White matter integrity and cognition have been found to decline with advancing adult age. Aerobic exercise may be effective in counteracting these declines. Generally, white matter integrity has been quantified using a volumetric measure (WMV) and with tensor-based parameters, such as fractional anisotropy (FA) and mean diffusivity (MD), the validity of which appears to be compromised in the presence of crossing fibers. Fixel-based analysis techniques claim to overcome this problem by yielding estimates of fiber density (FD), cross-section (FC), and their product (FDC) in multiple directions per voxel. In a sample of 61 healthy older adults (63–76 years old), we quantified changes in white matter integrity following an aerobic exercise intervention with the commonly used volumetric and tensor-based metrics (WMV, FA, MD) and with fixel-based metrics (FD, FC, FDC). We investigated the associations of changes in these white matter parameters with changes in cardiovascular fitness and Digit Symbol Substitution task (DSST) performance, a marker of perceptual speed. In line with previous findings, we observed maintained WMV in the corpus callosum of exercisers, and positive change-change correlations between WMV and fitness, and between WMV and perceptual speed. For FA and MD, group differences in change opposite to those hypothesized were found in the corpus callosum, posterior corona radiata, and superior longitudinal fasciculus at an uncorrected significance threshold. Likewise, regions in superficial WM in the prefrontal cortex showed group differences in FD and FDC change, uncorrected, with more positive change in controls and more negative change in exercisers. Finally, changes in FD and FDC were found to be inversely correlated to changes in fitness and DSST performance. The present results corroborate previous findings of WMV changes, but cast doubt on current physiological interpretations of both tensor-based and fixel-based indicators of white matter properties in the context of exercise intervention studies.