The authors argue that the current classification system is based on an inmate's sentence length and that such an indicator provides little or no predictive value for inmate misconduct; therefore, it is not meeting the policy objectives assigned to classification and risk assessment. The authors used a large data set (n=13,000) based on incident reports from 1992 to 1994 to test the hypothesis that the length of an inmate's sentence is positively related to the rate of serious misconduct. The measure used as the dependent variable was the inmate's cumulative serious infraction score. This variable was a measure of both the frequency of infractions committed by the inmate over his incarceration period and the severity of each infraction as weighted by the institution. A number of control variables were also included to assess whether other factors besides sentence length were related to the inmates' infraction scores. These variables included age, high school education, employment stability, marital status, time spent in prison, and current custody level. The regression analysis found that the strongest predictive variables for misconduct were age and whether the prisoner had graduated from high school; length of sentence had almost no predictive value and usually showed a negative relationship with infraction and assault rates. The study also found that the predictive value of the current system of classification is limited but that the system is able to adjust by including prior positive and negative behavior of the inmate. This paper does not recommend that California's entire inmate classification system be dismantled, but rather that some of the underlying factors that are used at inmate admission be reassessed. 5 tables, 4 notes, and 17 references