Anhedonia is the clinical word for it: things that used to be rewarding stop being rewarding. Not sadness exactly — an absence where anticipation and pleasure used to be. The standard explanation has been that the brain’s reward circuit is running quiet, and it is a satisfying explanation because it maps the symptom directly onto a mechanism. A meta-analysis of the imaging literature suggests that it is too simple, and possibly wrong in an interesting way.

The Standard Account

Many neuroimaging studies have investigated reward-processing dysfunction in major depressive disorder, and they have led to what the meta-analysis calls a common idea: that depression is associated with blunted responses within the reward circuit, particularly in the ventral striatum.

The ventral striatum is a reasonable place to look. It is central to how the brain registers reward and, importantly, how it anticipates one — and anticipation is much of what appears to be missing in anhedonia.

The trouble was that the link between depression and reward-related responses in other regions kept coming back inconclusive, which is a signal that the frame might be wrong rather than merely incomplete.

It is also worth noting how a summary like that becomes standard. Studies that look where previous studies found something tend to find something there too, and a region that acquires a reputation acquires a disproportionate share of the analyses. That is not misconduct; it is how a literature concentrates, and it is a reason to treat a widely repeated summary as a hypothesis rather than a settled result.

What the Meta-Analysis Did

It pooled 41 whole-brain neuroimaging studies of reward-related responses — a total of 794 patients with major depressive disorder and 803 healthy controls — using a coordinate-based method that asks where results converge across studies rather than reanalysing each one (Translational Psychiatry, 2019).

Whole-brain matters here. A study that examines only the ventral striatum can confirm or fail to confirm something about the ventral striatum; it cannot discover that the interesting thing is happening somewhere else.

The Result That Does Not Fit the Summary

The authors state it directly: their findings argue against the common idea that major depressive disorder is primarily linked to deficits within the reward system.

What they found instead was opposing abnormalities inside the same circuit — hypo-responses in the ventral striatum and hyper-responses in the orbitofrontal cortex.

The orbitofrontal cortex is worth a sentence on its own. It is heavily involved in evaluating how much something is worth — comparing options, updating value when circumstances change, deciding whether an outcome met expectation. An increased response there is not obviously a deficit of anything.

Two regions moving in opposite directions is a different kind of finding from one region going quiet. It suggests the problem may be less about how much signal there is and more about the relationship between parts — which is why the authors propose dysregulated corticostriatal connectivity as the underlying account rather than a regional deficit.

A Second Analysis, Narrowing It Further

A 2026 meta-analysis approached reward processing from a different angle: distinguishing bipolar disorder from major depressive disorder, on the grounds that misdiagnosing one as the other delays appropriate treatment. It searched eight databases and identified 23 fMRI studies totalling 1,264 participants (Psychiatry Research: Neuroimaging, 2026).

Its finding for depression was strikingly specific. Major depressive disorder showed blunted responses only in the bilateral caudate, and only during reward delivery — not during anticipation.

Bipolar disorder, by contrast, showed its differences during reward anticipation: hyperactivity in the left inferior frontal gyrus and insula, with reduced activity on the right, and no consistent pattern during delivery.

Anticipation and delivery being separable is the useful part. ‘Reward processing’ is at least two different operations, and a condition can differ on one without differing on the other.

Why This Matters More Than a Technical Correction

Because the simple version has a moral shadow. If the reward system is simply switched off, then motivation is gone and nothing can be done about it from the inside — a framing that lands badly on someone already being told to try harder.

A dysregulated relationship between regions is a different picture. It does not make anything easier, and it is not a reason for optimism about any individual. But it describes a system doing something disordered rather than a system doing nothing, and those are genuinely different claims about what is happening.

It also explains why motivation can be neurologically difficult rather than absent — why the machinery of wanting can be present and still not deliver.

There is a practical version of this too. A model that locates the problem in one structure invites interventions aimed at that structure; a model that locates it in the relationship between structures points somewhere different. Which model is right is an empirical question that is not yet settled, and the honest position is that the field is still arguing.

What Neither Analysis Establishes

Neither is a diagnostic tool. Both are group comparisons pooled across studies, with wide individual variation inside every group. No imaging test for depression exists, and the 2026 paper’s framing as a search for potential biomarkers is a statement of aspiration, not of current practice.

Neither establishes direction. Whether altered reward processing precedes depression, follows from it, or shares a cause with it is not answered by pooling cross-sectional comparisons.

Meta-analysis inherits what it pools. Coordinate-based methods depend on published peaks, which depend on what was reported and what was published — and a literature that expected to find striatal blunting is a literature that reported striatal results.

What both support is narrower: reward processing in depression is measurably atypical, it is regionally specific, it is not a uniform flattening, and anticipation and consumption behave differently. The broader question of whether a system that has learned one pattern can learn another is covered separately.