Fifty-fifty=50%?

Several recent surveys have asked respondents to estimate the probabilities of relatively unlikely events, such as dying from breast cancer and smokin g. Examination of their response distributions reveals a seemingly inapprop riate 'blip' at 50. The two studies reported here indicate that it reflects a response artifact associated with open-ended probability scales. The bli p vanishes when a response scale with explicit response options is offered. Apparently, the open-ended format leads some people to use the 50% option as 'fifty-fifty', an expression of having no idea as to the answer. As a re sult, the accuracy of people's reported beliefs depends on the response sca le used, as well as on how it evokes and channels such feelings of epistemi c uncertainty. (C) 1999 John Wiley & Sons, Ltd.