Diagnostic comparisons of performance on parts A and B of the Trail Making Test (TMT) assume that path structure in the two parts is equivalent but that task complexity is greater for B. The two parts are shown to differ with respect to length and angular variability. However, measures of fractal dimension show no difference in structural complexity between paths A and B. This analysis suggests a principled method for generating alternative pathways, varying in complexity, and opens the way for a systematic study of path-following. It also suggests that path-following may be interpretable within a general approach, in which perceptual, linguistic, reasoning and motor processes are seen as related through different groups of geometric transformations.