Evaluating curricula for the teaching of programming

• Authors: Ronald G. Ragsdale
• Venue: ACM SIGCUE Outlook - Topics in instructional computing
• Year: 1978

Despite the great and continuing interest in methods of effective teaching of computer science concepts, especially programming, there is little hard data to back up the opinions that are so vigorously expressed. Moreover, most of those studies that have attempted to investigate programming instruction have been sufficiently flawed so as to render them of little use. The first section of this paper deals with many of the flaws commonly found in studies of instruction in programming. These include non-random assignment of subjects, assuming identical instruction to different groups, experimenter bias, use of "available resources", short experimental treatments, home grown tests, and questionable analysis techniques. The second section contains a description of the principles of good experimental design, using the concepts of internal and external validity. An experiment usually requires a compromise between the internal validity, which relates to control of experimental conditions, and the external validity, which has to do with the generalizability of the results. The final section contains conclusions concerning initial steps that might be taken to rectify the problem caused by a lack of valid data. These steps include better and more objective quantification of the attributes of good programs and more integrative techniques for analyzing the results of multiple studies.