The first three samples used for the survey data analyses of the training evaluation were subsets of recent (1976-77) basic training graduates. The representativeness of these samples was assessed in relation to a population of recruits trained annually (1976-77). Three other samples were compared to State agency distributions. The three samples compared to the annual sample of recruits are respresentative of the annual sample on agency type and size, but recruits from more than 150 miles from the metro area are disproportionately represented. These samples underrepresent the smallest communities, tend to overrepresent regions E and 10, and tend to underrepresent regions D and 9. Three samples compared to statewide agency distributions appear not to be representative of the State on all variables. One finds more sheriffs' offices, region G (metro area) agencies, communities over 1,000, and agencies of five or more persons represented than would be expected on the basis of statewide population data. Some of the differences can be explained by the small community exemption which would eliminate more police agencies, outstate agencies, small agencies, and small communities from samples derived from basic training graduates. Thus, samples would not be so unrepresentative of agencies required to train their personnel as they are of all State agencies. Since exempt agencies can train their personnel and might someday be required to train them, opinions from exempt agencies which have been largely excluded from the various samples are still useful to obtain. Hence, the control group deliberately was selected from a set of agencies not required to train their personnel; it thus taps opinions and perspectives of those in small outstate agencies and communities. The biases identified can be noted or controlled when they are likely to affect statistical results. Tabular data are provided. (Author summary modified)