FAILURE RATE ANALYSIS HAS BEEN USED EXTENSIVELY IN THE ENGINEERING APPLICATION OF PROBABILITY AND STATISTICAL THEORY TO EQUIPMENT RELIABILITY PROBLEMS AND IN BIOMEDICAL SURVIVAL STUDIES. STUDIES INCLUDE STOLLMACK AND HARRIS (1974) ON THE POWER OF FAILURE RATE TECHNIQUES FOR EVALUATING RECIDIVISM DATA; TURNBULL (1977) ON THE USE OF NONPARAMETRIC METHODS TO ANALYZE RECIDIVISM DATA; AND THE CONNECTICUT PAROLEE DATA STUDY WHICH ILLUSTRATES THE POTENTIAL USEFULNESS OF THE FAILURE RATE REGRESSION MODEL. THE SAMPLE FOR THE STUDY CONSISTED OF 37 AND 71 MAXIMUM-SECURITY MALE OFFENDERS PAROLED FROM TWO CORRECTIONAL INSTITUTIONS IN CONNECTICUT. INFORMATION WAS THEN COLLECTED FROM FILES KEPT BY EACH PAROLE OFFICER. THE DEPENDENT VARIABLE MEASURED WAS TIME FROM RELEASE UNTIL FIRST ARREST. USING THE MANTEL-HAENSZEL TEST AND THE LOG-LIKELIHOOD RATIO TEST BASED ON THE COX MODEL, A CONSIDERABLE DIFFERENCE APPEARED BETWEEN THE ARREST RATES FOR THE TWO INSTITUTIONS. HOWEVER, AFTER ADJUSTING VARIOUS COVARIATES, THE DIFFERENCE BETWEEN INSTITUTIONS APPEARED FAR LESS SIGNIFICANT. ALTHOUGH THIS STUDY WAS NOT INTENDED TO THOROUGHLY ANALYZE THE DATA DUE TO THE STUDY'S SMALL SAMPLE SIZE, FURTHER NONPARAMETRIC STUDIES SHOULD ENABLE THE RESEARCHER TO CHOOSE AN APPROPRIATE PARAMETRIC MODEL TO INCREASE THE SMALL SAMPLE POWER OF TESTING PROCEDURES. REFERENCES AND TABULAR DATA ARE INCLUDED. THE COX MODEL IS APPENDED. (WJR)