Whenever people try to recollect an earlier given estimate after they have received feedback about the true solution, they tend to overestimate what they had known in foresight. This phenomenon is known as ''hindsight bias''. This paper reports three attempts to eliminate hindsight bias by labelling the feedback value as another person's estimate (instead of as the solution) and by providing extremely incorrect (instead of the true) values as feedback. Both variations, however alone and in combination failed to reduce hindsight bias. Only when the data were separated according to whether participants considered the feedback value plausible or not did cases of unbiased recollections emerge: Feedback values that were labelled as estimates of another person and found to be implausible did not lead to hindsight bias. This finding argues against the view that hindsight bias is an automatic and unavoidable effect of feedback presentation.