The last fifty years have seen rapid growth in student enrollment in online courses. However, few systematic investigations have been utilized to identify best practices in online education experimentally. Skinner (1958) laid out a science of teaching derived from the principles of operant conditioning, and methods for adopting programmed instruction into the evolving technology of his time. In what he termed a "Teaching Machine," automated instruction programmed contingencies for the student with self-paced, carefully designed sequences towards mastery of the material. This series of investigations evaluated the efficacy of programmed instruction in online courses, as measured by quiz performance, the frequency of discussion posts, instructor time commitment, generalization, maintenance, and student perceptions of the online modalities used. The online classrooms were all conducted through Adobe Connect Meeting Software (2017) to include both asynchronous and synchronous online arrangements. Experiment 1 compared the effects of on-campus delivered lectures and online delivered lectures on weekly quiz performance, percentage correct on within assessments forms, the frequency of questions asked, participant preference, and generalization measures. Experiment 2 compared the effects of lectures delivered exclusively online and module packets, designed with components of Skinnerian programmed instruction, on weekly quiz performance, instructor time commitment, participant preference of both experimental conditions, and generalization measures. Experiment 3 compared the effects of online lectures + discussion and module packets + chat on weekly quiz performance, participant preference, and generalization measures. With the increasing demand for university courses delivered exclusively online, results are discussed on the viability of automated, programmed instruction to teach course material exclusively online.