THE NATURE OF BLOCK GRANT PROGRAMS LIKE LEAA'S IS SUCH THAT EVALUATORS' ATTEMPTS TO ACHIEVE PROGRAM ACCOUNTABILITY CAN CREATE FEDERAL-STATE TENSIONS IF THEY ARE REGARDED AS AN ENCROACHMENT ON STATES' DECISIONMAKING AUTONOMY. EVALUATORS MUST ALSO DEAL WITH THE DIFFUSE QUALITY OF THE LEAA BLOCK GRANT PROGRAM, IN WHICH LARGE NUMBERS OF RELATIVELY SMALL GRANTS ARE DISPERSED BOTH GEOGRAPHICALLY AND THROUGHOUT THE CRIMINAL JUSTICE SYSTEM AND PRIVATE SECTOR. BUT MORE SIGNIFICANT THAN THE STRUCTURE OF THE GRANT PROGRAM IS THE PROBLEM OF CONCEPTUALIZING WHAT IT IS THAT LEAA SHOULD BE DOING. FOR EVALUATION PURPOSES, GLOBAL STATEMENTS OF GOALS MUST BE TRANSLATED INTO A SYSTEM OF OBSERVABLE MEASURES OF CHANGE, AND THIS PROCESS IS FAR FROM TRIVIAL EVEN WITHIN A SINGLE PROGRAM AREA. LEAA STARTED OUT WITH A CLEAR UNDERSTANDING THAT ITS GOALS WAS CRIME CONTROL--REDUCING STREET CRIME. BUT THE GOAL HAS CHANGED TO 'SYSTEM IMPROVEMENT,' AND VERY LITTLE IS KNOWN ABOUT THE MOST BASIC ASPECTS OF THE SYSTEM THAT IS TO BE IMPROVED--ITS INTERNAL DYNAMICS OR ITS EFFECTS ON CRIME. IN ADDITION TO POLITICAL AND CONCEPTUAL PROBLEMS, LEAA EVALUATORS HAVE DATA PROBLEMS STEMMING FROM THE FACT THAT CRIMES ARE DEFINED DIFFERENTLY IN DIFFERENT JURISDICTIONS, AND THAT ELEMENTS OF THE CRIMINAL JUSTICE SYSTEM DO NOT WORK FOR THE FEDERAL GOVERNMENT AND ARE NOT OBLIGED TO PROVIDE EVALUATORS WITH DATA. A BRIEF DESCRIPTION OF THE OFFICE OF EVALUATION'S STRUCTURE AND FUNCTIONS IS ALSO PRESENTED. (LKM)