The present work is based on the property of some models of analysis to provide inter subjects representations to explain Music. It is part of a number of experiments (Martínez, Shifres; Shifres, Martínez; Martínez, Shifres; Shifres, Martínez) that aims to study the reality of certain cognitive assumptions belonging to the schenkerian theory. One of the most important contributions of this theory is the idea of structure and prolongation as attributes of tonal music and the relationship between them (Salzer, [1962]-1990; Forte & Gilbert, [1982]-1992). Some aspects of the theory have been tested experimentally (Serafine, Glassman & Overbeeke, 1989) providing evidence for the cognitive reality of musical structure. The nature of the relationship between surface and structure presents a problem to accomplish an experimental study. The events belonging to the musical structure are represented at the level of the musical surface. Therefore, although it is possible to modify the surface level without changing the structure, it is impossible to modify the structure without modifying at the same time the musical surface. This difficult has been treated by Serafine (1988) who studied the cognitive reality of structure using compound melodies and by Dibben (1994) who studied the abstraction of structure in terms of the time-span reduction (Lerdhal & Jackendoff). However, this phenomena has not been experimentally researched enough in terms of the dialectic relationship between both components: musical surface and musical structure. It is possible that the surface attributes contend with the structural attributes in the priority to process the tonal information. The main body of experimental studies concerning the musical surface have devoted to the analysis of melodic contour (Dowling, Trehub, Edworthy ). These studies would provide methodological tools to cope the problem of the “competition” mentioned above. In order to study the understanding of structure by listening, a number of tests were run. It was studied the implicit use of structure in similarity judgement tasks between pairs of melodic stimuli. The musical materials consisted on a melody model (m) and two comparison melodies, one of them sharing the structure (cm1) and the other do not (cm2). The model melodies were selected from the academic repertoire of Western Tonal Music and the two comparison melodies were composed trying to minimise surface differences. Subjects had to listen to the melodies in pairs (m-mc1 or m-mc2) and to judge their similarity. We hypothesised that they would rate the melody which shared the structure as the most similar to the model. The main problem in this design consisted on the isolation of both components: structure and surface. Therefore, the composition of the comparison melodies was done following a number of constraints and controls post hoc were done. One of these post hoc controls consisted on the comparison of the similarity of melodic contours in pairs (according to the measures for melodic contour by Edworthy (1985) and Dowling (1988)) running a Pearsons’s correlation coefficient for each of them. From the results obtained it is possible hypothesise that: i) when the highest correlation is found between the model and the cm2 (different structure) subjects tend to do a surface based judgement and ii) when the highest correlation is found between the cm 1 and the cm2, the differences between then would not be enough to do the judgement; in this case subjects would make a judgement based on structure. Since the experimental condition of this design did not provide the opportunity to empirically contrast the similarity of the comparison melodies, because each pair always had the model melody as one the members, then it is necessary to create a new experimental condition in order to further investigate the reliability of the correlation coefficient measure. This paper presents the results of an experiment in which subjects had to judge -while listening- the similarity between trios of melodies. One might expect that if subjects actually make use of structure while listening, their similarity ratings will reflect a competence between surface and structure components, and not only the rates of melodic contour similarity that are provided by the correlation coefficients. Results are being processed at this moment and they will be presented at the conference. Implications to the study of musical structure will be discussed.