Study five put the mediation model to a causal test by experimentally influencing people’s perceptions of how well-read they are. A new sample of 1,500 American adults drawn to match the demographics of the U.S. were given a false-feedback manipulation. ( ) After filling out the ART, participants were randomly told they did very well (top 15% of the population) or very poorly (bottom 15% of the population). Afterwards, they completed the same procedure as used in study four with the exclusion of the ‘adults these days’ measure as it was not part of the tested causal chain. The manipulation successfully changed participants’ subjective belief in how well-read they are (β = -.119, p < .001, 95%CI = -.17 to -.067), how much they ‘recalled’ enjoying reading as a child (β = -.055, p = .022, 95%CI = -.104 to -.009), but not how much they recalled their peers enjoying reading (β = -.023, p > .39). Crucially, we observed the predicted indirect effect of reducing subjective beliefs in how well-read someone believes they are causing a reduction in KTD through ‘memory’ of how much one recalls enjoying reading as a child (β = -.009, p = .04, 95%CI = -.02 to -.002). Furthermore, there was no direct effect of the feedback manipulation on KTD not passing through a mediator (β = -.023, p > .4). The absence of a direct effect of the false feedback manipulation on the KTD effect argues against the possibility that it occurs because people subjectively high in a trait set higher standards ( , , ). Instead, the fact that experimentally influencing people’s self-perceptions indirectly affects the KTD effect through people’s altered recollections, indicates that the KTD effect is driven (at least in part) by people’s biased memory of themselves as children. Additional Details First, there was no interaction between need for achievement and any of the DVs (all ps > .06), so we dropped nfa and the interaction between with the bad feedback. The results from the model are as follows: We built the causal mediation model as seen in study four but without the ‘adults these days’ variable. We also included the bad feedback effects on subjective judgment of how well-read people think they are (enjoyadult), how much they ‘recalled’ enjoying reading as a child (enjoychild) and how much the ‘recalled’ their peers enjoyment of reading (enjoypeers). First, to verify the manipulation through false feedback worked, we examined in the model whether participants believed they were less well-read; they did (β = -.119, p < .001, 95%CI = -.17 to -.067). Looking at the other parts of the model, we can see that the manipulation also had a direct effect on how much they ‘recalled’ enjoying reading as a child (β = -.055, p = .022, 95%CI = -.104 to -.009). There was no effect of the false feedback manipulation on how much they recalled their peers enjoying reading (β = -.023, p > .39). The main test of the model is whether the false feedback had a mechanistic effect on KTD through our proposed mediators. The predicted indirect was the effect of the feedback on subjective judgments, ‘recalled’ judgments, and the effect on ‘recollection’ of peers enjoyment of reading on KTD. As the manipulation did not alter judgments of peers enjoyment, we do not discuss any indirect effects. The indirect effect of the feedback on KTD through subjective assessments of the current self and through ‘memory’ of how much one recalls enjoying reading as a child was statistically significant and in the predicted negative direction (β = -.009, p = .04, 95%CI = -.02 to -.002). Meaning, making people believe they are currently less well read caused a decrease in KTD by not only making them believe they currently enjoy reading less but also making them believe as children they enjoyed reading less.