DISPLAYING UNCERTAIN INFORMATION ABOUT PROBABILITY: EXPERIMENTAL EVIDENCE