These investigations were conducted between 1983 and 1985 in the states of New York, New Jersey, Pennsylvania, Florida, and Texas. Names of approximately 24,000 migrant children obtained from annual migrant education censuses were individually cross-referenced with the appropriate State data bases to determine if they had been involved in a confirmed incident of maltreatment. The information acquired was converted to incidence estimates that were contrasted with the rates for all children in the respective states and were decomposed to identify high-risk cohorts within the migrant population. One finding common to all five assessments was that migrant children were significantly more likely to be maltreated than other children, although these incidence rates varied appreciably from one state to another. The emphasis of this paper is on the unique methodology employed in the research, issues pertaining to provisions for accessing central registers and protecting confidentiality of subjects, the generalizability of the findings, and cross-state incidence differentials for both migrants and children from nonmigrant families. 4 tables, 1 figure, 19 references. (Author abstract)