Traditional evaluative learning paradigms require participants to passively view and learn about pairs of stimuli, stripping people of the autonomy they typically have in everyday life. However, recent work incorporating autonomy into an evaluative conditioning task suggests autonomy can have consequential impacts on evaluations: sampling an initially-neutral conditioned stimulus (CS) more frequently predicted a positive shift in high-autonomy participants’ evaluations of it, regardless of whether the CS was consistently paired with positive or negative unconditioned stimuli (USs). The current work tested whether this sampling decision effect extended to clearly valent stimuli by modifying the paradigm to also measure participants’ evaluations of the USs. Two experiments (N = 553) replicate the sampling decision effect for CSs but find it does not extend to the paired USs. The experiments also rule out an alternative explanation for the effect on CSs, demonstrating that it cannot by explained by idiosyncratic variance in high autonomy participants’ evaluations of the USs. As such, this work further suggests the act of choosing to sample a stimulus more frequently predicts a positive evaluative shift toward it.