Speech production in noise with and without hearing protectiona) Jennifer B. Tuftsb) and Tom Frank Department of Communication Disorders, The Pennsylvania State University, University Park, Pennsylvania 16802
共Received 23 August 2002; revised 25 April 2003; accepted 19 May 2003兲 People working in noisy environments often complain of difficulty communicating when they wear hearing protection. It was hypothesized that part of the workers’ communication difficulties stem from changes in speech production that occur when hearing protectors are worn. To address this possibility, overall and one-third-octave-band SPL measurements were obtained for 16 men and 16 women as they produced connected speech while wearing foam, flange, or no earplugs 共open ears兲 in quiet and in pink noise at 60, 70, 80, 90, and 100 dB SPL. The attenuation and the occlusion effect produced by the earplugs were measured. The Speech Intelligibility Index 共SII兲 was also calculated for each condition. The talkers produced lower overall speech levels, speech-to-noise ratios, and SII values, and less high-frequency speech energy, when they wore earplugs compared with the open-ear condition. Small differences in the speech measures between the talkers wearing foam and flange earplugs were observed. Overall, the results of the study indicate that talkers wearing earplugs 共and consequently their listeners兲 are at a disadvantage when communicating in noise. © 2003 Acoustical Society of America. 关DOI: 10.1121/1.1592165兴 PACS numbers: 43.70.Gr, 43.50.Hg, 43.66.Vt 关DKW兴
I. INTRODUCTION
Among the many hazards that workers face on their jobs, exposure to excessive noise is one of the most pervasive. An estimated 30 million American workers are exposed to hazardous noise levels on their jobs 共NIOSH, 1996兲 and are at risk for developing noise-induced hearing loss. Hearing protection devices 共HPDs兲, such as earplugs or earmuffs, can reduce the danger of incurring noise-induced hearing loss if worn correctly and consistently. Wearing HPDs does not impair the ability of normal-hearing people to understand speech in noise levels above 85 dBA 共Berger, 1980; Abel et al., 1982; Nixon and Berger, 1991兲, and may even improve it 共Kryter, 1946; Chung and Gannon, 1979; Rink, 1979兲. However, workers often resist the use of HPDs because they claim that they have difficulty understanding speech while wearing them 共Royster and Holder, 1982; Helmkamp, 1986; Suter, 1992兲. A possible reason for this contradiction is that some workers may have pre-existing hearing loss, which, when combined with the attenuation of their HPDs, may make parts of the speech signal inaudible. In fact, speech understanding when wearing HPDs is often degraded for people with sensorineural hearing loss 共Lindeman, 1976; Chung and Gannon, 1979; Abel et al., 1982; Bauman and Marston, 1986兲. Another possible reason is that speech may be less intelligible if the talker is wearing HPDs, because wearing HPDs tends to induce changes in speech production 共Kryter, a兲
Portions of this work were presented at the annual meeting of the National Hearing Conservation Association, Dallas, TX, February 2002, and at the annual meeting of the American Academy of Audiology, Philadelphia, PA, April 2002. b兲 Current affiliation: Army Audiology & Speech Center, Walter Reed Army Medical Center, 6900 Georgia Ave. NW, Washington, DC 20307-5001. Electronic mail:
[email protected] J. Acoust. Soc. Am. 114 (2), August 2003
1946; Howell and Martin, 1975; Martin, Howell, and Lower, 1976兲. Kryter 共1946兲 reported that the intelligibility of male talkers reading monosyllabic word lists in a background of simulated engine room noise at levels of 75 to 105 dB SPL was slightly lower when they wore earplugs, compared with an open-ear condition. Further, the talkers lowered the level of their speech by 1 to 2 dB in all of the noise levels when they wore earplugs. Howell and Martin 共1975兲 measured the intelligibility of four male talkers reading monosyllabic word lists in a background of broadband noise at levels of 93 and 54 dB SPL while wearing earmuffs, earplugs, or no HPDs. In noise at 93 dB SPL, the percentage of phonemes correctly heard by listeners was 56% when the talkers were not wearing HPDs. This percentage decreased by 15% when the talkers wore earmuffs and by 26% when the talkers wore earplugs. In addition, the talkers lowered their speech levels by an average of 2.7 dB when they wore earmuffs and 4.2 dB when they wore earplugs. In noise at 54 dB SPL, all of the intelligibility scores approached 100%, although they were slightly lower when the talkers wore earplugs. At this noise level, the difference in the talkers’ speech levels was negligible 共0.3 dB兲 when the talkers wore earmuffs and was 1.8 dB lower with earplugs, compared with the no-HPD condition. In a follow-up study, Martin et al. 共1976兲 recorded overall and octave-band speech levels and measured the intelligibility of eight male talkers reading monosyllabic word lists while wearing earplugs, earmuffs, or no HPDs in a background of broadband noise presented at 67, 77, 87, and 95 dBA. They found that the percentage of phonemes correctly understood by listeners decreased when the talkers wore HPDs, particularly if the listeners also wore HPDs. Further, the overall speech level of the talkers was 2 to 3 dB lower
0001-4966/2003/114(2)/1069/12/$19.00
© 2003 Acoustical Society of America
1069
when they wore HPDs compared with the open-ear condition. The octave-band analysis revealed no significant changes in the talkers’ speech spectra when wearing HPDs other than would be expected from the change in their speech level. Martin et al. 共1976兲 surmised that when the talkers wore HPDs, changes in their voice quality might have occurred that were not reflected in the octave-band speech spectrum but that might have been detrimental to intelligibility. To evaluate this possibility, the recordings of the eight male talkers reading the word lists with and without earplugs in the background noise at 87 dBA were mixed with noise at a ⫹3-dB speech-to-noise ratio 共SNR兲 and presented to four male listeners at approximately 70 dB SPL. Any differences in intelligibility between speech produced with and without HPDs would therefore be due to factors other than speech level and SNR of the presentation. For seven of the eight talkers, no differences were observed in the intelligibility of their speech. Thus, the difference in intelligibility between talkers wearing and not wearing HPDs largely disappeared when speech level and SNR were controlled. Previous studies, including those just described, have shown that talkers lower the level of their voices in noise by 1 to 6 dB on average when they wear HPDs compared with an open-ear condition 共Kryter, 1946; Howell and Martin, 1975; Martin et al., 1976; Hormann et al., 1984; Casali, Horylev, and Grenell, 1987兲. Two of these studies 共Kryter, 1946; Casali et al., 1987兲 also reported that talkers raise their voices by approximately 4 dB when they speak in quiet while wearing HPDs, perhaps to compensate for the attenuation of the air-conduction component of their speech. Two explanations have been offered as to why people speak more softly when wearing HPDs in noise. First, complete occlusion of the ear canal improves bone-conduction hearing at frequencies below 2000 Hz. Due to the occlusion effect created by wearing HPDs, talkers’ voices will seem louder to them, and they may lower the level of their voices in response 共Howell and Martin, 1975; Martin et al., 1976; Casali et al., 1987; Suter, 1992兲. Second, since HPDs attenuate ambient noise, talkers wearing HPDs may lower their voices in response to the decrease in the perceived ambient noise level 共Kryter, 1946; Lindeman, 1976; Foreshaw and Cruchley, 1982兲. Aside from the changes in overall level, little is known about the speech of talkers wearing HPDs. In general, when speech is produced in a background of noise, without HPDs, several of its acoustic and prosodic characteristics 共e.g., overall level, fundamental frequency, formant frequencies, and spectral composition兲 change relative to speech produced at a normal conversational level in quiet. For example, for every 10-dB increase in the background noise level, talkers raise their voices by 1 to 6 dB, a phenomenon known as the Lombard effect 共Lane and Tranel, 1971兲. In addition, talkers produce higher fundamental frequencies when they speak in a background of noise 共Summers et al., 1988; Bond, Moore, and Gable, 1989; Junqua, 1993; Letowski, Frank, and Caravella, 1993兲. The frequencies of the first two formants also change, but in less consistent ways. 共Summers et al., 1988; Bond et al., 1989; Junqua, 1993; Tartter, Gomes, and Litwin, 1993兲. In general, relatively more high1070
J. Acoust. Soc. Am., Vol. 114, No. 2, August 2003
frequency energy is contained in speech that is produced in noise versus speech that is produced in quiet 共Webster and Klumpp, 1962; Summers et al., 1988; Junqua, 1993; Tartter et al., 1993; Pittman and Wiley, 2001兲. The spectral and prosodic differences that occur when talkers produce speech in noise and when they produce speech at greater vocal efforts in quiet are similar 共Bond et al., 1989兲. However, Letowski et al. 共1993兲 found that speech produced in noise contains more high-frequency energy than loud speech produced in quiet. Perhaps due in part to these changes, several investigators have found that speech produced in noise is more intelligible than speech produced in quiet, when both types of speech are presented in noise at the same SNR 共Dreher and O’Neill, 1957; Summers et al., 1988; Pittman and Wiley, 2001兲. Summers et al. 共1988兲 also reported that the relative improvement in the intelligibility of speech produced in noise increases as the SNR decreases. Most of the research concerning the speech of talkers wearing HPDs in noise has been limited to measurements of the overall SPLs, as described previously. Further, these studies were typically carried out in background noise levels that were restricted in either the number of discrete levels presented or the range they encompassed. In most cases, the speech stimuli were monosyllabic words, which are not representative of most communication situations. Further information about the speech of talkers wearing HPDs in noise may provide a basis for evaluating the validity of workers’ complaints and for selecting HPDs for particular work environments. The main purpose of the present study was to obtain overall and one-third-octave-band SPL measurements of connected speech produced by men and women wearing two types of earplugs and wearing no earplugs 共open ears兲. Measurements were made in quiet and in several noise levels. Differences between the two types of earplugs in terms of their effect on speech production were also assessed. Additionally, the Speech Intelligibility Index 共SII兲 共ANSI, 1997兲 was used to determine if the amount of speech information available to a hypothetical listener changed as a function of background noise level and whether or not the talkers were wearing earplugs. The SII is especially useful for evaluating communication systems when it is not practical to use actual listeners for that purpose, and it has previously been used as a means of ranking HPDs for maximum speech communication ability in noise 共Williams and Michael, 1991兲. II. METHOD A. Subjects, HPDs, and test environment
Thirty-two subjects 共16 males, 16 females兲 ranging in age from 19 to 35 years 共mean⫽25.4 years, s.d.⫽4.9 years兲 were enrolled in the study and received $20.00 for their participation. Each subject was a fluent speaker of American English, and had normal air-conduction thresholds 共⭐20 dB HL from 0.25 to 8 kHz re: ANSI S3.6-1996兲 and a normal tympanogram 共Margolis and Heller, 1987兲 in each ear. For part of the test session, each subject wore one of two types of insert HPDs commonly used in industry. One type was a J. B. Tufts and T. Frank: Speech production and hearing protection
user-molded foam earplug 关EAR Classic兴 and the other was a premolded triple flange earplug 关Bilsom 556兴. All testing was conducted in an 18⫻18⫻6-ft. sound-treated room having ambient noise levels suitable for ears-open testing from 0.125 to 8 kHz 共re: ANSI S3.1-1991兲. B. Stimulus characteristics
1. Speech passages
During the test session, each subject was required to read twelve speech passages from the Speech Intelligibility Rating 共SIR兲 test 共McDaniel and Cox, 1992兲. The 12 passages were chosen from the 20 passages of the SIR test by measuring the average one-third-octave-band spectra of the commercially available CD recordings of all 20 passages, computing the slopes of regression lines fitted to the spectra, and then selecting those passages whose slopes fell within 1 standard deviation of the mean. Each passage consisted of approximately 110 words on a familiar topic, and all were equated on the basis of reading level, vocabulary, and sentence structure. 2. Background noise
Background noise was presented to the subjects via TDH-49 earphones mounted in circumaural enclosures as they read the speech passages. The background noise was a CD recording of pink noise having equal power within each one-third-octave band, as measured in an IEC 318 artificial ear 关Bruel & Kjaer, 4153兴. Pink noise was chosen because its spectral level decreases with increasing frequency, similar to many industrial noises. To create a flat one-third octave-band spectrum 共i.e., pink noise兲 as measured in the artificial ear, the following steps were taken. Initially, a CD recording of pink noise was created by looping a pink-noise wave file and writing the wave file to a CD using an editing program 关SOUND FORGE, v4.5兴. The pink-noise CD was then played from a CD player 关Optimus, CD-1850兴 to the right channel input of an audiometer 关Madsen, OB 822兴. The audiometer output was split and directed to two TDH-49 earphones, located in the test booth. Each TDH-49 earphone was mounted in a circumaural enclosure having an oval cushion 关David Clark, H10– 40兴. Next, each earphone was placed on a flat plate adaptor 关Bruel & Kjaer type 1, DB0843兴 and loaded with a static force of 10 N. The flat plate adaptor was seated on the IEC 318 artificial ear, which was equipped with a 21-in. microphone 关GRAS, 40AF兴 and preamplifier 关01 dB, PRE 12H兴 that was connected to a real-time spectrum analyzer 关01 dB Symphonie兴. The audiometer output was adjusted to 100 dB SPL as measured in the IEC 318 coupler. Once this was done, the one-third-octave-band spectrum of the pink noise was averaged over a 1-min period, with samples taken once every 10 ms. Using the measured spectrum as a guide, the spectrum of the original pink-noise wave file was modified using the one-third-octave-band graphic equalizer in the SOUND FORGE editing program, and then rerecorded onto a new CD. This procedure was repeated until the one-thirdoctave-band spectrum measured in the coupler was essentially flat 共⫾1 dB from 0.16 to 2.5 kHz; ⫾3 dB from 3.15 to J. Acoust. Soc. Am., Vol. 114, No. 2, August 2003
5 kHz; ⫾1 dB from 6.3 to 8 kHz兲. During data collection, the flat-spectrum pink-noise CD was played from the CD player to the audiometer and then to the circumaural earphones at 60, 70, 80, 90, and 100 dB SPL as measured in the IEC 318 coupler. The coupler SPLs remained stable throughout data collection 共⫾0.5 dB兲, as did the linearity of the audiometer output from 60 to 100 dB SPL 共⫾0.5 dB兲. The difference in the coupler output between the two earphones was ⬍1 dB at each one-third-octave-band from 0.16 to 8 kHz. C. Measurements
1. Speech
As each subject read the SIR test passages, his or her overall and one-third-octave-band speech levels were measured with a type 1 sound-level meter 共SLM兲 关Casella CEL, 593兴 mounted on an adjustable tripod and equipped with a 1 2-in. omnidirectional sound-field microphone 关Casella CE, 3122兴 having a flat response 共⫾1 dB兲 from 0.05 to 12 kHz. The SLM was interfaced with a laboratory computer 关Dell, Dimension 4100兴 in an adjacent control room. A commercial software program 关Casella CEL, 6695兴 was used for remotely controlling the SLM and for downloading and storing each subject’s speech measurements. The SLM was programmed to average each subject’s overall and one-thirdoctave-band speech levels from 0.08 to 20 kHz, using a fast response time 共0.125 ms兲, a sampling rate of 75.6 kHz, and a dynamic range of 25–100 dB SPL. The SLM was calibrated with an acoustical calibrator 关Casella CEL, 284/2兴 and remained stable throughout data collection 共⫾0.1 dB兲. 2. Earplug attenuation
The attenuation provided by the earplugs was measured with a Fitcheck measurement system 关Michael & Associates, 700兴 consisting of both hardware and software. The hardware consisted of a 40-dB fixed attenuator and a 70-dB variable attenuator, a subject response switch, and circumaural earphones, which were the same earphones used to transduce the pink noise. The software was used to control the attenuator settings and the test signals, record the subject’s responses, and perform calculations. The Fitcheck unit was connected to the laboratory computer located in the control room, while the response switch and the circumaural earphones were located in the test booth. The software instructed the sound card of the computer to generate onethird-octave-bands of noise centered at 0.25, 0.5, 1, 2, and 4 kHz, which were directed to the Fitcheck unit and then to the circumaural earphones. The software also determined the status of the response switch, increasing the attenuation of the one-third-octave-band noise test signals when the response switch was pressed, and decreasing the attenuation when the response switch was released. The variable attenuator had a step size of 1.5 dB and an attenuation rate of 3 dB/second. Each subject’s one-third-octave-band noise 共1/3 OBN兲 thresholds were obtained for each center frequency while he or she wore the circumaural earphones without the earplugs 共open-ear 1/3-OBN thresholds兲 and with the earplugs 共closed-ear 1/3-OBN thresholds兲 using the Fitcheck automated threshold-tracing procedure, which was controlled by
J. B. Tufts and T. Frank: Speech production and hearing protection
1071
the software. This was done by having the subject continuously press the response switch whenever the 1/3-OBN test signal was audible and release the switch whenever the signal was inaudible. For each 1/3-OBN test signal, eight threshold reversals 共four peaks and four valleys兲 were obtained. The subject’s threshold for each 1/3-OBN test signal was calculated automatically by the software as the midpoint between the average of the peaks and the average of the valleys, with the first two reversals ignored. The output of the Fitcheck measurement system was calibrated by adjusting its variable attenuator to an arbitrary level at each 1/3-OBN center frequency, and measuring the output voltage at the earphone terminals 共with the earphones in line兲 using a Hewlett-Packard 400GL voltmeter. The output voltage remained stable 共⫾1.3 dB兲 throughout data collection. 3. Occlusion effect
Pure-tone bone-conduction thresholds 共BCTs兲 at frequencies of 0.25, 0.5, and 1 kHz were obtained for each subject. The BCTs were measured using forehead placement of a Radioear B-71 bone vibrator. The pure tones were generated by the audiometer 关Madsen, OB 822兴 and directed to a 20-dB fixed-pad attenuator in the test booth and then to the bone vibrator. The bone vibrator was fitted in the center of each subject’s forehead and held in place with a headband adjusted to produce ⬎300 grams of force as measured by a strain gauge 关Abbott, 4580兴. The output of the bone vibrator was calibrated by setting the audiometer output to an arbitrary level at each test frequency, and measuring the output voltage at the terminal of the bone vibrator 共with the bone vibrator in line兲 using a Hewlett-Packard 400 GL voltmeter. The output voltage remained stable 共⫾0.7 dB兲 throughout data collection. During the test session, BCTs were measured in several conditions: with no earplugs or earphones worn, with the earphones only, with the earplugs only, and with the earplugs and earphones worn. This was done to determine the magnitude of the occlusion effect due to the earphones alone, the earplugs alone, the earplugs and earphones together, and the earplugs and earphones relative to the earphone-only condition. Of particular interest was the latter measurement. Since the subjects were wearing earphones in the control condition and earplugs and earphones in the experimental condition, the difference between the BCTs at each test frequency in these two conditions represented the magnitude of the occlusion effect experienced by the subjects when they wore earplugs relative to the control condition. D. Procedure
Sixteen subjects were tested with the foam earplugs and 16 subjects were tested with the flange earplugs. Within each earplug group 共foam or flange兲, speech measurements were obtained for eight subjects 共four males, four females兲 first while wearing the earplugs and then without the earplugs. The other eight subjects 共four males, four females兲 were tested first without the earplugs and then while wearing the earplugs. Each subject received a randomized order for the 1072
J. Acoust. Soc. Am., Vol. 114, No. 2, August 2003
12 reading passages, a randomized order for the six noise conditions 共quiet, 60, 70, 80, 90, and 100 dB SPL兲 for testing with the earplugs, and a different randomized order for the six noise conditions for testing without the earplugs. Prior to the start of each subject’s test session, the subject was seated in the test booth and the microphone of the SLM was adjusted to be level with and 1 m away from his or her mouth at 0° azimuth. The face of the microphone was positioned at 90° relative to the subject’s lips so that his or her voice reached the microphone with a grazing incidence. The procedures carried out during data collection are described below. The order in which they are listed corresponds to the order of procedures for the 16 subjects who were tested first with no earplugs, then with earplugs. 共1兲 Without earplugs, speech in quiet and in five levels of pink noise: The subject was required to read several of the speech passages while different levels of noise 共60, 70, 80, 90, or 100 dB SPL兲, or quiet, were presented over the circumaural earphones. One noise level or quiet was presented per passage. The subject was instructed to read each passage in such a way that a conversation partner listening in noise at the same level would be able to understand what was said. Further, the subject was instructed to read continuously until told to stop, even if he or she made a mistake or had to begin reading the passage again. Once told to stop, the subject was instructed to turn to the next reading passage and wait for the next noise level to be presented. The subject was allowed to look over each new passage and, if necessary, take a drink of water before continuing. For each speech passage, the investigator adjusted the audiometer to provide the appropriate level of pink noise, or quiet, according to the randomized order for that subject. As soon as the subject began reading each passage, the SLM was activated to record his or her overall and one-third-octave-band speech levels for a period of 45 s. This procedure was repeated until all of the noise level conditions were completed. 共2兲 Open-ear 1/3-OBN thresholds: The subject’s binaural open-ear 1/3-OBN thresholds were measured with the Fitcheck system at the center frequencies of 0.25, 0.5, 1, 2, and 4 kHz, in ascending order by frequency. 共3兲 Insertion of the earplugs: Following a demonstration by the investigator, the subject inserted the earplugs. If the earplugs appeared to be poorly inserted 共e.g., most of the earplug was visible outside the ear canal兲, the investigator asked the subject to remove and reinsert them. 共4兲 With earplugs, speech in quiet and in five levels of pink noise: The instructions for reading the speech passages were repeated. The speech measurements were obtained in quiet and in noise levels of 60, 70, 80, 90, and 100 dB, with the subject wearing the earplugs in addition to the circumaural earphones. 共5兲 Closed-ear 1/3-OBN thresholds: The subject’s closed-ear 1/3-OBN thresholds were obtained with the Fitcheck system. 共6兲 BCTs: The subject’s BCTs were measured at 1, 0.5, and 0.25 kHz, in descending order by frequency. Threshold J. B. Tufts and T. Frank: Speech production and hearing protection
at each test frequency was defined as the mean of four ascending thresholds using a 2-dB step. The BCTs were measured in four conditions: while wearing both the circumaural earphones and earplugs, while wearing the earplugs without the earphones, while wearing neither the earplugs nor the earphones, and while wearing the earphones only. The order of procedures was reversed for the 16 subjects who were tested first with earplugs, then without earplugs, except that this group of subjects inserted the earplugs halfway through the first procedure 共i.e., step 6兲. 1. Dependent measures
Three types of dependent measures were obtained or calculated for each talker in each background noise level and in each ear condition 共with versus without earplugs兲. These included 共1兲 two measures of speech level; 共2兲 one measure that described the speech spectrum; and 共3兲 one measure of predicted intelligibility. The first measure of speech level was the overall speech level 共in dB SPL兲. The second was the speech-to-noise ratio 共SNR; in dB兲. The SNR was defined as the overall speech level minus the overall noise level presented to the subject. Since the speech and the noise were measured under different conditions 共i.e., the speech was measured in the sound field, and the noise was measured in an IEC 318 artificial ear兲, the difference between the two is not, strictly speaking, a true SNR. However, due to the necessity of measuring the speech without contamination by the background noise, the SNR was calculated in this way. The measure that described the speech spectrum was the spectral center of gravity 共SCoG; in Hz兲. This was defined as the center frequency of the one-third-octave-band between 0.16 and 8 kHz, inclusive, containing the frequency above and below which there were equal amounts of power in the speech signal. The SCoG was determined by successively adding the power in each consecutive one-third-octaveband, in order of increasing center frequency, and then choosing the one-third octave band whose addition caused the running sum to first exceed one-half of the total power in the signal. Last, the SII was calculated assuming a hypothetical listener with open ears 共i.e., without HPDs兲 and normal hearing threshold levels, and using the one-third-octave-band procedure and the frequency importance function for continuous discourse described in ANSI S3.5-1997. In essence, an SNR was calculated in each one-third-octave-band and then weighted by the frequency importance function for continuous discourse. These weighted SNRs were summed to produce a number giving the proportion of speech information available to a hypothetical listener. The SII ranges from a minimum of 0.0 to a maximum of 1.0. III. RESULTS
Three separate analyses were conducted on the data. The first analysis was done to examine the differences between the ear conditions 共with versus without earplugs兲. For this analysis, the data were collapsed over earplug type 共foam and flange兲. A four-way mixed-factors ANOVA was conJ. Acoust. Soc. Am., Vol. 114, No. 2, August 2003
FIG. 1. Mean overall speech levels 共in dB SPL兲 and standard errors as a function of noise level, with open ears and with foam and flange earplugs.
ducted on each dependent measure, with background noise level and ear condition as within-subjects factors, and testing order and sex as between-subjects factors. The Huynh–Feldt epsilon was used to correct the probabilities for the main effect of background noise level and its interaction with ear condition, due to the violation of the sphericity assumption 共Huynh and Feldt, 1976兲. The second analysis examined the differences between the earplug groups 共foam versus flange兲. Using only the data obtained when the earplugs were worn, a four-way mixedfactors ANOVA was conducted on each dependent measure, with background noise level as the within-subjects factor, and earplug group, testing order, and sex as the betweensubjects factors. As before and where applicable, the probabilities were corrected for the violation of sphericity using the Huynh–Feldt epsilon. The third analysis was conducted to ensure that the two earplug groups were similar with respect to the dependent measures when the earplugs were not worn. For each dependent measure, a four-way mixed-factors ANOVA was conducted on the data obtained in the open-ear condition. For each ANOVA, the main effect of earplug group and all of its interactions were nonsignificant (p⬎0.05). Thus, any differences between the earplug groups when the earplugs were worn could be attributed to the earplug type. A. Overall speech level and SNR
Figures 1 and 2 illustrate the mean overall speech levels and SNRs, respectively, and their standard errors, for the talkers with and without earplugs in quiet 共overall speech level only兲 and in each background noise level. In quiet, the mean overall speech levels were very similar between ear conditions 共67.5 dB SPL without earplugs versus 66.6 and 67.2 dB SPL for the foam and flange earplug groups, respectively兲. In the presence of background noise, the talkers raised the level of their voices. However, the increase in the overall speech level did not match the increase in the background noise level. That is, although the background noise level increased 40 dB from 60 to 100 dB SPL, the overall
J. B. Tufts and T. Frank: Speech production and hearing protection
1073
FIG. 2. Mean SNRs 共in dB兲 and standard errors as a function of noise level, with open ears and with foam and flange earplugs.
speech level only increased by 5 to 13 dB, depending on the ear condition. Therefore, the SNR decreased as the background noise level increased. When the talkers wore earplugs, the mean overall speech level and SNR were consistently lower in each background noise level compared with the open-ear condition. In fact, the mean differences in both the overall speech levels and the SNRs between the two ear conditions increased from 4 to 11 dB as the background noise level increased from 60 to 100 dB. In background noise at 100 dB SPL, the overall speech levels were 84.4 dB for the talkers with open ears, and 71.9 and 74.3 dB for the foam and flange earplug groups, respectively. The corresponding SNRs were ⫺15.6, ⫺28.1, and ⫺25.7 dB. For the first analysis on the overall speech level 共with versus without earplugs兲, the main effects of background noise level and ear condition, and their interaction, were statistically significant 关noise level: F(5,140)⫽326.517, p ⬍0.001; ear condition: F(1,28)⫽226.956, p⬍0.001; interaction: F(5,140)⫽106.826, p⬍0.001]. In addition, the interaction of background noise level and testing order was significant 关 F(5,140)⫽4.327, p⫽0.015]. For the second analysis 共foam versus flange earplug groups兲, the main effects of background noise level and earplug group, and their interaction, were significant 关noise level: F(5,120) ⫽119.726, p⬍0.001; earplug group: F(1,24)⫽10.757, p ⫽0.003; interaction: F(5,120)⫽4.389, p⫽0.008]. The interaction of background noise level and testing order was also significant 关 F(5,120)⫽7.398, p⬍0.001]. Similarly, for the first analysis on the SNR, the main effects of background noise level and ear condition, and their interaction, were significant 关noise level: F(4,112) ⫽3041.646, p⬍0.001; ear condition: F(1,28)⫽4965.188, p⬍0.001; interaction: F(4,112)⫽60.568, p⬍0.001]. Further, the three-way interaction of background noise level, ear condition, and sex was significant 关 F(4,112)⫽2.776, p ⫽0.034]. For the second analysis, the main effects of background noise level and earplug group, and their interaction, were significant 关noise level: F(4,96)⫽3478.527, p ⬍0.001; earplug group: F(1,24)⫽11.620, p⫽0.002; inter1074
J. Acoust. Soc. Am., Vol. 114, No. 2, August 2003
FIG. 3. Mean spectral centers of gravity 共in Hz兲 and standard errors as a function of noise level, with open ears and with foam and flange earplugs.
action: F(4,96)⫽3.642, p⫽0.023]. In addition, the interaction of background noise level and testing order was significant 关 F(4,96)⫽5.487, p⫽0.003]. These results indicate that the level of a talker’s voice 共and consequently the SNR兲 is significantly lower on average when earplugs are worn in the presence of background noise compared with an open-ear condition. In addition, the foam earplug group had slightly lower overall speech levels and SNRs than the flange earplug group. However, it should be noted that the increases in overall speech level in the presence of background noise were not sufficient to maintain a favorable SNR either with or without earplugs.
B. Spectral center of gravity
Figure 3 illustrates the mean frequencies of the spectral center of gravity 共SCoG兲 and their standard errors, for the talkers with and without earplugs, in quiet and in each background noise level. In quiet, the SCoGs were very similar between ear conditions 共475 Hz for the talkers without earplugs versus 414 and 491 Hz for the talkers with foam and flange earplugs, respectively兲. In the presence of background noise, the SCoG increased in frequency. However, the frequency of the SCoG was consistently lower in each background noise level when the talkers wore earplugs. The differences between the two ear conditions increased as the noise level increased. In background noise at 100 dB SPL, the mean SCoGs were 972 Hz for the talkers without earplugs and 624 and 756 Hz for the talkers with foam and flange earplugs, respectively. The effect on the frequency spectrum of an increase in the SCoG is shown more clearly in Fig. 4, which illustrates the one-third-octave-band speech spectra for 共A兲 men and 共B兲 women, respectively, in quiet and in background noise at 100 dB SPL, with and without earplugs. Inspection of Figs. 4共A兲 and 共B兲 reveals that the speech spectra in quiet were very similar in level and shape between ear conditions. In background noise at 100 dB SPL, the speech spectra changed in both level and shape, reflecting the increase in the overall speech level as well as the shift of speech energy to the J. B. Tufts and T. Frank: Speech production and hearing protection
FIG. 4. The one-third-octave-band speech spectra of 共A兲 male and 共B兲 female talkers with and without earplugs in quiet and in background noise at 100 dB SPL. The parameter is background noise level.
higher frequencies. However, these changes were very different in magnitude between the ear conditions. For both men and women, the average speech spectrum of the talkers without earplugs was higher in level and showed a greater shift of energy to the higher frequencies than the average speech spectrum obtained when the talkers wore earplugs. In general, women had more high-frequency energy in their speech than men. For example, the SCoGs in quiet, without earplugs, were 520 Hz for the women and 430 Hz for the men. In background noise at 100 dB SPL, without earplugs, the corresponding SCoGs were 1200 and 743 Hz. This finding was not unexpected, given that women have generally higher fundamental and formant frequencies than men. For the first data analysis 共with versus without earplugs兲, the main effects on the SCoG of background noise level and ear condition, and their interaction, were significant 关noise level: F(5,140)⫽73.438, p⬍0.001; ear condition: F(1,28) ⫽64.652, p⬍0.001; interaction: F(5,140)⫽9.407, p ⬍0.001]. The main effect of sex and its interaction with J. Acoust. Soc. Am., Vol. 114, No. 2, August 2003
background noise level were also significant 关sex: F(1,28) ⫽51.221, p⬍0.001; interaction: F(5,140)⫽11.156, p ⬍0.001]. For the second analysis 共foam versus flange earplug groups兲, the main effects of background noise level and earplug group were significant 关noise level: F(5,120) ⫽33.853, p⬍0.001; earplug group: F(1,24)⫽17.866, p ⬍0.001]. In addition, the main effect of sex, its two-way interaction with background noise level, and its three-way interaction with background noise level and earplug group were significant 关 F(1,24)⫽44.994, p⬍0.001; two-way interaction: F(5,120)⫽6.006, p⬍0.001; three-way interaction: F(5,120)⫽2.370, p⫽0.043]. The two-way interaction of background noise level and testing order was also significant 关 F(5,120)⫽4.746, p⫽0.001]. These results indicate that the talkers produced relatively more high-frequency energy in their speech as the background noise level increased. However, the increase in highfrequency energy was not as pronounced when the talkers wore earplugs, particularly for the foam earplug group. In
J. B. Tufts and T. Frank: Speech production and hearing protection
1075
FIG. 5. Mean Speech Intelligibility Index values and standard errors as a function of noise level, with open ears and with foam and flange earplugs. Note that the three symbols overlap in the ‘‘quiet’’ condition.
FIG. 6. Mean attenuation values 共in dB兲 and standard errors as a function of one-third octave-band center frequency for the foam and flange earplugs. The mean PAR is also shown for each earplug type.
quiet and in background noise, women produced more highfrequency energy in their speech than men.
background noise levels 共60 to 90 dB SPL兲. The SII was consistently lower in each background noise level for the foam earplug group compared with the flange earplug group, particularly in moderate background noise levels 共60 to 80 dB SPL兲. The differences in the SII values between the ear conditions and between the earplug groups were smaller in the higher background noise levels than in the more moderate background noise levels. This occurred because in the calculation of the SII, it is assumed that no speech information is received by a listener in a particular frequency band if the SNR in that band is less than ⫺15 dB. Thus, in the higher background noise levels, where the SNRs were extremely poor, the differences between the groups were effectively ignored.
C. Speech Intelligibility Index „SII…
Figure 5 shows the mean SII values and standard errors for the talkers with and without earplugs, in quiet and in each background noise level. In quiet, the talkers achieved a mean SII of 1.00 both with and without earplugs 共0.99 and 1.00 for the talkers in the foam and flange earplug groups, respectively兲. The SII decreased to nearly zero in the higher background noise levels both with and without earplugs. In the background noise levels of 60, 70, 80, and 90 dB, the SII was consistently lower when the talkers wore earplugs. Because the SII data were less variable at each end of their range than in the middle, an arcsin transformation was applied before the analyses were conducted. For the first analysis 共with versus without earplugs兲, the main effects of background noise level and ear condition, and their interaction, were significant 关noise level: F(5,140)⫽2546.759, p ⬍0.001; ear condition: F(1,28)⫽212.351, p⬍0.001; interaction: F(5,140)⫽91.338, p⬍0.001]. In addition, the twoway interaction of background noise level and sex was significant 关 F(5,140)⫽3.966; p⫽0.002], as well as the threeway interaction of background noise level, ear condition, and testing order 关 F(5,140)⫽2.776, p⫽0.029]. For the second analysis 共foam versus flange earplug groups兲, the main effects of background noise level and earplug group were significant 关noise level: F(5,120)⫽1945.291, p⬍0.001; earplug group: F(1,24)⫽10.690, p⫽0.003]. The two-way interactions of background noise level with testing order and with sex were also significant 关noise level and testing order: F(5,120)⫽3.822, p⫽0.004; noise level and sex: F(5,120) ⫽3.626, p⫽0.006]. These results indicate that less and less speech information was available from the talkers as the background noise level increased. Further, when the talkers wore earplugs, the amount of available speech information was reduced relative to the open-ear condition, particularly in moderate to high 1076
J. Acoust. Soc. Am., Vol. 114, No. 2, August 2003
D. Attenuation and occlusion effect
Figure 6 shows the mean attenuation values and their standard errors at each one-third-octave-band center frequency for the two earplug types. The mean attenuation characteristics of the two earplug types were typical of passive insert-type HPDs, with greater attenuation of the higher frequencies 共⬎2 kHz兲 than of the lower frequencies 共⬍1 kHz兲. Also shown in the figure is the mean Personal Attenuation Rating 共PAR; Michael, 1999兲 of each earplug type and the associated standard errors. The PAR is a single-number rating of the attenuation of HPDs similar in derivation to the Noise Reduction Rating 共NRR兲. The mean PARs were 22.7 dB for the flange earplugs and 29.8 dB for the foam earplugs. 共The published NRRs for these earplugs are 27 and 29 dB, respectively.兲 Collapsed over earplug type, the mean values of the occlusion effect experienced by the subjects when they wore earplugs relative to the control 共earphone-only兲 condition at 0.25, 0.5, and 1 kHz were 2.8, 7.4, and 4.6 dB, respectively.1 The magnitude of the occlusion effect was similar between the two earplug types 共⭐1.5-dB mean difference at each test frequency兲. J. B. Tufts and T. Frank: Speech production and hearing protection
IV. DISCUSSION A. The effects of wearing earplugs on speech production
Speech communication is a two-way process involving a talker and a listener. To understand how HPDs affect that process, it is important to assess their effects on both the talker and the listener. However, much of the research to date on the effects of HPDs on speech communication has focused on the listener. These studies have generally involved a comparison of the ability to understand speech in a background of noise with open ears and with ears occluded by HPDs. In contrast, the purpose of this study was to assess the effects of HPDs on speech production, viz., the overall level and spectral characteristics of speech. Several speech measures were obtained for talkers with and without earplugs in quiet and in a background of pink noise presented at several levels. In quiet, occluding the talkers’ ear canals with earplugs resulted in a slight decrease of 0.6 dB in the mean overall speech level. Navarro 共1996兲 reported a similar decrease of 0.7 dB for talkers wearing EAR foam earplugs in quiet. These two findings suggest that talkers do not significantly alter the level of their voices when they wear earplugs in quiet. This conclusion contrasts with that reached by Kryter 共1946兲 and Casali et al. 共1987兲, who reported that talkers raise the level of their voices by approximately 4 dB when they wear HPDs in quiet. It is possible that, for the talkers in this study 关and perhaps the Navarro 共1996兲 study兴, the effects of the attenuation of the air-conduction component and the enhancement of the bone-conduction component of their speech offset one another. This may have occurred if the earplugs provided an acoustic seal but were not deeply inserted.2 In the presence of background noise, the talkers raised the level of their voices. However, depending on the noise level, the mean overall speech levels of the talkers in background noise were 4 to 11 dB lower when they wore earplugs compared with the open-ear condition. These differences were generally larger than the differences of 1 to 6 dB reported in previous research 共Kryter, 1946; Howell and Martin, 1975; Martin et al., 1976; Hormann et al., 1984; Casali et al., 1987兲. The variation across studies may be due to different amounts of HPD attenuation, different test room conditions 共e.g., reverberant vs anechoic兲, and different methods of measuring the speech level. Differences in the background noises presented to talkers may have also affected the speech levels. Casali et al. 共1987兲 showed that talkers wearing earmuffs in background noise having a highfrequency bias produced lower speech levels than talkers in background noise having a low-frequency bias. This may have occurred because the earmuffs attenuated the highfrequency noise to a greater extent than the low-frequency noise, and/or because the high-frequency noise provided less masking of the speech signal. The pink noise used in this study likely resulted in speech measures more representative of an environment with low-frequency, steady-state noise than other kinds of noise 共e.g., intermittent noise or noise with a high-frequency emphasis兲. Despite the methodologiJ. Acoust. Soc. Am., Vol. 114, No. 2, August 2003
cal variations, the talkers in this and previous studies reliably lowered the level of their voices while wearing HPDs in noise, although the magnitude of the effect varied across studies. As the background noise level increased, the amount of high-frequency energy in the talkers’ speech increased, as evidenced by the upward shift in the SCoG. This spectral change is consistent with previous research on speech produced in noise 共Webster and Klumpp, 1962; Summers et al., 1988; Junqua, 1993; Tartter et al., 1993; Pittman and Wiley, 2001兲 and speech produced with greater vocal effort in quiet 共Lazarus, 1990; Traunmu¨ller and Eriksson, 2000兲. Further, the SCoGs of the male talkers were lower in frequency than those of the female talkers in quiet and in each background noise level, both with and without earplugs. This finding was expected, given that men produce more low-frequency speech energy on average due to their lower fundamental frequencies. No other consistent differences between men and women were found for any other speech measure. When the talkers wore earplugs, the mean SCoG decreased in frequency by 147 to 327 Hz compared with the open-ear condition, depending on the background noise level. Thus, in a given background noise level, the talkers produced less highfrequency speech energy when they wore earplugs. In general, the increase in high-frequency energy that occurs when speech is produced in a background of noise may be beneficial for 共normal-hearing兲 listeners. Several investigators have found that speech produced in noise is more intelligible than speech produced in quiet, when both types of speech are mixed with noise at the same SNR 共Dreher and O’Neill, 1957; Summers et al., 1988; Pittman and Wiley, 2001兲. Apparently, the changes that occur in the frequency spectrum serve to increase the intelligibility of speech produced in noise relative to speech produced in quiet, when level and SNR are controlled. Thus, talkers wearing earplugs 共and consequently their listeners兲 may be at an additional disadvantage, since the shift of speech energy to the higher frequencies is not as pronounced for them. Although the talkers in this study raised the level of their voices and produced more high-frequency speech energy in background noise, any benefit to intelligibility that might have occurred as a result of these changes was more than offset by the decrease in the SNR. This conclusion is supported by the marked decrease in the SII as the background noise level increased. Further, these effects were even more pronounced when the talkers wore earplugs. Since the SNR and SII are important predictors of intelligibility 共Bradley, 1986兲, these data suggest that the intelligibility of the talkers would decrease as the background noise level increased, and that intelligibility would be further compromised when the talkers wore earplugs. Indeed, previous research has shown that talkers wearing HPDs are less intelligible in noise than talkers without HPDs 共Kryter, 1946; Howell and Martin, 1975; Martin et al., 1976兲. Despite the differences in the speech spectra between talkers with and without earplugs in the same background noise level, the shape of a talker’s speech spectrum does not appear to depend on whether the talker is wearing earplugs,
J. B. Tufts and T. Frank: Speech production and hearing protection
1077
but on the level of the talker’s voice. As an illustration, the talkers without earplugs in background noise at 70 dB SPL and the talkers with earplugs in background noise at 100 dB SPL had overall speech levels of 73.8 and 73.1 dB SPL, respectively, and corresponding SCoGs of 675 and 690 Hz. These measures were similar despite the fact that they were obtained in different ear conditions. Previously, Martin et al. 共1976兲 found no differences in the octave-band speech spectra of talkers with and without HPDs 共V51-R earplugs and Amplivox Sonogard earmuffs兲. These data suggest that earplug attenuation and the occlusion effect, in and of themselves, do not cause a talker to alter his or her speech spectrum. Instead, the changes in the speech spectrum are driven by the changes in the overall level of the talker’s voice. Thus, talkers with and without earplugs should be equally intelligible in background noise, provided their overall speech levels are the same. Unfortunately, the desired effect of wearing earplugs, the attenuation of ambient noise, creates an undesirable effect, the reduction of overall speech levels. This lowers the SNR and reduces the potentially beneficial changes in spectral shape that occur with an increase in the overall speech level. As a result, talkers wearing earplugs in background noise provide less speech information to listeners than talkers without earplugs. Small but consistent differences in the speech measures were observed between the earplug groups. The mean overall speech levels, SNRs, SCoGs, and SII values of the foam earplug group were slightly lower than those of the flange earplug group. These data suggest that, within the constraints of this study, talkers wearing foam earplugs may be slightly less intelligible than talkers wearing flange earplugs. However, the small differences between the earplug groups in this study would probably be of little or no consequence in a real-world setting, in which many other variables impact the communication process. Although it was not a significant main effect in any of the analyses that were conducted, the factor of testing order appeared in a small number of significant interactions with the factor of background noise level. Because its appearance was not consistent across dependent measures, and because the pattern of each dependent measure across background noise levels was very similar for both testing orders, it did not affect the interpretation of the results. The interactions involving the factor of sex for the SII were unexpected. However, the largest differences between the SII values for men and women, which occurred in the background noise levels of 60, 70, and 80 dB, were on the order of several hundredths and thus are not predictive of substantial differences in intelligibility between men and women. The significant main effect of sex and its interactions for the SCoG were expected due to the inherent differences in the speech spectra of men and women. The relationship between the attenuation and occlusion effect caused by HPDs on the one hand, and speech production on the other, is often mentioned in passing in the literature. In order to examine this relationship more thoroughly, the speech measurements were reanalyzed with attenuation and occlusion effect included as covariates. If, after the effects of attenuation and occlusion were removed, the main 1078
J. Acoust. Soc. Am., Vol. 114, No. 2, August 2003
effect of ear condition, its interaction with background noise level, or both were no longer significant, this indicated that one or both of the covariates accounted for the differences in the dependent measures between the ear conditions. In that case, follow-up ANOVAs were then conducted with each covariate separately. This reanalysis revealed that the combination of the earplug attenuation and occlusion effect mediated the differences in the overall speech levels, SNRs, and SII values between the ear conditions. Earplug attenuation alone mediated the differences in the SCoG between the ear conditions. However, these results do not imply a causal relationship. Instead, it is reasonable to assume that the earplug attenuation and the occlusion effect changed the talkers’ perception of the ambient noise level, and perhaps the loudness of their own voices relative to the ambient noise level. As a result, the talkers adjusted the level of their voices, which in turn drove the differences in the SNR, SCoG, and SII between the ear conditions. Previously, it was shown that the mean overall speech levels, SNRs, SCoGs, and SII values of the talkers wearing foam earplugs were slightly lower than those of the talkers wearing flange earplugs. It is tempting to ascribe these differences to the greater attenuation provided on average by the foam earplugs. However, a similar reanalysis of the data between the earplug groups revealed that neither the earplug attenuation nor the occlusion effect accounted for the differences in any of the speech measures between the foam and flange earplug groups. It is not clear whether other factors were responsible for the differences in the speech measures between the earplug groups, or if perhaps the PAR was not a sensitive enough measure to account for the differences. Further research is necessary to evaluate the effects of attenuation and the occlusion effect on the overall level of talkers’ speech when HPDs are worn.
B. Communication in the workplace
Communicating in a noisy environment can be frustrating even for people with normal hearing who are not wearing HPDs. One of the greatest challenges in hearing conservation is to protect workers’ hearing without further impacting their ability to communicate. Unfortunately, these two goals sometimes appear to be at odds. The results of this study suggest that talkers wearing earplugs alter their speech in ways that are deleterious to intelligibility. However, several variables in addition to those manipulated in this study may also contribute to the success or failure of speech communication in a given situation. For this reason, several caveats should be kept in mind. First, the SII was calculated under the assumption that potential listeners would have normal hearing and would not be wearing HPDs. In a noisy work environment, listeners may have noise-induced hearing loss and/or may wear HPDs. The combined effects of cochlear distortion and audibility loss in a listener would probably further reduce the intelligibility of the talker. Second, the contributions of talker motivation, feedback from a conversation partner, and visual and contextual cues, were not assessed in this study. However, these factors would probably improve the intelligibility of talkers J. B. Tufts and T. Frank: Speech production and hearing protection
in a real-world setting. Third, the speech measures obtained from talkers producing connected speech while wearing foam or flange earplugs may not be generalizable to studies using other types of speech, such as single words or short phrases, or to other types of HPDs, such as earmuffs or flatattenuation earplugs. In addition, the speech measures may have been affected by the characteristics of the background noise presented to the talkers, the size and absorptive characteristics of the test room, and the talker–listener distance implied by the distance of the microphone from the talker. Further research is necessary to evaluate the effects of these additional variables on speech production. Therefore, in generalizing the results of this study to the real world, the speech measures are best viewed in terms of the trends they reveal across background noise levels and ear conditions, not in terms of their absolute values. A thorough assessment of communication problems in the workplace should consider not only the listener, but also the talker. One way to improve workplace communication may be to train workers to raise their voices when wearing HPDs. However, because changes in one’s speech level occur unconsciously in response to changes in the perceived ambient noise level, this would be very difficult 共Casali et al., 1987兲. HPDs that provide different attenuation characteristics may offer another solution. For example, in environments with sufficiently low noise levels, the use of HPDs having minimal attenuation may induce talkers to raise their voices while also eliminating the need to remove the HPDs to hear what others are saying. More generally, a research strategy that combines qualitative and quantitative methodologies may be the best way to pinpoint and evaluate workplace communication problems. Examining the conditions in particular work environments 共e.g., noise characteristics, HPD types, talker–listener distance, available visual cues, etc.兲 could lead to the development of unique strategies for improving communication in these settings. V. CONCLUSIONS
In quiet, the talkers’ mean overall speech levels, SCoGs, and SII values were very similar with and without earplugs. In the presence of background noise, talkers automatically raised the level of their voices, as expected. Further, the relative amount of high-frequency energy in their speech increased, as evidenced by the upward shift of the SCoG. However, the mean SNR and SII decreased markedly, indicating that less and less speech information was available to a potential listener as the background noise level increased. These trends occurred regardless of whether or not the talkers were wearing earplugs, and are consistent with previous research on the effects of noise on speech production and intelligibility. Although the trends in the speech measures were similar with and without earplugs, the differences that were observed would be of potential significance for communication. The overall speech levels and SNRs were 4 to 11 dB lower when the talkers wore earplugs, compared with the open-ear condition. In addition, the upward shift of the SCoG was much smaller. Finally, in moderate to high background noise levels 共60 to 90 dB SPL兲, less speech information was availJ. Acoust. Soc. Am., Vol. 114, No. 2, August 2003
able from the talkers when they wore earplugs, as quantified by the SII. The differences that were observed in the speech measures between the foam and flange earplug groups were statistically significant, but small in magnitude and probably of little real-world significance. No consistent differences in the speech measures were noted between men and women with the exception of the SCoG. The results of the study indicate that talkers wearing earplugs 共and consequently their listeners兲 are at a disadvantage when communicating in noise. The challenge to hearing conservationists is to reconcile the twin goals of protecting workers’ hearing and preserving their ability to communicate.
ACKNOWLEDGMENTS
This research was conducted as part of a doctoral dissertation under the supervision of Tom Frank. It was supported by a grant from the Grants and Contracts Office of the College of Health and Human Development at Penn State University, by matching funds from the Department of Communication Disorders at Penn State University, and by HHS/ CDC U60/CCU315855-02. The authors are grateful for the advice and assistance offered by Ingrid Blood, Marjorie Leek, Kevin Michael, Toby Mordkoff, and Robert Prosek. The authors would also like to thank the staff of the Research Section of the Army Audiology & Speech Center at the Walter Reed Army Medical Center and three anonymous reviewers for their helpful comments on an earlier version of this manuscript. The opinions or assertions contained herein are the private views of the authors and are not to be construed as official or as reflecting the views of the Department of the Army or the Department of Defense. 1
The mean values of the occlusion effect created by the earphones at 0.25, 0.5, and 1 kHz were 20.2, 4.2, and 2.1 dB, respectively. At the same test frequencies, the mean values of the occlusion effect created when the earplugs were worn without the earphones were 22, 10.8, and 6.9 dB, respectively 共collapsed over earplug type兲. When both earplugs and earphones were worn, the mean values of the occlusion effect were 23.2, 11.8, and 6.6 dB, respectively. The occlusion effect created by wearing both earphones and earplugs was similar in magnitude to the occlusion effect created by the earplugs alone. Each of the occlusion effects mentioned above was calculated relative to a completely unoccluded ear canal, with no earplugs or earphones worn. However, the occlusion effect of interest here was that created by wearing the earphones and earplugs together relative to the earphone-only condition. These values, reported in the text, were obtained by subtracting, at each test frequency, the BCT with earplugs and earphones from the BCT with earphones only. Alternatively, these values could be obtained by subtracting, at each test frequency, the occlusion effect due to the earphones alone from the occlusion effect due to wearing the earplugs and earphones 共with very minor corrections for rounding errors兲. Additional measurements of the talkers’ speech in quiet were made to assess the impact of the occlusion effect due to the earphones themselves on the overall speech levels. These measurements were made while the talkers wore neither the earplugs nor the earphones, and while they wore the earplugs without the earphones. The mean overall speech level when neither earplugs nor earphones were worn was 66.4 dB, compared with a mean overall speech level of 67.5 when earphones only were worn. When the earplugs were worn, the mean overall speech levels were 66.9 and 67.0 dB with and without earphones, respectively. The fact that the talkers were wearing earphones during the test session appears to have had minimal to no impact on their overall speech levels. 2 This explanation was suggested by Elliott Berger, E-A-R/Aearo Company, Indianapolis, IN.
J. B. Tufts and T. Frank: Speech production and hearing protection
1079
Abel, S. M., Alberti, P. W., Haythornthwaite, C., and Riko, K. 共1982兲. ‘‘Speech intelligibility in noise with and without ear protectors,’’ in Personal Hearing Protection in Industry, edited by P. W. Alberti 共Raven, New York兲, pp. 371–385. ANSI 共1991兲. ANSI S3.1-1991, ‘‘American National Standard: Maximum Permissible Ambient Noise Levels for Audiometric Test Rooms’’ 共American National Standards Institute, New York兲. ANSI 共1996兲. ANSI S3.6-1996, ‘‘American National Standard: Specifications for Audiometers’’ 共American National Standards Institute, New York兲. ANSI 共1997兲. ANSI S3.5-1997, ‘‘American National Standard: Methods for Calculation of the Speech Intelligibility Index’’ 共American National Standards Institute, New York兲. Bauman, K. S., and Marston, L. E. 共1986兲. ‘‘Effects of hearing protection on speech intelligibility in noise,’’ Sound Vib. 20, 12–14. Berger, E. H. 共1980兲. EARLog 3: The Effects of Hearing Protectors on Auditory Communications 共Aearo Company, Indianapolis, IN兲. Bond, Z. S., Moore, T. J., and Gable, B. 共1989兲. ‘‘Acoustic-phonetic characteristics of speech produced in noise and while wearing an oxygen mask,’’ J. Acoust. Soc. Am. 85, 907–912. Bradley, J. S. 共1986兲. ‘‘Predictors of speech intelligibility in rooms,’’ J. Acoust. Soc. Am. 80, 837– 845. Casali, J. G., Horylev, M. J., and Grenell, J. F. 共1987兲. ‘‘A pilot study on the effects of hearing protection and ambient noise characteristics on intensity of uttered speech,’’ in Trends in Ergonomics/Human Factors IV, edited by S. S. Asfour 共Elsevier Science, North Holland兲, pp. 303–310. Chung, D. Y., and Gannon, R. P. 共1979兲. ‘‘The effect of ear protectors on word discrimination in subjects with normal hearing and subjects with noise-induced hearing loss,’’ J. Am. Aud. Soc. 5, 11–16. Dreher, J. J., and O’Neill, J. J. 共1957兲. ‘‘Effects of ambient noise on speaker intelligibility for words and phrases,’’ J. Acoust. Soc. Am. 29, 1320–1323. Foreshaw, S. E., and Cruchley, J. I. 共1982兲. ‘‘Hearing protector problems in military operations,’’ in Personal Hearing Protection in Industry, edited by P. W. Alberti 共Raven, New York兲, pp. 387– 402. Helmkamp, J. D. 共1986兲. ‘‘Why workers do not use hearing protection,’’ Occup. Health Saf. 55, 52. Hormann, H., Lazarus-Mainka, G., Schubeius, M., and Lazarus, H. 共1984兲. ‘‘The effect of noise and the wearing of ear protectors on verbal communication,’’ Noise Control Eng. J. 23, 69–77. Howell, K., and Martin, A. M. 共1975兲. ‘‘An investigation of the effects of hearing protectors on vocal communication in noise,’’ J. Sound Vib. 41, 181–196. Huynh, H., and Feldt, L. S. 共1976兲. ‘‘Estimation of the Box correction for degrees of freedom from sample data in randomized block and split-plot designs,’’ J. Educ. Stat. 1, 69– 82. Junqua, J. 共1993兲. ‘‘The Lombard reflex and its role on human listeners and automatic speech recognizers,’’ J. Acoust. Soc. Am. 93, 510–524. Kryter, K. D. 共1946兲. ‘‘Effects of ear protective devices on the intelligibility of speech in noise,’’ J. Acoust. Soc. Am. 18, 413– 417. Lane, H., and Tranel, B. 共1971兲. ‘‘The Lombard sign and the role of hearing in speech,’’ J. Speech Hear. Res. 14, 677–709. Lazarus, H. 共1990兲. ‘‘New methods for describing and assessing direct speech communication under disturbing conditions,’’ Environ. Int. 16, 373–392.
1080
J. Acoust. Soc. Am., Vol. 114, No. 2, August 2003
Letowski, T., Frank, T., and Caravella, J. 共1993兲. ‘‘Acoustical properties of speech produced in noise presented through supra-aural earphones,’’ Ear Hear. 14, 332–338. Lindeman, H. E. 共1976兲. ‘‘Speech intelligibility and the use of hearing protectors,’’ Audiology 15, 348 –356. Margolis, R. H., and Heller, J. W. 共1987兲. ‘‘Screening tympanometry: Criteria for medical referral,’’ Audiology 26, 197–208. Martin, A. M., Howell, K., and Lower, M. C. 共1976兲. ‘‘Hearing protection and communication in noise,’’ in Disorders of Auditory Function, edited by S. D. G. Stephens 共Academic, New York兲, pp. 47– 62. McDaniel, D. M., and Cox, R. M. 共1992兲. ‘‘Evaluation of the Speech Intelligibility Rating 共SIR兲 test for hearing aid comparisons,’’ J. Speech Hear. Res. 35, 686 – 693. Michael, K. 共1999兲. ‘‘Measurement of insert-type hearing protector attenuation on the end-user: A practical alternative to rely on the NRR,’’ Spectrum 16, 13–17. National Institute for Occupational Safety and Health 共NIOSH兲 共1996兲. Preventing Occupational Hearing Loss—A Practical Guide, DHHS 共NIOSH兲 Publication No. 96-110 共U.S. Department of Health and Human Services, Public Health Service, Centers for Disease Control, National Institute for Occupational Safety and Health, Cincinnati, OH兲. Navarro, R. 共1996兲. ‘‘Effects of ear canal occlusion and masking on the perception of voice,’’ Percept. Mot. Skills 82, 199–208. Nixon, C. W., and Berger, E. H. 共1991兲. ‘‘Hearing protection devices,’’ in Handbook of Acoustical Measurements and Noise Control, 3rd ed., edited by C. M. Harris 共McGraw-Hill, New York兲, pp. 21.1–21.24. Pittman, A. L., and Wiley, T. L. 共2001兲. ‘‘Recognition of speech produced in noise,’’ J. Speech Lang. Hear. Res. 44, 487– 496. Rink, T. L. 共1979兲. ‘‘Hearing protection and speech discrimination in hearing-impaired persons,’’ Sound Vib. 13, 22–25. Royster, L. H., and Holder, S. R. 共1982兲. ‘‘Personal hearing protection: Problems associated with the hearing protection phase of the hearing conservation program,’’ in Personal Hearing Protection in Industry, edited by P. W. Alberti 共Raven, New York兲, pp. 447– 470. Summers, W. V., Pisoni, D. B., Bernacki, R. H., Pedlow, R. I., and Stokes, M. A. 共1988兲. ‘‘Effects of noise on speech production: Acoustic and perceptual analyses,’’ J. Acoust. Soc. Am. 84, 917–928. Suter, A. H. 共1992兲. Communication and Job Performance in Noise: A Review 共American Speech-Language-Hearing Association, Rockville, MD兲, Monograph No. 28. Tartter, V. C., Gomes, H., and Litwin, E. 共1993兲. ‘‘Some acoustic effects of listening to noise on speech production,’’ J. Acoust. Soc. Am. 94, 2437– 2440. Traunmu¨ller, H., and Eriksson, A. 共2000兲. ‘‘Acoustic effects of variation in vocal effort by men, women, and children,’’ J. Acoust. Soc. Am. 107, 3438 –3451. Webster, J. C., and Klumpp, R. G. 共1962兲. ‘‘Effects of ambient noise and nearby talkers on a face-to-face communication task,’’ J. Acoust. Soc. Am. 34, 936 –941. Williams, D., and Michael, K. 共1991兲. ‘‘Maximizing communication ability in the selection of hearing protective devices,’’ J. Acoust. Soc. Am. 89, 1905.
J. B. Tufts and T. Frank: Speech production and hearing protection
