The authors of the Rand Corp. study meant, no doubt, to offer reassurance that youthful offenders can be rehabilitated. What they have done is to raise some scary possibilities that most of us hadn't even thought of.

Peter Greenwood and Franklin Zimring, in a study sponsored by the Justice Department's delinquency prevention office, said improper use of statistics masks the success of many rehabilitative programs. For instance, a researcher may lump together all programs using one technique, find no significant effect on overall recidivism, and conclude that the particular technique is worthless.

After a while, the impression gets about that nothing works, when in fact several programs, employing many different techniques, may be working reasonably well.

That's the first reason the asay we shouldn't give up on the possibility of rehabilitation. It's the second one they give that scares me: our new and improved ability to predict which delinquents will become chronic offenders and, eventually, adult criminals.

The more accurate our predictions, the authors argue, the more we will be able to focus scarce resources on high-risk youngsters.

Appealing, right? But just suppose that the high-risk delinquents don't respond to the rehabilitation. Isn't the next step simply to lock the little monsters up for longer stretches, sending the low-risk offenders home to make room for them if necessary?

That's how it strikes me, at least, and that is how it also seems to have struck the authors, who offer this interesting research finding: a selective incapacitation policy (doubling prison terms for delinquents rated as high risks to become chronic offenders) would result in a 6 percent increase in inmate populations and a 5 percent decrease in overall crime. Not a bad bargain when younote (as the authors do) that for a rehabilitation program to produce that same 5 percent reduction, it would have to reduce post-treatment offenses by a pie-in-the-sky 37 percent.

Any reasonably humane official would, of course, prefer rehabilitation to incarceration. But common sense suggests that those youngsters predicted to develop into hard-core offenders might also be the least likely candidates for rehabilitation. If that proves to be true, then we are likely to wind up salting youngsters away in prisons not for what they have done but for what the experts predict they will do in the future.

What are those predictions based on? According to the study, the number of arrests a youngster experiences is just one predictive criterion. Others include: biological and physical impairments and learning disorders; such pre-delinquent behavior as lying, fighting, truancy, smoking and promiscuous sex; low academic achievement and poor verbal skills.

And how accurate are the projections? Taking all these factors into account and integrating them with actual delinquent records, experts claim they can predict chronic offenders with an accuracy of some 50 percent.

That may strike Greenwood and Zimring as wonderful. What strikes me is the prospect of computer-bound experts making predictions of future criminality that will send some youngsters home to their parents and others to the slammer for "rehabilitative treatment" -- knowing that the predictions will be flat wrong about half the time.