Test and measurement bias in special education diagnosis has been well documented. Boys, linguistic and ethnic minorities, and children with behavior problems are among those overrepresented in several handicap categories, including speech impairment. Recent evidence indicates that variables associated with test interpretation or diagnostician background (non-measurement factors) may be better predictors of diagnostic bias. This study investigated the ability of non-measurement factors to predict the diagnostic decisions made for 345 speech impaired children enrolled in Head Start. Computer Assisted Telephone Interviews and Case File Review were used to collect data about (1) child characteristics, (2) diagnostic criteria, and (3) diagnostician background. Hierarchical regression procedures were used to test the predictive power of these three blocks of variables and of specific variables within each block. The diagnostician's rating of articulation severity was the criterion variable. The test score was the best predictor of articulation severity rating; non-measurement factors were not found to be effective predictors of the articulation component of speech diagnosis. There was some indication, however, that non-measurement factors may be related to the language component of speech assessment. Further investigation of the language severity rating and of other non-measurement factors was suggested.