This study relates individual and organizational characteristics to programmer performance and satisfaction. The study displays many of the problems that have been characteristic of research on programming reported in the industrial psychology and organizational behavior literatures. Most of the data reported are subjective (i.e., collected on questionnaires). The author displays little knowledge in either the paper or its references of the behavioral or measurement research on programming. As a result, the conclusions rest on weak measurement techniques. For instance, programmer productivity was measured as a supervisor’s rating on a 5 point scale. This measure does not even take advantage of current performance scaling techniques developed by industrial psychologists for removing some of the subjective bias in supervisory ratings [1].

Because of measurement problems, it is difficult to know how reliable some of the results are. For instance, the failure to find that mathematical aptitude correlates with programmer performance may well result from any of three measurement problems. First, the IBM Programmer Aptitude Test is, at best, a crude measure of mathematical aptitude (and the author admits he may be stretching a point). Second, no job analysis has been performed to separate the myriad tasks loosely clustered as “programming.” The 149 participants in this study may have been doing very different types of jobs. Thus, those aspects of programming which should be theoretically related to mathematical aptitude are mixed with those aspects which may not be. Finally, supervisory ratings are too subjective and fail to distinguish between the myriad aspects of programming performance.

The author concludes that one of his most important findings is that experience (defined as number of years employed as a programmer) positively affects productivity. This result contradicts results, replicated several times by Sheppard and her colleagues [2] and others, that length of experience was not related to performance, although breadth of experience was. Cheney’s results may be accurate in the organizations where his data were collected, or they may reflect supervisors’ tendency to rate experienced employees more highly.

The results relating job perceptions to supervisory ratings and satisfaction are interesting. However, since the author only presents significance levels rather than the percent of variance accounted for, it is difficult to determine the practical significance (size of the impact) of his results. For important relationships to be explained in terms relevant to software engineering, greater effort needs to be spent in defining appropriate measures than was evident in this study.