The shorthand for bipolar disorder is two states and a switch. Up, down, and an implied normal in between where nothing much is happening. Almost every part of that is a simplification, and the most consequential simplification is the bit in the middle — because the clinical description of the condition explicitly includes something that persists when the episodes are not.

Step One: What the Clinical Description Actually Says

A 2026 systematic review opens by characterising bipolar disorder as a chronic psychiatric illness involving recurrent episodes of mania and depression with intervening euthymic periods — and persistent cognitive and affective dysfunction (Brain Imaging and Behavior, 2026).

Euthymic means the mood is within a typical range. It does not mean nothing is different.

That single word ‘persistent’ is what the two-state picture omits, and it changes the shape of the condition: not an oscillation between two abnormal states with normality between them, but an ongoing difference with episodes superimposed.

It also changes what counts as a good outcome. If the goal is understood as returning to a baseline that the description says is not fully intact, then the measure is set against something the condition does not offer — and the periods between episodes get read as failure rather than as the part of the condition they are.

Step Two: Looking at the Brain at Rest

If something persists between episodes, resting-state imaging is the obvious place to look — measuring how regions communicate when a person is not doing any particular task.

Resting-state imaging asks a specific question. Rather than showing someone a task and measuring the response, it records activity while a person lies still and does nothing in particular, then examines which regions fluctuate together. What it captures is the intrinsic organisation of the system rather than its reaction to a demand — which is the right target for something described as persisting between episodes.

The 2026 review synthesised that literature systematically: PRISMA and Cochrane methods, a PROSPERO-registered protocol, searching PubMed/MEDLINE, Embase and Scopus from inception to 25 February 2025, including adults diagnosed by DSM criteria against healthy controls, across any functional-connectivity analytic approach.

Its central observation about the state of the evidence is the one worth carrying: findings remain heterogeneous across analytic approaches.

That is not a throwaway caveat. It means the answer you get depends partly on the method you use to ask — which is a serious problem for a field trying to identify a reliable signature, and a good reason to distrust any single confident claim about ‘the bipolar brain’.

Step Three: A Question With a Cleaner Answer

Where the imaging literature does converge is on a narrower question, and it is a clinically pointed one: how to distinguish bipolar disorder from major depressive disorder.

That distinction is not academic. A 2026 meta-analysis states plainly that misdiagnosis of bipolar disorder as major depressive disorder delays appropriate treatment — which is why identifying differentiating markers has attracted so much effort.

It searched eight databases and pooled 23 fMRI studies totalling 1,264 participants, focusing on reward processing (Psychiatry Research: Neuroimaging, 2026).

Step Four: The Phase Distinction

The finding turns on a separation that is easy to state and easy to overlook. Reward processing is not one operation — anticipating a reward and receiving one are different, and can be measured separately.

In bipolar disorder, the differences appeared during anticipation: hyperactivity in the left inferior frontal gyrus and insula, with reduced activity in the right inferior frontal gyrus and insula. During reward delivery, no consistent amplification or attenuation was observed.

In major depressive disorder, the pattern was the other way around: blunted responses only in the bilateral caudate, and only during reward delivery.

So two conditions that can present similarly in a depressive episode differ on when in the reward sequence their signature appears. That is a more useful kind of difference than a difference in magnitude, because it does not depend on how severe someone happens to be on the day of scanning.

One caution on reading those regional details. Left-versus-right differences in a pooled meta-analysis are exactly the kind of finding that can reflect how studies were analysed and reported rather than a stable biological fact, and the connectivity review’s warning about heterogeneity across analytic approaches applies here too.

Step Five: What This Does Not Deliver

Not a diagnostic test. The authors frame their findings as offering potential neurobiological biomarkers for differentiation and as possibly contributing to precision psychiatry. Potential and possibly are the operative words; these are group-level averages pooled across studies, with wide variation inside each group.

Not a resolved picture at rest. The connectivity review’s own summary is that findings remain heterogeneous across analytic approaches. A field that gets different answers from different methods has not converged.

Not an account of cause. Every study pooled in both reviews compares groups at a point in time. Nothing establishes whether the differences precede the condition, follow from it, or accompany its treatment.

And not a description of any individual. Neither review supports interpreting one person’s scan, and no imaging is used to diagnose bipolar disorder.

Why the Fuller Description Is Worth Having

Because the two-state picture sets the wrong expectation for the periods between episodes. If euthymia is understood as ‘well’, then persistent cognitive difficulty during it reads as a personal failing rather than as part of what the clinical description already includes.

It also reframes what stability means. A condition with an ongoing component alongside episodic ones is one where the between-times matter — and where the things that act on the whole system, including sleep and routine, are part of the picture rather than peripheral to it.

The reward-processing findings in depression, and why the standard ‘blunted reward’ summary turned out to be too simple there as well, are covered in what the meta-analysis actually found.