It has been said often enough now that it has the weight of fact: a relative handful of "career" criminals are committing such a disproportionate number of offenses that we could substantially reduce the crime rate just by locking a few hundred of them up.

The theory, which may actually be true, is appealing on a number of counts. To begin with, almost anything is appealing that promises a substantial reduction in crime. Even more so is anything that promises to reduce crime without the necessity of massive outlays for new prisons or courts or police officers. And anything that promises to do it by cracking down on the irredeemable career criminals, without jamming the jails with casual crooks capable of rehabilitation, can fill even a cynic with hope.

The snag is: how do you tell the irredeemably bad from the salvageable? How do you tell which few hundred to keep in prison for a long time and which are more suitable for lighter punishment? At least that used to be the snag. Now the Rand Corp., in two studies funded by the National Institute of Justice, says it can separate the wayward sheep from the intractable goats--the "violent predators," as Jan and Marcia Chaiken call them in their report.

The Chaikens said those who should be considered for long-term incarceration are usually involved repeatedly in three "defining" crimes--robbery, assault and drug dealing -- although they may also have engaged in burglaries, thefts and other property crimes. In the other study, Peter Greenwood used the same data used by the Chaikens to devise a scale to identify the high-rate burglars and robbers -- the ones whose incarceration would render the streets measurably safer. Greenwood's checklist asks whether a particular criminal has:

1) Been imprisoned more than half of the two-year period preceding the most recent arrest; 2) a prior conviction for the crime; 3) a juvenile conviction prior to age 16; 4) Been committed to a state or federal juvenile facility; 5) used heroin or a barbiturate in the two years preceding the current arrest; 6) used the same drugs as a juvenile, or 7) been employed for less than half of the preceding two-year period.

Racking up four or more "yes" answers identifies a subject as a high-rate criminal -- one of the "violent predators."

Greenwood said his review of California data demonstrates that "a selective incapacitation strategy that reduced terms for low (zero to 1 point on his scale) or medium-rate (2 or 3 points) robbers, while increasing terms for high-rate robbers, could achieve a 15 percent reduction in the robbery rate with only 95 percent of the current incarcerated population level for robbery."

Unfortunately, things aren't quite that simple. The authors of the two studies caution that inadequate record-keeping or faulty analysis might lead to the long-term imprisonment of relatively innocuous offenders. But the problems go deeper than that. How can a judge justify differential sentences for two offenders because one of them scores higher on some sociological profile? How can he add time to an adult's sentence because of an earlier juvenile offense, or because the convicted person had been unemployed for 13 months? Despite its obvious attractiveness, the approach runs head-on into one of the basic tenets of the criminal justice system: that offenders should be punished for what they do, not for what they are.