Textual Tree (Prolog) Tracer: An Experimental Evaluation
Mukesh J. Patel, Chris Taylor, Benedict du Boulay
Abstract: We re11ort the findings or the effect of Prolog trace outputs' format and information content on simple Prolog problem solving performance of novice Prolog programmers. In this study trace outputs based on Transparent Prolog Machine (TMP*), Spy (based on Byrd Box) and Textual Tree Tracer (TTT) arc evaluated for effectiveness in providing information about Prolog program executions; the last one is a new (prototype) tracer developed at The University or Sussex which was designed to partly overcome some of the shortcoming of other tracers including those evaluated in a previous similar study (Patel, du Boulay and Taylor, 1991a). Subjects (n=13) solved simple Prolog programming problems with the aid of a trace output from one of the tracers. The results show that there was little overall difference between TTT and TPM* based trace outputs. Analysis of responses reveal relative strengths of TTT and confirm weaknesses of TMP* trace outputs observed in a similar previous study (Patel et al 1991c). Implications of similarities and differences of findings from both studies are discussed, as are issues related to the effect of format (or notation) on access to information and subjects' comprehension of Prolog.