Eighty-seven preschoolers with autism spectrum disorders who were initially nonverbal (under 6 words in language sample and under 21 parent-reported words said) were assessed at five time points over 16 months. Statistical models that accounted for the intercorrelation among nine theoretically- and empirically-motivated predictors, as well as two background variables (i.e., cognitive impairment level, autism severity), were applied to identify value-added predictors of expressive and receptive spoken language growth and outcome. The results indicate that responding to joint attention, intentional communication, and parent linguistic responses were value-added predictors of "both" expressive and receptive spoken language growth. In addition, consonant inventory was a value-added predictor of "expressive" growth; early receptive vocabulary and autism severity were value-added predictors of "receptive" growth.