The review by Zhou and Zhao of the Department of Rehabilitation at Children’s Hospital of Nanjing Medical University focused on children and adolescents. Its authors searched PubMed/MEDLINE, Web of Science, Scopus and PsycINFO for English-language research published from database inception through June 2026. Eligible evidence included systematic reviews, meta-analyses, randomized and prospective studies, diagnostic-accuracy studies, cross-sectional investigations and methodological publications. Because the objective was a conceptually structured narrative synthesis rather than a formal systematic review, the authors did not conduct quantitative pooling or a formal risk-of-bias assessment. They compared paradigms by cognitive target, task design, psychometric properties, ecological resemblance, clinical application and evidence maturity.

Continuous performance tests began as vigilance tasks in which participants monitored changing stimuli for infrequent targets over an extended period. Modern versions commonly assess sustained attention, response inhibition and response consistency. Omission errors, meaning failures to respond to targets, are commonly interpreted as lapses in task engagement, while commission errors, or inappropriate responses to non-targets, may reflect reduced inhibition or a liberal response strategy. Mean reaction time describes average speed, whereas reaction-time variability captures fluctuations from trial to trial. That variability was among the most consistently replicated group-level findings in ADHD, appearing in a meta-analysis spanning 319 studies, although the measures used to calculate it differ across platforms and are not numerically interchangeable.

The state-regulation account proposed in the review helps explain why group differences can be statistically clear while individual diagnostic accuracy remains modest. In this framework, ADHD can involve unstable allocation of attention and effort rather than a uniformly reduced capacity to concentrate. Performance may rise and fall with arousal, motivation, incentive, fatigue or other changing states. A large study can detect an average difference between groups, but a single testing session samples only one instance of that variable process. The same distinction is important when interpreting omissions, speed and variability: an unusually variable or slow performance cannot automatically be read as a stable, ADHD-specific deficit. The review instead places CPTs within a multimodal assessment framework that includes developmental history, functional impairment and information from multiple settings.

Traditional standardized CPTs currently have the strongest normative and psychometric foundations. The Conners CPT-3 and TOVA are visually structured target-detection tasks with established administration procedures, age-referenced standards and substantial review-level support. IVA-CPT incorporates visual and auditory targets and non-targets and generates platform-specific composite scores for vigilance, focus, speed, prudence, consistency and stamina. Those composites summarize performance under IVA-specific definitions and transformations, so they cannot be treated as equivalents of reaction-time standard deviation, coefficients of variation or variability scores from other systems. Signal-detection analysis can also separate perceptual sensitivity, represented by d′, from response bias, commonly expressed as β or criterion c. Children with ADHD generally show reduced perceptual sensitivity in meta-analytic evidence, but response-bias findings are less consistent, and neither measure reliably distinguishes inattentive from hyperactive–impulsive presentations in an individual child.

Technology is extending the range of behavior that can be recorded during a CPT. The MOXO-dCPT adds controlled visual, auditory or combined environmental distractions to a computerized target-detection task and reports platform-specific indices for attention, timing, impulsivity and excessive or repeated key pressing. QbTest combines an attention and impulse-control task with infrared motion tracking. It has age- and sex-referenced norms, regulatory clearance as an aid in ADHD assessment, and support from systematic-review, meta-analytic and randomized clinical-impact evidence, although its diagnostic discrimination remains modest across heterogeneous clinical samples. By contrast, Kinect-based CPTs record dominant-hand movements used to respond to targets rather than unrestricted whole-body activity; their pediatric evidence remains preliminary and largely confined to a small proof-of-concept study. These systems should therefore not be treated as interchangeable simply because all incorporate motion-related data.

Virtual-classroom systems such as AULA Nesplora and ClinicaVR embed target detection in simulated lessons with controlled visual, auditory and contextual distractors. Their outcomes can include conventional omissions, commissions, reaction times and variability, as well as head movement, orientation and responses to environmental distractions. Such tasks may resemble a classroom more closely than a quiet laboratory test, but resemblance is not the same as demonstrated validity. The review distinguishes ecological resemblance, meaning that a task looks or sounds like a real setting; experimental complexity, meaning that it includes multiple sensors, data streams or manipulated distractions; and demonstrated ecological validity, meaning direct evidence that test performance predicts classroom behavior, academic function or everyday impairment better than traditional CPT performance. For none of the reviewed paradigms, including virtual-reality systems, has sufficient direct evidence established that third form of validity.

Eye-tracking and machine-learning approaches remain largely investigational. Eye-tracking CPTs can quantify fixation duration, dwell time, gaze allocation, saccades and off-task attention, extending measurement beyond what accuracy and reaction time alone reveal. Some studies have reported improved group classification after gaze measures were added, and machine-learning models have also produced promising results in individual datasets. Most findings, however, come from small, single-center, retrospective or proof-of-concept investigations with internal validation. Hardware, calibration, sampling rates, task implementations and algorithms differ across systems, while large, independently replicated, age-stratified and cross-cultural normative datasets remain scarce. Better classification within a research dataset does not establish externally validated sensitivity, specificity, predictive values or incremental usefulness in routine care. The added complexity of a modern platform must therefore be evaluated separately from its scientific novelty.

Low diagnostic specificity is the central limitation across CPT platforms. The cognitive processes they measure are transdiagnostic: attentional and inhibitory difficulties also occur in autism spectrum disorder, anxiety, sleep disturbance, intellectual disability, learning difficulties and other conditions. Much of the published diagnostic evidence compares children with ADHD with typically developing peers, who are unlikely to show the same errors for any reason. Such designs can overestimate discrimination in real clinical settings, where children undergoing evaluation often have overlapping or competing explanations. Comorbidity makes the problem more pronounced. Studies summarized in the review found that some performance abnormalities in children with autism and ADHD resembled those seen in ADHD alone, while intellectual functioning independently influenced several CPT measures. A child with ADHD may also produce a normal result, particularly when intellectual ability is higher, creating false-negative as well as false-positive possibilities.

Interpretation is further affected by sleep, fatigue, anxiety, emotional state, medication, language and sensory abilities, comprehension, cooperation and motivation. Clinicians should examine the configuration and temporal course of several indices and task blocks rather than treat one score as decisive. Omissions accompanied by marked variability may be consistent with intermittent engagement, while uniformly slow but stable responses may indicate broader processing-speed limitations. Commission errors must be considered alongside reaction time, response bias, target frequency and testing conditions rather than automatically labeled as impulsivity. Developmental stage is equally important: attention and inhibitory control change substantially through childhood and adolescence, so norms or demands developed for older children can inflate misclassification among preschoolers. Evidence should receive less weight when another explanation is stronger or when abnormalities appear only under unusual testing conditions.

Treatment response provides a separate application that must also be interpreted carefully. The review notes that treatment-related cognitive change has been studied more directly than diagnostic use, with stimulant studies reporting reductions in omission and commission errors and sometimes reaction-time variability. Methylphenidate-sensitive changes have been observed in small studies using VR-CPT, TOVA and non-VR CPTs, and one direct comparison reported greater improvements with methylphenidate than atomoxetine across several CPT measures. These results concern cognitive task performance, not automatically rated symptoms or real-world functioning. A statistically reliable change exceeding measurement error indicates that task performance changed, but it does not establish clinically meaningful recovery. Repeated testing must also account for platform-specific reliability, practice effects, developmental change and the interval between assessments, making CPT-based monitoring a complement to symptom ratings, functional assessment and clinical evaluation rather than a replacement for them.

The review’s overall conclusion is that CPTs can characterize the temporal dynamics and moment-to-moment variability of attention and response control in ways that retrospective ratings and interviews may miss, but their clinical value depends on the information they add to established assessment. Conners and TOVA retain the most mature normative and psychometric foundations, and QbTest has comparatively stronger evidence among technology-assisted systems. MOXO-dCPT and virtual-classroom systems remain developing, while eye-tracking, Kinect-based and machine-learning-assisted approaches are primarily research tools. Future studies should harmonize task and outcome definitions, build large age-stratified and cross-cultural norms, manage practice effects and test whether specific paradigms improve diagnostic certainty, assessment efficiency, treatment decisions, functional prediction or healthcare outcomes. Until that evidence exists, technological sophistication cannot substitute for demonstrated clinical utility.

Subject of Research: Continuous performance testing as an objective adjunct in pediatric ADHD assessment

Article Title: Continuous performance tests in the assessment of children with attention-deficit/hyperactivity disorder: a narrative review

Article References: Zhou, Y., & Zhao, X. (2026). Continuous performance tests in the assessment of children with attention-deficit/hyperactivity disorder: a narrative review. Frontiers in Neuroscience, 20, Article 1962306. https://doi.org/10.3389/fnins.2026.1962306

Image Credits: AI Generated

DOI: 10.3389/fnins.2026.1962306

Keywords: attention-deficit/hyperactivity disorder, continuous performance tests, pediatric assessment, reaction-time variability, diagnostic specificity, normative data, virtual reality, eye tracking, clinical assessment, psychometrics, Continuous, performance