THE MAJOR FOCUS OF THIS REPORT IS AN EXAMINATION OF THE LIMITATIONS OF PREDICTION INSTRUMENTS. AS ILLUSTRATIONS, TWO RECIDIVISM STUDIES OF ADULT OFFENDERS ARE PRESENTED. THE FIRST STUDY EXAMINED 15,000 PERSONS PAROLED OR DISCHARGED FROM WASHINGTON STATE CORRECTIONAL FACILITIES DURING THE 10-YEAR PERIOD FROM 1965 TO 1976. EACH YEAR'S POPULATION WAS EXAMINED IN TERMS OF THE RELATIONSHIP BETWEEN RECIDIVISM AND VARIABLES SUCH AS INSTITUTION OF RELEASE, AGE AT COMMITMENT, ETHNIC GROUP, AND ALCOHOL GROUP. IT WAS FOUND THAT OVER THE 10-YEAR PERIOD RECIDIVISM RATES WERE REDUCED. THIS FINDING IS ATTRIBUTED PARTLY TO THE ESTABLISHMENT OF ONSITE HEARINGS. IT IS REASONED THAT HEARINGS IN THE CORRECTIONAL FACILITIES MAKE IT EASIER TO REVOKE PAROLE FOR OFFENDERS. THE GREAT INCREASE IN THE NUMBER OF DRUG OFFENDERS AND THE LOW RETURN RATE OF THESE NEWER DRUG OFFENDERS ALSO ACCOUNTS FOR THE DECREASED RECIDIVISM RATE. THE SECOND STUDY HAD AS ITS SAMPLE 4,709 PERSONS RELEASED FROM WASHINGTON STATE FACILITIES OVER THE 3 YEARS FROM 1973 TO 1976. THE MAJOR FINDING OF THIS STUDY WAS THAT OFFENDERS WITH HISTORIES OF ALCOHOL ABUSE HAD HIGHER FAILURE RATES ONLY DURING THE FIRST 12 TO 24 MONTHS FOLLOWING RELEASE. THE EXAMINATION OF PREDICTION STUDIES LED TO THE FORMULATION OF FOUR CONCERNS REGARDING THE USE OF STATISTICAL PREDICTION INSTRUMENTS. THE FIRST CONCERN IS THAT ADEQUATE TIME BE GIVEN FOR FOLLOWUP STUDIES. FOLLOWUP PERIODS OF LESS THAN 2 YEARS SHOULD BE VIEWED CRITICALLY. SECOND, IT MUST BE REALIZED THAT STATISTICAL SIGNIFICANCE IS NOT ALWAYS SYNONOMOUS WITH PRACTICAL SIGNIFICANCE. THIRD, DIFFERENTIAL SELECTION OF OFFENDERS MUST BE TAKEN INTO ACCOUNT WHEN COMPARING GROUPS. FOURTH, BASE EXPECTANCY SCORES SHOULD ONLY BE USED FOR RESEARCH AND LIMITED ADMINISTRATIVE PURPOSES. A LIST OF REFERENCES IS PROVIDED. (MLC)