Metacognition is defined as the capacity to monitor and control one's own cognitive processes. Recently, Carpenter and colleagues (2019) reported that metacognitive performance can be improved through adaptive training: healthy participants performed a perceptual discrimination task, and subsequently indicated confidence in their response. Metacognitive performance, defined as how much information these confidence judgments contain about the accuracy of perceptual decisions, was found to increase in a group of participants receiving monetary reward based on their confidence judgments over hundreds of trials and multiple sessions. By contrast, in a control group where only perceptual performance was incentivized, metacognitive performance remained constant across experimental sessions. We identified two possible confounds that may have led to an artificial increase in metacognitive performance, namely the absence of reward in the initial session and an inconsistency between the reward scheme and the instructions about the confidence scale. We thus conducted a preregistered conceptual replication where all sessions were rewarded and where instructions were consistent with the reward scheme. Critically, once these two confounds were corrected we found moderate evidence for an absence of metacognitive training. Our data thus suggest that previous claims about metacognitive training are premature, and calls for more research on how to train individuals to monitor their own performance.