Acceptable Noise Level, Phoneme Recognition in Noise, and Measures of Auditory Efferent Activity

Transcription

1 J Am Acad Audiol 16: (2005) Acceptable Noise Level, Phoneme Recognition in Noise, and Measures of Auditory Efferent Activity Ashley W. Harkrider* Steven Brad Smith Abstract Acceptable noise level (ANL) is unrelated to sentence recognition in noise but may be related to phoneme recognition in noise (PRN). Individual differences in efferent activity in medial olivocochlear bundle (MOCB) and acoustic reflex (AR) pathways may account for intersubject variability in ANL and PRN. Monotic and dichotic ANL, monotic PRN, contralateral suppression of transient evoked otoacoustic emissions, and ipsilateral and contralateral acoustic reflex thresholds were measured in 31 adults with normal hearing. Results indicate that monotic ANL and PRN are unrelated. Monotic and dichotic ANL are related, suggesting that nonperipheral factors mediate ANL. Intersubject variability in ANL cannot be accounted for by individual differences in MOCB or AR activation. Intersubject variability in PRN cannot be accounted for by individual differences in MOCB or contralateral AR activation. It may be influenced by the ipsilateral AR pathway. Efferent activity in the contralateral AR arc is correlated with efferent activity in the MOCB. Key Words: Acceptable noise levels, acoustic reflex thresholds, contralateral suppression of transient evoked otoacoustic emissions, phoneme recognition in noise Abbreviations: ANL d = acceptable noise level measured dichotically; ANL m = acceptable noise level measured monotically; AR = acoustic reflex; ART = acoustic reflex threshold; BBN = broadband noise; BNL = background noise level; CSTEOAE = contralateral suppression of transient evoked otoacoustic emission; MCL = most comfortable level; MOCB = medial olivocochlear bundle; PRN = phoneme recognition in noise; SNR = signal-to-noise ratio; SPIN = Speech-in-Noise test; TEOAE = transient evoked otoacoustic emission Sumario El nivel aceptable de ruido (ANL) no se relaciona con el reconocimiento de frases en ruido pero puede estar relacionado con el reconocimiento de fonemas en ruido (PRN). Las diferencias individuales en la actividad eferente del fascículo olivo-coclear medial (MOCB) y de las vías del reflejo acústico (AR) pueden ser responsables de variaciones intersujeto en ANL y PRN. Se midieron ANL monóticos y dicóticos, PRN monóticos, supresión contralateral de las emisiones otoacústicas evocadas por transitorios, y reflejos acústicos ipsi y contra-laterales en 31 sujetos con audición normal. Los resultados indican que el ANL y el PRN monóticos no están relacionados. El ANL monótico y dicótico sí están relacionados, lo que sugiere que factores no periféricos median en el ANL. La variabilidad intersujeto en ANL no puede ser explicada por diferencias individuales en el MOCB o en la activación del AR. La variabilidad inter-sujeto en PRN no puede ser explicada por diferencias individuales en el MOCB o en la activación contralateral del AR. Sí podría estar influida por la vía ipsilateral del AR. La actividad eferente en el AR contralateral está correlacionada con la actividad eferente en el MOCB. *Department of Audiology and Speech Pathology, University of Tennessee, Knoxville, TN; Albuquerque Speech, Language, and Hearing Center, Albuquerque, NM Ashley W. Harkrider, University of Tennessee, Department of Audiology and Speech Pathology, 457 South Stadium Hall, Knoxville, TN 37996; Phone: ; Fax: ; aharkrid@utk.edu 530

2 ANL, PRN, and Efferent Activity/Harkrider and Smith Palabras Clave: Niveles aceptables de ruido, umbrales de reflejo acústico, supresión contralateral de la emisiones otocústicas evocada por transitorios, reconocimiento de fonemas en ruido Abreviaturas: ANL d = nivel aceptable de ruido medido dicóticamente; ANL m = nivel aceptable de ruido medido monóticamente; AR = reflejo acústico; ART = umbral del reflejo acústico; BBN = ruido de banda ancha; BNL = nivel de ruido de fondo; CSTOAE = supresión contralateral de emisión otoacústica evocada por transitorios; MCL = nivel de comodidad; MOCB = fascículo olivococlear medial; PRN = reconocimiento de fonemas en ruido; SNR = tasa de señal/ruido; SPIN = Prueba de Lenguaje en Ruido; TEOAE = emisión otoacústica evocada por transitorios Acceptable noise level (ANL) is one type of speech-in-noise measure that is ascertained by having listeners selfselect an acceptable level of background noise (e.g., Nabelek et al, 1991; Crowley and Nabelek, 1996; Rogers et al, 2003; Nabelek, Tampas, et al, 2004). ANL characterizes the maximum level of background noise an individual is willing to accept while listening to running speech without becoming tense or tired. It is calculated by finding the difference between the most comfortable level (MCL) for the speech stimulus and the maximum level of the noise accepted (BNL), or ANL = MCL - BNL. This approach is not concerned with quantifying speech perception but more with describing an individual s willingness to listen to speech at a self-chosen maximum background noise level without fatiguing. Determining this self-selected acceptance for background noise provides predictive information that is related to successful hearing aid use (Nabelek et al, 1991; Crowley and Nabelek, 1996). Using a predictive model based on the data from 191 listeners with hearing impairment, Nabelek, Burchfield, et al (2004b) determined that people with low ANLs (<8 db), who accept a lot of background noise, will be successful users of hearing aids. People with mid to high ANLs ( 8 db), who do not accept as much background noise, are less likely to be successful users. Thus, as ANL increases, chances of success with hearing aids decrease. It follows that finding a way to reduce listeners ANLs would increase their chances of benefiting from aural habilitation. In order to determine effective strategies for decreasing ANLs, it is important to understand the psychological and physiological processes that mediate ANL. Interestingly, ANL is not related to performance on a different type of speech-innoise measure, the Speech-in-Noise test (SPIN), in listeners with hearing impairment (Crowley and Nabelek, 1996; Nabelek et al, 2004a). The SPIN is designed to assess word recognition in noise by presenting sentences with and without linguistic, contextual cues (Kalikow et al, 1977). Because of the lack of correlation between ANL and SPIN, it has been concluded that the perceptual tasks required by ANL measurement are not directly analogous to those required by the SPIN test. Specifically, the SPIN test measures comprehension of sentences in noise (a complex linguistic process), while ANL measures willingness to tolerate background noise while listening to the words of a story (no linguistic process) (Nabelek et al, 2004a). Consequently, ANL and SPIN tests each provide unique information about a patient s abilities. The relation between ANL and other speech-in-noise tasks has not been investigated. In the current study, a link between ANLs and the percentage of words or phonemes correctly perceived when presented at a particular signal-to-noise ratio (SNR) (e.g., Nilsson et al, 1991, 1992, 1993; Studebaker et al, 1994; Beattie et al, 1997, Wilson and Strouse, 2002) was sought. A large correlation between ANL and phoneme recognition in noise (PRN) was not expected due to the lack of correlation of ANL with SPIN scores and the fact that the perceptual requirements of ANL and PRN differ. However, the perceptual tasks required during PRN are not as linguistically demanding (versus the SPIN) and so, performance on PRN and ANL tests may be more similar than performance on SPIN and ANL measures. Additionally, both ANL (e.g., Franklin et al, 2001; Rogers et al, 2003; Nabelek et al, 2004a) and PRN (e.g., Cooper and Cutts, 1971; Suter, 1985) measures have high intersubject variability, which exists even in young, normal-hearing listeners. 531

3 Journal of the American Academy of Audiology/Volume 16, Number 8, 2005 Thus, even though the perceptual demands of ANL and PRN differ, there may be common psychological or physiological variables that influence overall performance on these tasks within a listener (e.g., low ANL corresponds to better performance on PRN), as well as result in common findings regarding intersubject variability within a population. If so, one could choose to use ANL or PRN clinically to provide similar information about an individual patient (e.g., predicted success with hearing aids) or population. However, if the behavioral measures are unrelated, it would suggest that unique information would be provided by each and both should be implemented to characterize a patient s overall speech-in-noise performance. Further, a lack of correlation between ANL and PRN would suggest that the intersubject variability in ANL was not caused by the same combination of psychological or physiological factors contributing to the intersubject variability of PRN. Because intersubject variability persists in ANL and PRN measures among audiometrically matched subjects, it can be inferred that these tasks of speech-in-noise performance involve variables that have little relationship with audiometric results. Several other possible factors contributing to the intersubject variability in ANL have been evaluated, including age (Nabelek et al, 1991), noise type (Nabelek et al, 1991; Crowley and Nabelek, 1996), uncomfortable loudness level (Franklin et al, 2001), and listener s gender (Rogers et al, 2003). However, none of these variables were shown to be significantly related to ANL and, therefore, are not likely to be contributing to the individual differences in ANLs. Generally, it is believed that the factors listed above are most likely to impact responses from the peripheral auditory system, causing one to wonder if more central portions of the auditory system are responsible for individual performance and intersubject variability on these measures. In an attempt to better define the levels of the auditory system and physiological processes that contribute to ANL and PRN, the study, first, compares traditional PRN and ANL measures (monotic) with dichotic ANL measures (speech in one ear, noise in the opposite ear) from the same group of subjects. Presumably, the dichotic ANL (ANL d ) would reflect processing from the superior olivary complex (the first place in the auditory system where information from the two ears is processed simultaneously) and higher. Thus, if monotically measured PRN or ANL (ANL m ) and dichotic ANL (ANL d ) are related, it would follow that the subject s PRN and ANL m was mediated in the central auditory nervous system, not the periphery. Second, the possibility is investigated that individual differences in the level of efferent activity in the lower brainstem are contributing to the wide range of ANL and PRN scores within a given population. Specifically, activity levels in two efferent systems are evaluated, the acoustic reflex (AR) and the medial olivocochlear bundle (MOCB) pathways. The AR consists of ipsilateral and contralateral lower brainstem pathways whose activation causes a contraction of the middle ear muscles, typically occurring in response to a moderateto-loud ipsilateral, contralateral, or bilateral sound, or before and during vocalization (Borg and Zakrisson, 1975a). The overall effect is a stiffening of the ossicular chain and an outward rotation of the stapes footplate in the round window resulting in an increase in impedance and a corresponding decrease in transmission of low-frequency sounds (e.g., for review, see Møller 1984; Mahoney et al, 1979; Pang and Peake, 1986). Activation of the efferent MOCB pathway, originating in the superior olivary complex in the lower brainstem, has an inhibitory effect on primarily contralateral outer hair cell responses, and the subsequent responses of primary afferent neurons (Guinan et al, 1983; for review, see Sahley, 1997). The MOCB inhibits these auditory responses by limiting the amount of gain that is provided by the cochlear amplifier (Davis, 1983; for review, see Patuzzi, 1996). Clinically, the level of MOCB activity can be noninvasively evaluated in humans by suppressing the amplitude of otoacoustic emissions (OAEs) with the introduction of an ipsilateral, contralateral, or bilateral stimulus (for review, see Robinette and Glattke, 2002). Both the AR and MOCB efferent systems have been shown to contribute to speech reception and signal detection in noise in both physiological and behavioral studies. For example, it has been suggested that one function of the AR is to improve speech discrimination by decreasing the masking of high-frequency sounds by low-frequency maskers (Simmons, 1964; Borg and Zakrisson, 1973, 1975b; Mahoney et al, 1979; Dorman et al, 1987; Pang and Guinan, 1997). 532

4 ANL, PRN, and Efferent Activity/Harkrider and Smith Borg and Zakrisson (1975b) reviewed the methods and results from three of their studies (1973, 1974, 1975a) that investigated the effects of stapedius muscle contraction on speech perception and detection of masked pure tones in 35 humans subjects. Using immittance measurements and electromyography, they showed that the stapedius muscle is activated during the vocalization process and in response to moderate- to high-intensity maskers. Using behavioral measures, it was established that the activation of the stapedius muscle improved the recognition of intense speech, and monosyllabic recognition and pure-tone detection masked by low-frequency noise. Similar to the findings of Borg and Zakrisson (1973, 1975b), Mahoney et al (1979) reported a positive correlation between AR activation and speech discrimination. High-level speech was transferred through the cat middle ear to activate the AR and was recorded as cochlear potentials that were then presented as speech stimuli to human listeners. Results demonstrated that the active AR enhanced speech discrimination in quiet and in noise in humans, although this conclusion was limited to high-level speech sounds. In accord, Wormald et al (1995) documented a decrease in speech recognition at high levels ( db SPL) in 80 audiometrically normal patients with paralysis of the stapedius reflex due to Bell s palsy. Seventy percent of the subjects with an absent stapedius reflex showed word recognition scores with significant (49%) rollover in performance-intensity functions, with mean scores decreasing from 98 to 49%. In the same subjects, no significant rollover in word recognition was evident following recovery of the facial nerve palsy and the subsequently normal stapedius reflex. These authors suggested that 65% of the decline in speech recognition scores at high intensities in the patients with Bell s palsy was due to the absence of an AR since the normal-hearing listeners, under a simulated condition of an absent AR, showed the same amount of decline as those with Bell s palsy. However, in contrast to the above-mentioned studies, Cox and Greenberg (1977) reported a decrement in speech discrimination in noise with simultaneous activation of the AR by a contralateral 2000 Hz tone in normal listeners. Danaher and Pickett (1972) also reported that the AR deteriorated speech discrimination in subjects with moderate to severe hearing loss. Additionally, Waas (1976) found that, although listeners with normal hearing with unilateral stapedial paralysis had better discrimination scores to high-level speech introduced to the contralateral, intact ear, the difference in performance between the two ears was not significant. Thus, the importance of AR activation in speech perception remains a topic for debate. With regard to the MOCB pathway, it is hypothesized that one function is to improve the detection of signals in background noise (Guinan, 1996; for review, see Sahley, 1997). Giraud et al (1997) assessed the strength of MOCB suppression via contralateral suppression of OAEs and correlated these values with perceptual performance in recognizing monosyllabic words in broadband noise (BBN) with and without the addition of a contralateral suppressor. Measures were obtained for a group of 20 subjects with normal auditory systems and a group of five patients with vestibular neurotomies. Results from this study showed that the presence of a contralateral stimulus significantly reduced the rate of decline in speech-in-noise performance with increasing levels of ipsilateral noise in subjects with normal hearing. In patients with vestibular neurotomies, the effect was observed on the healthy side, but not the de-efferented side. Measurements of speech perception on the surgically altered side showed significantly more rapid degradation in performance with increased ipsilateral masker levels. Additionally, for audiometrically normal ears, improvements in phoneme recognition resulting from a contralateral stimulus were significantly correlated with more robust contralateral suppression of OAEs. In summary, there is a growing body of evidence to support the hypothesis that one function of the AR and MOCB pathways is to improve the detection of signals in background noise. It is possible that reduction in the effects of background noise may be proportional to the amount of AR and/or MOCB activation. If so, behavioral measures of speech-in-noise performance like ANL m, ANL d, and PRN should be correlated with acoustic reflex thresholds (ARTs) or contralateral suppression of transient evoked otoacoustic emissions (CSTEOAEs). Moreover, if objective measures correlate with ANL in particular, they could be implemented into clinical routines when attempting to determine hearing aid success. 533

5 Journal of the American Academy of Audiology/Volume 16, Number 8, 2005 This study was designed to investigate the possibility that (1) individual performance on subjective measures of speech-in-noise performance involving different perceptual requirements are related and (2) individual differences in levels of efferent activity in the AR arc and/or MOCB pathways account for well-documented but unexplained intersubject variability in these distinctly different perceptual tasks involving auditory performance in noise. Specifically, the aims were (1) to investigate relations among the three speech-in-noise performance measures (monotic ANL, dichotic ANL, monotic PRN); (2) to determine if three different measures of speech-in-noise performance are related to two measures of efferent, suppressive feedback to the cochlea (ART, CSTEOAE); and (3) to investigate relations between the two measures of efferent auditory activity. Comparing the behavioral tasks allows one to determine if the perceptual requirements associated with monotic ANL are the same as those required by dichotic ANL and/or monotic PRN. Additionally, knowledge will be gained about the levels of the auditory system and physiological factors that mediate the responses of one or all of the behavioral measures. Comparisons of the two measures of efferent auditory activity will allow us to determine if individuals with high levels of activity in the AR arc also have high levels of activity in the MOCB pathway. Subjects METHODS Thirty-one individuals between the ages of 19 and 40 years had pure-tone thresholds better than 20 db HL at octave frequencies between 250 and 8000 Hz and at 6000 Hz bilaterally with the exception of one subject who had one threshold of 25 db. (Note: This subject was included in the study because the threshold was believed to be due to excessive noise during the audiometric screening.) All subjects had unremarkable auditory histories, otoscopic, and acoustic immittance results (tympanometry and ipsilateral/contralateral AR). Experimental Design Four measures were selected to provide data relating to the perception of speech in noise and the two primary efferent feedback systems. All measurements were obtained in a sound-treated booth with permissible ambient noise levels (re: ANSI, 1996). Measures were made for the subjects right ears; however, many of the measures involved the simultaneous acoustic stimulation of the left ear. The effects of noise on a listener were evaluated for each subject using three different behavioral measures: (1) the measure of the maximum level of noise acceptable to a subject while listening to running speech was calculated for a monotic condition (ANL m ) by subtracting each subject s ipsilaterally accepted background noise (multitalker babble) from their MCL to running speech; (2) a dichotic ANL (ANL d ) procedure was conducted, which was modified from the previously mentioned procedure by simultaneously presenting the running speech and the competing noise (multitalker babble) to opposite ears; and (3) speech perception in noise was assessed with a phoneme recognition task, which involved the presentation of 50 monosyllabic words (NU 6) embedded in the same multitalker babble stimulus used in the two ANL measures. Ipsilateral and contralateral ARTs were obtained in response to BBN. Each subject s MOCB efferent activity was indirectly measured by quantifying the suppression of TEOAEs resulting from the introduction of a contralateral BBN. The experimental session lasted approximately 1.25 hours. The University of Tennessee Institutional Review Board approved the protocols. Apparatus and Test Materials Monotic ANL Measures for calculating ANL were obtained for each subject based on the method described by Nabelek et al (1991). The test materials used in this procedure consisted of a Cosmos recording of running male speech (Cosmos Distr. Co.) and a competing stimulus of multitalker babble from the revised SPIN test (Bilger et al, 1984). The stimuli were presented from a compact disc to a Madsen OB822 audiometer and delivered to the right ear of the subject via an insert earphone (Etymotic, ER3A). Each subject was 534

6 ANL, PRN, and Efferent Activity/Harkrider and Smith instructed to signal the examiner, using hand gestures (thumb up = increase level; thumb down = decrease level), to adjust the volume of the speech or speech babble either up or down. The subject provided the examiner with verbal confirmation following appropriate adjustment. Dichotic ANL Using the same test materials and equipment described above, a contralateral ANL measure was obtained. This measure was termed the dichotic ANL (ANL d ) and was determined with the speech signal presented to the subject s right ear and the speech babble noise presented to the subject s left ear. Phoneme Recognition in Noise The method for measuring PRN was adapted from Wilson and Strouse (2002). Fifty phonetically balanced, monosyllabic words from lists 1A or 2A of the NU 6 were presented in the presence of an ipsilaterally competing stimulus. The NU 6 words were presented at 55 db HL using an Auditec recording of male speech. The competing stimulus was the same multitalker babble recording that was used in the ANL procedures. The level of the ipsilateral multitalker babble was 55 db HL (0 db SNR) during the presentation of the 50 NU 6 words. Both stimuli were delivered from a compact disc to a Madsen OB822 audiometer and delivered to the right ear via insert earphones (ER-3A). After each word was presented, subjects were asked to verbally repeat and write the response. The examiner also manually recorded each response during the test. Acoustic Reflex Ipsilateral and contralateral ARTs were obtained bilaterally with a Grason-Stadler GSI 33 Middle Ear Analyzer. A 226 Hz probe tone was used with a starting pressure of +200 dapa. The reflex-eliciting test stimulus was BBN for ipsilateral and contralateral recordings. The BBN was briefly presented at intensities ranging from 50 to 110 db HL to determine the ART. All signals were presented to the subjects via appropriately fitting Grason Associates, Inc. Single Use Eartips. The size and insertion depth of the ear-tip was standardized for all subjects based on anatomical landmarks. Recording of Otoacoustic Emissions TEOAE recordings were obtained using the Otodynamics Ltd. ILO88/92 Otoacoustic Emission System (Kemp et al, 1990). All stimuli presentation and data collection were accomplished with this system using a standard acoustic probe containing a receiver and a microphone, which was inserted into the subject s right ear canal with Otodynamics Ltd. foam probe tips. TEOAEs were obtained for the right ear, using click stimuli (80 msec rectangular electrical pulses) presented linearly at a rate of 50/sec. The click level was adjusted to 60 db peak SPL (±3 db) for all but three subjects. For these three subjects, click levels of db SPL were used due to reduced OAE amplitudes. The linear click mode, as defined by the ILO-88 system, was employed in an effort to maximize the OAE response obtained at the low-click presentation levels. TEOAE responses were analyzed in a 20 msec epoch following the onset of stimulation. The residual response contained the energy of the TEOAE between 0 and approximately 5000 Hz. Responses were summed and stored alternatively in one of two buffers (A or B). Recording was complete when the responses to 260 of the stimuli groups (4 clicks) were summed in each buffer. Incoming noise was updated dynamically during the recording. If the noise in the external auditory canal exceeded the pre-established rejection level (45 52 db pspl), the recording paused and resumed once the noise level dropped back below the rejection level. If the accepted number of sweeps was below 85%, the TEOAE waveform was rerecorded. TEOAE Suppression OAE data were collected as detailed above for two conditions with and without contralateral BBN. For the noise condition, constant BBN was delivered to the contralateral ear at 5 db above the click level as recommended by Berlin (1999) to optimize TEOAE suppression. Thus, for most subjects, the click stimulus level was 60 db SPL and the contralateral BBN level was 65 db SPL. 535

7 Journal of the American Academy of Audiology/Volume 16, Number 8, 2005 For the three subjects who required the use of a higher click stimulus level to obtain a reliable OAE response, the BBN was delivered at db SPL to ensure that the BBN contralateral suppressor level was 5 db higher than the click level. The BBN was generated by a Madsen Audiometer (OB822) and delivered through an ER3A insert earphone. Test Procedures Monotic ANL Prior to obtaining measurements for calculating the ANL m, each subject received verbal and written instructions explaining the experiment and his/her task. The starting level for determining each subject s MCL for speech was 10 db HL. The level was increased in 5 db steps as the subject signaled for an increase in level. Once the level of the speech surpassed the subject s MCL, the subject signaled to decrease the level. At this point in the procedure, increments of 2 db steps were used until the MCL was established. You will be listening to a story in your right ear. After you listen for a few moments, you will be asked to adjust the loudness that you like. You will signal with your thumb pointing either up (louder) or down (softer) to allow you to adjust the story louder and softer in small steps. Please signal the volume to be turned up to a level that is too loud and down to a level that is too soft, and then select your comfortable listening level. After establishing the MCL, the background noise was added to the same ear. The subject was given instruction for adjusting the level of the background noise while listening to the ongoing speech at MCL. The background noise started at 10 db HL and was increased in 5 db steps until the subject signaled for the level to be reduced. Then the level of the background noise was adjusted in 2 db steps as signaled by the subject until the maximum amount of accepted background noise was established. I will now add some background noise and ask you to signal with your thumb either pointed up (louder) or down (softer) to adjust the loudness of the background noise to a level which you would be willing to accept or "put up with" without becoming tense and tired while listening to and following the words of the story. First, turn the noise up until it is too loud and then down until the story becomes very clear. Finally, adjust the noise (up and down) to the level that you would "put up with" for a long time while following the story. The ANL m was calculated by finding the difference between the MCL for the speech stimulus and the maximum level of the noise accepted (BNL), or ANL = MCL - BNL. Dichotic ANL The ANL d procedure differed from the ANL m procedure in that the speech and the background noise were delivered to opposite ears, right and left ears, respectively. The subjects were informed that the noise and the speech would be in opposite ears for this measurement. The ANL d is otherwise procedurally identical to the ANL m. Measurements and calculations of the ANL d were conducted for all subjects in the manner detailed above. Phoneme Recognition in Noise To assess PRN, subjects were instructed to verbally repeat their identification of each monosyllabic word to the best of their ability. Subjects were encouraged to respond by verbally repeating any sound they may have heard, even if they were not fully confident about their correct identification of the sound. Competing multitalker babble was simultaneously presented to the right ear at 0 db SNR for 50 NU 6 test items. Verbal responses from each subject were manually recorded during this procedure. The percentage of phonemes correctly perceived was calculated for each subject by dividing the total number of phonemes in the word list by the total number of phonemes correctly identified. Acoustic Reflex Thresholds ARTs were obtained ipsilaterally and contralaterally for each subject using BBN. The ART levels were determined using a 536

8 ANL, PRN, and Efferent Activity/Harkrider and Smith simple up and down procedure. The lowest stimulus level that repeatedly elicited an AR (0.02 ml) was documented as the ART. TEOAEs and Suppression of TEOAEs Methods for quantifying contralateral suppressive effects on TEOAEs were based on those proposed by Brashears and Hood (2003). Three response waveforms were collected for each condition (with and without a masker). The order of presentation for the two TEOAE conditions was interleaved across the six test runs to control for any order effect. Then, the three recordings from each condition were compared for similarities in terms of SNR, click level, click stability, and waveform repeatability. The two most similar waveforms (one from each condition) were selected for the final analysis. Contralateral suppressive effects were evaluated by comparing the TEOAE waveform obtained without contralateral noise to the TEOAE waveform obtained with contralateral noise. (Note: Including all waveforms by averaging the three TEOAE recordings without noise and comparing them to the average of the three TEOAE recordings with noise did not alter the findings.) Analysis was performed using the Echomaster software program developed by Wen et al (1993). A single number value representing an overall suppressive effect was derived from the responses that occurred between 8 and 18 msec. The number represented the TEOAE response mean without the suppressor minus the TEOAE response mean with the suppressor for that time interval. Statistical Analysis Pearson product-moment correlations were obtained among a total of eight variables. Three of these were speech in noise variables (ANL m, ANL d, and PRN). The remaining variables were measures of efferent activity. These included four ARTs (ipsilateral right, ipsilateral left, contralateral right, contralateral left) to BBN and one measure of contralateral suppression of the TEOAE measured in the right ear for the 8 18 msec time increment. RESULTS Data from 31 adults with normal hearing were collected for three measures of speech-in-noise performance (ANL m, ANL d, and PRN) and one measure of efferent activity (ARTs). For the second measure of efferent activity, CSTEOAE, data are available from 30 subjects due to a technical problem with the measurement of one subject. The number of subjects, means, ranges, and standard deviations for these measures are summarized in Table 1. Correlation coefficients were computed for all pairings of the variables listed in Table 1 using Pearson product-moment correlations. In order to minimize the chances of making a Type I error, correlations with a p value of less than 0.05 were required also to have an effect size of 0.35 or greater to be considered significant. Typically, for the behavioral sciences, correlation coefficients of 0.1, 0.3, and 0.5 are interpreted as small, medium, and large, respectively (Green et al, 2002). Significant results of the correlation analyses are presented in Table 2 and show that two of the correlations were statistically significant at p < 0.01 (2-tailed), three at p < 0.05 (2-tailed), and all five of the correlation coefficients were greater than or equal to 0.35, indicating a medium or large effect size. Although not relevant to the goals of the study, as expected, AR measures (ipsilateral/contralateral, RE/LE) were significantly correlated with each other. Any Table 1. Descriptive Statistics Measure N Min Max Mean SD RE Ipsi AR threshold to BBN (db HL) LE Ipsi AR threshold to BBN (db HL) RE Contra AR threshold to BBN (db HL) LE Contra AR threshold to BBN (db HL) CSTEOAE (db SPL) ANL m (db) ANL d (db) PRN (% correct)

9 Journal of the American Academy of Audiology/Volume 16, Number 8, 2005 Table 2. Pearson Product-Moment Correlations Correlated Measures Significance Correlation Effect Level (p) Coefficient (r) Size ANL m and ANL d < Large LE Contra AR threshold (BBN) and CSTEOAE < Medium RE Contra AR threshold (BBN) and CSTEOAE < Medium ANL d and PRN < Medium RE Ipsi AR threshold (BBN) and PRN < Medium remaining correlations tended to be smaller and not significant. There was a positive correlation between measures of dichotic and monotic ANL, indicating that, while listening to speech, subjects who accepted more noise in the monotic condition were able to accept more noise in the dichotic condition as well. A negative correlation was found between ANL d and PRN, indicating that subjects accepting higher levels of BBN while listening to speech in the opposite ear performed better on phoneme recognition in noise. A negative correlation between contralateral ARTs and CSTEOAE indicated that subjects with high levels of contralateral AR pathway activation (low contralateral ARTs) also demonstrated high levels of MOCB activation (greater amounts of contralateral TEOAE suppression). This was true when contralateral ARTs were tested with probe right, BBN left and probe left, BBN right. A negative correlation also was found between ipsilateral ARTs and PRN, both tested at the right ear, suggesting that high levels of ipsilateral AR pathway activation corresponded to better performance on the phoneme recognition in noise task. Relevant correlations that were not significant included CSTEOAE to ANL m, ANL d, or PRN, indicating that the level of activity in the MOCB, as assessed by contralateral suppression of OAEs, does not affect the amount of noise accepted monotically or dichotically while listening to speech or the accuracy of recognition of phonemes in ipsilateral noise. Also not significant were correlations between ipsilaterally or contralaterally measured ART and ANL m or ANL d, indicating that the level of activity in the AR pathways does not affect the amount of noise accepted monotically or dichotically while listening to speech. ANL m was not correlated with PRN, suggesting that the two monotic tasks have different perceptual requirements. DISCUSSION This study was designed to investigate the possibility that measures of speech-innoise performance involving different perceptual tasks are related and that individual differences in levels of efferent activity in the AR arc and/or MOCB pathways may account for the unexplained intersubject variability in these subjective tasks. The aims were (1) to investigate relations among the three speech-in-noise performance measures (monotic ANL, dichotic ANL, monotic PRN); (2) to determine if three different measures of speech-in-noise performance are related to two measures of efferent, suppressive feedback to the cochlea (ART, CSTEOAE); and (3) to investigate relations between the two measures of efferent auditory activity. Relations among Performance-in-Noise Measures Results related to the first aim indicated that the amount of background noise subjects were willing to accept monotically (ANL m ) was correlated with the amount of background noise subjects were willing to accept dichotically (ANL d ). Additionally, subjects accepting more background noise dichotically tended to perform better on the monotic task of recognizing phonemes in noise (PRN). Both findings suggest that ANL and PRN are mediated, to some extent, at a level at or beyond the superior olivary complex where binaural processing first occurs. However, the correlation between ipsilateral ARTs and PRN imply that this performance-in-noise measure has contributions from below the superior olivary complex as well. The large correlation between the ANL m and ANL d suggest that similar processing is involved for the dichotic and monotic conditions. In a questionnaire administered after the experimental session, most subjects (79%) reported dichotic and monotic ANL to 538

10 ANL, PRN, and Efferent Activity/Harkrider and Smith be very similar tasks, and most (87%) said that they used the same strategy for selecting the background noise level that was acceptable. In comparing the difficulty of the ANL m and ANL d tasks, 28 subjects reported the task difficulty to be within two rank order levels (scale = 1 10 for increasing difficulty), and one subject was within three rank order levels, perhaps indicating that the internal processes required to perform the ANL m and ANL d tasks were similar. Based on the findings that sounds are easier to identify when they are separated in space (Yost, 1994), one might expect ANLs to be smaller (subjects accepting more background noise while following the words of the story) for the dichotic ANL task versus the monotic ANL task. In the dichotic condition, the two signals are separated in space, which would involve localization, and are presented independently to separate ears, which involves lateralization. However, Figure 1 illustrates that subjects did not consistently accept more background noise in a dichotic condition. In fact, many subjects accepted similar amounts of noise for the two conditions as revealed by the number of symbols that fell near the solid line in Figure 1. This line represents equal ANL m and ANL d values, while the dashed line depicts the line < Figure 1. Relation between ANL m and ANL d. Individual ANL in a monotic condition (RE) plotted against ANL in a dichotic condition (story RE, babble LE). The dashed line represents the line of best fit. The solid line represents equal ANL m and ANL d values at any given point. The ANL d and ANL m measures were significantly correlated (r = 0.685, p 0.01). of best fit for the data. In terms of absolute values, 12/31 (39%) of subjects accepted more noise for the dichotic condition, and 19/31 (61%) accepted more noise for the monotic condition. The range of ANL d was larger than for the ANL m measure. Yet, even with the relatively high intersubject variability of the ANL d, the relationship between the ANL d and ANL m is robust. The fact that a dichotic listening condition does not improve acceptance of background noise is further evidence that the psychological factors required for ANL, whether in a monotic or dichotic listening condition, are distinctly different than those required for typical speech-in-noise measures. Previous findings with ANL (Nabelek et al, 2004a) suggest that this measure provides valuable information about a listener s performance in noise abilities that is not provided by speechrecognition-in-noise tasks, and the findings of this study appear to support this conclusion. No significant correlations were obtained between monotic ANL and PRN. No significant correlations between ANLs obtained monotically (Crowley and Nabelek, 1996) or binaurally in soundfield (Nabelek et al, 2004a) and SPIN scores have been found in listeners with hearing impairment. Thus, ANL appears to involve a different combination of perceptual variables than the SPIN or PRN and is possibly tapping into a different perceptual phenomenon. Curiously, the PRN task in a monotic condition did correlate with the ANL measure performed in a dichotic condition. There was a significant inverse relationship between the amount of background noise subjects were willing to accept in a dichotic condition (ANL d ) and performance in the monotic phoneme recognition task (PRN), indicating that subjects with normal hearing accepting a higher level of background noise introduced to the contralateral ear also performed better in noise in the ipsilateral ear. It is possible, considering the large number of correlations run, that the correlation between the ANL d and the PRN (p < 0.05) may be due to inflated Type I error. However, the effect size was medium to large (0.455), and so the association cannot be dismissed outright. Future research obtaining ANL m and ANL d in listeners with and without hearing impairment simultaneously participating in SPIN or PRN tasks would be helpful in clarifying the 539

11 Journal of the American Academy of Audiology/Volume 16, Number 8, 2005 relations, or lack thereof, between ANL and identification in noise performance. Speech in Noise Measures and Efferent Activity With regard to the second aim, intersubject variability in ANL m and ANL d could not be accounted for by individual differences in the level of activity in the AR pathway measured by ipsilateral or contralateral ARTs. There are no previous studies examining relations between ARTs and ANL or any other measure of acceptance or tolerance of background noise. The lack of correlation is consistent with the hypothesis that ANL is influenced by more central regions of the auditory system. Although possible, it does not seem likely that ARTs were not a sensitive enough physiological measure to correlate with a behavioral measure like ANL, because ipsilateral ARTs did correlate with PRN scores. Subjects with lower ipsilateral ARTs did tend to have better PRN scores, indicating that the ipsilateral AR may contribute to speech recognition in noise presented at moderate intensity levels (speech = 55 db HL, babble = 55 db HL) in the same ear. It is not surprising that contraction of the AR pathway on one side may reduce the masking effects of the babble on monosyllabic words on the same side. This is in accord with previous studies showing the AR to decrease the effects of background noise on signal detection and speech performance (Borg and Zakrisson, 1973; Mahoney et al, 1979; Wormald et al, 1995). However, this role of the AR is thought to be most effective for highintensity signals and maskers ( 90 db SPL) due to the high thresholds of the AR (Borg and Zakrisson, 1973; Mahoney et al, 1979; Wormald et al, 1995). Borg and Zakrisson (1973) did show some contributions of the AR in improving speech understanding in noise at lower intensity levels (75 90 db SPL) for three of four subjects whose mean ARTs were 97 db SPL to speech stimuli. The authors conclude that AR contributions to speech understanding in noise could have occurred at levels near or below the clinically recorded ARTs to speech. Likewise, in the current study, ipsilateral ARTs to BBN were above 55 db HL for all but two cases (one subject to right ipsilateral BBN), with a mean of 69 db HL. Comparatively, for the PRN measure the monosyllabic words and the speech babble were presented at 55 db HL. It was expected that the speech-babble stimuli used in the PRN would evoke similar ARTs as the BBN since all of these signals are, by definition, broadband. If this assumption is valid, the inverse relation between the ART and the PRN could indicate that AR contractions were occurring at a level below the ARTs that were measured by the immittance bridge for the great majority of subjects. Thus, activation of the AR pathway may occur to sounds lower in intensity than those predicted by typical immittance measures and may explain how the AR could play a role in improving speech perception in noise at moderate-intensity noise levels and not just high-intensity ones. If so, individual differences in PRN performance could be due, in part, to intersubject variability in AR pathway activation. The fact that there was no measure of AR activity that correlated with ANL m or ANL d but one that did correlate with PRN, suggests that primary contributions to ANL versus PRN could be from different areas of the auditory system. No correlations were found between contralateral suppression of TEOAEs and any of the speech-in-noise performance measures. These results reject the hypotheses that individual differences in efferent activity of the MOCB contribute to the intersubject variability in the amount of background noise accepted while listening to monotic or dichotic speech or the intersubject variability in speech recognition in monotic noise. As with ARTs, there are no previous studies examining relations between MOCB activity and ANL or any other measure of acceptance of background noise and only one (Giraud et al, 1997) reporting relations between MOCB activity and phoneme recognition in noise. Contrary to the findings of the present study, Giraud et al (1997) reported correlations between CSTEAOEs and a phoneme recognition-in-noise task. The conflicting data may be attributed to methodological differences in the word recognition task itself and in the analysis of the psychophysical data. CSTEOAEs were measured in a similar manner in both studies. Giraud et al (1997) measured monaural word recognition in continuous ipsilateral noise with and without the simultaneous introduction of contralateral BBN. The 540

12 ANL, PRN, and Efferent Activity/Harkrider and Smith contralateral BBN was introduced to elicit the MOCB response during the word recognition task. The slope of the psychometric function for each subject was measured as the SNR of the monosyllabic words, and the masker in the ipsilateral ear was varied. Slopes were then compared to the amount of TEOAE suppression. A steeper slope was obtained following the introduction of contralateral noise, indicating a reduction in the effective ipsilateral masking of the phonemes in the presence of the contralateral noise. The authors attributed this improvement to the contralateral noise activating the MOCB. However, even in the absence of the contralateral noise, the MOCB could have been activated by the ipsilateral masker, because ipsilateral stimuli have been shown to produce significant suppression of OAEs (Berlin et al, 1995). The baseline phonemerecognition scores used in the control condition (ipsilateral masker only) may have also been affected by the MOCB, even in the absence of contralateral suppression. Adding the contralateral masker during the ipsilateral phoneme recognition-in-noise task created the condition of binaural masking signals. In contrast, the current study evaluated phoneme recognition only in a monotic condition. If there were MOCB affects on phoneme recognition-in-noise performance, it is possible that a binaural masking condition would have resulted in greater improvements in phoneme recognition performance than an ipsilateral masking condition. This possibility is supported by the findings of Berlin et al (1995), who demonstrated greater binaural versus ipsilateral suppression of OAEs. MOCB effects on phoneme recognition in noise in the current study would have only resulted from the less robust ipsilateral MOCB activation, while the MOCB effects in the Giraud et al (1997) would have resulted from the more robust bilateral MOCB activation. Perhaps binaural activation of the MOCB is necessary to produce a significant improvement in the phoneme recognition scores, and so a measurable correlation between CSTEOAEs and PRN. It may also be the case that binaural activation of the MOCB is necessary to influence ANL measures and thus produce a measurable correlation between CSTEOAEs and monotic or dichotic ANL. It is also possible that the results of the present study accurately reflect a lack of relations between MOCB activity and ANL or PRN. If so, these performance-in-noise measures may be influenced by primarily cortical regions of the auditory system. In this case, individual differences in the level of MOCB activity would not be expected to contribute to individual differences in the performance-in-noise measures. The strong positive correlation between ANL m and ANL d suggests that ANL is mediated at the level of the superior olivary complex or beyond, because dichotic auditory information is not available until this level of the brainstem. Interestingly, preliminary evoked potential data from our laboratory measured in females with low and high ANLs appear to support this hypothesis. The correlation between ANL d and PRN suggests that phoneme recognition in a monotic noise condition also may be mediated, at least in part, at or above the superior olivary complex. However, unlike ANL m or ANL d, PRN was correlated with AR activity, suggesting that it may be influenced by more peripheral auditory structures as well. Relations between Measures of Auditory Efferent Activity With reference to the third aim, subjects with lower contralateral ARTs had significantly greater CSTEOAEs. It is possible that middle ear impedance affected the measures from the two efferent pathways similarly and/or that the contralateral suppressor of the TEOAEs was at a level that unexpectedly activated the AR, which added to the overall amount of suppression measured. It is also possible that there was no methodological confound and subjects with stronger efferent activity in one pathway simply have stronger efferent activity in the other. Middle ear impedance characteristics have been shown to have an influence on the measurements of ARTs (Silman, 1984) and TEOAEs (for review, see Margolis, 2002). Higher ARTs and smaller OAE amplitudes coincide with greater immittance of the middle ear system. Higher middle ear immittance ultimately reduces the amount of energy that is transduced from the external stimulus to the inner ear and back from the inner ear and middle ear to the recording microphone. A subject with relatively greater acoustic immittance would be expected to have increased contralateral ARTs (Silman, 1984), and this same subject should also have 541