Training methods naturally employed by trainers were analyzed and compared to systematic structured training procedures. Trainers were observed teaching retarded subjects how to assemble a bicycle brake, roller skate, carburetor, and lawn mower engine. Trainers first taught using their own (personal) method, which was recorded in terms of types of prompts given, timing of prompts (before or after responses), and type of task sequencing (whole task or chaining methods). Trainers were then required to teach by using either a Structured Whole Method (whole task presentation with prompts given for error correction) or Backward Chaining (last parts taught first using prompts to cue subject responding). Three trainers' personal methods consisted of whole task presentation, one trainer used forward chaining, and two trainers used a combination of whole and chaining. The most effective procedure was the Backward Chaining Method, followed by the forward chaining procedure used spontaneously by one trainer, and the combination of whole and chaining methods. The least effective procedure was the Structured Whole Method. Results indicated that the effectiveness of training procedures was due to both the timing of prompts and type of task sequencing. (Author)