Opinion is basic to long-range developmental forecasting. The difficulties (such as the influence of dominant individuals, noise, and group pressure for conformity) of obtaining a group opinion through traditional face-to-face interaction led to the development of the Delphi procedures, which are described in this paper. The characteristics of these procedures — anonymity, iteration with controlled response, and statistical group response — derive from RAND's discovery that simply the average of individual opinions, without discussion, tends to be more accurate than group opinion resulting from discussion. The experiments that led to these results involved almanac questions, such as, how many votes did Kennedy receive in the 1960 Presidential election in Texas? An initially wide range of answers were found to gradually converge, improving in accuracy in the majority of cases; the pattern of responses resembled a log-normal curve. Further studies will attempt to dampen the effect of group pressure while amplifying accuracy.