Prefrontal contributions to the stability and variability of thought and conscious experience

This review (2021) examines how the prefrontal cortex and other brain networks influence the variability and stability of mental phenomena, such as executive functions, mind-wandering, and psychedelic experiences. Specifically, they highlight how different brain networks contribute to these dynamics in the short and long term while acknowledging that the stability of conscious experiences are also contingent upon the stability or variability of the internal and external environments. Since most research on psychedelics has mostly focussed on investigating large-scale brain networks, the authors conclude that future research should also study how specific regions contribute to the variability and stability of conscious experiences depending on their functional specialization.


“The human prefrontal cortex is a structurally and functionally heterogenous brain region, including multiple subregions that have been linked to different large-scale brain networks. It contributes to a broad range of mental phenomena, from goal-directed thought and executive functions to mind-wandering and psychedelic experience. Here we review what is known about the functions of different prefrontal subregions and their affiliations with large-scale brain networks to examine how they may differentially contribute to the diversity of mental phenomena associated with prefrontal function. An important dimension that distinguishes across different kinds of conscious experience is the stability or variability of mental states across time. This dimension is a central feature of two recently introduced theoretical frameworks—the dynamic framework of thought (DFT) and the relaxed beliefs under psychedelics (REBUS) model—that treat neurocognitive dynamics as central to understanding and distinguishing between different mental phenomena. Here, we bring these two frameworks together to provide a synthesis of how prefrontal subregions may differentially contribute to the stability and variability of thought and conscious experience. We close by considering future directions for this work.”

Authors: Andre Zamani, Robin Carhart-Harris & Kalina Christoff



The prefrontal cortex is a broad swath of brain tissue encompassing numerous cytoarchitecturally and functionally heterogenous subregions. These subregions are thought to influence cognition in relatively distinct ways, and can be distinguished based on their neural, cognitive, and phenomenological correlates.

The DFT and REBUS models distinguish between stable and variable thought dynamics, and suggest that serotonergic psychedelics induce a heightened variability of conscious experience through their effect on brain-wide hierarchical information flow.

We combine insights afforded by two frameworks to understand how different prefrontal subregions may contribute to increased stability or variability in thought and conscious experience.


The dynamic framework of thought (DFT) is a conceptual framework for understanding human thought that takes into account the dynamics of mental states, in addition to their contents.

The DFT proposes two general ways to alter the dynamics of thought: deliberate and automatic constraints. These constraints increase the stability or variability of thought and conscious experience.

Deliberate constraints are exemplified by cognitive control and executive processes that are supported by the brain’s control networks. These constraints increase stability of thought over time by restricting thought contents and the transitions between them, typically in the service of an explicit goal.

In certain contexts, such as creative thinking, deliberate constraints may contribute to an increased variability of thought, but only indirectly. These constraints may be related to executive capacity and the ability to reject uncreative ideas or to implement more creative but also more executively taxing idea generation strategies.

Automatic constraints, such as affective salience, sensory salience, and habits, can constrain thought variability by making certain thought contents persist across multiple mental states, but can also increase the variability of thought over time by supporting rapid switches of attentional focus between different stimuli.

When both deliberate and automatic constraints are relatively low, thoughts arise more “spontaneously” and are marked by variability. These thoughts are generated through wayward transitions in the stream of thought, which may generate a stronger subjective experience of spontaneity.

Inates the thought stream, a wider range of mental states are likely to occur with less predictable transitions between them. This may result in increased variability of thoughts over time.

The DFT model proposes that dynamic interactions between different large scale brain networks underlie deliberate and automatic constraints, as well as spontaneous thought processes.

Relaxed beliefs under psychedelics (REBUS)

The REBUS model proposes a unifying account of how serotonergic psychedelic compounds affect conscious experience. These compounds include lysergic acid diethylamide (LSD), psilocybin, and dimethyltryptamine (DMT).

Psychedelics are thought to increase the variability of conscious experience by significantly altering the brain’s neurocognitive hierarchies of information flow. This is accomplished by relaxing the constraining influence that beliefs at higher levels of the brain’s neurocognitive hierarchies have on bottom-up information.

REBUS uses the term “belief” more broadly, as a synonym for the more technical Bayesian term ‘prior’, referring to predictions encoded in neuronal connections and activity.

The brain is a predictive system that uses numerous hierarchical levels to predict the input it is about to receive from its respective lower levels. The brain minimizes prediction error by adjusting its predictions or adjusting incoming information so that it may better fit its predictions.

The entropic brain hypothesis states that the degree of entropy in spontaneous brain activity correlates with the diversity of subjective conscious experience. Psychedelic states are thought to have highly entropic underlying brain activity, whereas clinically relevant states such as rumination and obsessive thought have low entropy.

Psychedelics increase bottom-up information flow by decreasing the compressing influence of implicit top-down beliefs, leading to an overall increased variability of conscious experience.

Synergies between the frameworks

The DFT and REBUS model both describe cognitive control as a process that stabilizes thought and conscious experience, and both posit that there are processes outside of cognitive control that contribute to increased stability of thought through some form of automatic influence.

The presence of increased constraints or implicit beliefs, such as affective salience, does not necessarily result in decreased variability of thought and conscious experience. Instead, some automatic constraints, such as affective salience, may contribute to increased variability of thought and conscious experience.


We organize the evidence in response to this question around the central concepts from DFT and REBUS: deliberate constraints, automatic constraints, and top-down beliefs. We focus on prefrontal subregions that are part of the frontoparietal control network.

Even though we focus on specific prefrontal subregions, each large-scale brain network is affiliated with multiple regions and brain structures inside as well as outside of the PFC that also make crucial contributions to stability and variability of thought.

Deliberate constraints

The brain’s frontoparietal and cingulo-opercular control networks support the flexible, top-down implementation of cognitive control and executive function through dynamically coupling with other brain networks.

Deliberate constraints may contribute to increased stability of thought by recruiting multiple prefrontal control network regions, including the dorsolateral PFC, rostrolateral PFC, and dorsal anterior cingulate cortex.

Frontoparietal control network may contribute to stronger deliberate constraints through the maintenance and implementation of rules, which are implemented through the PFC’s biasing influences on the neural activity of other regions throughout the brain.

Some rules are linked to memory contents rather than external environmental contextual cues, and thus depend upon memory retrieval for their implementation. The DLPFC and RLPFC support this online maintenance of rule-outcome associations.

The DLPFC is associated with anticipating cue onset during rulebased behavior, and with the ability to realize delayed intentions. It may also underlie the deliberate shielding of long-term goals from possible distractors.

The RLPFC appears to act as an intermediary between the frontoparietal and cingulo-opercular control networks, representing multiple rule-outcome associations at once and helping to guide adjustments to cognitive control. It is also associated with anticipating and planning for future failures of rule-adherence.

The DLPFC and RLPFC may work together to increase deliberate constraints upon thought by implementing memory-derived rules based upon the expected value of their outcomes. Thought suppression may be one example of this.

In one study, participants with high-constraint nouns displayed increased functional connectivity between the DLPFC and default network core, which was interpreted as evidence of increased top-down inhibition to reject uncreative verbs.

Other studies show the DLPFC is closely associated with thought suppression, and that it influences the hippocampus to bias the stream of thought toward contents in line with current goals and task demands.

The RLPFC may affect thought dynamics through its representation of multiple rule-outcome associations. Increased RLPFC activation and functional connectivity with the anterior medial PFC node of the default network core are associated with increased creativity.

DLPFC and RLPFC are connected differently with the default network core, and this difference is consistent with the previously described difference in function between DLPFC and RLPFC according to level of abstraction in cognitive control.

The frontoparietal control network may contribute to increased variability in thought during creative thinking by driving brain dynamics toward unique network configurations that favor the generation of highly creative ideas, although the control networks may also contribute to increased stability at more global timescales.

The cingulo-opercular control network supports deliberate constraints in ways that are distinct yet complimentary to that of the frontoparietal control network. It may contribute to increased stability in thought and conscious experience by scaffolding mental contents through abstract goal representations.

The dACC is thought to be involved in deliberate constraints, including monitoring performance, responding to conflict, and adjusting behavior. This function is enabled by the dACC’s interactions with control, motor, and visceral neural systems.

The dACC is connected to several cortical motor regions as well as the spinal cord, and its neural activation exhibits a somatotopic organization. The dACC may also interact with the anterior insula to bias viscero-somatic processing to prepare the body for actions.

Anorexia nervosa patients exhibit increased connectivity between the default network and the retrosplenial cortex of the medial temporal lobe, which may account for their increased tendency to engage in outcome-oriented imagination about their body.

In summary, the dACC may contribute to increased stability of conscious experience by aligning outward actions with one’s goals. However, in an uncertain or rapidly changing environment, the dACC may contribute to increased variability of conscious experience.

Automatic constraints

Automatic constraints are associated with multiple brain networks, including the default network, the salience network, and the ventral attentional network.

Around the turn of the 21st century, a set of brain regions was identified that underlie a “default mode” of human brain function.

Scientific understanding of the default network has evolved considerably in the last two decades, and now includes a core subcomponent, medial temporal lobe subcomponent, and third (or dorsomedial) subcomponent.

The core subcomponent of the default network (DNCORE) is highly interconnected with many other brain regions and is associated with a diverse range of functions, including mnemonic elaboration and self-referential thinking.

The anterior medial PFC is the anterior-most node of the default network core and is closely associated with abstract self-referential processes. Its activation increases as a function of the self-relatedness of ongoing processing and is diminished when considering an adjective that applies to an intimate other.

The aMPFC is involved in the processing of relatively abstract self-related information, such as adjectives and personality traits, and may thus underlie self-processing at a more schematic level.

Schemas are complex and abstract knowledge structures formed by extracting and generalizing statistical regularities across previous individual experiences. The aMPFC may contribute to an increased stability of thought and conscious experience by automatically constraining thought contents using abstract self-referential schemas toward personally significant information.

Functional connectivity between the aMPFC and DNMTL has been observed during self-focused rumination in healthy controls and depressed individuals, suggesting that the aMPFC may increase automatic constraints on thought.

The medial temporal lobe component of the default network is associated with episodic thinking, the generation of specific, contextual, and visuo-spatially detailed thought contents, as well as the generation of spontaneous thoughts.

The ventromedial prefrontal cortex (vMPFC) may support increased automatic constraints on thought, in contrast to the medial temporal lobe structures.

The mOFC appears to underlie emotional evaluations of internally generated events based upon their relevance to current goals and needs. This view is supported by multiple lines of evidence, including strong anatomical connections to memory-related regions in the DNMTL and increased neural activation during internally-oriented cognitions.

Activation in the mOFC correlates with the degree of rated familiarity and anticipated pleasantness for future episodic simulations, rated pleasantness for recalled episodic memories, and reported motivation to engage in autobiographical reflection. This suggests that the mOFC supports the elaboration of thoughts in value-dependent ways.

The mOFC is closely linked to schema-guided memory processes, including autobiographical memories and fictitious thinking. Lesions to the mOFC disrupt the ability to elaborate cued memories into extended sequences of thought.

Thought that is heavily schema-dependent may rely on interactions between the mOFC and the hippocampus during initial construction, and may also help bolster hippocampal encoding for uncertain memories.

We suggest that the mOFC may contribute to increased stability in thought and conscious experience by automatically constraining cognition and behavior around highly valued schemas.

The third subcomponent of the default network, the dorsomedial subcomponent, is involved in conceptual processing, mentalizing, and constructive mental simulation, of which abstract social cognition is just one.

The dorsomedial PFC is activated during abstract sociocognitive processes, including visual perspective taking, evaluating why versus how someone performed a behavior, differentiating in-group versus out-group individuals, and forming impressions of others. However, it is also activated during abstract non-social processes.

The dMPFC appears to underlie abstract, highly constructive processing more broadly, such as the abstraction of meaning from memories, and especially their emotional meaning. Its activation is associated with greater vividness of recollection when recalling negatively-valenced episodic memories in younger adults but positively-valenced episodic memories in older adults.

The dMPFC is associated with automatic constraints on thought, and the gist of a person is used to guide the internal stream of thought when simulating another person in imagination. This may contribute to increased stability in thought and conscious experience.

In schizophrenia, the automatic constraints afforded by the dMPFC may contribute to increased variability in thought and conscious experience, such as hyper-associative thinking, auditory hallucinations, and an attenuated self-other boundary.

Schizophrenic individuals have hyperactivity in the dMPFC, which may underlie an attenuated self-other boundary, and may process self and other perspectives concurrently, which may facilitate or impair behavior depending upon the context.

Schizophrenic individuals may not segregate high-construal self- and other-information to the same degree as people without schizophrenia, which may disrupt the stabilizing effect that a normative sense of self has over cognition and behavior.

The salience and ventral attention networks are considered to support the early automatic identification of salient information originating exteroceptively, interoceptively, or within the thought stream. This allows important information to become amplified throughout neurocognitive hierarchies through increased attention and cognitive control.

There is considerable overlap in the subregions and functions of the salience network and ventral attention network, but here we treat them as separate networks when discussing how their PFC nodes may contribute to mental state dynamics.

The salience network is composed of the anterior insula and the anterior cingulate cortex. It is thought to support appraisals of viscero-sensory signals based on self-referential and conceptual knowledge, and may contribute to an increased stability of thought and conscious experience.

The rACC exhibits strong anatomical connections to brain regions involved in processing physiological signals, as well as strong anatomical connections to two main DNCORE regions, the anterior medial PFC cortex and the posterior cingulate cortex. This connectivity may uniquely allow the rACC to integrate viscero-sensory signals with self-referential autobiographical knowledge.

A disconnect between viscero-sensory processing and autobiographical self-knowledge may underlie the diminished interpretability of viscero-sensory signals associated with alexithymia, and may lead to increased sensitivity to viscero-sensory signals and anxiety.

The rACC may contribute to greater stability or variability of conscious experience depending on the level of variability in viscero-sensory signals.

The ventral attention network comprises the anterior insula, ventral extent of the temporoparietal junction, and inferior frontal gyrus, and is the prefrontal node of the ventral attention network. The inferior frontal gyrus is involved in the inhibition of motor responses and task-sets.

The right IFG is involved in the detection of unattended and unexpected stimuli in the external environment, and may support automatic constraints for salience detection for information in the external environment.

The right IFG has downstream effects upon the primary motor cortex and subthalamic nucleus of the basal ganglia, and may underlie its role in behavioral inhibition and supporting associations between rules and their expected outcomes toward the deployment of goal-directed behaviors.

The right IFG may contribute to increased variability of thought and conscious experience in environments marked by frequent changes in the salience and relevance of spatially localized external stimuli.

In circumstances where the external environment is stable and relatively invariable, the right IFG may contribute to reducing the variability of conscious experience by directing attention away from the internal stream of thought and toward the external task at-hand.

Top-down beliefs as constraints

Beliefs increase the brain’s processing efficiency by sending unpredicted portions of incoming information upwards along neurocognitive hierarchies. This is why beliefs contribute to an increased stability in thought and conscious experience, and why ego-dissolution is crucial to the increased variability of conscious experience engendered by psychedelic compounds.

Psychedelics disrupt the normative sense of self by attenuating mental and bodily self-experience and by releasing information from lower levels of the neurocognitive hierarchy from the constraining influence of top-down beliefs.

Psilocybin and LSD administration decrease functional connectivity between the aMPFC and posterior cingulate cortex of the DNCORE, and decrease aMPFC activation and blood flow, which are important neural correlates underlying the subjective effects of psychedelics.

Early clinical research on psychedelics suggests that the bodily sense of self is subdued before mental aspects of self-experience become attenuated, and that the rACC may be involved in the phenomenology of self-experience under psychedelics.

The early relaxation of bodily self-beliefs may play a causal role in the subsequent relaxation of mental self-beliefs, and the dual-relaxation of mental and bodily self-beliefs will likely be associated with the deepest form of subjective ego-dissolution.

Psychedelics reduce the capacity for goal-related beliefs to support an increased stability of thought and conscious experience. Goal-related beliefs constrain lower levels of the neurocognitive hierarchy to process incoming information relative to the expectation of goal-completion.

LSD and psilocybin appear to reduce the functional integrity of the control networks, which in turn increases the communication between the control networks and other brain networks, thereby contributing to an increased variability of thought and conscious experience.

Psychedelics seem to reduce the constraining influence of mental self-beliefs, bodily self-beliefs, and goal-related beliefs over lower levels of the neurocognitive hierarchy, thereby increasing variability of thought and conscious experience.


Recent theoretical advances have highlighted the importance of understanding mental dynamics, and the present discussion suggests future directions for research.

The neural constraints imposed by the dACC, rACC, and inferior frontal gyrus may not necessarily result in increased stability of conscious experience over time.

Regional differences in contributing to greater stability or variability in thought and conscious experience could be understood further by examining possible gradients of specialization and functional organization across adjacent cortical regions. For example, future work could examine the potential information processing gradient along the ventral-dorsal axis of the ACC.

To make stronger regionally specific predictions regarding specific prefrontal subregions, neuroscientific work related to large-scale brain networks should use more fine-grained and regionally specific analysis approaches. In particular, psychedelics research should use more specific discrimination between different brain networks and regions.

There are functional gradients of cortical organization and evolutionary gradients of cortical expansion that may hold insights into the phylogenetic development of stability and variability in thought and conscious experience.

The DFT and REBUS model have already led to new assessment tools and interventions for various clinical conditions, including a behavioral assessment for detecting early-stage Alzheimer’s disease and a new model for psychedelic assisted therapy.

A more complete understanding of brain function and organization will require combining different levels of explanation, such as large-scale network analyses and regionally specific functional localizations of brain function.


Authors associated with this publication with profiles on Blossom

Robin Carhart-Harris
Dr. Robin Carhart-Harris is the Founding Director of the Neuroscape Psychedelics Division at UCSF. Previously he led the Psychedelic group at Imperial College London.