Research Design: This study consisted of a within-participant design with repeated measures across test conditions.

J Am Acad Audiol 25:1022 1033 (2014) Evaluation of Wideband Frequency Responses and Nonlinear Frequency Compression for Children with Cookie-Bite Audiometric Configurations DOI: 10.3766/jaaa.25.10.10 Andrew John* Jace Wolfe Susan Scollie Erin Schafer Mary Hudson* Whitney Woods* Julie Wheeler* Krystal Hudgens Sara Neumann Abstract Background: Previous research has suggested that use of nonlinear frequency compression (NLFC) can improve audibility for high-frequency sounds and speech recognition of children with moderate to profound high-frequency hearing loss. Furthermore, previous studies have generally found no detriment associated with the use of NLFC. However, there have been no published studies examining the effect of NLFC on the performance of children with cookie-bite audiometric configurations. For this configuration of hearing loss, frequency-lowering processing will likely move high-frequency sounds to a lower frequency range at which a greater degree of hearing loss exists. Purpose: The purpose of this study was to evaluate and compare the effects of wideband amplification and NLFC on high-frequency audibility and speech recognition of children with cookie-bite audiometric configurations. Research Design: This study consisted of a within-participant design with repeated measures across test conditions. Study Sample: Seven children, ages 6 13 yr, with cookie-bite audiometric configurations and normal hearing or mild hearing loss at 6000 and 8000 Hz, were recruited. Intervention: Participants were fitted with Phonak Nios S H2O III behind-the-ear hearing aids and Oticon Safari 300 behind-the-ear hearing aids. Data Collection: The participants were evaluated after three 4-to 6-wk intervals: (1) Phonak Nios S H2O III without NLFC, (2) Phonak Nios S H2O III with NLFC, and (3) Oticon Safari 300 with wideband frequency response extending to 8000 Hz. The order in which each technology was used was counterbalanced across participants. High-frequency audibility was evaluated by assessing aided thresholds (db SPL) for warble tones and the high-frequency phonemes /sh/ and /s/. Speech recognition in quiet was measured with the University of Western Ontario (UWO) Plurals Test, the UWO Distinctive Features Difference (DFD) Test, and the Phoneme Perception Test vowel-consonant-vowel nonsense syllable test. Sentence recognition in noise was evaluated with the Bamford-Kowal-Bench Speech-In-Noise (BKB-SIN) Test. *Department of Communication Sciences and Disorders, University of Oklahoma Health Sciences Center, Oklahoma City, OK; Hearts for Hearing, Oklahoma City, OK; School of Communication Sciences and Disorders and National Centre for Audiology, Faculty of Health Sciences, University of Western Ontario, London, Ontario, Canada; Department of Speech and Hearing Sciences, University of North Texas, Denton, TX Jace Wolfe, 3525 NW 56th Street, Suite A-150, Oklahoma City, OK 73112; E-mail: jace.wolfe@heartsforhearing.org This study was funded by a grant from Phonak AG to Andrew John and Jace Wolfe. 1022

NLFC for Cookie-Bite Audiograms/John et al Analysis: Repeated-measures analyses of variance were used to analyze the data collected in this study. The results across the three different conditions were compared. Results: No difference in performance across conditions was observed for detection of high-frequency warble tones and the speech sounds /sh/ and /s/. No significant difference was seen across conditions for speech recognition in quiet when measured with the UWO Plurals Test, the UWO-DFD Test, and the Phoneme Perception Test vowel-consonant-vowel nonsense syllable test. Finally, there were also no differences across conditions on the BKB-SIN Test. Conclusions: These results suggest that NLFC does not degrade or improve audibility for and recognition of high-frequency speech sounds as well as sentence recognition in noise when compared with wideband amplification for children with cookie-bite audiometric configurations. Key Words: Hearing aids, moderate hearing loss, nonlinear frequency compression Abbreviations: ANOVA 5 analysis of variance; BKB-SIN 5 Bamford-Kowal-Bench Speech-In-Noise; BTE 5 behind-the-ear; DFD 5 Distinctive Features Differences; HSD 5 honestly significant difference; NLFC 5 nonlinear frequency compression; PPT 5 Phoneme Perception Test; PTA 5 pure-tone average; SNR 5 signal-to-noise ratio; UWO 5 University of Western Ontario; VCV 5 vowel-consonant-vowel; WB 5 wideband INTRODUCTION Frequency-lowering hearing aid technology shifts high-frequency sounds to a lower frequency range where patients presumably possess better hearing sensitivity. Frequency lowering is also used to overcome bandwidth limitations inherent in many contemporary behind-the-ear (BTE) hearing aids when worn on the ear with a custom earmold. A number of different frequency-lowering approaches exist in commercially available hearing aids; these are nicely summarized in existing literature (Galster et al, 2011; Glista et al, 2009; Simpson, 2009). Several recent publications have demonstrated that children and adults experience improvements in audibility and speech recognition associated with the use of nonlinear frequency compression (NLFC). Simpson and colleagues (2005) evaluated word recognition in a group of adults who had moderate sloping to severe to profound high-frequency hearing loss and used a noncommercial body-worn hearing aid providing conventional amplification,nlfc,andwideband amplification (i.e., frequency response extended to 8000 Hz). They reported that wideband amplification provided no improvement in speech recognition compared with conventional amplification. However, participants did show significant improvements in word recognition, consonant recognition, and fricative identification with NLFC use compared with both wideband and conventional amplification. In a follow-up study, Simpson et al (2006) evaluated word recognition in a group of adults with precipitously sloping hearing losses that were severe to profound above 1500 Hz. No improvements in speech recognition were seen with NLFC compared with conventional amplification. Simpson et al (2006) attributed this finding to the fact that the participants in the latter study had severe to profound hearing loss over most of the speech frequency range; therefore, the full spectrum of mid- and highfrequency speech sounds simply could not be compressed enough to restore audibility without substantial distortion of the input signal. Glista et al (2009) evaluated NLFC in a group of children and adults who had high-frequency hearing loss ranging from moderate to profound in degree. The authors objectives were to determine the benefits and limitations of NLFC, to determine appropriate methods for optimizing various NLFC parameters (crossover frequency, compression ratio, etc.) to optimize user performance, and to identify methods to verify outcomes with NLFC technology in realistic situations. Glista and colleagues reported that NLFC provided an improvement in the group score for consonant and plural (e.g., /s/) recognition when compared with conventional amplification, but no change was observed for vowel recognition. Additionally, the authors noted greater improvements in performance with NLFC for children compared with adults; furthermore, more NLFC benefit was typically associated with greater degrees of high-frequency hearing loss. Substantial variability in NLFC benefit was noted across participants; some experiencing no benefit from NLFC, whereas others received significant improvement in speech recognition and detection of high-frequency sounds with NLFC. Glista and colleagues suggested that the participants degree of high-frequency hearing loss, age, and a potential need for a period of acclimatization may have influenced the benefit individuals received from NLFC. Wolfe et al (2010; 2011) evaluated the potential advantages and limitations of NLFC for a group of 15 schoolaged children with moderate high-frequency hearing loss. Specifically, Wolfe et al (2010) measured sound detection and speech recognition in quiet and in noise after each of two 6 wk periods: one with NLFC and one without NLFC. The order in which children used the NLFC feature was counterbalanced to prevent an 1023

Journal of the American Academy of Audiology/Volume 25, Number 10, 2014 order effect from biasing results. Wolfe et al (2010) reported that the children achieved better detection of high-frequency warble tones and the phoneme /s/ with the use of NLFC relative to conventional amplification. Better speech recognition of the plurals /s/ and /z/ in the final position with NLFC enabled was also seen, as was as better recognition of the phoneme /s/when spoken by a female (center frequency near 9000 Hz) on a vowelconsonant-vowel (VCV) nonsense syllable test. The authors found no difference between the NLFC on-and-off conditions for the phonemes /sh/, /f/, /t/, /k/, or the phoneme /s/ when spoken by a male (center frequency near 6000 Hz) on the VCV nonsense syllable test. Likewise, they also reported no difference between NLFC on-and-off conditions for sentence recognition in noise. Wolfe et al (2011) also evaluated speech recognition in quiet and in noise after the children used NLFC for an additional 6 mo to evaluate whether an extended period of NLFC use facilitated acclimatization to the frequency-compressed speech, allowing for further improvements in performance. After 6 mo of NLFC use, the participants showed continued improvement of recognition of the high-frequency phonemes on the VCV phoneme test. In particular, the authors reported significant improvement with NLFC for both the female (center frequency at 9000 Hz) and male (center frequency at 6000 Hz) /s/, as well as for the /t/, /f/, and /d/. Furthermore, the children also achieved significantly better speech recognition in noise after 6 mo of NLFC use compared with their performance with NLFC off. These reports by Wolfe and colleagues were clinically significant because they indicated that children with moderate high-frequency hearing loss may achieve better detection and recognition of high-frequency speech sounds with NLFC, and also that children may need several months of NLFC use before they realize the full benefit of the technology. Wolfe et al (in press) recently evaluated NLFC use for a groupof11school-agedchildrenwithmildhigh-frequency hearing loss (,50 db HL pure-tone average [PTA] and,50 db HL threshold at 8000 Hz for the better ear). The children were evaluated with use of three hearing aid technologies: (1) the Phonak S Nios H2O BTE hearing aid without NLFC, (2) the Phonak S Nios H2O BTE hearing aid with NLFC enabled, and (3) the Oticon Safari 300 BTE with a wideband frequency response (e.g., a reported frequency response extending to 8000 Hz). The children used each of the aforementioned hearing aid technologies for 4 6 wk, and their performance was evaluated at the end of each interval. The order in which the children used each technology was counterbalanced to avoid biasing from an order/learning effect. Wolfe et al (submitted) reported no differences between NLFC and the wideband frequency response conditions for recognition of plurals (/s/ and /z/) in the final position, recognition of consonants on the University of Western Ontario (UWO) Distinctive Features Differences (DFD) nonsense syllable test (e.g., acil), nor for sentence recognition in noise. However, use of NLFC provided better detection for high-frequency warble tones (6000 and 8000 Hz) as well as better recognition of the phoneme /s/ when spoken by a male (center frequency of 6000 Hz) and female (center frequency of 9000 Hz) on an adaptive assessment of the recognition threshold of consonants in a VCV nonsense syllable test. Wolfe et al (submitted) concluded that the use of NLFC is not advantageous or detrimental for children with mild hearing loss as they listen to speech at sufficient sensation levels. However, they did suggest that NLFC does improve mildly hearing-impaired children s detection and recognition of low-level, high-frequency speech sounds. Currently there are no published studies evaluating NLFC for children with audiometric configurations in which hearing thresholds are better in the high frequencies (4000 8000 Hz) than the mid frequencies (1000 4000 Hz), such as rising audiograms or cookiebite/u-shaped configurations. For example, no studies have been published evaluating the effects of NLFC use for children with cookie-bite audiometric configurations (i.e., an audiometric in which hearing thresholds in the low frequencies (,1000 Hz) and high frequencies ($4000 Hz) are better than hearing thresholds in the mid frequencies (1000 4000 Hz). It is reasonable to suppose that NLFC may degrade performance by lowering high-frequency speech sounds to a lower frequency range with relatively poorer hearing acuity than the original source region. On the contrary, one may propose that bandwidth limitations inherent in contemporary hearing technology (even possibly in hearing aids with a reported wideband response) may preclude sufficient audibility of high-frequency speech sounds. Indeed, Wolfe et al (submitted) have demonstrated that even children with mild hearing loss experienced some improvement in high-frequency phoneme detection and recognition with use of NLFC vs wideband amplification. Published research has shown that a fairly large proportion of children with hearing loss possess u-shaped or cookie-bite audiometric configurations (from this point forward, referred to as cookie-bite audiograms) (Macrae and Dillon, 1996; Pittman and Stelmachowicz, 2003). For instance, Pittman and Stelmachowicz (2003) studied audiometric configuration in a group of 227 children (age 6 yr) and reported that one of five children possessed a cookie-bite audiogram. Pittman and Stelmachowicz (2003) defined the cookie-bite configuration as an audiogram in which one or more adjacent thresholds between 500 4000 Hz were at least 20 db poorer than the threshold at 250 or 8000 Hz. They also noted that the rounded, cookie-bite configuration was more common in the pediatric population they sampled compared with a group of 248 adults between 60 and 61 yr old. Macrae and Dillon (1996) also noted a similar 1024

NLFC for Cookie-Bite Audiograms/John et al difference in audiometric configurations between children and adults. Pittman and Stelmachowicz (2003) note the paucity of published data validating hearing aid fitting and signal-processing options for patients with cookie-bite audiograms and note the need for additional research with this patient population. The primary objective of the present study was to evaluate the potential advantages and limitations of one signal processing option, NLFC, for children with cookie-bite audiograms. Specifically, we evaluated the detection of high-frequency warble tones and speech sounds as well as speech recognition in quiet and in noise for a group of children who possessed cookie-bite audiometric configurations and who used two sets of hearing aids, one programmed both with and without NLFC, and another set of hearing aids featuring a wideband frequency responses (e.g., reported bandwidth extending to 8000 Hz). METHODS Participant Characteristics Nine children recruited from the patient population of the Hearts for Hearing Foundation in Oklahoma City, Oklahoma, USA, served as participants. However, one child had to withdraw from the study because the family could not arrange transportation to travel to testing sessions and another child was withdrawn because of the inability to follow study instructions. Informed consent was obtained from all participants and parents before data collection. The study was approved as human subject research by the University of Oklahoma Health Sciences Center Institutional Review Board, IRB #14439. All children had cookie-bite hearing losses (see Figure 1 for mean audiometric thresholds). In order to be classified as having a cookie-bite audiogram for inclusion in the study, the children met the following criteria: 1. Pure-tone thresholds at 250 and 8000 Hz at least 15 db better than the poorest threshold from 1000 3000 Hz, 2. Pure-tone threshold at 8000 Hz not exceeding 45 db HL in the better ear, 3. PTA poorer than 30 db HL but better than 70 db HL, 4. Participant between 6 and 13 yr old, 5. Participant consistently using amplification but without prior experience with NLFC, 6. Use of English-spoken language as primary mode of communication, 7. No cognitive or behavioral disabilities that would prevent completion of the tests used in the study, and 8. No conductive hearing loss or auditory neuropathy spectrum disorder hearing loss. The mean age of the children who participated in the study was 10.8 yr old (age range 5 6.8 13.7 yr, SD 5 2.4 yr). The mean PTA for the right and left ears was 39.8 and 37.1 db HL, respectively, and the mean air conduction pure-tone threshold at 8000 Hz for the right and left ear was 20.6 and 18.8 db HL, respectively. All participants were fit binaurally with Oticon Safari 300 and Phonak Nios S III BTE hearing aids. The Oticon Safari 300 hearing aid possesses a reported Figure 1. Mean audiogram (n 5 8). 1025

Journal of the American Academy of Audiology/Volume 25, Number 10, 2014 bandwidth of 8000 Hz, which refers to the upper limit of the digital signal processor (2 cc coupler response: 100 6800 Hz) (Oticon, 2013). The reported bandwidth for the Phonak Nios S H2O III device is 7100 Hz (2 cc coupler response: 100 7100 Hz) (Phonak, 2013). Hearing Aid Fitting Procedure The real-ear-to-coupler difference was measured for each child. Then, for both the Oticon and Phonak hearing devices, simulated probe microphone measures (with the measured real-ear-to-coupler difference) were conducted on the Audioscan Verifit hearing aid analyzer to evaluate the output of the hearing aids. Simulated probe microphone measures were used in this study to reduce the chances of test-retest error associated with fitting two different hearing instruments with in situ probe microphone measures (Dillon, 2001). Simulated measurements performed in a 2 cc coupler also allowed the clinician to examine the hearing aid frequency response to 8000 Hz. The fitting clinician matched, as closely as possible, the hearing aid output to the DSL v5.0 target (65 db) for the target speech signal (presented at 55, 65, and 75 db SPL). Furthermore, the maximum output was measured to an 85 db SPL swept pure tone, and the clinician ensured that it did not exceed the DSL v5.0 targets for the 85 db SPL swept-tone. For the Phonak device, these initial probe microphone measurements were made while NLFC was disabled. Then, probe microphone measures were performed with NLFC enabled to ensure that the 6300 Hz frequencylowering verification stimulus of the Audioscan Verifit was sufficiently audible (i.e., the compressed signal met or exceeded the DSL v5.0 target at the destination frequency). When NLFC was enabled, the clinician ensured that the output for the 6300 Hz frequency-lowering verification stimulus was positioned within the frequency response of the hearing aid when NLFC was disabled (i.e., the output for the 6300 Hz FLVS was positioned before the rolloff of the frequency response in the NLFC-off mode). Based on probe microphone measures, the crossover frequency of NLFC was set lower than the manufacturer s default for three of the children. The crossover frequency is the initial frequency at which NLFC is applied. To ensure that NLFC parameters were fitted appropriately for the children, the examiners made certain that the children could discriminate between /sh/ and /s/ without visual cues when spoken at an average conversational speech level from 12 feet away. None of the children exhibited confusion of the /s/ for /sh/ phonemes, and none reported poor sound quality with NLFC enabled. This finding is in contrast to Glista and colleagues experiences using NLFC with persons with severe to profound highfrequency hearing loss, who occasionally showed /sh/ for /s/ confusion when NLFC parameters were set aggressively (Glista et al, 2009). Study Design The study included three 4 6 wk trial periods in which the participants used the hearing aids in one of three conditions: (1) Oticon Safari 300 wideband (WB), (2) Phonak Nios S H2O III with NLFC disabled ( NLFC OFF ), or (3) Phonak Nios S H2O III with NLFC enabled ( NLFC ON ). The order in which the NLFC was used was randomly counterbalanced across participants. Recognition of speech sounds in quiet, speech recognition in noise, and aided thresholds were evaluated following each 4 6 wk trial to allow for a comparison of performance with wideband amplification to performance versus NLFC enabled versus NLFC disabled. Study Measures All audiometric outcome measures were repeated at each of the three test intervals of the study (i.e., after a 4 6 wk period with NLFC enabled in the Phonak hearing aid, after a 4 6 wk period with NLFC disabled in the Phonak hearing aid, and after a 4 6 wk period with the Oticon wideband hearing aid). All assessments were performed in the binaural-aided condition using sound-field stimuli. These assessments consisted of the following: Assessment of Aided High-Frequency Tone Thresholds Aided sound-field thresholds were measured for warble tones centered at 4000, 6000, and 8000 Hz. Aided threshold measurements were measured using a Madsen Orbiter 922 type 1 audiometer routed through a calibrated Crown D75A two-channel amplifier to a JBL Control 5 loudspeaker. The speaker was located at head level, 1 m away from and at 0 azimuth relative to the test participant. The Hughson-Westlake ascending method of threshold determination was used to assess aided thresholds. Specifically, the threshold search at each frequency involved ascending in level in 5 db steps until the stimulus was detected. At that point, the level was decreased by 10 db, and an up 5 / down 10 db approach was used until the participant responded twice at the same level while the stimulus level was ascended. Then, an up 2 / down 4 db approach was used, and threshold was recorded as the lowest level at which the participant responded twice while the stimulus level was ascending. Assessment of High-Frequency Speech Detection and Recognition Consonant detection and recognition were evaluated using three tasks: the UWO Plurals Test (Glista and Scollie, 2012), the UWO-DFD Test (Cheesman and Jamieson, 1996), and the Phoneme Perception Test (PPT) (Boretzki and Kegel, 2009). These are described below. 1026

NLFC for Cookie-Bite Audiograms/John et al The UWO Plurals Test is an open-set, speech-recognition task developed by researchers at UWO to evaluate a hearing aid wearer s ability to hear the high-frequency phonemes /s/ and /z/ as bound morphemes indicating plurality (Glista et al, 2009). In the present study, the UWO Plurals Test was administered across all three hearing aid conditions. This test contains 15 separate words that should be familiar to the typical school-aged child. The words are included in both the singular and plural form (ant[s], balloon[s], book [s], butterfly[/ies], crab[s], crayon[s], cup[s], dog[s], fly [/ies], flower[s], frog[s], pig[s], skunk[s], sock[s], and shoe[s]). Test items are spoken by a female talker and are presented in recorded format, so it is imperative that children have access to acoustic energy beyond 6000 Hz for correct identification of the plural in each word. In fact, the centroid frequency of the /s/ phoneme in the final position of the target words occurs at approximately 7050 Hz. The UWO Plurals Test was presented at a level of 50 dba. The test stimuli were generated from a JVC XLFZ-258 CD player, which was coupled to the Madsen Orbiter 922 audiometer by an interface cable. The test stimuli were then presented from a JBL Control 5 loudspeaker located 1 m directly in front of the child and at the approximate plane of the child s head. A 1000 Hz calibration tone provided with the UWO Plurals Test was initially used to set the V.U meter to 0 db. Then, a type 2 sound-level meter was used at the beginning of the assessment period with each child to confirm that the calibration signal provided on the UWO Plurals Test CD corresponded to the desired presentation level (when measured with the microphone placed at the location of the child s head during the test session) of each speech stimulus. Participants provided oral responses to the test stimuli, which were scored in open-set by the investigators as correct or incorrect. The children were instructed to repeat the words that they heard, and they were not aware that their task was to recognize the bound morpheme /s/ or /z/ that may exist at each word. Two 30-word lists of the UWO Plurals Test were presented for a total of 60 words in each condition. The UWO-DFD task was presented using test software installed on a Linux-enabled laptop PC. Signals were routed from the lineout channel of the laptop via a Crown D-75A two-channel amplifier to a JBL Control 5 loudspeaker located 1 m directly in front of the child and at the approximate plane of the child s head. Closed-set test stimuli, which consisted of 16 phonemes embedded in medial position within the nonsense token /a il/, were presented at 60 dba by alternating male and female recorded voices. The participant s task was to use the computer mouse to select the consonant heard in the stimulusmedial position from a set of 21 consonants and consonant clustersdisplayedonamonitor(b,ch,d,f,g,h,j,k,l, M, N, P, R, S, SH, T, TH, V, W, Y, Z). A practice list was administered prior to beginning scoring to orient participants to the task. The UWO-DFD Test was completed in the NLFC On and Oticon wideband conditions for all participants. Technical difficulty with the test computer prevented the UWO-DFD Test from being completed with some of the children who were scheduled for assessment after the 4 6 wk interval in which they used the Phonak aid without NLFC. As a result, a full dataset was not collected for the Phonak NLFC Off condition. A total of 42 stimuli were presented in each condition, and a percentagecorrect score was calculated. For a more complete description of the UWO-DFD, the reader is referred to Cheesman and Jamieson (1996). Consonant detection and recognition thresholds were also evaluated using the detection and recognition tasks, respectively, of the PPT. The PPT was developed by hearing scientists and audio engineers of Phonak AG (Boretzki and Kegel, 2009) and features three subtests: (1) a detection task, (2) a recognition task, and (3) an identification task. Only the detection and recognition tasks were used in the present study. For these two subtests of the PPT, stimuli were presented by a Windowsenabled computer via lineout to a Crown D-75A two-channel amplifier and a JBL Control 5 loudspeaker located at head level 1 m directly in front of the participant. A type 2 sound-level meter was used at the outset of the assessment period with each participant to confirm that the broadband calibration signal produced by the PPT software output was set to the software-indicated 70 dba. Subsequent sound-level measurements were conducted to verify the appropriate presentation level of the stimuli at regular intervals. The detection task involves the determination of threshold (in db SPL) for four phonemes: (1) /sh/ low, (2) /sh/ high, (3) /s/ low, and (4) /s/ high. The /sh/ low stimulus is intended to acoustically approximate the /sh/ when spoken by an adult male and possesses a center frequency at approximately 3000 Hz. The /sh/ high stimulus possesses a center frequency near 5000 Hz in order to simulate the /sh/ phoneme when spoken by a female. The /s/ low and /s/ high stimuli possess center frequencies of 5000 and 9000 Hz, respectively, in an attempt to approximate the male and female /s/, respectively. A Hughson-Westlake clinical method of ascending limits was used to evaluate detection thresholds for each of the aforementioned stimuli. Stimuli were adjusted in an up 5 / down 10 db approach until two responses were obtained at the same level in an ascending presentation level. Next, the stimuli were adjusted in an up 1 / down 2 db approach, and threshold was defined as the lowest level in which two responses were obtained in an ascending presentation level. The recognition subtest of the PPT involved the adaptive presentation of seven nonsense tokens (/asa/, /ada/, /afa/, /aka/, /asha/, /aha/, and /ama/) presented by a female speaker and included at the end of the carrier phrase My name is [token]. The center frequency of 1027

Journal of the American Academy of Audiology/Volume 25, Number 10, 2014 the /sh/ and /s/ phonemes in the asha/ and /asa/ were positioned at approximately 5000 and 9000 Hz, respectively, which is typical for female speech. Two extra stimuli, referred to as asha5k and asa6k, were filtered so that the center frequencies were positioned at 3000 and 6000 Hz, respectively, which more closely approximates the center frequency of /sh/ and /s/ when spoken by a male. The PPT was administered using a computer-controlled adaptive procedure in which the listener was asked to respond to stimuli by selecting (via the use of a wireless computer mouse) one of the seven tokens displayed on a monitor in the test booth as buttons labeled ADA, AFA, AKA, ASA, ASHA, AMA, and AHA. The PPT adaptive procedure is described by Boretzki and Kegel (2009) and Wolfe et al (2010). Briefly, the PPT test uses an ascending/descending procedure to determine the presentation level for 50% correct performance. For each test administration, approximately 100 130 stimuli are presented, with the actual number varying based on the reliability of participant responses. The PPT test software then calculates the participant s threshold in db SPL for 50% identification of /s/ low, /s/ high, /sh/ low, and /sh/ high nonsense syllables. Assessment of Speech Recognition in Noise The signal-to-noise ratio (SNR) required for 50% correct sentence recognition was determined by administering two list pairs from the Bamford-Kowal-Bench Speech- In-Noise (BKB-SIN) Test (Bench et al, 1979). Testing was conducted according to the procedures described in the BKB-SIN test manual (Etymotic Research, 2005). Specifically, the sentences were generated from a JVC XL-FZ258 CD changer routed to a Madsen Orbiter 922 audiometer, Crown D-75A two-channel amplifier, and JBL Control 5 loudspeaker located at head level at 1 m directly in front of the participant. Presentation of BKB-SIN stimuli was at a fixed level (50 db HL in this study), and the level of the multitalker competing noise signal increased throughout each list of the test. The sentences and competing noise were both presented from the same loudspeaker, located at head level, 1 m directly in front of the participant, and the SNR for 50% correct performance was determined. Study measures were conducted during 60 90 min sessions. The order of task presentation was randomly assigned across and within each child. Children were given the opportunity to take breaks throughout each test session. A type 1 sound-level meter (Ivie IE-45 audio analyzer with Bruel & Kjaer model 4940 1/2-inch microphone) was used to verify and calibrate presentation levels for signals used in all study measures before data collection on each testing day. RESULTS Mean thresholds for high-frequency warble tones (4000, 6000, and 8000 Hz) as a function of amplification condition are displayed in Figure 2. A 3 3 3 analysis of variance (ANOVA) was performed to test for the effect of tone frequency (4000, 6000, and 8000 Hz) and amplification condition (Oticon Safari 300 WB, Phonak Nios S H2O III with NLFC enabled [NLFC ON], and Phonak Nios S H2O III with NLFC disabled [NLFC OFF] on thresholds for high-frequency warble tones. This analysis revealed significant effects of tone frequency (F [2,12) 5 4.44, p 5 0.04, h 2 p 5 0.43) and condition (F [2,12] 5 4.26, p 5 0.04, h 2 p 5 0.42), and the interaction term (F [4,24] 5 4.66, p 5 0.006, h 2 p 5 0.44). Post hoc testing using the Tukey honestly significant difference (HSD) test found thresholds for the 8000 Hz tone to be significantly higher (worse) than thresholds Figure 2. Mean aided thresholds (db SPL) for detection of high-frequency warble tones by hearing aid condition. 1028

NLFC for Cookie-Bite Audiograms/John et al Figure 3. Mean percent-correct scores on the UWO Plurals Test by hearing aid condition. for the 4000 Hz tone (p 5 0.007); no other comparisons were significant. Post hoc testing using the Tukey HSD test found no significant differences among amplification conditions, although comparisons between NLFC ON and WB (p 5 0.056) and between NLFC OFF and WB (p 5 0.059) approached significance. UWO Plurals Test Mean percent-correct scores on the UWO Plurals Test as a function of amplification condition are displayed in Figure 3. A repeated-measures ANOVA for the effect of amplification across the three fitting conditions on UWO Plurals Test performance was not significant (F [2,12] 5 0.58, p 5 0.57). UWO-DFD Test Mean percent-correct scores on the UWO-DFD Test as a function of amplification condition are displayed in Figure 4. A repeated-measures ANOVA for the effect of amplification between the WB and NLFC ON conditions on UWO-DFD test performance was not significant (F [1,6] 5 0.04, p 5 0.85). Examination of specific consonant errors did not reveal any important patterns that could be related to frequency compression: overall performance and fricative-specific errors were essentially unchanged between the two listening conditions. Note that the UWO-DFD Test was not administered in the NLFC OFF condition. One participant did not complete this test. PPT Detection Test Mean sound-field detection thresholds for highfrequency consonant tokens as a function of amplification condition are displayed in Figure 5. A 4 3 3 ANOVA was performed to test for the effect of test token (/s/ low, /s/ high, /sh/ low, /sh/ high) and the three amplification conditions on high-frequency phoneme detection threshold. Figure 4. Mean percent-correct scores on UWO-DFD Test by hearing aid condition. 1029

Journal of the American Academy of Audiology/Volume 25, Number 10, 2014 Figure 5. Mean aided thresholds (db SPL) for detection of high-frequency consonant tokens on the PPT Detection subtest by hearing aid condition. This analysis revealed significant effects of token (F [3,18) 5 8.63, p 5 0.001, h p 2 5 0.59), and the interaction term (F [6,36] 5 5.95, p, 0.001, h p 2 5 0.50). There was no significant effect of amplification condition (F [2,12] 5 0.91, p 5 0.83). Post hoc testing using the Tukey HSD found thresholds for the /s/ high token to be significantly higher (worse) than thresholds for the /s/ low (p 5 0.002), /sh/ high (p 5 0.009), and the /sh/ low (p 5 0.018) tokens. No significant difference was seen among the /s/ low, /sh/ high, and /sh/ low tokens (all p. 0.05). PPT Recognition Test Mean sound-field detection thresholds for high-frequency consonant tokens as a function of amplification condition are displayed in Figure 6. A 4 3 3 ANOVA was performed to test for the effect of test token (/s/ low, /s/ high,/sh/low,/sh/high)andthethreeamplification conditions on high-frequency phoneme detection threshold. This analysis revealed significant effects of token (F [3,18) 5 3.96, p 5 0.03, h p 2 5 0.40), and the interaction term (F [6,36] 5 3.38, p 5 0.01, h p 2 5 0.36). There was no significant effect of amplification condition (F [2,12] 5 0.29, p 5 0.75). Post hoc testing using the Tukey HSD found thresholds for the /s/ high token to be significantly higher (worse) than thresholds for the /s/ low (p 5 0.009); no other comparisons were significant (all p. 0.05). BKB-SIN Test Mean SNR thresholds for 50% correct identification of BKB-SIN sentences in noise as a function of amplification condition are displayed in Figure 7. A repeatedmeasures ANOVA for the effect of amplification across Figure 6. Mean aided thresholds (db SPL) for closed-set recognition of high-frequency consonant tokens on the PPT Recognition subtest by hearing aid condition. 1030

NLFC for Cookie-Bite Audiograms/John et al the three fitting conditions on BKB-SIN performance was not significant (F [2,12] 5 0.09, p 5 0.91). DISCUSSION Speech recognition in quiet and in noise as well as detection of high-frequency warble tones and speech sounds were assessed while children with cookie-bite audiograms used an Oticon Safari 300 BTE hearing aid with a wideband frequency response (extended to 8000 Hz) and a Phonak Nios S H2O BTE hearing aid with NLFC disabled and NLFC enabled. The hearing aids were fit using procedures similar to those typically used by clinicians who fit hearing aids with frequencylowering technology for children with hearing loss. No significant differences were observed among the three conditions for detection of high-frequency sounds, recognition of high-frequency speech sounds, and sentence recognition in noise. Specifically, speech recognition in quiet and in noise was not compromised by NLFC nor was it improved with the use of NLFC. This finding is in contrast with what was observed in previous studies of performance with NLFC for children with mild and moderate highfrequency hearing loss (Wolfe et al, 2010; 2011; submitted). Wolfe et al (2010) evaluated the effects of NLFC in hearing aids for a group of 15 children with moderate hearing loss and found significant improvement in the detection of low-level, high-frequency sounds. They reported that children with moderate hearing loss achieved substantially better speech understanding in quiet and in noise with use of NLFC compared with their performance with wideband amplification. These improvements were maintained or increased during a long-term trial with NLFC (Wolfe et al, 2011). Wolfe et al (submitted) also evaluated the influence of NLFC on speech recognition in quiet and in noise for children with mild, high-frequency hearing loss. They concluded that children with mild hearing loss showed no improvement in speech recognition in quiet or in noise relative to performance with wideband amplification when the presentation level of test stimuli was set to provide sufficiently robust audibility through the high frequencies. However, Wolfe and et al (submitted) did find that children with mild hearing loss did experience better detection and speech recognition in quiet when the presentation level of the signal of interest is low. To summarize, Wolfe and colleagues have demonstrated significant benefit for NLFC in recognition of high-frequency speech sounds in children with moderate and mild hearing losses. In general, these benefits appeared to be increased for greater degrees of hearing loss (i.e., greater NLFC benefit was evident in children with moderate hearing loss compared with children with mild hearing loss). There are multiple reasons to explain why the children with cookie-bite audiograms in this study did not achieve the improved performance with NLFC seen in these previous studies. First, many of the children in the present study had already achieved ceiling-level performance without NLFC. Several of the measures used in this study have been shown to be sensitive to small differences that exist in performance between NLFC and wideband amplification, but it is not possible to show improvement with NLFC when participants score near 100% with the wideband frequency response on the nonadaptive tests used in this study. One test used adaptive procedures for measurement of both speech sound detection and recognition, but also showed no improvement or decrement with NLFC. Therefore, we can also consider that most of the children in this study had normal or near-normal hearing sensitivity in the high-frequency range. In fact, the mean audiometric threshold at 8000 Hz was approximately 20 db HL. Given that most participants in this study possessed normal to near-normal hearing sensitivity in the high Figure 7. Mean SNR for 50% correct performance (SNR-50) on the BKB-SIN Test by hearing aid condition. 1031

Journal of the American Academy of Audiology/Volume 25, Number 10, 2014 frequencies, it is not surprising that they achieved nearceiling or equivocal adaptive test performance for detection and recognition of high-frequency sounds of speech without NLFC. Finally, NLFC typically shifted highfrequency signals of interest from a region with normal to near-normal to a region with significant hearing loss. It is possible that poorer spectral and temporal resolution associated with greater hearing loss in the mid-frequency range may limit the benefit received by NLFC. The findings of this study do have clinical relevance, particularly in the context of findings of previous studies with NLFC in pediatric hearing aid fittings. Because NLFC did not seem to provide better detection of highfrequency sounds or improvement in speech recognition in quiet and in noise, the authors conclusion is that NLFC should be disabled in default fittings for children with cookie-bite audiograms. However, it is wise to consider the degree of high-frequency hearing loss when determining whether NLFC should be used. If a child has a cookie-bite audiogram with (for example) mid-frequency thresholds approximately 80 db HL and high-frequency thresholds roughly 60 db HL, then it is possible that the user s moderate high-frequency hearing loss may hinder the audibility of high-frequency speech sounds such as /s/. As a result, it may be necessary to use NLFC to overcome insufficient high-frequency audibility that is known to exist for children with moderate, high-frequency hearing loss (Wolfe et al, 2010; 2011). In the present study, audiometric configurations did not sample this type of fitting scenario. With that in mind, we recommend that NLFC not be ruled out as a fitting option for children who have cookie-bite audiograms or rising audiograms but also possess moderate, high-frequency hearing loss. The field of audiology is faced with the challenge of developing measures that may adequately determine whether NLFC is needed to optimize detection and recognition of high-frequency speech sounds. Indeed, many of the measures used in this study, such as detection of high-frequency speech sounds /s/, /sh/, etc., the UWO Plurals Test, the UWO-DFD nonsense word recognition test, and the PPT (a VCV nonsense word recognition test) may eventually serve as good tools for clinicians to use to determine whether NLFC is needed in the clinical setting and also to determine how NLFC parameters should be selected to optimize user benefit. The developers of these assessments have been diligently working to make these tests clinically available. It is recommended that clinicians consider the fitting strategies used in this study to fit hearing aids for children with cookie-bite audiograms. Probe microphone measures allow clinicians to observe the function of frequency-lowering amplification and also to determine whether high-frequency speech sounds are audible without NLFC. The clinicians may also choose to use many of the informal (e.g., asking the children to identify /s/ and /sh/, which was used in this study to ensure that there was no confusion between the /sh/ and /s/ phonemes with use of NLFC) and formal (e.g., UWO-DFD) measures we used in this study. Previous studies have indicated that these measures are able to identify instances in which the audibility of important speech sounds may be compromised without NLFC (Glista et al, 2009; Wolfe et al, 2010; 2011). Additional research is needed to determine more effective measures that may be used clinically to determine, on an individual basis, whether a user should be fitted with NLFC. CONCLUSIONS 1. Use of NLFC appears to provide no improvement or degradationinthedetectionoflow-level,high-frequency speech sounds and warble tones for children with cookie-bite audiograms relative to their performance with wideband amplification. 2. No differences in speech recognition in quiet or in noise were observed between NLFC on and wideband amplification conditions for children with cookie-bite audiograms. 3. No deterioration in performance was observed with the use of NLFC. 4. NLFC should be disabled by default for children who have cookie-bite audiograms and normal to near-normal hearing sensitivity in the high frequencies. 5. On the basis of previous research findings with children with flat and sloping audiograms, NLFC should still be considered for children who have cookie-bite configurations or rising audiograms and hearing thresholds in the mild to moderate range and poorer in the high-frequency range. 6. Clinicians should use contemporary fitting and verification procedures to optimize NLFC settings for children with cookie-bite hearing losses. 7. A need exists for the development of commercially available tests that will assist clinicians in determining whether NLFC is needed and, if so, how the parameters should be selected. REFERENCES Bench J, Kowal A, Bamford J. (1979) The BKB (Bamford-Kowal- Bench) sentence lists for partially-hearing children. Br J Audiol 13(3):108 112. Boretzki M, Kegel A. (2009) The Benefits of Nonlinear Frequency Compression for People with Mild Hearing Loss. Phonak AG: Phonak Field Study News. Cheesman MF, Jamieson DG. (1996) Development, evaluation and scoring of a nonsense word test suitable for use with speakers of Canadian English. Can Acoust 24(1):3 11. Dillon H. (2001) Hearing Aids: A Comprehensive Text. New York, NY: Boomerang Press and Thieme. Etymotic Research. (2005) Bamford-Kowal-Bench Speech-in-Noise Test (Version 1.03). IL: Elk Grove Village. 1032

NLFC for Cookie-Bite Audiograms/John et al Galster J, Valentine S, Dundas A, Fitz K. (2011) Spectral IQ: Audibly improving access to high-frequency sounds. Starkey White Paper, Starkey Laboratories, Inc., Eden Prairie, MN. Glista D, Scollie S. (2012) Development and evaluation of an English language measure of detection of word-final plurality markers: the University of Western Ontario Plurals Test. Am J Audiol 21(1):76 81. Glista D, Scollie S, Bagatto M, Seewald R, Parsa V, Johnson A. (2009) Evaluation of nonlinear frequency compression: clinical outcomes. Int J Audiol 48(9):632 644. Macrae JH, Dillon H. (1996) Gain, frequency response, and maximum output requirements for hearing aids. J Rehabil Res Dev 33(4):363 376. Pittman AL, Stelmachowicz PG. (2003) Hearing loss in children and adults: audiometric configuration, asymmetry, and progression. Ear Hear 24(3):198 205. Simpson A. (2009) Frequency-lowering devices for managing highfrequency hearing loss: a review. Trends Amplif 13(2):87 106. Simpson A, Hersbach AA, McDermott HJ. (2005) Improvements in speech perception with an experimental nonlinear frequency compression hearing device. Int J Audiol 44(5): 281 292. Simpson A, Hersbach AA, McDermott HJ. (2006) Frequencycompression outcomes in listeners with steeply sloping audiograms. Int J Audiol 45(11):619 629. Wolfe J, John A, Schafer E, et al. (2011) Long-term effects of nonlinear frequency compression for children with moderate hearing loss. Int J Audiol 50(6):396 404. Wolfe J, John A, Schafer E, Nyffeler M, Boretzki M, Caraway T. (2010) Evaluation of nonlinear frequency compression for schoolage children with moderate to moderately severe hearing loss. J Am Acad Audiol 21(10):618 628. Wolfe J, John AB, Schafer E, Hudson M, Woods W, Wheeler J, Hudgens K, Neumann S. (in press). Evaluation of wideband frequency responses and nonlinear frequency compression for children with mild to moderate high-frequency hearing loss. Int J Audiol. 1033