Attention is an intuitive concept that is considered a core component of cognition and everyday functioning (Baddeley, 1996; Duncan, 1986; Norman and Shallice, 1986; Posner and Petersen, 1990), exhibiting a clear trajectory of early life maturation and later-life decline (Fortenbaugh et al., 2015; Tipper et al., 1989). Physiological factors such as fitness are well-known to have a general influence on cognition (Colcombe and Kramer, 2003) and are thought to impact performance through improvements in attentional control (Colcombe et al., 2004, 2006; Prakash et al., 2007). Across the lifespan, attentional processes are linked to the successful navigation of a host of everyday behaviors (Barriga et al., 2002; Bogdanova et al., 2016; Gross, 2015; Halperin, 1991; Kinsella, 1998; Racer and Dishion, 2012; Stierwalt and Murray, 2002); and like many apparently simple behaviors, it can be challenging to define and measure (Anderson, 2021; Shi et al., 2019; Unsworth et al., 2024; von Bastian et al., 2020; Yangüez et al., 2024).

Experimental and clinical work focused on the measurement of sustained attention has produced a wide selection of continuous performance tasks (e.g., Continuous Performance Test: CPT (Beck et al., 1956); AX-CPT (Servan-Schreiber et al., 1996); Psychomotor Vigilance Test: PVT (Dinges and Powell, 1985); Paced Auditory Serial Attention Test: PASAT (Gronwall, 1977); Test of Variables of Attention: TOVA (Leark et al., 1997); Mackworth Clock Test (Mackworth, 1948); Sustained Attention to Response Test: SART (Robertson et al., 1997); Cambridge Neuropsychological Test Automated Battery: CANTAB (Sahakian and Owen, 1992); Continuous Visual Attention Test: CVAT (Schmidt et al., 2024); Gradual-onset Continuous Performance Test: GradCPT (Rosenberg et al., 2013). These tasks are tuned to capture behavioral features thought to contribute to successful performance or identify specific areas of deficit, based on the paradigm and study population of interest (Conners et al., 2003; Cooper et al., 2017; Homack and Riccio, 2006; Klee and Garfinkel, 1983; Lopez-Garcia et al., 2016). Given the complexity of attentional processes and limitations inherent in any one particular paradigm, the development of a large corpus of measures with different features and tuning will facilitate continued knowledge building and translation to practical applications.

As outlined below, opportunities exist to augment or improve upon existing paradigms through novel behavioral sampling, dynamically adaptive assessment, and task calibration approaches - amongst others. Here we describe the Continuous Performance Critical Stability Task (cpCST), which modifies an established sensorimotor integration task to create a novel attention task featuring high-density behavioral sampling, dynamic adaptation, and effective behavioral calibration across the lifespan. We first describe these new features, not readily available in current paradigms, and the proposed advantages of these enhancements. We then present a preliminary psychometric evaluation of the cpCST’s primary outcome metric (instantaneous reaction time; iRT). We also examine predictive validity of the cpCST iRT to flanker task performance, Woodcock-Johnson Intellectual Ability and Achievement scores, as well as a measure of cardiorespiratory fitness (VO2max), before discussing the advantages of the cpCST paradigm in relating physiological and brain timeseries to participant behavior.

In most continuous performance tests, attention is probed via button press responses at discrete intervals ranging from roughly one to several (10+) seconds apart c.f. (DiFrancesco et al., 2019; Dinges and Powell, 1985; Homack and Riccio, 2006); attentional lapses are inferred on the occasion of delayed, missed, or incorrect responses. Despite the relatively sparse sampling of behavior (< 1 Hz - once every second or longer), these response time studies have demonstrated attentional fluctuations over time (Castellanos et al., 2005; Decker et al., 2023; Di Martino et al., 2008; Esterman et al., 2013; Jackson and Balota, 2012; Smallwood et al., 2004); however, higher density sampling of behavior may more effectively characterize the maintenance of focus over time, moment-to-moment fluctuations, and/or lapses in attention. While many established tasks require continuous monitoring of stimuli (e.g., CPT and PVT), they do not sample behavior continuously. Our primary goal for the development of the cpCST was to create a task that sampled behavior at a much higher rate (30 times per second) than existing tests. Further, unlike a stop-signal (Logan and Cowan, 1984) or gradual-onset continuous performance test (Rosenberg et al., 2013), the cpCST does not include the feature of building up a prepotent response as the result of higher frequency responding. This allows for the assessment of continuous attention under qualitatively different conditions than many tests with higher responding rates.

Administering reaction time tasks to older adults, children, or clinical populations often requires adjustments to various parameters such as stimulus type, stimulus modality (audio vs. visual), presentation and response durations, response interval, interstimulus intervals, stimulus set sizes, or proportion of trial types (see Cowan et al., 2010; Craik, 1986; de Souza Almeida et al., 2021; Reuter-Lorenz and Cappell, 2008; Rueda et al., 2004). These adjustments are motivated by group differences in sensorimotor speed, working memory, auditory or visual acuity, etc., (Cerella, 1990; Conway et al., 2003; Denckla, 1996; Jacobson et al., 2011; Kail, 1991; Owsley, 2016; Salthouse, 1996; Verhaeghen and Cerella, 2008; Wingfield et al., 2005). While accommodations such as these allow for versions of standard neurocognitive tasks to be applied across a wider range of individuals, they raise concerns regarding the comparability of results across test variants (Best and Miller, 2010; Cooper et al., 2017; Hedge et al., 2018; Parsons et al., 2019). Additionally, these changes are applied under the assumption that the altered parameters are uniformly appropriate to the group in question [e.g., trading arrow shapes for cartoon fishes in the ANT-C task or an increase in presentation duration for older adults (Gershon et al., 2010; Rueda et al., 2004)], despite well-documented heterogeneity within groups (Fair et al., 2012; Lindenberger and Baltes, 1997; Logan et al., 2023). As part of the NKI-RS2 lifespan characterization study, our goal was to develop a test that did not require different versions across the ages 9–75 years. Our focus was to use simple stimuli and an intuitive response modality to decrease instructional or proficiency barriers. Further we adopted a closed-loop system developed for a sensorimotor paradigm (Jex et al., 1966), described in detail below, that is calibrated to the individual’s own motor performance to equate individual performance differences into a uniform task design. We are not aware of any other continuous performance tasks that incorporate this design feature.

Rather than assuming that a single set of task adaptations will be equally appropriate across a given group (e.g., older adults or children), fully adaptive paradigms individualize task parameters for each participant by dynamically altering key task features in response to ongoing task behavior. Some tasks are explicitly designed to be adaptive [e.g., Stop-Signal Reaction Time (Logan and Cowan, 1984)]. More recent approaches overlay adaptive procedures that alter task features such as presentation time, response windows, or set size, in response to participant performance in real time as the task evolves (Barbey et al., 2022; Draheim et al., 2024, 2021; Schneiders et al., 2011). As such, each participant’s task is custom tailored to their individual performance on that task through approaches such as staircase or Bayesian-based adaptive algorithms (Farahbakhsh et al., 2019; Manning et al., 2018). These approaches are more efficient (Attarha et al., 2024; Davis et al., 2002; Gibbons et al., 2024; Sorrel et al., 2020), and can be leveraged not only in assessment, but also training protocols (e.g., Roheger et al., 2020). However, they also suffer from drawbacks such as edge case and small sample size failures, induction of artifactual oscillatory “yo-yo” patterns in difficulty, as well as the additional complexity involved in dynamically adapting task parameters in real time (García-Pérez, 2011; Kontsevich and Tyler, 1999; Treutwein, 1995).

One promising approach is to leverage the best of both dynamic adaptive approaches and fixed stable approaches. Participant ability is assessed via adaptive staircase or Bayesian methods in a calibration phase. During the test phase, difficulty is set to a fixed level matching the participant’s individual ability (e.g., the stimulus onset asynchrony that resulted in > 70% accuracy) (Chen et al., 2017; Fleming et al., 2010; Lindfield et al., 1994). Under this approach, difficulty is individually tuned - thus avoiding assumptions about the appropriateness or comparability of group-specific stimulus changes. And the difficulty is fixed during the testing phase, reducing computational complexity and artifactual issues such as induced oscillatory behavior or algorithmic failure. For the cpCST, we leveraged such a hybrid approach by incorporating a calibration phase and a test phase. The goal was to maximize our ability to individually calibrate performance on the sensorimotor component of the task so that performance adjustments related to drifting attention could be better compared across individuals who differ in performance on the sensorimotor integration dimension of the task.

The cpCST design is intended to capture attentional dynamics from moment to moment using a simple, relatively short duration, high information density, individualized difficulty approach in order to maximize the detection of attentional control performance differences across a lifespan sample. The cpCST uses the hybrid-calibrated approach to first assess visuomotor ability, replicating the Critical Stability Task developed by Jex et al. (1966), and then employs a fixed difficulty visuomotor continuous performance task based on the individual’s motor stability threshold (MST; see below). Thus the cpCST retains Jex et al.’s (1966) Critical Stability Task name and adds a new continuous performance phase. The cpCST additionally possesses useful features such as simple task instruction and continuous sampling of behavior (@30 Hz; i.e., 30 times per second), providing a robust complement to existing neurocognitive tools.

The cpCST is an extension of a psychomotor tracking task (Critical Stability Task; CST) developed for NASA by Jex et al. (1966) to evaluate pilot performance under unstable control conditions. This simple and elegant foundational paradigm has subsequently been adopted by human and non-human primate laboratories to develop and refine human-machine interfaces (Quick et al., 2018), understand the neural mechanisms of sensorimotor coordination (Sadeghi et al., 2024), and examine drug induced motor control disruption (Ramaekers et al., 2006). We propose that the CST also provides a strong foundation upon which to build a continuous self-calibrated task to assess dynamic fluctuations in attentional control. Specifically, we employed a variant of the original CST to serve as a calibration phase that established a participant’s individual MST. We then used that individualized score to set the difficulty level for that participant’s continuous performance phase. The goal for this new continuous performance phase was to maintain attention on a relatively easy task (set as 30% of MST, based on internal pilot testing) for a 10 min period to capture attentional drift and the latency to respond to a drifting stimulus. It may be useful to clarify that there are a number of continuous tracking tasks in which participants use devices such as joysticks, trackballs, etc., to align a cursor with a spatially moving target item (Colino et al., 2017; Ewolds et al., 2017; Huang et al., 2005; Snow et al., 2016; Yang et al., 2021). The main difference between these sorts of continuous tasks and the cpCST is that the cpCST is based on Jex et al.’s (1966) closed-loop paradigm in which the goal is to maintain the target stimulus at the center of the screen, rather than track, for example, a vertically oscillating target. Additionally, in both the Jex et al. (1966) task and the cpCST, the target’s stability is solely dependent on the user’s movements – there is no externally driven movement of a target object for the user to track. See Figure 1, described in more detail below.

To evaluate this novel task, we embedded the cpCST within the Nathan Kline Institute Rockland Sample II (NKI-RS2), a large-scale, community-based lifespan study. The NKI-RS2 was designed to support the development and validation of next-generation tools for phenotyping normative brain-behavioral associations and investigate the underlying neural and physiological mechanisms that promote mental health across the lifespan. This context offered an opportunity to examine individual differences in attentional control across a wide age range using a task that prioritizes continuous behavioral sampling, individualized calibration, and high-density data collection. We characterized cpCST performance in relation to broader indices of cognitive function and health. In this preliminary analysis, we report behavioral data from the cpCST from a subset of participants to describe the development of a key task performance metric (instantaneous reaction time; iRT) and establish its reliability and preliminary predictive validity on cognitive and physiological indices.

Participants (N = 166) ranged in age from 18 to 76 years (M = 51.61, SD = 16.36) and reported 12 to 20 years of formal education (M = 15.81, SD = 2.11). The sample was 66% female (n = 110) and 34% male (n = 56). In terms of race, 81% identified as White (n = 134), 10% as Black or African American (n = 16), 5% as Asian (n = 8), 2% as American Indian or Alaska Native (n = 3), and 3% as multiracial (n = 5). Regarding ethnicity, 86% were Not Hispanic or Latino (n = 143), 13% were Hispanic or Latino (n = 22), and 0.6% preferred not to answer (n = 1).

We introduced the Continuous Performance Critical Stability Task, offering high temporal precision of continuous psychomotor control across the lifespan. This report provides preliminary evidence for the reliability, predictive validity, and temporally efficiency of the cpCST – a potentially valuable complement to existing attention assessment paradigms. Below, we summarize key methodological innovations and psychometric properties, followed by implications for future research and clinical applications.

The Continuous Performance Critical Stability Task introduces methodological advances in the assessment of attention. Its exceptional reliability, age invariance, predictive validity, and temporal efficiency address limitations in existing measures. Future validation efforts integrating physiological measures, computational modeling, and diverse clinical applications will further establish cpCST’s utility as an essential tool for attention research and assessment.