The impact of navigational alarms in ECDIS training: assessing physiological signals

The increasing reliance on ECDIS in maritime operations underscores the need to understand how navigational alarm affect human performance, yet evidence on their physiological impact during training remains limited. The study aimed to explore the effects of a stressful event on general autonomic arousal during navigational operations and how this relationship changes over the course of ECDIS training. A longitudinal design was employed with nine novice participants (N = 9), who completed repeated simulator sessions while physiological signals (EDA, ECG, PPG) were recorded and combined into a composite arousal index. The results showed that alarms significantly increased physiological arousal, with an initial rise followed by gradual attenuation across sessions. The findings indicate adaptation to repeated exposure and demonstrate that objective physiological measures capture dynamic training responses, supporting the need for their integration into simulator-based training to improve situational awareness, workload regulation, and maritime safety.

The learning process on a nautical simulator often leads to strain and fatigue, which affect the ability to retain information (Liu et al., 2020). Similarly, a long sea voyage itself also leads to fatigue, while the use of modern navigational instruments on ship bridges adds additional workload (Main et al., 2017). The mental effort is considered a major factor in maritime incidents (Ronca et al., 2023) and is frequently associated with the emotional and psychological challenges faced by officers of the watch (OOW). Emotional effort, exacerbated by long working hours, isolation, unpredictable environmental conditions, and the high risks of navigation, can affect cognitive functions such as situational awareness, attention, and problem-solving (Xue et al., 2024). Additionally, the detailed accident analysis presented in an EMSA report (European Maritime Safety Agency, 2022) implicated that some working procedures on the navigation bridge could contribute to accidents, revealing that OOWs often do not fully exploit the potential of modern equipment such as ECDIS and radar when monitoring navigation and carrying out risk assessments, particularly in collision avoidance (Vojković et al., 2024). On the other hand, over-reliance on a single device has also been shown to impair situational awareness in some cases of collisions and groundings (Marine Accident Investigation Branch, 2014a), where crew members relied too heavily on electronic navigation devices and neglected important visual and acoustic cues from their surroundings. Dependence on a single electronic device, especially in poor visibility or congested waters, has led to misinterpretation of data and errors in adapting to dynamic conditions (Norwegian Safety Investigation Authority, 2022). Over-reliance, compounded by the stress or fatigue, negatively affects decision-making processes and increases the risk of accidents (Marine Accident Investigation Branch, 2014b). For the above-mentioned reasons, nautical simulators are an indispensable tool for maritime training, as they provide a controlled environment for the assessment of navigational skills and decision-making abilities. In this context, recent research has focused on measuring participants’ physiological response and workload levels during simulator-based training to better understand the impact on route planning, performance and safety. In the studies, workload is often assessed using physiological signal data, typically EDA and heart rate variability, electroencephalogram (EEG) measurements, and also salivary cortisol levels (Main et al., 2017; Liu et al., 2018).

Stress resilience during navigation training has been shown to affect both performance and route planning. Analyses of physiological signal data, combined with statistical methods and machine learning models, demonstrate that challenging conditions such as restricted visibility, equipment failure, and adverse weather can elevate stress response levels and impair decision-making (Main et al., 2017; Xue et al., 2024). Differences in stress-related responses such as indecisiveness or impulsive decision-making, multitasking difficulties, and anxiety, are also evident between novice and experienced navigators, where experienced participants showed more stable patterns of physiological responses. The findings suggests that physiological signal assessment can provide valuable insights into individual training needs and performance capabilities (Bokau et al., 2025). Studies suggest that sound alarm is an effective way to trigger participants’ stress response (Hall et al., 2016).

To address the impact of sound alarms or similar stressors on learning and operational performance, several approaches have been developed. Real-time monitoring systems based on physiological signals allow instructors to track mental workload, emotional states, and stress response of participants throughout simulator exercises. Such systems could potentially support training adjustments (e.g., whether a participant is ready to progress, requires retraining, or is at risk of failure). The feedback emphasizes the importance of stress response management in skill acquisition training and highlights the role of adaptive training scenarios to improve safety and efficiency (Virdi et al., 2022; Xue et al., 2021). For example, the studies on professional pilots performing tasks have shown significant stress-related physiological responses, regardless of the task complexity. Results indicate that even highly trained individuals are susceptible to stress, underlining the need for training environments that prepare operators to cope effectively with high-pressure situations (Cao et al., 2019).

The study aimed to explore the effects of a stressful event on general autonomic arousal during navigational operations and how this relationship changes over the course of ECDIS training (International Maritime Organization, 2017). In operational terms, the stressful event was triggered by a loud voice alarm, and the general autonomic arousal was assessed as a physiological response to this alarm. Physiological measures of stress-related responses were used because they provide an objective, real-time assessment of stress responses, as opposed to subjective psychological scales (Arsalan et al., 2023). To increase the validity and reliability of the measure of the stress response, we combined data from EEG, PPG and EDA, as explained in greater detail in the Methods section. Participants were followed over a series of training sessions using a within-subject design to test whether training mediates the relationship between the sound alarm and the physiological stress response. In addition to identifying patterns in the effects of training on stress responses in novice. The final aim of this study was to test the limitations of navigation technology, paving the way for safer and more effective training practices, while enabling the simultaneous monitoring of nine participants to continuously assess their states and improve adaptive responses in real time.

Participants

Participants were volunteers from the ECDIS navigation course at the Faculty of Maritime Studies and Transport. A total of nine senior students (seven male, two females; mean age 21.1 years, SD = 3) took part. None had prior maritime navigation experience, ensuring a homogeneous novice sample. Before the study, the participants were briefed on the objectives, experimental procedures, and data collection protocols. They were informed about the simulator tasks, the physiological signals to be recorded (PPG, ECG, EDA), and their possibility to withdraw at any time without consequences. Written informed consent was obtained from all participants. Ethical approval was granted by the Commission for Research and Development Activities (KRRD) of the Faculty of Maritime Studies and Transport during its seventh correspondence session on 5 February 2025. The study complied with GDPR and institutional research protocols. The data were anonymized, securely stored, and was accessible only to authorized researchers, to ensure the protection of participants’ rights and privacy.

Instruments

The study was conducted using the Wärtsilä Navi Trainer Pro 6000 ECDIS simulator in desktop configuration (NT Pro 6,000) (Wärtsilä Corporation, 2023). The bridge environment included three monitors displaying the navigation area, ECDIS simulator, and conning station, along with a hardware console for telegraph, bow thruster, and steering system (Figure 1). The simulator provided a realistic and immersive environment by replicating key operational conditions and integrating electronic navigation tools such as radar overlay, conning station, and multifunctional displays. Trainees had access to essential data on vessel position, speed, course, environmental conditions, and proximity to other vessels, closely mirroring the decision-making challenges encountered in real maritime operations.

To record physiological arousal, we used a Senstream ring-type device with photoplethysmography, electrocardiography, and an electrodermal activity sensor, enabling real-time monitoring of autonomic nervous system function and cardiovascular dynamics (Senstream, 2025). The use of these sensors allowed continuous and relatively non-invasive data collection, ensuring minimal movement artifacts and high temporal resolution (Figure 2). The ring transmits data wirelessly via Bluetooth to the acquisition software, enabling real-time monitoring and synchronized recording across participants. The sampling rate of the device is 400 Hz ensured high resolution and accurate detection of responses.

In addition to physiological measures, we also used subjective scales, including the NASA-TLX Index (Hart, 2006), and the Positive and Negative Affect Schedule (PANAS) (Watson et al., 1988), to assess subjective perceptions of workload and emotional responses. Firstly, the NASA task load index was originally developed in aviation research (Hart et al., 1988) and is a widely used instrument for assessing perceived workload, meeting the required criteria (e.g., reliability and validity) of psychological testing (Hoonakker et al., 2011; Crawford and Henry, 2004). The scale consists of six subscales: mental-, physical-, and temporal demand, frustration, effort, and performance. Secondly, PANAS consists of 20 items rated on a five-point Likert scale, with scores summarized into two subscales, positive and negative emotions. Previous studies showed that PANAS also meets the criteria for psychological testing such as validity or reliability (Tuccitto et al., 2010; Krieglstein et al., 2023). Finally, in addition to NASA-TLX and PANAS, we used a set of custom Likert-scale questions specifically designed to capture participants’ subjective reflections on task performance. The questions that have already been used in some previous studies (Košir et al., 2024; Žagar, 2025), provided additional insights into perceived stress response, effort, and overall task experience. This list of questions is provided in the Appendix.

Research protocol

Using a combination of methods, the study aimed to analyze the effects of sound alarm exposure to physiological arousal during seven sessions of ECDIS training (Lei et al., 2022). The experimental protocol followed Bridge Alert Management (BAM) criteria for emergency alarms (International Maritime Organization IMO, 2010). The ECDIS simulation sessions were integrated into the participants’ structured academic training and aligned with the STCW requirements for Officer of the Watch certification (International Maritime Organization, 2017). The primary objective of these sessions was to replicate real-world maritime conditions, foster the development of navigation skills, and assess participants’ responses to sound alarms, as well as their physiological and subjective reactions under simulated operational pressure.

Prior to the experiment, the participants completed a structured 10-h introductory program (familiarization), integrated into their academic curriculum. The introductory program provided the students with a basic skill to handle simulator’s hardware and software. The participants were trained in the fundamental functions of the ECDIS system, including navigation, chart interpretation, collision avoidance manoeuvres, and responding to navigational alarms. The familiarization ensured a standardized baseline level of competence for managing the simulator in participants.

Following the familiarization phase, the participants took part in a 6-week data collection. Each week, one simulation session was conducted, during which multiple navigational scenarios were presented. This training program ensured the collection of a robust dataset for repeated measures analyses. The sessions were designed to test participants’ skill to perform collision avoidance tasks using ECDIS and required real-time decision-making under varying operational conditions. The difficulty level and stressors within the scenarios were kept consistent across all sessions to allow meaningful comparisons of the reactivity to stress over time. For this reason, all simulation scenarios included similar core tasks, such as route planning, route monitoring, collision avoidance with a single target vessel, and return to the planned route. Environmental and operational parameters, including traffic density, navigational complexity, and vessel handling requirements, were not systematically quantified; however, they were carefully controlled to maintain a consistent and comparable level of task difficulty across all conditions. We acknowledge, however, that perfect equivalence cannot be fully guaranteed due to inherent variability in navigation scenarios and individual differences in perception of task difficulty.

A loud emergency alarm was introduced in each session to serve as a stressor. Within the regulatory framework defined by IMO MSC.302 (87), MSC.232 (82), and IEC 61174 (International Maritime Organization IMO, 2010), bridge alerts are explicitly designed as structured mechanisms that support response to navigational hazards. BAM performance standards require alerts to be hierarchically organized, to avoid overload and ensure effective operator attention management. Although the sound does not replicate a specific ECDIS type, it captures key properties of the alert with the highest priority and is treated as a standardized proxy for real-world alarms. (Motz et al., 2009). The alarm was identical across all sessions, and was manually triggered by the instructor at a predefined time during each session. The volume settings were kept constant across all sessions, ensuring consistency of exposure. Participants were informed in advance that alarms could occur during the simulation as part of the training environment; however, they were not informed about the exact timing of the alarm, preserving its element of unpredictability and ensuring a naturalistic response.

The experimental design included seven sessions of the training process. The first session served as a baseline measure (control session), whereas the second, third, fifth, and sixth sessions consisted of training (experimental) exercises involving realistic scenario-based navigation tasks using ECDIS simulator. The fourth session functioned as an internal practical examination on the ship bridge simulator; here, the students were supervised by instructors but not externally evaluated, and the seventh (final) session consisted of the formal (external) examination.

At the beginning of each session, participants were briefed on the objectives, tasks, and safety protocols. They were informed about their simulated environment, including the geographical area and vessel specifications. Scenarios were designed to reflect real-world maritime operations, including navigation in narrow waters such as the Port of Koper, high-traffic areas such as the English Channel and the Strait of Messina, and approaches to large ports such as Rotterdam and Antwerp. Each session required participants to operate different vessel types, including container ships, bulk carriers, and tankers, under various navigational constraints such as traffic separation schemes, port approaches, and restricted visibility conditions. Each session included four exercise scenarios involving route planning, activation of navigational instruments, monitoring of the route, execution of a collision avoidance manoeuvre with one target vessel, and a return to the planned route. The duration of each exercise scenario was approximately 15–20 min. The difficulty of the tasks was deliberately kept constant across all sessions.

During the sessions, the participants’ physiological responses were continuously recorded. Immediately after each session, participants completed a set of subjective scales that were used to evaluate perceived workload, task difficulty, emotional response levels, certainty, and perceived effort. In addition, video recordings of the sessions were collected for post hoc behavioural analysis, which helped us clarify anomalies and outliers in the dataset.

Data analysis

The study aimed to maintain consistency in training conditions; however, individual learning variations posed a challenge. Differences in adaptation rates and situational awareness influenced both physiological and subjective responses. Subjective data were analysed using paired t-tests and correlation analyses to examine the relationship between perceived difficulty, stress response, and confidence across sessions. Data analysis employed a repeated-measures approach to evaluate trends in EDA responses and questionnaire data across sessions. By comparing the second and sixth sessions, the study showed whether the participants exhibited reduced effort and emotional general autonomic arousal as they became more proficient in ECDIS-assisted navigation. Mean and standard deviation values were calculated to identify overall patterns in emotional responses, including peak responses and fluctuations in skin conductance levels. Outlier detection was performed using the interquartile range (IQR) method to minimize distortions in the dataset. Temporal trends in physiological responses were visualized with line plots, which facilitated the interpretation of changes in stress response and effort over time.

Statistical analyses were conducted on aggregated participant-level data. Raw data from multiple. acq files (BIOPAC Systems and Inc. AcqKnowledge File Portability, 2004) were loaded using the Bioread library (Fitch et al., 2025). For each participant, ECG and PPG signals were smoothed with a moving average using a window of 5,000 samples (Table 1). Each signal was normalized individually to the [0,1] range using min–max normalization across the entire dataset (Equation 1). The moving average window length was selected after empirical testing of multiple candidate window sizes, with the aim of balancing noise reduction and preservation of meaningful physiological dynamics. The objective was to examine sustained physiological arousal across training sessions, therefore no separation into tonic and phasic components was performed. The analysis relied on the smoothed signal to capture overall changes in electrodermal activity, and moving average window emphasizes slower, tonic-like trends in autonomic activation, while attenuating phasic responses. The selected window length acts as a filter, where the effective cut-off frequency is inversely related to the window length. Long window reduces high-frequency and emphasize changes in autonomic activation associated with task engagement. We acknowledge that such a window may attenuate rapid phasic responses to discrete events; however, the aim of the study was to capture overall arousal trends rather than precise event-locked dynamics.

The ECG, PPG, and EDA were then aggregated to compute an index, reflecting physiological general autonomic arousal (Fitch et al., 2025). The pre-processing pipeline was intentionally kept minimal, relying on smoothing and normalization to ensure consistency across modalities and participants. While bandpass filtering can improve signal quality in certain contexts, the present approach prioritized a uniform processing framework across signals to support the construction of a composite index. After normalization, the signals were combined additively (Equation 2). The model assumes that each modality contributes linearly to overall arousal, on the basis that heightened sympathetic activity is typically expressed simultaneously across both cardiovascular and electrodermal systems (Sosa et al., 2026; Joudeh et al., 2023).

The result was a scalar time series reflecting normalized physiological excitation per participant (Fitch et al., 2025). Group-level arousal was then calculated by averaging across participants, providing a unified measure of arousal that captured both individual fluctuations and shared group dynamics during the simulator sessions, allowing for the interpretation of how the stress response influenced the physiological states. With respect to artifacts, interquartile range was implemented. a) The measurements were obtained using a ring-type wearable sensor and designed to minimize movement artifacts, and the simulator environment itself required relatively limited physical movement from participants. b) The application of a relatively long moving average window further attenuated high-frequency fluctuations associated with motion artifacts. We acknowledge that our approach does not completely eliminate artifacts and have clarified this as a limitation.

The results are organised in three sections: (a) the control session, (b) the experimental session, focusing on arousal within ±2 min of the alarm and the mean value of the participants’ physiological responses together with the overall trend of arousal, and (c) subjective measures. As mentioned, the training process had seven sessions: the first was a control session, the second, third, fifth and sixth were experimental training exercises with operational complexity, whereas the fourth and seventh sessions served as grading sessions and were therefore excluded from the experimental dataset. Sampling rate was 400 Hz, so each session contained over half a million samples, making the averages highly stable.

Figure 3 represents the control session (Session 1), where no alarms were triggered during the simulation. The participants were not exposed to sudden or unexpected stressors. The observed average normalized level of 1.72 indicates sustained physiological autonomic arousal, reflecting a tonic elevation likely associated with task engagement, ongoing performance monitoring, and evaluative apprehension. The pattern aligns with the psychological concept of state anxiety, where individuals exhibit heightened autonomic arousal in anticipation of a challenging or evaluative situation, even in the absence of additional stressors.

The trends of the experimental sessions are presented in Figure 4, illustrating the responses recorded across representative tasks, providing an insight into the training process. It captures both the earlier stage of familiarization and the more demanding stage, including the stressor. The red vertical line marks the moment when a random audible alarm was triggered during the middle of the simulated navigational task. The average autonomic arousal across ±2 min around the alarm is illustrated on Figure 5. The red vertical line marks the onset of the audible alarm during the simulated navigational task. Notably, some increase is evident even before the alarm is triggered. It is likely related to the participants noticing the instructor approaching the main simulator console and preparing to start the alarm, which led them to anticipate the sound of the alarm.

The mean value of the participants’ physiological response after the alarm is considerably higher than lower before the alarm compared to the response after immediately after the alarm, indicating a substantial rise in physiological autonomic arousal (Figure 6). Similarly, the median and mode value, are also higher after the alarm. The average values of the autonomic arousal, significantly increase after the alarm compared to the period before the alarm, suggesting that participants consistently respond to the alarm event with a measurable change in their physiological state. A paired-samples t-test was conducted to compare the mean normalized excitation levels before and after the alarm event. The analysis revealed a statistically significant difference between the two conditions, t (3) = 4.1976, p = 0.024669, indicating that the observed change is unlikely to be attributable to random variation alone. The magnitude of the effect, quantified by Cohen’s d = 2.0988, indicates a large effect size, suggesting that the difference is not only statistically significant but also practically substantial. The assumption of normality of the difference scores was assessed using the Shapiro–Wilk test, yielding W = 0.9848 and p = 0.929,406. Since the p-value is well above the 0.05 threshold, there is no evidence to reject the null hypothesis of normality.

When comparing experimental sessions, the overall trend of arousal increased during the first four training sessions and then gradually decreases in the later sessions (Figure 7), reflecting a shift from novelty to habituation. At the beginning, participants face an unfamiliar simulator environment, new navigational tasks, and performance pressure, all of which heighten arousal and stress responses. A quadratic model was fitted to the session means, yielding y = −0.1296 x 2 + 1.1243 x − 1.4641. The negative quadratic term (a < 0) indicates a concave-down trend with a mid-session peak, implying that due repeated exposure, the novelty effect diminishes, and tasks become more automated. The coefficient of determination (R 2 = 0.980) indicates that the model explains 98% of the variance in the mean values, but high R 2 should be interpreted with caution as it is based on a limited number of aggregated data points (n = 4) and is therefore presented primarily as a descriptive indicator.

In addition to the change in physiological response over sessions, which showed a steady decrease in arousal, the subjective measures yielded no such change in individuals’ perceptions of task load and affect. Table 2 shows that mental demand decreases from the first to the fifth simulation and then slightly increases toward the sixth sessions. Physical demand rises steadily, indicating that later simulations required more physical or operational effort. Temporal demand drops early but climbs again in later sessions, reflecting higher time pressure in advanced tasks. Performance remains relatively high and stable across all sessions. Effort decreases until mid-program and then increases again, while frustration follows a similar pattern with the lowest value around the third simulation and a marked increase by the final session.

The NASA TLX results was significant on two measures: temporal demand, and load index. Mental load index was decreasing across sessions, but increased in the last one. The same trend was visible with participants’ perception of their own performance. The performance scale showed decrease through sessions and showed increase in the last session too. Temporal demand yielded no obvious trend throughout the sessions. The subjective measures of performance as well as the PANAS scales yielded no change across sessions.

The discussion first addresses physiological responses across sessions, then insights from subjective measures, followed by fine-grained within-session physiological responses, and finally the study’s limitations and the questions raised by the results. The analysis yielded insights into both dynamic (physiological) and static (subjective) responses across the training sessions, where the experimental design allowed us to assess immediate physiological response of the students’ both between and within sessions. The novelty of this approach lies in excitement index and its integration of multiple physiological modalities into a single composite measure of general autonomic arousal. Rather than relying on a single signal, the index combines cardiac activity, peripheral blood flow, and electrodermal activity, each capturing complementary aspects of sympathetic nervous system activation. The multimodal aggregation provides a sensitive indicator of momentary arousal, reducing the limitations and noise associated with individual signals. In this research, the index is not intended to represent a specific emotional state, but a generalized physiological activation level, enabling consistent quantification of responses to salient events. It offers a standardized, reproducible proxy for real-time autonomic reactivity that aligns with the need to assess attention and alertness in maritime contexts.

Physiological responses varied across training sessions. Early sessions were characterized by higher variability and more pronounced peaks, reflecting participants’ limited familiarity with the ECDIS environment and heightened sensitivity to unexpected stimuli. In the later sessions, responses became more moderated, suggesting adaptation to task demands, improved coping strategies, and reduced startle reactivity, suggesting progression in the interplay between learning, workload management, and resilience development within the simulator environment. Based on a subjective report across sessions, we assumed that task subjective indicators of the task load would be higher in the beginning of the training procedure and would decrease across sessions. However, the results only partly supported this prediction. Whereas the average task load, as indicated by NASA TLX, showed the expected decrease alongside the training sessions, but unexpectedly increased in the final session, the rest of the task-load measures showed no obvious trend, or yielded no change. The same pattern was evident with the participants’ subjective measure of emotional response as assessed by PANAS: the data showed no change across sessions. Whereas physiological responses did show increase in the beginning of the training process and decrease thereafter, the subjective measures failed to mirror this pattern. Although the observed results provide encouraging evidence, the present work, due to the limited sample size, should be considered as a preliminary investigation, which limiting the generalizability of the findings. The statistical assumptions such as normality of residuals and sphericity may not be fully met, the results should be interpreted with caution, which may affect the robustness of the significance levels. We acknowledge that the findings should be considered exploratory. The inconsistency between physiological and subjective measures was reported in previous studies (Košir et al., 2024), suggesting that physiological and subjective measures might reflect different kinds of psychological experience of both cognitive load and emotional arousal (Boucsein, 2012).

When interpreted through ECDIS framework, the results support the role of alarms as attention-orienting mechanisms within bridge operations. Across sessions, the observed reduction in physiological variability and peak responses is consistent with progressive familiarization and improved management of alarm-related workload. Early heightened responses reflect sensitivity to novel alerts, whereas later moderation suggests adaptation and more efficient allocation of attention—aligning with BAM principles that emphasize prioritization and minimization of unnecessary alarm burden. The increase in arousal during and immediately after the alarm is consistent with BAM’s emphasis on ensuring that critical alerts effectively capture operator attention. The continued elevation following the event suggests a shift toward an operationally alert state, although it remains unclear whether this reflects adaptive vigilance, stress, or both. We acknowledge that the study does not address full alarm management behaviour or decision-making outcomes.

The experiment was carefully designed to assess students’ responses to their daily training tasks, both across and within training sessions; however, several limitations were encountered. The most notable constrain is probably the small number of the students participating in the training sessions, and the temporal mismatch between physiological and subjective data, requiring cautious interpretation. Nevertheless, the findings highlight the potential value of integrating physiological monitoring into simulator training, particularly for assessing how trainees respond to alerts and for supporting the development of effective alarm handling in line with ECDIS principles. Whereas it is a limitation in the context of statistical analyses, the sampling in fact included all students assigned to the course. From this perspective, we believe that the results are generalizable beyond the sample of the current study. The next limitation is related to the fact that physiological responses were recorded throughout the sessions whereas subjective ones were only assessed after each session, which imply a limitation to their direct comparisons. One way to address this problem in the future would be to use quick visual scales (Joudeh et al., 2023) rather than PANAS and NASA TLX. The experimental design provided fine-grained dynamic insights into the participants’ physiological responses to the alarm. The results showed that physiological response to the alarm was immediate and consistent throughout the sessions. In fact, the arousal was shown to have increased when the instructor approached to the alarm, and continued to increase while the alarm was sound. The whole alarm event lasted approximately 10 s in total. The participants were aware of the importance of treating the simulation with the same seriousness as in a real navigational watch. Physiological responses after the alarm indicated a shift in participants’ mindset from routine engagement toward a more alert and operationally focused posture, which is consistent with previous research and showing that authority on the ships’ bridge can influence physiological readiness (Žagar, 2025). The participants began to understand the alarms’ significance in safety-critical scenarios. This marked a transition from passive observation to more active and responsible engagement with the navigation task. It remains unclear, however, whether increased indicators of general autonomic arousal after the alarm reflect adaptive vigilance, performance-related stress, or a combination of both. It is also unclear how these responses are related to inter-individual differences. Alarm response is defined as the immediate physiological and attentional reaction to a discrete alarm event, as captured in the present study, whereas alarm fatigue refers to a longer-term phenomenon characterized by desensitization to repeated or excessive alarms (Magli et al., 2020). While the simulator environment offers high validity in terms of task structure and system interaction, it cannot fully replicate real-world maritime conditions, including operational pressure, team communication, environmental uncertainty, and the consequences of decision errors. The controlled and predictable structure of simulation scenarios, as well as the presence of instructors, may influence participant behaviour and physiological responses, which limit direct generalization to real-world operations.

Although the simulator environment and recorded sessions would allow the objective performance analyses, the performance metrics were here at this project not extracted from the simulator logs. Thus, future work would examine how alarm-induced arousal interacts with actual task performance and decision-making accuracy, as well as whether repeated exposure leads to habituation or sustained sensitivity in longer training programs. Further studies should also require larger sample sizes and the consideration of alternative statistical approaches, including non-parametric methods or corrections for sphericity, to ensure greater robustness of the results.

The study demonstrates that participants’ physiological states evolved dynamically across simulator training sessions, reflecting processes of adaptation, anticipatory anxiety, and resilience development. Initial variability and heightened reactivity indicated limited familiarity with the ECDIS environment and increased sensitivity to novel or unexpected stimuli. Over time, responses became more regulated, suggesting habituation, improved coping strategies, and greater workload control. Alarm events consistently elicited strong autonomic arousal, highlighting the distinction between written and simulator-based examinations, which underscored the influence of perceived evaluative threat and contextual framing on emotional responses.

The contribution of the study lies in its longitudinal demonstration that physiological adaptation during ECDIS simulator training follows a dynamic trajectory that may not completely be mirrored by subjective workload or emotional self-reports. By showing a systematic dissociation between autonomic regulation and stable subjective ratings, the study provides empirical evidence that physiological measures capture fast, automatic adaptation and anticipatory processes that remain largely outside trainees’ conscious appraisal. The consistent physiological reactivity to alarm events highlights how evaluative context and perceived threat uniquely shape stress response responses in simulator-based assessments, indicating the need to integrate objective biomarkers into maritime training design to better support resilience and decision-making under pressure. The results support the view that stress response and arousal in high-stakes maritime operations are not merely reactive but are also shaped by anticipatory mechanisms, authority cues, and the cognitive appraisal of situational demands. Understanding these mechanisms is important for enhancing training programs that aim to build emotional resilience, reduce cognitive overload, and promote adaptive performance under pressure. The revealed patterns in physiological and subjective responses raise important questions for future research. A notable finding is the divergence between physiological adaptation across sessions and the relative stability of subjective indicators of workload and emotional arousal. While physiological measures reflected learning-related change, the subjective measures did not yield parallel trends, raising the question of whether physiological and subjective indicators capture different aspects of the students’ cognitive and emotional experience, as suggested with the previous literature. One can hypothesized that physiological responses reflect fast, automatic processes, whereas subjective reports represent reflective, post hoc appraisals that may be influenced by social filters such as training norms, expectations, or self-presentation biases. Future research should investigate whether trainees are fully aware of their internal states during simulator training or OOWs are aware of their states being on duty on the bridge. This is an important question for the students’ training because the gap between physiological (dynamic) and subjective part of the experience of cognitive or emotional overload in OOWs under critical circumstances may potentially lead to critical mistakes in the decision-making process, and eventually to maritime incidents. The findings from future studies might therefore provide training tailored to address these issues in particular.

Another area requiring further investigation concerns emotional responses during training procedures. Although the current tasks appear sufficient to elicit robust physiological activation, participants may cognitively reframe the simulator context as a low-consequence or “safe” environment, analogous to a video-game setting in which errors are reversible and personal stakes are limited. Such cognitive distancing and perceived invulnerability can attenuate conscious threat appraisal, leading to emotional downregulation at the reflective level despite ongoing autonomic arousal. As a result, fear, effort, and stress response responses that would normally be consciously experienced and reported in real-world operations may fail to reach awareness or be acknowledged in self-report instruments such as TLX or PANAS in our case. This challenge suggests that standard affective scales may lack sensitivity to context-specific, safety-critical emotional processes and to implicit coping mechanisms such as minimization, normalization, or defensive appraisal. Future studies should therefore consider integrating momentary affect assessments, stress response appraisal measures, and situation-specific emotional and cognitive probes that can be administered during task execution (Fairclough, 2009). In this vein, cognitive load or emotional arousal might be induced experimentally, with carefully designed tasks, to evoke psychological processes and assess their psychological responses in real time. Thus, future research will provide a comprehensive account of cognitive-emotional dynamics during simulator training, where improved longitudinal designs will clarify how resilience and coping strategies consolidate over extended training periods and whether anticipatory responses diminish with expertise.