Using Verbal Data to Measure the Ecological Validity of Sound Reproduction Systems C. Guastavino*, B. Katz**, J-D. Polack***, D. Levitin* & D. Dubois*** 25th November 2003 *McGill University 1 , Department of Psychology, 1205 Dr. Penfield Avenue, Montreal, QC, H3A 1B1, Canada, Catherine.Guastavino@mail.mcgill.ca ** LIMSI-CNRS, Perception Située, BP 133, F91403, Orsay, France. *** Laboratoire d’Acoustique Musicale, Université Paris 6, CNRS (UMR 7604), 11 rue de Lourmel, F75015 Paris, France. Abstract A methodology based on the linguistic exploration of spontaneous descriptions was developed to investigate the sound quality of urban environments. A mail sur- vey was first conducted to better understand how low frequencies affect people in their everyday life. Three other experiments were carried out to explore the eco- logical validity of experimental settings in laboratory conditions. The reference study consisted of interviews conducted in actual environments, which were also recorded simultaneously. The recordings were used for two listening tests, the first one using stereophonic reproduction and the second one using multichannel re- production. The comparison of the verbal data collected in the different contexts sketches some theoretical and methodological issues concerning the reproduction of everyday life scenes in laboratory conditions. The linguistic analyses indicate that the “same” acoustic phenomenon gives rise to different cognitive represen- tations, depending on the spatial presentation of the stimuli. It follows that the quality of the reproduction system must be adapted to specific properties of mental representations (here, spatial immersion vs. source identification). The analysis 1 Work performed while at the Laboratoire d’Acoustique Musicale, Université Paris 6, CNRS (UMR 7604), 11 rue de Lourmel, F75015 Paris, France. 1