Behavior Modeling Training and Generalization: Interaction of Learning Point Type and Number of Modeling Scenarios

Article excerpt

Two decades ago, Goldstein and Sorcher (1974) introduced behavior modeling as a way to train supervisors in effective interpersonal and management skills. Behavior modeling training (BMT) consists of observation and reproduction of a sequence of new behaviors to be learned by watching another person engage in that sequence of behaviors. Because of its usefulness in training complex behavioral routines, BMT has become one of industry's most widely used training methods (Decker & Nathan, 1985). This training method is grounded in Bandura's (1977) social learning theory and is described in terms of four component processes: attention, retention, reproduction, and motivation to transfer. Baldwin (1992) summed up the goals of BMT as "hav[ing] people observe a model, remember what the model did, do what the model did, and finally, use what they have learned on the job" (p. 147).

Over the years, behavior modeling training has been subjected to rigorous evaluation studies in a variety of settings (Baldwin, 1992; Burke & Day, 1986; Latham & Saari, 1979; Meyer & Raich, 1983). In terms of Kirkpatrick's (1967) traditional categories for evaluating training programs (e.g., reactions, learning, behavior change, and organizational outcomes), BMT has been shown to result in both positive trainee reactions and learning (Robertson, 1990). In terms of behavior changes and organizational outcomes, BMT has produced mixed results (McGehee & Tullar, 1978; Russell, Wexley, & Hunter, 1984). Unfortunately, the optimal approach to ensuring this transfer of training to the workplace (generalization) is not yet determined (Baldwin, 1992; Clark & Voogel, 1985; Decker, 1984).

The most robust finding, in terms of a behavioral outcome, is that BMT leads to the faithful reproduction of modeled behavior (Decker, 1980, 1984). More problematic is the delineation of those variables that result in generalization of appropriate responses to similar stimuli (Baldwin, 1992). In this context, generalization entails the derivation of a general relationship between categories of stimuli and categories of appropriate responses. Here, the trainee must more than just "reproduce" the behavior. Indeed, the trainee must develop a generalized relationship for linking a stimulus set to an appropriate response repertoire.

Decker (1980, 1982) demonstrated that generalization within the training setting can be enhanced by accompanying the modeling display with written descriptions of the key behaviors, which serve to cue the trainee as to what behaviors are most important, called learning points. Decker (1984) compared the differential effectiveness of three learning point types: behavioral (a detailed behavioral description of the modeled behavior), summary label (a short, global description of key behaviors), and rule code (a description and rationale for key behaviors). Results indicated that the use of summary labels and rule codes significantly improved generalization over both behavioral learning points and a control condition; rule codes were not significantly better than summary label learning points.

These results shed light on the generalization process. The failure of behavioral learning points to improve generalization was not surprising. Behavioral learning points help focus the trainee's attention, but do not aid in retention because they do not help the trainee to develop a general rule for linking a stimulus set to an appropriate response repertoire. Strictly behavioral learning points are too stimulus specific to give a clue about a higher order generalization rule. But both summary labels and rule codes, by focusing on categories of stimuli and describing how to respond to them in more abstract terms, orient the trainee toward thinking in more general terms. This predisposes the trainee to begin deriving the rule that governs the generalization process.

Simply knowing a rule, however, may not be enough to produce generalization; if it were, an instruction booklet would suffice. …