Authors: Armin Toghi, Azar Mohammadzadeh, Zahra Alemi, Anahita Khorrami Banaraki
Categories: Systematic Review, Eye-tracking, Inattention, Psychiatric disorders, Oculomotor metrics, Transdiagnostic biomarkers
Source: BMC Psychiatry
Authors: Armin Toghi, Azar Mohammadzadeh, Zahra Alemi, Anahita Khorrami Banaraki
Attention impairment is a dimensional and heterogeneous trait, distributed continuously across the population. Understanding the pathophysiology of eye movement control offers valuable insight into attention dysfunctions and their underlying neural circuits. This systematic review aims to map different oculomotor paradigms and metrics to key attention systems across a range of psychiatric conditions, highlighting their potential for identifying biological markers of attention.
We conducted a systematic search on PubMed, Scopus, and Web of Science databases using keywords related to ‘eye-tracking,’ ‘(in)attention,’ and ‘mental disorders.’ Seventy-five studies were included, categorized into three core domains of Selective Attention (spatial/feature), Sustained Attention, and Executive Control, based on the associated oculomotor paradigms. These studies covered various psychiatric conditions, including Attention Deficit Hyperactivity Disorder (ADHD), Autism Spectrum Disorder (ASD), Tourette Syndrome (TS), Obsessive-Compulsive Disorder (OCD), Borderline Personality Disorder (BPD), Developmental Coordination Disorder (DCD), and Schizophrenia Spectrum and Other Psychotic Disorders (SSD).
The findings highlighted impairments in several key oculomotor metrics shared across psychiatric conditions, and moderately correlated with self-reported and behavioral measures of inattention symptoms. These include antisaccade direction errors, fixation durations on task-relevant/irrelevant areas, the latency to first fixation, and anticipatory and intrusive saccades. These metrics are particularly important in paradigms requiring endogenous shifts of attention.
Eye-tracking metrics could serve as transdiagnostic biomarkers for identifying inattention symptoms across normative and psychiatric conditions. Further research is needed to transition eye-tracking from a research tool to a clinically actionable biomarker for personalized assessment and intervention.
The online version contains supplementary material available at 10.1186/s12888-025-07415-w.
Attention impairments are commonly associated with the clinical definition of Attention Deficit Hyperactivity Disorder (ADHD) [1]. However, such impairments also manifest to varying degrees in normative populations and across a wide range of mental health conditions, including Autism Spectrum Disorder (ASD) [2–6], Tourette Syndrome (TS) [7], Major Depressive Disorder (MDD) [8], Borderline Personality Disorder (BPD) [9], Developmental Coordination Disorder (DCD) [10], and Schizophrenia Spectrum and Other Psychotic Disorders (SSD) [11–13]. The heterogeneous nature of attention deficits across psychiatric conditions underscores the importance of identifying biological markers of attention and moving beyond ADHD-specific clinical measures.
Converging evidence indicates that eye movement and attention control are intricately linked [14]. The fundamental neural circuitry underlying oculomotor tasks substantially overlaps with attention control networks [15, 16]. Notably, mental health conditions characterized by attention deficits exhibit abnormal oculomotor behavior, which correlates with both behavioral and self-reported measures of inattention [17, 18]. As such, observing oculomotor patterns during visual exploration provides a compelling avenue for identifying transdiagnostic neurobehavioral biomarkers of attention [19]. Such biomarkers offer a more nuanced understanding of inattention across diverse populations, presenting promising avenues for examining comorbidities in psychiatric conditions, identifying high-risk individuals, and monitoring disease progression [4, 20].
The emergence of eye-tracking technology has provided a promising approach to studying oculomotor behavior [21]. With its sub-millisecond temporal resolution [22], high spatial accuracy [23] operational convenience setup [24] and applicability in natural environments [23, 25], eye-tracking has become a popular tool for investigating cognitive processes. Moreover, exploring the neural correlates of oculomotor parameters could further underscore its utility as a “brain mapper” in neuropsychological testing, offering indirect yet valuable insights into the abnormal neural circuits implicated in cognitive dysfunctions [26]. This technology opens a new way of studying psychopathologies in terms of pathophysiology, with low cost and high precision. In line with this, recent studies increasingly integrate behavioral measures of attention with eye-tracking metrics to enhance the prediction of psychopathological symptoms [18, 27–29].
Oculomotor tasks using eye-tracking technology have been extensively studied in various mental health conditions. While different eye-tracking paradigms engage distinct dimensions of attention and cognitive control, mapping these paradigms and their associated parameters to attention systems could facilitate progress in identifying the physiological mechanisms underlying attentional dysfunctions in psychiatric conditions [1]. This approach aligns with the National Institute of Mental Health’s (NIMH) Research Domain Criteria (RDoC) project, which emphasizes studying psychopathologies through their underlying pathophysiology beyond the strict symptom criteria of the categorization systems [30]. Unlike behavioral measures, which are often diverse and difficult to standardize [31, 32], oculomotor metrics align with the RDoC goal of developing reliable, circuit-based measures of mental constructs [30]. According to the RDoC framework, paradigms and metrics incorporating the concept of capacity limitation and competition are linked to the Attention construct. This term directly relates to the definition of Selective Attention, which refers to the selective processing of specific stimuli over others. At the same time, other paradigms currently listed under Cognitive Control—especially those assessing Sustained Attention and Executive Control—remain crucial for clinical and neuropsychological evaluations of inattention. These paradigms involve intertwined inhibitory and non-inhibitory processes (e.g., attention control), which are difficult to disentangle [33].
Selective attention refers to the preferential processing of attention that prioritizes sensory inputs by selecting based on their location (Spatial Domain) or visual features (Feature Domain-color or shape) [34]. Selective attention can be driven by exogenous mechanisms, which are stimulus-driven and automatic (e.g., triggered by sudden or salient external events), or by endogenous mechanisms, which are goal-directed and under voluntary control (e.g., guided by internal expectations or instructions). The prosaccade task is a widely used eye-tracking paradigm to assess basic oculomotor metrics associated with exogenous spatial attention, such as latency and accuracy. In this task, participants focus on a fixation marker; after a brief interval, a peripheral target appears, eliciting involuntary saccades toward the target. Endogenous cues presented before stimulus onset, such as in spatial cueing tasks, can also predict the target’s location and induce voluntary shifts of attention. Manipulations of the location, duration, and timing of fixation markers (Gap, Step, and Overlap conditions) and peripheral targets could further examine the priming, orienting, and shifting of visual attention in psychiatric conditions [35]. In contrast, targets can also be attended to based on visual features rather than spatial location. Tasks such as visual search and change detection guide attention toward specific stimulus features and are commonly used to study feature-based attention [1]. Eye-tracking metrics in these tasks extend beyond simple reaction times to include measures of search initiation (e.g., latency to the first fixation), search patterns (e.g., Scanning length), and fixation durations on target versus distractors. Neuroimaging studies repeatedly support the involvement of the lateral frontal eye field (FEF), supplementary eye field (SEF), and posterior parietal cortex (PPC) in visually guided saccades [36]. These regions control saccades with direct connections to the superior colliculus (SC). Complementarily, lobules VI and VII of cerebellar vermis contribute to the coordination of visually guided saccades with interactions with brainstem circuitry and SC (Fig. 1.A) [37]. However, the neural mechanism under feature-based attention is less investigated. While feature-based attention shared a core neural mechanism with spatial attention, primate studies suggested the further involvement of the inferior frontal junction and the inferior temporal (IT) cortex [1].Fig. 1Neural circuits underlying oculomotor tasks across different visual attention domains. Abbreviations: ACC Anterior cingulate cortex, CN Caudate nucleus, CRBL (V6–7) Cerebellar vermis lobules VI and VII, dlPFC Dorsolateral prefrontal cortex, IC Insular cortex, IT Inferior temporal cortex, LC Locus coeruleus, LFEF Lateral frontal eye field, mFEF Medial frontal eye field, PN Pontine nuclei, PPC Posterior parietal cortex, pre-SMA Pre-supplementary motor area, SC Superior colliculus, SEF Supplementary eye field, SNpr Substantia nigra pars reticulata, THA Thalamus
Sustained attention is the ability to maintain focus on a task over an extended period without significant declines in performance. It is partially mediated by the arousal system and gated by the Locus coeruleus–norepinephrine (LC–NE) system [38], but it also needs top-down inhibition of prepotent responses. Performance in sustained attention tasks (e.g., Continuous Performance Tasks (CPT)) is closely tied to the ability to detect a rare target stimulus while ignoring non-targets. Eye-tracking technology can measure tonic (resting) alertness by assessing pupil diameter during simple fixation tasks and phasic alertness by introducing cues or warning signals. These measures can provide insights into potential aberrations in LC–NE function in psychiatric conditions and their links to deficits in sustained attention [39, 40]. Fixation metrics, such as fixation duration on target-relevant versus irrelevant areas and intrusive saccades during task execution, provide further insights into attention control and correlate with self-reported inattention in both normative and ADHD populations [18, 27, 41]. In terms of neural circuits, multiple brain regions support visual sustained attention, which is dissociable from executive control [42] and spatial attention [1]. As said, the LC-NE system modulating arousal state influences sustained attention performance. Furthermore, the brain network of visual sustained attention consists of the anterior cingulate cortex (ACC), dorsolateral prefrontal cortex (dlPFC), thalamus, insula, caudate nucleus, and IT (Fig. 1.B) [42, 43].
Eye-tracking paradigms that assess executive control examine how participants deal with conflicts [34]. Performance in these paradigms requires two interdependent processes, starting with the inhibition of the unwanted reflexive saccade to the target, No-Go or stop signal, and the consecutive triggering of a voluntary correct saccade made in the direction opposite to the target, or the target for Go Signals [44]. Although inhibitory control is central to these paradigms, performance is closely linked to attentional processes [13], with attention distribution to the cue and stimulus target influencing directional errors [12, 33]. Meta-analysis of oculomotor fMRI studies further supports this view, showing that activation of area 7 A in the PPC plays a crucial role in shifting the locus of spatial attention in antisaccade paradigms [36]. Furthermore, response inhibition tasks elicit stronger activation in medial FEF, SEF, dlPFC, and dorsal ACC, which are associated with motor preparation and inhibition [36, 44] (Fig. 1.C). These findings underscore how response inhibition deficits may reflect inefficient interactions between exogenous and endogenous attention systems [33].
In the current study, we aimed to map eye-tracking paradigms and associated parameters to attention systems across various psychiatric conditions. To achieve this, we conducted a systematic review of articles that employed eye-tracking technology to extract oculomotor parameters in non-valenced attention-demanding tasks. We categorized the studies based on their oculomotor tasks into one of three attention systems—sustained attention, selective attention, or executive control—as shown in Table 1. An overview of the included oculomotor paradigms, potential experimental manipulations, and related metrics is provided in Tables 2 and 3.Table 1Included oculomotor task/metrics and their relation to the visual attention system and related RDoC constructsOculomotor taskRDoC ConstructVisual Attention SystemsEye tracking metricsProsaccade TaskSpatial Cueing TaskAttentionSelective Attention(Spatial Domain)Latency, Accuracy (gain), Direction errors, Peak velocity, Anticipatory saccade, Intra-subject latency variability, Gap/Overlap effectVisual Search TaskChange Detection TaskAttentional BlinkSelective Attention(Feature Domain)Latency, Fixation duration, Fixation count, Scanning time, Scanning length, Latency to first fixation/saccade, Variability in latency to first fixation/saccade, Visual span size, Gaze dispersion, Post-search durationContinuous Performance TestFixation TaskSmooth Pursuit TaskSustained Attention (Without Competition-Distractor)Pupil diameter, Intrusive saccade, Fixation duration, Fixation count, Gaze dispersion, Accuracy (gain), Gaze variability, Catch-up saccadeCognitive Control/Response Selection, Inhibition/SuppressionSustained Attention (With Competition-Distractor)Stroop TaskAntisaccade TaskCountermanding TaskGo/No Go TaskDelayed Oculomotor Response TaskExecutive ControlLatency, Accuracy (gain), Direction errors, Fixation duration, Fixation count, Latency to first fixation/saccade, Intra-subject latency variability, Anticipatory saccadeTable 2Overview of included oculomotor paradigms and their experimental manipulationsOculomotor paradigmsDescriptionExperimental ManipulationProsaccade TaskParticipants fixate on a central marker. After a brief interval, a peripheral target appears, eliciting a reflexive saccade toward the target.The fixation marker may disappear before (Gap), at (Step), or after (Overlap) stimulus onset.Spatial Cueing TaskParticipants fixate on a central marker. After a brief interval, a cue (e.g., a central arrow) appears before the peripheral stimulus to induce voluntary shifts of attention.The cue and target may appear at the same location (Valid), opposite locations (Invalid), or without a preceding cue (Neutral).Visual Search TaskParticipants are instructed to detect a specific target feature among distractors in a visual scene.Target–distractor similarity (e.g., color, shape, or complexity) is varied. Target presence may also be manipulated.Change Detection TaskParticipants fixate on a central marker. After a brief interval, two real-world images are shown in sequence, separated by a blank screen. Participants must detect any changes.Changes may occur in central or marginal features.Attentional BlinkParticipants view a rapid sequence of letters. In probe-only blocks, they identify a probe letter; in probe-plus-target blocks, they identify both a target and a subsequent probe.The temporal interval between the target and the probe is varied.Continuous Performance TestParticipants fixate on a task area and are instructed to detect rare target stimuli among non-targets.Visual and/or auditory distractors may be present or absent. Stimulus presentation duration may also be manipulated.Fixation TaskParticipants maintain fixation on a central or peripheral marker for a sustained period.Presence or absence of distractors during the fixation period.Smooth Pursuit TaskParticipants are instructed to follow a moving target as accurately as possible.Target speed (velocity) is manipulated.Stroop TaskParticipants are shown a series of color words and must name the color while ignoring the word itself.Word meaning and color may be congruent (aligned) or incongruent (conflicting).Antisaccade TaskParticipants fixate on a central marker. After a brief interval, a peripheral target appears, and they are instructed to saccade in the opposite direction of the peripheral stimulus.The fixation marker may disappear before (Gap), at (Step), or after (Overlap) stimulus onset.Countermanding TaskParticipants initiate a saccade toward a peripheral stimulus but must inhibit it if a stop signal is presented.Stop-signal delay and type (visual or auditory) are manipulated.Go/No Go TaskParticipants are instructed to make a saccade toward a Go stimulus and to inhibit it in response to a No-Go stimulus.Inter-stimulus interval and stimulus type are manipulated.Delayed Oculomotor Response TaskParticipants are instructed to delay their saccade to a peripheral target until the stimulus disappears.The delay interval (the duration participants must wait before initiating a saccade) is varied.Table 3Summary of oculomotor metricsOculomotor metricsDescriptionSaccade latencyReaction time between target appearance and saccade initiationSaccade accuracy (gain)The ratio of eye movement amplitude to target displacementIntra-subject latency variabilityStandard deviation of saccadic latency across trialsPeak velocityMaximum speed achieved during a saccadeGap effectSaccade latency difference between gap and overlap conditionsOverlap effectIncreased saccade latency when fixation point remains during target onsetDirection errorProportion of incorrect saccades made toward distractors or in the opposite direction of instructed targetsAnticipatory saccadePremature eye movements initiated before target onsetFixation durationTime spent fixating on a specific locationFixation countNumber of discrete fixationsLatency to first fixation/saccadeTime from stimulus onset to the first fixation/saccadeVariability in latency to the first fixation/saccadeStandard deviation of latency to first fixation/saccade across trials.Scanning timeDuration between the initial fixation following stimulus onset and the fixation on the targetScanning lengthTotal distance covered by eye movements during exploration; sum of all saccadic amplitudesPost-search durationTime spent fixating on the target after it has been identifiedVisual span sizeThe region of the visual field from which information is extracted during a single eye fixation.Gaze dispersionSmallest convex area enclosing all fixation pointsGaze variabilityStandard deviation of gaze coordinates (gaze SD) during explorationIntrusive saccadesInvoluntary eye movements that disrupt fixationCatch-up saccadeSaccade during smooth pursuit that corrects for the mismatch between eye and moving target’s location.Pupil diameterPupil size changes reflecting tonic and phasic pupillary system activity
While we discuss our findings in relation to previous meta-analyses on oculomotor metrics in psychiatric conditions [6, 45–48], our systematic review differs in several important
We imposed no restrictions on the type of psychiatric condition, allowing us to include studies across a broad range of conditions with inattention symptoms.Our review includes a wide variety of oculomotor tasks only if they directly manipulate attention (e.g., assessments of capacity, attentional load, or direction of attention) and align with one of the defined attention systems.We exclusively included studies that used eye-tracking instruments to measure oculomotor metrics. Studies using alternative methods, such as electrooculography (EOG), were excluded due to their lower spatial and temporal resolution.
To evaluate the psychopathology of attention in mental disorders with eye-tracking parameters, we conducted a systematic review with the standards of the Preferred Reporting Items for Systematic Review and Meta-Analysis (PRISMA) (Fig. 2**)** [49].Fig. 2PRISMA flowchart. Graphical representation of the number of papers retrieved, screened, and included in the systematic review [50]
Our primary objective focused on three main ‘eye-tracking,’ ‘(in)attention,’ and ‘mental disorder.’ The search string was constructed using combinations of keywords from these core concepts, as detailed in the Supplementary Material. Following PRISMA 2020 guidelines, relevant keywords were derived from peer-reviewed systematic reviews and meta-analyses [51]. The search was conducted in PubMed, Scopus, and Web of Science databases. The initial search was completed on May 3, 2023. An updated search rerun prior to journal publication, covering studies published between May 3, 2023 and June 29 2025. Additionally, a snowball search was performed by reviewing references of identified systematic and meta-analyses and using PubMed to screen studies citing them.
Studies were included if they met all the following (1) Used eye-tracking technology to extract oculomotor parameters. (2) Participants had a mean age above 7 years, with no restrictions on the type or severity of the psychiatric condition. (3) Investigated attention functions (e.g., Selective attention, Sustained attention, and Executive control) using oculomotor tasks that directly manipulated attention. (4) Peer-reviewed studies published in English.
We excluded studies published before 2000, reviews, non-human studies, and studies involving neurological disorders (e.g., Alzheimer’s, Parkinson’s, aphasia, hemineglect), brain injuries, or concussions. Studies focusing on attentional biases related to emotional stimuli or addiction were excluded, as these impairments are typically secondary to external substance responses. No restrictions were applied regarding geographic location or sample size.
A total of 2,127 records were extracted after the initial search on May 3, 2023. The updated search during the peer-review process resulted in the addition of 70 more records. Duplicates were removed using EndNote and Manually checked by two reviewers. After removing duplicates, 1,557 records were imported into ASReview software, an open-source machine learning-assisted tool for screening titles and abstracts [52]. Active learning techniques were employed to prioritize relevant records, with manual input from reviewers. Three reviewers (AT, ZA, AM) worked in independent groups to screen records. Reviewers alternated screening in 24-hour shifts followed the same active learning model and data, but used different prior records. Relevant articles were compiled in an Excel file after screening. After applying eligibility criteria, 89 studies proceeded to full-text screening using Rayyan, enabling reviewers to review and annotate studies collaboratively [53]. Discrepancies were resolved through discussion, and final decisions were consulted with the domain expert (AK).
Following full-text screening, 75 studies were included in the review. Reasons for exclusions at this stage are detailed in the Supplementary Material. Three authors (AT, ZA, AM) manually extracted data on key variables, including author, publication year, sample size, developmental stage, diagnostic manual (i.e., Diagnostic and Statistical Manual of Mental Disorders (DSM), or International Classification of Diseases (ICD)), rating scale (self-reports and clinical evaluations), oculomotor task, experimental design, eye-tracking model, and main findings (Table 4). Each study was categorized into one of three attention systems—sustained attention, selective (Spatial) attention, selective (Feature-based) attention, or executive control (Response inhibition)—based on the oculomotor task.Table 4Description of 75 included studies in the systematic reviewAuthor (Year)Sample Size(Age)Diagnostic ManualRating Scale (cutoff)Developmental StageOculomotor TaskEye Tracking (Sampling Rate)Main Finding Ono et al., 2021 [54]19 ADHD (9.1), 18 ASD (8.4), 30 Control (8.2)DSM-5WISC/WPPSIChildren and AdolescentsProsaccade, AntisaccadeEyelink 1000 Plus system (500 Hz)The ADHD group has lower peak velocity in the overlap Pro task, while the ASD group has lower corrective saccade in the step Pro task. Fernandez-Ruiz et al., 2020 [55]22 ADHD-C (12.6), 20 Control (12.9)DSM-IVNAChildren and AdolescentsProsaccade, AntisaccadeISCAN ETL-400 camera (120 Hz)The ADHD group has more DE and longer SRT in the AS task and hyperactivation in dlPFC in the preparatory phase. Jung et al., 2015 [56] 9 Tourette (14.1), 30 Control (14.1)DSM-IV-TRWASI, YGTSSChildren and AdolescentsProsaccade, AntisaccadeOber saccadometer (position 1000 Hz)The Tourette group exhibited longer SRT and shorter peak velocity in the Pro-only task but demonstrated enhanced task-switching performance in both Pro and Anti-mixed trials. Klein et al., 2003 [57] 46 ADHD (11.3), 46 Control (11.3)ICD 10, DSM-IV (DIPS)RPMChildren and AdolescentsProsaccade, AntisaccadeIris scaler (512 Hz)In the ADHD group, higher DE and fewer express saccades were independent of age, while the augmentation of SRT during the AS compared to Pro was dependent on age. Huang & Chan, 2020 [58] 12 ADHD-C (8.8), 12 Control (9.1)DSM-5NAChildren and AdolescentsProsaccade, AntisaccadeSMI Hi-Speed eye tracker (500 Hz)In the Pro task, the ADHD group was faster and more accurate, but in AS trials, they exhibited more DE compared to the control.Parvinchi and Sandor et al., 2013 [7]9 TS + ADHD (12.1), 7 TS + ADHD + OCD (14.4), 6 Control (11)DSM-III-RYGTSS, Y-BOCS, CGIChildren and AdolescentsProsaccade, AntisaccadeEl-Mar Series 2020 Eye-Tracker (120 Hz)TS-ADHD group showed longer saccadic latency. Karatekin et al., 2010 [59] 26 ADHD (12), 29 Psychosis (14.5), 48 Control (14.8).DSM-IV (K-SADS)WISC (> 70), SANS, SAPSChildren and AdolescentsProsaccade, AntisaccadeISCAN Eye Tracking Laboratory, Model ETL-400 (60 Hz)Unlike ADHD patients, individuals with youth-onset Psychosis, had more AS error rates. Both clinical groups showed more variable SRT in both the Pro and AS tasks. Karatekin, 2006 [60] 10 ADHD (14.25), 15 Control (10.25), 18 Control (19.1), 15 Control (15)DSM-IV (K-SADS)WISC, YCI, CBCChildren and AdolescentsProsaccade, AntisaccadeISCAN Eye Tracking Laboratory, Model ETL-400 (60 Hz)SRT and accuracy in AS tasks were impaired in the ADHD group, but this effect diminished with different AS manipulations. Carr et al., 2010 [61] 34 ADHD-I (15.3), 37 ADHD-C (15), 71 Control (15.5)DSM-IV (K-SADS)WISC (> 75), CBC, CPRS-R: SChildren and AdolescentsProsaccade, AntisaccadeISCAN, ETL-400 (240 Hz)ADHD combined type made more DE in AS tasks. Caldani et al., 2020 [62] 32 ASD (12.1), 32 Control (11)DSM-5 (ADOS, ADI-R)WISCChildren and AdolescentsProsaccade, AntisaccadeMobile EBT (300 Hz)Saccadic latency, error rate, and gain were intact in the ASD group compared to the control. The ASD group made more express and anticipatory saccades than the control. Amestoy et al., 2021 [35]35 ASD (29), 29 ASD (14.1), 33 ASD (7.9), 49 (21.2)DSM-5 (ADOS, ADI-R, K-SADS, DIGS)WISC (> 70), BRIEF, ADHD-RSChildren, Adolescents and AdultsProsaccade, AntisaccadeTobii Pro TX300 (300 Hz)Intact Pro metrics in ASD children, adolescents, and adults. There was higher DE in the ASD group compared to the control. Calancie et al., 2024 [9]25 BPD (16.4), 24 BPD/ADHD (16.3), 53 Control (15.7)DSM-5 (SCID-PD-5)BISChildren, AdolescentsProsaccade, AntisaccadeEyelink 1000 Plus (500 Hz)The BPD and BPD/ADHD groups showed greater variability in saccade latency and reduced fixation on task cues. Unlike the BPD group, individuals with BPD/ADHD also made more anticipatory saccades and AS DE. Kleberg et al., 2020 [63] 24 ADHD (10.5), 47 Control (10.4)DSM-5WISC, SNAP, SDQChildren and AdolescentsProsaccadeTobii TX120 (60 Hz)Slower saccadic latency in the ADHD group was compensated after introducing a phasic alerting auditory cue. Kleberg et al., 2023 [64] 54 ADHD (T1: 10.55, T2, 12.53)DSM-5ADHD-RS, SNAP, SCASChildren and AdolescentsProsaccadeTobii TX120 (60 Hz)Decreasing saccade latency after phasic alerting cues predicts ADHD symptoms over a 2-year period. Wilmut et al., 2007 [10]7 DCD (7.5), 18 Control (3.1), 18 Control (4.1), 10 Control (3.1)NAWISC/WPPSI, MABC (< 10th percentile)Children and AdolescentsProsaccadeInfrared camera (-)Saccade latency and attention disengagement index in the look and hit condition, but not the look condition, were higher in the DCD group. Matsuo et al., 2015 [65] 37 ADHD (7.9), 88 Control (7.8)DSM-IVADHD-RS, CBCLChildren and AdolescentsProsaccadeT.K.K.2930a infrared limbus detection (1000 Hz)The ADHD group had a lower gap effect in the Pro task compared to the controls. The gap effect didn’t correlate with ADHD rating scales.Bilbao & Pinero., 2021 [66]7 Dyslexia, 6 DCD, 4 ADHD (8.2), 15 Control (9.3)DSM-5NAChildren and AdolescentsProsaccadeTobii Eye X (-)The neurodevelopmental disorder group had higher hypometric saccades and regression compared to controls.Sanchez et al., 2020 [67]14 ADHD (11.1), 158 Monozygotic, 162 Dizygotic Control (11.1)NA.WISC, CPRSChildren and AdolescentsAntisaccadeTobii T120 (120 Hz)Anticipatory eye movement, but not DE in the AS task, was positively correlated with the inattentive trait. Long et al., 2015 [68]121 Healthy (14)DSM-IV (K-SADS)BRIEFChildren and AdolescentsAntisaccadeOptics eye-tracking system (-)AS performance correlated with disruptive behaviour disorder symptoms. Schwarz et al., 2015 [69]11 ADHD (10.4), 11 Control (9.6)DSM-IV-TRASEBA (> 63), VADPRS (> 6 symptoms)Children and AdolescentsAntisaccadeIView X MRI-LR, SensoMotoric InstrumentsAlthough the ADHD group did not significantly differ in AS performance, they exhibited hyperactivation in the right dlPFC and caudate nucleus. Bucci et al., 2017 [70] 31 ADHD (9.8), 31 Control (9.7)DSM-5 (K-SADS)WISC, ADHD-RS, MABCChildren and AdolescentsProsaccade, Antisaccade, FixationMobil EBT (300 Hz)The ADHD group has shorter latency in the overlap/step Pro task, more DE in the AS task, and increased intrusive saccades during fixation task. Hanisch et al., 2006 [71]22 ADHD (12), 22 Control (11.8)DSM-IV (K-DIPS)WISC (> 85), CBC, FBB-HKSChildren and AdolescentsProsaccades AntisaccadeFixationCountermandingEyelink2 (250 Hz)The ADHD group had a lower percentage of correct saccades in countermanding tasks compared to the control. Loe et al., 2009 [72]14 ADHD-C (10.2), 12 ADHD-I (10.4), 33 Control (10.4)DSM-IVWISC, VADPRS (rate 2 or 3 in 6 of 9 inattention symptoms), CBCLChildren and AdolescentsProsaccade, Antisaccade, Fixation, Memory-guided saccadesApplied Science Laboratories model 504 (60 Hz)The ADHD group has more errors in AS and fixation tasks. The ADHD-C group has more variability in SRT with respect to the interstimulus fixation period. O’Driscoll et al., 2005 [73]12 ADHD-I (12.74), 10 ADHD-C (12.38), 10 Control (12.66)DSM-IVWISC (> 85), CPRS-R, CSI-4Children and AdolescentsProsaccade, Antisaccade, Predictable taskEyelink (250 Hz)ADHD-C group, but not the ADHD-I group, have lower predictive saccade and more DE in the AS task. Bucci et al., 2014 [74]28 ADHD (9.6), 14 Control (9.75)NACPRS, CTRSChildren and AdolescentsProsaccade, Antisaccade,Fixation,Pursuit eye movementMobile EBT (300 Hz)Intrusive saccades during fixation and latency in the AS task were higher in the ADHD group. Goldberg et al., 2002 [75]11 ASD (HFA) (13.8), 11 Control (14.4)DSM-IV(ADI-R, ADOS-G)WISC (> 80)Children and AdolescentsProsaccade, Antisaccade, Predictive saccade, Memory guided saccadeBinocular infrared oculography (250 Hz)The ASD group has more DE in the AS task, longer latency in the MGS task, and fewer predictive saccades in the predictive saccade task. Caldani et al., 2019 [76]21 ADHD (8.1), 21 Control (8.6)DSM-5 (K-SADS)WISC, ADHD-RS (< 10)Children and AdolescentsFixationMobile EBT (300 Hz)The ADHD group made more errors during the fixation task and showed more difficulty in the simple fixation task compared to fixation with distractors. Gould et al., 2001 [77]53 ADHD-C (9.4), 44 healthy (9.8)DSM-IVNAChildren and AdolescentsFixationOber2 infrared orbital scanning system (600 Hz)The number of saccades during fixation did not differ between ADHD and control. ADHD children made larger saccades than healthy controls. Kim et al., 2022 [2]24 ASD (11.6), 27 Control (11.3)DSM-5 (ADI-R, ADOS-G)WISC, CPRS, SRS, SCQChildren and AdolescentsFixation, OddballEyeLink 1000 (500 Hz)Compared to the control group, ASD children have a larger resting pupil diameter during the fixation task and a lower pupil dilation response for target stimuli in the oddball task. Castellanos et al., 2000 [78]32 ADHD (8.8), 20 Control (9.6)DSM-IV (DICA)WISC (> 80), CTRSChildren and AdolescentsSmooth Pursuit eye movements, Go-no-go, Memory-guided taskOber2 infrared orbital scanning system (600 Hz)Girls with ADHD made more commission and intrusion errors in the go-no-go tasks. Li et al., 2003 [14]11 ADHD-C (12.2), 12 Control (11.8)DSM-IV (K-SADS)CBCLChildren and AdolescentsSpatial cueing taskEyelink (250 Hz)The magnitude of inhibition of return in the ADHD group was comparable to the controls. Caldani et al., 2020 [15]31 ADHD (9.1), 30 Control (9.5)DSM-5 (K-SADS)WISC, ADHD-RS, MABCChildren and AdolescentsSpatial cueing taskMobile EBT (300 Hz)The ADHD group has higher saccade error and saccade latency in the endogenous condition, neutral condition, and invalid sub-condition, respectively. Connolly et al., 2016 [79]14 ADHD-C (11.4), Control (9.8)DSM-5WISC (> 80), CRS-R, SRSChildren and AdolescentsSpatial cueing taskEyelink 1000The ADHD group had a larger saccade amplitude and peak velocity. These saccadic measures correlated with Conners’ hyperactivity subscale but not the inattention subscale. Lee et al., 2023 [28]30 ADHD (8), 30 Control (8.1)DSM-5ADHD-RS, CBCLChildren and AdolescentsContinuous performance testHappymind (30 Hz)ADHD children have less fixation on the relevant AOI and higher gaze variability. Redondo et al., 2020 [17]23 ADHD (10.1), 31 Control (10.6)DSM-5WISC (> 85)Children and AdolescentsContinuous performance testAuto Refractometer WAM-5500Pupil diameter and accommodation variability during the execution of the CPT task were correlated with behavioural performance in the ADHD group. Demirdöğen et al., 2022 [80]30 ADHD (10.2), 30 Control (10.3)DSM-5WISC (> 80), CPRS, CTRSChildren and AdolescentsClass-flow video taskSMI RED250 (-)The ADHD group had lower total fixation in the relevant area of interest (e.g., whiteboard) and higher fixation on irrelevant areas than the control group. Levy et al., 2025 [81]24 ADHD (26.7), 25 Control (25.7)NAASRS-v1.1Children and AdolescentsVirtual Reality ClassroomEye tracker embedded in the VR headset (-)The ADHD group had a stronger neural response (larger N1 response) to distractors. Keehn et al., 2023 [40]24 ASD (11.6), 24 Control (11.1)DSM-5 (ADOS-2, ADI-R)WASI, SCQChildren, AdolescentsVisual search taskEyeLink 1000 Plus (500 Hz)Larger resting pupil size in ASD children was associated with superior search performance and longer fixation durations. Hochhauser et al., 2022 [82]44 ADHD (14.6), 40 Control (14.6)DSM-5DBD-RS (> 6 symptoms endorsed as “pretty much”/”very much”)Children and AdolescentsChange detection taskTobii X2-60The time to first fixation, fixation duration, and scan path were intact in ADHD adolescents.Turkan et al., 2016 [83] 24 ADHD (8), 24 Control (8)DSM-IVWISC (> 80), CBCL, T-DSM-IV-SChildren and AdolescentsChange detection taskTobii T60 (60hz)ADHD children have less fixation on changed areas, and their first fixation duration was longer compared to the control group. Bast et al., 2023 [39]31 ASD (13.1), 28 ADHD (13.4), 31 Control (14.3)DSM-5 (K-SADS-PL, ADOS, ADI-R)WAIS (> 70), SRS, SCQ, FBB-ADHDChildren, AdolescentsVisuospatial reaction-time taskTobii X30 eye tracker (30 Hz)ASD symptoms are associated with larger baseline pupil size. Unlike the ADHD group, the ASD group failed to adapt the LC–NE system in response to changes in task utility.Schwerdtfeger et al., 2013 [84]14 ADHD-C (29.5), 14 Control (29.6)DSM-IVNAAdultsProsaccade, AntisaccadeISCAN ETL-400 camera (120 Hz)ADHD participants made more DE and longer and variable SRT during AS trials. There was less activity in FEF, SEF, PEF, dlPFC, and CN in AS preparation. Reilly et al., 2008 [12]24 SCZ (25.8), 30 Control (22.2)DSM-IV (SCID)QT, BPRS, SAPS, SANS, HMAD-24AdultsProsaccade, AntisaccadeInfrared reflection sensors (500 Hz)SCZs were faster in step Pro, made more DE in AS gap and overlap trials, and had a lower overlap effect associated with DE in the AS task. Nigg et al., 2002 [85]22 ADHD (23.1), 21 Control (21.6)DSM-IV (DIS)WISC (> 70), CAARS (> 65)AdultsProsaccade, AntisaccadeISCAN RK-416 eye movement monitor (120 Hz)The ADHD group had more DE than the control group in the AS condition. Ettinger et al., 2018 [86]25 SCZ (37.7), 25 Control (37.4), 27 Relatives (39.7)DSM-IV (SCID)MWT-B, GAF, SPQ, PANSSAdultsProsaccade, AntisaccadeEyeLink 1000There was more AS DE in the SCZ group compared to relatives and the control group, which was correlated with the SPQ disorganized dimension score. Carr et al., 2006 [87]23 ADHD-I, 27 ADHD-C (23.1), 67 Control (24.2)DSM-IV (K-SADS-PL)ADHD-RSAdultsProsaccade, AntisaccadeISCAN ETL-400 (240 Hz)The ADHD group showed greater anticipatory saccades than controls, irrespective of subtype or recovery status. Ma et al., 2025 [88]71 ADHD (22.6), 71 Control (22.2)DSM-5CAARSAdultsProsaccade AntisaccadeMobile EBT (300 Hz)The ADHD group showed more anticipatory saccades in the overlap Pro task, as well as higher direction errors, anticipatory saccades, and longer latencies in the AS task. These measures did not differ between ADHD subtypes. Theleritis et al. 2014 [89] 49 SCZ (23.6), 34 OCD (23.2), 39 Control (11.6) 50 Control (23), 67 Control (48)DSM-IV (M.I.N.I)Y-BOCSAdultsProsaccadeIRIS SCALAR (600 Hz)Modelling saccadic reaction time showed higher variability of the decision signal in the SCZ group compared to the control group, while this parameter did not differ between OCD and the control group. Thomas et al. 2018 [13]54 SCZ (41.39), 137 Control (30.13)DSM-IV (M.I.N.I)WTAR, PNASSAdultsAntisaccadeEyeLink II (500 Hz)The SCZ group had a higher AS error rate, which correlates with attentional and spatial working memory functions. Aichert et al., 2012 [90]504 Healthy (-)NAMWT-B, BIS-11AdultsAntisaccadeEyeLink 1000 (1000 Hz)DE in the AS task was significantly correlated with self-reported impulsivity level.Shmukler et al. 2021 [91]93 SCZ (29.3), 41 Control (28.6)ICD-10 (M.I.N.I, CDSS)SAS, BARS, YMRS, PANSSAdultsAntisaccade, Go/No Go taskSMI RED (500 Hz)The SCZ patients made slower responses and higher errors in AS and Go/No Go tasks compared to the controls. Feifel et al., 2004 [92]12 ADHD (31), 12 Control (33)DSM-IV (SCID)WURS, BADDSAdultsProsaccade, Antisaccade, Fixation, Predictive saccadeEye Track Model 210 (256 Hz)ADHDs generated significantly more anticipatory saccades in the gap Pro task and more DE in the AS task. Shirama et al., 2016 [3]16 ASD (26.3), 16 Control (27.7)DSM-IV-TRWAIS, GSQAdultsAntisaccade, FixationView Point MHU03 (220 Hz)The ASD group had worse fixational stability, while their AS performance was intact. Canu et al., 2022 [93]20 SCZ (19.8), 28 ADHD (19.9), 26 ASD (19.7), 29 Control (19.8)DSM-IV (ADOS, ADI-R)WISC, CAARS, SRS, DBD-RSAdultsProsaccade, Antisaccade, Fixation, MGS, Go/No GoEyeLink 1000 (1000 Hz)Inhibition-associated metrics are impaired in all clinical groups compared to controls and are more pronounced in SCZ than in ADHD and ASD. Obyedkov et al., 2019 [94]156 SCZ (34.4), 42 CHR (clinical high risk) (36.4), 61 Control (21.8)ICD-10SANS, SAPSAdultsProsaccade,Antisaccade,Predictive saccadeICS Chartr 200 VNG/ENGThe SCZ group was less accurate in the Pro task, with higher AS DE. Canu et al., 2022 [19]20 SCZ (19.8), 28 ADHD (19.9), 26 ASD (19.7), 29 health (19.8)DSM-IV (ADOS, ADI-R)WISC, CAARS, SRSAdultsVisual search taskEyeLink 1000, Plus (1000 Hz)Almost all search variables differ between SCZ and the control group. The ADHD group had higher intra-subject variability in the pre-search stage, and the ASD group was impaired in post-search parameters. Elahipanah et al., 2011a [95]32 SCZ (31), 26 Control (29)DSM-IV-TR (SCID)WISC, PNASSAdultsVisual search taskEyeLink 1000The SCZ group had an inflexible and smaller visual span than the control group. Elahipanah et al., 2011b [96]28 SCZ (31.1), 26 Control (29.6)DSM-IV-TR (SCID)WISC, PNASSAdultsVisual search taskEyeLink 1000The proportion of eye fixations and time to first target fixation is intact in the SCZ group in a serial visual search task. Schwab et al., 2015 [97]13 SCZ (33.8), 10 relatives (62.3), 24 Control (37.9)ICD-10MRS, PNASSAdultsVisual search task (Landolt orientation task)iView X HED-MHT (200 Hz)Unlike healthy controls or SCZ relatives, saccadic latencies in the SCZ group didn’t modulate with the task’s changing cognitive demands. Zhang et al., 2021 [98]86 SCZ (47.9), 80 Control (40.4)DSM-IV (SCID)SAPS, SANS, MDRS-2AdultsVisual search (Exploratory Eye Movement test)DEM-2000In the SCZ group, a total of eye scanning length parameters correlated with cognitive impairments, including attention. Poynter et al., 2013 [41]40 Healthy (25.5)NACAARSAdultsFixation, Scan-Identify, Search, StroopTobii TX 300 EyeTracker (300 Hz)The attention subscale of the CAARS correlated with individual fixation duration parameters. Bey et al., 2019 [99]168 OCD (33.3), 93 relatives (46.6), 171 Control (34)DSM-IV (SCID)Y–BOCS, OCI-RAdultsSmooth pursuit eye movementsEyeLink II system (250 Hz)/EyeLink 1000Both standard and predictive smooth pursuit were unaffected in the OCD group compared to healthy controls and relatives. Ross et al., 2000 [100]17 ADHD (37), 49 SCZ (38), 37 Control (37)DSM-IV (SCID)WURSAdultsSmooth pursuit eye movementsInfrared photoelectric limbus (500 Hz)Unlike the ADHD group, the SCZ group’s SPEM gain and the frequencies of anticipatory and leading saccades were impaired compared to the control. Hutton & Tegally, 2005 [101]40 Healthy (21.4).NANAAdultsSmooth pursuit eye movementsSR-Research Eyelink II (500 Hz)Velocity and position error parameters of smooth pursuit tasks significantly change after introducing attentionally demanding tasks. Elbaum et al., 2020 [18]43 ADHD (23.84), 42 Control (23.8)DSM-5 (SCID)ASRS, WURSAdultsCPTEyeLink 1000 (250 Hz)The ADHD group spent more time looking at the distractibility area of interest than the control. Selaskowski et al., 2023 [102]18 ADHD (36.1), 18 Control (25.9)DSM-5 (IDA-R)ADHS-SBAdultsCPT in virtual seminar roomInfrared-based Tobii eyetracker (50 Hz)The ADHD group spent more time fixating on distractors. Saccade duration was correlated with CPT omission errors and gaze wandering. Wiebe et al., 2023 [103]24 ADHD-unmedicated(31.6), 25 ADHD-medicated (31.4), 25 Control (31)DSM-5 (IDA-R)ADHS-SB, DASSAdultsCPT in virtual seminar roomInfrared-based Tobii eyetracker (50 Hz)The unmedicated ADHD group spent more time gazing at distractors and made more omission and commission errors than the control group. Wiebe et al., 2024 [29]43 ADHD (20–62), 43 Control (18–50)DSM-5 (IDA-R)ADHS-SB, DASSAdultsCPT in virtual seminar roomInfrared-based Tobii eyetracker (50 Hz)A model incorporating four eye-tracking features, three CPT features, and one head actigraphy feature classified the ADHD group with 81% accuracy in the independent test dataset. Drescher et al., 2023 [104]40 High-ADHD (21.7), 36 Low-ADHD (22.6)NAASRS (High: >14, Low < 7), DASSAdultsCPTEyeLink 1000 desktop-mount eyetracker (1000 Hz)Pupil metrics did not differ between groups with high and low ADHD symptom levels. Lev et al., 2022 [27]33 ADHD (23.3), 33 Healthy (23.3)DSM-5 (SCID)ASRS, WURSAdultsCPTEyeLink 1000 (250 Hz)The ADHD group spent more time looking at the distractibility area of interest than the control. Vakil et al., 2019 [105]30 ADHD (24.4), 30 Control (23.7)NAWAISAdultsStroop testISCAN Eye Tracking Laboratory (120 Hz)The ADHD group spent more time on the target than on the distracter and had a more pronounced Stroop effect compared to the control group. Roberts et al., 2018 [106]14 ADHD, 14 Control (31)DSM-IV-TR (SCID)CAARSAdultsSpatial cueing taskEyeLink 1000Perceptual effects of endogenous and exogenous attention did not differ between the ADHD group and controls. Adams et al., 2011 [107]30 ADHD (21.1), 27 Control (22)DSM-IVCAARS, AASRS, BISAdultsDelayed oculomotor response taskModel 504 (60 Hz)The ADHD group had higher premature saccades during the task compared to the control group, and this parameter correlates with reaction time variability on the stop-signal task. Armstrong & Munoz 2003b [108]14 ADHD (34.6), 14 Control (34.4)DSM-IVBADDS (> 50)AdultsCountermanding taskEyelink video-based eye-tracker (250 Hz)The ADHD group has lower SRT compared to the control, especially if the stop signal appears at the periphery. Armstrong & Munoz 2003a[109]15 ADHD (29.1), 15 Control (33.1)NABADDS (> 50)AdultsAttentional blinkEyelink SR Research (250 Hz)ADHD adults have more saccades during probe-plus-target blocks, which was correlated with probe errors.Abbreviations: AASRS ADD/H Adolescent Self Report Scale—Short Form, ADI-R Autism Diagnostic Interview Revised, ADOS-G Autism Diagnostic Observation Schedule Generic, ADHS-SB Self-Rating Behavior Questionnaire, ADHD Attention Deficit Hyperactivity Disorder, ASEBA The Achenbach System of Empirically Based Assessment, ASRS Adult ADHD Self-Report Scale, v1.1, AS Antisaccade, BADDS Brown’s Attention Deficit Disorder Scale, BARS Barnes Akathisia Rating Scale, BIS Barratt Impulsivity Scale, BPD Borderline Personality Disorder, BRIEF Behavior Rating Inventory of Executive Function, BPRS Brief Psychiatric Rating Scale, CAARS Conners’ Adult ADHD Rating Scales, CBC Child Behavior Checklist, CDSS Calgary Depression Scale for Schizophrenia, CRS-R Conners Rating Scale-Revised, CPRS-R: S Conners’ Parent Rating Scale Revised Short Form, CTRS Conners’ Teacher Rating Scales, DASS Depression, Anxiety, and Stress Symptoms, DICA Diagnostic Interview for Children and Adolescents, DIGS Diagnostic Interview for Genetics Studies, DIPS Diagnostic Interview for Psychiatric Disorders, DIS Diagnostic Interview Schedule for DSM-IV, DBD-RS Disruptive Behavior Disorders Rating Scale, DSM Diagnostic and Statistical Manual of Mental Disorders, FBB-ADHD German ADHD Rating Scale for Parents, FBB-HKS German ADHD Rating Scale, FEP First-Episode Psychosis, FHR–P Familial High-Risk for Psychosis, GAF Global Assessment of Functioning, GSQ Glasgow Sensory Questionnaire, HAM-D-24 24-item Hamilton Depression Rating Scale, HFA High Functioning Autism, ICD-10 International Classification of Diseases, 10th Revision, IDA-R German Clinical Interview for the Integrated Diagnosis of ADHD in Adulthood, K-SADS-PL Schedule for Affective Disorders and Schizophrenia for School-Age Children–Present and Lifetime, MABC Movement Assessment Battery for Children, MDRS-2 Revised Mattis Dementia Rating Scale, M.I.N.I Mini International Neuropsychiatric Interview, MRS Modified Rogers Scale, MWT-B Mehrfachwahl-Wortschatz-Intelligenztest, NA Not Available, OCI-R Obsessive-Compulsive Inventory-Revised, OCD Obsessive-Compulsive Disorder, Pro Prosaccade, QT Ammon’s Quick Test, RPM Raven’s Progressive Matrices, SANS Schedules for the Assessment of Negative Symptoms, SAPS Schedules for the Assessment of Positive Symptoms, SAS Simpson–Angus Scale, SCID-PD-5 Updated Structured Clinical Diagnostic Interview for Personality Disorders, SCAS Spence Children’s Anxiety Scale, SCQ Social Communication Questionnaire, SCZ Schizophrenia, SPEM Smooth Pursuit Eye Movement, SPQ Schizotypal Personality Questionnaire, SRT saccade reaction time, SRS Social Responsiveness Scale, SNAP Swanson Nolan and Pelham Questionnaire, T-DSM-IV-S Turgay DSM-IV Based Child and Adolescent Behavior Disorders Screening and Rating Scale, UHR–P Individuals with Ultra-High Risk for Psychosis, WASI Wechsler Abbreviated Scale of Intelligence Vocabulary, WAIS Wechsler Adult Intelligence Scale, WAIS-R Wechsler Adult Intelligence Scale–Revised, WISC Wechsler Intelligence Scale for Children, WPPSI Wechsler Preschool and Primary Scale of Intelligence, WTAR Wechsler Test of Adult Reading, WURS Wender Utah Rating Scale, Y-BOCS Yale-Brown Obsessive-Compulsive Scale, YCI Yale Children’s Inventory, YMRS Young Mania Rating Scale, YGTSS Yale Global Tics Severity Scale
The risk of bias for included studies was assessed using a modified Newcastle-Ottawa scale for case-control, cross-sectional, and cohort studies [110]. Each study was assigned a score out of 9, with scores below 5 indicating a high risk of bias. Assessment was conducted by three authors (AT, ZA, AM).
The final sample comprised 75 studies. The studies covered a broad range of psychiatric conditions, including ADHD, ASD, DCD, OCD, BPD, SSD, dyslexia, TS, and non-clinical population with symptom evaluations. Each study was reviewed based on its oculomotor tasks, categorized into one of the attention systems, and its developmental stage. The distribution of studies across attention systems and developmental stages is illustrated in Fig. 3A, and the distribution of covered parameters is illustrated in Fig. 3B. The Risk of Bias table of the included studies is provided in the Supplementary Material. Among 75 included studies, three [65, 77, 101] received a risk of bias score below 5 and should therefore be interpreted with caution.Fig. 3Study Selection (A) Distribution of studies and reports across attention systems in children/adolescents (≤ 18 years) and adults (> 18 years). (B) Number of reports for each oculomotor metric. Abbreviations: ADHD, Attention-Deficit/Hyperactivity Disorder; ASD, Autism Spectrum Disorder; SSD, Schizophrenia Spectrum Disorder; TS, Tourette Syndrome; OCD, Obsessive-Compulsive Disorder; DCD, Developmental Coordination Disorder; BPD, Borderline Personality Disorder
Selective attention in the spatial domain involves directing attention to specific locations in the visual field through involuntary (exogenous) or voluntary (endogenous) processes [1, 111].
Saccade latency is the most frequently reported variable in the prosaccade task. Findings in ADHD children are mixed, with most studies reporting no significant differences in latency compared to controls [7, 14, 15, 54, 55, 57, 59, 61, 71–74, 79]. However, two studies reported shorter latencies [58, 70], while others found longer latencies [63, 65]. Similarly, for other psychiatric groups, results were inconsistent. For example, Ono et al. [54] and Caldani et al. [62] found no differences between ASD, ADHD, and the neurotypical group in prosaccade latency, while Goldberg et al. [75] reported longer latencies in children with high-functioning autism. Studies on TS also showed variability, with one study observing longer latencies [56] and another reporting no significant differences [7]. Children with DCD and adolescents with BPD showed no significant differences in latency compared to controls [9, 10]. The gap effect, measuring reduced saccade latency in gap conditions compared to overlap conditions, was generally intact in ASD children [35, 62, 75]. In ADHD children, the gap effect was preserved in most studies [63, 71], with one exception reporting an attenuated effect [65].
Direction errors in the prosaccade task were comparable between ADHD, ASD, and neurotypical controls in most studies [14, 15, 35, 54, 55, 57, 59, 61, 72]. However, Caldani et al. [15] reported higher error rates in ADHD children during exogenous cueing (neutral and invalid conditions) and endogenous cueing (neutral condition) in the spatial cueing task, and another study observed lower successful prosaccade rates in the ASD group during step conditions but not gap and overlap conditions [54].
Findings for peak velocity were inconsistent. Lower peak velocity was observed in the prosaccade task for ADHD [54] and the TS groups [56], but not in ASD [54], and BPD [9]. Conversely, Connolly et al. [79] reported larger peak velocities in ADHD children, while O’Driscoll et al. [73] found no differences between ADHD and neurotypical controls.
Saccadic accuracy was generally preserved in ADHD, TS, and DCD groups [10, 54, 56, 63, 72], though one study reported reduced accuracy in ASD children [35], and one study reported hypometric saccade in neurodevelopmental disorders, including ADHD, DCD, and dyslexia [66]. Lastly, anticipatory saccades were higher in ASD children [35, 62], and BPD adolescents with comorbid ADHD [9].
In adult populations, prosaccade latency was generally preserved across clinical groups, with no significant differences observed for schizophrenia (SCZ) [12, 86, 89, 93] but not [94], ADHD [84, 85, 87, 92, 93, 106], ASD [35, 93], and healthy controls. However, intra-subject variability (ISV) in prosaccade latency, which reflects trial-to-trial consistency, was elevated in SCZ patients compared to controls [89, 93]. In a study comparing prosaccade ISV across SCZ, ASD, ADHD, and healthy adults, SCZ showed the highest ISV, significantly more than ASD and controls, while ADHD fell between SCZ, ASD, and TD groups without significant differences from either [93]. The prosaccade gap and overlap effect were intact in both ASD [35, 93] and ADHD adults [93]. However, increased anticipatory saccades were observed in SCZ [93] and ADHD adults [85, 87, 88, 92], though one study found no significant differences in ADHD [93]. In ASD, anticipatory saccades did not differ significantly from controls [93]. The prosaccade accuracy was largely preserved in ADHD [85, 92] but impaired in SCZ compared to controls [94].
In children and adolescents, prosaccade latency shows mixed findings across ADHD, ASD, TS, and SCZ, whereas in adults, latency is generally comparable to neurotypical controls. The gap and overlap effects remain relatively preserved across clinical groups. However, SCZ adults consistently exhibit greater ISV in prosaccade latency. Prosaccade direction error rates are generally intact in ADHD and ASD. Peak velocity findings lack a clear pattern across studies. Anticipatory saccades and saccade accuracy are impaired in ASD children but not in ASD adults, while both ADHD and SCZ adults demonstrate elevated anticipatory saccade rates. Lastly, the gap effect in the ASD group and anticipatory saccade in the ADHD and control group correlate with the inattention symptoms (See Table 5).Table 5Eye-tracking parameters correlated with Self-report or behavioral measures of attentionStudyAnalysis (Sample size)Oculomotor taskOculomotor parameterSelf-report or behavioral measures of Attentionr**P(value) Kim et al., 2022 [2]ASD (24) + NT (27)FixationPupil diameterInattention score of C3-P/S0.310.03 Poynter et al. 2013 [41]NT (40)FixationFactor scores (fixation rate, size, duration, saccade amplitude, microsaccade rate)T-scores on the attention subscale of CAARS0.360.03Visual Search0.390.02 Elbaum et al., 2020 [18]ADHD (43) + NT (42)CPTFixation duration on task-relevant areasCorrect responses of CPT (Attention index)0.37< 0.01Fixation duration on distractor areasCorrect responses of CPT (Attention index)0.36< 0.01 Redondo, et al. 2020 [17]ADHD (23) + NT (31)CPTPupil diameterOmission errors of CPT0.36< 0.05 Selaskowski et al. 2023 [102]ADHD (18)CPT in virtual realitySaccade durationOmission errors of CPT0.25–0.5< 0.01 Demirdöğen et al., 2022 [80]ADHD (30)Class-flow video taskFixation duration on task-relevant areasInattention/Cognitive score of (CPRS/CTRS-R)−0.43< 0.05 Zhang et al., 2021 [98]SCZ (37 DS, 49 NDS)Exploratory eye movementFrequency of eye fixation in the DS groupContinuous vigilance/attention of MDRS-2−0.54< 0.001Total eye scanning lengthInattention score of SANS0.33< 0.05 Amestoy et al., 2021 [35]ASD (97)ProsaccadeGap effectInattention score of ADHD-RS IV0.270.03 Nigg et al., 2002 [85]ADHD (22), Healthy (21)ProsaccadeAnticipatory saccadeNumber of endorsed DSM-IV inattention symptoms0.430.003 Thomas et al., 2019 [13]SCZ (54)AntisaccadeDirectional errorCPT−0.430.004Abbreviations: ADHD Attention Deficit Hyperactivity Disorder, ADHD-RS IV Attention-Deficit/Hyperactivity Disorder Rating Scale, ASD Autism Spectrum Disorder, CAARS Conners’ Adult ADHD Rating Scales, C3-P/S Conners 3 Parent Short Form, CPRS/CTRS-R Conners’ Parent and Teacher Rating Scale–Revised Short Forms, CPT Continuous Performance Test, DSM-IV Diagnostic and Statistical Manual of Mental Disorders, 4th Edition, DS Deficit Schizophrenia, MDRS-2 Revised Mattis Dementia Rating Scale, NDS Non-Deficit Schizophrenia, NT Neurotypical, SANS Scale for the Assessment of Negative Symptoms, SCZ Schizophrenia
Selective attention in the feature domain involves allocating attention to specific features, such as color, orientation, or shape, to identify targets among distractors [1].
Feature-based attention in children and adolescents with ADHD has been examined using change detection tasks. Hochhauser et al. [82] reported intact fixation metrics, including fixation duration, fixation counts, and latency to first fixation, in ADHD adolescents compared to healthy controls. However, another study found that children with ADHD exhibited shorter fixation durations, particularly for marginal changes, and lower latency to first fixation compared to controls [83]. Despite conflicting findings on fixation metrics, both studies reported inefficient search patterns in ADHD children. Hochhauser et al. [82] observed increased gaze dispersion, while Turkan et al. [83] noted faster oculomotor responses but reduced detection accuracy in ADHD children. One study on visual search in ASD children reported that the ASD group demonstrated significantly faster and more efficient search performance, driven by fewer fixations rather than shorter fixation durations during search trials [40].
Canu et al. [19] examined oculomotor metrics during a visual search task in ASD, SCZ, and ADHD adults. SCZ patients showed the most impairment, deviating significantly from controls in almost all visual search metrics. ADHD participants exhibited variable latency of the first saccade and fixation duration during the search phase, but their performance was similar to controls in other metrics impaired in SCZ. ASD participants were the fastest to locate the target, with less variability in search and scanning times compared to ADHD, SCZ, and control groups. However, they were impaired in the initiation and post-search phases, showing slower and more variable latency to the first saccade and prolonged post-search duration. Across all three clinical groups, shared impairments were observed in search-phase metrics, including longer and more variable fixation durations, and in search initiation, with greater variability in latency to the first saccade.
In another study, Elahipanah et al. [95] used a visual search task requiring participants to locate a red/blue cross among distractors with varying target-to-distractor ratios. SCZ participants demonstrated intact peripheral detection, as measured by the proportion of eye fixations and time to first target fixation. In another study from this team, they conducted a visual search experiment with varying task difficulty to examine visual span size and the modulation of visual span as a function of task difficulty in SCZ patients [96]. In the difficult discrimination condition, distractors and targets were highly similar in their shapes (named similarity task), or distractors were unfamiliar shapes (named familiarity task), and in the low demand condition, distractors and targets were less similar, or distractors were familiar. They found that SCZ participants exhibited slower searches across all conditions, smaller visual spans during low-similarity conditions, and reduced flexibility in visual span modulation based on task difficulty. These results suggest impaired attentional resources in SCZ. These findings are further supported by Zhang et al. [98], who reported fewer fixations and shorter eye scanning lengths in an exploratory eye movement task, where participants were asked to find differences between three “S” shaped figures. Moreover, the eye scanning length in SCZ patients correlated with inattention and memory scores on the Revised Mattis Dementia Rating Scale [98].
Schwab et al. [97] investigated the modulation of saccadic latency by task demand in SCZ patients and their relatives. Participants were asked to determine whether two objects were identical in terms of color (low demand) or orientation (high demand). SCZ patients showed reduced saccadic latency differences between tasks compared to controls and relatives, which was interpreted as a deficit in top-down saccade generation [97]. In another study, saccades were measured during an attentional blink task, where participants attended to a rapidly presented list in two blocks [112]. ADHD adults exhibited attentional blink deficits, with higher saccade rates during probe-plus-target trials correlating with increased probe errors. The authors concluded that gaze instability during task execution contributed to poorer performance in attention-demanding tasks [112].
Feature-based attention studies in children and adolescents with ADHD show conflicting results for fixation metrics but consistently highlight inefficient search patterns. In adults, SCZ tends to exhibit the most pronounced impairments, with longer or more variable fixation durations and reduced flexibility when task difficulty increases; ADHD adults mostly show impairments in search initiation (e.g., longer or more variable first‐fixation latencies) and fixation duration, whereas ASD children and adults are faster in locating targets but share difficulties in the initiation and post-search phases. Together, there is a consistent pattern of search-phase impairments across clinical groups, particularly related to fixation duration and latency to first fixation.
Sustained attention is the ability to maintain focus on a task over an extended period without significant declines in performance.
An increase in intrusive saccades during fixation tasks in children with ADHD compared to age-matched controls was observed in five out of six studies [70, 72, 74, 76, 77], but not in [71]. One study reported that ADHD children made fewer errors during fixation tasks with distractors compared to simple fixation tasks [76], and another reported no differences between ADHD subtypes (inattentive and hyperactive) in the proportion of errors during fixation tasks [72].
By analyzing participants’ visual fields across areas of interest (AOIs), researchers compared the allocation of attention to task-relevant and task-irrelevant areas. One study evaluating ADHD children’s performance during a classroom video task and another assessing ASD children during fixation tasks found reduced fixation duration in task-relevant areas for both groups compared to controls [2, 80]. Another study integrated a CPT task with eye tracking to evaluate its predictive utility for ADHD symptoms [28]. Results showed that ADHD children had shorter fixation durations and greater gaze variability in task-relevant areas than controls [28]. Combining eye-tracking metrics with behavioral parameters improved sensitivity and specificity in identifying ADHD children (from 0.76 AUC to 0.88 AUC) [28]. A recent study retrieved after our initial search explored the neurophysiological mechanisms underlying this aberrant attention allocation [81]. It showed that children with ADHD had heightened neural responses to irrelevant sounds in a virtual reality classroom, despite showing comparable behavioral performance and fixation durations to task-relevant areas (e.g., the teacher) [81].
Five studies examined pupil diameter during visual attention tasks. One study found that children with ASD exhibited larger resting pupil diameters during fixation tasks [2]. In children with psychosis, phasic pupil dilation was reduced during pro/antisaccade tasks [59]. However, other studies reported no significant differences in pupil dilation between ADHD and the neurotypical group during CPT [17] or pro/antisaccade tasks [59]. These studies demonstrated that pupil dilation metrics correlated with attention functions measured by behavioral tasks [17] and self-report questionnaires [2]. Two additional studies explored the effects of phasic pupil dilation in response to auditory warning cues before prosaccade. While no significant relationship between pupil dilation and ADHD symptoms was found, phasic alerting cues reduced saccadic latencies in the ADHD group, with the degree of reduction predicting ADHD symptom levels after a two-year follow-up [63, 64]. Two studies retrieved in the updated search examined tonic and phasic pupil size in children with ADHD and ASD [39, 40]. Both found that baseline pupil size was larger in children with ASD [39, 40]. Moreover, unlike children with ADHD, those with ASD failed to adjust LC–NE activity in response to changes in task difficulty [39].
Lastly, two studies with ADHD children reported no significant differences in smooth pursuit eye movement (SPEM) parameters, including Catch-up saccades and saccadic gain, compared to controls [74, 78].
Three studies examined fixation tasks in clinical and control populations. Feifel et al. [92] found no differences in intrusive saccades during fixation tasks with distractors between ADHD adults and controls. However, Canu et al. [93] reported more intrusive saccades in SCZ and ASD participants (but not ADHD) in the distractor direction, with no differences observed in blocks without distractors. Shirama et al. [3] investigated fixational stability using metrics such as the number of saccades and gaze dispersion from its mean position during fixation tasks with or without a fixational target. While fixational stability was intact in the ASD group when a fixational target was present, they had difficulty maintaining gaze in tasks without a fixation target. One study performed factor analysis on various oculomotor metrics (e.g., fixation rate, fixation duration, fixation size) during a sustained fixation task [41]. Results showed a moderate correlation between factor scores and attention deficit T-scores from the Conners’ Adult ADHD Rating Scales (CAARS) in the normative population [41].
The relationship between SPEM metrics and attention in normative adult populations was explored by Hutton and Tegally [101]. They found that non-spatial tone discrimination tasks disrupted smooth pursuit by reducing velocity gain and increasing position error. Attention deficits associated SPEM changes, including velocity gain and anticipatory saccades, varied across ADHD, OCD, and SCZ groups. One study found impaired SPEM velocity gain in SCZ but not in ADHD adults compared to controls [100], while another study involving 168 OCD patients and their relatives reported no impairment in velocity gain during standard and predictive SPEM tasks [99].
Lastly, six studies combined conventional CPT metrics with eye-tracking instruments, three of them in virtual reality environments, to assess their potential in improving group prediction for ADHD adults [18, 27, 29, 102–104]. These tasks included blocks with and without distractors (auditory and/or visual). Higher fixation duration to distractors is repeatedly reported in ADHD adults [18, 27, 29, 102, 103], correlating with CPT behavioral measures [18, 27]. One study further reports that ADHD symptom levels did not relate to pupil indices [104].
In children and adolescents with ADHD, increased intrusive saccades and higher fixation on task-irrelevant areas are relatively common, whereas in adults with ADHD, fixation deficits persist, but intrusive saccade findings are less consistent. Unlike the ADHD population, the ASD and SCZ groups show more consistent impairments in pupil diameter. SCZ is more reliably associated with impaired smooth pursuit metrics, including velocity gain and anticipatory saccades, while these parameters are generally intact in ADHD and OCD groups. Lastly, fixation duration on task-relevant/irrelevant areas and pupil diameter correlate with attention scores from both behavioral tasks and self-report measures in clinical and non-clinical populations (See Table 5).
Executive control (response inhibition) involves the ability to suppress reflexive or prepotent responses.
The proportion of directional errors or percentage of correct trials is the most frequently reported variable in response inhibition tasks. In ADHD studies, the results varied based on experimental design. Almost all studies using the antisaccade gap task found significantly more direction error in ADHD participants compared to controls [7, 55, 57, 58, 61, 67, 70, 72, 74]. However, results from the step and overlap tasks were mixed, with most studies reporting no significant group differences in direction error [54, 59, 69, 71, 73]. Studies on ASD children also showed inconsistent results, with two studies reporting higher direction error [35, 75] and others finding no differences compared to controls [54, 62]. In the TS group, fewer direction error were observed compared to both TS + ADHD participants [7] and healthy controls [56], while BPD adolescents with comorbid ADHD showed higher direction errors compared to controls [9]. Direction errors were also linked to symptoms of disruptive behavior disorders [68] and executive function measures [35] but showed no significant correlation with ADHD traits, such as inattention or hyperactivity/impulsivity [67]. However, a new study retrieved after an updated search reports no correlation between direction error and the Barratt Impulsiveness Scale in the BPD and BPD/ADHD group [9].
Antisaccade latency findings were similarly mixed. Six studies with ADHD children found no significant differences compared to controls [54, 55, 69–71, 74]. However, four studies reported longer latencies for ADHD participants [57, 60, 61, 79]. Results for the ASD group were similarly inconsistent, with reports of intact [62], prolonged [75], and shorter [35] antisaccade latencies compared to controls. The TS group generally demonstrated intact or shorter antisaccade latencies compared to controls [7, 56]. Accuracy findings were mixed, with most studies reporting no significant differences in antisaccade gain for ADHD children compared to controls [54, 58, 72], but not [60].
Apart from the Antisaccade task, the countermanding, Go/No-Go, and delayed oculomotor tasks are key paradigms that integrate eye-tracking to measure inhibitory control. The Countermanding task revealed elevated error rates in ADHD children [71]. ADHD girls also showed higher commission errors in the Go/No-Go task [78].
Antisaccade direction errors were consistently higher in SCZ [12, 13, 86, 91, 93, 94] and ADHD adults [84, 85, 87, 88, 92, 93] compared to controls, regardless of gap or overlap manipulations. In ASD, findings were one study found no significant difference [3], while others reported elevated direction error compared to controls [35, 93]. In SCZ, the overlap effect negatively correlated with direction error, suggesting the contribution of diminished attentional engagement to error rates [12]. Among clinical groups, SCZ patients exhibited the most pronounced deficits, with the highest error rates [86, 93]. A head-to-head comparison of SCZ, ASD, and ADHD groups showed that SCZ patients made significantly more directional errors than ASD but not ADHD [93].
Additionally, a study on SCZ subgroups and the clinical high-risk (CHR) group for psychosis found that CHR participants also showed elevated directional error compared to controls. Within SCZ subgroups, those with predominantly disorganization symptoms had the highest error rates, exceeding those with predominantly negative or positive symptoms [94]. In a large normative sample (n = 504), Aichert et al. [90] demonstrated a strong correlation between the Barratt Impulsiveness Scale and antisaccade errors. Additionally, Thomas et al. [13] reported that antisaccade errors in SCZ were significantly correlated with attentional processes (measured via Continuous Performance Tests) and working memory scores (from the Wechsler Memory Scale).
Findings on antisaccade latency were mixed across clinical groups. Prolonged latencies were reported for SCZ [91, 93] and ADHD adults [84, 87, 88], while other studies found no significant differences for SCZ [12, 86] or ADHD adults [85, 92, 93] compared to controls. These discrepancies were not clearly related to fixation manipulations (e.g., gap/overlap, step paradigms) or whether pro/antisaccades were interleaved or presented in block designs. The antisaccade gap effect remained intact in ADHD adults [92, 93] but was not specifically examined for other psychiatric conditions. Antisaccade latencies in ASD participants were generally intact across studies [3, 35, 93].
Antisaccade accuracy deficits were primarily reported in SCZ participants, who showed hypometric saccades compared to controls [86]. Accuracy was generally preserved in ADHD and ASD adults [3, 87, 92]. Antisaccade ISV was significantly elevated in SCZ participants [89, 93], but was intact in ADHD and ASD groups [93]. Anticipatory saccades were higher in SCZ participants [93]. For ADHD, results were two studies reported increased anticipatory saccades [87, 88], while another found intact anticipatory saccades [93].
Vakil et al. [105] integrated a Stroop test with eye-tracking to investigate attentional control in ADHD adults. Metrics included fixation duration on targets and distractors, number of fixations, latency to first fixation on the target, and shifts between targets and distractors. While both ADHD and control groups showed similar increases in fixation duration between congruent and incongruent conditions, ADHD participants had longer and more frequent fixations on targets and distractors and delayed time to first fixation on targets compared to controls. Furthermore, one study using a delayed oculomotor task showed that ADHD participants exhibited elevated premature saccades, which correlated with reaction time variability in the Stop Signal task [107]. Another study employed a modified countermanding task that included central and peripheral stop signals. While central stop signals supported oculomotor countermanding, peripheral stop signals required greater inhibitory control. ADHD adults were less accurate and slower than controls only when the stop signal was peripheral rather than central [109]. Lastly, two studies investigated the Go/No-Go task in ADHD, ASD, and SCZ adults. Commission errors were significantly higher in SCZ and ADHD adults but not in ASD participants [91, 93]. Other impairments, such as saccade latency and omission errors, were observed only in SCZ patients [91, 93].
In children and adolescents, ADHD exhibits elevated directional errors under gap-based antisaccade tasks but not in overlap or step conditions; among adults with ADHD, these errors remain elevated regardless of fixation manipulations. ASD findings are mixed across both youth and adult groups, whereas SCZ consistently emerges as the most impaired group in antisaccade tasks, characterized by high error rates, intra‐subject variability, and reduced saccade accuracy. On Go/No‐Go and countermanding tasks, both SCZ and ADHD tend to show more commission errors, while ASD generally presents fewer inhibitory deficits.
This study aimed to map oculomotor paradigms to attention systems and evaluate their implications across psychiatric conditions using eye-tracking technology, focusing on three dimensions of selective (feature/spatial), sustained, and executive control. This review included studies involving children, adolescents, and adults with various psychiatric conditions. This approach provides a comprehensive overview of how attention dysfunction can be studied through oculomotor metrics and highlights shared neurobehavioral dysfunction across psychiatric conditions.
The most frequently reported oculomotor metric in our review, implicated in both neurodevelopmental and schizophrenia spectrum disorders, was antisaccade direction errors. Four previous meta-analyses on ADHD children, adolescents, and adults have consistently reported higher antisaccade direction errors with moderate effect sizes [45, 46, 48, 113]. Other meta-analyses also provide evidence for elevated direction errors in OCD, individuals at familial high risk for psychosis, and first-episode psychosis [47, 114]. A recent large meta-analysis of 146 studies with 13,807 participants further showed that impairment in this metric was preserved across psychiatric conditions [48] (Table 6). This trend of higher direction errors supports its potential as a transdiagnostic biomarker, although the degree of impairment and the implicated neurobehavioral circuits vary across psychiatric conditions. Studies comparing antisaccade direction errors across adults with SCZ, ASD, and ADHD show that SCZ patients have the highest direction error, significantly more than ASD patients [93]. Additionally, the degree of impairment appears to depend on developmental stage. ADHD adults consistently show elevated direction error, regardless of fixation manipulation, while ADHD children exhibit higher direction error primarily under gap conditions. Complementary, fMRI studies have shown that ADHD children exhibit higher activation in the dlPFC before antisaccade execution compared to controls [55, 69]. In contrast, ADHD adults show hypoactivation in the dlPFC, FEF, SEF, and posterior eye fields during the antisaccade preparatory phase [84]. These findings suggest that while both children and adults with ADHD experience higher direction errors due to inadequate preparatory mechanisms, the underlying cause may differ.Table 6Meta-analysis of oculomotor studies in different psychiatric conditionsMeta-analysisClinical groupDevelopmental StageOculomotor taskImpaired oculomotor parameter (Sample size)Model (Effect size) Maron et al., 2021 [46]ADHDChildren, Adolescents, Adults (no age limit)ProsaccadeAntisaccadeFixationSPEMMGSIntrusive saccade (114 ADHD, 105 control)RA (Hedges’ 1.37)Antisaccadedirection errors (150 ADHD, 232 control)RA (Hedges’ 0.65) Sherigar et al., 2023 [45] ADHDChildren, Adolescents (< 18)ProsaccadeAntisaccadeFixationSPEMMGSProsaccade Latency (75 ADHD, 76 control)RA (mean −27.38)Antisaccade direction errors (199 ADHD, 191 control)RA (mean 15.36) Chamorro et al., 2022 [113] ADHDChildren, Adolescents, Adults (no age limit)ProsaccadeAntisaccadeFixationMGSCountermanding TaskIntrusive saccade (325 ADHD + control)RA (Hedges’ 1.11)Antisaccade direction errors (739 ADHD + control)RA (Hedges’ 0.79)Johnson et al., 2016 [6]ASDChildren, Adolescents, Adults (no age limit)ProsaccadeAntisaccadeFixationSmooth pursuit eye movementStandard deviation of saccade gain (93 ASD and 147 control)RA (Hedges’ 0.87)RA (Hedges’ 1.71)/ RA (Hedges’ 1.35)/ RA (Hedges’ 1.87)Antisaccade direction errors (161 ASD, 182 control) Open loop gain in pursuit eye movement (101 ASD, 140 control) Closed loop gain in pursuit eye movement (114 ASD, 154 control) O’Driscoll & Callahan 2008 [115]SCZAdultsSmooth pursuit eye movementMaintenance gain (1547 SCZ, 1471 control)Cohen’s d (−0.87)Leading saccade rate (291 SCZ, 317 control)Cohen’s d (1.31) Ekin et al., 2024 [47]FEP, UHR–P, FHR–PAdultsAntisaccadeDirection errors in FEP (440 FEP, 436 control)RA (Hedges’ 1.16)Direction errors in FHR–P (212 FHR, 231 control)RA (Hedges’ 0.58) Bey et al., 2018 [114]OCDAdultsAntisaccadeDirection errors (189 OCD, 204 control)RA (Hedges’ 0.48) Breuer et al., 2024 [48]SCZ, ASD, ADHD, MDD, OCD, SUD, ED, PD, AD, BPAdultsAntisaccadeDirection errors (337 ADHD, 68 ASD, 509 BD, 91 ED, 362 MDD, 473 OCD, 194 SUD, 4993 SCZ)RA (Hedges’ ADHD: 0.74, ASD: 0.76, BD: 0.75, ED: 0.48, MDD: 0.56, OCD: 0.46, SUD: 0.58, SCZ: 1.14)Latency (315 ADHD, 257 BD, 466 OCD, 3687 SCZ)RA (Hedges’ ADHD: 0.47, 0.61, OCD: 0.42, SCZ: 0.68)Abbreviations: AD Anxiety Disorder, ADHD Attention Deficit Hyperactivity Disorder, ASD Autism Spectrum Disorder, BD Bipolar Disorder, ED Eating Disorder, FEP First-Episode Psychosis, FHR–P Familial High Risk for Psychosis, MDD Major Depressive Disorder, MGS Memory-Guided Saccade, OCD Obsessive-Compulsive Disorder, PD Personality Disorder, RA Random Effects, SCZ Schizophrenia, SPEM Smooth Pursuit Eye Movement, SUD Substance-Use Disorder, UHR–P Ultra-High Risk for Psychosis
The correlation between antisaccade direction errors and self-reported or behavioral clinical measures has been inconsistent across psychiatric conditions. While studies have shown that higher direction errors are associated with impulsivity in ASD [35], and with schizotypy, attention, and working memory impairments in SCZ [12, 13, 86], Sanchez et al. [67] found that neither inattention nor hyperactivity/impulsivity traits in ADHD adults could predict antisaccade direction errors. Furthermore, biometric modelling studies of twins and their parents suggest that variance in antisaccade responses is attributable to genetic and non-shared environmental factors [116]. Together, antisaccade errors might be a potential endophenotype for sub-typing or sub-grouping mental health conditions that share similar neurobehavioral dysfunctions, or ‘biotypes’. However, this does not speak to the uniqueness of this metric for inattention. Instead, there seems to be an intertwined relationship between direction error and inhibition, attention, and spatial working memory functions in normative and psychiatric populations [47].
Aberrant fixation duration and latency to first fixation are also commonly observed across neurodevelopmental and schizophrenia spectrum disorders. These metrics, especially fixation duration on task elements (denoted as AOI) and latency to first fixation, are frequently used in tasks requiring participants to detect a target among non-target distractors. These tasks span across sustained attention paradigms such as the CPT, feature-based attention such as visual exploration, and executive control, including the Stroop test and the Go/No-Go task.
Latency to first fixation in the search initiation phase can be influenced by salient stimulus features, which require an exogenous shift of attention, or by top-down attention, especially in more difficult task conditions (e.g., high similarity between the target and distractors). Longer and more variable times to first fixation have primarily been reported in SCZ and ASD, and to a lesser extent in ADHD. However, basic oculomotor metrics, such as prosaccade latency, were found to be comparable across these conditions and the normative population, indicating that initial shifts of attention and basic oculomotor circuitry are intact. Instead, the impairment appears to be in the top-down control of attention, especially before shifts in challenging task conditions, which are linked to difficulties in search initiation [19].
On the other hand, fixation duration at task elements is influenced by multiple factors, including task complexity, stimulus size, individual interest, and top-down attention [117]. Among various oculomotor metrics, fixation duration has been frequently associated with self-reported and behavioral measures of inattention (Table 5). For example, inattention scores on the Conners Rating Scale have been shown to correlate with fixation duration on task-relevant areas in both normative and psychiatric conditions. In particular, fixation duration in a visual search task correlates with inattention scores in the normative population [41], in ADHD children during class-flow video tasks [80], and in SCZ patients during the exploratory eye movement task [98]. Additionally, fixation duration on both target and distractor areas correlates with behavioral measures of attention, including correct responses of the CPT in ADHD and controls [18]. Although the strength of the correlation is typically weak to moderate (ranging from 0.3 to 0.5) (Table 5), it underscores the utility of fixation duration as a potential transdiagnostic feature for predicting inattention symptoms. Notably, ADHD children, adolescents, and adults tend to show less fixation on targets in visual search [19], CPT [18, 27, 29, 102, 103], and Stroop [105]. Integrating eye-tracking fixation metrics with behavioral measures from the CPT has improved the prediction of psychopathological symptoms in children and adults with ADHD [18, 27–29, 88].
Increased anticipatory saccades in Gap pro/antisaccade paradigms are also shared across psychiatric conditions and have been associated with inattention symptoms. Elevated anticipatory saccades have been reported in ASD children [35, 62], ADHD [85, 87, 88, 92], and SCZ adults [93]. Anticipatory saccades in the Gap prosaccade task correlate with endorsed inattention symptoms of the DSM-IV in both ADHD and healthy controls [85]. Furthermore, one study demonstrated a phenotypic association between anticipatory eye movements in the antisaccade task and parent-reported inattentive behaviors [67]. This study also reported a moderate genetic correlation between inattention and the proportion of anticipatory eye movements [67]. Physiologically, fixation and saccade cells located in the SC, FEF, and dlPFC have been linked to anticipatory saccades [92, 118]. These regions are part of the frontoparietal and dorsal attentional networks, which neuroimaging studies have shown to be altered in ADHD, ASD, and SSD [5, 119, 120].
Intrusive saccades during fixation represent another prevalent oculomotor metric associated with poor sustained attention. Two meta-analyses reported significantly higher intrusive saccades in ADHD groups during simple fixation tasks (without distractors), with large effect sizes [46, 113]. Within our included studies, elevated intrusive saccades were more pronounced in ADHD children than in adults. Notably, ADHD patients exhibited higher intrusive saccades in simple fixation tasks but performed comparably to controls in tasks with distractors. This suggests that in the absence of competing stimuli, individuals with ADHD struggle more to maintain fixation, further supporting the idea that sustained attention, rather than selective attention, is particularly impaired in the ADHD population [106]. However, we have found few studies examining this metric in other psychiatric conditions.
Other oculomotor parameters yielded mixed results across psychiatric conditions. For instance, resting pupil diameter appears intact in ADHD but is enlarged in children with ASD. SPEM metrics, such as gain and velocity, are generally preserved in ADHD and OCD populations, but are consistently impaired in SCZ patients [115]. Basic oculomotor metrics, including saccadic latency, velocity, and gain in pro/antisaccade paradigms, also show variability. While prosaccade task metrics (e.g., latency, accuracy) remain largely intact across ADHD, ASD, SCZ, TS, BPD and DCD populations, antisaccade task performance is more inconsistent. These findings suggest that endogenous (voluntary) shifts of attention, rather than exogenous (reflexive) ones, are broadly impaired to varying degrees in psychiatric conditions and may better align with psychopathological symptoms.
Together, this study aimed to map oculomotor paradigms and metrics to key attention systems and highlight their potential for identifying neurobehavioral biomarkers of inattention. Our findings suggest that impairments in fixation duration, latency to first fixation, direction errors, anticipatory saccades, and intrusive saccades are shared across diverse psychiatric conditions. However, these metrics exhibit only moderate correlations (approximately r = 0.3–0.5) with inattentive symptoms, and few studies have validated such associations in normative populations. Therefore, the current evidence is insufficient to conclude that these metrics constitute clinically applicable biomarkers. Future longitudinal studies in healthy populations are needed to further clarify which oculomotor metrics are associated with the development of attention networks and how these relate to self-reported measures of inattention.
Moreover, the relationship between oculomotor metrics within each paradigm and the corresponding neurobehavioral dysfunctions in psychiatric conditions remains poorly understood. For instance, elevated antisaccade direction errors may arise from poor inhibitory control over prepotent responses, attention allocation to cue and target, or reduced working memory capacity for holding task instructions. These mechanisms may account for the heterogeneous correlations observed between this metric and both self-reported and behavioral measures of cognitive processes. To address such discrepancies, future studies are encouraged to complement basic oculomotor indices (e.g., latency and direction errors) with measures of attentional distribution (e.g., fixation duration on task elements) and assessments of individual differences in working memory capacity.
This study holds two potential implications for the RDoC framework. First, inattention symptoms may be viewed as a dimensional, heterogeneous trait distributed continuously in the population, which can be studied using various paradigms that target different visual attention systems. Mapping these paradigms to key attention systems within the RDoC constructs could advance our understanding of the neurophysiology of inattention by linking these traits to specific neural circuits [20, 121]. Second, among the included studies, there is consistent evidence that integrating behavioral metrics with eye-tracking parameters improves ADHD diagnostic precision by increasing both sensitivity and specificity [18, 27–29, 88]. Notably, eye-tracking metrics have been shown to contribute more significantly to model prediction than behavioral and EEG data features [29, 88]. Our study may contribute to this line of research by highlighting several oculomotor metrics that are shared across psychiatric conditions and associated with inattention symptoms. Future eye tracking studies using a dimensional approach are further needed to elucidate the utility of attention-related eye tracking parameters in sub-typing psychiatric conditions based on attention problems.
This systematic review has important limitations. First, the conclusions are drawn from an overview of findings and the use of crude summary statistics, which should be interpreted with caution. Second, attention-related oculomotor impairments are also relevant to conditions not covered in our review, such as major depressive disorder [8, 122], bipolar disorder [123], and anxiety disorder [48]. Their absence may reflect our search strategy’s focus on paradigms excluding valenced stimuli (e.g., Face). Lastly, there was a clear bias in our study pool toward the ADHD diagnostic category, likely due to the inclusion of the “inattention” related keywords. This is not surprising, given that attention impairments are historically inherent in the ADHD diagnostic criteria and reflected in its name. However, this left other psychiatric conditions with few studies, which might compromise our aim to identify transdiagnostic oculomotor features.
Basic oculomotor metrics in paradigms related to the spatial domain of selective attention (e.g., Prosaccade) were largely intact across psychiatric conditions. However, results from paradigms assessing sustained attention and executive control showed mixed outcomes within and between psychiatric conditions, depending on oculomotor metrics, task manipulations, and developmental stage. Impairments in fixation duration, latency to first fixation, direction errors, anticipatory saccades, and intrusive saccades are shared across diverse psychiatric conditions and moderately correlate with inattention symptoms.
Supplementary Material 1.