Speech understanding and extended high-frequency hearing sensitivity in blast-exposed veterans^a)

It is not uncommon for clinical audiologists to encounter patients who report significant hearing difficulties despite having normal- or near-normal-hearing sensitivity (Billings , 2018; Edwards, 2020; Koerner , 2020). This issue is especially prevalent in Service Members and Veterans who have increased risk of damage to the central nervous system from occupational hazards such as traumatic brain injury or exposure to high-intensity blasts (Grant , 2021; Tepe , 2020). Indeed, normal-hearing Veterans with a history of blast exposure commonly report auditory difficulties such as issues understanding speech in complex listening environments with background noise or reverberation, listening to rapid speech, or listening over the telephone (Papesh , 2019; Papesh , 2021; Saunders , 2015). This patient population also tends to demonstrate poorer performance on behavioral tests of central auditory function, including tests of sound localization and speech understanding in complex listening conditions, compared to individuals with no history of head injury (Gallun , 2012; Kubli , 2018). However, results have not been consistent across studies that used different measures, such as those that assess the perception of speech in noise, rapid time-compressed speech, or the perception of speech in noise in spatialized auditory scenes (Papesh , 2019; Papesh , 2021; Saunders , 2015). These discrepancies necessitate a more thorough investigation into the effects of blast exposure on speech perception using a comprehensive battery of tests to identify measures that may be clinically useful for this patient population.

Whereas findings from the previous studies discussed above suggest that hearing difficulties in this normal-hearing patient population likely stem from damage to central rather than peripheral auditory structures, the majority of these studies have only considered hearing thresholds at audiometric frequencies between 250 and 8000 Hz. Measuring pure-tone thresholds in this frequency range is standard clinical practice for diagnosing hearing impairment because it is traditionally thought that frequencies below approximately 8 kHz contain most of the acoustic energy important for speech perception. However, it is well established that humans can perceive acoustic information up to 20 kHz, and recent work has demonstrated that acoustic energy in extended high-frequency (EHF) regions beyond 8 kHz is present and audible in speech (e.g., Monson and Caravello, 2019; Moore , 2008; Motlagh Zadeh , 2019). Evidence suggests that EHF hearing is also important for localization and speech understanding in background noise (Badri , 2011; Best , 2005; Flaherty , 2021; Levy , 2015; Mishra , 2022; Monson , 2019; Motlagh Zadeh , 2019; Polspoel , 2022; Trine and Monson, 2020; Yeend , 2019). Therefore, it is possible that degradation of hearing sensitivity in the EHF range could limit access to acoustic cues that are important for speech perception.

There has also been increasing interest in the use of EHF hearing thresholds for predicting subclinical damage in the standard audiometric range, which may also underlie speech understanding in noise difficulties in normal-hearing patients. For instance, it is thought that damage to the peripheral auditory system from ototoxins (Blankenship , 2021; Konrad-Martin , 2010) or noise exposure (Ahmed , 2001; Le Prell , 2013; Liberman , 2016; Mehrparvar , 2011) may be apparent in EHF regions before any measurable pure-tone hearing deficits appear in lower-frequency regions. Recent efforts have examined the use of EHF thresholds as an indicator of cochlear synaptopathy stemming from noise exposure (Liberman , 2016; Bharadwaj , 2019), which is thought to impact suprathreshold auditory function, particularly speech understanding in noise, and is not reliably detected in a standard pure-tone audiogram measured using stimuli at or near threshold (Kujawa and Liberman, 2009). It is, therefore, possible that individuals with peripheral hearing loss in the EHF region could experience speech understanding difficulties resulting from reduced access to and perception of acoustic speech cues in the EHF range, subclinical deficits in lower-frequency regions, or a combination of both.

While the relationship between EHF hearing and speech understanding in noise is still under investigation (see Hunter , 2020), evidence from available research in this area raises important questions about the source of auditory difficulties commonly reported by normal-hearing Service Members and Veterans with a history of blast exposure who, in addition to head injury, also have greater risk of exposure to extreme levels of noise from weaponry and machinery (Tepe , 2020). Although previous evidence points toward central rather than peripheral auditory deficits in this population, it is unclear whether EHF hearing sensitivity is impacted in blast-exposed Veterans and whether variability in EHFs may, at least partially, explain auditory difficulties in this normal-hearing patient population. Therefore, the current retrospective study was designed to (1) further examine the effects of blast exposure on speech understanding abilities in a variety of listening conditions, (2) determine whether EHF hearing thresholds from a sample of normal-hearing blast-exposed Veterans are significantly different from those of a control group with no history of head injury, and (3) determine how variability in EHF hearing sensitivity impacts speech understanding in complex auditory environments and whether this contributes to the effects of blast exposure on speech understanding in noise.

This study represents a retrospective analysis of data collected from 41 blast-exposed Veterans [mean age, 37.7 years old; standard deviation (SD), 11.0 years of age; range, 24–65 years of age; biological sex, male = 40, female = 1] and 31 control participants (mean age, 37.6 years old; SD, 13.9 years of age; range, 22–67 years of age; biological sex, male = 21, female = 10) that were recruited from the Veterans Affairs Portland Health Care System (VAPORHCS) for data collection between 2012 and 2015. A subset of these participants and their performances on several behavioral central auditory processing tests were previously described in detail (Gallun , 2016). To summarize, participants were included in the blast-exposed group if they indicated that they had been exposed to at least one high-intensity blast (within 50 m of a large explosion) within 10 yrs of the start of the study. Enrollment of Veteran and non-Veteran participants with no history of blast exposure or neurological injury into the control group was delayed, resulting in a participant sample that was aligned in age distribution to the blast-exposed group.

Participants were included in the current dataset if they had no evidence of outer- or middle-ear pathology and normal or near-normal hearing, which was defined by a four-frequency pure-tone average (standard PTA) of 0.5, 1, 2, and 4 kHz of 35 dB hearing level (HL) or better when averaged across the right and left ears. These audiometric inclusion criteria are less strict than those reported by Gallun (2016) to better represent the wide range of normal- or near-normal-hearing patients who commonly present to audiology clinics with complaints of auditory difficulties. The Institutional Review Board at the VAPORHCS approved this project. All participants provided verbal and written informed consent before participation and were provided compensation for their time.

Behavioral pure-tone hearing thresholds were collected bilaterally at 0.25, 0.5, 1, 2, 3, 4, 6, and 8 kHz using ER-2 insert earphones (Etymotic Research Inc., Elk Grove Village, IL) and 9, 10, 11.2, and 12.5 kHz using Sennheiser HAD 200 circumaural headphones (Old Lyme, CT) connected to a Grason-Stadler (GSI-61; Eden Prairie, MN) audiometer. A modified Hughson-Westlake procedure was used to establish thresholds (Carhart and Jerger, 1959). Speech reception thresholds (SRTs) were also measured for each participant using standard clinical procedures. All of the participants were tested in a sound treated and electrically shielded booth.

Participants completed a large battery of behavioral central auditory and speech understanding measures, many of which were discussed previously by Gallun (2016). Data from the following measures that assessed speech understanding abilities in complex auditory environments have not been previously reported. These measures represent listening conditions which are thought to reflect auditory complaints commonly reported by blast-exposed Veterans, including speech understanding in background noise, spatialized listening conditions, and degraded listening conditions such as when the speech is time compressed or low-pass filtered. All behavioral speech understanding measures described below were administered through insert headphones.

• Quick speech-in-noise test (QuickSIN). The QuickSIN (Etymotic Research, Inc.; Killion , 2004) is a behavioral measure of speech understanding in noise. Participants were presented binaurally with IEEE sentences with five key words per sentence, spoken by a target speaker in the presence of four-talker babble. The presentation level of the target speaker began at 70 dB HL and was progressively decreased by 5 dB over the course of six sentences while the background speech babble was held at a constant level. This procedure resulted in six signal-to-noise ratio (SNR) conditions from 25 to 0 dB SNR. The number of key words correctly repeated was then subtracted from 25.5 dB to estimate the SNR at which approximately 50% of words were correctly repeated. Results were averaged over a total of two QuickSIN lists.

• Words-in-Noise (WIN). The WIN (Wilson, 2003) was employed as an additional measure of speech understanding in noise that used monosyllabic words presented binaurally at 80 dB HL. The level of speech babble background noise was progressively increased in 4-dB steps, resulting in seven SNR conditions ranging from 24 to 0 dB SNR. Similar to the QuickSIN, this measure estimates the SNR at which a participant can correctly repeat 50% of the target words, averaged over 2 runs of 35 words each.

• Compressed speech test (CST). The CST required participants to repeat NU-6 words that were 45% time compressed and presented in separate 50-word lists to the right and left ears at a level of 50 dB above each individual's SRT. This resulted in a 55–80 dB HL range of presentation levels across participants. The CST was scored as percent correct averaged across right and left ear presentations.

• Spatial release from masking (SRM). The SRM measure used in the current study has been previously described in detail by Gallun (2013). This task used sentences from the coordinate response measure (Bolia , 2000) that are in the form of “ready (CALL SIGN) go to (COLOR)(NUMBER) now” and were presented binaurally over headphones. Target sentences were presented at 75 dB HL. Participants were asked to select the color-number combination associated with a particular call sign located at 0° in the presence of two competing background talkers that were either colocated at 0° or spatially separated at +45° and −45°. Target-to-masker ratios (TMR) were decreased from 10 to −8 dB in 2-dB steps. The number of correct responses was subtracted from the starting TMR of 10 dB to approximate the point at which performance was 50% in the colocated and spatially separated listening conditions (Gallun , 2013). Performances between the colocated and spatially separated conditions were then subtracted to also estimate SRM threshold (in dB), which represents a listener's ability to benefit from the spatial separation of background talkers.

• SCAN-A. The SCAN-A (Keith, 1995) was designed as a screening tool for auditory processing disorders in adults. It consists of four subtests that assess speech understanding abilities in complex auditory environments: filtered words, auditory figure-ground, competing words, and competing sentences. All target speech stimuli were presented at a level of 50 dB above each individual's SRT, which resulted in a 55–80 dB HL range of presentation levels across participants. For the filtered words and auditory figure-ground subtests, participants were asked to repeat 40 monosyllabic words, where 20 were presented to each ear separately. The filtered words subtests used stimuli that were low-pass filtered at 750 Hz, whereas the auditory figure-ground subtest used words presented in multi-talker speech babble at a constant 4 dB SNR. Each test was scored as the percent of words correctly repeated across the right and left ear presentations.

  The competing words and competing sentences subtests are measures of binaural listening abilities. The competing words subtest is a measure of binaural integration in which participants were presented with 15 sets of 2 simultaneous monosyllabic words, with 1 presented to each ear. Participants were asked to repeat the words presented to each ear, starting with the right ear first for list one and the left ear first for list two. In contrast, the competing sentences subtest is a measure of binaural separation in which participants were presented with a pair of sentences simultaneously to each ear and asked to repeat the sentence heard in one ear while ignoring the other. Participants were scored based on the percent of words or sentences correctly repeated from both ears.

A standard PTA was calculated as the average threshold across ears at 0.5, 1, 2, and 4 kHz, whereas an EHF PTA was calculated as the average threshold across ears at 9, 10, 11.2, and 12.5 kHz. Linear regression models were created in R (R Core Team, 2014) using the nlme package (Pinheiro , 2016) to assess potential differences between participant group (blast-exposed vs control) in age, standard PTA, and EHF PTA, with α = 0.05.

In addition, participant age, standard PTA, EHF PTA, and participant group (blast-exposed vs control) were added as fixed effects in separate linear regression models to determine whether age, standard PTA, and EHF PTA influenced behavioral results and whether results were significantly different between blast-exposed and control participants, with α = 0.05. Given the range of ages and standard PTAs across blast-exposed and control participants, it was predicted that these factors may have significant effects on any of the behavioral speech understanding measures used in the current study. Therefore, each of the four variables (age, standard PTA, EHF PTA, and participant group) was tested sequentially such that potential effects of participant age and standard PTA were considered before determining whether EHF PTA explained additional variance in behavioral results. In this same manner, the effect of participant group was only tested after accounting for the effects of the other three variables to ensure that potential group effects were not driven by differences in participant age or hearing sensitivity.

Pure-tone hearing thresholds for the left and right ears averaged across participants within the blast-exposed group and control group are presented in Fig. 1. Analysis revealed that there was a significant difference in standard PTA between the blast-exposed (mean = 12.4 dB HL, SD = 7.1 dB HL) and control (mean = 8.6 dB HL, SD = 4.0 dB HL) participant groups [F(1,70) = 7.03, p = 0.01]. However, there was no statistically significant difference in EHF thresholds between the two participant groups[F(1,70) = 2.41, p = 0.12], although the blast-exposed participant group had a higher mean EHF PTA (21.2 dB HL, SD = 20.2 dB HL) compared to the control group (14.0 dB HL, SD = 18.5 dB HL). Standard PTA was significantly related to EHF PTA [F(1,70) = 22.94, p < 0.001] such that individuals with higher standard PTAs tended to also have higher EHF PTAs. There was no significant difference in age between the blast-exposed and control participant groups[F(1,70) = 0.00, p = 0.98]. However, participant age was significantly related to standard [F(1,70) = 10.52, p = 0.002] and EHF PTA [F(1,70) = 97.57, p < 0.001] across study participants, with older participants tending to have poorer hearing sensitivity across frequency.

Figure 2 displays the distribution of results from each behavioral measure for the blast-exposed and control participant groups as well as mean performance averaged across individuals in each participant group. Given that not all participants completed every measure, sample sizes for each participant group are also provided for each measure in Fig. 2. Table I provides results from the multiple linear regression models for each speech understanding measure, including the colocated and spatially separated conditions of the SRM task. Residuals for each linear regression model described below did not show signs of heteroscedastic variance and followed assumptions for normality except for several cases that contained extreme values, which were defined as studentized residuals with an absolute value greater than or equal to 3. Supplementary Table II is provided to show the results of these linear regression models with these data points removed for comparison to results from the full data set described below.¹

As depicted in Table I, analysis of the full data set revealed that participant age had a significant effect on performance on the CST and colocated and spatially separated conditions of the SRM task such that older participants tended to have poorer performance on these tasks. There was not a statistically significant effect of age on speech understanding abilities as measured by the QuickSIN, WIN, SRM, or SCAN-A subtests. Standard PTA had a significant effect on the QuickSIN, WIN, CST, and the filtered words, auditory figure-ground, and competing words subtests of the SCAN-A. In general, participants with poorer hearing thresholds captured by the standard PTA tended to have poorer speech understanding abilities as measured by these tests. There was no statistically significant effect of standard PTA on the competing sentences subtest of the SCAN-A or SRM measures. Analysis also revealed that EHF PTA did not have a significant effect on any of the behavioral speech understanding measures used in the current study.

Participants in the blast-exposed participant group performed more poorly, on average, across each speech understanding measure compared to participants in the control group (see Fig. 2). After accounting for potential effects of participant age and hearing sensitivity, the results of the multiple linear regression models provided in Table I revealed an effect of participant group (i.e., whether or not a participant had a history of blast exposure) on the CST, the spatially separated listening condition of the SRM task, and the filtered words, auditory figure-ground, and competing words subtests of the SCAN-A. There were no statistically significant effects of participant group on the QuickSIN, WIN, or competing sentences subtest of the SCAN-A.

This study was designed to examine performance on several measures of speech understanding to identify those that are sensitive to the effects of blast exposure. After accounting for effects of participant age and hearing sensitivity, the normal- and near-normal-hearing Veterans with blast exposure had poorer performance on several measures that represented a range of listening conditions commonly reported as troublesome for this patient population. These included measures that assess understanding of rapid speech, degraded speech, and speech in noise, as well as speech understanding in conditions that require accurate binaural processing. Evidence from previous studies suggests that these types of auditory difficulties are likely the result of deficits in central auditory processing rather than damage to peripheral auditory mechanisms. However, recent work has shown that variability in EHF hearing sensitivity may contribute to difficulties understanding speech in noise even in individuals with normal pure-tone hearing thresholds in the standard audiometric range (Badri , 2011; Best , 2005; Flaherty , 2021; Levy , 2015; Mishra , 2022; Monson , 2019; Motlagh Zadeh , 2019; Polspoel , 2022; Trine and Monson, 2020; Yeend , 2019). Given that blast-exposed Veterans are at higher risk of exposure to hazardous levels of noise that may impact EHF hearing regions (Tepe , 2020), this work also aimed to determine whether these participants had poorer EHF thresholds compared to control participants with no history of head injury and whether variability in EHF sensitivity may be driving differences in speech understanding abilities between participant groups.

Previous work suggests that speech understanding deficits may stem from (1) the reduced audibility of important EHF speech cues due to EHF hearing loss (Best , 2005; Flaherty , 2021; Monson , 2019; Monson and Caravello, 2019; Motlagh Zadeh , 2019), and/or (2) subclinical damage to lower-frequency regions that is signaled by a loss of hearing sensitivity in the EHF range (Badri , 2011; Bharadwaj , 2019; Liberman , 2016; Mishra , 2022). Results from the current study suggest that neither appears to the be case for the sample of participants tested here. Although the difference was not statistically significant, average EHF hearing thresholds (PTAs of 9, 10, 11.2, and 12.5 kHz) were poorer for the blast-exposed participant group compared to the control group. However, results from the present study showed that this variability in EHF thresholds did not explain performance on any of the behavioral speech understanding measures after accounting for effects of participant age and standard PTA. Although this finding is inconsistent with those from the studies mentioned above, it aligns well with others that have revealed no relationships between EHF hearing sensitivity and speech understanding (Smith , 2019; Stiepan , 2020).

However, interpretation of these null results is complicated by the stimulus materials and equipment calibration used in the current study. Previous work revealing significant relationships between access to EFH speech cues and speech understanding in noise used stimuli with demonstrated energy in the EHF range (e.g., Monson , 2019; Motlagh Zadeh , 2019). Results have been inconsistent in studies that have used clinical measures similar to those included in this study, such as the QuickSIN (Smith , 2019; Stiepan , 2020), where the bandwidth of stimulus recordings may be limited. Recent work by Monson and Buss (2022) highlighted this issue by comparing the long-term average speech spectra (LTASS) of several common speech corpora to examine energy in the EHF range. Indeed, it was exhibited that recordings of speech stimuli from the QuickSIN have a bandwidth that begins to roll off beyond 8 kHz, limiting access to potentially useful EHF cues (Monson and Buss, 2022). While the filtered words subtest from the SCAN-A is specifically designed to be low-pass filtered at 750 Hz, the bandwidths of other recordings used in the present study are unknown. Therefore, it is possible that current results were influenced by the stimuli used in the chosen speech understanding measures.

To explore this possibility, LTASS were plotted for a sample of speech recordings used in this study, including sentences from the coordinate response measure used in the SRM task and competing sentences subtest of the SCAN-A. Figure 3 shows that the version of the coordinate response measure sentences had a steep cutoff at 8 kHz, while the speech samples from the SCAN-A had a slightly wider bandwidth that continued to roll off through 10 kHz. This analysis is consistent with findings from Monson and Buss (2022) and raises the possibility that a lack of significant relationship between variability in EHF thresholds and speech understanding abilities may be due to the absence of audible EHF information contained in speech recordings. It is also important to note that even if the stimuli used in the current study did have an extended bandwidth, the insert headphones used for data collection were not specifically calibrated to ensure accurate presentation of EHF information to participants (Bharadwaj , 2019; Hunter , 2020; Lough and Plack, 2022). Furthermore, estimating access to potential EHF cues available from extended bandwidth stimuli would be further complicated because of the limited range of hearing thresholds collected here, which did not extend beyond 12.5 kHz.

The issue described above limits our ability to interpret how acoustic cues in EHF regions may or may not have influenced speech understanding abilities in the current study. Additional limitations related to the sample of participants tested and types of audiometric measures used here also exist, which make it difficult to interpret findings in the context of studies that suggest that reduced EHF hearing sensitivity may signal subclinical deficits in lower-frequency regions. Although participants in this study were restricted to those with standard PTAs less than or equal to 35 dB HL, representing clinically “normal” to “mild” sensorineural hearing loss, analysis revealed that standard PTAs were significantly different between participant groups and this variability in standard PTA influenced most of the speech understanding measures used in the present study. This significant difference in standard audiometric hearing thresholds between participant groups may suggest that any possible damage present in the standard audiometric range was no longer “subclinical” in nature at the time of testing.

In addition, current results also revealed a strong relationship between standard and EHF PTAs across participants. While participants in the current study had a wider range of ages and pure-tone hearing thresholds in standard audiometric frequencies compared to previous work, significant relationships between standard and EHF thresholds have also been reported by previous studies that enforced a stricter audiometric criterion, typically by excluding participants with thresholds >20 dB HL (Mishra , 2022; Motlagh Zadeh , 2019). Taken together, these findings suggest that variability in hearing sensitivity that is traditionally considered to be in a clinically normal or “near-normal” range for standard audiometric frequencies can still signal auditory deficits that may impact speech understanding abilities in complex environments, and this variability may be, at least partially, driving the effects of EHF thresholds observed in previous work. Given the significant difference in standard PTAs between participant groups in this study, this possibility should be considered in future studies that examine auditory difficulties in normal-hearing blast-exposed Veterans. In addition, future research in this area should consider including measures that may be more sensitive to damage in lower-frequency regions of the auditory periphery than pure-tone audiometry, such as otoacoustic emissions (Mishra , 2022; Stiepan , 2020) or electrophysiological measures, which are commonly used to investigate the presence of noise-induced cochlear synaptopathy in humans (Bramhall , 2019; Bramhall , 2021; Liberman , 2016).

Results from the current study showed a lack of statistically significant differences in EHF hearing sensitivity between blast-exposed Veterans and control participants. This work also suggests that variability in EHF hearing sensitivity is not driving poorer speech understanding abilities in normal-hearing blast-exposed Veterans. However, several important study limitations, including limited bandwidth of the speech stimulus recordings, complicate interpretation of this finding and should be considered in the design of future studies in this area. Although differences in hearing sensitivity in standard audiometric regions did have a significant effect on many of the speech understanding measures used in this study, group-related differences in performance remained significant after controlling for the effects of hearing sensitivity. These findings support previous work which suggests that auditory difficulties in blast-exposed Veterans, at least partially, stem from factors unrelated to peripheral hearing sensitivity, such as central auditory or cognitive processing (Gallun , 2012; Gallun , 2016; Hoover , 2017; Kubli , 2018; Saunders , 2015; Turgeon , 2011). These findings also support the use of speech understanding measures specifically designed to identify central auditory processing deficits, such as the SCAN-A test battery, as they are reflective of the types of difficulties commonly reported by blast-exposed Veterans and appear to be particularly sensitive to auditory difficulties in this patient population.

This material is the result of work supported with resources and the use of facilities at the Veterans Affairs Rehabilitation Research and Development Service (VA RR & D) National Center for Rehabilitative Auditory Research (NCRAR; Center Award No. C2361C/I50 RX002361) at the VA Portland Health Care System in Portland, Oregon. The views expressed are those of the authors and do not represent the views of the U.S. Department of Veterans Affairs or the United States Government. The authors would like to thank Dr. Marjorie Leek and Dr. M. Samantha Lewis for their original support of this project through the following awards from the United States Department of VA RR & D: Merit Review Award No. C7755I [principal investigators (PIs): F.J.G. and M. R. Leek], Senior Research Career Scientist award to M. R. Leek (Award No. C4042L), and a Career Development II award to M. S. Lewis (Award No. C7067W). The current work was also supported by a Career Development I award to T.K.K. (Award No. RX003187). The authors would also like to thank Dr. Melissa A. Papesh, Heather Belding, and Michele Hutter for their contributions to this project.