Nature and origins of virtual environments: a bibliographical essay

Perceptions may be compared with hypotheses in science. The methods of acquiring scientific knowledge provide a working paradigm for investigating processes of perception. Much as the information channels of instruments, such as radio telescopes, transmit signals which are processed according to various assumptions to give useful data, so neural signals are processed to give data for perception. To understand perception, the signal codes and the stored knowledge or assumptions used for deriving perceptual hypotheses must be discovered. Systematic perceptual errors are important clues for appreciating signal channel limitations, and for discovering hypothesis-generating procedures. Although this distinction between 'physiological' and 'cognitive' aspects of perception may be logically clear, it is in practice surprisingly difficult to establish which are responsible even for clearly established phenomena such as the classical distortion illusions. Experimental results are presented, aimed at distinguishing between and disconvering what happens when there is mismatch with the neural signal channel, and when neural signals are processed inappropriately for the current situation. This leads us to make some distinctions between perceptual and scientific hypotheses, which raise in a new form the problem: What are 'objects'?

This article describes techniques used to synthesize headphone-presented stimuli that simulate the ear-canal waveforms produced by free-field sources. The stimulus synthesis techniques involve measurement of each subject's free-field-to-eardrum transfer functions for sources at a large number of locations in free field, and measurement of headphone-to-eardrum transfer functions with the subject wearing headphones. Digital filters are then constructed from the transfer function measurements, and stimuli are passed through these digital filters. Transfer function data from ten subjects and 144 source positions are described in this article, along with estimates of the various sources of error in the measurements. The free-field-to-eardrum transfer function data are consistent with comparable data reported elsewhere in the literature. A comparison of ear-canal waveforms produced by free-field sources with ear-canal waveforms produced by headphone-presented simulations shows that the simulations duplicate free-field waveforms within a few dB of magnitude and a few degrees of phase at frequencies up to 14 kHz.

Listeners reported the apparent spatial positions of wideband noise bursts that were presented either by loudspeakers in free field or by headphones. The headphone stimuli were digitally processed with the aim of duplicating, at a listener's eardrums, the waveforms that were produced by the free-field stimuli. The processing algorithms were based on each subject's free-field-to-eardrum transfer functions that had been measured at 144 free-field source locations. The headphone stimuli were localized by eight subjects in virtually the same positions as the corresponding free-field stimuli. However, with headphone stimuli, there were more front-back confusions, and source elevation seemed slightly less well defined. One subject's difficulty with elevation judgments, which was observed both with free-field and with headphone stimuli, was traced to distorted features of the free-field-to-eardrum transfer function.