The paper summarizes preliminary results from a study of practices with language minority (Hispanic) students referred for special education in large urban districts. A data collection instrument was developed to record information on such topics as student characteristics, family background, referral, assessment, individualized education program (IEP) information, and language assessment data. General characteristics of the referral process in one district (background, language usage, grades, current placements, IEP team participation, and group differences) are reported. A path model was developed using causal modeling techniques to examine possible factors leading to diagnoses of either learning disabilities or language impairment. Five variables in the model were found to have a direct effect on placement category after the effect of all the other variables had been controlled: sex, language scale, psychometric team, IQ test administered, and age. Findings revealed a surprising frequency of the language impaired designation. In general, findings were replicated in a second district. (CL)