How good are psychiatrists at predicting future mental problems? A 35-year study concludes that they may underestimate people's abilities to change for the better.

Between 1946 and 1949, psychiatrists examined 230 first-year medical students and attempted to predict how they would fare in school and in life. In 1980 and 1982, Dr. C. Knight Aldrich of the Virginia School of Medicine tracked down 196 of them to check out the predictions.

In general, the psychiatrists of the 1940s were better at predicting success than failure. Of those given a "good prognosis" in the 1940s, 95 percent finished medical school, and, 35 years later, 95 percent had no serious emotional problems.

Predictions of trouble were less accurate. Even of those given a "poor prognosis," 72 percent finished medical school, and 74 percent were free of serious problems 35 years later.

"Psychotherapy did not appear to affect the outcome," Aldrich reported in the American Journal of Psychiatry. "As many did better than predicted without psychotherapy as with psychotherapy."

Are predictions today better than they were in 1946? That won't be known for another 35 years, Aldrich wrote. But he adds: "Prediction in psychiatry . . . may resemble prediction in the stock market, in that too many variables exist for it ever to be accurate in individual cases."