Page:FBI Law Enforcement Bulletin 54 (8).pdf/19

 {| !Agent !Cases Classified !Case Agreed !Percent Agreement
 * +Figure 3 Agreement of Agents' Homicide Type Classifications With Presenter's Classification
 * 1||62||48||77.4
 * 2||40||35||87.5
 * 3||55||45||81.8
 * 4||30||23||76.7
 * 5||27||25||92.6
 * }
 * 4||30||23||76.7
 * 5||27||25||92.6
 * }
 * 5||27||25||92.6
 * }

RESULTS Type of Crime Not all participating Agents were available to classify each of the 64 homicide types (sexual, nonsexual, or unknown). In total, the 6 Agents made 285 classifications, 64 of which were made by the Agent presenting the case. Thus, there were 221 classifications that couid be used for comparison with the presenter's classifications. Of these, 180 classifications (81.4 percent) agreed with the presenter's classification.

Of the 6 Agents, 1 Agent made 57 (89 percent) of the presentations. Because the percentage of his presentations was so large, comparing his classifications with the presenter's would not be informative. The agreement rate for the other five Agents and the number of cases they classified are shown in figure 3. Given the minimal amount of information supplied by the presenter, these agreement rates are high.

When the classifications of each Agent were compared with those of any other Agent, the agreement rate ranged from 77 percent to 100 percent. Again, these are high agreement rates.

Structure of Crime Scene There were 220 classifications of the structure of the crime scene (organized, disorganized, mixed, unknown) that could be used for comparison with the presenter's classification. Of these, 163 (74.1 percent) agreed with the presenter. (See fig. 4.)

The agreement rate between any two Agents ranged from 45 percent to 89 percent. The agreement rates of Agents with the presenter and with each other varied substantially. This appears due mainly to variation in experience and involvement with the process of classifying crime scenes. The agreement rates among the three Agents routinely involved with this process ranged from 62 percent to 80 percent. Given the minimal data supplied by the presenter, these agreement rates must be considered good. However, classification in any field is a skill learned and reinforced by continuous involvement. In the medical field, for example, the diagnosis of a patient's medical condition is similarly learned and reinforced through continuous involvement.

The interrater reliability study evaluated the agreement of Agents in classifying homicide by the type of crime and by the structure of the crime scene. In particular, the classification of crime scenes as organized has proven to be useful in profiling offenders in unsolved and motiveless murders.

This study demonstrated that there is reliability in the classification of crime types and scenes by BSU Agents. Given only minimal information about the crime, agreements of Agents with respect to crime types was high (at least 77 percent). Agreement of Agents with respect to classifying the crime scene, while not as high as the crime-type agreement, appeared to be related to Agent experience and involvement in the classification process. For experienced and active Agents, who were given only minimal information about the crime scene, agreement rates ranged from 62 percent to 80 percent. More information would certainly have improved the agreement rates. FBI

Footnote Robert R. Hazelwood and John Douglas "The Lust Murderer." FBI Law Enforcement Bulletin, April 1980, p. 6.

August 1985 / 17