Day and Kasperczyk (1985) and Day, Watson, and Jolly (1986) claimed that the Poggendorff illusion, in a Tolansky-type vertical bisection task, is no smaller when the line transversals are replaced with dots and the oblique parallels are removed, despite other evidence that transversals and parallels contribute significantly to the illusion. Wenderoth, O'Connor, and Johnson (1986) showed indirectly that both factors are important by instructing subjects to avoid a misunderstanding of the task, which was inferred to have occurred in the experiments of Day and his colleagues. In Experiment 1 of the present study, we demonstrated directly that this misunderstanding occurred when Day's instructions were used. In Experiment 2 we used a different, matching method for measuring standard Poggendorff effects, a method that avoids possible misinterpretations of the bisection task. In Experiment 3 we used this method to show directly that both line (vs. dot) transversals and oblique parallels (vs. no parallels) significantly increase illusions. The method of measurement used here should prove to be a useful tool in testing and placing constraints on theories of the Poggendorff effect.