Auditory detection of flowing water by humans in a natural habitat: Psychophysical and modelling data

Knowledge is lacking about the way humans process acoustic information conveyed by flowing water in natural settings. To address this issue, the sound of a river was recorded in a Mediterranean forest three times per day and for two seasons along a 100 m transect from a river segment and at two further locations (150 and 200 m) on a different transect. Stimuli were presented diotically to listeners tasked to detect presence of water. A two-interval, two-alternative forced-choice paradigm was used to measure detection performance, where the target and standard stimuli corresponded to samples from the 100 m transect and samples from the two further locations on the other transect, respectively. Each listener was tested on a single trial only. Absolute level cues were removed. Detection performance measured for this habitat varied mainly with distance from the river segment and season. The empirical data were compared to those simulated by a model calculating time-averaged texture statistics at the output of a cochlear and a modulation filterbank and predicting performance via a template-matching decision strategy. The results indicate that the sparsity of envelope fluctuations and the coordination of temporal-envelope fluctuations across perceptual channels are diagnostic cues for water detection in natural environments.