There is a consensus among social skills trainers and researchers that broad intervention effects, such as response generalization, are both attainable and desirable, yet recent reviews of the literature have revealed that little is known about collateral effects of procedures designed to effect social interaction. A multiple baseline, multiple probe design was used to evaluate the effects of a social skill intervention for three elementary school age children with behavioral and learning disor ders. An observation system utilizing time-sampling and duration recording was developed to assess the acquisition of target social behaviors, the concomitant changes in nontargeted behaviors, and changes in peer and teacher responses to target students as a result of the training package. The results indicated success in increasing student acquisition of targeted social behaviors which in turn led to lengthier peer interactions for all three students. Additionally, all three target students demon strated increases in nontargeted responses. The results are discussed in light of social skill training, response covariation, collateral effects, and generalization. I Students with behavioral disorders often demonstrate difficulties in interpersonal relation ships, particularly peer interactions (Gresham, 1988). For most children with behavioral disor ders, the lack of adequate social skills coupled with the presence of interfering behavior, such as aggression, appears to be the foundation for peer interaction problems. The bulk of social skill intervention research has focused on dem onstrating that students with behavioral disor ders can acquire adequate social skills and re duce their display of behaviors that interfere with positive peer interactions (Asher & Coie, 1990; Gresham, 1990; McConnell, 1987). While the social skills literature documents that the peer interactions of students with so cial problems improve following social skills training, less evidence exists that these gains generalize or result in a substantial improve ment in the student's overall social competence (Chandler, Lubeck, & Fowler, 1992; Gresham, 1988; Kohler & Greenwood, 1986; Mahur & Rutherford, 1991 ; McConnell, 1987; Newcomb, Bukowski, & Pattee, 1993). It is often difficult to determine if the lack of evidence for positive long-term effects of social skills training is miss ing due to a lack of treatment integrity (Gre sham, Gansle, & Noell, 1993), the strength of the treatment package (Stokes & Osnes, 1989; Strain, 1993), or limited knowledge regarding the durability of trained skills in the natural en vironment once the intervention has been with drawn (Kohler & Greenwood, 1986; McCon nell, 1987; Voeltz & Evans, 1988). Review of social skills training literature re veals numerous attempts to assess strategies for programming generalization. Chandler, Lubeck, and Fowler (1992), for example, examined var ious social skill generalization strategies with withdrawn preschoolers and found that al though generalization strategies frequently are incorporated into social skills training, the ef fects often are not evaluated. Schloss, Schloss, Wood, and Kiehl (1986) in a comprehensive review of 25 studies, examined the effects of social skills training with students with behav ioral disorders. Nearly half of the studies dem onstrated lasting results during a maintenance assessment. However, the type of generalized responding most often reported was stimulus generalization (i.e., across persons, environ ments, events). Very few demonstrations of re Behavioral Disorders, 1994, 20(1 ), 35-50 35 This content downloaded from 207.46.13.57 on Mon, 08 Aug 2016 05:02:47 UTC All use subject to http://about.jstor.org/terms sponse generalization (i.e., across related be haviors) have been reported. Nonetheless, both sets of authors concluded that programming for generalization of social skills is limited and that the generalization effects of most treatments remain unknown. In 1991 Mahur and Rutherford examined the efficacy of peer-mediated interventions by analyzing the effects of 21 social skill interven tion studies. Within the peer-mediated social skills training literature, these authors found that certain treatment components, such as incor porating naturally occurring reinforcers, were more effective than others in promoting gener alized effects. These reviewers and others (e.g., Kohler & Greenwood, 1986; McConnell, 1987; Stokes, 1992; Stokes & Osnes, 1989) have con cluded that what is known about broad inter vention effects is insufficient to either predict or control for generalization. We concur with Kohler and Greenwood (1986) and Stokes and Osnes (1989) who recommend systematic in vestigation to specifically analyze generaliza tion effects. Sasso, Melloy, and Kavale (1990) examined the generalized effects of an inter vention package that included modeling, re hearsal, feedback, and use of peer confederates. The treatment increased the frequency of tar geted social responses; produced generalization across situations; and decreased nontargeted, aberrant behaviors. The generalized effect of decreased aberrant behavior led Sasso et al. to conclude that social behavior interventions may have a generalized impact on behaviors seem ingly unrelated to the target behavior. In the context of social skills training, two alternative mechanisms may underlie response generalization: response covariation (Kazdin 1982a) and collateral behavior change (Voeltz & Evans, 1988). Response covariation occurs when two or more behaviors are directly or in versely related; when one behavior changes, the other behavior changes as well (Kazdin, 1982a). Several authors (Neef, Shafer, Egel, Cataldo, & Parrish, 1983; Parrish, Cataldo, Kol ko, Neef, & Egel, 1986) have documented the direct relationship between compliance and inappropriate behavior in children with devel opmental delays. Decreased levels of aggres sion and other forms of disruptive behavior of ten have been hypothesized to covary with improved social skills (Brockman, 1985; Gre sham, 1990; Sasso et al., 1990). Haring, Roger, Lee, Breen, and Gaylord Ross (1986) investigated the behavioral interre lationship of the acquisition of conversation initiations by children with moderate disabili ties and the unexpected increase in the nontar geted skill of emitting expansions within con versations with peers without disabilities. Once the target students reliably emitted the targeted communication skills (initiations), rapid and expansive displays of nontargeted communica tion skills (conversation expansions) also oc curred. Similar effects were reported by Koegel and Frea (1993) who modified the social be havior of children with autism. In addition to generalized effects for target children, an evaluation of the multiple effects of an intervention also might determine the collateral effects for students not included in the intervention. Collateral behavior change refers to increases in nontargeted positive ef fects associated with the acquisition of certain critical behaviors (Voeltz & Evans, 1988). Sas so and Rude (1987), for example, developed a peer-initiation package for students with severe disabilities. Of interest was the result that there were substantial increases in initiations to tar get students by untrained peers without disabil ities. The challenge in evaluating the multiple effects of social skills interventions has both pragmatic and methodologie considerations. First, most social skills training occurs in the classroom setting and not typically in situations such as recess, when lengthier peer interactions are more likely to occur (Brockman, 1985; Car tledge & Milburn, 1983; Gresham, 1988,1990; Kohler & Greenwood, 1986). Thus, the possi bility of any generalization effects being ob served may be limited. Second, many social skills training programs include multiple target behaviors, some of which do not promote peer interactions. For example, teacher-preferred social skills (i.e., adult-oriented responses) are commonly taught and may not change the lev el of peer interactions; in fact, they may decrease socialization with peers (Gresham, 1988). Even if the pragmatic concerns just noted are controlled, the evaluation of the possible multiple intervention effects remains problem atic due to the sheer number of behaviors and situations that warrant assessment. The purpose of the present investigation was to provide a descriptive evaluation of the multiple effects of a social skills intervention conducted within a special education classroom. A structured learn ing approach was employed to teach social ini tiation and sustained responding to three stu dents. The intervention included modeling, coaching, instruction, and feedback. In addi 36 Behavioral Disorders This content downloaded from 207.46.13.57 on Mon, 08 Aug 2016 05:02:47 UTC All use subject to http://about.jstor.org/terms tion to the components of the structured learn ing model, generalization training strategies (Stokes & Osnes, 1986) were used. The strate gies included teaching multiple examples of the target behavior, using common stimuli, and incorporating natural contingencies. Finally, the observation of concurrent changes in in appropriate behavior, such as aggression, en abled us to determine if prosocial responding and inappropriate behavior are responses that covary.