REPRESENTATIVES FROM THE UNIVERSITY OF CALIFORNIA'S LAWRENCE LIVERMORE LABORATORY USED CRIME STATISTICS FROM THE SAN DIEGO POLICE DEPARTMENT TO DETERMINE HOW EFFECTIVE THE LABORATORY'S COMPUTERIZED PATTERN RECOGNITION PROGRAM (PATTER) WOULD BE IN PREDICTING THE LIKELIHOOD THAT A GIVEN CRIME WOULD BE SOLVED. THE EXPERIMENT SHOWED THAT PATTERN RECOGNITION WAS A FEASIBLE APPROACH TO CRIME ANALYSIS AND BROUGHT OUT SOME OF THE PROPERTY AND VARIABLE RELATIONSHIPS INVOLVED IN SUCH APPLICATIONS. PHASE 2 OF THE RESEARCH FOUND THAT PATTER'S PREDICTIVE ACCURACY COULD BE INCREASED BY INCLUDING MORE VARIABLES AND THAT THE PREDICTIVE CAPABILITY COULD BE USEFUL IN POLICE OPERATIONS. FOR EXAMPLE, IF PATTERN RECOGNITION METHODS CAN ESTABLISH A PRIORITY LIST FOR CASE ASSIGNMENTS, POLICE SUPERVISORS WILL BE FREED FROM A LARGE PART OF THEIR ROUTINE ADMINISTRATIVE BURDEN. A SUBROUTINE OF PATTER WAS USED TO RANK 200 ACTUAL CRIMES IN ORDER OF PROBABILITY OF SOLUTION. THE 50-PERCENT POINT (I.E., 50-PERCENT CHANCE OF SOLUTION) FELL BETWEEN CASE NUMBERS 76 AND 77. IN REALITY, 34 OF THE 200 CASES HAD BEEN SOLVED. OF THE SOLVED CASES, PATTER HAD LISTED 20 IN ITS FIRST 76. THE RELATIVE COST-EFFECTIVENESS OF SOLVING 20 OUT OF THE ADVANTAGES OF THE AUTOMATED METHOD. PHASES 1 AND 2 OF THE RESEARCH UNDERSCORED THE IMPORTANCE OF CHOOSING APPROPRIATE VARIABLES WHEN APPLYING PATTERN RECOGNITION TO CRIME ANALYSIS. IN PHASE 3, DATA COLLECTED BY THE SAN DIEGO DEPARTMENT BUT NOT NORMALLY STORED IN ITS AUTOMATED DATA-PROCESSING SYSTEM WERE EXAMINED TO DETERMINE THEIR POTENTIAL USE IN PATTERN RECOGNITION. FOR THE CRIME OF BURGLARY, THE FOLLOWING VARIABLES WERE FOUND TO CONTAIN THE INFORMATION NEEDED FOR PREDICTION: ARREST INFORMATION; NUMBER OF WITNESSES; EAST-WEST COORDINATES ON THE CARTESIAN MAP; VEHICLE INFORMATION; AND HOUR OF DAY. THE PREDICTION ALGORITHM DEVELOPED IN THE RESEARCH IS SIMPLE ENOUGH TO BE IMPLEMENTED ON PORTABLE, PROGRAMMABLE CALCULATORS.