Many experiments are not reliable

The middle segment (from 23 minutes) of this great Slate discussion considers findings that many experimental results in the Human Sciences, and also in some branches of medicine, have proven not to be reproducible. One of my favourite results, about the reduction in willpower when short-term memory is in use, was not repeated when other experimenters tried it. The panel discusses sample size, experimental methods and the cultures of different scientific and social science disciplines.
I can embed the audio. It's at this link (click below the picture). Start at 23 minutes.

Slate podcast link