This thesis presents the results of two studies that investigate the question of what interruption-styles are most appropriate for end-user programmers who are debugging programs. In the studies, end-user programmers are presented with surprises that encourage them to investigate, use, and learn about debugging devices in their programming environment. We used various interruption-styles to present these surprises to the end-user programmers, and we evaluated how they affected the end-user programmers ability to learn about the debugging features, their accuracy at debugging their programs, and how accurate they were at judging how well they had debugged their programs. The three styles we compared were immediate-style interruptions (which force the user to acknowledge them), low-intensity negotiated-style interruptions (which do not force the user to acknowledge them, but rather use visual elements such as red circles around cell values in order to notify users that there is something for them to do), and high-intensity negotiated-style interruptions (which are the same as low-intensity negotiated-style interruptions except that the visual elements are more intense, e.g. they are larger and they blink). We found that low-intensity negotiated-style interruptions best supported end-user programmers debugging, learning, and self-assessment. We also found that immediate-style and high-intensity negotiated-style interruptions had very similar effects on the end-user programmers. / Graduation date: 2004

The psychological characteristics of programmers are believed to be important determinants of programming productivity. However, little evidence is available to support this contention. This investigation, motivated by the lack of such evidence, was concerned with determining the influence of the programmer's cognitive complexity (differentiation and integration) and experience on comprehending and modifying programs of different levels of complexity. Data were collected from ninty-three graduate and undergraduate students in a classroom experimental setting. In the first phase of the experiment, a background questionnaire was administered in order to collect experience and other demographic information. Also, a domain-specific Role Construct Repertory (REP) Test was administered to collect cognitive complexity information. In the second phase, the subjects were randomly assigned to either the program comprehension group or to the program modification group. Both groups used two COBOL programs of differing levels of complexity to do comprehension and modification exercises. Three sets of hypotheses were tested. The first set of hypotheses was designed to evaluate the direction and strength of the relationship between cognitive complexity and program comprehension and modification. The second set of hypotheses was designed to evaluate the combined influence of cognitive complexity and program complexity on the comprehension and modification of the programs. The third set of hypotheses was designed to evaluate the moderating effect of experience on the relationship of cognitive complexity to program comprehension and modification. Cognitive integration was found to have a significant and positive nonlinear relationship only with the relatively complex program modification scores. The subjects who were ranked high in cognitive integration performed better than those ranked low in modifying the relatively complex program; but they performed the same in modifying the relatively simple program. Cognitive differentiation was found to have no significant relationship with either comprehension scores or modification scores. Experience of the subjects did not significantly moderate the relationship of cognitive complexity and program comprehension and modification.