Background. During the last decade an important body of research in Psychology of Music has been devoted to provide empirical evidence of certain assumptions of Music Theory, related to the study of musical structure. Musical structure has been approached from different perspectives. A previous study investigated the role of structure -in terms of the reductional representation, as proposed by some approaches- in perceptual similarities between melodies. A rivalry between components of surface and deeper structural levels was found when listeners judged the similarity of unaccompanied melodies with same or different structure- derived from voice leading principles - and similar surface - understood as melodic contour. Aims. The purpose of this paper is to test the pertinence of some theoretical models -which study both musical structure and surface- as predictors of the found rivalry. Method. Stimuli used in the mentioned study -consisting of fifteen trios of tonal melodies- were analysed according to the principles of some presented by Quinn (1999), Schmuckler (1999), Krumhansl (1990) and Bharucha (1984) and a correlated measure of the contour similarity (Shifres & Martínez, 1999). These analysis were compared with the responses given by the listeners in that study (N=146). Results. None of the cited approaches appeared to be either a linear predictor or a complete explanation of the perceptual similarities. However, both Bharucha’s model and the correlated measure of contour similarity were better than the others to explain the named rivalry, when structure was analysed according to both Krumhansl’s tonal weights and the voice leading principles. Conclusions. These results would suggest that listeners are sensitive to the presence of the deeper structural information arising from the principles of voice leading. Further relationships between Bharucha’s and Krumhansl’s models with principles of the voice leading in tonal music are discussed.