Expectations, or the prior probability of a choice outcome, are powerful sources of evidence for improving decision making under uncertainty. Most expectations in the real world are learned implicitly on the basis of statistical properties of observers’ environments. However, most studies investigating effects of expectations on perceptual decisions explicitly instruct observers on prior probabilities within the experiment, and thus fail to capture the experience-dependent uncertainty of real-world expectation learning. Here, we report data from a novel expectation-guided perceptual decision making task specifically designed to address this gap. Human observers (n=21) learned, through experience, probabilistic relationships between cues and images. Then, they explicitly reported both an estimate of each cue’s prediction and a confidence rating in that estimate before performing a cued perceptual decision task. We find that, although these measurements are highly correlated, confidence in an explicit report is the primary factor that interacts with implicit expectations to shape perceptual decisions.