a b s t r a c t Acoustic backscatter mosaics derived from multibeam and sidescan sonars are often used to estimate seafloor type and composition, which are important parameters in the description of benthic habitats. However, due to limitations of the mosaicking technique, backscatter mosaics are restricted in their capacity to unambiguously discriminate seafloor properties. A more adequate technique to characterize the seafloor is the analysis of backscatter angular response, since this response is an intrinsic property of the seafloor. Angular response analysis sometimes lacks spatial resolution, however, as this approach is limited to the swath width of the sonar. In this paper, we propose an approach to combine mosaicking and angular response analysis techniques in an attempt to take advantage of both the spatial resolution of the mosaic, and the angular resolution derived from the angular response analysis. The proposed method for analyzing the backscatter mosaic together with the backscatter angular response is applied to the acoustic backscatter acquired by a Simrad EM1002 multibeam sonar (95–98 kHz) on Stanton Banks, to the west of Scotland. First, a normalized acoustic backscatter mosaic is prepared for the survey area. Then, visual interpretation of the mosaic produces areas on the seafloor with similar textural pat-terns that we call acoustic themes. Finally, the average backscatter angular response of all the backscatter samples that fall within the same acoustic theme, regardless of the acquisition line they belong to, is com-pared to a formal mathematical model that links acoustic backscatter observations to seafloor properties. The inversion of the model is accomplished with a constrained iterative method known as angular range analysis (ARA), which produces estimates of the same acoustic impedance, roughness and the mean grain size of the insonified area of the seafloor. The results of the ARA inversion are compared to bottom pho-tographs acquired in the area delimited by the acoustic theme, showing a very good correlation. The abil-ity to discriminate benthic habitats may therefore be improved using this approach.