The present study utilized meta-analytic procedures to estimate the diagnostic validity of instruments used to screen young children, ages 1.5-5 years, for autism. Five scales met inclusion criteria, and data from 18 studies contributed the meta-analysis. Results revealed that 4 of 5 scales met criteria for "good" validity, including two broad band scales (instruments not restricted to screening for autism). The current results suggest that validity differences might be a function of how instruments sample across the DSM content domains. Specifically, high validity instruments included a higher proportion of items assessing social interaction skills. The availability of valid broad- and narrow-band instruments, as well as implications for constructing future screening instruments, is discussed.