Objective: The National Research Council identified inadequate research definitions for abuse and neglect as barriers to research in child maltreatment. We examine the concordance between child protective services (CPS) classifications of maltreatment type with the determinations of type from two research coding systems. We contrast the two coding systems and the CPS classification, in their ability to predict subsequent difficulties in the psychological functioning of maltreated children at age 8. Method: The sample included 545 children who were enrolled in Longitudinal Studies of Child Abuse and Neglect (LONGSCAN) with a report of child maltreatment, had data collected at approximately 4 and 8 years of age, and had a lifetime review of CPS records to age 8. CPS Maltreatment reports were coded using LONGSCAN's modification of the Maltreatment Classification System (MMCS) and the Second National Incidence Study maltreatment coding system (NIS-2). The first analyses used reports as the unit of analysis to examine agreement between CPS and research determinations of allegation type. Validation analyses examined outcomes for each type of maltreatment experienced after age 4 under each coding system using the Child Behavior Checklist (CBCL), Trauma Symptom Checklist-Alternative form, and the Vineland Screener as the measures of outcome. Control variables were the CBCL and Battelle Developmental Screener, measured at age 4. Results: There were a total of 1980 reports of maltreatment for 545 study children although only 1593 CPS reports specified at least one type of maltreatment. There were differences between the type of maltreatment recorded in child protective service records and the conclusions reached by either research classification system. CPS classifications were most discordant with the research systems for emotional abuse and neglect. Nearly 10% of physical and sexual abuse reports, as determined by the MMCS, were classified as neglect by the child protective service agencies. The NIS-2 system and the MMCS had very high Kappa statistics for agreement for physical and sexual abuse. The validity of the research definitions for physical and sexual abuse was demonstrated in models predicting children's functioning at age 8. Prediction of child functioning was significantly but modestly improved in several domains compared to the CPS classifications. Conclusion: Both research classification systems moderately improved on the prediction of the adverse effects of maltreatment compared to the characterization of a maltreatment exposure as recorded by CPS.