(back to SpatialLocalizationPaper)

Much of our focus will be directed at the interplay between our two ears' sensations, and the mental picture that results. However, we first look at the brain's ability to localize sound from just one ear. This section also considers by extension the effect of sounds in the median plane, which being the locus of points equidistant from both ears, naturally eliminates binaural processing from the auditory events. Our understanding can then be generalized to the binaural case.

Monaural localization depends entirely on the seemingly vestigial flaps of cartilage and skin known as the pinnae. For many years, before their role was thoroughly researched, no theory existed either explaining monaural localization or even acknowledging its existence. However, it is now apparent that the pinnae's primary purpose is in fact localization -- despite its shape, its ability to funnel sounds into the ear canal (i.e. enhance the ear's overall sensitivity) has been proven insignificant (Gelfand, 47). With this information in hand, there are several prominent models to explain the pinnae's true role.

Most basically, the pinnae cast an acoustic shadow. Like any object, they have vastly different mechanical impedance from the vibration medium (air), and a size corresponding to certain wavelengths of sound. Waves that are small compared with an average pinna have frequencies above 3000 Hz, and it is at this point that we begin to observe the reflections and diffractions that characterize an acoustic shadow* (Gulick, 339). Since the pinnae protrude in the frontal plane, the shadow is cast such that high-frequency sounds emanating from the rear have diminished intensity compared with those from the front, with the effect graduated in between. Thus we have the genesis of a spectral discriminant, a low-pass filter that varies with position to give us information about the front/back localization.

The auditory events corresponding to sounds originating in the median plane provide a testing ground for this theory. The distance between the ears is equal, negating any interaural effects, and the variation in position is front-to-back, as we predict the pinnae shadow effect to be.** Indeed, the localization blur with white noise ranges from about 4-10 degrees of arc, the most difficult identification being directly above the head. Even finer distinctions can be made when signals of smaller bandwidth are used (Blauert, 310).

A more in-depth treatment of the pinnae finds a series of temporal alterations to the perceived sound, generated by interactions with its "corrugations." When the wavelength of a sound is small compared with these features (>6 KHz), a delayed reflection from the concha can be 80-15 microseconds and from the helix 100-300 microseconds, when azimuth and elevation are respectively varied. These hypothesized delays are very small, below the masking threshold usually attributed to (monaural) hearing, but an alternative theory is easily formulated: since they reflect sound waves that interfere with direct sources, the time delays correspond to phase and frequency changes easily detectable by the ear (Gulick, 340; Moore, 161). With these considerations in effect, we can thus discern a wide range of positions in the horizontal and frontal planes.

Furthermore, we recognize additional components of the incoming signal that affect our monaural accuracy. The described filtering effects of the head and pinnae each apply to rather narrow frequency bands. With broadband signals, thus, a greater number of informative distortions reaches the ear canal; indeed, experiments with narrow-band noise finds localization blur much increased. For example, signal components above 7KHz must be present to minimize blur with respect to elevation, while providing signals devoid of high-frequency information invites front/rear confusion (Blauert, 102). The remaining factor to consider is familiarity with the source. Correct judgement of position becomes more frequent when (e.g.) a familiar voice is used, particularly reducing front/rear ambiguities; similar performance gains have been reported with white noise when the subjects were allowed a trial period (Blauert, 104).

In sum, the measured localization blur in the median plane is about ? 9 degrees in front when the head is immobilized, varying up to about 22 degrees above the head. The former figure can vary up to 17 degrees when the program is unfamiliar speech, or down to 4 degrees for white noise (Blauert, 44).

*Note that positively locating a source based on the shadow effect alone requires a reference for comparison. In the text cited listeners were given multiple trials; in the real world, the ambiguity might be resolved by head movements. I raise this point to emphasize that allowing comparisons with known stimuli and/or head movements greatly increases the resolution of virtually every monaural and binaural process we discuss here. (Unfortunately, it greatly increases the difficulty of artificially reproducing a 3D sound environment.) The remaining experimental results to be discussed should assume an unfamiliar source and/or room in addition to an immobilized head, unless otherwise noted.

**As proof that no binaural processing is at work, we cite an experiment where sounds in the median plane were recorded in one ear with the other plugged, then played back into both ears after appropriate equalization. No appreciable differences in localization were found (Blauert, 306).

onward to BinauralLocalization