# Forty volunteers completed 960 simulated submersible tasks and a BMC Psychology model found simulated volunteer performance peaked inside narrow visual, auditory, cognitive and movement workload bands before falling again, while the result still needs validation with actual oceanauts in a manned submersible

> A BMC Psychology paper asked a practical question for deep-sea operations: how much mental workload helps a pilot stay sharp before the burden starts to drag performance down? The study gathered results from 40 male volunteers who each completed 24 tests modeled...

Canonical URL: https://www.argo.net/forty-volunteers-completed-960-simulated-submersible-tasks-and-a-bmc-psychology-model-found-simulated-volunteer-performance-peaked-inside-narrow-visual-auditory-cognitive-and-movement-workload-bands/
Byline: ARGO.net Editorial Team
Published: 2026-08-25T19:25:02+00:00
Categories: Explainer, Oceans

![A person's hand gripping an aircraft control inside a cockpit during daytime](https://www.argo.net/wp-content/uploads/2026/08/manned_submersible_cockpit.jpg)

A [**BMC Psychology paper**](https://pmc.ncbi.nlm.nih.gov/articles/PMC11789376/) asked a practical question for deep-sea operations: how much mental workload helps a pilot stay sharp before the burden starts to drag performance down? The study gathered results from **40 male volunteers** who each completed 24 tests modeled on submersible work, which produced **960 data points** before outlier removal. The team then used those results to build a prediction model for oceanaut cognitive performance.

The paper stays clear about the setting. These participants were volunteers acting through a **simulated mission experiment** and the discussion section says they were not actual oceanauts in a manned submersible. That limit matters because the experiment still tried to imitate real duties, including interface checks, control-panel actions, voice instructions, posture demands and comfort ratings, yet it did not place trained oceanauts inside a live dive.

The reason to study this at all is easy to grasp. The authors note that earlier sea-trial statistics tied 70 to 80 percent of manned deep-submergence accidents to human factors and broader workload research has already shown in settings such as driving that performance often slips when mental demand moves too high or too low, as summarized in [a Frontiers in Psychology review on mental workload and driving](https://doi.org/10.3389/fpsyg.2014.01344). Submersible pilots work in cramped cabins, long shifts, darkness, noise and vibration, so even small changes in attention or response speed can matter.

## How the 960 records were built

The experiment started with a task analysis built around the **VACP workload method**, which breaks job demands into visual, auditory, cognitive and psychomotor channels. The authors used it to score **four typical tasks** for oceanaut work. Task 1 covered position checking, Task 2 covered control-panel commands, Task 3 covered voice instructions followed by action and Task 4 covered voice information followed by question answering. Each participant completed 11 versions of Task 1, 6 of Task 2, 4 of Task 3 and 3 of Task 4, for 24 tests total.

The tasks represented several concrete parts of simulated submersible work. In the first task, participants memorized distance values, heading angle and dive depth from a display, rated comfort, judged whether a later value was correct and then entered their answer with a button. In the second task, they inspected an operation panel, memorized a manipulation sequence and pressed the proper controls for hovering and sampling actions. The design aimed to mirror the kinds of visual checking and ordered button work that can fill a pilot's shift.

The audio-heavy tasks covered another side of submersible work. Task 3 asked volunteers to listen to voice instructions under four noise frequencies, remember the key details and press the matching control. Task 4 asked them to listen to spoken position data, retain it and answer interface questions from memory. Reaction time and accuracy were recorded for every test, then combined into a performance score. The paper cites work on [integrated speed and accuracy measures](https://doi.org/10.5334/joc.154) because a fast answer is only useful when it is also correct.

## Where the model found performance peaks

After the team removed eight outliers, the model was trained on 952 remaining records. Their chosen **RBF neural network** outperformed a back-propagation neural network in the paper's validation table, with RMSE of 0.219, MAE of 0.046 and MSE of 0.007. Those error values describe prediction accuracy rather than human skill by themselves, but they do show the fitted model tracked the experimental data much more closely than the comparison model.

The more interesting result came from the workload curves. The model suggested **workload sweet spots** rather than a simple rule that less demand is always better. Visual workload improved predicted performance until about 80, then performance dropped after the load moved higher, which led the authors to recommend a visual band of 80 to 90. Auditory workload improved performance up to about 50, with an optimal band of 40 to 50. Cognitive workload rose toward a band of 120 to 130 before declining, while psychomotor workload worked best around 30 to 40.

The same model also pointed to effects outside the task channels. Predicted performance was highest when participants reported **subjective comfort** at level 5 or 6 on the seven-point scale, which the paper labels relatively comfortable and very comfortable. The strongest single age point in the model was a **29-year-old profile**, while the lowest was 21 years old. Higher education also tracked with higher predicted performance, with doctorate holders scoring above lower degree levels in the model outputs.

## Why moderate workload beat minimal workload

The authors argue that a pilot does not benefit from stripping every channel down to the lowest possible load. Some demand appears to keep attention engaged, while excessive demand starts to crowd out memory, judgment, or response control. That idea fits a wider literature: [an IISE Transactions on Occupational Ergonomics and Human Factors study](https://doi.org/10.1080/24725838.2020.1770898) reported that transitions between low and high mental workload can improve multitasking performance, which helps explain why the best point may sit in a middle band instead of at the floor.

Comfort showed a similar pattern. The volunteers performed best when conditions felt reasonably pleasant, yet the model predicted a decline once comfort moved to the most satisfied end of the scale. The paper links that result to environmental research on temperature and humidity, including [an Indoor Air study on cognitive performance under extreme heat](https://doi.org/10.1111/ina.12755). In plain terms, people working in confined technical spaces seem to do better when the cabin feels manageable, without assuming that maximum ease always brings maximum focus.

The education finding also fits a broader research base, though it should be read carefully. The model captured a correlation inside this study's sample, which was highly educated from the start: 5 volunteers held undergraduate degrees, 20 held master's degrees and 15 held doctorates. That pattern lines up with evidence that education can support later cognitive performance, including [a Proceedings of the National Academy of Sciences paper on education and later-life cognition](https://doi.org/10.1073/pnas.1811537116). Even so, correlation inside a simulation is still different from proving that a degree alone would improve real dive performance.

## What the model still cannot claim

The biggest limit is right in the paper's own discussion. The study never tested actual oceanauts during real submersible operations. It tested volunteers who met selection rules such as daily mental work, long sitting time and no recent neck or back pain. All 40 were male, their mean age was 27.89 and the sample was concentrated in a narrow educational range. That gives the model a tidy dataset, yet it narrows how far the results can be generalized.

The experiment also simplified the mission environment while trying to preserve its essential demands. Another participant sat beside the subject to imitate a natural working arrangement, the tasks used realistic interfaces and voice conditions and the comfort form captured how each test felt in the moment. Even with those efforts, the evidence still belongs to a lab-like mission simulation. It does not yet tell us how a trained crew member would perform after hours inside a real manned submersible, under true operational pressure, with live mission consequences.

That leaves the study in a useful middle ground. It offers a structured map of how workload, comfort, age and education may interact during oceanaut-like tasks and it suggests that task design should aim for balanced demand rather than constant minimization. At the same time, the paper's next step is obvious because the authors say it themselves: future work should bring actual oceanauts into simulation experiments or verify the findings during real operations. Until that happens, the clearest reading stays close to the evidence: 40 volunteers completed 960 simulated tasks, a model found narrow performance peaks across four workload channels and the result remains an informed guide rather than a final rule for deep-sea crews.