Exercise or not? An empirical illustration of the role of behavioral alternatives in exercise motivation and resulting theoretical considerations

The study addresses why intentions to exercise often fail to translate into behavior, focusing on in-the-moment (situated) decision processes when individuals must choose between exercising and a competing non-exercise alternative. While social-cognitive approaches emphasize intentions derived from expectations and goals, a persistent intention–behavior gap suggests additional processes are at play. Dual-process models (e.g., Affective-Reflective Theory of Physical Inactivity and Exercise, ART; Theory of Effort Minimization in Physical Activity, TEMPA) posit that automatic, situation-dependent processes can either facilitate or hinder exercise behavior. ART emphasizes learned automatic affective valuations of exercise; TEMPA posits a default tendency to minimize effort, biasing toward inactivity. Prior work shows immediate automatic reactions to exercise stimuli and links individual differences in automatic and controlled processes to exercise behavior, but little is known about the situated attentional processes that occur precisely when choosing between exercise and a specific alternative. This study investigates gaze behavior (first gaze, fixations) during forced choices between exercise and non-exercise options, and relates these situated processes to interindividual differences in automatic valuation (AMP), controlled evaluation (self-reported feelings), and usual exercise volume.

The paper situates its research within dual-process frameworks that integrate automatic and controlled processes in exercise behavior. ART posits learned automatic affective valuations of exercise that can facilitate or inhibit reflective intentions; TEMPA suggests an inherent tendency to minimize effort creating a default bias against physical activity. Empirical studies show immediate affective and approach/avoidance responses to exercise cues and associations between implicit/explicit measures and later behavior. Harris and Bray (2019, 2021) demonstrated that mental fatigue reduces likelihood of choosing exercise in single-shot choices, highlighting situated influences. Cheval et al. (2020) found active individuals exhibit attentional bias toward physical activity stimuli using eye-tracking, though not in a forced-choice context. Process tracing from consumer psychology shows mixed links between first gaze and choice but stronger associations between total fixations and choice, with debate on causality. The current study extends this literature by examining gaze during actual forced choices between mutually exclusive exercise vs. non-exercise alternatives and linking these situated attentional processes to interindividual automatic and controlled evaluations and typical exercise behavior.

Design and participants: Cross-sectional lab study using eye-tracking during a computerized forced-choice task. Participants were 101 university students (Mage=23.6, SD=3.6; 48.5% female) from the University of Potsdam; most studied sport science (n=80) or psychology (n=21). Five participants were excluded due to technical issues. Inclusion criteria included normal or corrected-to-normal vision and no color blindness; participants avoided confounding activities prior to testing. Ethics followed institutional guidelines; written informed consent obtained. Data and materials are available at https://osf.io/ubrj7/. Behavioral alternatives task: Participants read five everyday vignettes describing scenarios involving a choice between exercising and a non-exercise alternative (two alone, two with others, one ambivalent). For each vignette, five randomized pairs of grayscale pictures were presented (25 trials total). Each pair showed an exercise and a non-exercise option side-by-side with randomized left-right positions. Participants had up to 10 s to choose using keyboard keys. A 5 s fixation cross preceded each trial. Choices were coded 1=exercise, 0=non-exercise. Pictures: 25 exercise and 25 non-exercise images matched on brightness/contrast, perspective, number of people, and absence of facial expressions, sexual stimuli, or labels; resolution ≥1024×768. Exercise images represented common moderate/vigorous activities (e.g., biking, running, fitness, swimming, rollerblading); non-exercise included leisure activities (reading, listening to music, lying in the park). Images sourced from a license-free database (pixabay.com) and four author-taken. Eye-tracking and gaze metrics: Gazepoint GP3 eye-tracker (60 Hz) recorded gaze. First gaze was coded 1 if the first fixation landed on the exercise image, 0 for non-exercise. Fixations were identified using the I-VT algorithm (velocity threshold 30°/s). Number of fixations was computed separately for exercise and non-exercise images per trial. Interindividual measures: - Automatic valuation of exercise: An affect misattribution procedure (AMP) using the same exercise and non-exercise images as primes and Chinese ideographs as targets. Primes: 75 ms, 125 ms blank, ideograph 200 ms, followed by mask until response (pleasant/unpleasant). 100 randomized trials. AMP score = proportion pleasant after exercise primes minus after non-exercise primes (range -1 to 1), then z-transformed. Split-half reliability p=0.81. - Controlled evaluation: Single-item self-report of overall feelings toward exercising on a 7-point continuum (absolutely negative to absolutely positive); z-standardized. - Self-reported exercise volume: From IPAQ-short adapted for deliberate exercise sessions (frequency × duration per session), expressed as minutes per week. One outlier (360 min/session) excluded from analyses involving volume. Procedure: Participants completed AMP first, then eye-tracker calibration and the behavioral alternatives task in a dim lab setting (22" monitor, 60 cm viewing distance). After the task, they completed a questionnaire on demographics, potential confounders, controlled evaluation, and exercise behavior. Statistical analysis: Generalized mixed-effects models (lme4 in R) with crossed random effects for subjects and trials. Logistic mixed models predicted first gaze (exercise vs. non-exercise). Linear mixed models predicted number of fixations on exercise and non-exercise images. Analysis steps: (1) unconditional means, (2) add choice (0/1) to test gaze–choice associations, (3) add each interindividual variable (AMP, controlled evaluation, exercise volume) separately to examine subject-level differences in gaze. Power simulations indicated ≥90 participants with 25 trials provide ~80% power for medium effects in crossed designs. Significance assessed via 95% CIs and p-values.

- Choice preference: Across all trials, participants were more likely to choose exercise than non-exercise (OR=1.85, 95% CI [1.39, 2.47], p<0.001), corresponding to ~65% exercise choices. - Correlates of choosing exercise (subject level): - Self-reported exercise volume: r=0.43, 95% CI [0.20, 0.53], p<0.001. - Controlled evaluation (positive feelings toward exercise): r=0.58, 95% CI [0.43, 0.70], p<0.001. - Automatic valuation (AMP): r=0.20, 95% CI [0.00, 0.38], p=0.05. - First gaze: No overall bias to fixate exercise vs. non-exercise first (OR=1.29, 95% CI [0.89, 1.88], p=0.18). First gaze was more likely to land on the ultimately chosen alternative (OR=1.30, 95% CI [1.04, 1.62], p=0.02). First gaze was not significantly related to exercise volume (OR=1.00, p=0.39), controlled evaluation (OR=0.99, p=0.77), or AMP (OR=0.98, p=0.66). - Fixations: Participants made more fixations on the chosen alternative than the non-chosen one. Non-exercise fixations increased when non-exercise was chosen (bnon-ex=1.07, 95% CI [0.78, 1.36], p<0.001) and exercise fixations increased when exercise was chosen (bex=-0.79, 95% CI [-1.05, -0.53], p<0.001). Mean per-trial fixations: exercise 3.99 (95% CI [3.66, 4.31]); non-exercise 3.90 (95% CI [3.56, 4.24]). No overall difference in fixations between exercise and non-exercise across all trials (b=-0.09, 95% CI [-0.34, 0.17], p=0.51). - Interindividual variables and fixations: AMP, controlled evaluation, and exercise volume generally did not predict number of fixations on exercise or non-exercise. A small effect indicated more negative self-reported feelings were associated with more fixations on non-exercise (b=-0.27, 95% CI [-0.53, -0.00], p=0.05). Overall, gaze behavior tracked in-the-moment choices rather than general preferences or usual behavior.

The findings indicate that when individuals face a concrete choice between exercising and a specific non-exercise alternative, their attention (initial gaze and fixations) aligns with what they are about to choose, rather than with their general automatic or controlled evaluations of exercise or their typical exercise volume. This supports dual-process perspectives emphasizing the importance of situated, momentary processes in shaping behavior, while challenging assumptions that inherent (TEMPA) or learned (ART) automatic biases directly manifest as early attentional biases in such choice contexts. The absence of a general first-gaze bias toward either option and the weak links between interindividual measures and gaze suggest that task demands and the specific alternatives at hand predominantly drive attentional allocation during decision-making. The robust association between fixations and choice is consistent with the gaze cascade effect, implying that attention may actively participate in constructing preferences at the moment of choice. These results highlight that beyond cultivating positive general attitudes or intentions toward exercise, interventions should address the salience and appeal of competing alternatives at the time decisions are made. The study thus contributes to refining dual-process models by demonstrating the partial independence of state-like situated processes from trait-like general preferences in exercise-related choices.

Interindividual differences in general exercise preferences (automatic-affective valuation, controlled evaluation, and usual exercise behavior) relate to overall choice tendencies between exercise and non-exercise options. However, situated gaze behavior during these choices does not reflect these interindividual preferences and instead depends on the specific alternatives present, aligning with the option ultimately chosen. This underscores the pivotal role of situated processes in exercise decision-making and suggests that theories and interventions should better incorporate the influence of competing behavioral alternatives at the moment of choice to more effectively promote physical activity.

- Ecological validity: Choices were made in hypothetical scenarios; real-life decisions may differ. Future work should employ ecological momentary assessment or naturalistic tasks to capture in-situ processes. - Causality: The within-subject design is observational; causal effects of attention on choice cannot be inferred. Experimental manipulation of attention is needed. - Sample characteristics: Predominantly university students, many in sport-related programs, and relatively active participants may bias toward exercise choices, limiting generalizability. - Number of trials: Only 25 trials; while suitable for within-trial process analyses with crossed random effects, more trials would be needed to assess stable general preferences. - Measurement constraints: The AMP measured automatic valuation once, not situationally for each choice; it may not capture moment-to-moment automatic processes. Limited variability in exercise volume may also have attenuated associations. - Task demands: Forced-choice instructions and time limits may accentuate task-driven attention over individual preferences, potentially differing from free-viewing contexts.