Résumé: The avoidance reaction by fish in front of an approaching vessel is a major source of bias in direct biomass assessment and ecological studies based on fisheries acoustics data. An experiment was carried out to compare echosounder data obtained using a small speedboat and a research fisheries vessel generating significant higher noise above conventional reduced-noise standard. The results show that there was no significant difference between the individual fish target strength distributions, and the numbers of schools recorded by both boats, these schools having similar areas and perimeters. However, the schools detected by the noisier vessel were significantly deeper, and unexpectedly had a significantly higher energy level. These findings suggest that noise-reduced vessels trigger a different vessel avoidance reaction. The noise-reduction standard is not sufficient to reduce avoidance behavior. It is also to take into consideration the ambient noise, which could impair perception of the platform by the fish, and the probability that the acoustic stimuli could be less important than visual perception under some local conditions. The paper introduces the concept of partial avoidance and presents a conceptual diagram of the strength of the avoidance reaction. Last, it is not recommended, because of noise reasons, that vessels routinely used for pelagic stock assessment surveys be changed. Indeed standardized time series, which could be disrupted when switching to a new vessel, are more important than the hypothetical gain from change to quieter vessels. Obviously, all long-term surveys must change vessels; best practice will be to estimate the vessel effect before any change to avoid disrupting the time series and/or perform vessel intercalibration surveys.