Real-Time Cognitive State Monitoring via Physiological Signals in Commercial Aviation: A Systematic Literature Review with Reasoned Snowballing Expansion

Abstract

Aviation safety depends critically on pilots’ mental and cognitive states, particularly in high-stakes and complex operational environments where human errors cause most safety events today. This paper reviews current advances in real-time monitoring of commercial pilots’ cognitive states through physiological and neurophysiological signals and identifies methods applicable to enhance aviation safety and efficiency. In an increasingly complex and congested system, it is essential to investigate the relationships between pilots’ mental workload, stress, startle effect, and physiological parameters to highlight cognitive overload or deficiencies in real time. This systematic literature review was conducted according to PRISMA 2020 guidelines, using Google Scholar, Scopus, and PubMed, and identified 26 eligible studies. A targeted backward citation search screened 17 additional records, and two studies were added to the initial set. Twenty-eight records were therefore included and the review highlights a range of biometric indicators of pilots’ mental states with varying degrees of validity and operational applicability. Collectively, these studies offer a clear overview of state-of-the-art approaches, while also evidencing constraints related to intrusiveness and real-world feasibility. Physiological monitoring holds strong promise for enhancing pilot performance and safety by detecting early signs of overload and stress. However, its integration into operational aviation remains limited. Future research should prioritise longitudinal, in situ evaluations, multimodal data fusion, and pilot-centred design to ensure practical applicability, non-intrusiveness, and regulatory compliance, ultimately bridging the gap between academic research and cockpit reality.

1. Introduction

Airline pilots perform their duties within highly complex operational environments, where any deviation or improper execution of tasks may pose a significant risk to aviation safety [1,2]. Cognitive Load Theory posits that excessive demands on working memory impair human performance, particularly during tasks that require high levels of cognitive processing [3,4,5]. Monitoring and supporting pilots’ cognitive state in real time has become essential in searching for safer and more efficient flight systems [1,4,6,7,8,9]. Non-invasive physiological modalities are increasingly used to quantify mental workload, fatigue, startle and surprise responses, stress, and emotional states in simulated and real flight contexts. These modalities offer valuable insights into human performance and enable proactive interventions to mitigate cognitive overload risks and startle effects in pilots [9,10,11].

The aim of the paper is to review the literature on real-time monitoring of cognitive load and stress in pilots, and we intend to make three contributions to the literature. First, we intend to contribute to the literature on pilot error prevention and spur future research on effective real-time identification of cognitive load and stress markers, as today, 60–80% of aircraft accidents are attributed, at least in part, to pilot error rather than mechanical failure [2,12,13,14], with cognitive factors responsible for nearly 26% of them [11]. Secondly, our review aims to provide actionable insights to airlines regarding hiring and training pilots, given the projected global need for 612,000 new pilots by 2040 [15], pushing the industry towards decreasing the crew complements in the cockpit in the future. Thirdly, we intend to extend insights on how to provide advanced assistance to human-end users in aviation [9], as well as in optimising the design of aeronautical interfaces, because although technological advancements have enhanced safety and efficiency over time, they have also introduced new challenges, increasing pilots’ cognitive demands [16]. The proliferation of digital interfaces necessitates greater mental effort to interpret and manage vast amounts of information, significantly elevating workload [6,11].

Beyond addressing the review’s primary aims, the findings present objective metrics that can inform the design of aeronautical systems and human–machine interfaces and support the development of training strategies intended to mitigate the adverse effects associated with cognitive and mental workload degradation [3,17].

2. Cognitive Overload, Stress and Pilots’ Competencies

Excessive workload can lead to impaired situation awareness [18], decision-making, delayed reactions, stress, and operational errors, particularly in high-pressure situations [2,19]. Pilots engage in multiple concurrent activities as piloting an airliner involves the ability to allocate attention to control the flight path and its parameters [12,20], communicate inside and outside the cockpit [3,21], and constantly adapt to dynamic conditions [5,22]. Such situational complexity increases the cognitive demands of management activities, which consume attentional resources and elevate mental workload [5,23] and stress [24].

Stress is a physiological and psychological response to situational demands that exceedingly tax or exceed human coping resources, affecting cognitive and emotional behavioural responses [25,26,27]. It is a natural adaptation mechanism to internal and external stimuli, enabling individuals to respond to environmental challenges [28,29]. As a general appraisal, stress was categorised as eustress (positive stress that mobilises resources to engage with tasks and actions) and distress (negative stress). Distress is associated with adverse outcomes, including impaired cognitive functioning and decreased work [27,30]. Svensson et al. [31] emphasise that stress significantly impacts decision quality, increases the likelihood of errors, and reduces safety margins. It may contribute to spatial disorientation, accounting for approximately 33% of all aviation accidents, associated with an almost 100% fatality rate [20].

In aviation, stress is a significant factor contributing to cognitive overload, amplified by the unique challenges pilots face beyond those of ordinary workers [25], such as advanced cockpit systems [21], fatigue, sleep deprivation, irregular schedules, low oxygen levels [28], and stressors like noise, vibrations, congested traffic environments, adverse weather or technical malfunctions [26,28].

Cognitive overload arises when task demands exceed an the cognitive representational and processing capacity of the pilots, such that the volume or complexity of information surpasses available working memory and attentional resources, leading to degraded performance and reduced safety [2,3,5]. For pilots, this condition is particularly critical during demanding flight phases such as takeoff, landing [12], and abnormal, emergency, or uncertain operational situations, where it may result in inappropriate responses, impaired decision-making, diminished situational awareness, and increased error risk [5,32,33].

Overload, cognitive failures and stress significantly influence human performance and operational effectiveness, thereby contributing to errors that could become irreparable or fatal [19,21,28,34,35,36,37]. Mental overload, stress, and surprise affect situational awareness and attention [18,29] and impair decision-making [38], problem-solving [39], communication [34], and reaction times [29], ultimately increasing the likelihood of performance decline.

In the high-stakes environments of airliner operations, a critical interplay exists between cognitive impairment or degradation and human performance that remains insufficiently explored and inadequately addressed [29,40,41].

The early detection of any alterations in pilots’ psychophysical state is essential for maintaining safety and preventing the escalation of adverse conditions linked to human factors [2,7,12,21,27,40,41,42].

The growing capabilities of biosensors and methodologies to monitor cognitive failures offer promising avenues for enhancing human reliability in complex, dynamic, and high-risk operational domains [7,8,43]. Addressing the mental consequences of unbalanced workload and stress is pivotal to sustaining pilot performance and fostering operational resilience, which are fundamental to flight safety in the digital era [2,7,26,40].

This literature review aims to synthesise current research on using neuro-psychophysiological response signals to monitor cognitive functioning in commercial pilots in real time and intends to make three important contributions. The primary goal is to evaluate the potential of contemporary biosensing technologies to identify cognitive weakening during flight operations. In doing so, the review examines which physiological signals have been employed to assess pilots’ mental states, particularly those parameters that demonstrate the highest reliability in detecting cognitive failure and stress and enable timely interventions. Second, we aim to critically explore the extent to which the monitoring methods described in the literature are practically usable in real aviation contexts, meeting criteria such as non-invasiveness, wearability, and unobtrusiveness, so as not to interfere with pilots’ operational duties. Finally, the review addresses the potential future advancements that may arise from integrating existing scientific knowledge on real-time cognitive state monitoring with airline pilots’ practical, domain-specific expertise, to foster the development of ecologically valid, pilot-centred solutions that enhance safety and resilience in commercial aviation. To guide the review, the following research questions were formulated:

RQ1: What physiological signals are currently used to monitor commercial pilots’ cognitive states in real time?

RQ2: What biomarkers could be utilised to monitor commercial pilots’ cognitive states in real time to enhance safety, efficiency, and effectiveness?

RQ3: Which physiological parameters can offer high reliability and effectiveness for real-time detection of cognitive failure, stress, and mental impairment?

RQ4: What future advancements might result from integrating existing knowledge on real-time monitoring of pilots’ cognitive states through physiological signals with the practical, domain-specific expertise of airline pilots?

By addressing these questions, the review aims to map the state of the art in cognitive state monitoring for pilots, identify methodological trends and limitations, and explore how physiological data can support adaptive training and performance management in safety-critical contexts.

3. Methods

This review was conducted per PRISMA 2020 guidelines [44] to identify studies examining the real-time monitoring of commercial airline pilots’ cognitive states through physiological biomarkers, based on the inclusion and exclusion criteria as per Appendix A,

Table A1. Inclusion/exclusion criteria.

Criteria	Included	Excluded
Texts	Academic peer-reviewed papers, conference and meeting proceedings, working papers, books, university dissertations.
Population	Commercial airline pilots.	Military pilots, flying club pilots, helicopter pilots, remote pilots.
Measurement	Real-time monitoring of cognitive or mental states: overload, stress, startle through biometric or physiological measurements.	Fatigue, drones.
Time frame	Up to July 2025.
Device	Non-invasive methodologies.	Intrusive equipment impeding normal cockpit operations and non-aviation domains unless analogously.
Environment	Realistic operational scenarios that require demanding cognitive involvement in commercial full flight simulators.	Laboratory-only if unrealistic (e.g., MRI, tethered systems).
Study type	Experimental, observational, validation studies.	Theoretical or conceptual studies.

. Searches were conducted in Google Scholar, Scopus, and PubMed using Boolean operators (Appendix A,

Table A2. Query used in the databases.

Database	Search Query
Google Scholar	(“cognitive overload” OR “cognitive load” OR “stress” OR “startle” OR “mental overload” OR “mental load”) AND “real-time monitoring” AND (“biometric” OR “biomarkers” OR “sensors” OR “physiological”) (“commercial pilots” OR “airline pilots”)-fatigue-drones-evtol
Scopus	(“cognitive overload” OR “cognitive load” OR “stress” OR “startle” OR “mental overload” OR “mental load”) AND “real-time monitoring” AND (“biometric” OR “biomarkers” OR “sensors” OR “physiological”) AND (“commercial pilots” OR “airline pilots”) AND NOT TITLE-ABS-KEY (fatigue OR drones OR evtol)
PubMed	(“cognitive load” OR “mental load” OR “stress”) AND “real-time monitoring” AND (“biometric” OR “biomarkers” OR “sensors” OR “physiological”) AND (“commercial pilots” OR “airline pilots”)

), yielding 43 records from Google Scholar, 18 from Scopus, and 12 from PubMed. Following the initial screening, 26 studies met the eligibility criteria (for details, see Figure A1 in Appendix A). Based on this initial search we concluded that the aviation-focused body of research was rather limited. Most studies examined only a narrow range of biomarkers, often using trainees or low-experience pilots in settings lacking operational realism, as professional full-flight simulator environments were rarely employed. Furthermore, the research landscape is fragmented, with limited cross-study integration and few developments progressing towards practical tools, methodologies, or technologies. Given the limited evidence base in commercial aviation contexts, the review was broadened through reasoned snowballing, enabling the use of studies from non-aviation domains that allowed us to get a more comprehensive view on the nature of physiological biomarkers used to capture cognitive states. During the revision rounds and following suggestions made by the reviewers, we expanded the search and identified a recent systematic literature review on real-time, non-intrusive cognitive workload assessment across different operational domains [45]. Based on the initial evaluation, none of the 26 studies included in our review overlap with the 50 studies included in Basnet and Zahabi [45]. Based on the targeted backward citation search of Basnet and Zahabi [45], several aviation or military domain studies were identified; however, all involved non-commercial pilot samples (e.g., UAV operators, air-traffic controllers, and marine pilots or soldiers) and no further studies were retained for inclusion. One study with the most promising aviation leads [46] was further analysed and from this backward citation analysis two additional papers that included commercial pilots were identified [47,48] and included in our systematic literature review. Only the 28 studies identified through the systematic search, as described above and further elaborated in Appendix A, constitute the evidence base for this review, yet relevant cross-domain studies are also discussed in order to better contextualise the empirical evidence for the physiological markers used to evaluate cognitive states.

4. Results on Assessing Cognitive Load on Pilots

Engineering psychologists have developed approaches to evaluate mental workload directly or indirectly, typically through three methods: subjective evaluations, behavioural observations, and physiological or biometric indicators [3,9,29,49,50]. This review focuses specifically on physiological and biometric measures, which constitute the central interest of this research.

4.1. Subjective Evaluations

Subjective evaluations are self-reported, post-performance assessments of cognitive, emotional, or physical states, typically gathered through questionnaires, rating scales, or interviews [3,9,50]. They offer direct insight into perceived experiences—such as mental stress, fatigue, startle effect, situational awareness or cognitive load—which may not be reliably inferred from physiological or behavioural data alone. The most utilised procedures are the NASA-TLX, subjective workload assessment technique (SWAT), Cooper–Harper scales [2,5,19,21,29,40,43,50], Subjective Units of Distress Scale (SUDS) for stress and anxiety, Self-Assessment Manikin (SAM) for emotional valence and arousal, Situation Awareness Rating Techniques (SARTs), Participant Situation Awareness Questionnaire (PSAQ), and Situation Awareness Global Assessment Technique (SAGAT) for situational awareness [51,52]. Although valuable, these measures are typically administered after task completion, making them susceptible to biases, memory limitations, and individual perceptions [2,52]. They also impede monitoring performance changes over time, since collecting fresh data requires interrupting the operational scenario [27,53]. Furthermore, in-task data collection is intrusive, disrupts attention and workflow, and may reflect perceived difficulty rather than actual workload or performance [5,19,50]. Subjective evaluations should be complemented by physiological indicators to comprehensively assess the effects of cognitive load, stress, startle, and surprise in aviation contexts [8,33,38,40].

4.2. Behavioural Observations

These refer to systematically monitoring and analysing observable actions and performance indicators to infer underlying cognitive or emotional states. Cognitive workload is thus inferred indirectly through evaluating performance deficits, such as the frequency of errors or omissions in task execution [3,9]. Unlike physiological or neurophysiological measures, behavioural indicators provide an indirect yet ecologically valid assessment of mental states during task execution [24,54]. Specifically, they involve assessing how well an operator performs a given task, typically in terms of reaction time, accuracy, error rate, or secondary task performance. A deterioration in these behavioural metrics is generally interpreted as a sign of increased mental workload, stress or diminished cognitive reserve, offering valuable insight into an individual’s ability to cope with task demands under real-world conditions. In contrast, Rai and Singh [27] contend that behavioural observations lack continuity, are inherently subjective, and are inadequate for monitoring.

4.3. Physiological and Biometric Measures

Physiological and biometric measures have become increasingly accessible in recent years, owing to advances in computational processing, electronics miniaturisation, and physiological sensor technology, which now enable their real-time acquisition in operational environments [8]. These measures are utilised either independently or in combination to monitor and identify variations in cognitive state [7,10,19,43,52]. They are currently favoured for their simplicity, quantifiability, and greater objectivity than self-reports or behavioural assessments [53]. These measures offer key advantages: they require no active input from operators, enable passive and continuous data collection, support a multidimensional performance assessment, and allow for the development of generalisable models applicable across diverse tasks. However, psychophysiological measures bring several challenges: current wearable sensors may be intrusive and impair operational performance; they need to separate mental states that produce similar physiological patterns, such as workload, fatigue, and trust; sensor data are prone to noise and artefacts, necessitating advanced filtering; extensive model validation is required; statistical methods may overlook individual variability; system calibration requires baseline collection; and concerns remain regarding user privacy and operator acceptance [50].

This technological progress has enabled a growing body of research to quantify mental impairment by evaluating autonomic nervous system (ANS) activation in response to stress and cognitively demanding conditions [55]. The ANS plays a key role in regulating physiological responses to work overload, primarily through sympathetic activation and parasympathetic withdrawal, triggering measurable physiological changes. Under high workloads, operators must invest more mental effort to maintain an adequate level of performance [56]. Such effort decreases parasympathetic and increases sympathetic activity [57], which results in peripheral reactions in biometric markers [49].

Among the most widely used physiological parameters for assessing stress levels are heart rate (HR), heart rate variability (HRV), blood pressure (BP), skin conductance, respiration rate (RR), oculometric features, brain activity, muscle activation, and, to a lesser extent, body temperature and prosodic features of the voice [10,26,41]. While the former are well established, body temperature and voice prosody remain less standardised and are still considered emerging indicators in this context. These biomarkers provide insight into autonomic regulation and have been increasingly integrated into wearable devices, making stress monitoring more accessible in everyday life [10]. Today’s wearable devices can measure some of these parameters, providing reliable real-time, non-invasive feedback on physiological changes [5]. Despite these advances, research remains active, driven by the need to improve sensors’ accuracy further and fully exploit the potential of stress monitoring in today’s complex working environments.

Current limitations stem from analysing these parameters individually, which often fails to assess stress or cognitive load levels adequately. Consequently, efforts must be focused on improving data integration and refining algorithms to achieve more reliable and holistic stress detection systems. This review focuses exclusively on the latter, specifically physiological indicators suitable for real-time measurement through wearable, non-invasive, and unobtrusive devices of cognitive load, anxiety, or stress, particularly when they become excessive and unmanageable during piloting operations.

4.3.1. Electrodermal Activity

Electrodermal activity is a key marker of emotional and cognitive load in complex operational settings, as the skin, the body’s largest organ, plays a central role in immune defence, sensory–motor processing, thermoregulation, and vitamin synthesis, as well as emotional expression. Its functional complexity is mirrored by dense innervation, particularly from the autonomic nervous system, which modulates sweat gland activity [58]. This activity gives rise to measurable fluctuations in skin conductance, known as electrodermal activity (EDA) [58] or galvanic skin response (GSR). For these reasons, EDA is a widely studied psychophysiological parameter for assessing emotional arousal, stress, cognitive demand [10,23,27,55,59,60], and the startle effect [33].

EDA is exclusively regulated by the sympathetic branch of the ANS, which governs the body’s “fight or flight” responses to external and internal stimuli [9,29]. It denotes the fluctuation in the skin’s electrical conductance resulting from sweat gland activity [58,61] and is therefore closely linked to psychological processes such as emotional activation and mental workload [9,33,62]. In high-demand contexts such as commercial aviation, where operators are exposed to time pressure and rapidly changing scenarios, physiological signals have been shown to reflect variations in pilots’ cognitive and emotional states during realistic flight tasks [6], supporting their potential use for real-time assessment of psychophysiological conditions.

EDA consists of two distinct components: the skin conductance level (SCL), a slow-varying tonic signal that indicates baseline arousal over longer time scales, and the skin conductance response (SCR), a fast phasic component representing transient reactions to stimuli or task events [9]. The SCR is particularly sensitive to short-term fluctuations in cognitive and emotional load, making it a valuable marker for evaluating acute stress responses during complex operational tasks [62]. Research has shown that increased cognitive workload is frequently accompanied by heightened SCR frequency and amplitude, indicating increased sympathetic nervous system activation [59].

EDA is typically measured using sensors placed on the fingers, palms, or wrists—locations with high sweat gland density. These sensors detect changes in the skin’s electrical conductance caused by moisture from eccrine sweat glands. Commercial devices today provide both high-resolution EDA signals and integration with other physiological data streams [60].

Despite its advantages, EDA has several limitations [60]. Data collection requires a stable environment, careful electrode placement, proper calibration, and minimisation of motion artefacts. Wearable devices are particularly vulnerable to noise from electronic interference or variations in skin–electrode contact caused by pressure, movement, or device adjustments, which can distort the signal and be misclassified as skin conductance responses if not properly corrected [62].

In addition, EDA exhibits low specificity: it reliably indicates physiological arousal but cannot distinguish between cognitive load, emotional stress, or external stimuli [58]. Similar patterns may result from physical exertion or psychological stress, and measurements are further influenced by individual differences (e.g., baseline sweating) and environmental factors such as temperature, humidity, or respiration [62]. Sensors’ placement, long-term wearability, and data integrity during flight present practical challenges in aviation settings. Moreover, synchronising EDA data with specific flight phases or task segments is essential to contextualise conductance fluctuations relative to operational demands.

In conclusion, while EDA provides valuable insight into pilots’ emotional and cognitive states, especially under stress or high workload, it should ideally be integrated with multimodal monitoring systems—combining heart rate variability, respiration, or dry electrode electroencephalography (EEG)—to enhance reliability and interpretability in real-world aviation settings.

4.3.2. Cardiac Activity

Cardiac activity and cardio-respiratory measures have been employed in aviation research since the 60s to assess cognitive states, operational performance, stress, and fatigue [50,55]. The heart responds to environmental changes by regulating its rate and contractility, a process controlled by the ANS [13]. Its activity can be monitored using either electrocardiography (ECG) or photoplethysmography (PPG) [2,50,63]. Electrocardiography (ECG) remains the most reliable and widely used method for real-time measurement of cardiac activity [6,29,43]. By capturing the heart’s electrical activity across the entire cardiac cycle, including both systolic and diastolic phases [13], ECG provides not only heart rate data but also additional clinically and operationally relevant information [29]. It records heart activity through sensors in contact with the skin [7], allowing for the precise calculation of heart rate and RR intervals (the time intervals between two successive R-waves), which provide the primary input for HRV analysis.

Several studies have shown that HRV, calculated from ECG signals, is a sensitive index of cognitive load, mainly when assessed using measures that capture variations over time and across frequency bands [2,50,64], as well as of stress [27] and fatigue [65]. Photoplethysmography (PPG) is a non-invasive optical method that measures volumetric changes in peripheral blood using light sensors, such as light-emitting diodes (LEDs) and photodiodes, typically applied to the fingers, earlobes, toes or wrists, which provide favourable conditions for monitoring peripheral blood flow dynamics [66]. PPG signals allow the extraction of parameters such as heart rate and an estimate of HRV (also known as Pulse Rate Variability, PRV), which is well correlated to ECG data in low-noise conditions [67]. PPG devices such as smartwatches and biometric bracelets are particularly suitable for operational settings due to their non-invasive and wearable nature [63].

Real-time monitoring of the cardiovascular system provides important information about an operator’s cognitive and mental stress load. HR, HRV, task-evoked cardiac response, and BP can be non-invasively monitored to track physiological changes associated with cognitive load.

Heart Rate (HR)

Heart rate is one of the most extensively used physiological indicators for assessing cognitive load due to its sensitivity to mental effort and stress, even in the aviation context [14,21,27,29,54,68,69]. It is the number of heartbeats per minute and is directly influenced by ANS activity, particularly under conditions requiring sustained attention or decision-making [64]. HR tends to increase with rising cognitive demands, reflecting the sympathetic activation of the ANS in response to mental workload [13,17,23,43,49,55,70].

Studies in aviation and human–computer interaction have consistently shown HR increases during high-demand tasks, such as simulated flight operations where pilots or air traffic controllers are purposely subjected to complex, cognitively demanding scenarios [49,54,71]. Kim et al. [13] report that brain blood flow, corresponding to the peak in HR, is highest during the landing and lowest in the post-flight phase. Wickens et al. [3] argue that HR increase reflects the heightened arousal, concentration, and anxiety experienced by the pilot during approach and landing, as a consequence of the activation of the sympathetic nervous system, elevating HR and enhancing cerebral blood flow.

Elevated mental workload in pilots is typically associated with increased EEG power in the theta band (4–8 Hz) and a concurrent reduction in alpha-band power (8–15 Hz) [2,21]. Hidalgo-Muñoz et al. [6] report that although HR tends to increase as task difficulty rises, it decreases with training.

This physiological response is interpreted as an adaptive mechanism, mobilising energy and resources in situations of high cognitive and emotional demand [62]. As a result, HR monitoring has become a standard approach in ergonomics and applied cognitive research for indexing real-time operator workload.

However, despite its widespread use, HR as a standalone metric of cognitive load has been criticised for its limited specificity. Changes in HR can result from various confounding factors unrelated to cognitive processing, such as physical movement, emotional arousal [8], circadian rhythms, respiration or even individual baseline differences [13,14,49,72]. Moreover, although HR offers high temporal resolution, with changes detectable within milliseconds, its main limitation lies in interpretive specificity, since variations may reflect not only cognitive load but also emotional arousal, physical activity, or other physiological processes. As a result, HR is less reliable than more sophisticated measures like HRV or EEG for distinguishing between different types or intensities of cognitive activity. Therefore, while HR can provide valuable gross-level insights into autonomic arousal, combining it with other physiological or behavioural indicators is generally recommended for a more accurate assessment of mental workload [32].

Heart Rate Variability (HRV)

Heart rate variability is a widely recognised non-invasive biomarker for assessing ANS activity and inferring psychophysiological states such as stress, mental workload, cognitive impairment, startle reactions, and fatigue [3,5,10,27,29,33,43,49,50,63,64,65,69]. HRV is significant because it reflects cardiac function and provides insights into brain and nervous system activity, enabling the classification of distinct cognitive workload levels [73].

HRV is the dynamic outcome of the interaction among multiple regulatory systems (autonomic, endocrine, and central) operating across different time scales, supporting flexible adaptation to internal and external demands [74]. Unlike a metronomic rhythm, a healthy heart exhibits complex oscillations that enable rapid physiological responses to physical and psychological stressors [64]. HRV emerges from the non-linear, reciprocal interactions between the heart, brain, and ANS, and its intrinsic variability underpins adaptability and resilience [65,74].

HRV is particularly useful for monitoring cardiac responses during pilots’ cognitive performance, reflecting sympathetic–parasympathetic balance and sensitivity to mental workload [5,6,14,50]. Spectral analysis enables the assessment of ANS activity: high-frequency (HF) power reflects parasympathetic modulation, whereas low-frequency (LF) power, influenced by both branches, is often associated with sympathetic activity under stress. During cognitive load, HF power typically decreases and LF power may increase, with the LF/HF ratio commonly used to indicate autonomic balance [7,75]. However, the LF/HF ratio should be interpreted cautiously, as it may provide an oversimplified view of autonomic function [35,75].

Elevated stress and cognitive overload generally reduce HRV, especially HF power, while a higher resting HRV—particularly vagally mediated components—is linked to enhanced executive functioning and emotional regulation [5,6,64,74]. Phase-specific changes have been observed in aviation contexts: HF power decreases during landing and LF power increases during takeoff, with post-flight HF recovery reducing the LF/HF ratio [13].

Both ECG and PPG can measure HRV, although results may vary depending on the method and individual factors such as sex, age, and fitness [64]. PPG offers a practical, non-intrusive option for continuous, real-world monitoring, but its sensitivity to motion artefacts reduces the number of valid inter-beat intervals, potentially compromising accuracy; post-processing methods can mitigate these effects [63,65].

Although HRV is widely used to assess cognitive workload, heart rate often responds more rapidly to fluctuations in mental demand. For example, HR shows marked changes across flight phases [54], whereas HRV may remain relatively stable under prolonged cognitive load or task engagement [6,14,76]. Successful application of HRV in measuring cognitive load is generally limited to short, controlled tasks, as respiratory influences and other confounders affect HRV [32].

Despite these limitations, reductions in HF power consistently occur under cognitive demand and stress, confirming HRV’s utility as a sensitive physiological marker of mental workload and resilience [73,76].

Blood Pressure (BP)

Blood pressure—particularly systolic blood pressure (SBP)—has shown promise as a physiological marker of cognitive load, anxiety and stress [4,27,55,68,69]. SBP is sensitive to changes in sympathetic nervous system activity and has been found to increase during cognitively demanding tasks such as mental arithmetic, memory recall, and dual-task performance [23,70,77]. This increase reflects the cardiovascular system’s adaptation to heightened mental effort and stress, suggesting that SBP can indirectly indicate cognitive workload. Unlike HR, which captures overall autonomic activation, SBP may provide more specific insight into sympathetic vasomotor control under mental strain [78].

Recent advances in wearable sensor technologies have further enabled the continuous and non-invasive measurement of blood pressure in real-time, making it feasible to integrate BP monitoring into applied cognitive workload assessment, particularly in operational environments such as aviation, driving, and healthcare.

However, despite its potential, using BP as a cognitive load indicator presents notable limitations. BP is influenced by cognitive factors and a range of confounding variables, including posture, respiration, caffeine intake, emotional state, and individual cardiovascular fitness [77,79]. Additionally, while promising, continuous BP monitoring with wearable devices still faces technical challenges related to motion artefacts, calibration drift, and lower temporal resolution compared to other biosignals [71]. Furthermore, SBP responses can vary considerably between individuals, making it difficult to generalise across populations or task types without baseline normalisation or multimodal approaches. Consequently, BP is often recommended as a complementary measure rather than a standalone indicator of mental workload [69].

4.3.3. Body Temperature

Body temperature is one of the most well-known and extensively applied biosignals in the aviation domain [9,29,80]. Its relevance stems from its capacity to reflect ANS activity, particularly the sympathetic branch. Variations in skin temperature are mainly driven by changes in peripheral blood flow, which are modulated by sympathetic vasomotor responses to environmental [81], cognitive [62,82], and stress-related stimuli [27].

Infrared thermography (IRT), a non-contact and non-invasive imaging technique, has proven helpful in detecting temperature fluctuations. Research has shown that stimuli activating the sympathetic nervous system can lead to measurable reductions in cutaneous microcirculation, as indicated by a decrease in surface temperature [81]. Moreover, more recent studies confirm that skin temperature decreases under acute stress and increases cognitive workload due to stress-induced vasoconstriction [9].

The advantages of using skin temperature as a psychophysiological indicator include its unobtrusive measurement, high temporal resolution, and the ability to capture real-time changes in the autonomic state without requiring direct contact. Furthermore, because thermal responses are relatively involuntary, they offer a reliable reflection of underlying physiological processes [80,82].

Skin temperature, on the other hand, has some limitations. It is influenced by external factors such as ambient temperature, humidity, airflow, and physical movements, which can confound the interpretation of thermal data. Moreover, normal physiological processes, including circadian rhythms—particularly relevant in airline pilots—and age-related factors, can also contribute to fluctuations in body surface temperature [83]. Additionally, the latency of thermal responses may pose a challenge for applications that require the rapid detection of cognitive or emotional shifts. Finally, while useful in controlled environments, its reliability in dynamic and operational settings, such as aviation, remains among the most extensively studied and applied biosignals among the wide range used in psychophysiological monitoring.

4.3.4. Respiration Monitoring

Respiration monitoring is increasingly recognised as a valuable non-invasive tool for assessing emotional states, stress levels, startle response, fatigue, and cognitive load in complex operational environments, including aviation [7,19,21,50,65,84]. Respiratory parameters—such as respiratory rate (RR), tidal volume, and variability—are closely linked to ANS functioning and are known to respond rapidly to psychological and cognitive demands [8,23,43,49,50,55,85], and startle reactions [33].

In aviation, pilots under stress or high workload conditions often exhibit increased respiration rates and reduced respiratory variability, reflecting heightened sympathetic activation and reduced parasympathetic control [43,85]. It makes respiration a useful real-time indicator of psychophysiological strain, particularly in high-stakes contexts. Advances in wearable respiratory sensors, such as thoracic bands and capacitive plethysmography systems, have facilitated continuous monitoring during flight operations, making respiratory analysis increasingly accessible for in situ assessments [84]. In commercial aviation, where situational awareness and decision-making under pressure are critical, monitoring respiration may contribute to detecting signs of overload or stress.

Nonetheless, implementing respiratory monitoring in operational contexts also presents certain limitations. Respiratory patterns can be influenced by various non-task-related factors such as speech, body movement, altitude, fatigue, or even individual differences in breathing style [86].

Moreover, the interpretation of respiratory signals can be complex: for example, high cognitive load and emotional arousal, such as anxiety, can result in similar respiratory patterns, making it difficult to disentangle the underlying causes without multimodal data integration [87]. Furthermore, according to Ayres et al. [8], verbal activity—such as talking—can interfere with the reliability of respiratory measures by altering breathing patterns. Additionally, motion artefacts may compromise the accuracy of the wearable devices’ respiratory signal acquisition and analysis, potentially resulting in erroneous or incomplete data interpretations [88].

Respiratory techniques have been investigated for potential benefits in supporting cognitive performance and mitigating stress. Bernardi et al. [89] highlight the influence of controlled breathing on HRV, proposing it as a viable strategy for modulating arousal levels under high-stress conditions. Veltman and Gaillard [49] assert that HRV and HR are significantly influenced by respiratory activity, with slow breathing patterns contributing markedly to HRV—particularly following periods of elevated workload, such as post-landing phases.

Therefore, while respiration offers a promising method for continuous workload assessment, it is generally recommended that it be used with other physiological measures to enhance interpretability and robustness.

4.3.5. Eye Activity

Eye activity, evaluated through eye tracking, is increasingly applied in aviation to enhance safety, performance, and cockpit design, offering significant advantages [10,21,50,68,86,90]. Eye activity constitutes a physiological response commonly assessed through eye-tracking technologies, which provide metrics such as fixations, saccades (including micro-saccades), dwell time, scan paths, corneal and retinal reflections, changes in pupil diameter (dilation and constriction), blink frequency and latency, and PERCLOS (the proportion of time in which the eyelid covers ≥ 80% of the pupil) [3,10,23,47,48,50,53,55,86,90,91,92,93]. The key eye-related parameters that can be used to infer mental stress and anxiety include pupillary dilatation [91], blink rate, fixation duration, and saccadic velocity [53]. Eye tracking captures both involuntary responses (e.g., pupil dilation and blinks) and voluntary behaviours (e.g., fixations), making it a valuable tool for assessing various aspects of cognitive load [8,19,29,68].

Pupillary dilation is often considered a reliable index to measure cognitive processing, emotional arousal and sympathetic nervous system activation. Increased task difficulty is associated with elevated tonic pupil diameter [6,33,86,91], even though direct correlations between eye activity and startle responses remain underexplored [33]. Nevertheless, pupillary dilation should be interpreted with caution, as it is a sensitive but non-specific marker, strongly influenced by luminance, emotional arousal, and fatigue.

Blink rate and fixation metrics can reflect attentional engagement or fatigue, while saccadic dynamics may change in response to elevated cognitive demands [47,48,54,65,93]. Eye fixations, more consciously controlled than pupil dilation or blinks, are commonly used to assess cognitive load through fixation frequency, duration, and gaze transitions. Fixation frequency and duration are inversely related, and their interpretation depends on the task type. In visual search tasks, a higher frequency indicates a greater load, while in tasks requiring deep processing, a longer duration suggests an increased load [8]. Eye trackers are usually mounted on fixed surfaces (in front of the pilots) and require stable head positioning. In contrast, wearable or head-mounted devices—such as smart glasses or integrated cockpit-mounted systems—allow for natural head movements and are better suited for dynamic environments like aircraft cockpits [7,47,48,92]. However, Guo et al. [65] and Ryffel et al. [47] contend that the use of glasses may interfere with the accurate acquisition of eye movement data.

Data preprocessing typically includes artefact rejection, primarily to address signal loss caused by head movements, occlusions from eyeglasses [53], or calibration drift. In aviation settings, lighting conditions, reflexes, pupillary dilatation, gaze position, and vibrations pose additional challenges, requiring the use of luminance correction algorithms and robust calibration protocols [8,11,40,65,92]. Another limitation is that eye blinks are modulated by the task’s visual demands rather than workload levels. Specifically, the duration and the number of eye blinks decline as visual task demands intensify [49].

Many studies combine eye activity measures with complementary signals such as ECG, EEG, or GSR to improve validity, allowing for multimodal inference of pilot mental state [86].

While eye tracking provides high-resolution, non-intrusive access to cognitive-affective processes during task performance, its methodological complexity and confound sensitivity necessitate careful implementation and multimodal integration, especially in high-stakes domains such as commercial aviation.

4.3.6. Brain Activity

Brain activity is evaluated as a component of neuroergonomics, a growing field that studies the brain during the execution of complex tasks in real-world settings, by combining insights from psychology, engineering, and neuroscience [3,17]. Its main goal is to assess mental states accurately and understand the ‘brain at work’ [42] to enhance human performance [17]. Brain activity offers a direct window into the mental processes underlying cognitive workload, attention, anxiety, frustration, decision-making, deception, emotional regulation and situational awareness [11,17,29,42,52]. By analysing neural dynamics, researchers aim to detect early signs of mental overload, fatigue, or stress that may compromise performance in the aviation context [7,17,21,22], to enhance efficiency and safety [42].

In commercial aviation, brain activity is increasingly recognised as a critical biomarker for assessing emotional state, mental workload, stress, and startle reactions [33,54]. The rationale is rooted in the premise that cognitive and affective processes are directly reflected in the dynamics of neural activation, particularly in the prefrontal and parietal cortices, which are associated with working memory, decision-making, and attentional control [22]. Neurophysiological changes in these brain areas can indicate increased cognitive demands, emotional arousal, or overload, making brain activity a valuable indicator for the real-time assessment of operator state.

The increasing availability of portable, low-cost, and minimally intrusive brain-sensing technologies has advanced applied cognitive research, particularly through methods such as Event-Related Potentials (ERPs), EEG, functional near-infrared spectroscopy (fNIRS), and Transcranial Doppler Sonography (TCDS) [10,50]. Among these, EEG and fNIRS have emerged as prominent tools, enabling real-time monitoring of mental workload and alertness in operational settings, including flight deck simulations [3,6,11,19,22,23,27,29,31,36,37,52,68].

EEG is the most widely employed technique in operational contexts due to its high temporal resolution, portability, relative non-invasiveness, and wireless capability [10,17,42]. It detects electrical potentials generated by neuronal activity at the scalp level and can be used to compute a range of metrics [2,3,23,27]. As mental workload intensifies, frontal theta power, linked to working memory and executive functioning, increases. Conversely, parietal alpha power, which reflects cortical desynchronisation in regions engaged in task-relevant cognitive processing, decreases [2,3,19]. Wearable EEG devices used in flight simulators or operational flight environments provide high-quality signals and reliability and can be integrated with other biosensors to offer a comprehensive overview of pilots’ cognitive systems [3,10,17,19,22]. However, methodological challenges persist.

From a methodological standpoint, EEG data collection typically requires baseline calibration under resting conditions before real-time data acquisition during task performance. Its electrodes are susceptible to artefacts caused by internal and external signal contamination, head movement, muscular contractions, heartbeat, eye movement, environmental noise, and electromagnetic interference from the surrounding environment, which are difficult to eliminate in real-world settings [4,7,19,23,27]. Another limitation is the discomfort and intrusiveness of its headsets [22]. Nevertheless, as wearable neurotechnologies advance and analytic models improve, EEG-based monitoring is poised to play an increasingly significant role in adaptive automation and neuroergonomics.

fNIRS is a non-invasive neuroimaging technique that enables the real-time monitoring of brain activity, and mental workload during task performance without the need for bulky equipment [3,4,36,37,94]. It offers better spatial resolution (approximately 1 cm², depending on sensor geometry) [42,95], as well as portability, rapid setup on the participant’s head, minimal calibration requirements [17,95], tolerance to movement, reduced susceptibility to electromagnetic interference commonly present in aircraft environments, and an enhanced signal-to-noise ratio [4,23,42,51]. All these requisites make fNIRS well-suited for use in flight simulator scenarios [11] and real flights [42]. fNIRS measures cortical oxygenation changes by emitting specific wavelengths of infrared light that interact with oxygenated and deoxygenated haemoglobin in capillary beds. It allows for the measurement of hemodynamic responses in the superficial layers of the cerebral cortex, particularly in the prefrontal region, and can be employed to assess working memory load [4,23,42].

EEG and fNIRS can provide reliable assessments of mental workload in operational contexts, such as aviation, particularly when integrated with behavioural performance data or other physiological measures [96]. Despite its promise, brain activity monitoring still faces regulatory concerns [97] and ethical considerations regarding the continuous tracking of cognitive states in safety-critical occupations [22].

EEG and fNIRS signal processing involve filtering, artefact removal, and spectral analysis or machine learning to detect perturbations in flight situations [22]. Event synchronisation with flight phases or task markers is essential for accurate temporal alignment. Future research increasingly points to multimodal approaches, integrating EEG, fNIRS, and behavioural performance data, as the most robust strategy for reliably assessing cognitive workload and stress in operational aviation settings.

4.3.7. Muscle Activation

Typically measured through electromyography (EMG), muscle activation is increasingly recognised as a useful physiological parameter for assessing emotional arousal, mental stress, and cognitive overload in high-demand operational environments. EMG detects the electrical potentials of skeletal muscle fibres during contraction, offering insight into neuromuscular activity in response to mental and emotional stimuli [10,70].

In aviation, where pilots must maintain prolonged vigilance, make high-stakes decisions, and manage dynamic flight conditions, muscle tension may be a proxy for internal stress or mental workload, affecting flight task performance. For instance, neck pain and the activity of the trapezius muscle have been correlated with heightened mental workload, cognitive effort, stress and fatigue [70].

Surface EMG (sEMG) is the most common technique for non-invasively capturing muscle activation. It involves placing electrodes on the skin above target muscles, often requiring minimal preparation and offering real-time data acquisition [98].

Muscle selection must be theoretically grounded. Facial muscles typically indicate emotional valence—though sociocultural norms can suppress both emotions and their expressions [99]—while neck and shoulder muscles are associated with postural tension or mental effort [70]. In addition to these, Tichon et al. [68] also include measurements of jaw clenching, wrist flexion, calf tension, and thigh tension.

However, several perturbations limit the standalone use of EMG in assessing emotional and cognitive states. A significant concern is ambiguity in interpretation since the human face engages in numerous non-emotional activities. Muscle tension may result from cognitive effort, emotional stress, physical discomfort, or poor posture, making it challenging to isolate psychological causes [100]. Moreover, motion artefacts, noise and external vibrations—common in cockpit settings—can contaminate EMG signals, reducing their reliability [33]. Sensor placement and comfort may also affect long-term wearability and pilot compliance. Furthermore, EMG captures only localised muscle activity and does not offer the broader physiological perspective that can be achieved by integrating additional measures such as HRV, skin conductance, EEG, and respiration [101].

In conclusion, muscle activation offers a valuable yet rather complementary indicator of operator state in aviation contexts. When integrated with other physiological and behavioural data, EMG can enhance the robustness of cognitive workload assessments, particularly in real-time and ecologically valid settings.

4.3.8. Voice Pattern

Part of the speech production system has emerged as a valuable method for assessing emotional states, stress levels, and cognitive workload in individuals operating in high-stakes, cognitively demanding environments such as commercial aviation [29,50,72,102,103]. Voice patterns offer several advantages: dedicated devices are unnecessary, they are naturally non-intrusive, and they allow real-time data acquisition [7]. Voice data can be captured using onboard microphones and analysed through machine learning algorithms or signal processing software that extracts prosodic and spectral features. Methodologically, analysis requires high-quality audio recordings and consistent sampling conditions, particularly in noisy environments like aircraft cockpits.

Vocal attributes—such as fundamental frequency (pitch), amplitude (loudness), speech rate, frequency jitter, and energy distribution—are sensitive to physiological changes induced by stress and mental effort, reflecting underlying modulations in ANS activity [72,102,103]. Under cognitive stress, voice acoustic parameters significantly change, particularly with altered pitch, reduced speech rate, increased vocal tremor, and decreased articulation precision [103]. These acoustic modulations can therefore act as indirect but robust indicators of psychological strain. Based on the analysis of aviation incidents, Brenner [72] reports that vocal parameters can serve as relevant indicators of acute stress and cognitive overload in high-demand operational contexts.

Increased fundamental frequency, greater pitch variability, and reduced phrase length are commonly associated with elevated mental workload [102]. Conversely, although no study explicitly states that speech rate is a less reliable measure, based on ICAO [104] guidelines—which instruct pilots to maintain a natural tone, constant volume, and even speaking rate—the author argues that speech rate, being subject to voluntary control, influenced by training, and further constrained by the restricted vocabularies used in aeronautical communications [103], may not represent a robust indicator of cognitive overload or stress in pilots. Overall, changes in specific speech patterns may serve as a non-invasive behavioural marker for monitoring cognitive states under high workload conditions [102].

If, on the one hand, voice measures are unobtrusive and do not interfere in any way with the activities in the cockpit [102], several limitations complicate the application of voice analysis in operational settings. Contextual factors such as communication type, background noise, and microphone type and quality can significantly influence voice parameters, leading to measurement distortions [105]. Furthermore, inter-individual differences in vocal baseline—due to age, gender, native language, or fatigue—require personalised calibration, which may limit scalability. Hecker et al. [106] investigated task-induced stress and reported significant interindividual variability in acoustic manifestations. Their findings suggest that stress’s physiological and behavioural indicators are distinctly pronounced in certain individuals, while being less discernible in others.

Ethical and privacy concerns also arise regarding continuous speech monitoring. The voice serves as a primary medium of human communication and, by its very nature, inherently conveys personal and potentially sensitive information [107].

Moreover, voice pattern analysis remains an indirect indicator that may need to be triangulated with other physiological measures, like HR and RR, EMG data [102], and behavioural metrics, to ensure validity and robustness.

In conclusion, while not without limitations, voice pattern analysis offers a promising, non-invasive avenue for assessing pilot cognitive state, particularly when used in multimodal frameworks that combine physiological, behavioural, and contextual data.

In order to consolidate the descriptive analysis presented above and to provide a comparative overview of the physiological markers and their potential use in commercial aviation, we organise the main markers in

Table A4. A summary comparison of physiological markers and their possible use in commercial aviation.

Modality	Physiological Markers	Validity	Intrusiveness	Cockpit Feasibility	Maturity	Strengths and Limitations
ECG/HRV	HR, HRV	High	Medium	Medium	High	Strong workload/stress sensitivity; confounded by posture/movement; chest straps/patches can be acceptable but still burdensome.
PPG/PRV (smartwatch)	Pulse rate, PRV	Medium	Low	High	Medium–High	Very deployable but motion artefacts; PRV less robust than ECG in dynamic conditions.
EDA/GSR	SCR/SCL	Medium	Medium	Medium	Medium	High sensitivity to arousal; low specificity (stress vs. workload vs. emotion); electrode contact stability is a challenge.
Respiration	RR, variability	Medium	Medium	Medium	Medium	Sensitive to workload/stress/startle; confounded by speech and movement; chest bands can be intrusive.
Eye tracking	Fixations, saccades, dwell, scan path	High	Low–Medium	Medium	Medium–High	Strong for attention/workload; sensitive to calibration lighting and type of glasses; good cockpit relevance yet practical constraints persist.
Pupillometry	Pupil diameter/variability	Medium–High	Low–Medium	Medium	Medium	Highly workload sensitive; strong luminance/confound sensitivity; needs robust correction.
EEG (wearable)	Theta/alpha power, etc.	High	High	Low–Medium	Medium	Strong construct validity, but comfort, setup time, artefacts, and acceptance/regulatory concerns reduce feasibility.
fNIRS	HbO/Hb changes	Medium–High	High	Low–Medium	Medium	Better motion tolerance than EEG; still head-worn + setup; limited cockpit operational demonstrations.
EMG (sEMG)	Trapezius/face	Low–Medium	Medium	Low–Medium	Low–Medium	Useful complementary stress/tension marker yet interpretation ambiguous and sensible to vibration/movement noise.
Blood pressure	SBP/DBP trend	Low–Medium	Medium	Low–Medium	Low–Medium	Potential stress/workload sensitivity but calibration drift and device practicality remain issues.
Skin temperature/IRT	Peripheral temperature, facial thermography	Low–Medium	Low	Medium	Low–Medium	Non-contact possible; heavily confounded by environment/cabin conditions; latency.
Voice acoustics	Pitch, jitter, amplitude.	Medium	Very low	High	Medium	Extremely deployable (existing microphones), but noise/privacy/context confounds; best used multimodally.
Deep Metabolic-processes Assessment (DMA)	Metabolic indicators	Medium (empirical evidence in aviation is limited)	High (yet current implementations require biochemical or advanced biosensing)	Low	Low	Conceptually promising for capturing deeper physiological stress and allostatic load beyond surface autonomic measures; however, current sensing technologies are not yet practical for real-time cockpit deployment.
Multimodal fusion (recommended)	E.g., ECG/PPG + eye + EDA + respiration (+EEG/fNIRS)	Very High	Variable	Medium	Medium	Highest robustness/generalisation potential, but synchronisation, calibration, and integration complexity are key barriers as stated both in our review and Basnet and Zahabi.

presented in the Appendix. We review the degree of intrusiveness and usability in the cockpit as well as the level of technological maturity and the amount of empirical evidence we extracted from our review. In this overview we clarify why multimodal integration is identified in our review, as well as in Basnet and Zahabi [45], as a necessary condition for advancing research and application on wearables in aviation. In what follows, we integrate the theoretical insights into a comprehensive dual-path model of eustress–distress in pilots.

5. An Integrative, Dual Model of Stress Reactivity in Pilots

In order to integrate insights derived from this review, we propose an integrative framework in which situational constraints (workload peaks, automation anomalies, unexpected events or operational uncertainty) are the primary triggers of psychophysiological responses. These environmental and task demands are appraised through cognitive and emotional processing that are inherently influenced by individual differences in experience, coping capacity, stress reactivity and resilience. The appraisal process results in two possible pathways of stress response, namely eustress and distress. On the one hand, eustress represents adaptive and possible performance-enhancing activation that supports focus, alertness and problem-solving efficiency, while distress reflects cognitive overload, leading to performance decline through attentional narrowing, impaired reasoning, delayed responses and reduced situational awareness. It is our contention that these two pathways are indicated by distinct patterns measurable via the multimodal biomarkers described and summarised in this review. This integrative perspective is presented in Figure 1 depicting a dual-pathway model of stress reactivity in pilots.

Based on our literature review, we present situational constraints as objective conditions faced by pilots during flight (including emergencies), and we argue that they appraise these conditions in cognitive and emotional terms. Individuals’ differences in cognitive functions, genetic predispositions in stress reactivity and emotionality impact the way in which different pilots perceive similar situational constraints, and their reactivity to these situations is reflected as eustress (positive, motivational state) or distress (negative, disengaging state) [30,108,109]. We argue that these two states are differentially reflected by evaluative cognitions and behavioural and biometric markers [30] and we summarise in

Table 1. An overview of cognitive, behavioural and biometric markers of distress and eustress in pilots.

	Distress	Eustress
Evaluative cognitions	Negative self-talk and negative ruminative tendencies such as catastrophic thinking (“I might lose control of the aircraft”); threat appraisal (“This situation is dangerous, and it exceeds my abilities”); perceived lack of control.	Positive self-talk reflected as challenge appraisal (“This is difficult but manageable”); heightened motivation and engagement (“I can handle this situation”); focused attention with positive anticipation; optimistic outlook concerning the task.
Behavioural	Increase error rate in checklists and ATC communication; slower and erratic reaction times; confused speech; communication breakdown with others in the cockpit.	Faster and accurate reaction times; clear and concise ATC communication; smooth checklist execution without errors; adaptive task switching and proactive error-checking.
Biometric	Substantially elevated heart rate; increased skin conductance (phasic spikes in EDA); rapid and shallow breathing and decrease respiratory variability; enlarged pupil dilatation sustained beyond task demands; elevated trapezius EMG activity; raised systolic blood pressure.	Moderate increase in heart rate with preserved or slightly reduced heart rate variability; mild and transient increase in skin conductance; slight and controlled increase in respiration rate with maintained respiratory variability; pupil dilatation proportional to task demands with flexible recovery; increased frontal midline theta (EEG) reflective effective activation of executive control; stable muscle activation without excessive posture control.

a list of plausible indicators of eustress and distress under strong situational constraints faced by pilots (such as emergency situations or unforeseen flight circumstances).

6. Future Research Directions

Real-time monitoring of pilots’ cognitive states in high-complexity operational environments holds considerable promise for advancing human–machine teaming while enhancing pilots’ ability “to operate safely, effectively and efficiently in the commercial air transport environment” [110]. Despite sustained calls for reliable and objective real-time workload detection, practical implementation remains limited. A recent systematic literature review [45] synthesised 50 studies on real-time, non-intrusive cognitive workload monitoring across different operational domains. While this review provides a valuable cross-domain taxonomy of sensing modalities and modelling approaches, our systematic literature review diverges in scope and inclusion criteria as it focuses specifically on commercial aviation contexts. As a results, the two systematic reviews have minimal overlap between the included studies, which highlights a domain-specific evidence gap, showing that the translation of real-time cognitive state monitoring into commercial aviation contexts remains underdeveloped. Importantly, both reviews converge on three key conclusions. First, ocular metrics and cardiovascular markers dominate the field, while EEG and other modalities are less frequently used to evaluate cognitive load in realistic cockpit environments. Second, both reviews also identify a clear need for multimodal integration of physiological markers in order to improve the evaluation accuracy, generalizability and ecological validity of the results. A third common conclusion is that substantial challenges persist in bridging the gap between laboratory validation and operational deployment of physiological measurements, particularly regarding calibration and integration with real-time, real-life decision-support systems. While Basnet and Zahabi [45] highlighted this requirement across domains, our systematic review complements these findings and pinpoints an even more pronounced evidence gap for commercial aviation research. Although notable progress has been achieved in fields such as autonomous vehicles, artificial intelligence, and medical technologies, the widespread adoption of these approaches within commercial aviation and other high-stakes environments is still lacking [40]. Bridging this gap will require not only further technological maturity but also careful alignment with aviation’s operational, ergonomic, and regulatory constraints.

Another meaningful line for future research is to advance understanding of the neurocognitive mechanisms that underpin adaptive problem-solving, strategic planning, and the regulation of emotional or automatic responses under pressure in pilots. Pursuing this line of inquiry may also illuminate individual differences in stress exposure, physiological reactivity, recovery processes, and resilience, thereby providing a foundation for more finely tailored approaches to workload management and training in operational settings [110].

Another promising research direction aims to improve the understanding of the increase in mental workload and automation complacency during interactions with advanced technologies. In particular, the targeted use of biomarkers could help differentiate the cognitive processes underlying these phenomena [111]. Moreover, although we acknowledge that cognitive workload can occur at any stage during a flight, additional studies could explore the use of physiological markers to capture cognitive states in specific situations such as, for example, in the event of bad weather along the route; concerns related to traffic preventing the planned cruise altitude (causing an increase in fuel consumption); turbulence leading to over- or low-airspeed; weather at the destination or at alternate airports, or along the route; extensive mountainous areas that prevent descent in the event of an emergency descent to the safety altitude of 10,000 feet); marginal fuel quantities; fatigue; the time of day; or minor system malfunctions, as cognitive impairment can occur in any of these situations, as well as during specific flight phases.

A promising direction for future research lies in combining the biometric indicators discussed in this review with biological markers capable of detecting early signs of stress and cognitive overload in a non-invasive manner. Building on our integrative framework, future research should disentangle markers of eustress and distress and novel methodologies such as Deep Metabolic-processes Assessment (DMA^®, [26]) could offer innovative pathways for capturing bioelectrical responses associated with stress-induced metabolic alterations that could distinguish between eustress and distress. However, evidence for DMA in aviation is currently limited and novel validation studies are required before any clear claims of its utility for cockpit deployment can be supported. We thus call for multimodal integrative approaches that could further provide discriminative patterns of physiological markers that are indicative of eustress and distress in pilots.

7. Limitations

The key limitation of our study is that based on the sample descriptions found in the 28 papers included in the review, we could not find solid evidence that these studies used real professional pilots (a pilot is considered professional when he has 1500 h of experience and has “unfrozen” his professional licence, ATPL). We believe this is the case because a pilot grounded in a simulator is very expensive (and impractical for airlines), and professional time in certified commercial simulators is also pricy. Future research could further attempt to expand on the empirical evidence on the use of physiological markers as indicators of cognitive states in real pilots. Another limitation of this review is that it did not include other potential measures of cognitive states, such as head movements, facial expressions, and hormonal fluctuations, which have been proposed in the literature as additional indicators of operator status [50]. Instead, the decision was to focus on methods and markers that are more widely used and more readily applicable within the operational context of aviation. Our results also do not account for gender differences or other forms of social differentiation. Although the population of pilots in commercial aviation is not gender balanced, future research could further explore gender-related, racial, cultural, or age differences at the interface between physiological markers and cognitive and emotional states during flight. Another limitation concerns the methodological approach adopted. The study was based on a systematic literature review, subsequently expanded through the snowballing technique. While this strategy ensured a focused and relevant body of evidence, it also resulted in a relatively limited number of eligible studies. A broader review, incorporating multiple databases, would be necessary to provide a more comprehensive overview of the state of the art in real-time cognitive state monitoring through physiological signals in commercial aviation.

8. Conclusions

This systematic literature review underscores the critical importance of real-time cognitive state monitoring in commercial aviation as both a safety enabler and a pathway toward enhanced operational resilience. By synthesising evidence across a diverse range of physiological and neurophysiological modalities, the review reveals that non-invasive biometric sensing holds considerable promise for detecting early signs of cognitive overload, stress, and performance degradation in pilots. Such capabilities could enable adaptive system responses, support ecologically valid pilot training, and guide the development of resilient human–machine interfaces for stress-induced and cognitive load impairments. However, despite technological advances in biosensor accuracy, miniaturisation, and data processing, the integration of these monitoring systems into real-world cockpit operations remains limited. Practical implementation will require bridging the gap between scientific research, which has often remained abstract and confined to academia as an end in itself, and its practical integration in the complexity of the operative environment. Even if massively more expensive, the involvement of professional pilots with thousands of hours of airliner flight experience is essential to finding applicable solutions and making the research relevant in the real context. Moreover, trials cannot be conducted on students or pilots with limited, and in any case, non-airline experience, sitting in front of computers positioned on a table to attempt to simulate situations that are significantly more complex in the reality of the operative context. Furthermore, solutions must be practicable and exportable to real-world contexts, capable of meeting specific requirements regarding non-intrusiveness, ergonomics, and regulatory compliance. Future research should therefore focus on longitudinal in situ evaluations of sensor technologies, the fusion of multimodal physiological data, and the joint design of monitoring systems with pilots to maximise acceptance, trust, and effectiveness. By aligning scientific innovation with the reality of cockpit operations, the aviation industry can leverage real-time cognitive monitoring to improve human performance, reduce the likelihood of error, and maintain safety margins in an increasingly complex operating environment.