autism, it is useful to characterize the active ingredients of treatment approaches along a continuum—from traditional, discrete trial approaches to more contemporary behavioral approaches that used naturalistic language teaching techniques to developmentally oriented approaches (Prizant and Wetherby, 1998; Anderson and Romanczyk, 1999; Prizant and Rubin, 1999). The earliest research efforts at teaching speech and language to children with autism used massed discrete trial methods to teach verbal behavior by building labeling vocabulary and simple sentences. Lovaas (1977, 1981) provided the most detailed account of the procedures for language training using discrete trial approaches. Outcomes of discrete trial approaches have included improvements in IQ scores, which are correlated with language skills, and improvements in communication domains of broader measures, such as the Vineland Adaptive Behavior Scales (McEachin et al., 1993). A limitation of a discrete trial approach in language acquisition is the lack of spontaneity and generalization. Lovaas (1977) stated that “the training regime…its use of ‘unnatural’ reinforcers, and the like may have been responsible for producing the very situation-specific, restricted verbal output which we observed in many of our children” (p. 170). In a review of research on discrete trial approaches, Koegel (1995) noted that “not only did language fail to be exhibited or generalize to other environments, but most behaviors taught in this highly controlled environment also failed to generalize” (p. 23).
There is now a large body of empirical support for more contemporary behavioral approaches using naturalistic teaching methods that demonstrate efficacy for teaching not only speech and language, but also communication. These approaches include natural language paradigms (Koegel et al., 1987), incidental teaching (Hart, 1985; McGee et al., 1985; McGee et al., 1999), time delay and milieu intervention (Charlop et al., 1985; Charlop and Trasowech, 1991; Hwang and Hughes, 2000; Kaiser, 1993; Kaiser et al., 1992), and pivotal response training (Koegel, 1995; Koegel et al., 1998). These approaches use systematic teaching trials that have several common active ingredients: they are initiated by the child and focus on the child’s interest; they are interspersed and embedded in the natural environment; and they use natural reinforcers that follow what the child is trying to communicate. Only a few studies, all using single-subject designs, have compared traditional discrete trial with naturalistic behavioral approaches. These studies have reported that naturalistic approaches are more effective at leading to generalization of language gains to natural contexts (Koegel et al., 1998; Koegel et al., 1992; McGee et al., 1985).
There are numerous intervention approaches based on a developmental framework (e.g., Greenspan and Wieder, 1997; Klinger and Dawson, 1992; Wetherby et al., 1997; Prizant and Wetherby, 1998). While