Introduction
The elderly population aged 65 years or older in Korea is 17.5% of the total population, and is projected to increase to 25.14% in 2025, entering a super-aged society, and exceeding 37.61% in 2035 and 56.40% in 2050 [1]. As the elderly population grows, the difficulties that older individuals experience in everyday life become increasingly important. Age-related hearing loss typically presents as a bilateral rather than unilateral hearing impairment, which is largely attributed to systemic biological aging processes [2-5] and the fact that both ears are equally exposed to environmental factors such as accumulated noise [6,7], medications [7-9]. Thus, participants with bilateral sensorineural hearing loss often struggle to localize sounds accurately and understand the target speech in complex, noisy environments [10-12].
Sound localization relies on the time difference between the ears for low frequencies and the level difference between the ears for high frequencies [13,14]. Although sound localization is essential for navigating everyday environments, older adults with hearing loss often show reduced accuracy in localizing sounds across multiple spatial dimensions, including the horizontal (left-right) plane [15,16], the front-back dimension [17-20], and the vertical plane [11,21-23]. Moreover, in individuals with sensorineural hearing loss, damage to the cochlea impairs its ability to finely discriminate sound patterns—an essential function for distinguishing subtle differences in sound. This impairment reduces the ability to filter speech from background noise [24,25].
Hearing aids are commonly used to improve speech perception not only in quiet conditions but also in noisy environments. Enhancing sound localization is one of the important strategies for improving speech perception in noise. However, studies comparing sound localization before and after hearing aid use have reported inconsistent findings. Some studies have shown that optimizing hearing aid technologies as microphone placement [26], wireless synchronization [27], and directional processing [28], can significantly enhance spatial hearing abilities, particularly in reducing front-back confusions. Others have found no significant change [29-31], while some have even reported deteriorated horizontal sound localization immediately after hearing aid fitting compared to unaided conditions [28,32,33].
In contrast, more consistent positive outcomes have been reported when evaluating speech recognition in noise. Furthermore, several studies suggest that the ability to understand speech in noise improves over time following hearing aid use [34-36]. Nevertheless, some evidence indicates that this improvement may not be progressive, but rather that hearing aid users maintain a stable level of benefit without further enhancement over time [37-39].
Accordingly, in the present study, the authors aimed to observe how sound localization and speech perception in noise change not only immediately after hearing aid use but also over time with continued use. The researchers had previously conducted a study using disyllabic words to assess speech perception in noise over one year following hearing aid use and found no significant changes from one month to 12 months after fitting [37]. In the present study, we aimed to re-examine the changes in speech perception in noise over time using the K-Matrix sentence test, which better reflects real-life listening situations.
Subjects and Methods
Subject
The study included 13 elderly participants with bilateral sensorineural hearing loss (mean age=73.69±6.75 years; 7 males and 6 females). All of whom were first-time hearing aid users and wore their devices for an average of at least 8 hours per day during the study period. The pure-tone averages (0.5, 1, 2, and 4 kHz) of the participants without hearing aids at the time of pre-fitting and with hearing aids at 1, 3, and 6 months after fitting are presented in Table 1. All participants used bilateral Phonak hearing aids; seven wore receiver-in-canal (RIC), two wore completely-in-canal (CIC), and four wore in-the-canal (ITC). Hearing aids were programmed to fit the target specified by the NAL-NL2 formula for each hearing-impaired participant and verified using Audioscan Verifit (Etymotic Design Inc.). Participants had at least three follow-up visits at 2-week intervals during the initial adaptation period, with the gain set to 100% of the target value based on each participant’s audiogram. The adaptive directional microphone, noise reduction, and feedback management functions were set to their default settings. The details of the hearing aid types and the number of channels for hearing aids are presented in Table 1.
Participants with chronic otitis media, retro-cochlear lesions, endolymphatic hydrops, or hearing loss with a conductive or surgically correctable component were excluded. The study was conducted by the Declaration of Helsinki and the recommendations of the Institutional Review Board of the Nowon Eulji Medical Center, with written informed consent from all participants (EMCS 2022-11-005).
Procedure
In this study, sound localization and speech recognition in noise were assessed at four times, pre-fitting and post-fitting at 1-, 3-, and 6-months, in participants with no prior hearing aid experience. The tests were conducted in a soundproof room to evaluate the effects of hearing aid use and adaptation over time.
Sound localization test
The experiment was conducted in a soundproof booth equipped with seven speakers arranged in a semicircle, each placed at 30-degree intervals, as shown in Fig. 1. The speakers were positioned 1 m away from the participants. The stimulus used in the evaluation was speech shape noise with a level of 65 dB SPL. Participants were seated in a chair in the booth following the researcher’s instructions. To identify the location of each speaker, the corresponding speaker number was displayed in the upper right corner, and participants were instructed to verbally report the number of the speaker from which the sound was presented. Sounds were randomly presented three times from each of the seven speakers (seven speakers×three presentations each), resulting in 21 sound presentations in total. The experimenter outside the sound booth documented participants’ responses, which were evaluated using three performance metrics: 1) the percentage of correct responses, which indicates the proportion of accurately identified stimuli; 2) the mean absolute error, which represents the average of the absolute differences between predicted and actual values; and 3) the root mean square error, which quantifies the overall magnitude of error by assigning greater weight to larger errors, thus providing a comprehensive measure of accuracy.
Speech recognition in noise
To measure sentence recognition in noise, we used the Korean Matrix sentence recognition test [40-42]. All the Korean Matrix sentences used are semantically unpredictable, but they have the same grammatical structure (name, adjective, object, numeral, and verb) because each sentence was generated using a 5×10 base word matrix (10 names, 10 adjectives, 10 nouns, 10 numerals, and 10 verbs). The general principles and applications of the Matrix sentence-in-noise recognition tests are described in previous studies [43-45]. We utilized two types of noise in the Korean Matrix sentence recognition test: speech-shaped noise and the International Speech Test Signal. The speech-shaped noise was generated by superimposing the Korean Matrix sentences, so the long-term spectrum of speech and speech-shaped noise was the same. The international speech test signal noise [46] is considered non-intelligible speech noise because it consists of randomly remixed speech segments (100-600 ms) from six languages, which are spoken by six different female talkers reading “The North Wind and the Sun.” The Korean Matrix sentence recognition test was conducted using Oldenburg Measurement Applications software (HörTechg GmbH). The test sentences and noise were presented through a Fireface UCX digital-to-analog converter (RME), and the stimuli were delivered by a loudspeaker located 1 m in front of the participants. During the test, the noise level was fixed at 65 dB SPL while the sentence level was adjusted according to the participant’s response based on the maximum likelihood estimator [47]. Consequently, we measured the speech reception thresholds (SRTs) of 50% intelligibility by measuring the signal-to-ratio required to achieve 50% recognition.
Statistical analysis
A one-way repeated measures analysis of variance (RM ANOVA) was conducted to evaluate the effects of hearing aid usage on sound localization and speech recognition in noise over time. Post-hoc paired t-tests, significance levels were set at 0.05 for multiple comparisons after applying Bonferroni’s correction to the p-values. The within-participants factor was time, with four levels: pre-fitting, 1, 3, and 6 months post-fitting. All statistical analyses were performed using IBM SPSS software (ver. 30.0; IBM Corp.).
Results
Sound localization test
Fig. 2 illustrates individual changes in localization accuracy across four time points—pre-fitting, and 1-, 3-, and 6-months post-fitting—demonstrating varied but generally improving trends following hearing aid use. Participants who achieved 100% accuracy at the pre-fitting session (S9, S11, and S12) were excluded from statistical analysis due to a ceiling effect. RM ANOVA revealed a significant main effect of time on percent correct total (F (2.358, 21.461)=4.152, p=0.024), mean absolute error (F (2.436, 21.920)=4.558, p=0.017), and root mean square error (F (2.574, 23.169)=4.117, p=0.022) (Fig. 3 and Table 2). Post-hoc pairwise comparisons revealed a significant improvement in the percent correct total between the pre-fitting and 6-month conditions (p=0.047). The mean absolute error and root mean square error showed trends toward significance between the pre-fitting and 6-month conditions (p=0.053 and 0.054, respectively). Bubble charts were utilized to visualize the distribution of listeners’ responses to auditory stimuli and to represent both localization accuracy and response bias intuitively. Bubble charts were utilized to visualize the distribution of listeners’ responses to auditory stimuli and to intuitively represent both localization accuracy and response bias. By mapping stimulus azimuth and response azimuth on a two-dimensional plane, and representing the frequency of each stimulus-response pair with the size of the bubble, this method effectively conveys not only the correctness of responses (via diagonal alignment) but also the direction and magnitude of errors. Fig. 4, which illustrates responses under four listening conditions (pre-fitting, 1, 3, and 6-month aided), demonstrates a visual trend toward enhanced spatial mapping over time. Initially, responses were widely dispersed and misaligned with stimulus azimuths. As hearing aid use progressed, response distributions increasingly aligned along the stimulus-response diagonal, a pattern that was particularly evident at the 6-month mark.
Speech recognition in noise
The RM ANOVA showed a significant effect of time for speech-shaped noise (F (3, 36)=10.216, p<0.001) and the International Speech Test Signal (F (3, 36)=6.338, p=0.001). Post-hoc analyses for speech shape noise showed a significant improvement in SRT between the pre-fitting condition and at 1 month (p=0.014), 3 months (p=0.025), and 6 months (p=0.008) postfitting. Additionally, post hoc analysis of the International Speech Test Signal revealed a significant improvement in SRT between the pre-fitting and 1-month (p=0.022) and 3-month (p=0.044) post-fitting conditions. However, no significant differences were observed between time points after hearing aid fitting (p>0.05) (Fig. 5 and Table 3).
Discussion
This study investigates the effects of hearing aid use on sound localization and speech perception in noise among elderly individuals with bilateral sensorineural hearing loss. Longitudinal assessments were conducted multiple times: without a hearing aid at pre-fitting and at 1-, 3-, and 6-month post-fitting. Sound localization performance showed a general trend of improvement over time, as indicated by a significant main effect of time in all three metrics—percent correct, mean absolute error, and root mean square error—and by increasingly aligned response patterns along the stimulus-response diagonal over time in the bubble chart. Meanwhile, speech recognition in noise improved immediately after hearing aid fitting; however, no further improvement was observed over time.
Theoretically, wearing hearing aid is expected to improve the sound localization performance by providing reliable and specific timing and intensity cues across the frequency spectrum and by restoring symmetry of hearing between both ears. Some studies have shown that front-back confusion in sound localization can be significantly reduced by optimizing microphone placement on hearing devices. Positions near the ear canal entrance or within the concha demonstrate superior performance compared to behind-the-ear placements [26]. Activation of wireless synchronization in bilateral hearing aids has also been found to significantly reduce front-back localization errors for broadband stimuli in hearing-impaired listeners, indicating improved directional hearing [27]. Additionally, the use of cardioid directional microphones in hearing aids has been shown to significantly improve horizontal localization performance over time by reducing front-back confusions [28]. Furthermore, integrating motion-based beamformer steering into hearing aids significantly enhances speech understanding and sound localization during walking, thereby improving the overall listening experience in real-world environments [48].
However, there is no evidence showing improvement in horizontal plane localization that reflects real-life situations. Using single words in speech-shaped noise at 0 dB signal-to-noise ratio, no statistically significant differences in localization performance were found between the pre-fitting and post-fitting aid conditions after three weeks of use [49]. Similarly, short sentences filtered into low-frequency (200-600 Hz) and high-frequency (1500-4500 Hz) bands were used as auditory stimuli to compare sound localization performance, and the results also indicated no significant differences between the pre-fitting and post-fitting conditions [30]. Horizontal localization performance under various hearing aid conditions was compared in adults with mild-to-moderate sensorineural hearing loss, and it was found that bilateral hearing aid use did not significantly improve localization accuracy compared to the unaided condition [31]. In contrast, one study reported that sound localization performance could even deteriorate with the use of hearing aids. This study compared three listening conditions in older adults—adaptive-directional mode, omnidirectional mode, and unaided listening—and found that localization performance was significantly reduced when hearing aids were worn, particularly under the adaptive directional microphone configuration [33].
Studies investigating whether localization accuracy improves after an adaptation period following hearing aid fitting have also failed to show positive results. A study found no significant improvement in localization accuracy following hearing aid use, and the root mean square error measured at 2 weeks and 2 months post-fitting showed no significant change over time [28]. Similarly, another study reported no statistically significant improvement in horizontal localization accuracy either immediately after fitting or after a 3-week adaptation period [32].
In this study, the use of hearing aids yielded more positive results in horizontal sound localization. Although statistical significance was only observed for percent correct between the unaided and 6-month conditions, the consistent downward trends in both mean absolute error and root mean square error scores over time suggest a potential improvement in sound localization performance with continued hearing aid use and auditory adaptation.
Localization performance was notably lower in the Omni microphone and Inter-Ear Compression conditions compared to the unaided condition, suggesting that certain hearing aid configurations may degrade spatial perception rather than improve it [50]. Identifying the location of sound involves not only amplification but also auditory training. Previous studies have demonstrated that both laboratory- and home-based localization training programs can significantly improve frontback localization in individuals with hearing impairment, suggesting that hearing aids alone may be insufficient for achieving accurate spatial perception in complex listening environments [51]. These findings emphasize the importance of integrating hearing aid technology with targeted auditory rehabilitation to achieve optimal spatial.
It is well established that hearing aids improve speech perception in noise. Especially directional microphones significantly improved speech perception in noise compared to omnidirectional microphones and digital noise reduction alone. While one study suggests that the combination of directional microphones and digital noise reduction did not provide a statistically significant improvement over using directional microphones alone [52], another study reported better performance with the combined strategy [53]. However, whether the improvement in speech perception in noise provided by hearing aids increases over time remains unclear, as studies have reported mixed results. For example, Gatehouse found that while speech perception remained stable in the unaided ear, a significant improvement was observed in the aided ear between 4 and 12 weeks after fitting [54]. A similar pattern was reported by Munro and Lutman [35], who found that improvement in speech perception benefit over time occurred only in the fitted ear, with no change in the unfitted ear. Notably, these effects were evident only at the highest presentation level (69 dB SPL), likely because the amplified speech at this intensity was unfamiliar to participants with hearing impairment. In contrast, another study conducted in a similar setting found no evidence of auditory acclimatization in hearing-aided users. No improvement over time (from baseline to 12 weeks) was detected in the fitted ears of new hearing-aided users relative to that in the unfitted ears of the same users or to that in the fitted ears of experienced users [55]. In the present study, while a difference in speech perception was observed before and after hearing aid use, no change was found over time. This finding is consistent with the results of a previous study conducted in our laboratory with a 12-month follow-up period, which used disyllable words rather than sentences [37].
The current study has several limitations: the sample size was relatively small; the follow-up period was limited to 6 months; and, to avoid the ceiling effect in sound localization, three participants who scored 100% on the pre-fitting percent correct total were excluded. As a result, only 10 participants were included in the analysis, which may have weakened the statistical power of ANOVA. Furthermore, the hearing aid types varied among participants, including both RIC, CIC, and ITC models. Nevertheless, given the common challenge of participant attrition in longitudinal studies, the current prospective design remains valuable in that it successfully tracked a meaningful cohort of hearing aid users over time.
In conclusion, this study found that sound localization tended to improve gradually over time following hearing aid use. In contrast, speech perception in noise showed significant gains during the initial phase of hearing aid use but then reached a performance plateau. These findings emphasize the necessity for additional interventions to maximize long-term advantages. Future research should include a larger cohort with diverse hearing profiles, incorporate more naturalistic listening conditions, and evaluate the influence of hearing aid processing features under user-customized settings.