Evidence has suggested that cognitive decline may be a risk factor for freezing of gait (FOG) in Parkinson's disease (PD). Complex and challenging exercises have been suggested as potential rehabilitation strategies to decrease FOG severity and improve cognition; however, it is unknown whether improvement in cognition would explain decreased FOG severity following exercise. In this secondary analysis, we evaluated the effects of the adapted resistance training with instability (ARTI-complex and challenging exercises) compared with traditional motor rehabilitation (TMR-without challenging exercises) on cognitive function in people with FOG of PD. We also verified whether cognitive improvement explains the decrease in FOG previously published. Participants were randomized to either the experimental group (ARTI, n = 17) or the active control group (TMR, n = 15). Both training groups exercised 3 times a week for 12 weeks (80-90 minute each session). FOG severity (FOG ratio from inertial sensors during a 360° turning-in-place task), frontal lobe function (Frontal Assessment Battery [FAB]), global cognition (Montreal Cognitive Assessment [MoCA]), and attention and psychomotor speed (Digit Symbol Substitution Test [DSST]) were evaluated before and after interventions. Only the ARTI group improved FAB, MoCA, and DSST scores at posttraining. In addition, ARTI was more effective than TMR in improving FAB scores at posttraining. The changes in FAB scores explained the changes in FOG ratio following ARTI (R2 = .43, P < .01). This pilot study suggests that ARTI, a complex and challenging training, improves cognition in people with FOG of PD. Improvements in frontal lobe function with ARTI help explain decreased FOG severity.