Return to search

Probing the human auditory system with reverse correlation

Once the sound reaches the ear, hearing can no longer be exclusively described as a me- chanical phenomenon of wave propagation. As we follow the auditory pathways deeper into the brain, neuronal action potentials shape our perception of sound. But how exactly do the spectrotemporal characteristics of the sound wave a ect perception? We investigated human auditory perception and decisional behaviour with reverse correlation. This technique yields richer datasets than classical methods based on performance metrics alone, providing classi - cation images (CIs) that display observers' task-dependent strategies while potentially serving as templates for computational modeling. We found that observers use the same strategies to detect peaks and dips in sound pressure on a narrow time scale but rst and second-order CIs reveal di erent temporal dynamics within each strategy. When observers detected a speci c frequency on a similar timescale, we were able to expose signatures of neuronal-like spectrotemporal tuning. Detailed modeling of the results showed that observers were not able to rely on the explicit output of these channels. In auditory motion experiments, CIs presented distinct spectrotemporal dynamics between sounds moving from one side of the observer to the other and sound moving towards or away from the observer. In stark contrast, an artificial detector program returned identical CIs. When stimuli were embedded in fragments of human speech and natural sounds, observers used a knowledge- based strategy; as long as fragments were perceived as meaningful, CIs displayed robust tuning e ects which diminished when speech was presented in a temporally reversed order. Overall, we can conclude that reverse correlation is a powerful tool for probing the human auditory system. It reflects task-dependent strategies imposed by the underlying neuronal circuitry rather than statistics or task speci cation.

Identiferoai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:573428
Date January 2013
CreatorsJoosten, Eva Rosalia Margaretha
PublisherUniversity of Aberdeen
Source SetsEthos UK
Detected LanguageEnglish
TypeElectronic Thesis or Dissertation
Sourcehttp://digitool.abdn.ac.uk:80/webclient/DeliveryManager?pid=195809

Page generated in 0.0021 seconds