In environmental epidemiology, studies rely on the quantification of subject'sexposures in a surface defined as the subject exposure area. For outdoor exposure, this area is often considered as the subject's neighbourhood. But, depending on the authors, the size and the nature of this neighbourhood differs, making difficult to compare results. In order to study the impact of the sampling surface on the noise exposure values affected to a subject, a high definition environmental noise model has been builton a middle-sized French city. Outdoor neighbourhood noise indices were computed at 10,394 residential buildings, using eight different sizes of buffers defined by different radius values: 50m, 100m, 150m, 200m, 250m, 300m, 350m and 400m. For each buffer, noise indices correspond to the average noise level in the buffer's area, in LAeq24h. Results were first compared using Friedman's test. Post-hoc pairwise comparison were then conducted using Wilcoxon test and Siegel & Castelanne adjusted P-values. Indice differences between the smallest and the biggest size of buffer were compared for each building by subtracting 50m buffer noise indice to 400m buffer noise indice (?400-50=LAeq24h-400m - Laeq24h-50m). Influence of urban environmental factors (distance to main noise source, urban morphology and deprivation index) was analysed using a multilevel modelling. The means of the eight indices range between 49.0 dB to 49.4 dB, despite samples appears to be significantly different (P<0.05). The ?400-50 values ranges between -16.1 dB to +16.9 dB, and are significantly associated with urban environmental factors (All P<0.01). The size of the sampled surface does not seem to influence the estimated average noise exposure. However, for each building, the size of the neighbourhood seriously influence the estimated values. The environmental composition also appears to be of important contribution and should be carefully considered in the choice of the neighbourhood size.