Operators of modern military aircraft must acquire and process large amounts of information to perform their tasks effectively. Traditionally, aircraft designers have relied heavily on visual displays as a means of presenting information. Audio displays also have considerable potential for conveying information within this environment. Auditory signals can capture attention and provide information without the need for continual visual monitoring. Well-designed audio displays could complement or replace visual displays and reduce the visual workload currently experienced by aircrew.

Substantial progress has recently been made toward the development of three-dimensional (3D) audio displays that generate sound that is heard in 3D space. These displays work by passing auditory signals through location-specific digital filters that modify these signals as the head and ears would. It is envisaged that 3D audio displays will inform future aircrew of such things as the nature and location of threats to their aircraft. Several recent studies have shown that these displays could assist aircrew in their search for small visual targets such as other aircraft.

How reliably 3D audio displays will function within the aviation environment will depend in part on the extent to which auditory perception is affected by hypoxia. Previous studies have shown that hypoxia impairs various aspects of visual perception and general cognitive functioning. Of particular relevance with respect to 3D audio displays, however, is hypoxia's potential effect on sound localisation.

Hypoxia's effect on sound localisation has been examined in only two studies. Rosenberg and Pollard measured the localisation accuracy of subjects involved in separate mountaineering expeditions to the Andes, Himalayas and Alps. Sounds to be localised consisted of short bursts of broadband noise presented over headphones. These sounds had been recorded at the eardrums of an acoustic manikin while broadband noise was presented from a speaker at different azimuthal locations in the manikin’s interaural horizontal plane. They therefore contained the sound location cues imposed on the noise by the manikin’s head and pinnae. Rosenberg and Pollard observed a significant deterioration in sound localisation accuracy for subjects involved in each expedition at altitudes ranging from 1455 to 5360 m. In a subsequent laboratory study, Rosenberg et al. measured the localisation accuracy of subjects at sea level and simulated altitudes of 1524, 2438, 3657 and 5486 m in a hypobaric chamber. As before, sounds to be localised consisted of short bursts of broadband noise pre-recorded through the ears of an acoustic manikin. The accuracy with which subjects localised these sounds was observed by Rosenberg et al. to be significantly lower at the two higher altitudes. At both of these altitudes many subjects displayed localisation errors of a directionally systematic nature but the direction toward which they mislocalised was inconsistent across subjects.

It is possible that the altitude at which hypoxia begins to affect sound localisation is lower than the studies by Rosenberg and Pollard and Rosenberg et al. suggest. Two types of localisation error are commonly distinguished. Errors of the first, and most common, type are relatively small and occur where the location of a sound source is more-or-less correctly perceived. Errors of the second type are known as front-back confusions and occur where the angle of a sound source with respect to the median plane is correctly perceived but the source is judged to be in the incorrect hemisphere (front or back). Rosenberg et al. (19) reported an average localisation error of about 24^O at sea level and a front-back confusion rate that averaged 42% and did not differ with altitude. Rosenberg and Pollard reported an average localisation error of 22-26^o at sea level and did not report front-back confusion rates. These values are extremely high in relation to those normally measured in studies in which sounds are presented from a speaker in a free field. The poor performance at sea level in Rosenberg and Pollard's and Rosenberg et al.'s studies may have produced a floor effect that made it difficult to demonstrate performance reduction at altitudes above sea level.

As described above, the sounds to be localised in the studies by Rosenberg and Pollard and Rosenberg et al. were recordings that had been physically filtered by the head and pinnae of an acoustic manikin. The cues used by humans to localise sound are known to be highly individualistic and it is likely that subjects in these two studies localised poorly because the cues provided by the manikin were not well matched to those normally available to them. It is also possible that the localisation performance of these subjects was hampered by a paucity of effective high-frequency content in the sounds to be localised. While the hearing of each subject in the study by Rosenberg et al., for example, was reported to be "acceptable" or "good" in accordance with RAF standards, hearing was tested for frequencies up to 6 kHz only. As the ages of subjects in that study ranged from 19 up to 43 years, it is likely that some had reduced sensitivity for frequencies above that value. Frequencies above 6 kHz are known to provide important cues to sound location.

Sound-source locations in Rosenberg and Pollard's (18) and Rosenberg et al.'s (19) studies were restricted to the 0^° elevation plane. The elevation of a sound source is determined using cues that differ from those used to determine its azimuth (see 13 for review). It is possible, therefore, that hypoxia has different effects on elevation and azimuth judgements. As it is envisaged that 3D audio displays will provide both elevation and azimuth information, the effect of hypoxia on the localisation of sound presented from a broader range of elevations needs to be examined.

In this study, hypoxia’s effect on sound localisation was examined for sound presented from a free-field source using subjects whose ability to accurately localise this stimulus at sea level had been established. This ensured that any reduction in performance resulting from hypoxia would be readily apparent. The range of locations for which hypoxia's effect on sound localisation has been examined was extended to include locations outside the 0^° elevation plane. 

Four healthy, right-handed volunteers, one female and three male, ranging in age from 25 to 40 years participated in this study. Informed consent was obtained from each. Hearing was assessed by measuring absolute thresholds for 1, 2, 4, 8, 10, 12, 14 and 16 kHz pure tones using procedures described in detail by Watson et al. For each subject all thresholds were lower than the relevant age-specific norm.

All subjects were allowed to practice sound localisation during several training sessions prior to data collection. During this period each subject's performance stabilised at a level that demonstrated his/her proficiency at the sound localisation task.

Localisation performance was measured at each of four simulated altitudes (0, 1200, 2400 and 3700 m above sea level). Each subject participated in four experimental sessions during each of which a different altitude was tested. The order in which altitudes were tested was counterbalanced across the four subjects using a Latin square.

Hypoxia was induced at sea level and ambient atmospheric pressure by having subjects breathe an appropriate gas mixture through an aviator's oxygen mask. Mixtures having oxygen concentrations of 21.0, 18.1, 15.6 and 13.3 % were used to simulate altitudes of 0, 1200, 2400 and 3700 m, respectively. All were blended from bottled air, oxygen and nitrogen no more than 30 minutes before the session in which they were used. These mixtures were stored in 100 litre Douglas bags until needed. Testing in each session did not begin until the subject had breathed the relevant gas mixture for at least 15 minutes and his/her blood oxygen saturation level, which was monitored throughout via non-invasive pulse oximetry, had stabilised at an altitude-appropriate level. The saturation levels at which individual subjects stabilised for each altitude are shown in Table 1. An interval of at least one hour separated all pairs of consecutive sessions. No subject experienced hypoxia for more than two sessions on any day.

Localisation Procedure

The subject was seated in a sound-attenuated anechoic chamber at the centre of a 1-m radius hoop on which a loudspeaker  was mounted. Background noise levels within this chamber were less than 10 dB SPL in all 1/3-octave bands with centre frequencies from 0.5 to 16.0 kHz. Loudspeaker movement was driven by programmable stepping motors that could position the loudspeaker anywhere from 0 to 359.9^o azimuth and —50 to +80^o elevation with .1^o resolution. A convention of measuring azimuth in the clockwise direction and describing elevation below the interaural horizontal plane as negative was followed. The subject's view of the hoop and loudspeaker was obscured by a cloth sphere supported by thin fibreglass rods. The cloth from which this sphere was constructed was acoustically transparent. The inside of this sphere was dimly lit to allow visual orientation. Subjects wore a headband on which a laser pointer and a magnetic tracker receiver were mounted.

At the beginning of each trial the subject fixated on an LED at 0^o azimuth and elevation. When ready, he/she pressed a hand-held button. An acoustic stimulus was then presented from the loudspeaker, provided the subject's head was stationary (its azimuth, elevation and roll did not vary by more than 0.2^o over three successive readings of the head tracker made at 20 ms intervals). This stimulus consisted of a fresh, 328-ms burst of broadband (.05-20 kHz) noise incorporating 20-ms cosine-shaped rise and fall times. Its level was set at 60 dB (A-weighted) using a microphone and sound-level meter positioned at the centre of the hoop. The subject was instructed to keep his/her head stationary during stimulus presentation. Following stimulus presentation, the subject directed the laser pointer beam at the precise point on the surface of the cloth sphere from which he/she perceived the stimulus to come. The location and orientation of the laser pointer was measured by the magnetic tracker so that the point where the beam intersected the sphere could be calculated geometrically. An LED attached to the centre of the loudspeaker was illuminated and the subject then directed the laser pointer beam at the LED. The angle between the two vectors extending from the centre of hoop to the perceived location of the stimulus source and the loudspeaker LED was defined as the localisation error.

The stimulus location for each trial was chosen following a pseudorandom procedure that ensured a more-or-less even spread within each session. The part-sphere from —47.6 to +80^o elevation was divided into 42 sectors of similar shape and area. One of these sectors was chosen without replacement on each trial and a location within it was chosen randomly. The loudspeaker was moved to the target location in two steps to reduce the likelihood of subjects discerning the location from the duration of movement. During the first step the loudspeaker was moved to a randomly chosen location at least 30^o away from the previous and subsequent target locations in both azimuth and elevation. During the second step it was moved to the target location.

For each subject the average localisation error was calculated for each simulated altitude. The average localisation error at each altitude averaged across all four subjects is shown in Figure 1. Errors for trials on which front-back confusions were made are included in these averages and have not been modified in any way. The average localisation error at sea level (0 m) was a little less than 14^o, reflecting the good localisation performance of these subjects. Average localisation errors at all higher altitudes were a little smaller and ranged from 12.6 to 12.9^o. The error bars shown in this figure represent one standard error of the average localisation error. Their small magnitude indicates that average localisation errors varied little across subjects.

Figure 1.

Average localisation error at each altitude averaged across all four subjects.

An analysis of variance incorporating a Greenhouse-Geisser correction indicated that the effect of altitude on average localisation error was not significant (F(1.36,4.08)=.54, p=.505). The power of this analysis was calculated following procedures outlined by Keppel (9) and found to be sufficient („.8) to detect an effect of 5 to 6^o. While it is possible that a smaller effect was present but failed to reach significance due to lack of power, the magnitude of the altitude effect, as measured by an estimate of omega-squared (9), was found to be negligible (ω²=-.095). 

The results of this study indicate that hypoxia induced by exposure to simulated altitudes as high as 3700 m has no discernible effect on sound localisation. This finding is in contrast with those of Rosenberg and Pollard and Rosenberg et al. who observed a significant deterioration of sound localisation ability in subjects exposed to altitudes as low as 1455 m.

It is unlikely that the null result obtained in this study resulted from our use of a relatively small number of subjects. Analysis of our data revealed the presence of sufficient power to detect a 5 to 6^o effect, which is of similar size to, or smaller than, the effects observed by Rosenberg and Pollard and Rosenberg et al.. In addition, the only trend evident in our data, which was in the direction of improved localisation performance at higher altitude, was shown to be of negligible magnitude.

It was suggested above that Rosenberg and Pollard and Rosenberg et al. might have overestimated the altitude at which hypoxia begins to have a deleterious effect on sound localisation. This suggestion was made because of the poor sea-level performance of their subjects and the possible existence of a floor effect that made it difficult to demonstrate performance reduction at higher altitudes. It was argued that our use of a free-field localisation task would avoid this possibility and provide a particularly sensitive measure of localisation accuracy. As reported above, the average localisation error at sea level for our subjects was 14^o. This, as expected, is considerably smaller than the localisation errors of subjects in the studies by Rosenberg and Pollard and Rosenberg et al. and is more in keeping with those observed in previous free-field localisation studies. The good sea-level performance of our subjects provided a baseline against which a reduction in performance at higher altitudes would have been readily apparent. Despite this, however, we did not observe any performance reduction at altitudes up to 3700 m.

It is not clear why Rosenberg and Pollard and Rosenberg et al.observed a deterioration in sound localisation performance at altitudes below 3700 m and we did not. It has already been noted that Rosenberg and Pollard and Rosenberg et al. presented stimuli via a non-individualised 3D audio display while stimuli in our study were presented via a free-field source. Another obvious difference between these studies concerns the ambient pressure conditions under which hypoxia was induced. Rosenberg and Pollard and Rosenberg et al. induced hypoxia under hypobaric conditions by exposing their subjects to high altitude environments or low pressure in a hypobaric chamber. We induced hypoxia under normobaric conditions by reducing the oxygen concentration in the gas mixture inspired by our subjects at sea level. Changes in ambient pressure create pressure differences across the eardrum that can reduce auditory sensitivity and distort sound localisation cues unless pressure is equalised by opening the Eustachian tube. It is possible that subjects in Rosenberg et al.'s study localised inaccurately at altitudes above sea level because the pressure across their eardrums was not completely equalised. This is unlikely with respect to subjects in Rosenberg and Pollard's study, however, as they had ample time to achieve equalisation during the gradual physical ascent to each higher altitude. The effect of ambient pressure change in the absence of hypoxia could be examined by exposing subjects to low pressure in a hypobaric chamber while allowing them to breath a gas mixture with an appropriately enhanced oxygen concentration.

The results of this study are encouraging with respect to the application of 3D audio displays within the aviation environment. Many operators within that environment are routinely exposed to low levels of hypoxia and all are at risk of exposure to much higher levels. In view of the fact that severe levels of hypoxia have been found to induce significant reductions in cochlear sensitivity in non-human subjects, it seems reasonable to assume that localisation will be compromised under conditions of extreme hypoxia. As hypoxia levels normally experienced by operators of modern aircraft are below the maximum in this study, however, our results indicate that localisation performance in that environment will not normally be affected by hypoxia.