Wavelet modelling of collinearity judgment error

Prior research has found that participants manifest complex profiles of error when they are asked to judge collinearity of stimulus elements. These studies used harmonic analysis to model the data, and found large departures from accurate collinear judgments, with the amplitude and specific angular position of departures from valid judgment varying from one participant to the next. The models appeared to be a composite of many large‐, medium‐ and small‐scale departures, but this could not be established with any certainty because of the global nature of harmonic model components. Wavelet modelling is better suited to answer the question of whether the error profile is produced by independent sources that vary in size and location. Here, we examined judgments of collinearity of dot pairs across 360° of angular position. A priori and post hoc wavelet modelling strategies were used to identify independent sources of error that could not be attributed to chance, and some new statistical protocols were applied. We found evidence of error sources at several levels of scale, and these results were confirmed by the application of cross‐validation methods that make no assumption about the nature of the error probability distribution.