Research Findings: We examined the reliability and validity of the language, literacy, and mathematics domains of "Work Sampling for Head Start" (WSHS), an observational assessment designed for 3- and 4-year-olds. Participants included 112 children who were enrolled over a two-year period in Head Start and a number of other programs sponsored by community-based organizations affiliated with a local school district. Teachers were trained to administer the WSHS checklist and to collect observational data about their children over the course of the children's year of enrollment. Outcome data were individually administered tests of early reading and early mathematics. Cronbach's alphas, correlations, regressions, and receiver operating characteristic curves were computed. Results indicated very high reliability of WSHS subscales. Findings also demonstrated moderate correlations between WSHS and the outcomes and unique contributions to the assessments of reading and mathematics by WSHS over and above demographic variables. The receiver operating characteristic curves showed that WSHS can be used accurately by teachers to predict children's early mathematics and reading performance. Practice and Policy: Discussion includes the role of observational versus norm-referenced tests in early childhood classrooms. Also discussed are such issues as variance in methods of assessment and the impact of high-stakes tests on young children. (Contains 6 tables and 2 figures.)