Comparison of contrasting Prolog trace output formats

This paper reports on a comparative study of three Prolog trace packages. Forty-three students of an introductory Prolog course solved five different Prolog programming problems in each of three different conditions (using isomorphic problem variants to disguise recurring tasks). Each of the three conditions provided subjects with static screen-snapshot-mockups derived from one of three different trace packages (“conventional” Spy; “graphical AND/OR tree-based” TPM*; “informative textual” EPTB). When traces explicitly displayed the information asked for in the problem, subjects solved the problems more quickly. Conversely, when trace output obscured the required information (or necessitated difficult detective work to uncover the information), solution times were longer and answers less accurate. Deciding on a “good” format for display is thus a task-dependent decision, and impacts directly on the user's cognitive ability to solve a problem.