What Mental Fatigue Actually Does. Part Three: What Changes in the Network

Abstract

The first two parts of this series described mental fatigue from the outside: what the body does, how long it takes, and whether output actually suffers. This part looks inward, at a 2025 study by Lou and colleagues in PLOS One that asked what happens to the organisation of brain activity itself.

Their approach treats the brain as a network and asks whether fatigue rewires it. The answer is more specific than expected: the global architecture holds, while particular regions lose efficiency. We also examine a second 2025 paper, by Ren and colleagues, which addresses the standardisation problem identified in Part One.

Both are small, recent, and single-laboratory studies. We rate the evidence here as preliminary and explain below exactly why — including a limitation in the network study that its own authors disclose.

1.Treating the Brain as a Network

Most electroencephalography asks how much activity of a given rhythm is present at a given place. Network analysis asks a different question: which regions are working together, and how efficiently is information moving between them.

Lou and colleagues took this approach, arguing that patterns of functional connectivity constitute the neural basis for cognition and behaviour, and that analysing changes in network structure is therefore an essential route to understanding cognitive fatigue. They measured coupling between regions using the weighted phase lag index, a method chosen because it is relatively resistant to a common artefact in which a single source appears at multiple electrodes and creates the illusion of connectivity.

The resulting network was then described using graph theory — the mathematics of nodes and connections. At the global level they examined global efficiency, local efficiency, the clustering coefficient, characteristic path length, and the small-world property.

2.What Small-World Means and Why It Matters Here

A small-world network combines dense local clustering with a handful of long-range shortcuts, so that any node can reach any other in a few steps while most processing stays local. Healthy brains show this signature consistently, and its loss is associated with several neurological and psychiatric conditions.

It is therefore a meaningful negative result that mental fatigue left it intact. The authors report that despite the presence of cognitive fatigue, the brain's functional network maintained its typical small-world topology, and that the balance between local processing efficiency and global integration did not substantially change.

We think this is the most important sentence in the paper, and it points in a reassuring direction. Whatever mental fatigue is, it is not a degradation of the brain's fundamental organisation. It is something narrower.

3.Where the Change Actually Appeared

The changes showed up at the local level. After Bonferroni correction — a conservative adjustment for testing many regions at once, which here set the threshold at p below 0.005 — significant variations in nodal efficiency were confined to four regions: frontal, fronto-central, central and centro-parietal cortices.

That distribution is not random. It covers the areas most associated with executive control and with the preparation and monitoring of responses — precisely the functions that a sustained, demanding task loads most heavily. The regions that were working hardest are the ones whose efficiency shifted.

The picture that emerges across this series is consistent: a system that remains structurally sound while specific control regions become less efficient, effort feels higher, and willingness to spend that effort declines.

4.The Limitation the Authors Disclose

Here is where careful reading pays. The normality testing that underpins the statistics could be performed for six cortical regions. For four others — prefrontal, occipital, temporal and temporoparietal — it was omitted because the sample was insufficient, with three or fewer electrodes contributing.

This means the study cannot speak to those regions. It is not evidence that nothing happens in prefrontal or temporal cortex under fatigue; it is an absence of coverage. Given how prominently prefrontal cortex features in most accounts of cognitive control, that gap is worth holding in mind before treating the four-region result as a complete map.

On sample size the study is more reassuring. The authors performed a power calculation in G*Power with alpha at 0.05 and power at 0.8, and, allowing for ten per cent attrition, arrived at a minimum requirement of thirty-seven participants. Stating a target in advance and meeting it is the correct way to do this, and it is not universal in the literature reviewed in Part One.

Participants were neurologically healthy, aged twenty to thirty-five, right-handed, with normal or corrected vision and no colour vision deficiency, recruited at Beijing Rehabilitation Hospital, Capital Medical University between February and September 2024. Fatigue was induced with a Stroop-type task requiring suppression of the automatic reading response. That is a narrow and specific population, and the result should be read as applying to it.

5.The Standardisation Problem, and an Attempt to Fix It

Part One closed on the finding that the field cannot agree on how to induce the state it studies. A second 2025 paper, by Ren and colleagues in Brain Sciences, addresses that directly, and it belongs in this series for that reason rather than for any result about the brain.

Their argument is that existing induction methods often fail because the tasks are too simple to impose a genuine cognitive load, while the long durations required to compensate introduce emotional disturbance as a confound. Common choices — Stroop, n-back, the psychomotor vigilance task — depend for their effect on task difficulty and duration alone.

Their alternative raises demand along a different axis: a paradigm integrating complex cognitive abilities — attention, working memory, inhibitory control and cognitive flexibility — with time pressure, alongside the conventional time-on-task effect. The motivation is explicitly applied. They note that many professionals, from pilots and drivers to athletes and military personnel, work under exactly that combination of complexity and time pressure.

Whether this becomes a standard remains to be seen; a single laboratory proposing a paradigm is the beginning of that process, not the end. But it is the right response to the problem, and it illustrates something about how fields mature. The unglamorous work of agreeing on methods usually has to happen before the interesting results can be trusted.

6.What This Series Does and Does Not Establish

Across three parts and five studies, the defensible summary is this. Mental fatigue is a real state with reproducible cardiovascular correlates. Roughly ten minutes of demanding cognitive work induces it, and around twenty minutes is enough for performance costs to appear on some tasks. Response inhibition is not required. The brain's global network organisation is preserved while specific control regions lose efficiency. And whether measurable output declines depends heavily on the task and the person.

What has not been established is equally worth stating. There is no validated single marker of mental fatigue. There is no agreed method of inducing it, which limits how far studies can be pooled. Most of this work involves small samples of young healthy adults in laboratories, performing tasks chosen for experimental control rather than resemblance to work. And nothing in this literature identifies anything that prevents or reverses the state.

That last point is the one we would most like readers to carry away. The science of mental fatigue is currently much better at describing the problem than at offering a solution, and any product or protocol claiming otherwise is drawing on evidence that does not yet exist.

Editorial Comment

MindHeaven® makes no claim that any product affects brain network efficiency, functional connectivity, or the onset of mental fatigue. The studies described in this series measured what happens to healthy volunteers performing laboratory tasks; none of them tested an intervention of any kind.

We rate this part as preliminary evidence deliberately, even though the underlying work is careful. Two 2025 papers from single laboratories, with modest samples and — in the network study — regions its own authors flag as inadequately covered, are a promising direction rather than a settled account. When the replications arrive we will say so, including if they point the other way.