No study has put fitness and dose in a direct head-to-head test
The question sounds simple, but answering it requires a study that measures both things—how fit someone becomes and how much of the prescribed exercise they actually did—and then asks which one better predicts cognitive improvement. None of the available studies does that. The closest is an eight-week aerobic training trial in sedentary older adults at risk of cognitive decline, which measured fitness (via heart-rate recovery after exercise) and exercise engagement (the percentage of workout time spent in the prescribed heart-rate zone) alongside cognition [1]. Its own verification notes that it does not directly compare the predictive power of fitness versus dose.
The other studies sit on one side of the comparison or the other. Some show that fitter people tend to perform better on cognitive tests [2][3], and one shows that fitness relates to resting brain-wave patterns in young adults [5]. But these are associations with fitness level, not tests of whether gaining fitness beats completing the dose. So the honest answer to the headline question is that the comparison has not been made—and the nearest evidence points the other way.
Why doing the workout may matter more than the fitness number
In the eight-week trial, 63.4% of participants spent at least 70% of their exercise time in the prescribed heart-rate zone, and this engagement differed sharply between people who trained in person and those who trained remotely—remote participants were more engaged [1]. The authors concluded that improvements in both cognition and fitness were driven by exercise engagement, meaning the amount of prescribed exercise people actually performed [1]. That is a claim about what moved the outcomes, not a formal contest between fitness and dose, but it places engagement upstream: you have to do the dose to get the fitness change.
The cognitive results themselves were modest. Participants improved on verbal fluency and verbal memory but got worse on working memory, and all of these changes were small (effect sizes below 0.5) [1]. Fitness improved too, again with small-to-moderate effects [1]. When everything moves a little, it becomes hard to crown one predictor as the winner—which is part of why the question remains open.
Fitness is linked to cognition, but the link is uneven
Where fitness has been studied as a predictor, the pattern is real but selective. In older adults with mild cognitive impairment, higher peak oxygen uptake—the standard laboratory measure of fitness—was associated with better executive function and episodic memory, but not with global cognition [2]. Those associations held for men but not women in that sample, a sex difference the authors flag as a reason to personalize exercise recommendations [2].
A separate study of cognitively normal adults aged 55–80 found that higher fitness predicted faster processing speed, and that this held independently of white-matter lesions—small areas of age-related damage to the brain's wiring [3]. Fitness and lesion burden each contributed on their own, with no sign that lesions weakened the fitness benefit [3]. Even further from behavior, fitness in healthy young adults predicted power in faster resting brain-wave bands (alpha-2 and beta), which the authors suggest may reflect a brain state ready for cognitive engagement [5]. These findings support fitness as a meaningful marker of brain health, but they describe fitness levels, not the cognitive payoff of a prescribed training dose.
What this means if you are choosing an exercise plan
For practical purposes, the evidence favors paying attention to whether you actually hit the prescribed intensity, not just to the fitness number on a test. The trial that measured both found that time spent in the target heart-rate zone tracked the improvements [1], and its authors framed regular commitment to the regimen as the driver. That is a suggestion from one eight-week study, not proof that dose always outranks fitness.
It also means a fitness test is a useful snapshot but an incomplete guide. Fitness correlates with executive function, memory, and processing speed in older adults [2][3], and with resting brain activity in young adults [5], yet none of these studies shows that chasing a fitness gain produces more cognitive benefit than simply completing the workouts as prescribed. The direct comparison—fitness change versus dose as competing predictors of cognitive gain—is the study still waiting to be done.
Sources used in this answer
Exercise engagement drives changes in cognition and cardiorespiratory fitness after 8 weeks of aerobic training in sedentary aging adults at risk of cognitive decline
In eight weeks of aerobic training for sedentary older adults, cognition and fitness improved modestly, and the authors concluded these improvements were driven by exercise engagement—the percentage of workout time spent in the prescribed heart-rate zone—rather than by fitness alone.
Sex Differences in the Associations of Cardiorespiratory Fitness with Executive function, Episodic Memory, and Global Cognition in Older Adults with Mild Cognitive Impairment
In older adults with mild cognitive impairment, higher peak oxygen uptake was associated with better executive function and episodic memory but not global cognition, with significant associations for males only.
35 Impact of Cardiorespiratory Fitness and White Matter Lesions on Processing Speed with Age
In cognitively normal adults aged 55–80, higher fitness predicted better processing speed independently of white-matter lesion volume, with no interaction between the two.
Cerebrovascular reactivity and the fMRI-BOLD response in cardiorespiratory fitness
The paper concerns cerebrovascular reactivity and fMRI-BOLD responses in relation to cardiorespiratory fitness and mentions exercise-related cognitive benefits, without comparing fitness to prescribed exercise dose.
The relationship between cardiorespiratory fitness and resting-state cortical activation: An EEG study on power spectrum density
In healthy young adults, higher maximal oxygen consumption predicted greater resting EEG power in the alpha-2 and beta bands but not in delta, theta, or alpha-1 bands.