E. K., & Herman, P. (2023). Working memory control dynamics follow principles of spatial computing. Nature Communications, 14(1), 1429. https://doi.org/10.1038/s41467-023-36555-4

Lundqvist, M., Herman, P., & Miller, E. K. (2018). Working memory: Delay activity, yes! Persistent activity? Maybe not. The Journal of Neuroscience, 38(32), 7013-7019. https://doi.org/10.1523/JNEUROSCI.2485-17.2018

Lundqvist, M., Rose, J., Herman, P., Brincat, S. L., Buschman, T. J., & Miller, E. K. (2016). Gamma and beta bursts underlie working memory. Neuron, 90(1), 152-164. https://doi.org/10.1016/j.neuron.2016.02.028

MacDonald, A. W., Cohen, J. D., Stenger, V. A., & Carter, C. S. (2000). Dissociating the role of the dorsolateral prefrontal and anterior cingulate cortex in cognitive control. Science, 288(5472), 1835-1838. https://doi.org/10.1126/science.288.5472.1835

Macmillan, M., & Lena, M. L. (2010). Rehabilitating Phineas Gage. Neuropsychological Rehabilitation, 20(5), 641-658. https://doi.org/10.1080/09602011003760527

Maia, T. V., & McClelland, J. L. (2004). A reexamination of the evidence for the somatic marker hypothesis: What participants really know in the Iowa gambling task. Proceedings of the National Academy of Sciences, 101(45), 16075-16080. https://doi.org/10.1073/pnas.0406666101

Mansouri, F. A., Koechlin, E., Rosa, M. G. P., & Buckley, M. J. (2017). Managing competing goals - a key role for the frontopolar cortex. Nature Reviews Neuroscience, 18(11), 645-657. https://doi.org/10.1038/nrn.2017.111

Mante, V., Sussillo, D., Shenoy, K. V., & Newsome, W. T. (2013). Context-dependent computation by recurrent dynamics in prefrontal cortex. Nature, 503(7474), 78-84. https://doi.org/10.1038/nature12742

Menon, V., & D'Esposito, M. (2022). The role of PFC networks in cognitive control and executive function. Neuropsychopharmacology, 47(1), 90-103. https://doi.org/10.1038/s41386-021-01152-w

Miller, E. K., & Cohen, J. D. (2001). An integrative theory of prefrontal cortex function. Annual Review of Neuroscience, 24, 167-202. https://doi.org/10.1146/annurev.neuro.24.1.167

Minzenberg, M. J., Laird, A. R., Thelen, S., Carter, C. S., & Glahn, D. C. (2009). Meta-analysis of 41 functional neuroimaging studies of executive function in schizophrenia. Archives of General Psychiatry, 66(8), 811-822. https://doi.org/10.1001/archgenpsychiatry.2009.91

Miyake, A., & Friedman, N. P. (2012). The nature and organization of individual differences in executive functions: Four general conclusions. Current Directions in Psychological Science, 21(1), 8-14. https://doi.org/10.1177/0963721411429458

Miyake, A., Friedman, N. P., Emerson, M. J., Witzki, A. H., Howerter, A., & Wager, T. D. (2000). The unity and diversity of executive functions and their contributions to complex "frontal lobe" tasks: A latent variable analysis. Cognitive Psychology, 41(1), 49-100. https://doi.org/10.1006/cogp.1999.0734

Nee, D. E., & D'Esposito, M. (2016). The hierarchical organization of the lateral prefrontal cortex. eLife, 5, e12112. https://doi.org/10.7554/eLife.12112

Ott, T., & Nieder, A. (2019). Dopamine and cognitive control in prefrontal cortex. Trends in Cognitive Sciences, 23(3), 213-234. https://doi.org/10.1016/j.tics.2018.12.006

Padoa-Schioppa, C., & Assad, J. A. (2006). Neurons in the orbitofrontal cortex encode economic value. Nature, 441(7090), 223-226. https://doi.org/10.1038/nature04676

Padoa-Schioppa, C., & Conen, K. E. (2017). Orbitofrontal cortex: A neural circuit for economic decisions. Neuron, 96(4), 736-754. https://doi.org/10.1016/j.neuron.2017.09.031

Panichello, M. F., & Buschman, T. J. (2021). Shared mechanisms underlie the control of working memory and attention. Nature, 592(7855), 601-605. https://doi.org/10.1038/s41586-021-03390-w

Panichello, M. F., Jonikaitis, D., Oh, Y. J., Zhu, S., Trepka, E. B., & Moore, T. (2024). Intermittent rate coding and cue-specific ensembles support working memory. Nature, 636(8042), 422-429. https://doi.org/10.1038/s41586-024-08139-9

Parvaz, M. A., Moeller, S. J., & Goldstein, R. Z. (2016). Incubation of cue-induced craving in adults addicted to cocaine measured by electroencephalography. JAMA Psychiatry, 73(11), 1127-1134. https://doi.org/10.1001/jamapsychiatry.2016.2181

Parvaz, M. A., Moeller, S. J., d'Oleire Uquillas, F., Pflumm, A., Maloney, T., Alia-Klein, N., & Goldstein, R. Z. (2017). Prefrontal gray matter volume recovery in treatment-seeking cocaine-addicted individuals: A longitudinal study. Addiction Biology, 22(5), 1391-1401. https://doi.org/10.1111/adb.12403

Passingham, R. E., & Smaers, J. B. (2014). Is the prefrontal cortex especially enlarged in the human brain? Allometric relations and remapping factors. Brain, Behavior and Evolution, 84(2), 156-166. https://doi.org/10.1159/000365183

Petrides, M. (2005). Lateral prefrontal cortex: Architectonic and functional organization. Philosophical Transactions of the Royal Society B: Biological Sciences, 360(1456), 781-795. https://doi.org/10.1098/rstb.2005.1631

Rigotti, M., Barak, O., Warden, M. R., Wang, X.-J., Daw, N. D., Miller, E. K., & Fusi, S. (2013). The importance of mixed selectivity in complex cognitive tasks. Nature, 497(7451), 585-590. https://doi.org/10.1038/nature12160

Rose, N. S., LaRocque, J. J., Riggall, A. C., Gosseries, O., Starrett, M. J., Meyering, E. E., & Postle, B. R. (2016). Reactivation of latent working memories with transcranial magnetic stimulation. Science, 354(6316), 1136-1139. https://doi.org/10.1126/science.aah7011

Schuck, N. W., Cai, M. B., Wilson, R. C., & Niv, Y. (2016). Human orbitofrontal cortex represents a cognitive map of state space. Neuron, 91(6), 1402-1412. https://doi.org/10.1016/j.neuron.2016.08.019

Semendeferi, K., Lu, A., Schenker, N., & Damasio, H. (2002). Humans and great apes share a large frontal cortex. Nature Neuroscience, 5(3), 272-276. https://doi.org/10.1038/nn814

Shallice, T., & Burgess, P. W. (1991). Deficits in strategy application following frontal lobe damage in man. Brain, 114(2), 727-741. https://doi.org/10.1093/brain/114.2.727

Shenhav, A., Botvinick, M. M., & Cohen, J. D. (2013). The expected value of control: An integrative theory of anterior cingulate cortex function. Neuron, 79(2), 217-240. https://doi.org/10.1016/j.neuron.2013.07.007

Steingroever, H., Wetzels, R., Horstmann, A., Neumann, J., & Wagenmakers, E.-J. (2013). Performance of healthy participants on the Iowa Gambling Task. Psychological Assessment, 25(1), 180-193. https://doi.org/10.1037/a0029929

Stokes, M. G. (2015). 'Activity-silent' working memory in prefrontal cortex: A dynamic coding framework. Trends in Cognitive Sciences, 19(7), 394-405. https://doi.org/10.1016/j.tics.2015.05.004

Sydnor, V. J., Larsen, B., Bassett, D. S., Alexander-Bloch, A., Fair, D. A., Liston, C., Mackey, A. P., Milham, M. P., Pines, A., Roalf, D. R., Seidlitz, J., Xu, T., Raznahan, A., & Satterthwaite, T. D. (2021). Neurodevelopment of the association cortices: Patterns, mechanisms, and implications for psychopathology. Neuron, 109(18), 2820-2846. https://doi.org/10.1016/j.neuron.2021.06.016

Szczepanski, S. M., & Knight, R. T. (2014). Insights into human behavior from lesions to the prefrontal cortex. Neuron, 83(5), 1002-1018. https://doi.org/10.1016/j.neuron.2014.08.011

Tamnes, C. K., Herting, M. M., Goddings, A.-L., Meuwese, R., Blakemore, S.-J., Dahl, R. E., Güroğlu, B., Raznahan, A., Sowell, E. R., Crone, E. A., & Mills, K. L. (2017). Development of the cerebral cortex across adolescence: A multisample study of inter-related longitudinal changes in cortical volume, surface area, and thickness. The Journal of Neuroscience, 37(12), 3402-3412. https://doi.org/10.1523/JNEUROSCI.3302-16.2017

Vaidya, A. R., Pujara, M. S., Petrides, M., Murray, E. A., & Fellows, L. K. (2019). Lesion studies in contemporary neuroscience. Trends in Cognitive Sciences, 23(8), 653-671. https://doi.org/10.1016/j.tics.2019.05.009

van Erp, T. G. M., Walton, E., Hibar, D. P., Schmaal, L., Jiang, W., Glahn, D. C., Pearlson, G. D., Yao, N., Fukunaga, M., Hashimoto, R., Okada, N., Yamamori, H., Bustillo, J. R., Clark, V. P., Agartz, I., Mueller, B. A., Cahn, W., de Zwarte, S. M. C., Hulshoff Pol, H. E., ... Orhan, F. (2018). Cortical brain abnormalities in 4474 individuals with schizophrenia and 5098 control subjects via the Enhancing Neuro Imaging Genetics Through Meta Analysis (ENIGMA) consortium. Biological Psychiatry, 84(9), 644-654. https://doi.org/10.1016/j.biopsych.2018.04.023

Volkow, N. D., Chang, L., Wang, G.-J., Fowler, J. S., Franceschi, D., Sedler, M., Gatley, S. J., Miller, E., Hitzemann, R., Ding, Y.-S., & Logan, J. (2001). Loss of dopamine transporters in methamphetamine abusers recovers with protracted abstinence. Journal of Neuroscience, 21(23), 9414-9418. https://doi.org/10.1523/JNEUROSCI.21-23-09414.2001

Volkow, N. D., Fowler, J. S., Wang, G.-J., Hitzemann, R., Logan, J., Schlyer, D. J., Dewey, S. L., & Wolf, A. P. (1993). Decreased dopamine D2 receptor availability is associated with reduced frontal metabolism in cocaine abusers. Synapse, 14(2), 169-177. https://doi.org/10.1002/syn.890140210

Wallis, J. D., Anderson, K. C., & Miller, E. K. (2001). Single neurons in prefrontal cortex encode abstract rules. Nature, 411(6840), 953-956. https://doi.org/10.1038/35082081

Wang, J. X., Kurth-Nelson, Z., Kumaran, D., Tirumala, D., Soyer, H., Leibo, J. Z., Hassabis, D., & Botvinick, M. (2018). Prefrontal cortex as a meta-reinforcement learning system. Nature Neuroscience, 21(6), 860-868. https://doi.org/10.1038/s41593-018-0147-8

Wilson, R. C., Takahashi, Y. K., Schoenbaum, G., & Niv, Y. (2014). Orbitofrontal cortex as a cognitive map of task space. Neuron, 81(2), 267-279. https://doi.org/10.1016/j.neuron.2013.11.005

Wolff, M. J., Jochim, J., Akyürek, E. G., & Stokes, M. G. (2017). Dynamic hidden states underlying working-memory-guided behavior. Nature Neuroscience, 20(6), 864-871. https://doi.org/10.1038/nn.4546

Zou, X., Durazzo, T. C., & Meyerhoff, D. J. (2018). Regional brain volume changes in alcohol-dependent individuals during short-term and long-term abstinence. Alcoholism: Clinical and Experimental Research, 42(6), 1062-1072. https://doi.org/10.1111/acer.13757