Bipolar disorder is typically described in terms of what it looks like from the outside: a person cycling between elevated, expansive states and periods of profound low mood. The clinical categories are accurate. The picture they paint, however, is of a symptom pattern — not of what is happening in the brain that produces it.

Research from the last decade points to something more distributed. Four large-scale brain systems — the circadian timing system, the reward circuit, the default mode network, and the fronto-limbic circuit — are each disrupted in bipolar disorder, and they don’t all move in the same direction at the same time. The result is not one switch flipping between two positions. It is an interconnected architecture under strain at multiple points simultaneously.

The mood-swing frame: useful and incomplete

Describing bipolar disorder as extreme mood swings is clinically accurate as far as it goes. The Diagnostic and Statistical Manual identifies the core symptom clusters — elevated or expansive mood in manic or hypomanic episodes, depressed mood and loss of interest in depressive episodes — and those descriptions give clinicians a reliable way to recognize and classify the condition.

What the description doesn’t capture is mechanism. Naming a symptom pattern tells you what clinicians observe. It doesn’t explain why the same person’s sleep need drops dramatically during a manic episode, then rebounds to excessive in depression. It doesn’t explain why risk-taking accelerates in one state and motivation collapses in another. For those questions, you need to look at the underlying systems.

A 2022 review in *Translational Psychiatry* by Bi, Che, and Bai mapped the neural network picture of bipolar disorder across the neuroimaging literature. Their conclusion: the core neuropathology involves disruption of at least four large-scale networks — the default mode network, the salience network, the central executive network, and the sensorimotor network — alongside specific dysfunction in the fronto-limbic circuit connecting the prefrontal cortex and the amygdala (Bi et al., *Translational Psychiatry*, 2022). Mood swings are what you see at the surface. These network disruptions appear to be generating them.

The circadian system: timing that runs at the wrong speed

One of the most quantified features of bipolar disorder is its relationship to sleep — but the disruption runs deeper than sleep itself. A 2024 review in *Psychiatry and Clinical Neuroscience* by Tonon and colleagues examined the circadian dimension of bipolar disorder across clinical populations.

The figures are striking. During manic episodes, somewhere between 69% and 99% of people with bipolar disorder report a dramatically decreased need for sleep — not insomnia, but a reduction in the hours required to feel rested. During depressive episodes, the pattern reverses: more than 78% show hypersomnia, sleeping far more than usual. The same biological system, running in opposite directions (Tonon et al., *Psychiatry and Clinical Neuroscience*, 2024).

The research points to the body’s master clock as a core site of disruption. Clock genes — including CLOCK, PER1, and PER2 — regulate the 24-hour rhythms that govern not just sleep but hormone release, body temperature, and neural excitability. In bipolar disorder, those genes appear to function differently than in controls, producing what researchers describe as phase advances of up to seven hours during mania and phase delays of four to five hours during depression.

This matters because the circadian system is not just a sleep regulator. It is the temporal frame for the entire biological system. Mood, energy, appetite, and cognition all operate within circadian timing. When the clock runs at the wrong speed, the effects cascade through every process it governs.

The reward circuit: predictions that track the wrong signal

A second system disrupted in bipolar disorder is the reward circuit — specifically, how the brain updates its predictions about what will be rewarding.

In a 2024 fMRI study published in *Biological Psychiatry Global Open Science*, Moningka and Mason compared 21 people with bipolar disorder to 21 matched controls on a probabilistic reward task. The key finding involved the ventral striatum, a region central to reward processing. In the bipolar group, the ventral striatum tracked what the researchers called momentum-biased reward prediction errors — the brain’s estimates of upcoming reward were disproportionately shaped by recent streaks rather than by the underlying statistical probabilities of the task.

Controls showed no such effect. Their striatal signals tracked the task’s actual probability structure. In the bipolar group, recent outcomes carried outsized weight: a run of wins pushed expectations higher than the data warranted; a run of losses pushed them lower (Moningka and Mason, *Biological Psychiatry Global Open Science*, 2024).

Reward prediction errors are the signals the brain uses to update its beliefs about the world. When those signals are systematically biased toward momentum rather than reality, the downstream effects could include the amplified goal-pursuit of mania and the withdrawal characteristic of depression. The study also found that weaker connectivity between the insular cortex and the striatum was associated with more severe manic symptoms, suggesting the circuit’s miscalibration varies across individuals.

This is a different picture from saying that people with bipolar disorder simply feel more euphoric in mania. The reward system’s computational function — its capacity to learn accurately from feedback — appears to operate differently.

The default mode network: self-referential thought, disrupted

When the mind is not focused on an immediate task, a network of regions called the default mode network becomes active. It underlies autobiographical memory — the ongoing narrative of one’s own past — self-referential thinking, and the capacity to mentally simulate future scenarios. It is the system that maintains a coherent sense of self across time.

A 2024 resting-state fMRI study by Zhao and colleagues compared 56 people with bipolar disorder to 71 healthy controls. The bipolar group showed reduced activity and reduced connectivity within the default mode network, with the most consistent findings in the left precuneus — a hub region involved in self-representation and conscious awareness. They also found diminished negative connectivity between the precuneus and subcortical regions including the putamen and insula, a pattern that suggests the default mode network’s normal relationship with other brain systems is altered (Zhao et al., *BMC Psychiatry*, 2024).

The default mode network’s clinical relevance lies in what it does. When it functions normally, it supports the capacity to maintain a coherent self-understanding across different emotional states. In bipolar disorder, the difficulty of holding a stable sense of self across mood states — of recognizing, during a manic episode, that the experience is episodic rather than revelatory — may be connected to how this network is operating even at rest.

The Zhao study is cross-sectional: it captures a snapshot, not a trajectory. Whether the default mode network changes track with mood state transitions or represent a more stable background characteristic is an open question.

The fronto-limbic circuit: the governor under strain

Perhaps the most studied circuit in bipolar disorder is the connection between the prefrontal cortex and the amygdala — what researchers call the fronto-limbic circuit. The amygdala is often described as a fear center, but its role is better characterized as emotional salience: it flags information as emotionally significant, whether that’s a threat, an opportunity, or a social cue that warrants attention.

Under ordinary conditions, the ventrolateral prefrontal cortex (vlPFC) modulates amygdala activity — contextualizing emotional signals and moderating responses when the situation doesn’t call for them. In bipolar disorder, particularly during manic episodes, neuroimaging studies consistently find amygdala hyperactivity alongside reduced activation or functional connectivity in the vlPFC.

The Bi et al. review identifies this as the core neuropathological finding in bipolar disorder: fronto-limbic network disruption that results in emotional signals that are less effectively calibrated to context (Bi et al., *Translational Psychiatry*, 2022). The amygdala isn’t malfunctioning in isolation — it is responding. What’s disrupted is the regulatory circuit that normally integrates that response with what the situation actually calls for.

This framing explains something clinicians observe: the emotional intensity of manic or mixed episodes isn’t always characterized by euphoria. It can be irritability, agitation, or a sense of urgency that the social environment cannot account for. The emotional system is generating signals; the moderating circuit is less effective at grounding them in context.

Four systems, not one switch

The research described here covers four distinct levels of disruption. The circadian system sets the temporal frame of the brain’s operation. The reward circuit determines how the brain updates its expectations from experience. The default mode network underlies self-referential cognition and the continuity of self across time. The fronto-limbic circuit moderates the intensity and context-calibration of emotional responses.

These systems interact, but they are not the same system. Circadian disruption affects neural excitability and hormonal rhythms, which in turn influence reward processing and emotional regulation. Striatal reward prediction errors that amplify momentum could contribute to the accelerating, self-reinforcing quality of manic states. Default mode network disruption may contribute to the difficulty of self-monitoring during those same states — of recognizing that what is being experienced is a mood state rather than an accurate perception of reality.

What the four-system picture adds to the mood-swing framing is a picture of the mechanisms generating those cycles. The observable pattern — episodes of mania and depression — reflects multiple regulatory processes operating differently and interacting with each other in ways the research is still mapping.

For people experiencing symptoms associated with mood dysregulation, the multi-system picture has an implication that goes beyond labeling: the factors that influence each of these systems are not all the same. Sleep timing, how the brain responds to reward feedback, self-monitoring capacity, and emotional context each respond to different conditions. Understanding them as distinct — while recognizing how they interact — may be more informative than treating mood state as a single variable to be managed.

What the evidence does and doesn't establish

The studies cited here have real limitations that should be named directly.

Most neuroimaging studies of bipolar disorder use relatively small samples. The Moningka and Mason reward study had 21 participants per group; the Zhao default mode network study had 56 in the bipolar group. Effect sizes established in small samples often do not replicate at their original magnitude in larger ones. These findings are consistent with each other and with a broader literature, but comprehensive replication is not yet complete.

Cross-sectional designs capture a brain at one moment, not across transitions. Most imaging studies are conducted when participants are in a euthymic state — neither manic nor depressed — to control for acute mood effects. What the networks look like during the transition between states, which may be the period most clinically significant, is less well characterized.

Individual variation is large. Group-level fMRI averages describe the mean of a heterogeneous population. The same diagnostic category encompasses people with different phenotypes, genetics, episode histories, and treatments. A population-level finding does not describe any individual brain.

Finally, the causal direction of most findings is unknown. Disrupted default mode network connectivity in people with bipolar disorder does not, by itself, tell you whether the disruption precedes the disorder, results from it, is shaped by treatment, or reflects a shared upstream cause. The network associations are consistent and replicable; the mechanisms that explain them are still being worked out.