Auditory Spatial Discrimination and Sound Localization in Single-Sided Deaf Participants Provided with a Cochlear Implant Open Access

Spatial hearing is a fundamental ability of the auditory system. Representations of auditory space need to be computed from binaural and monaural cues in the auditory brainstem and cortex because space is not directly represented by the peripheral auditory system. As early as 1907, Rayleigh developed the duplex theory of spatial hearing. Studies on human auditory localization demonstrated that interaural time differences (lTDs) and interaural level differences (ILDs) specify the location of a sound source in the horizontal plane [Rayleigh, 1907; Stevens and Newman, 1936; Mills, 1958; Nordlund, 1962a, 1962b; Yost and Dye, 1991; Blauert, 1997; Wightman and Kistler, 1997; Recanzone et al., 1998; Carlile et al., 1999]. ITDs are the dominant cue at low frequencies <1.5 kHz, while ILDs are the dominant cue at higher frequencies >2.5 kHz [Blauert, 1997].

Auditory spatial discrimination is mostly studied by using the minimum audible angle (MAA) paradigm introduced by Mills [1958]. The MAA is defined as the smallest angular distance between two adjacent sound sources that can be discriminated and thus represents a threshold for auditory spatial acuity, an indicator of auditory spatial resolution. In contrast to localization, MAA is not only the task to detect a specific direction using ITD and ILD cues but essential to compare between small differences in the available cues. In normal-hearing (NH) adults, Mills [1958] reported best discrimination at central positions with thresholds of about 1°–2°. At lateral positions, spatial discrimination declines, and MAA thresholds increase to about 10° [Blauert, 1997]. Kühnle et al. [2013] and Freigang et al. [2014] investigated MAA in different age groups and found that MAA thresholds are similarly elevated in young children (6–7 years) and older listeners (65–83 years) compared to the other age groups [Freigang et al., 2015].

In persons with asymmetrical hearing, there is a discrepancy between auditory information from the left and right ears, and thus spatial hearing may be less accurate. The extreme cases are patients with single-sided deafness (SSD), who completely lack binaural input. A cochlear implant (CI) can partially compensate for asymmetrical hearing and thus restore localization to some degree [Van Hoesel and Tyler, 2003; Ching et al., 2004; Seeber et al., 2004; Dunn et al., 2008; Grossmann et al., 2016; Dillon et al., 2017a, 2017b]. Evidence suggests that bilateral CI users rely mostly on ILD [Van Hoesel and Tyler, 2003; Seeber and Fastl, 2008; Dorman et al., 2015], and their ITD sensitivity is generally supposed to be poor [Van Hoesel et al., 2009; Aronoff et al., 2010]. However, a role of the envelope ITD processing in spatial hearing with a CI is possible [Noel and Eddington, 2013; Todd et al., 2019] since envelope ITD cues are preserved during CI preprocessing.

Only a few studies showed that a hearing device can also restore auditory spatial discrimination. Beijen et al. [2010] tested children with a CI on one side and a hearing aid on the other side with a left-right discrimination task and showed that the children performed significantly better than chance. Zaleski-King et al. [2019] tested MAA via headphones in adult participants with a CI on one side and a hearing aid on the other side and found similar results. Bilaterally provided CI participants benefit from the second CI: Senn et al. [2005] showed that the participants performed better in the MAA task with two CI than with one CI. To the authors’ knowledge, to date, there are no studies examining MAA thresholds in CI-SSD participants. Auditory spatial discrimination has been commonly measured as discrimination between speech signals and concurrent noise presented at different or the same position. However, this task cannot be directly compared to MAA measurements because these tests measure different abilities (even though both rely on a combination of ITD and ILD).

The present study measures individual spatial discrimination ability using an MAA task to identify good and poor performers relative to a large cohort of age-matched controls. Signals were presented from three locations on the CI and NH sides. Low-pass and high-pass filtered noises were used to investigate potential differences between the value of ITD and ILD cues. Performance in the MAA test was compared to the performance of the same CI-SSD participants in the test for absolute sound localization from Ludwig et al. [2021]. We hypothesized that a CI can restore the spatial discrimination ability in SSD participants to some degree and that ILD and ITD contribute to this performance gain.

The study included 18 adults aged 24–81 years (mean age: 55.8, standard deviation (SD): 18.2 years, 13 females, five males). All CI-SSD participants were patients of the Clinic of Otorhinolaryngology at the Emergency Hospital Berlin and/or of the Hearing Therapy Center Potsdam. All participants were provided with a CI on one side and had normal hearing on the other side. The participants’ audiometric thresholds on the NH side were 20 dB hearing level or better, at octave frequencies from 250 through 8,000 Hz [ANSI, 1996]. Ten participants had a CI on the left and eight on the right side. The mean duration of SSD was 2.6 years (range: 0.3–8.7); the mean duration of CI usage was 11.3 months (range: 1.1–31.5) (values for all individuals are listed in Table 1).

Control participants comprised 120 healthy NH adults between 20 and 83 years. They were divided into the following age cohorts: 20–29 years (n = 22): 12 women, 24.1 years ±2.3; 30–39 years (n = 23): 13 women, 33.2 years ±2.5; 40–59 years (n = 20): 13 women, 51.1 ± 4.9; 65–83 years (n = 55): 30 women, 68.1 years ±5.5. The aim of the analysis was to compare single CI-SSD participant’s results to the results of an age-related control group because it is statistically better than comparing to a single control participant of exactly the same age since the brain age does not necessarily reflect the hearing age [Briley and Summerfield, 2014]. To account for age-related increases in hearing thresholds among some older adults, stimulus levels were adjusted during testing (see: “Individual determination of stimulus intensity” and Freigang et al. [2014]).

All participants gave written informed consent for participation. This study was conducted according to the World Medical Association Declaration of Helsinki and approved by the Local Ethics Committee of the University of Leipzig.

Participants were tested using their everyday CI setting. Fifteen participants had a Nucleus implant using an advanced combinational encoder strategy and a CP910 Nucleus 5^® speech processor from Cochlear™ (Hannover, Germany). Three participants had a Sonata implant using FS4 (fine-structure) coding strategy and an Opus 2 speech processor from Medel Ltd. (Innsbruck, Austria). The microphones of both processors were set to “omnidirectional”. Both processors transfer the envelope of the auditory signal. Only the Medel processor additionally transfers some of the temporal (low-frequency [LF]) fine-structure information of the stimuli from four electrodes. Auto-sensitivity (ASC, Cochlear™) or automatic gain control (AGC, Medel) were activated for all participants but did not influence our experimental stimuli because they activate at higher intensities. Both processors’ time constants (the time taken for the ASC or AGC to adapt) are on the order of seconds [Patrick et al., 2006; Medel, 2020] and far exceed the duration of the stimuli used in this study.

All psychoacoustic testing was performed in a darkened, anechoic, sound-attenuated room (40 m²; Industrial Acoustics, Niederkrüchten, Germany) free from distracting elements. Forty-seven custom-designed loudspeakers (Visaton FRS8) were arranged in a semicircular section (radius 2.35 m, with the subject in the center position) spanning the front of the subject from −98° to +98°; the angular separation between the loudspeakers was 4.3°. Acoustic stimuli with a defined azimuth were generated by the activation of a single loudspeaker. For positions between loudspeakers, the same signal was played from two adjacent loudspeakers at half intensity, resulting in the perception of a sound source between the two speakers, and thus resulting in an angular resolution of 2.15°. Each loudspeaker’s transfer function was equalized. For this, the transmission spectrum was measured using a Bruel and Kjaer (Naerum, Denmark) measuring amplifier (BandK 2,610) and microphone (BandK 2,669, preamplifier BandK 4,190) and a real-time signal processor (RP2.1; Tucker-Davis Technologies, TDT, Alachua, FL, USA). An inverse filter was computed and later used for generating acoustic stimuli with flat spectra across the stimulus frequency range (300–8,000 Hz). This calibration minimized spectral differences between loudspeakers and thus had no perceivable effect on signal level or frequency.

The entire loudspeaker array was covered by black, acoustic transparent gauze to prevent the participants from seeing the number, the location, and the spatial distribution of potential sound sources. The participants were seated in the center of the loudspeaker array in a comfortable seat equipped with a headrest, with the head oriented to the 0° azimuth indicated by a white LED light spot.

The speaker array was combined with an array of 188 white light-emitting diodes (LED, 2.52 lux, 0.6° visual angle) mounted in azimuthal steps of 1° at eye level. The LEDs were controlled by 51 printed circuit boards, which were arranged on top of the loudspeakers. Four infrared (IR)-sensitive phototransistors were mounted on each board, arranged at the same angular distances as the LEDs, but covering an additional 8° on both sides. A customized IR torch served as pointing device (IR-torch, Solarforce L2 with 3 W 850 nm NVG LED, Fulidat Electronics Limited, Kowloon, Hong Kong). The subtended angle of the IR light beam covered a maximum of 8° at the level of the LEDs. The mean location of all activated IR-sensitive phototransistors was computed online, and the corresponding LED flashed up as a visual feedback of the pointing direction for the participant.

Stimulus generation and test procedures were controlled by MATLAB^® (2007b; MathWorks Inc., Natick, MA, USA). Stimuli were digitally generated by two PC-controlled instruments from TDT (RX8 modules System III) devices (Tucker-Davis Technologies, TDT, Alachua, FL, USA).

Stimuli were low-frequency (LF: 0.3–1.2 kHz) or high-frequency (HF: 2–8 kHz) Gaussian noise bursts. These spectra were chosen to selectively address binaural signal processing based on ITD or ILD. Both noises had a bandwidth of two octaves. Signal duration was 500 msec. Signals were presented at 40 dB sensation level (see Individual Determination of Stimulus Intensity). The level of the stimuli was not roved in order to enhance comparability with earlier studies [Senn et al., 2005; Francart et al., 2018] and because there seems to be no influence of different levels on localization [Dillon et al., 2017b; Buss et al., 2018].

Depending on the age of the participants, their ability to concentrate, and the individual need for breaks, test sessions took about 2 h. The applied tests make only low demands on cognitive effort of the participants and have previously been successfully used for the evaluation of spatial hearing skills in adults with acquired brain lesions [Witte et al., 2012], as well as in schoolchildren starting at the age of 6 years [Kühnle et al., 2013].

Individual hearing thresholds for LF and HF signals were obtained from 0° at the beginning of each testing session using a staircase (heard/not heard) procedure. Starting at a level of 60 dB SPL, intensity was decreased or increased in 3 dB steps. A single test run was terminated after eight turn points. These respective threshold values for both frequency bands were used to set the presentation level for the subsequent tests at 40 dB sensation level. Presentation levels ranged from 50 dB SPL to 70 dB SPL (mean: 62 dB SPL).

MAAs were measured by applying a three-alternative forced-choice design using a 1-up/1-down paradigm aiming at a 50% correct response level on the psychometric function [Levitt, 1971]. In the three-alternative forced-choice testing, the participants were asked to differentiate between two reference signals with the same azimuth and one test signal with a different azimuth, with random order of reference and test signals in the stimulus triplet. Responses were given by pressing the appropriate one of three push buttons on a response box. Reference locations were at ±4°, ±30°, and ±60° azimuth. At the start of each test run, the deviant sound was presented with spatial disparity of 30°. Deviants were always located more laterally with respect to the reference location. Spatial disparity between reference and deviant sound was decreased after each correct response and increased after each incorrect one. Any change from a correct to an incorrect response and vice versa was marked as a turn point. The step size was reduced from 4.3° to 2.15° after two turn points. For a single reference location, the test run was terminated after eight turn points, and thresholds were calculated as the mean azimuthal difference between reference and deviant sounds at the last six turn points. The participants were instructed to face straight ahead and to stay in that position during all stimulus presentations. Prior to actual testing, the participants were given three practice trials to make sure they had understood the procedure and were able to correctly handle the buttons of the response box. In addition, the observer was sitting outside the arc behind the participant to control that the participants gave consistent responses and that the staircase procedure converged. A single test run, i.e., the acquisition of the MAA for one target location, took about 1–3 min. Test order of these conditions was varied pseudo-randomly over the group to avoid a potential presentation order bias.

The presentation conditions have been described in detail by Ludwig and colleagues [Ludwig et al., 2021]. Auditory localization was tested for six azimuthal locations ±4°, ±30°, and ±60°. Each location was tested five times in random order. LF and HF signals were tested separately, resulting in 60 signal presentations (6 locations × 5 repetitions × 2 stimulus conditions). The participants were instructed to face the 0° loudspeaker and look at a fixation point during stimulus presentation. After each signal presentation, the participants were asked to indicate the perceived sound location with the IR torch. For that, they were allowed to turn their head to the perceived sound location, after which they again faced straight forward. Prior to actual testing, participants were presented three practice trials to familiarize themselves with the procedure.

The MAA thresholds were analyzed using repeated-measures analysis of variance (ANOVA) with within factors “FREQUENCY” (LF vs. HF), “SIDE”, and “LOCATION”. Post hoc t tests or, in cases of failed normality tests, Wilcoxon signed rank tests were carried out to explore significant main effects or interactions.

The MAA thresholds were standardized (z-transformed): $z−score=x−x¯norm/stdxnorm$⁠, using mean and standard deviation of the corresponding age-matched control group. The control group consisted of adults aged 20–29, 30–39, 40–59, and 65–83 years. Z-scores above 1.64 indicate a significant one-tailed t test result at a 5% type I error rate. An ANOVA and a multiple regression model were computed to detect differences in localization ability on the MAA at locations 4°, 30°, and 60° on the CI and on the NH side.

All analyses were calculated for both frequencies and all signal locations on the CI side and on the NH side. p values were Bonferroni corrected.

On average, the group of participants showed a high variability of MAA thresholds, with interquartile ranges of 7.2°–35.8° on the CI side and 10.7°–36.4° on the NH side (Table 2). Control participants between 20 and 69 years showed a smaller interquartile range of 1.9°–22° (Table 3).

MAA did not increase with increasing laterality. For signals presented from 60°, MAA thresholds up to 40° indicate that some participants had difficulty differentiating between signals from 60° and for the loudspeaker furthest away from the midline in our array (98°, Fig. 1). The ANOVA yielded no main effects or interactions.

Due to the high interindividual variability, the group median scores do not reflect individual performance (Fig. 2). There were participants with thresholds comparable to the corresponding age group in the LF and also in the HF condition. However, there were also participants with thresholds up to 40° irrespective of the signal location or the frequency, which might indicate guessing in the absence of sound location perception.

Age is a main factor in explaining changes in MAA performance [Freigang et al., 2015]. Thus, we compared the present MAA results to the corresponding age-matched control data (Table 3; data for 20–29-year-olds and 65–83-year-olds from Freigang et al. [2014]).

The first visual observation revealed that the CI-SSD participants did not show elevated MAA in general and that it is more appropriate to analyze performance on the CI and NH sides separately and individually. Since elevated z-scores could have occurred by chance, participants were divided into two groups according to incidences of z-values in each frequency band: good performers, with MAA that did not differ significantly (z < 1.64) from the performance of NH controls at least at 4 of the 6 tested locations, and poor performers, with MAA that differed at more than two locations from the controls.

Eight participants were able to differentiate between locations as well as NH controls in at least one of the two frequency bands. Six participants showed good performance in the LF condition and eight in the HF condition. Most difficulties occurred at 4° and at 30° (Fig. 3).

Twelve participants had poor MAA results in the LF condition and 10 in the HF condition (Fig. 4). MAA thresholds were largest at 4° and 30°. In the HF condition, elevated MAA at 30° was more pronounced on the CI side than on the NH side, whereas in the LF condition, they were about equal at both sides.

Seven participants were tested twice at the same location once wearing the CI and once in an SSD condition without the CI. Three of them were tested at 4° on the CI side, and four of them were tested at 4° on the NH side. Five participants performed better in the binaural than in the monaural condition irrespective of the signal location and thus benefitted from wearing the CI (Fig. 5).

Most of the participants were able to differentiate between signals on the CI side and on the NH side in the localization task [Ludwig et al., 2021]. In the MAA task, however, about one-third of them were able to discriminate between adjacent stimuli on either side. Six (for LF signals) and seven (for HF signals) out of 18 participants showed good performance for MAA and localization (Table 4, lines 2–9). Five participants were bad performers in the MAA task but not in the localization task (Table 4, lines 11–15). Three participants showed variable performance regardless of the condition, and two participants had no benefit from the CI in all conditions.

An ANOVA with a multiple regression model was computed concerning the influence of absolute localization ability at three different locations on the MAA on the CI and on the NH side. The localization accuracy was related to MAA for LF signals (F_(3,17) = 6.883, p = 0.004). Post hoc linear regression revealed that less accurate localization at 4° on the CI side implied less accurate spatial discrimination at 4° on the CI side, as measured by the MAA (r = 0.772, t₁₈ = 2.843, p = 0.013, Fig. 6).

To our knowledge, the present study is the first investigation of MAAs in CI-SSD participants. MAA thresholds for nonspeech signals were estimated on the CI side and on the NH side. The single-subject analysis revealed elevated thresholds in particular at 30° and at 4° on both sides. At 60°, many participants had MAA in the range of the controls. The participants of the present study were already tested in the absolute localization task in the study of Ludwig and colleagues [Ludwig et al., 2021]. Six (in the LF condition) and seven (in the HF condition) of the participants who were able to localize signals in the absolute localization task were also able to discriminate spatially separated stimuli in the MAA task.

In NH children [Kühnle et al., 2013] and adult participants [Freigang et al., 2014], MAA systematically increased from frontal to lateral. Also, unaided hearing-impaired adults (measured via headphones) showed increasing MAA with increasing laterality [Cai et al., 2015]. Using a free-field left-right discrimination task, Senn et al. [2005] showed increasing MAA with increasing laterality in bilateral CI users. In contrast, hearing-impaired children tested without their hearing aids did not consistently benefit from frontal positions of the sound sources in the MAA task [Meuret et al., 2017]. Zaleski-King et al. [2019] measured MAA via headphones in bimodally fitted adults and showed no increasing MAA with increasing laterality. Most of the participants of the present study also gained no benefit from frontal signal presentation. The presence or absence of increasing MAA with laterality in the above results may be explained by different degrees of asymmetry in hearing thresholds between the ears. There is evidence that sound lateralization is encoded by the comparison between two opponent hemispheric channels with balanced activity at the midline [Stecker et al., 2005; Magezi and Krumbholz, 2010; Briley et al., 2013; Briley and Summerfield, 2014]. A large asymmetry between devices, or between a device and the NH side, may interfere with this encoding scheme and reduce the spatial resolution at the midline.

Ludwig et al. [2021] showed that 87% of their CI-SSD participants were able to use ITD information for localization of LF sounds. Only six of those participants also showed good performance in discrimination of LF sounds in the present study. This is in accordance with the study of Dorman et al. [2015], whose CI-SSD participants performed better than chance in the LF condition. Envelope ITD cues are preserved during CI preprocessing [Todd et al., 2019] and also support localization [Kerber and Seeber, 2013], lateralization [Majdak et al., 2006], and MAA [Senn et al., 2005] in bilateral CI users. Results in CI-SSD participants differ: Dirks and colleagues [Dirks et al., 2019], using low-pass filtered words providing primarily ITD cues, found no influence of envelope ITD on localization. In contrast, Bernstein and colleagues, using bursts of electrical pulse trains, found an influence on ITD detection [Bernstein et al., 2018]. Some of the present participants appeared to be able to use envelope ITD information to discriminate sound locations (MAA) since there was no appreciable ILD in the LF condition.

The six good performers had MAA in the normal range at frontal locations and thus seem to have adapted to use ITD cues, whereas the poor performers have not. Using white noise bursts, Senn et al. [2005] measured the left-right discrimination ability in bilaterally implanted listeners and found MAA thresholds between 3° and 8° at the front. The present median MAA at 4° was about 21°. The difference between the location of signal presentation and the threshold (4° vs. 25°) corresponds to 176 µs ITD difference. Median MAA at 60° was about 22°. The difference between these locations (60° vs. 82°) corresponds to only 125 µs. The larger ITD difference at frontal locations did not lead to better MAA in poor performers.

Just noticeable differences (JNDs) for ITD were measured in bilaterally implanted CI listeners: they ranged from 1,354 to 3,265 µs for wide-band noise [Laback et al., 2004]. For envelope ITD, JND ranged from 135 to 683 µs [Noel and Eddington, 2013] and 100–16,000 µs [Senn et al., 2005]. Note that the last value is far outside of the physiological range of ITD. In contrast, Francart and colleagues [Francart et al., 2018] measured JND-ITD in CI-SSD participants of 200–2,000 μs, with a median of 438 μs. This is in accordance with the present data since stimulus presentation at 60° corresponds to 468 µs, and all of the good performers and some of the poor performers correctly perceived the hemifield in which the stimuli were presented.

NH listeners can adapt within a few days to distorted ITD [Javer and Schwarz, 1995; Trapeau and Schönwiesner, 2015]. Patients would be expected to adapt to ITD shifts over time. However, the duration of CI usage was no reliable predictor for the adaptation to distorted envelope ITD. For example, ID 12_75 had 31.5 months of CI usage and was a poor performer, but ID 3_27 had 1.1 months of CI usage and was a good performer. Since some of the poor performers had normal or near-normal MAA at 60°, they may have used other cues than ITD at more lateral locations (see below).

A reason for the increased MAA compared to the results in the absolute localization in the LF condition could be differences in signal transmission between assisted and unassisted ear: (1) it has been shown that a delay in timing between a CI and the NH ear [Seebacher et al., 2019] or between a CI and a hearing aid in bimodal patients [Francart et al., 2009; Zirn et al., 2015, 2019] may influence localization. In the study of Seebacher et al. [2019], electrical stimulation was about 1–2 ms ahead of time for frequencies above 1.5 kHz, and the best performance in localization was reached with a 1 ms delay. (2) A CI also does not simulate the dispersion of transmission times between low and high frequencies due to the traveling wave delay in the cochlea [Dau et al., 2000]. There are interaural pitch mismatches, in which frequencies allocated to the electrodes (and contralateral HA or NH) differ from the electrically stimulated cochlear frequencies [Reiss et al., 2014; Bernstein et al., 2018], and there is some spread of excitation from the electrodes that further mismatches tonotopic alignment. Electrode arrays in CI are also typically shorter and too thick to reach the most apical regions of the basilar membrane. This results in the stimulation of the CI being temporally and spectrally compressed relative to the stimulation of the NH ear. Aronoff et al. [2015] simulated an either temporal or spectral compression of a unilateral CI in NH listeners and found that these compressions limit binaural fusion. They argued that both compressions may be factors decreasing the binaural abilities in CI-SSD patients [Aronoff et al., 2015]. The spectral compression seemed to be more detrimental than the temporal compression. The brain may adapt to the spectral mismatches by remapping the pitch over the course of the rehabilitation [Reiss et al., 2008; Svirsky et al., 2015]. However, we found no clear relationship between duration of CI use and MAA. Dirks and colleagues [Dirks et al., 2022] also found no clear relationship between the frequency-allocation shift and perceptual changes.

In NH children [Kühnle et al., 2013] and adults [Wightman and Kistler, 1997; Freigang et al., 2014, 2015], MAA thresholds for LF noise are better than for HF noise. This pattern is reversed in SSD participants provided with a CI [Firszt et al., 2012; Dorman et al., 2015; Mertens et al., 2016; Dirks et al., 2019] and in bilateral CI users [Seeber and Fastl, 2008; Van Hoesel et al., 2009; Aronoff et al., 2010], who localize HF signals better than LF signals, presumably because ITD are degraded. However, results of the present study suggest that some CI-SSD participants may have some residual access to ITD and can use that information to discriminate sound source locations.

ILD information is also degraded in these participants. The ILD for HF sounds as used in our study varies from about 3° at frontal to 15° at lateral locations in NH participants [Ludwig et al., 2021]. Correspondingly, MAA for these sounds increases from 3° at frontal to 20° at lateral locations in NH participants. ILD is severely compressed by the automatic gain control of CI [Francart and McDermott, 2013; Dorman et al., 2014]. In bilaterally implanted participants, an ILD difference of only about 1.2 dB has been measured between broadband sounds from 15° to 45° (Medel) [Dorman et al., 2014] and about 4 dB between sounds from 10° to 70° (Cochlear) [Potts et al., 2019]. In CI-SSD participants, Francart et al. [2018] found that ILD of 2 dB yielded the largest extent of lateralization and was the most symmetric. In our data, MAA for HF signals did not vary systematically with source azimuth, suggesting that the sensitivity to ILD did not change with source azimuth in these patients. Median MAA thresholds at 4° were about 16°, and at 60°, they were 22°. The elevated MAA and lack of variation with source azimuth are likely due to the ILD compression in the CI. Furthermore, the auditory system of NH persons adjusts quickly to distorted ILD [Kumpik et al., 2010; Keating and King, 2013], which may not be possible due to the compression.

Taken together, there were more good performers in the localization task than in the MAA task. Differences in signal transmission between the CI side and the NH side in CI-SSD participants may have thus played a minor role in sound localization in these participants but might have had a greater impact on spatial discrimination. Although CI-SSD participants were able to localize sounds, interaural differences were presumably too compressed to allow discrimination between spatially separated stimuli on either side. Although the LF and HF conditions do not perfectly separate interaural differences, this applies equally to ITD and ILD.

Monaurally tested blind individuals localize sounds more accurately than sighted participants, presumably due to greater sensitivity to spectral cues [Doucet et al., 2005; Voss et al., 2011]. Behavioral evidence in unilaterally deaf participants supports the idea that monaural spectral cues contribute to sound localization in the horizontal plane [Slattery and Middlebrooks, 1994; Van Wanrooij and Van Opstal, 2004; Agterberg et al., 2014]. Slattery and Middlebrooks [Slattery and Middlebrooks, 1994] used stimuli without interaural cues and investigated SSD participants and NH subjects with one ear plugged. They showed that NH listeners with a monaural earplug had a lateral response bias to the NH ear, whereas SSD participants localized both sides equally well and thus presumably had adapted to spectral cues. Out of the two participants of the present study with the longest duration of deafness, one was a good performer and the other was a poor performer; thus, the duration of deafness seemed to be no reliable predictor for a possible adaptation to spectral cues.

Five participants benefitted from the CI and improved MAA thresholds compared to the monaural SSD condition and thus may (apart from the use of ITD and ILD cues) have used binaural spectral cues resulting from the CI. Two participants performed better in the SSD condition than in the CI-SSD condition, which is in accordance with the study of Ausili and colleagues [Ausili et al., 2019], who tested spatial discrimination in NH listeners using real-time vocoders and showed that spatial discrimination is still possible on the acoustic hearing side.

The omnidirectional microphones used in CI lack direction-specific filtering of the external ear and thus provide very limited spectral cues. However, one of our participants may have possibly benefitted from spectral cues. This participant had no benefit from the CI in the absolute localization task of our previous study [Ludwig et al., 2021] but was a good performer in the current MAA task in the CI-SSD and SSD condition. The other participants did not appear to use spectral cues even in the NH ear because MAA thresholds were equally poor on both sides.

There is inconsistent empirical support for differential impact of spectral cues on frontal and lateral locations. For example, Häusler and colleagues [Häusler et al., 1983] hypothesized that the diminished ability to make use of spectral information leads to increased MAA thresholds at the sides. In contrast, the improvement in sound localization sometimes reported in blind individuals has been attributed to their greater sensitivity to spectral cues corresponding to lateral sound source locations [Doucet et al., 2005; Voss et al., 2011]. A study in mice [Ito et al., 2020] showed that neural responses to sounds from the frontal field depend on spectral cues, whereas responses to sounds from the lateral field depend on ILD. In our participants, MAA thresholds were normal or near normal at 60°, and most problems occurred at 30° and at 4°, suggesting that the participants may have used spectral cues at lateral source azimuths to some degree.

The present results show that a CI can, to some extent, restore the spatial discrimination ability in SSD participants. Eight out of 18 participants had MAA thresholds in the normal range. Even in the good performers, difficulties remained at frontal locations on the CI side as well as on the NH side. This applies equally to low- and high-frequency stimuli. Five out of 7 participants benefitted from the CI and improved MAA thresholds compared to a monaural SSD condition. Generally, there were more good performers in the localization task than in the MAA task. Although CI-SSD participants were able to localize a sound, it apparently required better balance between both sides to discriminate between spatially separated stimuli on the CI and on the NH side.

We would like to thank all SSD participants and control participants for their participation.

This study was approved by the Ethik-Kommission an der Medizinischen Fakultät der Universität Leipzig (Ethics Committee of the University of Leipzig) (decision No. 127/17-ek). All procedures were in accordance with the ethical standards of the Institutional and/or National Research Committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. Written informed consent was obtained from all individual participants included in the study.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. The study was supported by Leipzig University for Open Access Publishing.

Alexandra Annemarie Ludwig, Rolf-Dieter Battmer, and Arneborg Ernst designed the study. Alexandra Annemarie Ludwig and Rolf-Dieter Battmer conducted the experiments. Alexandra Annemarie Ludwig, Sylvia Meuret, and Michael Fuchs analyzed the results. Alexandra Annemarie Ludwig and Marc Schönwiesner wrote the manuscript.

All data generated or analyzed during this study are included in this article. Further inquiries can be directed to the corresponding author.