Published in Frontiers in Neuroergonomics, the research by Maria Rivas-Vidal and colleagues at Universidad Politécnica de Madrid, Indra Sistemas, and Universidad Miguel Hernández of Elche tackles a deceptively simple question: can we actually tell these cognitive states apart when they share overlapping behavioral and physiological signatures? The answer, according to their findings, is yes — but only if you hit the brain with the right combination of experimental levers.

The team recruited 67 participants across two studies, drawn from operational professions including firefighters, police officers, and military personnel. Study 1 involved 40 participants (29 men, 11 women, aged 25 to 45) who completed two sessions separated by no more than 14 days, with task difficulty systematically manipulated between easy and hard levels. Study 2 involved 27 participants (20 men, 7 women, aged 19 to 43) who performed the same tasks twice in a single day — once roughly two hours before their typical bedtime, and once immediately after being awakened from approximately two hours of sleep. This design allowed the researchers to independently target two complementary mechanisms: increased cognitive load and reduced alertness.

Participants cycled through three complementary tasks designed to probe different aspects of cognition. The Multi-Attribute Task Battery II (MATB-II) simulates the multitasking demands of an aircraft cockpit, requiring simultaneous management of tracking, resource management, communications, and system monitoring subtasks. The N-Back task challenges working memory by asking participants to identify when a digit matches one shown N steps earlier in a sequence. The Sustained Attention to Response Task (SART) presents rapid streams of digits and requires participants to withhold responses to a specific target, probing attentional lapses. Each session lasted approximately 90 minutes, bookended by baseline recordings with eyes open and closed.

The measurement arsenal was equally comprehensive. Subjective questionnaires included the NASA Task Load Index for workload, the Karolinska Sleepiness Scale and Stanford Sleepiness Scale for drowsiness, Visual Analog Scales for perceived alertness, and thought probes for mind wandering. Performance metrics were distilled from 24 indicators for MATB-II, 16 for N-Back, and 16 for SART down to six representative metrics per task. Although physiological signals — including 32-channel EEG, electrocardiography, electrodermal activity, blood pulse volume, respiration, and eye-tracking — were recorded throughout, the current analysis focused exclusively on behavioral and subjective measures, laying the groundwork for future multimodal modeling.

The results paint a strikingly clear picture of dissociation. When task difficulty increased, performance degraded robustly across all three tasks. In MATB-II, tracking error rose by 19.5 percent, resource management deviation by 37.6 percent, and system monitoring reaction time by 34.3 percent, all with medium to large effect sizes. N-Back accuracy dropped from 0.99 to 0.93, error rate surged by 582 percent, and reaction time variability jumped by 60 percent, all with large effect sizes. SART showed a similar pattern, with error rate climbing nearly 298 percent and F1 score falling by 25.1 percent. Crucially, perceived workload and fatigue also rose with difficulty, while subjective drowsiness actually decreased — harder tasks kept participants more alert.

Sleep restriction told a fundamentally different story. After the brief sleep period, subjective drowsiness increased significantly across all tasks, with large effect sizes for the first baseline, MATB-II, and N-Back, and medium effects for SART and the second baseline. Mind wandering followed suit, rising substantially after sleep, particularly during baseline conditions. Yet performance metrics remained remarkably resilient: N-Back showed no significant changes at all, and SART actually trended toward slight improvements in accuracy and reduced error rates, possibly reflecting a shift toward more cautious responding. MATB-II even showed modest improvements in some metrics, which the authors attribute to a complex interplay between partial recovery from the brief sleep period and potential sleep inertia upon waking.

This dissociation is the study’s central contribution. Task difficulty and sleep restriction do not simply push all cognitive states in the same direction — they selectively modulate different components of a shared cognitive and physiological landscape. Increased difficulty acts primarily on performance and workload-related fatigue, while sleep restriction acts primarily on subjective drowsiness and mind wandering. The authors suggest that subjective fatigue responses may depend on the relative balance between cognitive demand and sleep-related factors inherent to each task context, rather than reflecting a single unified construct.

The findings have immediate implications for operational monitoring. Real-time systems designed to detect operator state in aviation, defense, or emergency response currently struggle to distinguish between workload-driven strain and sleep-driven drowsiness, yet the appropriate interventions differ radically: reducing task load helps the former, while rest or alertness countermeasures help the latter. By demonstrating that complementary experimental manipulations can empirically separate these states using distinct behavioral and subjective signatures, the study provides a validated foundation for future predictive models that integrate physiological signals with performance and self-report data.

Limitations remain. The participant sample, drawn from specific professional groups, may exhibit characteristics associated with operational training that limit generalizability. The smaller sample in Study 2 may have reduced sensitivity to subtle performance effects, contributing to the high number of non-significant findings. Performing the same tasks twice in one day introduces potential practice effects that a between-subjects design could better control. Subjective measures remain susceptible to reporting biases. Nevertheless, the authors argue that the multimodal dataset — combining synchronized physiological, behavioral, and subjective recordings across three complementary tasks — represents a critical step toward robust, real-time cognitive state monitoring in high-demand operational contexts, where understanding not just that an operator is struggling, but why, could mean the difference between mission success and catastrophic failure.

Subject of Research: Experimental differentiation of mental workload, fatigue, drowsiness and mind wandering under task demand and sleep restriction

Article Title: Differentiating linked cognitive states under task demand and sleep restriction: toward an experimental framework for operational monitoring

Article References: Rivas-Vidal, M., Calvo Cordoba, A., Farfán, F. D., & García-Cena, C. E. (2026). Differentiating linked cognitive states under task demand and sleep restriction: toward an experimental framework for operational monitoring. Frontiers in Neuroergonomics, 7, Article 1909914. https://doi.org/10.3389/fnrgo.2026.1909914

Image Credits: AI Generated

DOI: 10.3389/fnrgo.2026.1909914

Keywords: mental workload, drowsiness, mental fatigue, mind wandering, sleep restriction, task difficulty, MATB-II, N-Back, SART, neuroergonomics, situational awareness, operational monitoring