Two methods of identifying language disordered children are examined. Traditional approaches require attention to relatively superficial morphological and surface syntactic criteria, such as, noun-verb agreement, tense marking, pluralization. More recently, however, language testers and others have turned to pragmatic criteria focussing on deeper aspects of meaning and communicative effectiveness, such as, general fluency, topic maintenance, specificity of referring terms. In this study, 54 regular K-5 teachers in two Albuquerque schools serving 1212 children were assigned on a roughly matched basis to one of two groups. Group S received in-service training using traditional surface criteria for referrals, while Group P received similar in-service training with pragmatic criteria. All referrals from both groups were reevaluated by a panel of judges following the state determined procedures for assignment to remedial programs. Teachers who were taught to use pragmatic criteria in identifying language disordered children identified significantly more children and were more often correct in their identification than teachers taught to use syntactic criteria. Both groups identified significantly fewer children as the grade level increased.