Although the demand for training individuals to implement discrete-trials teaching (DTT) is high, published studies on strategies to do so are few. We used a modified multiple-baseline design across participants to evaluate a training package for teaching university students to implement a 19-component DTT procedure to teach three tasks to a confederate role-playing a child with autism. Also, in an AB within-subject design with each participant, we compared the two components of the training package, a self-instructional manual and feedback plus demonstration. After an average of 2.6 h to master the training manual, participants' DTT accuracy while teaching a confederate improved from 34% in baseline to 66%. After one to three sessions of feedback plus demonstration of teaching one task, DTT performance averaged 92% on that task. Participants' DTT performance generalized to the other two tasks (mean accuracy of 94%) and to teaching the tasks to a child with autism (mean accuracy of 91%). (Contains 1 figure and 3 tables.)