Art.-Van Opstal (B)

417 Human auditory localization is remarkably accurate, even in complete darkness under open-loop conditions1–4 (Fig. 2; see also Methods). This performance may rival that of the barn owl, a well studied nocturnal hunter that relies heavily on its auditory system to capture prey. Yet, quite different localization mechanisms are used by each species. In barn owls, sound azimuth (horizontal direction) is derived from interaural timing differences, whereas sound elevation cues are provided by the interaural level differences that result from the up–down asymmetry of its facial ruff5. In humans, however, both interaural cues only relate to the sound-source azimuth. In contrast, sound elevation and front-back direction are determined on the basis of spectral cues generated by the direction-dependent filtering of the pinnae6–9. Sounds entering the ear via the pinna aperture have some frequencies amplified and some attenuated, with an effect that can be described mathematically by linear transfer functions (‘pinna filters’, Fig. 1)10. Sound elevation detection in humans and in many other mammalian species may therefore be considered as a spectral pattern-recognition problem. The importance of the pinnae in sound localization has been demonstrated in both pinna-occlusion experiments11, and in narrow-band sound localization studies1. Moreover, when localization is attempted ‘through another person’s ears’ (using virtual sound-source synthesis techniques), localization errors increase dramatically12. However, accurate localization on the basis of spectral cues poses constraints on the sound spectrum. A sound needs to be broad-band in order to yield sufficient spectral shape information, and the acoustic signal at the eardrum comprises the original source spectrum as modified by the linear pinna filter. Both spectral functions however, are a priori unknown to the auditory localization system, and the extraction of sound elevation and front–back direction is therefore not trivial1,2,13. It has been suggested that the auditory system may resolve this problem by assuming that real-life sounds do not contain the prominent peaks and notches of the different pinna filters. (Fig. 1a)1,2,13. Because sound localization relies on implicit physical cues, the auditory system must somehow transform the binaural differences and monaural spectral pinna cues into consistent spatial information. It is thought that the auditory system acquires these spatial relations through learning, and that the visual system may train and calibrate the acoustic localization process by providing accurate spatial feedback. Indeed, behavioral experiments with young barn owls have shown that the integrity of the visual system guides acoustic localization performance14,15: When reared with prisms, the owl’s auditory localization response shifts in the same direction as the altered visual representation, although the acoustic cues remain unchanged. Likewise, acoustic perceptual shifts that are induced by one-sided ear plugs are resolved by the availability of visual feedback, although some adjustment of the optic tectum’s spatial map also occurs after eyelid suture16. In addition, the formation of the auditory space map in the owl’s inferior colliculus has been shown to rely mostly on visual experience in early life17. Comparable results have been obtained for the spatial representation of sound in the midbrain superior colliculus of newborn mammals18,19. Conceivably, the human auditory localization system may develop by a similar learning process, as the subtle acoustic cues vary substantially during growth. However, no data are available that clearly demonstrate an adaptive capability of the human auditory localization system20. We therefore tested whether human subjects would be able to adapt to a consistent change in the spectral localization cues.