1
Neural and behavioral sensitivities to azimuth degrade
with distance in reverberant environments
Sasha Devore1, Antje Ihlefeld2, Barbara G. Shinn-Cunningham2 and Bertrand Del-
gutte1
1 Eaton-Peabody Laboratory, Massachusetts Eye and Ear Infirmary, Boston, MA
sashad@mit.edu, Bertrand_Delgutte@meei.harvard.edu
2
Hearing Research Center, Boston University, Boston, MA {ihlefeld, shinn}@bu.edu
1 Introduction
Reverberation poses a challenge to sound localization in rooms. In an anechoic
space, the only energy reaching a listener’s ears arrives directly from the sound
source. In reverberant environments, however, acoustic reflections interfere with
the direct sound and distort the ongoing directional cues, leading to fluctuations in
interaural time and level differences (ITD and ILD) over the course of the stimulus
(Shinn-Cunningham et al. 2005). These effects become more severe as the distance
from sound source to listener increases, which causes the ratio of direct to rever-
berant energy (D/R) to decrease (Hartmann et al., 2005; Shinn-Cunningham et al.,
2005).
Few neurophysiological and psychophysical studies have systematically exam-
ined sensitivity to sound source azimuth as a function of D/R (Rakerd and Hart-
mann, 2005). Here, we report the results of two closely-integrated studies aimed at
characterizing the influence of acoustic reflections like those present in typical
classrooms on both the directional sensitivity of auditory neurons and the localiza-
tion performance of human listeners. We used low-frequency stimuli to emphasize
ITDs, which are the most important binaural cue for sounds containing low-
frequency energy (MacPherson and Middlebrooks, 2002; Wightman and Kistler,
1992). We find that reverberation reduces the directional sensitivity of low-
frequency, ITD-sensitive neurons in the cat inferior colliculus (IC), and that this
degradation becomes more severe with decreasing D/R (increasing distance). We
show parallel degradations in human sensitivity to the azimuth of low-frequency
noise.

2
2 Single-unit neurophysiology
2.1 Methods
Methods for recording from low-frequency, ITD-sensitive neurons in the IC of
anesthetized cats were as described by Hancock and Delgutte (2004). We focused
on measuring neural responses as a function of source azimuth in simulated rooms.
Binaural room impulse responses (BRIRs) were simulated using the room-
image method (Allen and Berkeley, 1979) for a pair of receivers corresponding to
the left and right ears of a cat in the center of a simulated reverberant room (11 x
13 x 3 m). We did not include a model of the head in the simulations, so that the
resulting BRIRs contained ITD but essentially no ILD cues. BRIRs were calculated
for azimuths spanning the frontal hemifield (-90º to 90º) at distances of 1m and 3m
with respect to the midpoint of the receivers. Anechoic impulse responses were
created by time-windowing the direct wavefront from the 1m reverberant BRIRs.
The direct-to-reverberant energy ratio (D/R) was calculated as the ratio of the en-
ergy in the direct sound (time-windowed from the BRIR) to the energy of the re-
maining impulse response. An overall D/R was determined for each source dis-
tance by averaging across all azimuths. Virtual room stimuli were created by con-
volving the BRIRs with a reproducible 400-ms broadband noise burst. The first
400 ms of the resulting signals were presented to dial-in-urethane anesthetized cats
over calibrated, closed acoustic systems.
Neural responses were measured as a function of source azimuth for each vir-
tual room condition (anechoic, 1m, and 3m). We typically used 11 azimuths (15º
steps) or, occasionally, 7 azimuths (30º steps). The noise stimulus was repeated 16
times at each azimuth, with random order across azimuths. We computed the aver-
age firing rate by counting the number of action potentials over the stimulus dura-
tion.
2.2 Results
We measured neural responses as a function of azimuth for 25 IC units from 7 cats.
Rate-azimuth curves for two typical units are shown in Fig. 1A. Neural rate re-
sponses depend strongly on source azimuth in the anechoic condition (D/R = ∞
dB), with a preference for contralateral azimuths. Reverberation reduces the range
of firing rates over all source azimuths (a “demodulation”), although rates still vary
systematically with azimuth. To quantify these observations, we define the relative
response range as the range of firing rates for a given room condition normalized
by the range of firing rates in the anechoic condition. The relative range for the
anechoic condition is 1, by definition. For most neurons, the relative ranges in the
reverberant conditions are less than 1, indicating a compression of the rate-
azimuths curves; furthermore, relative range decreases with decreasing D/R (Fig.
1B).
Neural sensitivity to azimuth depends not only on the response range, but also
on the variability in responses at each azimuth. To quantify sensitivity, we com-
puted the mutual information (MI) between the source azimuth and the neural