Differential effects of three training procedures to teach sight words to 13 children with moderate to mild mental retardation were investigated in an alternating treatments design. Number of correct responses was assessed during probe sessions in word-alone (word was presented without picture), integrated-picture (word was presented with integrated picture, no fading), and picture-fading (integrated picture was faded out) conditions. Results show that most children learned to identify sight words fastest in the word-alone condition. Effects were largely maintained during follow-up at 2 to 5 weeks after training.