Marshmallow evidence
On the topic of "extraordinary claims require extraordinary evidence", I always thought that the claims made from Stranford's Walter Mischel's famous "Marshmallow Test" required better evidence. For example, most children in the experiment came from the Stanford University community, a slightly biased sample of 3 yr olds.
What are the claims? That a child's ability to delay gratification (and not eat the marshmallow) was a good predictor of success and wellbeing in education and adult life. "Studies famously linked the ability to delay gratification with a variety of positive life outcomes, such as higher SAT scores, better social skills, and lower body mass index (BMI). These findings have influenced both scientific thought and public policy, with early childhood interventions often designed to enhance self-control and delay of gratification, under the belief that these traits are critical for later success." (link)
Columbia University 's Tyler Watts has been at it for a number of years. The results are sobering. In the most recent publication with his colleagues, the claims were found, once again, to be not just exaggerated but plainly false. After controlling for factors like family background and early childhood environment, the correlations between waiting time and adult outcomes disappeared, except for BMI. Even "obvious" outcomes, such as impulsive and risk-taking behaviors, did not correlate. More generally, early traits -and early interventions, have short-lived correlations to outcomes.
Having said that, this does not take anything away from the wonderful late Susan Wojcicki, one of the children from the Palo Alto preschool selected for the test, who was, according to her mother, the one kid able to delay gratification the longest.