A defining feature of human cognition is the ability to behave in a goal-directed manner in the face of a myriad of competing, and often salient, alternatives (our ability to self-regulate) [1-3]. Behaviors that promote long-term health, such as physical activity (PA), are often described as “effortful” or “difficult” and thus require effective self-regulatory control to align momentary behavior (going for a walk) with long-term goals (to remain healthy) [4-8]. Regular PA is associated with a wide range of long-term health benefits, including reduced risk for preventable chronic disease (eg, cardiovascular disease and type 2 diabetes) and age-related neuropathology (eg, Alzheimer disease and related dementias) [9-14]. In addition, regular PA may support long-term cognitive health throughout the adult lifespan [15,16]. Thus, a virtuous cycle has been speculated to exist between cognition and PA behavior [5,17,18]. Critically, however, substantial individual differences exist in the ability to translate one’s motivation and goals to maintain or increase PA into overt daily behavior (known as the PA “intention-behavior gap”) [6,8,19,20]. Despite great efforts to understand why such a gap exists and how to successfully intervene, the antecedents and moderators of successful self-regulation remain poorly understood. Moreover, it remains unclear whether these factors play a stable role in PA self-regulation across the adult lifespan.

Recent attention has turned to cognitive ability, specifically executive control, as a possible source of individual differences in self-regulatory capacity [21-24]. Several recent studies evaluating the executive hypothesis, including our own work, support this claim showing that control over information processing moderates and independently contributes to health outcomes and health promotion behavior over and above self-reported intentions or tendency to self-regulate [6,21,22,25,26]. However, executive control abilities are not stationary and may vary as a function of time (eg, time of day) and context (eg, in response to a stressful interaction) [27-29]. Moreover, the underlying neurocognitive resources (the frontoparietal brain structure and function) are particularly subject to age-related decline and alteration [30-38]. These multitimescale alterations may influence when an individual is prepared to successfully self-regulate and may modify the process of self-regulation over the course of the adult lifespan. However, few studies have examined these mechanisms across these timescales or in an age-diverse sample.

The Physical Activity, Imaging, and Ambulatory Testing (PHIAT) project was a high-frequency, ambulatory assessment study conducted over the course of 2019-2024. The project was sponsored by the National Institute on Aging of the National Institutes of Health as part of an R00 Career Development Award (R00AG056670; principal investigator: JGH). The project was designed as an adult lifespan, microlongitudinal observational study of real-time variation in cognitive executive control and self-regulation of health-promoting behavior. The overall aims of the PHIAT project were as follows: (1) to examine variation in executive control and self-regulation of PA in real time and (2) to examine age-variant and invariant contributions of brain structure, function, and executive control in self-regulation of health-promoting behaviors. The overarching hypothesis examined by the PHIAT project was that executive control is a resource upon which successful self-regulation depends (an executive hypothesis of self-regulation; see also [8,22-24]).

To accomplish the PHIAT project aims, we chose an ecological momentary assessment (EMA) approach. The EMA protocol was conducted over a 14-day measurement burst (6 assessments/d). This approach was complemented by the contemporaneous use of multiple wearable activity monitors (to measure PA, sedentary behavior, and self-monitoring behavior) and the administration of ultra-brief, ambulatory cognitive assessments during EMA surveys to examine short-term variation (within- and between-day) in multiple domains of cognition.

These high-frequency data streams were designed to be integrated at multiple timescales to accomplish the study aims. A central focus of the project was on short-term variation in executive control. In a previous set of studies, we rigorously designed and parameterized ultra-brief adaptations of complex span working memory tasks for EMA administration [27]. Based on our development and testing, we selected the ultra-brief (~1 min) rotation span task as our primary indicator of executive control in PHIAT. We also included other ultra-brief experimental indices of attention and executive control, including a go/no-go task and a multiple object tracking task, as well as a previously validated indicator of processing speed (symbol search task) [39]. Each of these tasks was administered during 5 of the 6 daily EMA surveys (morning survey, plus 4 beeped surveys), providing a detailed time series of variation in multiple domains of cognition both within and between days in the study.

Participants were recruited from a combination of social media postings, internal institutional resources (eg, “StudyFinder”), secondary recruitment from other studies conducted at Penn State, and the community surrounding health care sites of Penn State Health system. Participants were included if they were aged 18 years and older, were ambulatory, were fluent in English, had access to a reliable cellular network or Wi-Fi connectivity for onboarding, and had an email account for communication of remote consent procedures. Participants were excluded if they had a motor or visual impairment that would interfere with the operation of the study-provided smartphone, had a history of neurological injury or disease or significant health problem that could be complicated by PA, were unable to consent, were younger than 18 years, or were a prisoner.

All participants provided written informed consent and received compensation for their participation. The compensation rate for the study was US $5 per day of participation (14 d × US $5 = US $70), with a US $30 incentive payment for completing >75% of expected surveys (maximum compensation: US $100).

All experimental procedures were approved by Pennsylvania State University’s Institutional Review Board for the ethical treatment of human participants under protocol number STUDY00013106. All participants provided electronic informed consent via REDCap prior to engaging with any study procedures. All study data were deidentified to ensure participant privacy and confidentiality.

The study protocol involved 3 phases: (1) participants completed a remote study onboarding visit where informed consent was documented, a telephone adaptation of the Montreal Cognitive Assessment [46] was administered, and participants received a study overview, study materials (study-provided smartphone, tablet, activity monitoring wearables, and EMA app), and instructions for completing the EMA surveys and ambulatory cognitive assessments; (2) questionnaires were administered on a study-provided tablet during the week leading up to the EMA measurement burst; and (3) participants completed a 14-day EMA measurement burst while wearing the activity monitors.

Participants were asked to carry a study-provided smartphone for a period of 14 days (Xiaomi Mi A2 model phones configured with Android One OS, v8.0; Xiaomi Corporation, Beijing, China). A custom Android EMA mobile app, the “Mobile Monitoring of Cognitive Change” (M2C2) app, was loaded onto the study-provided phones and used to administer the EMA protocol, including self-report survey items and 4 ambulatory cognitive assessments. Phones were configured into a “kiosk” mode where only the study app was available to the participant. EMA surveys were administered to each participant over 14 consecutive days in a measurement burst design. Each day involved 1 self-initiated morning survey, followed by 4 pseudorandomly signaled beeped surveys, and 1 self-initiated evening survey (Figure 1C). Participants were instructed to begin the morning survey as soon as they woke up and the evening survey before going to bed. At the start of the study, participants provided their typical waking time. From this information, the EMA protocol was initialized so that the first beeped survey would be delivered at a pseudorandomly determined time within 2 hours of their reported waking time. The remaining 3 beeped surveys were each separated by 3.75 hours and with a start time that was pseudorandomly jittered each day in the burst by up to 30 minutes. Beeped surveys “timed-out” (were unavailable) if the participant did not respond within 30 minutes of the notification. All ambulatory cognitive assessments were administered at the end of each morning and beeped survey but not during the evening survey to limit participant burden.

Brief ~1-minute self-report surveys with items covering the domains listed below were administered at the beginning of each survey administration (ie, session). EMA surveys were designed to provide an in-depth description of participants’ daily experiences, exposures, thoughts, and behaviors (all surveys), as well as their outlook and expectations for each day (during the morning surveys) and review and inventory of each day as a whole (during the evening surveys). A particular emphasis was placed on participation, and factors surrounding participation, in healthful behaviors (PA being of primary interest). For example, each morning survey included items such as “Today, I intend to be physically active.” These items were included to model daily within-person self-regulatory processes, including associations between daily variation in PA motivation and behavior (measured via activity monitors, see below). Throughout each day, exposures that might facilitate or act as a barrier to engagement in healthful behaviors (eg, opportunities for PA, exposure to stressful events) were sampled to model short-timescale variation in engagement and other processes (eg, moment-to-moment, within-day variation in PA behavior). Table 1 presents a complete list of all items included in the protocol.

The high-frequency sampling methodology inherent to EMA protocols results in complex nested data structures. Each observation (“moment”) can be considered as nested within a day of the protocol and, in turn, nested within an individual. Statistical analysis approaches that can adequately account for these complex data structures will be used to achieve the study aims. Mixed-effects modeling and dynamic structural equation modeling are two leading approaches to analyzing EMA data that will be used [68]. In each case, all observations will be “centered” in a manner that provides for unambiguous interpretation of effects at each “level” of the study design (ie, within-person, between-person) [69].

For analyses related to the dynamic action control framework, linear mixed-effects models will be estimated where day-to-day variation in behavior (eg, daily step counts) will be regressed on within- and between-person (ie, “characteristic”) variation in morning intention ratings. To examine how the characteristic level of, for example, PA intention strength modifies associations between daily variation in PA intention strength and daily PA behavior, cross-level interactions between characteristic PA intentions, and within-person daily variation in PA intention strength will be included in the model. For analysis of momentary contextual reactivity effects, linear mixed-effects models will be estimated to examine how momentary variation in the outcome of interest (eg, negative affect, PA behavior between surveys) is affected by exposure to a given context (eg, stressor, opportunity for PA), consistent with our previous work [28,70].

In each case (dynamic action control and momentary contextual reactivity), the central hypothesis being tested in the PHIAT project, the executive hypothesis of self-regulation, will be examined at both the between- and within-person levels. To test the hypothesis, within-person variation and characteristic level of executive control (momentary and average performance on the ambulatory cognitive assessments, respectively) will be included as moderators of action control and contextual reactivity effects. For example, we will include within and between interactions between momentary and characteristic working memory capacity (rotation span task performance) and PA intention strength in dynamic action control models to evaluate how within- and between-person executive control ability modifies associations between intentions to engage in PA and PA behavior.

We expect data generated by the PHIAT project to provide new insights into the intersection of cognitive executive control ability, self-regulation, and health-promoting behaviors. In particular, we anticipate that the ability to link the high-frequency assessment protocol and activity monitoring devices at multiple timescales will help bring the dynamic processes surrounding self-regulatory success and failure into sharper relief. Few previous studies have examined how short-timescale variation in these factors interacts in determining participation in health-promoting behavior, and the current study is an initial step in moving from a cross-sectional/individual difference approach to action control, contextual reactivity, and the executive hypothesis toward a within-person, process-level description.

The 3 activity monitoring devices selected for the study, the thigh-worn ActivPAL monitors, hip-worn ActiGraph GT3X devices, and the wrist-worn commercial fitness monitor, were each chosen to capture different aspects of PA and self-regulatory behavior. The long battery-life ActivPAL devices, being affixed to the thigh, allowed for near-continuous measurement of posture and movement throughout the 14-day protocol. From the ActivPAL devices, we expect to obtain high-fidelity estimates of total PA volume (eg, step counts) as well as estimates of time spent in various postures/behaviors (eg, stepping time, sitting time). The ActiGraph GT3X devices are a widely used tool for estimating PA at different intensity thresholds (eg, moderate-to-vigorous intensity PA) [71]. Because the sample includes adults throughout the lifespan, the selection of activity intensity cut points will require careful consideration [72]. Finally, while the activity data generated by the wrist-worn fitness watch is not expected to be an outcome of interest, participants’ use and interaction with the “check my steps” feature of the EMA application (which was connected via Bluetooth to this device) will be examined in studies of how self-monitoring of PA behavior is influenced by goal setting and variation in executive control.

The wide age range of the PHIAT study sample will allow for investigation into how age-associated changes in each factor either independently or mutually alter self-regulatory processing throughout the adult lifespan. Previous work has shown that many of the individual factors under study undergo normative age-associated change. For example, it is well-established that older age is associated with reductions in PA, increases in sedentary behavior, and reduced executive cognitive ability [73-79]. In addition, previous work on psychosocial aspects of aging (eg, affective and cognitive well-being) has demonstrated that factors such as perspective, world view, and cognitive capacity may play a role in altering the strategies we select and rely on to manage our thoughts, emotions, and behavior in older adulthood [24,80-82]. The PHIAT project offers an opportunity to connect this line of work from the fields of psychosocial aging and human development to these areas of health promotion theory.

While the strengths of the PHIAT project study design reviewed above offer the potential to gain new insights into self-regulation of health promotion, there are several limitations. While every effort was made to recruit a generalizable and representative sample, the final sample was disproportionately female and White. We suspect this was driven, in part, by our advertising and recruiting methods, which included a health system–associated marketing campaign over social media. In addition, despite recruiting community members throughout the region with our sampling methodology, the PHIAT project involves a volunteer convenience sample that may not fully generalize to the broader population. While the inclusion of an adult lifespan sample has significant advantages, as reviewed above, there are systematic changes in the factors under study that are known to occur with age. Careful consideration of how to bring chronological age into analytical models will need to be given (eg, as a continuous factor, through group stratification), as well as attention to how changes in the effective sample size within each strata (younger/older, decade by decade) affect statistical power. Finally, while EMA studies that examine how individuals behave in response to everyday exposures can be viewed in many ways as naturalistic experiments, the project still involves an observational study design, and findings will require further experimental work to tease apart causality and mechanism of action.

The PHIAT project involved a novel study design and an adult lifespan sample. The design included near-continuous sampling of thoughts, emotions, context, cognition, and behavior, and is amenable to a wide array of multilevel, time series, latent variable, and other modeling approaches for developing insights into these rich and nested data streams. The inclusion of community-dwelling adults ranging from 18 to 89 years of age will help provide insights into how self-regulation of health-promoting behaviors changes throughout the lifespan and help identify potential mechanistic targets for health promotion interventions.