Beyond memory and experimenter demand: Scientific consensus messages correct misperceptions

Scientific consensus messaging is among the most widely studied and cited approaches in climate change communication. Research consistently shows that scientific consensus messages (“97% of climate scientists agree that human-caused climate change is happening”) reduce misperceptions and enable climate change attitudes that reflect the scientific evidence. A concern, however, is that the effects of this and similar consensus messages may reflect memory and experimenter demand effects rather than genuine misperception correction. In this brief empirical note, we outline several theoretical arguments and reanalyze two large-scale secondary datasets from the United States ( n = 6,301; van der Linden et al., 2019) and 27 countries ( n = 7,000; Većkalov et al., 2024 ), providing evidence that misperception correction following scientific consensus messages is unlikely to be driven by memory and experimenter demand effects alone. First, we show that consensus messaging reduces misperceptions compared to a control condition, even in the highly unlikely scenario that every post-test response at or near 97% reflects memory or experimenter demand effects. Second, we demonstrate that participants report being more persuaded by the scientific consensus message and update their personal beliefs slightly more when they indicate consensus estimates of 97% ( vs . a different number) in the post-test. If the ‘97%’ responses were primarily driven by memory effects, we would expect the opposite pattern. Overall, scientific consensus messaging recalibrates scientific consensus perceptions and important attitudes, regardless of whether individuals report exactly 97% at post-test–reinforcing the importance of perceived scientific consensus as a ‘gateway’ belief to evidence-based climate change attitudes. • Some claim consensus messaging effects reflect memory/demand effects, not belief change. • We analyzed two secondary datasets from scientific consensus messaging experiments. • Misperception correction is robust to excluding post-test responses of ‘97%’. • Reporting ‘97%’ at post-test is linked to greater self-reported and actual persuasion. • Misperception correction is unlikely driven by memory or experimenter demand alone.