This paper will consider the suitability of four such debuggers for neophyte Prolog programmers. We shall present an analysis of an empirical investigation into the relative benefits of: a) Spy (Byrd, 1980), b) Prolog Trace Package (PTP) (Eisenstadt, 1984), c) Transparent Prolog Machine (TPM) (Eisenstadt and Brayshaw, 1988; Brayshaw and Eisenstadt, 1991) and d) the Textual Tree Tracer (TTT) (Taylor, du Boulay and Patel, 1992). Specifically the analysis will attempt to go further than a simple comparison of performance, although this also will be given. A more detailed account will present a protocol analysis of the information content, strategies and misunderstandings typical of each notation. We know that learning Prolog has its own individual pitfalls (Taylor, 1988). Additionally, we consider the effects of the context in which the learning takes place. An a priori assumption of the study reported here is that novices will have to come into contact with some form of debugger as part of this learning process. The study will provide empirical results to show how students cope with various debuggers. The overall result of this work not only has implications for tracer design but also raises pedagogical issues about how novices cope with the learning environments we provide. 2. Related work