View in EDS HTML Full Text PDF Full Text

Decoding Target Discriminability and Time Pressure Using Eye and Head Movement Features in a Foraging Search Task

Saved in:
Bibliographic Details
Title: Decoding Target Discriminability and Time Pressure Using Eye and Head Movement Features in a Foraging Search Task
Language: English
Authors: Anthony J. Ries (ORCID 0009-0004-3440-6276), Chloe Callahan-Flintoft, Anna Madison, Louis Dankovich, Jonathan Touryan
Source: Cognitive Research: Principles and Implications. 2025 10.
Availability: Springer. Available from: Springer Nature. One New York Plaza, Suite 4600, New York, NY 10004. Tel: 800-777-4643; Tel: 212-460-1500; Fax: 212-460-1700; e-mail: customerservice@springernature.com; Web site: https://link.springer.com/
Peer Reviewed: Y
Page Count: 19
Publication Date: 2025
Sponsoring Agency: US Army Futures Command, Combat Capabilities Development Command Soldier Center (DEVCOM)
Contract Number: W911NF2320205
Document Type: Journal Articles
Reports - Research
Descriptors: Stress Variables, Time Management, Decision Making, Reaction Time, Eye Movements, Human Body, Motion, Military Training, Weapons, Search Strategies, Assistive Technology, Computer Simulation, Discrimination Learning, Performance Based Assessment, Kinetics, Visual Perception, Visual Learning, Classification
DOI: 10.1186/s41235-025-00657-y
ISSN: 2365-7464
Abstract: In military operations, rapid and accurate decision-making is crucial, especially in visually complex and high-pressure environments. This study investigates how eye and head movement metrics can infer changes in search behavior during a naturalistic shooting scenario in virtual reality (VR). Thirty-one participants performed a foraging search task using a head-mounted display (HMD) with integrated eye tracking. Participants searched for targets among distractors under varying levels of target discriminability (easy vs. hard) and time pressure (low vs. high). As expected, behavioral results indicated that increased discrimination difficulty and greater time pressure negatively impacted performance, leading to slower response times and reduced d-prime. Support vector classifiers assigned a search condition, discriminability and time pressure, to each trial based on eye and head movement features. Combined eye and head features produced the most accurate classification model for capturing tasked-induced changes in search behavior, with the combined model outperforming those based on eye or head features alone. While eye features demonstrated strong predictive power, the inclusion of head features significantly enhanced model performance. Across the ensemble of eye metrics, fixation-related features were the most robust for classifying target discriminability, while saccadic-related features played a similar role for time pressure. In contrast, models constrained to head metrics emphasized global movement (amplitude, velocity) for classifying discriminability but shifted toward kinematic intensity (acceleration, jerk) in time pressure condition. Together these results speak to the complementary role of eye and head movements in understanding search behavior under changing task parameters.
Abstractor: As Provided
Entry Date: 2025
Accession Number: EJ1481527
Database: ERIC
Full text is not displayed to guests.
FullText Links:
  – Type: pdflink
    Url: https://content.ebscohost.com/cds/retrieve?content=AQICAHj0k_4E0hTGH8RJwT4gCJyBsGNe_WN95AvKlDbXJGqwxwGl-KqnkbUKE8g1NlYHLY7WAAAA4zCB4AYJKoZIhvcNAQcGoIHSMIHPAgEAMIHJBgkqhkiG9w0BBwEwHgYJYIZIAWUDBAEuMBEEDCB4qP-LTgxUR3035QIBEICBm-K9ECCrBL9GavczsZA4G-5bDVlfOXcu_mZ5ecs3mItz9pdx8O3N0Db1aYhIHvXhezWcPGE9QnDTQWqjvAcHRxtJMFjiWD-aai5mADiXJ_7iKsHzfbOxJ3S0GMjZtbyxmm3aOd6uBjJASQ1MVHf3rCTjYbVdSCsQNO4JLGF_OlegeQcl4YzmRY1R-c6VOQwNVN0-_x3--OeSz1iB
Text:
  Availability: 1
  Value: <anid>AN0187497571;[k1e6]22aug.25;2025Aug26.02:45;v2.2.500</anid> <title id="AN0187497571-1">Decoding target discriminability and time pressure using eye and head movement features in a foraging search task </title> <p>In military operations, rapid and accurate decision-making is crucial, especially in visually complex and high-pressure environments. This study investigates how eye and head movement metrics can infer changes in search behavior during a naturalistic shooting scenario in virtual reality (VR). Thirty-one participants performed a foraging search task using a head-mounted display (HMD) with integrated eye tracking. Participants searched for targets among distractors under varying levels of target discriminability (easy vs. hard) and time pressure (low vs. high). As expected, behavioral results indicated that increased discrimination difficulty and greater time pressure negatively impacted performance, leading to slower response times and reduced d-prime. Support vector classifiers assigned a search condition, discriminability and time pressure, to each trial based on eye and head movement features. Combined eye and head features produced the most accurate classification model for capturing tasked-induced changes in search behavior, with the combined model outperforming those based on eye or head features alone. While eye features demonstrated strong predictive power, the inclusion of head features significantly enhanced model performance. Across the ensemble of eye metrics, fixation-related features were the most robust for classifying target discriminability, while saccadic-related features played a similar role for time pressure. In contrast, models constrained to head metrics emphasized global movement (amplitude, velocity) for classifying discriminability but shifted toward kinematic intensity (acceleration, jerk) in time pressure condition. Together these results speak to the complementary role of eye and head movements in understanding search behavior under changing task parameters.</p> <p>Keywords: Virtual reality; Eye tracking; Head movements; Visual search; Task prediction; Cognitive state; Visual foraging; Multiclass classification; Machine learning; Support vector machine; Information and Computing Sciences Artificial Intelligence and Image Processing</p> <p>Supplementary Information The online version contains supplementary material available at https://doi.org/10.1186/s41235-025-00657-y.</p> <hd id="AN0187497571-2">Introduction</hd> <p>In modern military operations, making quick, accurate decisions in visually complex environments is critical to mission success. This is particularly true for soldiers in dynamic, high-stress scenarios, requiring rapid ability to locate targets (e.g., potential threats in the environment) and differentiate them from distractors, which are sometimes visually similar (e.g., friendlies among enemy soldiers). The significant consequences of delayed or incorrect responses in these situations underscore the critical role of a soldier's attentional focus and their capacity to manage competing, time-sensitive demands. Collectively, these factors shape how soldiers perceive and respond to information, ultimately affecting their decision-making and the effectiveness of their actions.</p> <p>The military has shown an increasing interest in leveraging the computational power of artificial intelligence (AI) to help soldiers in complex, time-constrained decision-making. However, seamless integration with human teams requires providing AI with ongoing contextual information, such as changes in mission parameters or operator goals. This contextual information can potentially be inferred from changes in observable behaviors and human–system interactions. For example, delays in operator response times or atypical information requests from the AI may indicate that task parameters or environmental conditions have changed. Having this information and drawing such inferences would enable the AI to provide tailored assistance without imposing additional cognitive load on the operator by requiring explicit reports of task parameter changes. Real-time measurement and prediction of task-induced changes in search performance could significantly enhance decision-making, enabling effective adaptation to situational demands (Brunyé et al., [<reflink idref="bib8" id="ref1">8</reflink>]). For instance, AI could identify imbalances in cognitive load across a team and recommend task reallocation. As the military integrates AI into combat scenarios, detecting fluctuations in search behavior could inform autonomous systems of changes in the soldier's cognitive state or mission parameters, allowing the AI to adjust its behavior for real-time support to improve team efficiency (Marathe et al., [<reflink idref="bib64" id="ref2">64</reflink>]; Metcalfe et al., [<reflink idref="bib66" id="ref3">66</reflink>]; Reis et al., [<reflink idref="bib76" id="ref4">76</reflink>]).</p> <p>Changes in search behavior often reflect underlying shifts in cognitive state (Boot et al., [<reflink idref="bib7" id="ref5">7</reflink>]; Doshi & Trivedi, [<reflink idref="bib23" id="ref6">23</reflink>]; Eckstein, [<reflink idref="bib28" id="ref7">28</reflink>]; HEnderson et al., [<reflink idref="bib40" id="ref8">40</reflink>]; MacInnes et al., [<reflink idref="bib62" id="ref9">62</reflink>]). These changes can originate from both endogenous factors such as fatigue and memory (Ganesan et al., [<reflink idref="bib35" id="ref10">35</reflink>]; Le-Hoa Võ & Wolfe, [<reflink idref="bib58" id="ref11">58</reflink>]) and exogenous factors of the search array itself, like visual saliency (Christ & Abrams, [<reflink idref="bib15" id="ref12">15</reflink>]; Itti & Koch, [<reflink idref="bib48" id="ref13">48</reflink>]; Theeuwes, [<reflink idref="bib87" id="ref14">87</reflink>]; Yantis & Jonides, [<reflink idref="bib95" id="ref15">95</reflink>]). Classically, search behavior differences due to changes in task parameters have been studied using central tendency metrics calculated from a block of trials (e.g., average response times are slower when targets are heterogeneous compared to homogeneous, (Duncan & Humphreys, [<reflink idref="bib26" id="ref16">26</reflink>]). While this method is effective and has provided insight into numerous cognitive mechanisms, it does not allow for a trial-by-trial assessment of task-induced behavioral variance. Understanding these task-induced changes in behavior are essential to infer or "predict" the current task parameters. It is this prediction that could have meaningful impact in the design of future AI integration into human teams. As such, the current work seeks to demonstrate whether head and eye movement data (either together or separately) can be used to infer the current parameters of the human's search task.</p> <hd id="AN0187497571-3">Eye and head movements</hd> <p>Human eye gaze provides numerous metrics for assessing changes in search performance and information processing demands (GilChrist & Harvey, [<reflink idref="bib37" id="ref17">37</reflink>]; König et al., [<reflink idref="bib51" id="ref18">51</reflink>]; Marshall, [<reflink idref="bib65" id="ref19">65</reflink>]). Saccade characteristics, such as frequency and peak velocity, predict target detection accuracy and reflect workload variations during visually complex decision-making tasks (Boot et al., [<reflink idref="bib7" id="ref20">7</reflink>]; Di Stasi et al., [<reflink idref="bib21" id="ref21">21</reflink>]). Fixation metrics, including duration, frequency, and sequence, are strong predictors of target selection in object search tasks (Huang et al., [<reflink idref="bib45" id="ref22">45</reflink>]) and are effective for distinguishing between classes of images (Karessli et al., [<reflink idref="bib49" id="ref23">49</reflink>]). Scanpath length captures search strategies, with longer scanpaths suggesting a global or ambient processing approach that prioritizes information outside the immediate field of view (Groner et al., [<reflink idref="bib39" id="ref24">39</reflink>]; Velichkovsky et al., [<reflink idref="bib89" id="ref25">89</reflink>]). Lower decision certainty is associated with fewer, longer fixations and slower saccades (Brunyé & Gardony, [<reflink idref="bib9" id="ref26">9</reflink>]). In high-stress shooting scenarios, elite police officers exhibit longer fixations and faster response times on relevant targets compared to novices, reflecting more efficient search strategies (Vickers & Lewinski, [<reflink idref="bib90" id="ref27">90</reflink>]). Additionally, fixation frequency, duration, and pupil diameter reliably track fluctuations in cognitive workload (Enders et al., [<reflink idref="bib30" id="ref28">30</reflink>]; Mallick et al., [<reflink idref="bib63" id="ref29">63</reflink>]; Pomplun & Sunkara, [<reflink idref="bib73" id="ref30">73</reflink>]; Van Orden et al., [<reflink idref="bib88" id="ref31">88</reflink>]). Collectively, these gaze-based metrics illustrate how both spatial and temporal characteristics of eye movements can be used to evaluate task-induced changes in search behavior across various scenarios.</p> <p>Head movements, though less frequently measured in visual search studies, complement eye-tracking metrics by providing unique information on the spatiotemporal characteristics of search behavior in visually complex and dynamic tasks (Agtzidis et al., [<reflink idref="bib1" id="ref32">1</reflink>]; Bischof et al., [<reflink idref="bib6" id="ref33">6</reflink>]; Callahan-Flintoft et al., [<reflink idref="bib12" id="ref34">12</reflink>]; Doshi & Trivedi, [<reflink idref="bib23" id="ref35">23</reflink>]). Head movements reveal broader shifts in attention and spatial orientation, especially during large gaze shifts or rapid reorienting (Pelz et al., [<reflink idref="bib70" id="ref36">70</reflink>]; Stahl, [<reflink idref="bib85" id="ref37">85</reflink>]). Day to day visual search often requires scanning large regions of the visual field, using peripheral vision to detect and respond to multiple, potentially hidden targets. Integrating head movement metrics into search performance assessments, therefore, may enhance predictive power, contributing to more comprehensive and ecologically valid models of perception and decision-making.</p> <p>Visual search in the real world occurs, most often, in a cluttered and dynamic three-dimensional environment where the head and eyes must move in concert to acquire necessary information for target detection. It is imperative then to explore these two movement systems in tandem for a deeper understanding of the cognitive processes underlying search behavior (Bischof et al., [<reflink idref="bib6" id="ref38">6</reflink>]; Sidenmark & Gellersen, [<reflink idref="bib80" id="ref39">80</reflink>]). Head and eye movements may differentially support cognitive function depending on whether task-relevant information lies in the peripheral or central regions of the current reference frame (David et al., [<reflink idref="bib20" id="ref40">20</reflink>]; Draschkow et al., [<reflink idref="bib24" id="ref41">24</reflink>]; Solman & Kingstone, [<reflink idref="bib82" id="ref42">82</reflink>]). For instance, when relevant information is spatially dispersed, participants may engage a trade-off between making head movements and engaging working memory storage (Draschkow et al., [<reflink idref="bib24" id="ref43">24</reflink>]). Additionally, individuals differ in their reliance on head versus eye movements during visual search, with some favoring head movements for reorienting; a distinction that may reflect different cognitive strategies, such as"head movers"versus"non-movers"(Fuller, [<reflink idref="bib34" id="ref44">34</reflink>]; Stahl, [<reflink idref="bib84" id="ref45">84</reflink>], [<reflink idref="bib85" id="ref46">85</reflink>]). Together, these findings speak to the importance of both eye and head movements in supporting visual search execution and suggest that both systems may be affected by changes in task parameters.</p> <hd id="AN0187497571-4">Virtual reality (VR) and visual foraging</hd> <p>Virtual reality (VR) offers unique capabilities to study visual search and other tasks that closely mimic real-world interactions (Bischof et al., [<reflink idref="bib6" id="ref47">6</reflink>]; Callahan-Flintoft et al., [<reflink idref="bib12" id="ref48">12</reflink>]; Clay et al., [<reflink idref="bib17" id="ref49">17</reflink>]). Traditional visual search paradigms, typically confined to two-dimensional displays, offer precise control over experimental variables but lack the immersive, spatial interactions essential for capturing natural head and eye movements. VR provides realistic, three-dimensional environments that replicate the spatial complexity of real-world search tasks, enabling the study of cognitive processes and the natural coordination of eye and head movements in controlled, yet realistic settings (Callahan-Flintoft et al., [<reflink idref="bib12" id="ref50">12</reflink>], [<reflink idref="bib13" id="ref51">13</reflink>]; Clay et al., [<reflink idref="bib17" id="ref52">17</reflink>]; Sidenmark & Gellersen, [<reflink idref="bib80" id="ref53">80</reflink>]). Capturing both eye and head movements through integrated head-mounted display (HMD) systems, VR provides data on focal and peripheral search behavior which are essential elements of situational awareness in military operations (Brunyé & Giles, [<reflink idref="bib10" id="ref54">10</reflink>]; Pettersson et al., [<reflink idref="bib71" id="ref55">71</reflink>]; Sidenmark & Gellersen, [<reflink idref="bib80" id="ref56">80</reflink>]).</p> <p>Similarly, visual foraging tasks provide an ecologically valid model for studying decision-making and attention allocation in multi-target scenarios (Cain et al., [<reflink idref="bib11" id="ref57">11</reflink>]; Kristjánsson et al., [<reflink idref="bib54" id="ref58">54</reflink>]; Wolfe, [<reflink idref="bib94" id="ref59">94</reflink>]). Unlike traditional single-target search tasks, visual foraging involves searching for and selecting multiple targets within a scene, reflecting real-world situations where multiple potential targets may arise. This complexity closely resembles challenges faced by soldiers who must identify and respond to multiple threats in combat, where prioritizing targets and filtering distractions is critical (Kristjánsson et al., [<reflink idref="bib55" id="ref60">55</reflink>]). Research shows that participants adapt their search strategies in foraging tasks based on target/non-target discriminability and time constraints, with more rapid switching between targets under high time pressure and less target switching when target discrimination is difficult (T. Kristjánsson et al., [<reflink idref="bib56" id="ref61">56</reflink>]). Such conditions are also associated with increased saccade amplitude and higher fixation frequency (Tagu & Kristjánsson, [<reflink idref="bib86" id="ref62">86</reflink>]). These findings highlight the flexibility of search strategies in foraging tasks and underscore the value of VR for examining cognitive processes underlying naturalistic visual search, particularly in environments where target discrimination and time pressure impose distinct task demands.</p> <hd id="AN0187497571-5">Current study</hd> <p>The current study explores the potential of eye and head movement metrics to predict changes in target discriminability and time pressure. Participants engaged in a naturalistic foraging search task within a VR forest environment using a head-mounted display (HMD) equipped with integrated eye and head tracking. Participants were required to identify and engage targets among distractors under varying levels of discrimination difficulty (easy vs. hard) and time pressure (low vs. high). By analyzing the collected eye and head movement data, we aimed to determine the predictive power of these metrics in classifying task parameter changes.</p> <p>Here, participants performed a foraging task while standing in one spot (i.e., without ambulation), enabling us to capture the natural coordination of eye and head movements during immersive search while still providing strict control on the spatial relationship with the viewer and the search array. This set-up serves as a midway point for ultimately understanding search in a fully ambulatory context. However, even though the participant is stationary, allowing the eyes and head to move freely during search provides a wealth of spatiotemporal features, any number of which could be influenced by changes in task parameters. As such, the current work adopts a data-driven approach in employing machine learning to disentangle whether eye and head movement is predictive of changes in search context. Support vector machines (SVM) and neural networks have been effective in classifying task-induced changes in behavior based on eye movement data (El Iskandarani et al., [<reflink idref="bib29" id="ref63">29</reflink>]; Hulle et al., [<reflink idref="bib46" id="ref64">46</reflink>]; Kothari et al., [<reflink idref="bib52" id="ref65">52</reflink>]; MacInnes et al., [<reflink idref="bib62" id="ref66">62</reflink>]). However, incorporating head movement metrics has the potential to significantly improve classification accuracy or provide an alternative for predicting behavioral changes when eye-tracking measurements are not feasible or practical (Hu et al., [<reflink idref="bib44" id="ref67">44</reflink>]). Recent evidence demonstrates that the temporal derivatives associated with head movements, velocity, acceleration, and jerk can predict task-relevant physiological responses in VR environments (Salehi et al., [<reflink idref="bib77" id="ref68">77</reflink>]). In addition to eye metrics, integrating head kinematic data, including rotation velocity, acceleration, jerk, and more traditional features such as amplitude, duration, and frequency, into machine learning models may capture complimentary information relevant to classifying search behaviors and provide insights that more classic analysis approaches could miss.</p> <hd id="AN0187497571-6">Method</hd> <p></p> <hd id="AN0187497571-7">Participants</hd> <p>The Institutional Review Boards at the U.S. Air Force Academy (USAFA) and the U.S. Army Combat Capabilities Development Command (DEVCOM) approved this study (ARL 21–132), and all procedures followed the guidelines of the Declaration of Helsinki. A total of 33 United States Air Force Academy (USAFA) cadets participated in the study. They were recruited through the Sona Systems subject pool and received course credit for participation. Two participants experienced technical difficulties with the system, and one experienced vertigo and did not complete the task. The data from these three participants were not analyzed. Of the 30 final participants, 24 were male, and 6 were female, average age of 19.1 years. Self-reported handedness showed 3 were left-handed, 26 were right-handed and 1 ambidextrous. Before beginning the experiment, participants provided written informed consent. To ensure participants had normal color vision they completed the Snellen chart (at least 20/40 vision required) and Ishihara color plates. Each of the 30 participants completed the vision test successfully without eyeglasses.</p> <hd id="AN0187497571-8">Materials</hd> <p>The experiment was programmed in Unity (2020.3.44f1) and presented in the HTC Vive Pro Eye VR headset (1440 × 1600 pixels per eye, 90 Hz refresh, 110 deg field of view, VIVE SRanipal SDK) with embedded Tobii eye tracking (120 Hz sampling rate, Tobii XR SDK). The experimental environment was presented using a Corsair One PC (Windows 10, Intel Core i9 CPU @ 3.6 GHz, 64-bit, Nvidia GeForce RTX2080Ti, 32 GB RAM). Two external lighthouses tracked the head, the torso using a Vive Tracker, and hand controller positioned in a ForceTube gunstock (ProTube VR) to simulate a weapon.</p> <hd id="AN0187497571-9">Design and procedure</hd> <p>After signing informed consent and undergoing the vision tests, participants completed a demographics form and the Simulator Sickness Questionnaire (SSQ) (Kennedy et al., [<reflink idref="bib50" id="ref69">50</reflink>]). A Vive tracker was then attached to the center of the chest, and the participant received general instruction on how to hold the weapon and perform the foraging task. Next participants were centered on a foam pad (2 × 3 ft) between the lighthouses. Participants could rotate their torso but were instructed to always keep their feet on the foam pad. The pad was used as it provided a proprioceptive reminder (i.e., the contrast between the foam and the floor) to participants to stay within the experimental space. Once positioned, the participant was handed the weapon controller to use in whatever hand they preferred and then the VR headset was donned and adjusted for fit and comfort. Next, participants adjusted the inter-pupil distance to ideal parameters determined by Tobii software and performed a 5-point eye-tracking calibration. After a successful calibration (determined by Tobii software), participants were given practice with the task to get familiar with the environment, the target and distractor objects, and aiming and shooting the weapon.</p> <p>Participants performed a foraging search task within a forested environment (Meadow Environment—Dynamic Nature Version 2.9.3 created by Nature Manufacture) by shooting targets while trying to not shoot distractors. Participants shot the weapon by depressing the trigger on the hand controller, which produced an auditory gunshot. A visual ray (red line) extended from the weapon into the environment to assist aiming. Targets were kneeling avatars holding a weapon and wearing a specific camouflage pattern (Fig. 1). Targets were intermixed with distractors. Distractors were avatars also holding a weapon, in the same kneeling positions as targets, but wearing a camouflage pattern that was either easy to discriminate (Easy condition) or difficult to discriminate (Hard condition) from the target camouflage pattern. Four dominant colors were used in each avatar.[<reflink idref="bib1" id="ref70">1</reflink>] Targets and distractors shot their weapons to generate background gunshot sounds but did not aim at the participant or fire projectiles in the virtual environment. If a target or distractor was shot, it fell to the ground and ceased shooting. The environment was rendered 360 degrees; however, camouflage targets and distractors were spawned without environment occlusion, only within 180 degrees of the game origin point (i.e., forward center position). Spawn points were a random distance between 10 (4.24 × 1.72 degrees of visual angle (dva)) and 100 m (1.1 ×.5 dva). Trials were completed with low time pressure (45 s in the Low condition) or high time pressure (18 s in the High condition). Hard discrimination difficulty and high time pressure levels were determined a priori through pilot testing to ensure they were challenging yet accomplishable. The angular separation between targets within the scene varied by trial with half of the trials having a target dispersion of 5 dva and the other half 10 dva. For the purposes of this study, trials were collapsed across dispersion levels, leaving a 2 × 2 within-subjects design with target discriminability (Easy vs. Hard) and time pressure (Low vs Hard) as the main factors.</p> <p>Graph: Fig. 1 Task environment. A. Screenshot of the easy condition in the foraging shooting task. The red ray extending from the weapon indicates current aiming position. B. Examples of the search target and distractors in the easy and hard conditions. C. Spatial coordinates of the VR environment and head position (+ X right, + Y up, and + Z forward, left-handed coordinate system) with corresponding yaw, pitch, and roll rotations. Eye position data were recorded in a right-handed coordinate system (+ X left, + Y up, and + Z forward)</p> <p>Each participant completed 4 blocks of 20 trials. Due to a pre-configuration error, one participant completed 6 additional trials having a target dispersion of 8 dva. These trials only occurred on the first block and were not included in the analysis. Each trial contained between 12 and 18 targets (uniformly distributed) and 12 distractors. This was implemented to prevent participants counting a fixed number of targets. Prior to each block a 5-point eye tracker calibration was performed. Each trial started with a blank background containing a plus sign + at the world forward position (0,0,1). When the gaze position remained continuously centered within 2dva and the head within 3dva of the central marker for 500 ms, the forested environment with targets and distractors was displayed. Participants were instructed to shoot all targets using the trigger button on the controller/weapon and press down on the controller track pad with their thumb when they were finished searching. If the trial time limit was reached (i.e., 45 or 18 s), the trial ended. Discriminability (as well as target dispersion) was randomized within each block with the constraint that each difficulty/dispersion trial combination was equally represented in the block. The first two blocks were low time pressure, and the last two blocks were high time pressure. Time pressure was blocked such that the first two blocks were always low pressure, and the final two blocks were high pressure. This fixed order was used to minimize carryover effects observed during pilot testing, where beginning with high pressure caused participants to adopt a sustained fast-paced search strategy that persisted into subsequent low-pressure blocks. Presenting low pressure first allowed participants to engage in more deliberate search behavior before transitioning into a more time-constrained strategy, better preserving the intended contrast between conditions. After the experiment the participants completed another SSQ as well as the three-component iGroup Presence Questionnaire (IPQ) (ScHubert, [<reflink idref="bib78" id="ref71">78</reflink>]). Summary results of the SSQ and IPQ can be found in Supplementary Materials.</p> <hd id="AN0187497571-10">Data acquisition</hd> <p>Each unity frame, eye, head, and controller samples were timestamped, recorded, and synchronized with other experimental events (e.g., trial start/stop, target/distractor labels and positions) through the Lab Streaming Layer (LSL) and Lab Recorder (Kothe et al., [<reflink idref="bib53" id="ref72">53</reflink>]). LSL serves as a framework for synchronizing time-series data from various sources and sampling rates, offering a language-independent, real-time communication protocol for data recording and sharing. Due to a coding error, the Vive tracker used to track torso rotation did not record properly for any participant. Unfortunately, as a result, these data could not analyzed.</p> <hd id="AN0187497571-11">Data processing</hd> <p>X, Y and Z coordinates from the left eye and head were used to calculate multiple metrics for the analyses (see Fig. 1 and Table 1) (Nyström & Holmqvist, [<reflink idref="bib67" id="ref73">67</reflink>]; Vlaskamp et al., [<reflink idref="bib92" id="ref74">92</reflink>]). Eye-tracking samples were deemed invalid if the pupil diameter was recorded as less than or equal to zero. If the time window of consecutive invalid samples was under 50 ms in duration, a linear interpolation was used to fill this gap. If the period extended longer than that, those samples, along with the neighboring 2 samples on either side of the drop-out, were marked as unusable and not included in saccade classification. Saccades and fixations were identified with a velocity-based algorithm adapted from the EYE-EEG toolbox (Dimigen et al., [<reflink idref="bib22" id="ref75">22</reflink>]; Engbert & Mergenthaler, [<reflink idref="bib31" id="ref76">31</reflink>]). Each participant's eye unit vector was used to calculate angular velocity. The average velocity and standard deviation in each direction (x, y, z) were calculated, and a 3-dimensional, elliptical velocity threshold was set at six standard deviations from the mean in each direction. In this way, each participant had an automated, custom saccade threshold. Saccades were rejected if they were under 12 ms in duration (Hooge et al., [<reflink idref="bib42" id="ref77">42</reflink>]).</p> <p>Table 1 Eye and head features used in classification models to infer task-induced effects of target discriminability and time pressure</p> <p> <ephtml> <table frame="hsides" rules="groups"><thead><tr><th align="left"><p>Metric</p></th><th align="left"><p>Definition</p></th><th align="left"><p>Significance</p></th><th align="left"><p>SupportingLiterature</p></th></tr></thead><tbody><tr><td align="left" colspan="4"><p>Fixation</p></td></tr><tr><td align="left"><p>Fixation Frequency</p><p>Avg Fixation Duration</p><p>Fixation Duration Proportion</p><p>Target Fixation Frequency</p><p>Avg Target Fixation Duration</p><p>Target Fixation Duration Proportion</p><p>Distractor Fixation Frequency</p><p>Avg Distractor Fixation Duration</p><p>Distractor Fixation Duration Proportion</p><p>Target Fixations per Target</p><p>Distractor Fixations per Distractor</p><p>Target Distractor Fixation Ratio</p></td><td align="left"><p>Total fixation count divided by trial time (sec)</p><p>Average fixation time (sec)</p><p>Sum of all fixation durations divided by trial time</p><p>Total target fixation count divided by trial time</p><p>Average target fixation time (sec)</p><p>Sum of all target fixation durations divided by trial time</p><p>Total distractor fixation count divided bytrial time</p><p>Average distractor fixation time (sec)</p><p>Sum of all distractor fixation durations divided by trial time</p><p>Total number of target fixations divided by total number of targets</p><p>Total number of distractor fixations divided by total number of distractors</p><p>Target fixation preference compared to other fixations</p></td><td align="left"><p>FixationFrequency: Reflects how often attention shifts to</p><p>specific visual location. Higher frequencies may indicate</p><p>increased visual samplingor search effort. Fixation Duration:</p><p>Longer fixations typically reflect deeper processing, uncertainty,</p><p>or hesitation; shorter durations may indicate more confident or</p><p>automatic responses. Fixation Duration Proportion: Indicates</p><p>how much of the trial is spent fixating; higher values suggest</p><p>greter engagment or deliberation. Target vs. Distractor</p><p>Fixations: Target-focused metrics reflect goal-directed</p><p>attention while distractor-focused metrics may indicate</p><p>inefficient search or difficulty in discrimination</p></td><td align="left"><p>Drew et al., (<xref ref-type="bibr" rid="bibr25">2017</xref>); Over et al., (<xref ref-type="bibr" rid="bibr42">2007</xref>); Reingold and Galhot, (<xref ref-type="bibr" rid="bibr75">2014</xref>); Lu and Sarter, (<xref ref-type="bibr" rid="bibr61">2019</xref>); Huanget al., (<xref ref-type="bibr" rid="bibr45">2015</xref>); Karessli et al., (<xref ref-type="bibr" rid="bibr49">2017</xref>); Iskandarani et al., (<xref ref-type="bibr" rid="bibr29">2024</xref>)</p></td></tr><tr><td align="left" colspan="4"><p>Saccade</p></td></tr><tr><td align="left"><p>Saccade Frequency</p><p>AvgSaccade Size</p><p>Saccade Size Per Second</p><p>AvgTarget Saccade Size</p><p>AvgDistractor Saccade Size</p></td><td align="left"><p>Total saccade count divided bytrial time (sacades/s)</p><p>Average size of a saccade(dva)</p><p>Sum of all saccade magnitudes divided bytrial time (°/s)</p><p>Average size of saccade to a target (dva)</p><p>Average size of saccade to a distractor (dva)</p></td><td align="left"><p>Saccade Frequency: Higher frequencymayreflect rapid</p><p>scanningor uncertaintyin searchstrategy. Saccade Size and</p><p>Size per Second: Larger or more frequent saccades may</p><p>suggest a global search strategywhile smaller saccades may</p><p>indicate more focused, precisesearch behavior. Target vs</p><p>Distractor Saccade Size: Differentiates the visual distance the</p><p>eye travels totask-relevant vs. irrelevant stimuli, potentially</p><p>reflectingdiscrimination efficiency</p></td><td align="left"><p>Watson et al., (<xref ref-type="bibr" rid="bibr93">2010</xref>); Over et al., (<xref ref-type="bibr" rid="bibr42">2007</xref>); Di Stasi et al., (<xref ref-type="bibr" rid="bibr21">2011</xref>); Velichkovski et al., (<xref ref-type="bibr" rid="bibr89">2002</xref>); Gordonet al., (<xref ref-type="bibr" rid="bibr38">2024</xref>)</p></td></tr><tr><td align="left" colspan="4"><p>Head</p></td></tr><tr><td align="left"><p>Head Movement Frequency</p><p>AvgHead Movement Size</p><p>Head Movement Size Per Second</p><p>AvgHead Fixation Duration</p><p>Head Fixation Duration Proportion</p><p>Head Fixation Frequency</p><p>Global Head Movement Per Second</p></td><td align="left"><p>Total head movement count divided bytrial time</p><p>Average size of a head movement (degrees of visual angle)</p><p>Sum of all head movement magnitudes divided bytrial time (°/s)</p><p>Average time the head remains below the threshold of a head movement (sec)</p><p>Sum of all head fixation durations divided bytrial time</p><p>Total head fixation count divided bytrial time</p><p>Sum of the angular distancethe head moved dividedbytrial time Independent of</p><p>head movement classification</p></td><td align="left"><p>Head Movement Frequencyand Size: Capturehow often and</p><p>how far users reorient their field of view; higher values may</p><p>relfect broader or more effortful searchstrategies. Head</p><p>FixationDurationand Proportion: Longer or more frequent</p><p>headfixations mayreflect more focused attention or reduced</p><p>exploratorybehavior. Global Head Movement: Indicates spatial</p><p>engagement withthe scene</p></td><td align="left"><p>Hu et al., (<xref ref-type="bibr" rid="bibr44">2023</xref>); David et al., (<xref ref-type="bibr" rid="bibr19">2018</xref>); Fuller, (<xref ref-type="bibr" rid="bibr34">1992</xref>); Stahl, (<xref ref-type="bibr" rid="bibr83">1999</xref>); Oommen and Stahl, (<xref ref-type="bibr" rid="bibr68">2005</xref>); Freedman and Sparks, (<xref ref-type="bibr" rid="bibr33">1997</xref>)</p></td></tr><tr><td align="left" colspan="4"><p>Eye</p></td></tr><tr><td align="left"><p>AvgEye in Head Angle</p><p>Global Eye Movement Per Second</p><p>AvgEye to Gaze Proportion</p></td><td align="left"><p>Average rotation of the eye within the head (dva)</p><p>Sum of the angular distancethe eye moved divided bytrial time. Independent of saccade classification</p><p>Average proportion of eye contribution to gazeangle. Gaze Angle = Eye Angle + Head Angle</p></td><td align="left"><p>Eye inHead Angle: Larger anglesmayreflect relianceon eye movements rather than head shifts. Eye-to-Gaze Proportion: important for undertandingsearch strategypreference (e.g."eye movers"vs."head movers")</p></td><td align="left"><p>Sidenmark & Gellersen, (<xref ref-type="bibr" rid="bibr80">2019</xref>, <xref ref-type="bibr" rid="bibr81">2023</xref>); Stahl, (<xref ref-type="bibr" rid="bibr84">2001a</xref>, <xref ref-type="bibr" rid="bibr85">2001b</xref>); Freedman, (<xref ref-type="bibr" rid="bibr32">2008</xref>); Zangemeister and Stark, (<xref ref-type="bibr" rid="bibr98">1982</xref>); Doshi and Trividi, (<xref ref-type="bibr" rid="bibr23">2012</xref>)</p></td></tr><tr><td align="left" colspan="4"><p>Kinematic</p></td></tr><tr><td align="left"><p>Head</p></td><td align="left" /><td align="left" /><td align="left" /></tr><tr><td align="left"><p>AvgHead Velocity</p><p>Std of Head Velocity</p><p>AvgHead VelocityX,Y,Z</p><p>Std of Head VelocityX,Y,Z</p><p>AvgHead Acceleration</p><p>Std of Head Acceleration</p><p>AvgHead Acceleratiaround the X,Y,Z</p><p>Std of Head Acceleratiaround the X,Y,Z</p><p>AvgHead Jerk</p><p>Std of Head Jerk</p><p>AvgHead Jerk X,Y,Z</p><p>Std of Head Jerk X,Y,Z</p><p>Eye</p><p>AvgEye Velocity</p><p>Std of Eye Velocity</p><p>AvgEye VelocityX,Y,Z</p><p>Std Eye VelocityX,Y,Z</p><p>AvgEye Acceleration</p><p>Std of Eye Acceleration</p><p>AvgEye Acceleratiaround the X,Y,Z</p><p>Std Eye Acceleratiaround the X,Y,Z</p><p>AvgEye Jerk</p><p>Std of Eye Jerk</p><p>AvgEye Jerk X,Y,Z</p><p>Std Eye Jerk X,Y,Z</p></td><td align="left"><p>Average headvelocity(°/s)</p><p>Standard deviation of head velocity(°/s)</p><p>Average headvelocityaround the X, Y, and Zaxes (°/s)</p><p>Standard deviation of head velocityaround the X, Y, and Zaxes(°/s)</p><p>Average headacceleration (°/s<sup>2</sup>)</p><p>Standard deviation of head acceleration(°/s<sup>2</sup>)</p><p>Average headacceleration around the X, Y, and Zaxes(°/s<sup>2</sup>)</p><p>Standard deviation of head acceleration around the X, Y, and Zaxes (°/s<sup>2</sup>)</p><p>Average headjerk (°/s<sup>3</sup>)</p><p>Standard deviation of head jerk (°/s<sup>3</sup>)</p><p>Average headjerk around the X, Y, and Zaxes (°/s<sup>3</sup>)</p><p>Standard deviation of head jerk around the X, Y, and Zaxes(°/s<sup>3</sup>)</p><p>Average eye velocity(°/s)</p><p>Standard deviation of eye velocity(°/s)</p><p>Average eye velocityaroundthe X, Y, and Zaxes (°/s)</p><p>Standard deviation eye velocityaround the X, Y, and Zaxes(°/s)</p><p>Average eye acceleration (°/s<sup>2</sup>)</p><p>Standard deviation of eye acceleration (°/s<sup>2</sup>)</p><p>Average eye acceleration around the X, Y, and Zaxes (°/s<sup>2</sup>)</p><p>Standard deviation eye acceleration around the X, Y, and Zaxes (°/s<sup>2</sup>)</p><p>Average eye jerk (°/s<sup>3</sup>)</p><p>Standard deviation of eye jerk (°/s<sup>3</sup>)</p><p>Average eye jerk around the X, Y, and Zaxes (°/s<sup>3</sup>)</p><p>Standard deviation eye jerk around the X, Y, and Zaxes(°/s<sup>3</sup>)</p></td><td align="left"><p>Velocity: Indicates speed of eye or head movement; higher</p><p>values mayreflect urgencyor less deliberate behavior</p><p>Acceleration: Measures the rate of movement change. Jerk:</p><p>Captures abruptnessor smoothness of movement; higher jerk</p><p>suggestsmore reactiveor less controlled movements</p><p>Variability(Standard Deviation): Grater variabilityacross</p><p>velocity, acceleration, or jerkmayreflect lessstabel or more</p><p>exploratorysearch strategies</p></td><td align="left"><p>Di Stasi et al., (<xref ref-type="bibr" rid="bibr21">2011</xref>); Salehi et al., (<xref ref-type="bibr" rid="bibr77">2024</xref>); Hu et al., (<xref ref-type="bibr" rid="bibr44">2023</xref>); Lisberger et al., (<xref ref-type="bibr" rid="bibr59">1981</xref>)</p></td></tr></tbody></table> </ephtml> </p> <p>To exclude biologically implausible saccades, an additional set of thresholds was applied. Specifically, saccades longer than 120 ms or with peak velocities below 25 deg/sec or above 1000 deg/sec were removed (Enders et al., [<reflink idref="bib30" id="ref78">30</reflink>]; Holmqvist et al., [<reflink idref="bib41" id="ref79">41</reflink>]; Long et al., [<reflink idref="bib60" id="ref80">60</reflink>]). Saccades and head movements with amplitudes ≤ 1 deg and fixations (eye and head) shorter than 75 ms were also excluded from the analysis (Hooge et al., [<reflink idref="bib43" id="ref81">43</reflink>]). Saccade and fixation-based metrics were calculated after removing these outlier events.</p> <p>To classify head movements, we applied an adapted algorithm from Chen and Walton ([<reflink idref="bib14" id="ref82">14</reflink>]), originally designed to classify head movements in macaques but shown to be highly reliable when classifying human head movements (Callahan-Flintoft et al., [<reflink idref="bib13" id="ref83">13</reflink>]). Head movements were detected by applying a sliding 100 ms window to the head's angular speed. Head motion onset began when at least 72% of data points within this window exceeded a threshold of 6°/s, with fewer than three consecutive time points falling below the threshold. Head movement onset was recorded as the first time point in the window exceeding the threshold. Similarly, motion offset was determined using a 22 ms window, where 72% of data points were below the threshold. The exact head movement offset time was defined as the first time point within this qualifying window that dropped below the threshold. Head 'fixations' were periods between classified head movements where the head sustained a low angular speed and remained relatively stable.</p> <hd id="AN0187497571-12">Metrics calculation</hd> <p>Trial completion time, average inter-target engagement time and d-prime were evaluated to assess behavioral performance. Target completion time was measured from scene onset to when the participant pressed the trial end button indicating search was complete or when time expired, whichever came first. Inter-target time was the average time taken between successive target hits. D-prime was used to distinguish between signal and noise by quantifying the difference in hit and false alarm rates. This provided a better measure of sensitivity to the differences between targets and distractors while accounting for potential response bias when compared to target accuracy alone. Analyses were performed in R (RStudio 2023.12.1 + 402"Ocean Storm"Release).</p> <p>Multiple metrics were computed using the rotational data of eye and head movements along the X (pitch), Y (yaw), and Z (roll) axes, with many derived across both effectors. We focused on core temporal and spatial characteristics of movement, including frequency, amplitude, and duration, features commonly used to characterize attentional engagement and search strategy in visually demanding tasks. For example, fixation duration and frequency have been linked to visual processing load and search efficiency (Drew et al., [<reflink idref="bib25" id="ref84">25</reflink>]; Reingold & Glaholt, [<reflink idref="bib75" id="ref85">75</reflink>]). Saccadic peak velocity has been proposed as an index for detecting variations in mental workload (Di Stasi et al., [<reflink idref="bib21" id="ref86">21</reflink>]), and saccade size has been shown to predict the onset time of object processing during visual search tasks (Gordon et al., [<reflink idref="bib38" id="ref87">38</reflink>]). Similarly, head movement amplitude and frequency reflect spatial reorientation and can signal increased search demands or target uncertainty (David et al., [<reflink idref="bib19" id="ref88">19</reflink>]; Freedman & Sparks, [<reflink idref="bib33" id="ref89">33</reflink>]; Oommen & Stahl, [<reflink idref="bib68" id="ref90">68</reflink>]). We also incorporated gaze and head kinematics which have shown promise in identifying cognitive state changes and user experience during complex VR tasks (El Iskandarani et al., [<reflink idref="bib29" id="ref91">29</reflink>]; Hu et al., [<reflink idref="bib44" id="ref92">44</reflink>]).</p> <p>To capture the dynamics of eye and head movements more comprehensively, we calculated kinematic features based on velocity, acceleration, and jerk (Culmer et al., [<reflink idref="bib18" id="ref93">18</reflink>]; Di Stasi et al., [<reflink idref="bib21" id="ref94">21</reflink>]; Hu et al., [<reflink idref="bib44" id="ref95">44</reflink>]; Kothari et al., [<reflink idref="bib52" id="ref96">52</reflink>]; Lisberger et al., [<reflink idref="bib59" id="ref97">59</reflink>]; Plamondon, [<reflink idref="bib72" id="ref98">72</reflink>]; Salehi et al., [<reflink idref="bib77" id="ref99">77</reflink>]). Velocity refers to the rate of change of angular position over time. It provides insight into how quickly the eyes and head move during gaze shifts, reflecting the speed of orienting behavior. Velocity was computed as the first derivative of the angular position with respect to time for each axis. Acceleration is the rate of change of velocity over time. It indicates how rapidly the speed of eye and head movements increases or decreases, offering information about the smoothness and control of movements. Acceleration was calculated as the first derivative of velocity or the second derivative of angular position. Jerk represents the rate of change of acceleration over time. It captures the abruptness or smoothness of movements, with higher jerk values indicating more abrupt changes. Jerk is particularly relevant for detecting sudden adjustments in gaze direction. Jerk was computed as the first derivative of acceleration or the third derivative of angular position.</p> <p>Eye-specific features were also considered, such as the average eye-in-head angle and the proportion of eye contribution to overall gaze, where gaze is defined as the sum of eye and head angles (Doshi & Trivedi, [<reflink idref="bib23" id="ref100">23</reflink>]; Freedman, [<reflink idref="bib32" id="ref101">32</reflink>]; Sidenmark & Gellersen, [<reflink idref="bib80" id="ref102">80</reflink>]; Sidenmark et al., [<reflink idref="bib81" id="ref103">81</reflink>]; Stahl, [<reflink idref="bib83" id="ref104">83</reflink>], [<reflink idref="bib85" id="ref105">85</reflink>]; Zangemeister & Stark, [<reflink idref="bib98" id="ref106">98</reflink>]). The proportion of eye contribution indicates the relative reliance on eye versus head movements during gaze shifts.</p> <p>To glean insights into differences in information processing, separate metrics for target and distractor processing were also calculated, such as fixation frequency, duration on targets and distractors, and target-distractor fixation ratio. These metrics were exclusively available for eye measurements as head movements lack the same spatial granularity.</p> <p>Where applicable, metrics were normalized by trial duration to account for differences in trial lengths. Each metric was calculated separately for each trial. See Table 1 for the complete list and descriptions of all computed metrics. Data files and processing scripts will be made available upon reasonable request.</p> <hd id="AN0187497571-13">Machine learning classification</hd> <p>The Boruta algorithm, a multivariant wrapper feature selection method (Kursa et al., [<reflink idref="bib57" id="ref107">57</reflink>]), was used to assess feature importance prior to building classification models. Boruta identifies the most relevant features in a model-agnostic manner by comparing the importance of actual features to shuffled versions of the same data, referred to as"shadow features."A random forest model was run for 50 iterations, and the importance of each original feature was compared to the maximum importance of the shadow features in each iteration. At the end of the iterations, a binomial distribution (α = 0.05) was used to determine whether each feature was significantly important. By applying Boruta prior to running machine learning pipelines, we eliminated redundant and irrelevant features. Using only the important and independent features to train models reduces noise, helps avoid overfitting issues, and improves the human interpretability of model features (Dwivedi et al., [<reflink idref="bib27" id="ref108">27</reflink>]).</p> <p>The subset of important features from Boruta were incorporated into a machine learning pipeline in Python using sklearn (Pedregosa et al., [<reflink idref="bib69" id="ref109">69</reflink>]) consisting of Standard Scaling and a support vector classifier (SVC) with a radial basis function kernel. Model validation was performed via a tenfold cross-validation method, which divides the data into 10 equal portions with roughly equivalent class distribution. In each fold, 9 of the portions are used to train the pipeline, and the last is used to test the model using classification accuracy as the primary evaluation metric.</p> <p>To compare model performance, nonparametric Wilcoxon tests were conducted using the SciPy library (Virtanen et al., [<reflink idref="bib91" id="ref110">91</reflink>]). The Bonferroni–Holm modifier was applied to avoid false positives, and effect sizes were calculated using Hedges' <emph>g.</emph></p> <p>While Boruta provides a binary indication of whether a feature is significant or not, it does not provide the importance ranking of these features on model predictions. Feature importance within the models was performed using Shapley additive explanations (SHAP) Python library (Scott & Su-In, [<reflink idref="bib79" id="ref111">79</reflink>]). This provided a ranking of the features selected through Boruta. As in prior research, we selected a 20% representative subset of samples to train SHAP (Christoph, [<reflink idref="bib16" id="ref112">16</reflink>]).</p> <p>To investigate the contribution of both head and eye movements to model performance, as well as to tease apart how task parameter changes might affect movement systems differently, the following models were run. First, a 4-way classification omnibus model was run using all eye and head features designated as significantly important (Eye and Head). This model labeled each trial based on predicted target discriminability (Easy vs Hard) and time pressure (Low vs High), such that any given trial could be classified as Easy-Low, Easy-High, Hard-Low, or Hard-High. Next, a 4-way classification (with the same labels listed above) model was fit using only features derived from eye-tracking data (Eye Only) and another model was fit using only features derived from head-tracking data (Head Only). Finally, to assess whether eye and head features differ in their ability to infer changes in search parameters two separate 2-way classification models were run for each feature subset, focusing separately on target discriminability and time pressure.</p> <hd id="AN0187497571-14">Results</hd> <p></p> <hd id="AN0187497571-15">Eye and head movement summary</hd> <p>To demonstrate data quality, Fig. 2 provides a quantitative summary of eye and head movement characteristics. In the left column, the distribution of saccade and head movement amplitudes reveals a bimodal pattern for saccades, with average saccade amplitudes exceeding those of head movements. The second column illustrates main sequence plots for both eye and head movements, showing a clear velocity–amplitude relationship typical of these effectors (Bahill et al., [<reflink idref="bib4" id="ref113">4</reflink>]; Zangemeister et al., [<reflink idref="bib97" id="ref114">97</reflink>]). The fixation duration distributions display similar patterns for eye and head, though head fixations were generally longer on average than eye fixations. Angular distributions in the top right and bottom right plots show that eye position, relative to the head, remains within approximately 15 degrees of head center, while head orientation is predominantly distributed along the horizontal plane.</p> <p>Graph: Fig. 2 Eye (top row) and head movement (bottom row) summary from 20,000 randomly selected trial events across all participants. Left column—Saccade (top) and head movement (bottom) amplitude distribution. Second column from left—Main sequence plots for saccades (top) and head movements (bottom), illustrating the relationship between movement amplitude and peak velocity. Third column from left—distribution of eye and head fixation durations. Right column—heat map showing the angular distribution of eye in head (top) and head in the world (bottom). deg = degrees; ms = milliseconds</p> <hd id="AN0187497571-16">Behavioral performance</hd> <p>Dependent measures were analyzed using a random effects repeated measures analysis of variance (ANOVA) with the 'ezANOVA'R package. The within-subjects factors were target discriminability (Easy, Hard) and time pressure (Low, High) with participant as a random effect (Fig. 3[<reflink idref="bib2" id="ref115">2</reflink>]). Trial completion analysis revealed a significant main effect for target discriminability F(<reflink idref="bib1" id="ref116">1</reflink>,<reflink idref="bib29" id="ref117">29</reflink>) = 155.8, <emph>p</emph> <.001, η<sups>2</sups><subs>g</subs> =.11, with easy discrimination trials completed faster than hard. The main effect of time pressure was also significant F(<reflink idref="bib1" id="ref118">1</reflink>,<reflink idref="bib29" id="ref119">29</reflink>) = 65.2, <emph>p</emph> <.001, η<sups>2</sups><subs>g</subs> =.42 with high time pressure trials completed faster than low time pressure. The discrimination by time pressure interaction was significant F(<reflink idref="bib1" id="ref120">1</reflink>,<reflink idref="bib29" id="ref121">29</reflink>) = 96.5, <emph>p</emph> <.001, η<sups>2</sups><subs>g</subs> =.04 indicating an increased impact of time pressure in the difficult with respect to easy discrimination trials. Similar findings were found for inter-target time which showed main effects for target discriminability F(<reflink idref="bib1" id="ref122">1</reflink>,<reflink idref="bib29" id="ref123">29</reflink>) = 227.22, <emph>p</emph> <.001, η<sups>2</sups><subs>g</subs> =.33 and time pressure F(<reflink idref="bib1" id="ref124">1</reflink>,<reflink idref="bib29" id="ref125">29</reflink>) = 97.88, <emph>p</emph> <.001, η<sups>2</sups><subs>g</subs> =.39 indicating the average time between target engagement was faster for easy discrimination and high time pressure targets relative to hard discrimination and low time pressure targets. The interaction was significant F(<reflink idref="bib1" id="ref126">1</reflink>,<reflink idref="bib29" id="ref127">29</reflink>) = 40.9, <emph>p</emph> <.001, η<sups>2</sups><subs>g</subs> =.06 suggesting a larger effect of time pressure in the hard discrimination condition. The d-prime analysis revealed a main effect for target discriminability F(<reflink idref="bib1" id="ref128">1</reflink>,<reflink idref="bib29" id="ref129">29</reflink>) = 43.5, <emph>p</emph> <.001, η<sups>2</sups><subs>g</subs> =.17 showing higher d-prime in the easy compared to hard discrimination condition. However, the main effect for time pressure and the interaction between discrimination difficulty and time pressure were not significant (<emph>p</emph> >.05). Together these results show that target discriminability and time pressure significantly influenced behavioral performance, with faster completion and inter-target times and higher d-prime under easier conditions.</p> <p>Graph: Fig. 3 Behavioral performance. Target discriminability and time pressure had a significant main an interactive effect on trial completion time (left) and inter-target engagement time (middle). Target discriminability significantly impacted d-prime (right)</p> <p>Behavioral data indicated that 38% of high time pressure trials reached the 18-s time limit, suggesting the trial ended automatically rather than by participant decision. However, this does not necessarily imply insufficient time to complete the task as many participants still achieved 100% hit rate, even at the maximum duration, indicating that they may have completed a first pass and continued scanning or verifying targets. Similarly, there was a wide range of hit rates in trials that ended earlier, suggesting that participants often had time remaining but chose to terminate. Notably, some of the lowest hit percentages occurred in trials participants ended on their own, further implying that time constraints were not the sole limiting factor. In contrast, only 3% of low time pressure trials ended due to the time limit, while 33% extended beyond 18 s, reinforcing the impact of time pressure on search behavior.</p> <hd id="AN0187497571-17">Classification</hd> <p>The Boruta algorithm identified the most significant features from each feature set, and these were used in the 4-way classification (Multiclass model; Fig. 4). Model performance when using 49 of the 76 combined eye and head features was significantly better than random chance <emph>(M</emph> = 0.66, <emph>SD</emph> = 0.03, <emph>p</emph> = <emph>.002 g</emph> = <emph>14.15).</emph> Similarly, when using 39 out of 45 Eye Only features, the model achieved significant classification above chance (<emph>M</emph> = 0.64, <emph>SD</emph> = 0.03, <emph>p</emph> = <emph>.002, g</emph> = <emph>7.70</emph>) and 17 of 31 Head Only features, also performed significantly above chance in the 4-way classification model (<emph>M</emph> = 0.39, <emph>SD</emph> = 0.02, <emph>p</emph> = <emph>.002</emph>, <emph>g</emph> = <emph>6.77</emph>). Notably, incorporating head features with eye features significantly improved classification accuracy relative to Eye Only features (<emph>p</emph> = <emph>0.03, g</emph> = <emph>0.93)</emph> highlighting the modest but complementary value of combining head movement features with eye movement data. Eye Only features significantly outperformed Head Only features (<emph>p</emph> = 0.019, <emph>g</emph> = 5.61).</p> <p>Graph: Fig. 4 Multiclass model accuracy. Significant detection of task-induced effects of target discriminability and time pressure was achieved across all feature sets, with combined eye and head features having the highest model accuracy</p> <p>The top 10 features, ranked in order of importance, for each model are included in Table 2.</p> <p>Table 2 Top 10 features and mean accuracy for each model</p> <p> <ephtml> <table frame="hsides" rules="groups"><thead><tr><th align="left" colspan="4"><p>4-Way Classification</p></th><th align="left" colspan="2"><p>Target Discriminability</p></th><th align="left" colspan="2"><p>Time Pressure</p></th></tr><tr><th align="left"><p>Feature Rank</p></th><th align="left"><p>Eye &Head M = 0.66</p></th><th align="left"><p>Eye M = 0.63</p></th><th align="left"><p>Head M = 0.39</p></th><th align="left"><p>Eye M = 0.85</p></th><th align="left"><p>Head M = 0.65</p></th><th align="left"><p>Eye M = 0.77</p></th><th align="left"><p>Head M = 0.64</p></th></tr></thead><tbody><tr><td align="left"><p><bold>1</bold></p></td><td align="left"><p>Target Fixations per Target</p></td><td align="left"><p>Target Fixations per Target</p></td><td align="left"><p>Global Head Movement per Sec</p></td><td align="left"><p>Distractor Fixation Frequency</p></td><td align="left"><p>Global Head Movement per Sec</p></td><td align="left"><p>Target Fixations per Target</p></td><td align="left"><p>Avg Head Acceleration X</p></td></tr><tr><td align="left"><p><bold>2</bold></p></td><td align="left"><p>Distractor Fixations per Distractor</p></td><td align="left"><p>Distractor Fixations per Distractor</p></td><td align="left"><p>Std Head Velocity X</p></td><td align="left"><p>Distractor Fixation Frequency</p></td><td align="left"><p>Avg Head Velocity Z</p></td><td align="left"><p>Target Fixation Frequency</p></td><td align="left"><p>Std Head Velocity Z</p></td></tr><tr><td align="left"><p><bold>3</bold></p></td><td align="left"><p>Saccade Frequency</p></td><td align="left"><p>Saccade Frequency</p></td><td align="left"><p>Avg Head Velocity X</p></td><td align="left"><p>Distractor Fixations per Distractor</p></td><td align="left"><p>Avg Head Fixation Duration</p></td><td align="left"><p>Avg Target Fixation Duration</p></td><td align="left"><p>Std Head Velocity X</p></td></tr><tr><td align="left"><p><bold>4</bold></p></td><td align="left"><p>Distractor Fixation Duration Proportion</p></td><td align="left"><p>Distractor Fixation Duration Proportion</p></td><td align="left"><p>Avg Head Acceleration X</p></td><td align="left"><p>Target Fixations per Target</p></td><td align="left"><p>Avg Head Jerk</p></td><td align="left"><p>Saccade Size per Sec</p></td><td align="left"><p>Head Amplitude per Sec</p></td></tr><tr><td align="left"><p><bold>5</bold></p></td><td align="left"><p>Distractor Fixation Frequency</p></td><td align="left"><p>Target Fixation Frequency</p></td><td align="left"><p>Head Amplitude per Sec</p></td><td align="left"><p>Target Fixation Duration Proportion</p></td><td align="left"><p>Head Movement Frequency</p></td><td align="left"><p>Saccade Frequency</p></td><td align="left"><p>Std Head Jerk X</p></td></tr><tr><td align="left"><p><bold>6</bold></p></td><td align="left"><p>Target Fixation Frequency</p></td><td align="left"><p>Distractor Fixation Frequency</p></td><td align="left"><p>Head Fixation Frequency</p></td><td align="left"><p>Fixation Duration Proportion</p></td><td align="left"><p>Head Fixation Duration Proportion</p></td><td align="left"><p>Avg Eye Acceleration X</p></td><td align="left"><p>Avg head Fixation Duration</p></td></tr><tr><td align="left"><p><bold>7</bold></p></td><td align="left"><p>Fixation Duration Proportion</p></td><td align="left"><p>Target Fixation Duration Proportion</p></td><td align="left"><p>Avg Head Fixation Duration</p></td><td align="left"><p>Avg Eye Velocity X</p></td><td align="left"><p>Std Head Velocity X</p></td><td align="left"><p>Target Fixation Duration Proportion</p></td><td align="left"><p>Head Fixation Frequency</p></td></tr><tr><td align="left"><p><bold>8</bold></p></td><td align="left"><p>Target Fixation Duration</p><p>Proportion</p></td><td align="left"><p>Fixation Duration Proportion</p></td><td align="left"><p>Avg Head Jerk</p></td><td align="left"><p>Avg Eye Jerk Y</p></td><td align="left"><p>Avg Head Velocity</p></td><td align="left"><p>Avg Eye Jerk X</p></td><td align="left"><p>Avg Head Jerk X</p></td></tr><tr><td align="left"><p><bold>9</bold></p></td><td align="left"><p>Avg Eye in Head Angle</p></td><td align="left"><p>Avg Eye in Head Angle</p></td><td align="left"><p>Avg Head Velocity X</p></td><td align="left"><p>Avg Eye Acceleration Y</p></td><td align="left"><p>Avg Head Velocity X</p></td><td align="left"><p>Distractor Fixations per Distractor</p></td><td align="left"><p>Avg Head Jerk</p></td></tr><tr><td align="left"><p><bold>10</bold></p></td><td align="left"><p>Saccade Size per Sec</p></td><td align="left"><p>Avg Eye Acceleration X</p></td><td align="left"><p>Avg Head Jerk X</p></td><td align="left"><p>Avg Saccade Size</p></td><td align="left"><p>Avg Head Movement Size</p></td><td align="left"><p>Avg Eye Velocity</p></td><td align="left"><p>Std Head Velocity Y</p></td></tr></tbody></table> </ephtml> </p> <p>Avg = Average; Std = Standard Deviation; Sec = Second; See Table 1 for feature definitions</p> <p>Target discriminability and time pressure were modeled separately using either Eye Only or Head Only features, and all models performed significantly above chance (Fig. 5). The target discriminability model trained on Eye Only features achieved a mean accuracy of 0.85 (SD = 0.03, <emph>p</emph> = 0.002, g = 7.41), utilizing 34 of the 45 available eye features. In comparison, the time pressure model using Eye Only features reached a mean accuracy of 0.77 (SD = 0.02, <emph>p</emph> = 0.002, g = 8.57), with 35 of 45 features selected. Similarly, Head Only feature models also exceeded chance performance: The target discriminability model attained a mean accuracy of 0.64 (SD = 0.04, <emph>p</emph> = 0.001, g = 1.93) using 12 of the 31 available head features, while the time pressure model achieved 0.65 (SD = 0.04, <emph>p</emph> = 0.002, g = 2.48) with 20 of the 31 features. Eye movement features were better at classifying the target discriminability of a trial (<emph>M</emph> = 0.85, <emph>SD</emph> = 0.03) compared to the time pressure (<emph>M</emph> = 0.77, <emph>SD</emph> = 0.02), <emph>p</emph> = 0.002, <emph>g</emph> = 2.46. When using only head movement features, classification accuracies between target discriminability (M = 0.64, SD = 0.04) and time pressure (M = 0.65, SD = 0.04) were not statistically significant (<emph>p</emph> > 0.5).</p> <p>Graph: Fig. 5 Classification accuracy for the target discriminability and time pressure models from Eye Only and Head Only features</p> <hd id="AN0187497571-18">Discussion</hd> <p>The primary objective of this study was to investigate how task parameters—target discriminability and time pressure—affects foraging behavior and whether eye and head movement metrics can be used to reliably infer these task-induced changes. By employing a naturalistic VR foraging task, we aimed to examine the extent to which these movement systems contribute independently and collectively to adaptive search strategies under varying task conditions.</p> <p>Behavioral results indicated that both target discriminability and time pressure significantly influenced search performance. Consistent with prior research, trials with easily discriminable targets from distractors (Easy condition) led to faster trial completion times, shorter inter-target times, and higher d-prime scores, indicating more efficient and selective search behavior (Kristjánsson et al., [<reflink idref="bib56" id="ref130">56</reflink>]; Tagu & Kristjánsson, [<reflink idref="bib86" id="ref131">86</reflink>]). In contrast, the hard condition was associated with slower response times and reduced d-prime, highlighting the greater perceptual demands required when targets were more visually similar to distractors. Time pressure also had a significant impact, with high-pressure trials showing faster completion and inter-target times particularly under harder discriminability. This interaction between difficulty and time pressure reflects the competing demands of speed and perceptual decision-making during visual search. Participants may shift toward a strategy that prioritizes speed over accuracy, suggesting a potential tradeoff between rapid decision-making and visual discrimination during search.</p> <p>While it is well known that changes in task parameters, as manipulated here, produce changes in behavior, these metrics are typically reported as descriptive statistics in post hoc analyses, after averaging across multiple trials. To provide adaptive support, any intelligent system will need to use moment-to-moment human behavior to infer the given task and environmental parameters. As an intermediate step toward such a system, the current work used a machine learning approach to classify the target discriminability and time pressure level of individual trials. Our classification approach demonstrated that eye and head movement features can be used to accurately detect the task conditions that shape search behavior. As expected, eye features alone achieved high classification accuracy. However, when combined with head movement features, classification accuracy improved significantly. By identifying which metrics carry predictive value across trials, this work establishes a foundational feature space that could be translated into finer temporal windows in future work.</p> <hd id="AN0187497571-19">The feasibility of using head tracking to infer task parameters</hd> <p>Modern eye-tracking devices, such as those in HMDs or mobile glasses, often include head-tracking capabilities, offering an additional resource for analyzing search behavior. By comparing Eye Only and Head Only models our findings suggest that head movement data alone can infer task-induced changes in search behavior, though with reduced accuracy compared to eye-tracking data. However, when combined with eye-tracking metrics, head movement data provided unique information, enhancing the overall predictive power of the model. This highlights the potential of head movement data as both a standalone substitute when eye tracking is unavailable and a complementary feature for enhancing the analyses of cognitive processes and search behavior in dynamic environments (Clay et al., [<reflink idref="bib17" id="ref132">17</reflink>]). Despite advances in camera-based mobile eye tracking, capturing relative or absolute gaze angle across varied real-world environments remains challenging due to factors like variable illumination or device slippage (Ghiani et al., [<reflink idref="bib36" id="ref133">36</reflink>]). In contrast, head-tracking sensors, such as accelerometers or IMUs, offer a more robust while less precise means to estimate gaze direction, making them an attractive alternative for capturing data in challenging environments. Our findings underscore the potential of head-tracking data to quantify task-induced changes in behavior, even when eye-tracking data are unavailable. Furthermore, the simplicity and durability of head-tracking sensors make them a practical solution for applications requiring lightweight, unobtrusive, and cost-effective systems.</p> <p>The results from our multiclass classification showed head movement metrics provided unique information not fully captured by eye metrics alone. The most informative features in the 4-way classification model from head-only data were the total angular distance of head movement during the trial, normalized for trial duration, followed by kinematic features such as velocity and acceleration. Notably, head movement distance and stability (i.e., head fixation duration) were more sensitive to task-induced changes for target discriminability, while kinematic features were ranked higher for time pressure. These findings align with prior research demonstrating the utility of head kinematics for inferring search behavior (Hu et al., [<reflink idref="bib44" id="ref134">44</reflink>]; Salehi et al., [<reflink idref="bib77" id="ref135">77</reflink>]). For example, Hu et al. ([<reflink idref="bib44" id="ref136">44</reflink>]) showed that head movement velocity and acceleration could reliably distinguish between different types of visual search, such as free viewing, target search, and tracking. Similarly, research by (Aivar et al., [<reflink idref="bib2" id="ref137">2</reflink>]), demonstrated that body movement velocity increases with repeated search in the same context, reflecting an optimization of movement strategies for faster and more efficient search execution. This, combined with the fact that there was no significant difference in performance between the 2-way classification models of target discriminability and time pressure using head data alone, suggests that head tracking may provide some, though limited insight into the parameters of an individual's search.</p> <hd id="AN0187497571-20">The role of eye movements in target discriminability and time pressure</hd> <p>In the target discriminability model, fixation behaviors emerged as the most informative features, including Distractor Fixation Frequency, Target Fixation Duration Proportion, and Distractor Fixation Duration Proportion. This pattern suggests that eye movement metrics are sensitive to the perceptual details between targets and distractors. Specifically, under more difficult conditions, participants likely increased feature-level inspection, fixating on distractors more frequently and for longer durations as they made their decision. This sensitivity indicates that eye movement behavior can serve as a dynamic, implicit signal of perceptual difficulty.</p> <p>Conversely, the time pressure model prioritized fixation frequency and saccadic behaviors, with top features including Target Fixations per Target, Target Fixation Frequency, Saccade Size per Second, and Saccade Frequency. This suggests that under high time pressure, participants may have adopted a faster, more efficient search strategy with frequent saccades and brief fixations, likely to maximize the rate of information acquisition. Therefore, these eye movement features provide the potential for tracking cognitive state changes associated with search urgency.</p> <p>The current results are consistent with previous findings that show eye movement metrics such as saccade amplitude, fixation duration, and gaze velocity are highly sensitive to changes in search demands and target discriminability. For example, increased saccade amplitude and fixation frequency have been linked to harder visual search tasks, where targets are more difficult to discriminate from distractors (Pomplun et al., [<reflink idref="bib74" id="ref138">74</reflink>]; Tagu & Kristjánsson, [<reflink idref="bib86" id="ref139">86</reflink>]). Similarly, fixation duration tends to increase as target discriminability becomes more challenging (Becker, [<reflink idref="bib5" id="ref140">5</reflink>]; Pomplun et al., [<reflink idref="bib74" id="ref141">74</reflink>]), while tasks with high perceptual and attentional demands decrease peak saccadic velocity (BacHurina & Arsalidou, [<reflink idref="bib3" id="ref142">3</reflink>]). Together our results are in agreement with current theories of visual search, such as the functional viewing field (FVF), which suggest that task demands, particularly search difficulty, modulate the amount of information processed during a fixation. These demands subsequently influence the spatial and temporal characteristics of saccades and fixations as participants adjust their strategies to effectively process visual information (Hulleman & Olivers, [<reflink idref="bib47" id="ref143">47</reflink>]; Young & Hulleman, [<reflink idref="bib96" id="ref144">96</reflink>]).</p> <hd id="AN0187497571-21">Practical implications</hd> <p>The ability to infer task conditions from combined eye and head movement metrics has meaningful implications for the development of adaptive systems that monitor user state. IMUs, used to track head movements, are affordable, lightweight, and robust sensor. The current work indicates that such a sensor, easily deployed to the field, could provide insight into target discrimination difficulty experienced by a soldier triggering adjustments in system support. For example, this information could be shared with other squad members or a commander to inform the reallocation of resources such as deploying autonomous agents to aid in search and target interrogation. While head movement data provide valuable insight into user state, the results indicate that this signal alone is not as informative as eye movement data. Our results show that eye, compared to head data, provide higher classification accuracy and should be prioritized when available. In contexts with degraded or unavailable eye tracking, due to occlusion, motion, lighting, etc., head-tracking data offer a viable alternative.</p> <p>Importantly, the utility of head movement data as a substitute will also depend on the search context. In constrained environments, such as narrow hallways, head movements are limited, and eye movements are more likely to carry the critical information needed for search. In contrast, in broader or more open environments where potential search targets fall beyond the customary oculomotor range (COMR), head movements become more informative and can better reflect search dynamics (Stahl, [<reflink idref="bib83" id="ref145">83</reflink>], [<reflink idref="bib85" id="ref146">85</reflink>]). Thus, inferring contextual information from head movement data is most appropriate when search demands extend beyond the natural range of eye movements alone.</p> <hd id="AN0187497571-22">Limitations and future work</hd> <p>A notable aspect of our approach was the inclusion of target- and distractor-specific metrics, such as fixation frequency and duration, which emerged as some of the most discriminative features in our models. For example, Distractor Fixation Frequency was a top-ranked feature for target discriminability, reflecting participants' increased reliance on distractor fixations when distinguishing targets in the hard condition. While the results remain meaningful and demonstrate that these features offer valuable insights into task-induced search behavior, environmental labels may not be readily available in real-world settings.[<reflink idref="bib3" id="ref147">3</reflink>] However, our approach is a critical step toward developing a robust inferential model of task-induced changes in foraging behavior. Quantifying and understanding the utility of all eye and head movement metrics, regardless of their accessibility in a real-world context, provides a broad foundation for multimodal application spaces. For example, augmented reality (AR) applications would potentially have some context as to the task-relevance of overlaid information (e.g., a UI display) as well as some understanding of the environment (e.g., geospatial location). Head and eye movement features could be combined with other sensors, such as cameras, to provide a rich understanding of how environmental factors change behavior of individuals and groups. Moreover, having such a model to characterize "hard" or "easy" visual search may provide AI systems valuable information about the effectiveness of the current AI support. For instance, if eye and head movement signatures indicate an operator is having difficulty in searching the environment, this may indicate too much visual clutter in the AR display.</p> <p>Several factors limit the direct applicability of our findings to more naturalistic scenarios. The current study was conducted in a stationary context, with the task confined to 180 degrees of space. While this setup allowed for controlled measurement of eye and head movements during the task, it may not fully capture the spatiotemporal demands encountered in all foraging search environments. Future work should examine the generalizability of these findings to more ambulatory tasks with natural navigation and dynamic, occluded targets to better reflect real-world search complexity. The reduced field of view imposed by the HMD likely influenced search behavior, leading to more exaggerated head movements as participants compensated for limited peripheral vision. Similarly, field testing scenarios should consider the effect of helmets or other gear on head movements, as such equipment could substantially alter behavioral patterns compared to those observed in the current laboratory environment.</p> <p>Our study used a generalized classification model which applied a uniform approach across all participants. While this is useful for identifying overarching patterns in search behavior, it may overlook individual differences in movement strategy, such as varying reliance on head versus eye movement (Fuller, [<reflink idref="bib34" id="ref148">34</reflink>]; Stahl, [<reflink idref="bib84" id="ref149">84</reflink>]). An individualized model that accounts for participant-specific movement patterns or preferences could improve predictive accuracy by incorporating eye–head coordination variability and other effector interactions. Refining classification models with personalized metrics would also provide a more detailed perspective on the cognitive mechanisms driving search behavior.</p> <p>In the present study, we analyzed behavior and eye/head movements across the entire duration of the search process within each trial, without segmenting distinct phases such as the time to first target acquisition, inter-target intervals during the middle of the trial, or the final search segment preceding the last response. This approach was intended to reflect the continuous nature of foraging behavior and to validate that our experimental manipulations influenced task performance. However, we acknowledge that the final segment of a search may involve qualitatively different decisional processes, potentially inflating inter-target times. Future work could examine subcomponents of the search process, such as time to first target, mid-trial search dynamics, or search termination behavior, to better isolate the distinct cognitive processes of each phase.</p> <hd id="AN0187497571-23">Conclusion</hd> <p>This study highlights the benefit of integrating eye and head movement metrics to infer task-induced changes in search behavior. The findings demonstrate that head movements, while often underutilized, offer unique and complementary information to eye metrics. These insights have practical implications for designing adaptive systems for near-real-time monitoring of operator states and informing adaptive systems designed to dynamically adjust their support.</p> <hd id="AN0187497571-24">Author contributions</hd> <p>AR, CCF, and AM conceived the study and methodology. AR, AM, CCF, and LD performed programming, data curation, and formal analysis. AR, CCF, LD, and JT wrote original draft. All authors read, edited, and approved the final manuscript. We extend our gratitude to Alex Stauff and Christian Barentine for their expertise and dedication in programming the experimental paradigm and C1C Dan Park, C1C Phil Luba, and C1C Khava Tsarni for their assistance with data collection.</p> <hd id="AN0187497571-25">Funding</hd> <p>This research was supported by U.S. Army DEVCOM U.S. ARL project number: W911NF-23–2-0205.</p> <hd id="AN0187497571-26">Availability of data and materials</hd> <p>The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.</p> <hd id="AN0187497571-27">Declarations</hd> <p></p> <hd id="AN0187497571-28">Ethics approval and consent to participate</hd> <p>Participants participated for extra course credit and provided informed consent. This work was approved by U.S. Army DEVCOM U.S ARL under protocol #ARL-21–132.</p> <hd id="AN0187497571-29">Consent for publication</hd> <p>All authors have approved the manuscript for review.</p> <hd id="AN0187497571-30">Competing interests</hd> <p>The authors declare that they have no competing interests.</p> <hd id="AN0187497571-31">Supplementary Information</hd> <p>Graph: Additional file 1</p> <p>Graph: Additional file 2</p> <p>Graph: Additional file 3</p> <p>Graph: Additional file 4</p> <hd id="AN0187497571-32">Publisher's Note</hd> <p>Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.</p> <ref id="AN0187497571-33"> <title> References </title> <blist> <bibl id="bib1" idref="ref32" type="bt">1</bibl> <bibtext> Agtzidis, I, Startsev, M, & Dorr, M. (2019). 360-degree Video Gaze Behaviour: A Ground-Truth Data Set and a Classification Algorithm for Eye Movements. Proceedings of the 27th ACM International Conference on Multimedia, 1007–1015. https://doi.org/10.1145/3343031.3350947</bibtext> </blist> <blist> <bibl id="bib2" idref="ref115" type="bt">2</bibl> <bibtext> Aivar MP, Li C-L, Tong MH, Kit DM, Hayhoe MM. Knowing where to go: Spatial memory guides eye and body movements in a naturalistic visual search task. Journal of Vision. 2024; 24; 9: 1. 10.1167/jov.24.9.1. 39226069. 11373708</bibtext> </blist> <blist> <bibl id="bib3" idref="ref142" type="bt">3</bibl> <bibtext> Bachurina V, Arsalidou M. Multiple levels of mental attentional demand modulate peak saccade velocity and blink rate. Heliyon. 2022. 10.1016/j.heliyon.2022.e08826. 35128110. 8800024</bibtext> </blist> <blist> <bibl id="bib4" idref="ref113" type="bt">4</bibl> <bibtext> Bahill AT, Clark MR, Stark L. The main sequence, a tool for studying human eye movements. Mathematical Biosciences. 1975; 24; 3: 191-204. 10.1016/0025-5564(75)90075-9</bibtext> </blist> <blist> <bibl id="bib5" idref="ref140" type="bt">5</bibl> <bibtext> Becker SI. Determinants of dwell time in visual search: Similarity or perceptual difficulty?. PLoS ONE. 2011; 6; 3. 10.1371/journal.pone.0017740. 21408139. 3050928e17740</bibtext> </blist> <blist> <bibl id="bib6" idref="ref33" type="bt">6</bibl> <bibtext> Bischof WF, Anderson NC, Kingstone A. A tutorial: Analyzing eye and head movements in virtual reality. Behavior Research Methods. 2024; 56; 8: 8396-8421. 10.3758/s13428-024-02482-5. 39117987</bibtext> </blist> <blist> <bibl id="bib7" idref="ref5" type="bt">7</bibl> <bibtext> Boot WR, Kramer AF, Becic E, Wiegmann DA, Kubose T. Detecting transient changes in dynamic displays: The more you look, the less you see. Human Factors. 2006; 48; 4: 759-773. 10.1518/001872006779166424. 17240723</bibtext> </blist> <blist> <bibl id="bib8" idref="ref1" type="bt">8</bibl> <bibtext> Brunyé TT, Brou R, Doty TJ, Gregory FD, Hussey EK, Lieberman HR, Loverro KL, Mezzacappa ES, Neumeier WH, Patton DJ, Soares JW, Thomas TP, Yu AB. A review of US army research contributing to cognitive enhancement in military contexts. Journal of Cognitive Enhancement. 2020; 4; 4: 453-468. 10.1007/s41465-020-00167-3</bibtext> </blist> <blist> <bibl id="bib9" idref="ref26" type="bt">9</bibl> <bibtext> Brunyé TT, Gardony AL. Eye tracking measures of uncertainty during perceptual decision making. International Journal of Psychophysiology. 2017; 120: 60-68. 10.1016/j.ijpsycho.2017.07.008. 28732659</bibtext> </blist> <blist> <bibtext> Brunyé TT, Giles GE. Methods for eliciting and measuring behavioral and physiological consequences of stress and uncertainty in virtual reality. Frontiers in Virtual Reality. 2023. 10.3389/frvir.2023.951435</bibtext> </blist> <blist> <bibtext> Cain MS, Vul E, Clark K, Mitroff SR. A Bayesian optimal foraging model of human visual search. Psychological Science. 2012; 23; 9: 1047-1054. 10.1177/0956797612440460. 22868494</bibtext> </blist> <blist> <bibtext> Callahan-Flintoft C, Barentine C, Touryan J, Ries AJ. A case for studying naturalistic eye and head movements in virtual environments. Frontiers in Psychology. 2021. 10.3389/fpsyg.2021.650693. 35035362. 8759101</bibtext> </blist> <blist> <bibtext> Callahan-Flintoft C, Jensen E, Naeem J, Nonte MW, Madison AM, Ries AJ. A comparison of head movement classification methods. Sensors. 2024; 24; 4: 1260. 10.3390/s24041260. 38400418. 10893452</bibtext> </blist> <blist> <bibtext> Chen LL, Walton MMG. Head movement evoked by electrical stimulation in the supplementary eye field of the Rhesus monkey. Journal of Neurophysiology. 2005; 94; 6: 4502-4519. 10.1152/jn.00510.2005. 16148273</bibtext> </blist> <blist> <bibtext> Christ SE, Abrams RA. Abrupt onsets cannot be ignored. Psychonomic Bulletin & Review. 2006; 13; 5: 875-880. 10.3758/BF03194012</bibtext> </blist> <blist> <bibtext> Christoph, M. (2020). Interpretable machine learning: A guide for making black box models explainable. Leanpub.</bibtext> </blist> <blist> <bibtext> Clay V, König P, Koenig S. Eye tracking in virtual reality. Journal of Eye Movement Research. 2019; 12; 1: 10-6910. 7903250</bibtext> </blist> <blist> <bibtext> Culmer PR, Levesley MC, Mon-Williams M, Williams JHG. A new tool for assessing human movement: The kinematic assessment tool. Journal of Neuroscience Methods. 2009; 184; 1: 184-192. 10.1016/j.jneumeth.2009.07.025. 19646475</bibtext> </blist> <blist> <bibtext> David, E. J, Gutiérrez, J, Coutrot, A, Da Silva, M. P, & Callet, P. L. (2018). A dataset of head and eye movements for 360° videos. Proceedings of the 9th ACM Multimedia Systems Conference, 432–437. https://doi.org/10.1145/3204949.3208139</bibtext> </blist> <blist> <bibtext> David E, Beitner J, Võ ML-H. Effects of transient loss of vision on head and eye movements during visual search in a virtual environment. Brain Sciences. 2020; 10; 11: 841. 10.3390/brainsci10110841. 33198116. 7696943</bibtext> </blist> <blist> <bibtext> Di Stasi LL, Antolí A, Cañas JJ. Main sequence: An index for detecting mental workload variation in complex tasks. Applied Ergonomics. 2011; 42; 6: 807-813. 10.1016/j.apergo.2011.01.003. 21316645</bibtext> </blist> <blist> <bibtext> Dimigen O, Sommer W, Hohlfeld A, Jacobs AM, Kliegl R. Coregistration of eye movements and EEG in natural reading: Analyses and review. Journal of Experimental Psychology: General. 2011; 140; 4: 552-572. 10.1037/a0023885. 21744985</bibtext> </blist> <blist> <bibtext> Doshi A, Trivedi MM. Head and eye gaze dynamics during visual attention shifts in complex environments. Journal of Vision. 2012; 12; 2: 9-9. 10.1167/12.2.9. 22323822</bibtext> </blist> <blist> <bibtext> Draschkow D, Kallmayer M, Nobre AC. When natural behavior engages working memory. Current Biology. 2021; 31; 4: 869-874.e5. 10.1016/j.cub.2020.11.013. 33278355. 7902904</bibtext> </blist> <blist> <bibtext> Drew T, Boettcher SEP, Wolfe JM. One visual search, many memory searches: An eye-tracking investigation of hybrid search. Journal of Vision. 2017; 17; 11: 5. 10.1167/17.11.5. 28892812. 5596794</bibtext> </blist> <blist> <bibtext> Duncan J, Humphreys GW. Visual search and stimulus similarity. Psychological Review. 1989; 96; 3: 433-458. 10.1037/0033-295X.96.3.433. 2756067</bibtext> </blist> <blist> <bibtext> Dwivedi R, Tiwari A, Bharill N, Ratnaparkhe M, Tiwari AK. A taxonomy of unsupervised feature selection methods including their pros, cons, and challenges. The Journal of Supercomputing. 2024; 80; 16: 24212-24240. 10.1007/s11227-024-06368-3</bibtext> </blist> <blist> <bibtext> Eckstein MP. Visual search: A retrospective. Journal of Vision. 2011. 10.1167/11.5.14. 22209816</bibtext> </blist> <blist> <bibtext> El Iskandarani, M, Atweh, J. A, & Riggs, S. L. (2024). Time Pressure Detection in a Visual Search Task Using Eye Tracking Metrics: A Virtual Reality Study. Proceedings of the Human Factors and Ergonomics Society Annual Meeting, 10711813241261288. https://doi.org/10.1177/10711813241261288</bibtext> </blist> <blist> <bibtext> Enders LR, Smith RJ, Gordon SM, Ries AJ, Touryan J. Gaze behavior during navigation and visual search of an open-world virtual environment. Frontiers in Psychology. 2021; 12: 681042. 10.3389/fpsyg.2021.681042. 34434140. 8380848</bibtext> </blist> <blist> <bibtext> Engbert R, Mergenthaler K. Microsaccades are triggered by low retinal image slip. Proceedings of the National Academy of Sciences. 2006; 103; 18: 7192-7197. 10.1073/pnas.0509557103</bibtext> </blist> <blist> <bibtext> Freedman EG. Coordination of the eyes and head during visual orienting. Experimental Brain Research. 2008; 190; 4: 369-387. 10.1007/s00221-008-1504-8. 18704387. 2605952</bibtext> </blist> <blist> <bibtext> Freedman EG, Sparks DL. Eye-head coordination during head-unrestrained gaze shifts in rhesus monkeys. Journal of Neurophysiology. 1997; 77; 5: 2328-2348. 10.1152/jn.1997.77.5.2328. 9163361</bibtext> </blist> <blist> <bibtext> Fuller JH. Head movement propensity. Experimental Brain Research. 1992; 92; 1: 152-164. 10.1007/BF00230391. 1486950</bibtext> </blist> <blist> <bibtext> Ganesan A, Alakhras M, Brennan PC, Mello-Thoms C. A review of factors influencing radiologists' visual search behaviour. Journal of Medical Imaging and Radiation Oncology. 2018; 62; 6: 747-757. 10.1111/1754-9485.12798. 30198628</bibtext> </blist> <blist> <bibtext> Ghiani A, Mann D, Brenner E. Methods matter: Exploring how expectations influence common actions. iScience. 2024. 10.1016/j.isci.2024.109076. 38361615. 10867666</bibtext> </blist> <blist> <bibtext> Gilchrist ID, Harvey M. Evidence for a systematic component within scan paths in visual search. Visual Cognition. 2006. 10.1080/13506280500193719</bibtext> </blist> <blist> <bibtext> Gordon SM, Dalangin B, Touryan J. Saccade size predicts onset time of object processing during visual search of an open world virtual environment. NeuroImage. 2024; 298. 10.1016/j.neuroimage.2024.120781. 39127183120781</bibtext> </blist> <blist> <bibtext> Groner, R, Walder, F, & Groner, M. (1984). Looking at Faces: Local and Global Aspects of Scanpaths. In A. G. Gale & F. Johnson (Eds.), Advances in Psychology (Vol. 22, pp. 523–533). North-Holland. https://doi.org/10.1016/S0166-4115(08)61874-9</bibtext> </blist> <blist> <bibtext> Henderson JM, Shinkareva SV, Wang J, Luke SG, Olejarczyk J. Predicting cognitive state from eye movements. PLoS ONE. 2013; 8; 5. 10.1371/journal.pone.0064937. 23734228. 3666973e64937</bibtext> </blist> <blist> <bibtext> Holmqvist, K, Nyström, M, Andersson, R, Dewhurst, R, Jarodzka, H, & Van de Weijer, J. (2011). Eye tracking: A comprehensive guide to methods and measures. Oxford University Press. <ulink href="http://dspace.ou.nl/handle/1820/3441">http://dspace.ou.nl/handle/1820/3441</ulink></bibtext> </blist> <blist> <bibtext> Hooge, I. Th. C, Vlaskamp, B. N. S, & Over, E. A. B. (2007). Chapter 27 - Saccadic search: On the duration of a fixation. In R. P. G. Van Gompel, M. H. Fischer, W. S. Murray, & R. L. Hill (Eds.), Eye Movements (pp. 581–595). Elsevier. https://doi.org/10.1016/B978-008044980-7/50029-X</bibtext> </blist> <blist> <bibtext> Hooge ITC, Niehorster DC, Nyström M, Andersson R, Hessels RS. Fixation classification: How to merge and select fixation candidates. Behavior Research Methods. 2022; 54; 6: 2765-2776. 10.3758/s13428-021-01723-1. 35023066. 9729319</bibtext> </blist> <blist> <bibtext> Hu Z, Bulling A, Li S, Wang G. EHTask: Recognizing user tasks from eye and head movements in immersive virtual reality. IEEE Transactions on Visualization and Computer Graphics. 2023; 29; 4: 1992-2004. 10.1109/TVCG.2021.3138902. 34962869</bibtext> </blist> <blist> <bibtext> Huang, C.-M, Andrist, S, Sauppé, A, & Mutlu, B. (2015). Using gaze patterns to predict task intent in collaboration. Frontiers in Psychology, 6. https://doi.org/10.3389/fpsyg.2015.01049</bibtext> </blist> <blist> <bibtext> Hulle, N, Aroca-Ouellette, S, Ries, A. J, Brawer, J, von der Wense, K, & Roncone, A. (2024). Eyes on the Game: Deciphering Implicit Signals to Infer Human Intentions, Trust, and Skill. RO-MAN, x–x.</bibtext> </blist> <blist> <bibtext> Hulleman J, Olivers CNL. The impending demise of the item in visual search. Behavioral and Brain Sciences. 2017; 40. 10.1017/S0140525X15002794. 26673054e132</bibtext> </blist> <blist> <bibtext> Itti L, Koch C. A saliency-based search mechanism for overt and covert shifts of visual attention. Vision Research. 2000; 40; 10: 1489-1506. 10.1016/S0042-6989(99)00163-7. 10788654</bibtext> </blist> <blist> <bibtext> Karessli, N, Akata, Z, Schiele, B, & Bulling, A. (2017). Gaze Embeddings for Zero-Shot Image Classification. 4525–4534. https://openaccess.thecvf.com/content_cvpr_2017/html/Karessli_Gaze_Embeddings_for_CVPR_2017_paper.html</bibtext> </blist> <blist> <bibtext> Kennedy RS, Lane NE, Berbaum KS, Lilienthal MG. Simulator sickness questionnaire: An enhanced method for quantifying simulator sickness. The International Journal of Aviation Psychology. 1993; 3; 3: 203-220. 10.1207/s15327108ijap0303_3</bibtext> </blist> <blist> <bibtext> König P, Wilming N, Kietzmann TC, Ossandón JP, Onat S, Ehinger BV, Gameiro RR, Kaspar K. Eye movements as a window to cognitive processes. Journal of Eye Movement Research. 2016; 9; 5: 25</bibtext> </blist> <blist> <bibtext> Kothari R, Yang Z, Kanan C, Bailey R, Pelz JB, Diaz GJ. Gaze-in-wild: A dataset for studying eye and head coordination in everyday activities. Scientific Reports. 2020; 10; 1: 2539. 10.1038/s41598-020-59251-5. 32054884. 7018838</bibtext> </blist> <blist> <bibtext> Kothe, C, Shirazi, S. Y, Stenner, T, Medine, D, Boulay, C, Grivich, M. I, Mullen, T, Delorme, A, & Makeig, S. (2024). The Lab Streaming Layer for Synchronized Multimodal Recording (p. 2024.02.13.580071). bioRxiv. https://doi.org/10.1101/2024.02.13.580071</bibtext> </blist> <blist> <bibtext> Kristjánsson Á, Jóhannesson ÓI, Thornton IM. Common attentional constraints in visual foraging. PLoS ONE. 2014; 9; 6. 10.1371/journal.pone.0100752. 24964082. 4071029e100752</bibtext> </blist> <blist> <bibtext> Kristjánsson T, Draschkow D, Pálsson Á, Haraldsson D, Jónsson PÖ, Kristjánsson Á. Moving foraging into three dimensions: Feature- versus conjunction-based foraging in virtual reality. Quarterly Journal of Experimental Psychology. 2022; 75; 2: 313-327. 10.1177/1747021820937020</bibtext> </blist> <blist> <bibtext> Kristjánsson T, Thornton IM, Kristjánsson Á. Time limits during visual foraging reveal flexible working memory templates. Journal of Experimental Psychology: Human Perception and Performance. 2018; 44; 6: 827-835. 10.1037/xhp0000517. 29809049</bibtext> </blist> <blist> <bibtext> Kursa MB, Jankowski A, Rudnicki WR. Boruta–A system for feature selection. Fundamenta Informaticae. 2010; 101; 4: 271-285. 10.3233/FI-2010-288</bibtext> </blist> <blist> <bibtext> Le-Hoa Võ M, Wolfe JM. The role of memory for visual search in scenes. Annals of the New York Academy of Sciences. 2015; 1339; 1: 72-81. 10.1111/nyas.12667. 25684693</bibtext> </blist> <blist> <bibtext> Lisberger SG, Evinger C, Johanson GW, Fuchs AF. Relationship between eye acceleration and retinal image velocity during foveal smooth pursuit in man and monkey. Journal of Neurophysiology. 1981; 46; 2: 229-249. 10.1152/jn.1981.46.2.229. 7264712</bibtext> </blist> <blist> <bibtext> Long RJ, Ariyo AG, Iwu ZI, Azari DP, Madison AM, Aroca-Ouellette S, Roncone A, Tossell CC, Ries AJ. Towards an adaptive system for investigating human-agent teaming: Let's get cooking!. Systems and Information Engineering Design Symposium (SIEDS). 2024; 2024: 1-6. 10.1109/SIEDS61124.2024.10534646</bibtext> </blist> <blist> <bibtext> Lu Y, Sarter N. Eye tracking: A process-oriented method for inferring trust in automation as a function of priming and system reliability. IEEE Transactions on Human-Machine Systems. 2019; 49; 6: 560-568. 10.1109/THMS.2019.2930980</bibtext> </blist> <blist> <bibtext> MacInnes W, Hunt AR, Clarke ADF, Dodd MD. A generative model of cognitive state from task and eye movements. Cognitive Computation. 2018; 10; 5: 703-717. 10.1007/s12559-018-9558-9. 30740186. 6367733</bibtext> </blist> <blist> <bibtext> Mallick R, Slayback D, Touryan J, Ries AJ, Lance BJ. The use of eye metrics to index cognitive workload in video games. IEEE Second Workshop on Eye Tracking and Visualization (ETVIS). 2016; 2016: 60-64. 10.1109/ETVIS.2016.7851168</bibtext> </blist> <blist> <bibtext> Marathe, A. R, Schaefer, K. E, Evans, A. W, spsampsps Metcalfe, J. S. (2018). Bidirectional Communication for Effective Human-Agent Teaming. In J. Y. C. Chen spsampsps G. Fragomeni (Eds.), Virtual, Augmented and Mixed Reality: Interaction, Navigation, Visualization, Embodiment, and Simulation (pp. 338–350). Springer International Publishing. https://doi.org/10.1007/978-3-319-91581-4_25</bibtext> </blist> <blist> <bibtext> Marshall SP. Identifying Cognitive State from Eye Metrics. Aviation, Space, and Environmental Medicine. 2007; 78; 5: B165-B175. 17547317</bibtext> </blist> <blist> <bibtext> Metcalfe, J. S, Perelman, B. S, Boothe, D. L, & Mcdowell, K. (2021). Systemic Oversimplification Limits the Potential for Human-AI Partnership. IEEE Access, 9, 70242–70260. IEEE Access. https://doi.org/10.1109/ACCESS.2021.3078298</bibtext> </blist> <blist> <bibtext> Nyström M, Holmqvist K. An adaptive algorithm for fixation, saccade, and glissade detection in eyetracking data. Behavior Research Methods. 2010; 42; 1: 188-204. 10.3758/BRM.42.1.188. 20160299</bibtext> </blist> <blist> <bibtext> Oommen BS, Stahl JS. Amplitudes of head movements during putative eye-only saccades. Brain Research. 2005; 1065; 1: 68-78. 10.1016/j.brainres.2005.10.029. 16300748</bibtext> </blist> <blist> <bibtext> Pedregosa F, Varoquaux G, Gramfort A, Michel V, Thirion B, Grisel O, Blondel M, Prettenhofer P, Weiss R, Dubourg V, Vanderplas J. Scikit-learn: Machine learning in python. The Journal of Machine Learning Research. 2011; 12: 2825-2830</bibtext> </blist> <blist> <bibtext> Pelz J, Hayhoe M, Loeber R. The coordination of eye, head, and hand movements in a natural task. Experimental Brain Research. 2001; 139; 3: 266-277. 10.1007/s002210100745. 11545465</bibtext> </blist> <blist> <bibtext> Pettersson K, Tervonen J, Heininen J, Mäntyjärvi J. Head-area sensing in virtual reality: Future visions for visual perception and cognitive state estimation. Frontiers in Virtual Reality. 2024. 10.3389/frvir.2024.1423756</bibtext> </blist> <blist> <bibtext> Plamondon R. A kinematic theory of rapid human movements: Part I. Movement Representation and Generation. Biological Cybernetics. 1995; 72; 4: 295-307. 10.1007/BF00202785. 7748959</bibtext> </blist> <blist> <bibtext> Pomplun, M, & Sunkara, S. (2019). Pupil Dilation as an Indicator of Cognitive Workload in Human-Computer Interaction. In Human-Centered Computing (pp. 542–546). CRC Press.</bibtext> </blist> <blist> <bibtext> Pomplun M, Garaas TW, Carrasco M. The effects of task difficulty on visual search strategy in virtual 3D displays. Journal of Vision. 2013; 13; 3: 24. 10.1167/13.3.24. 23986539. 3756766</bibtext> </blist> <blist> <bibtext> Reingold EM, Glaholt MG. Cognitive control of fixation duration in visual search: The role of extrafoveal processing. Visual Cognition. 2014; 22; 3–4: 610-634. 10.1080/13506285.2014.881443</bibtext> </blist> <blist> <bibtext> Reis J, Cohen Y, Melão N, Costa J, Jorge D. High-tech defense industries: Developing autonomous intelligent systems. Applied Sciences. 2021. 10.3390/app11114920</bibtext> </blist> <blist> <bibtext> Salehi, M, Javadpour, N, Beisner, B, Sanaei, M, spsampsps Gilbert, S. B. (2024). Cybersickness Detection Through Head Movement Patterns: A Promising Approach. In H. Degen spsampsps S. Ntoa (Eds.), Artificial Intelligence in HCI (pp. 239–254). Springer Nature Switzerland. https://doi.org/10.1007/978-3-031-60611-3_18</bibtext> </blist> <blist> <bibtext> Schubert TW. The sense of presence in virtual environments: A three-component scale measuring spatial presence, involvement, and realness. Z. Für Medienpsychologie. 2003; 15; 2: 69-71</bibtext> </blist> <blist> <bibtext> Scott M, Su-In L. A unified approach to interpreting model predictions. Advances in Neural Information Processing Systems. 2017; 30: 4765-4774</bibtext> </blist> <blist> <bibtext> Sidenmark L, Gellersen H. Eye, head and torso coordination during gaze shifts in virtual reality. ACM Transactions on Computer-Human Interaction (TOCHI). 2019; 27; 1: 1-40. 10.1145/3361218</bibtext> </blist> <blist> <bibtext> Sidenmark L, Prummer F, Newn J, Gellersen H. Comparing gaze, head and controller selection of dynamically revealed targets in head-mounted displays. IEEE Transactions on Visualization and Computer Graphics. 2023; 29; 11: 4740-4750. 10.1109/TVCG.2023.3320235. 37782604</bibtext> </blist> <blist> <bibtext> Solman GJF, Kingstone A. Balancing energetic and cognitive resources: Memory use during search depends on the orienting effector. Cognition. 2014; 132; 3: 443-454. 10.1016/j.cognition.2014.05.005. 24946208</bibtext> </blist> <blist> <bibtext> Stahl JS. Amplitude of human head movements associated with horizontal saccades. Experimental Brain Research. 1999; 126; 1: 41-54. 10.1007/s002210050715. 10333006</bibtext> </blist> <blist> <bibtext> Stahl JS. Adaptive plasticity of head movement propensity. Experimental Brain Research. 2001; 139; 2: 201-208. 10.1007/s002210100749. 11497062</bibtext> </blist> <blist> <bibtext> Stahl JS. Eye-head coordination and the variation of eye-movement accuracy with orbital eccentricity. Experimental Brain Research. 2001; 136; 2: 200-210. 10.1007/s002210000593. 11206282</bibtext> </blist> <blist> <bibtext> Tagu J, Kristjánsson Á. Dynamics of attentional and oculomotor orienting in visual foraging tasks. Quarterly Journal of Experimental Psychology. 2022; 75; 2: 260-276. 10.1177/1747021820919351</bibtext> </blist> <blist> <bibtext> Theeuwes J. Stimulus-driven capture and attentional set: Selective search for color and visual abrupt onsets. Journal of Experimental Psychology: Human Perception and Performance. 1994; 20; 4: 799-806. 10.1037/0096-1523.20.4.799. 8083635</bibtext> </blist> <blist> <bibtext> Van Orden KF, Limbert W, Makeig S, Jung TP. Eye activity correlates of workload during a visuospatial memory task. Human Factors. 2001; 43; 1: 111-121. 10.1518/001872001775992570. 11474756</bibtext> </blist> <blist> <bibtext> Velichkovsky BM, Rothert A, Kopf M, Dornhöfer SM, Joos M. Towards an express-diagnostics for level of processing and hazard perception. Transportation Research Part f: Traffic Psychology and Behaviour. 2002; 5; 2: 145-156. 10.1016/S1369-8478(02)00013-X</bibtext> </blist> <blist> <bibtext> Vickers JN, Lewinski W. Performing under pressure: Gaze control, decision making and shooting performance of elite and rookie police officers. Human Movement Science. 2012; 31; 1: 101-117. 10.1016/j.humov.2011.04.004. 21807433</bibtext> </blist> <blist> <bibtext> Virtanen P, Gommers R, Oliphant TE, Haberland M, Reddy T, Cournapeau D, Burovski E, Peterson P, Weckesser W, Bright J, Van Der Walt SJ. SciPy 1.0: Fundamental algorithms for scientific computing in Python. Nature Methods. 2020; 17; 3: 261-272. 10.1038/s41592-019-0686-2. 32015543. 7056644</bibtext> </blist> <blist> <bibtext> Vlaskamp BN, Over EA, Hooge IT. Saccadic search performance: The effect of element spacing. Experimental Brain Research. 2005; 167; 2: 246-259. 10.1007/s00221-005-0032-z. 16078032</bibtext> </blist> <blist> <bibtext> Watson MR, Brennan AA, Kingstone A, Enns JT. Looking versus seeing: Strategies alter eye movements during visual search. Psychonomic Bulletin & Review. 2010; 17; 4: 543-549. 10.3758/PBR.17.4.543</bibtext> </blist> <blist> <bibtext> Wolfe JM. When is it time to move to the next raspberry bush? Foraging rules in human visual search. Journal of Vision. 2013; 13; 3: 10. 10.1167/13.3.10. 23641077. 4521330</bibtext> </blist> <blist> <bibtext> Yantis S, Jonides J. Abrupt visual onsets and selective attention: Voluntary versus automatic allocation. Journal of Experimental Psychology. Human Perception and Performance. 1990; 16; 1: 121-134. 2137514</bibtext> </blist> <blist> <bibtext> Young AH, Hulleman J. Eye movements reveal how task difficulty moulds visual search. Journal of Experimental Psychology: Human Perception and Performance. 2013; 39; 1: 168-190. 10.1037/a0028679. 22642215</bibtext> </blist> <blist> <bibtext> Zangemeister WH, Jones A, Stark L. Dynamics of head movement trajectories: Main sequence relationship. Experimental Neurology. 1981; 71; 1: 76-91. 10.1016/0014-4886(81)90072-8. 7449898</bibtext> </blist> <blist> <bibtext> Zangemeister WH, Stark L. Gaze latency: Variable interactions of head and eye latency. Experimental Neurology. 1982; 75; 2: 389-406. 10.1016/0014-4886(82)90169-8. 7106221</bibtext> </blist> </ref> <ref id="AN0187497571-34"> <title> Footnotes </title> <blist> <bibtext> RGB color values. Target – (1) 129, 103, 81, (2) 82, 67, 60, (3) 60, 65, 63, (4) 59, 60, 51. Easy distractor – (1) 97, 71, 44, (2) 94, 93, 53, (3) 199, 169, 86, (4) 59, 62, 47. Hard distractor – (1) 113, 86, 68, (2) 78, 86, 56, (3) 88, 80, 68, (4) 77, 71, 63.</bibtext> </blist> <blist> <bibtext> box plots were created using the boxchart MATLAB (2021b) function. The horizontal line inside each box corresponds to the sample median while the top and bottom edges of the box are the upper and lower quartiles, respectively. Values that are more than 1.5 times the interquartile range (IQR) from the top or bottom are designated as outliers (points beyond whiskers). Each whisker above and below the box represents the maximum value that is not considered an outlier. The top and bottom edges of the notched region correspond to <emph>m</emph> + (1.57⋅<emph>IQR</emph>)/√<emph>n</emph> and <emph>m</emph> − (1.57⋅<emph>IQR</emph>)/√<emph>n</emph>, respectively, where <emph>m</emph> is the median, <emph>IQR</emph> is the interquartile range, and <emph>n</emph> is the number of data points. The green diamond represents the sample mean with gray error bars showing 95% confidence intervals.</bibtext> </blist> <blist> <bibtext> Top features were extracted and models evaluated without target/distractor feature information and showed the same pattern of results though classification accuracy was generally lower. See Supplementary Materials.</bibtext> </blist> </ref> <aug> <p>By Anthony J. Ries; Chloe Callahan-Flintoft; Anna Madison; Louis Dankovich and Jonathan Touryan</p> <p>Reported by Author; Author; Author; Author; Author</p> </aug> <nolink nlid="nl1" bibid="bib64" firstref="ref2"></nolink> <nolink nlid="nl2" bibid="bib66" firstref="ref3"></nolink> <nolink nlid="nl3" bibid="bib76" firstref="ref4"></nolink> <nolink nlid="nl4" bibid="bib23" firstref="ref6"></nolink> <nolink nlid="nl5" bibid="bib28" firstref="ref7"></nolink> <nolink nlid="nl6" bibid="bib40" firstref="ref8"></nolink> <nolink nlid="nl7" bibid="bib62" firstref="ref9"></nolink> <nolink nlid="nl8" bibid="bib35" firstref="ref10"></nolink> <nolink nlid="nl9" bibid="bib58" firstref="ref11"></nolink> <nolink nlid="nl10" bibid="bib15" firstref="ref12"></nolink> <nolink nlid="nl11" bibid="bib48" firstref="ref13"></nolink> <nolink nlid="nl12" bibid="bib87" firstref="ref14"></nolink> <nolink nlid="nl13" bibid="bib95" firstref="ref15"></nolink> <nolink nlid="nl14" bibid="bib26" firstref="ref16"></nolink> <nolink nlid="nl15" bibid="bib37" firstref="ref17"></nolink> <nolink nlid="nl16" bibid="bib51" firstref="ref18"></nolink> <nolink nlid="nl17" bibid="bib65" firstref="ref19"></nolink> <nolink nlid="nl18" bibid="bib21" firstref="ref21"></nolink> <nolink nlid="nl19" bibid="bib45" firstref="ref22"></nolink> <nolink nlid="nl20" bibid="bib49" firstref="ref23"></nolink> <nolink nlid="nl21" bibid="bib39" firstref="ref24"></nolink> <nolink nlid="nl22" bibid="bib89" firstref="ref25"></nolink> <nolink nlid="nl23" bibid="bib90" firstref="ref27"></nolink> <nolink nlid="nl24" bibid="bib30" firstref="ref28"></nolink> <nolink nlid="nl25" bibid="bib63" firstref="ref29"></nolink> <nolink nlid="nl26" bibid="bib73" firstref="ref30"></nolink> <nolink nlid="nl27" bibid="bib88" firstref="ref31"></nolink> <nolink nlid="nl28" bibid="bib12" firstref="ref34"></nolink> <nolink nlid="nl29" bibid="bib70" firstref="ref36"></nolink> <nolink nlid="nl30" bibid="bib85" firstref="ref37"></nolink> <nolink nlid="nl31" bibid="bib80" firstref="ref39"></nolink> <nolink nlid="nl32" bibid="bib20" firstref="ref40"></nolink> <nolink nlid="nl33" bibid="bib24" firstref="ref41"></nolink> <nolink nlid="nl34" bibid="bib82" firstref="ref42"></nolink> <nolink nlid="nl35" bibid="bib34" firstref="ref44"></nolink> <nolink nlid="nl36" bibid="bib84" firstref="ref45"></nolink> <nolink nlid="nl37" bibid="bib17" firstref="ref49"></nolink> <nolink nlid="nl38" bibid="bib13" firstref="ref51"></nolink> <nolink nlid="nl39" bibid="bib10" firstref="ref54"></nolink> <nolink nlid="nl40" bibid="bib71" firstref="ref55"></nolink> <nolink nlid="nl41" bibid="bib11" firstref="ref57"></nolink> <nolink nlid="nl42" bibid="bib54" firstref="ref58"></nolink> <nolink nlid="nl43" bibid="bib94" firstref="ref59"></nolink> <nolink nlid="nl44" bibid="bib55" firstref="ref60"></nolink> <nolink nlid="nl45" bibid="bib56" firstref="ref61"></nolink> <nolink nlid="nl46" bibid="bib86" firstref="ref62"></nolink> <nolink nlid="nl47" bibid="bib29" firstref="ref63"></nolink> <nolink nlid="nl48" bibid="bib46" firstref="ref64"></nolink> <nolink nlid="nl49" bibid="bib52" firstref="ref65"></nolink> <nolink nlid="nl50" bibid="bib44" firstref="ref67"></nolink> <nolink nlid="nl51" bibid="bib77" firstref="ref68"></nolink> <nolink nlid="nl52" bibid="bib50" firstref="ref69"></nolink> <nolink nlid="nl53" bibid="bib78" firstref="ref71"></nolink> <nolink nlid="nl54" bibid="bib53" firstref="ref72"></nolink> <nolink nlid="nl55" bibid="bib67" firstref="ref73"></nolink> <nolink nlid="nl56" bibid="bib92" firstref="ref74"></nolink> <nolink nlid="nl57" bibid="bib22" firstref="ref75"></nolink> <nolink nlid="nl58" bibid="bib31" firstref="ref76"></nolink> <nolink nlid="nl59" bibid="bib42" firstref="ref77"></nolink> <nolink nlid="nl60" bibid="bib41" firstref="ref79"></nolink> <nolink nlid="nl61" bibid="bib60" firstref="ref80"></nolink> <nolink nlid="nl62" bibid="bib43" firstref="ref81"></nolink> <nolink nlid="nl63" bibid="bib14" firstref="ref82"></nolink> <nolink nlid="nl64" bibid="bib25" firstref="ref84"></nolink> <nolink nlid="nl65" bibid="bib75" firstref="ref85"></nolink> <nolink nlid="nl66" bibid="bib38" firstref="ref87"></nolink> <nolink nlid="nl67" bibid="bib19" firstref="ref88"></nolink> <nolink nlid="nl68" bibid="bib33" firstref="ref89"></nolink> <nolink nlid="nl69" bibid="bib68" firstref="ref90"></nolink> <nolink nlid="nl70" bibid="bib18" firstref="ref93"></nolink> <nolink nlid="nl71" bibid="bib59" firstref="ref97"></nolink> <nolink nlid="nl72" bibid="bib72" firstref="ref98"></nolink> <nolink nlid="nl73" bibid="bib32" firstref="ref101"></nolink> <nolink nlid="nl74" bibid="bib81" firstref="ref103"></nolink> <nolink nlid="nl75" bibid="bib83" firstref="ref104"></nolink> <nolink nlid="nl76" bibid="bib98" firstref="ref106"></nolink> <nolink nlid="nl77" bibid="bib57" firstref="ref107"></nolink> <nolink nlid="nl78" bibid="bib27" firstref="ref108"></nolink> <nolink nlid="nl79" bibid="bib69" firstref="ref109"></nolink> <nolink nlid="nl80" bibid="bib91" firstref="ref110"></nolink> <nolink nlid="nl81" bibid="bib79" firstref="ref111"></nolink> <nolink nlid="nl82" bibid="bib16" firstref="ref112"></nolink> <nolink nlid="nl83" bibid="bib97" firstref="ref114"></nolink> <nolink nlid="nl84" bibid="bib36" firstref="ref133"></nolink> <nolink nlid="nl85" bibid="bib74" firstref="ref138"></nolink> <nolink nlid="nl86" bibid="bib47" firstref="ref143"></nolink> <nolink nlid="nl87" bibid="bib96" firstref="ref144"></nolink>
Header DbId: eric
DbLabel: ERIC
An: EJ1481527
AccessLevel: 3
PubType: Academic Journal
PubTypeId: academicJournal
PreciseRelevancyScore: 0
IllustrationInfo
Items – Name: Title
  Label: Title
  Group: Ti
  Data: Decoding Target Discriminability and Time Pressure Using Eye and Head Movement Features in a Foraging Search Task
– Name: Language
  Label: Language
  Group: Lang
  Data: English
– Name: Author
  Label: Authors
  Group: Au
  Data: <searchLink fieldCode="AR" term="%22Anthony+J%2E+Ries%22">Anthony J. Ries</searchLink> (ORCID <externalLink term="http://orcid.org/0009-0004-3440-6276">0009-0004-3440-6276</externalLink>)<br /><searchLink fieldCode="AR" term="%22Chloe+Callahan-Flintoft%22">Chloe Callahan-Flintoft</searchLink><br /><searchLink fieldCode="AR" term="%22Anna+Madison%22">Anna Madison</searchLink><br /><searchLink fieldCode="AR" term="%22Louis+Dankovich%22">Louis Dankovich</searchLink><br /><searchLink fieldCode="AR" term="%22Jonathan+Touryan%22">Jonathan Touryan</searchLink>
– Name: TitleSource
  Label: Source
  Group: Src
  Data: <searchLink fieldCode="SO" term="%22Cognitive+Research%3A+Principles+and+Implications%22"><i>Cognitive Research: Principles and Implications</i></searchLink>. 2025 10.
– Name: Avail
  Label: Availability
  Group: Avail
  Data: Springer. Available from: Springer Nature. One New York Plaza, Suite 4600, New York, NY 10004. Tel: 800-777-4643; Tel: 212-460-1500; Fax: 212-460-1700; e-mail: customerservice@springernature.com; Web site: https://link.springer.com/
– Name: PeerReviewed
  Label: Peer Reviewed
  Group: SrcInfo
  Data: Y
– Name: Pages
  Label: Page Count
  Group: Src
  Data: 19
– Name: DatePubCY
  Label: Publication Date
  Group: Date
  Data: 2025
– Name: SourceSuprt
  Label: Sponsoring Agency
  Group: SrcSuprt
  Data: US Army Futures Command, Combat Capabilities Development Command Soldier Center (DEVCOM)
– Name: NumberContract
  Label: Contract Number
  Group: NumCntrct
  Data: W911NF2320205
– Name: TypeDocument
  Label: Document Type
  Group: TypDoc
  Data: Journal Articles<br />Reports - Research
– Name: Subject
  Label: Descriptors
  Group: Su
  Data: <searchLink fieldCode="DE" term="%22Stress+Variables%22">Stress Variables</searchLink><br /><searchLink fieldCode="DE" term="%22Time+Management%22">Time Management</searchLink><br /><searchLink fieldCode="DE" term="%22Decision+Making%22">Decision Making</searchLink><br /><searchLink fieldCode="DE" term="%22Reaction+Time%22">Reaction Time</searchLink><br /><searchLink fieldCode="DE" term="%22Eye+Movements%22">Eye Movements</searchLink><br /><searchLink fieldCode="DE" term="%22Human+Body%22">Human Body</searchLink><br /><searchLink fieldCode="DE" term="%22Motion%22">Motion</searchLink><br /><searchLink fieldCode="DE" term="%22Military+Training%22">Military Training</searchLink><br /><searchLink fieldCode="DE" term="%22Weapons%22">Weapons</searchLink><br /><searchLink fieldCode="DE" term="%22Search+Strategies%22">Search Strategies</searchLink><br /><searchLink fieldCode="DE" term="%22Assistive+Technology%22">Assistive Technology</searchLink><br /><searchLink fieldCode="DE" term="%22Computer+Simulation%22">Computer Simulation</searchLink><br /><searchLink fieldCode="DE" term="%22Discrimination+Learning%22">Discrimination Learning</searchLink><br /><searchLink fieldCode="DE" term="%22Performance+Based+Assessment%22">Performance Based Assessment</searchLink><br /><searchLink fieldCode="DE" term="%22Kinetics%22">Kinetics</searchLink><br /><searchLink fieldCode="DE" term="%22Visual+Perception%22">Visual Perception</searchLink><br /><searchLink fieldCode="DE" term="%22Visual+Learning%22">Visual Learning</searchLink><br /><searchLink fieldCode="DE" term="%22Classification%22">Classification</searchLink>
– Name: DOI
  Label: DOI
  Group: ID
  Data: 10.1186/s41235-025-00657-y
– Name: ISSN
  Label: ISSN
  Group: ISSN
  Data: 2365-7464
– Name: Abstract
  Label: Abstract
  Group: Ab
  Data: In military operations, rapid and accurate decision-making is crucial, especially in visually complex and high-pressure environments. This study investigates how eye and head movement metrics can infer changes in search behavior during a naturalistic shooting scenario in virtual reality (VR). Thirty-one participants performed a foraging search task using a head-mounted display (HMD) with integrated eye tracking. Participants searched for targets among distractors under varying levels of target discriminability (easy vs. hard) and time pressure (low vs. high). As expected, behavioral results indicated that increased discrimination difficulty and greater time pressure negatively impacted performance, leading to slower response times and reduced d-prime. Support vector classifiers assigned a search condition, discriminability and time pressure, to each trial based on eye and head movement features. Combined eye and head features produced the most accurate classification model for capturing tasked-induced changes in search behavior, with the combined model outperforming those based on eye or head features alone. While eye features demonstrated strong predictive power, the inclusion of head features significantly enhanced model performance. Across the ensemble of eye metrics, fixation-related features were the most robust for classifying target discriminability, while saccadic-related features played a similar role for time pressure. In contrast, models constrained to head metrics emphasized global movement (amplitude, velocity) for classifying discriminability but shifted toward kinematic intensity (acceleration, jerk) in time pressure condition. Together these results speak to the complementary role of eye and head movements in understanding search behavior under changing task parameters.
– Name: AbstractInfo
  Label: Abstractor
  Group: Ab
  Data: As Provided
– Name: DateEntry
  Label: Entry Date
  Group: Date
  Data: 2025
– Name: AN
  Label: Accession Number
  Group: ID
  Data: EJ1481527
PLink https://search.ebscohost.com/login.aspx?direct=true&site=eds-live&db=eric&AN=EJ1481527
RecordInfo BibRecord:
  BibEntity:
    Identifiers:
      – Type: doi
        Value: 10.1186/s41235-025-00657-y
    Languages:
      – Text: English
    PhysicalDescription:
      Pagination:
        PageCount: 19
    Subjects:
      – SubjectFull: Stress Variables
        Type: general
      – SubjectFull: Time Management
        Type: general
      – SubjectFull: Decision Making
        Type: general
      – SubjectFull: Reaction Time
        Type: general
      – SubjectFull: Eye Movements
        Type: general
      – SubjectFull: Human Body
        Type: general
      – SubjectFull: Motion
        Type: general
      – SubjectFull: Military Training
        Type: general
      – SubjectFull: Weapons
        Type: general
      – SubjectFull: Search Strategies
        Type: general
      – SubjectFull: Assistive Technology
        Type: general
      – SubjectFull: Computer Simulation
        Type: general
      – SubjectFull: Discrimination Learning
        Type: general
      – SubjectFull: Performance Based Assessment
        Type: general
      – SubjectFull: Kinetics
        Type: general
      – SubjectFull: Visual Perception
        Type: general
      – SubjectFull: Visual Learning
        Type: general
      – SubjectFull: Classification
        Type: general
    Titles:
      – TitleFull: Decoding Target Discriminability and Time Pressure Using Eye and Head Movement Features in a Foraging Search Task
        Type: main
  BibRelationships:
    HasContributorRelationships:
      – PersonEntity:
          Name:
            NameFull: Anthony J. Ries
      – PersonEntity:
          Name:
            NameFull: Chloe Callahan-Flintoft
      – PersonEntity:
          Name:
            NameFull: Anna Madison
      – PersonEntity:
          Name:
            NameFull: Louis Dankovich
      – PersonEntity:
          Name:
            NameFull: Jonathan Touryan
    IsPartOfRelationships:
      – BibEntity:
          Dates:
            – D: 01
              M: 12
              Type: published
              Y: 2025
          Identifiers:
            – Type: issn-electronic
              Value: 2365-7464
          Numbering:
            – Type: volume
              Value: 10
          Titles:
            – TitleFull: Cognitive Research: Principles and Implications
              Type: main
ResultId 1