Reliability of Tinnitus Loudness Matches under Procedural ...Reliability of Tinnitus Loudness Matches under Procedural Variation James A. Henryt Christopher L. Flick Alison Gilbert*

J Am Acad Audiol 10 : 502-520 (1999)

Reliability of Tinnitus Loudness Matches under Procedural Variation James A. Henry*t Christopher L. Flick* Alison Gilbert* Roger M. Ellingson* Stephen A. Fausti*t

Abstract

Repeated tinnitus loudness matches (LMs) were obtained to determine response reliability using a computer-automated technique with two procedural variations, fixed or random step sizes, to increase output level during the initial ascending series of tones at each frequency. Twenty subjects with stable, tonal tinnitus were evaluated with both methods during each of two sessions . Response instructions were displayed on a portable computer, and a pen device was used to make response choices that appeared on the touch-sensitive video mon-itor. For each method, hearing thresholds and LMs were obtained, with 1-dB resolution, at '/s-octave frequencies from 1 to 16 kHz. Analyses revealed reliability of LMs to be equivalent between methods . LM data are reported in both dB SPL and dB SL, with the SPL values pro-viding greater reliability both within and between sessions (all is ? .889, p's <_ .0001) .

Key Words: Loudness, measurement, reliability, tinnitus

Abbreviations: ANOVA = analysis of variance, LAN = local area network, LM = loudness match

T

he study of tinnitus has become increas-ingly prevalent over the last few decades, especially with regard to treatments and

mechanisms (Tyler et al, 1992) . Significant progress has been made, and there is currently a high level of interest in conducting tinnitus research . These auspicious developments are providing hope and encouragement for patients whose lives are severely impacted by tinnitus (Fortune et al, 1999).

For tinnitus research involving humans, as well as for clinical tinnitus management, it is important to have methodology available for reliably quantifying the loudness and other per-ceptual attributes of tinnitus (Vernon and Fen-wick, 1984 ; Coles and Baskill, 1996 ; Goldstein, 1997). Although not yet demonstrated, these measurements should provide agreement with treatment outcome and be diagnostic for select-

*National VA RR&D Center for Rehabilitative Auditory Research, Portland VA Medical Center, Portland, Oregon ; tDepartment of Otolaryngology, Oregon Health Sciences University, Portland, Oregon

Reprint requests : James A . Henry, RR&D Center for Rehabilitative Auditory Research, Portland VA Medical Center (R&D-16), PO Box 1034, Portland, OR 97207

ing treatment strategies (Jastreboff, 1996). Many adaptations of psychoacoustic matching tech-niques have been described for this purpose, but none have received general acceptance . Tin-nitus "quantification" thus relies upon an indi-vidual's subjective report to determine if and how the tinnitus changes as the result of some exper-imental manipulation or treatment. The need exists for a technique to reliably identify acoustic stimuli that simulate an individual's tinnitus per-ception and that can be used in both research and clinical settings .

In the clinic, tinnitus matching measure-ments are often made in an attempt to define a tone, noise band, or other sound that mimics the acoustic sensation that a patient experiences as tinnitus . This type of technique was first described by Fowler (1943), and many subse-quent studies used the same basic technique with variations (Reed, 1960; Graham and Newby, 1962 ; Roeser and Price, 1980 ; Tyler et al, 1992). These and other studies have generally failed to show any relation between psychoacoustic tin-nitus measures and subjective improvement resulting from therapy (Hazell et al, 1985 ; Scott et al, 1985; Kuk et al, 1989; Murai et al, 1992 ; Jastreboff, 1996).

502

There has been little progress in the area of tinnitus measurement since the mid 1980s, which is somewhat inexplicable and unfortu-nate since tinnitus research involving humans requires reliable techniques to quantify the dis-order and to measure changes in the tinnitus per-ception over time. A battery of audiologic tests has long been standardized for the evaluation of auditory function ; however, a standardized test protocol does not exist for evaluating tinnitus . This is a detriment to both clinical care and to tinnitus research . It is the overall purpose of a series of ongoing experiments in this labora-tory to develop clinical methodology for con-ducting tinnitus matching that can lead to procedural standardization .

The requisite first step in developing a clin-ical technique for tinnitus matching is to deter-mine that the method obtains reliable responses . We have chosen to use computer automation to conduct such measurements, standardize the protocol, and minimize operator intervention and potential biasing characteristics . A prototype version of the automated system showed that reliable tinnitus loudness matches (LMs) were feasible using this approach (Henry et al, 1996) . The automated system has since received major new programming and hardware. The testing algorithms have been refined and programmatic changes have been implemented to improve testing efficiency and to allow all parameters to be easily adjusted .

There were two objectives for the present study: (1) to document test-retest reliability of tinnitus LMs at a series of 13 frequencies using the upgraded system and (2) to evaluate if there was any decrease in response reliability when randomization was applied to the start level of loudness matching tones and to the step sizes for increasing the level of the tones. In the testing algorithm for the original computerized system, loudness matching tones were raised above threshold in fixed step sizes until the subject (usually) reported that the tone was louder than the tinnitus . At that point, the tone was decreased in level as part of a bracketing pro-cedure . By presenting the initial series of ascend-ing tones in fixed decibel increments, the patient can, consciously or subconsciously, count the number of steps to arrive at a consistent LM. The question, then, is whether the reliability of LMs is improved through the use of a fixed order of stimulus presentation . If so, the reliability of such a technique is at least partially ascribable to cues provided by the testing protocol . Patients, therefore, may be more reliable with their

Tinnitus Loudness Matching/Henry et al

responses when the step sizes are presented in a series of predictable steps than if randomiza-tion is introduced into the protocol . To investi-gate this premise, tinnitus loudness matching was done using fixed and random methods, and these methods were evaluated independently and differentially for response reliability.

METHOD

Subjects

Twenty subjects participated in this study. They included two females and 18 males with a mean age of 58.0 years and an age range of 24 to 78 . These subjects were selected on the basis of having tonal, stable tinnitus in order to min-imize any variability in the tinnitus that might confound interpretation of the reliability analy-ses . Eleven of the subjects were previously patients, with the primary complaint of tinnitus, at the Tinnitus Clinic, Oregon Health Sciences University. They were thus experienced in tin-nitus evaluation using a clinical matching tech-nique . The remaining nine subjects did not have previous experience and consisted of four sub-jects from an earlier hearing aid study, three vet-erans referred from the Audiology Clinic at the Portland VA Medical Center, and two individu-als from the local community.

Hardware

The three major hardware components of the tinnitus testing system included (1) the main computer (Dell Dimension 166 MHz Pen-tium CPU) and signal generator ; (2) a custom-built signal conditioning module for signal mixing, attenuation, and headphone buffering (Oregon Hearing Research Center, Portland, OR); and (3) the subject computer (Compaq Concerto 4/25) . A block diagram of the system is shown in Figure 1 .

Acoustic Subsystem

Instrumentation for signal generation, con-ditioning, and transduction was configured in the signal path depicted in Figure l. Pure-tone sig-nal generation was accomplished using a preci-sion signal generator (National Instruments, AT-DSP2200-128k) housed in the main com-puter. The pure-tone signal was routed to the sig-nal conditioning module, which performed pulse shaping, attenuation, and headphone buffering

503

Journal of the American Academy of Audiology/Volume 10, Number 9, October 1999

Main Computer Acoustic Signal Path

Subject Instruction & Response Subsystem

Serial Interface

LAN

Subject Computer

Sound Booth

Signal Generator

Signal Conditioning

Module

ER-4 Insert Phones

Figure 1 Block diagram of hardware used for com-puter-automated tinnitus evaluation . LAN = local area network.

functions. Insert earphones (Etymotic Corp. ER-4) were used for transducers.

The signal generator was a 16-bit, high-performance, high-accuracy audio bandwidth digital to analog I/O board, which was used to generate stimulus waveforms directly from computer-generated digital data . The analog signals generated by the board were low-distortion, low-noise, and frequency-flat (http:// www.natinst.com). The signal generator board had both analog and digital anti-imaging fil-ters to remove unwanted glitches and high-frequency signals. These filters limited the bandwidth of the analog output signal to half the digital-analog conversion rate of 44,100 Hz.

A serial port from the main computer was used to control the signal conditioning module . This module provided programmable functions that included input mixing and gain, pulse win-dowing and muting, and overall attenuation. The attenuation range was 0.0 to 159.9 dB with 0.1-dB resolution .

The Etymotic ER-4 insert earphones fre-quency response characteristics met the needs of this study: >_ 100 dB SPL output from 1 to 16 kHz and <_ 3 percent harmonic distortion .

The main computer and the subject com-puter were connected via a local area network (LAN) interface using standard networking pro-tocols for two-way communication (see Fig. 1) . The Subject Instruction & Response subsystem was used to display prompts, messages, and instructions on the subject computer and to relay subject test responses back to the main computer. The subject computer was an IBM PC compatible, Microsoft Windows for Pen enabled, slate-type, notebook computer located inside the testing booth. The subject used a pen-pointing device to indicate responses by pen "touching" the appropriate buttons.

The subject computer was battery powered with a solid-state display to minimize poten-tially adverse acoustic and electromagnetic inter-ference inside the testing booth. The computer was upgraded to 20 Mb of memory, allowing the operating system's virtual memory feature to be disabled, which eliminated sound interference from hard drive disk use while managing sys-tem memory.

System Calibration

The acoustic subsystem, which consisted of the signal generation board, the signal condi-tioning module, and insert earphones, was cal-ibrated at the beginning of each test day. The same digital stimulus waveform generation soft-ware that was used during testing was used when calibrating the acoustic subsystem.

A custom automated calibration applica-tion was developed to provide a multifrequency, multilevel calibration set. A computer-controlled calibration routine was developed with serial interface control of a Bruel & Kjaer 2231 sound level meter and Type 1625 octave filter set (B&K; http ://www.bk.dk) . The ER-4 insert earphones were coupled to the sound level meter using a B&K Type 4157 ear simulator.

The calibration values were stored in a data-base and later accessed while testing to provide attenuation settings for calibrated sound pres-sure stimulus levels .

Software

Both the main and subject computers used the Microsoft Windows 95 operating system . All software developed for this study was designed to be Microsoft Windows 95 compatible . Microsoft

504


Windows OLE remote automation technology was used to provide the virtual connection between the main computer and the subject computer at the application level for passing control, instruction, and response messages.

Software was developed to simultaneously control (1) temporal parameters of pure-tone stimuli; (2) stimulus frequency and attenuation settings ; (3) display of subject instructions and response touch-buttons; (4) recording of subject responses for program control; (5) logging of subject responses into data files; and (6) main computer screen display for monitoring test sta-tus, progress, and results.

The main computer software program con-sisted of dialog forms for examiner entry of sub-ject information, test session information, parameters for the tinnitus-testing control loop, and visual displays for monitoring testing sta-tus, progress, and results . During testing, the

control loop established frequency and attenu-ation settings, formulated and routed instruction

messages to the subject computer, acquired the

subject response, adjusted the stimulus wave-

form frequency and levels according to the

response, and logged results . The examiner had the capability of overriding subject responses,

pausing, or stopping testing during the test ses-

sion . Overriding subject responses was not nec-

essary for any of the subjects reported in this

study. The only times testing was interrupted was when the subject touched the "Help" button, which was visible at all times on the subject

computer display. This paused the test and allowed the examiner to answer the subject's question(s) . Testing would then resume at the point of interruption .

Database-Driven Architecture

The main computer control program was designed from the ground up to be entirely dri-ven by a Microsoft Access desktop database . Information stored in database tables controlled the main program operation including para-meters for testing protocols, timing intervals, frequency sets, level-based step sizes, random-ization ranges, numbers of steps, numbers of averages, instrumentation used, etc. The test-ing system was therefore easily configurable by nonprogrammers for modification of experi-mental protocols. Detailed testing data were logged directly into tables during program oper-ation . Data logged included stimulus presenta-tion history, responses, and intermediate and final averaged results. Results and associated

testing parameters were easily queried out fol-lowing testing for multiway variable analysis .

Procedures

All procedures were conducted over two testing sessions, separated by 1 to 14 days (one subject completed both sessions in 1 day, sepa-rated by 2 hours) . Session 1 required 11/z to 2 hours of a subject's time, and Session 2 required about 11/a hours . During the first session, an initial evaluation was done to obtain a case his-tory, gather noise exposure and tinnitus infor-mation, determine middle-ear function and hearing sensitivity, and identify the "tinnitus ear."

Two experimental LM protocols, fixed and random, were completed during each session . Between protocols, the earphones were removed and the subject took an approximately 10-minute

break. Testing order for the two protocols was

counterbalanced by alternating the order for

each successive subject. In that way, the fixed

protocol was tested first for 10 of the subjects and

the random first for the other 10 . For each sub-

ject's second visit, the order of testing was reversed .

During each session, otoscopy was done prior to placing earphones for audiometric test-ing to confirm unoccluded ear canals and the nor-mal appearance of tympanic membranes.

Initial Evaluation (First Session Only)

At the start of the first session, a short case history was obtained to provide information regarding demographics, auditory and vestibu-

lar disorders, and family history of hearing loss .

Subjects were also asked if they had been exposed to significant noise and, if so, they com-pleted a noise exposure questionnaire . All sub-

jects also completed a tinnitus questionnaire . For each questionnaire, the attending audiologist asked the questions and subjects responded verbally.

Tympanometric screening was performed with a Grason-Stadler GSI-37 Auto Tymp to rule out active middle-ear pathology. Prior to test-ing with the automated technique, hearing thresholds were obtained manually with a clin-ical, high-frequency-capable audiometer (Vir-tual Corp., Model 320) at octave frequencies from 0.25 to 8 kHz, and at 1.5, 3, 6, 9, 10, 11 .2,

and 12.5 kHz. Instrumentation and procedures for manual threshold evaluation were as described in Fausti et al (1990) .

505


Selection of Tinnitus Ear. One ear was cho-sen as the "tinnitus ear" for each subject. If one ear had more predominant tinnitus, that ear was chosen as the tinnitus ear. If the subject had symmetric tinnitus, the tinnitus ear was chosen randomly.

Selection of Test Ear. Matching tones can be presented ipsilaterally or contralaterally to an ear with tinnitus . Both methods seem to have been used with about equal frequency, and we know of only one report comparing the two meth-ods within patients (Ishikawa et al, 1991). In that study, it was seen that measurements were essentially equivalent when hearing was sym-metric, although with asymmetric hearing sen-sitivity, LMs tended to be larger on the side with better hearing.

For the present study, the ear contralateral to the tinnitus ear was chosen as the test ear. Contralateral presentation of pure tones was chosen partly to minimize the potential of pro-ducing residual inhibition, which is the com-mon phenomenon of temporary tinnitus suppression following presentation of a suprathreshold stimulus (Vernon, 1987 ; Mitchell et al, 1993). This was a particular concern due to the large number of stimulus presentations each subject received . Furthermore, matching to contralateral tones has been reported to be a less difficult task than ipsilateral matching (Good-win and Johnson, 1980).

a

1 . You will be listening for soft pulsing tones . I

2. When you hear a tone, touch the 'I Hear It' box.

3. To start the test, touch'Go'.

Help

Experimental Protocols (Both Sessions)

Instructions to Subjects. Because of the alter-nating nature of the response tasks, instruc-tions for responding were required each time the task changed. This was accomplished for the automated procedure by displaying instruction screens prior to each task, as shown in Figures 2a and 3a .

Test Frequencies. Test frequencies for hearing thresholds and tinnitus LMs (using the auto-mated technique) were in the range of 1 to 16 kHz and separated by 1/s octaves (rounded to accommodate the 20-Hz resolution of the test equipment) . Frequencies included 1, 1.26, 1.6, 2, 2.52, 3.18, 4, 5.04, 6.36, 8, 10.08, 12.7, and 16 kHz, and testing proceeded in a stepwise fash-ion in this frequency order. LMs were not obtained below 1 kHz primarily because, for the majority of patients, tinnitus pitch matches are above 3 to 4 kHz, with very few below 1 kHz (Meikle and Walsh, 1984 ; Meikle et al, 1995).

b

Figure 2 Screen displays on subject's notebook com-puter for hearing thresholds : (a) instructions, (b) response screen .

Thus, the frequency range of 1 to 16 kHz would be most pertinent for clinical tinnitus matching, and additional frequencies would have increased subject fatigue in this study.

Automated Hearing Thresholds. As part of both the fixed and random protocols, automated hearing thresholds were obtained, followed by loudness matching, at each frequency. The goal for obtaining hearing thresholds with the auto-mated system was not to obtain hearing thresh-olds as normally defined (i .e ., 50% response level) . Rather, "threshold" was operationally defined as the average of two minimum response levels determined using an adaptation of the modified Hughson-Westlake audiometric test technique (Carhart and Jerger, 1959). The two responses were obtained during presentation of tones in ascending 1-dB increments . The threshold at each frequency served as the level from which to begin tinnitus loudness matching at that frequency, and also as the point from which to calculate LM sensation levels .

506


Tm,ftm Loud. ... f arch a

!. t . You will hear a single tone lasting several seconds.

2. Decide how the loudness of the tone compares to the loudness of your tinnitus.

3 . Touch the box which indicates your response choice .

b

Go

Tinnhus Loudnessld. .h

Tone On

How should the tone's loudness be changed to

make it match the loudness of your

tinnitus?

Much Louder

Louder

Equal

Softer

Much Softer

Help

Repeat

Help

Figure 3 Screen displays on subject's notebook com-puter for tinnitus loudness matching: (a) instructions, (b) response screen .

At each new test frequency, subjects were first shown instructions for threshold testing. When they had read and understood the instruc-tions, they touched the "Go" button on the screen with the pen device, the threshold testing screen appeared (see Fig. 2b), and testing proceeded.

At each test frequency, the initial presen-tation level was fixed at 60 dB SPL. Tone pre-sentation levels were then raised or lowered, dependent upon whether the subject responded to the stimulus during a specific time window. Pulsed pure tones of 400-msec duration and rise-fall times of 25 to 45 msec were presented at a 50 percent duty cycle . Each series of pulsed

tones was presented in a 2.4-sec window. Inter-vals between tone series were randomized from between 1 and 4 sec following a response and fixed at 1 sec following a no response .

The algorithm for determining an auto-mated hearing threshold consisted of a series of three bracketing procedures, each providing progressively smaller step sizes to finally result in threshold responses with 1-dB resolution (threshold-seeking algorithm summarized in Table 1) . The initial bracketing series (Series 1) used step increments of up 10 dB, down 20 dB to quickly bracket the threshold level to within 10 dB . To minimize the possibility of presenting tones at uncomfortably loud levels, the ascend-ing step size during Series 1 (and Series 2) was limited to 3 dB at levels ? 90 dB SPL. Subsequent bracketing series used algorithms of up 5 dB, down 10 dB (Series 2) and up 1 dB, and down 2 dB (Series 3) .

Each series was conducted until a mini-mum response level (lowest level where a

response was obtained) was determined . One response was required for Series 1, and two responses were required for each of Series 2 and 3 (see Table 1) . When two responses were required during a series, the responses were averaged to obtain the minimum response level for that series .

Following the response for Series 1, the ini-tial presentation level for Series 2 equaled the minimum response level determined from Series 1 minus 10 dB (see Table 1) . For Series 3, the initial level was the minimum response level from Series 2 minus 2 dB . Only the Series 3 min-imum response level was used for subsequent loudness matching and for data analysis .

Design Considerations for Automated Tin.

nitus Loudness Matching. There is currently no standard procedure for matching the loudness

of tinnitus using pure tones at a series of fre-quencies . A well-defined clinical protocol was described by Vernon and Meikle (1981, 1988), who recommended first presenting the loudness matching tone at threshold, then increasing the

Table 1 Level Step-Size Parameters for Pure Tones during Automated Threshold Testing

Brac Serie

ket s Start Level

Ascending Step Size (dB)

Descending Step Size (dB ) Repeated?

1 60 dB SPL +10 (+3 if level ? 90 dB SPL) -20 No

2 Mean of previous series Result minus 10 dB +5 (+3 if level ? 90 dB SPL) -10 Yes

3 Mean of previous series Result minus 2 dB 1 -2 Yes

507


tone in 1-dB increments until the patient reports a LM. Such a gradual increase was intended to reduce the possibility of producing residual inhi-bition (tinnitus suppression due to presenta-tion of suprathreshold stimuli) . However, for frequencies <_ 2 kHz, LMs have been shown to be typically obtained at approximately 20 dB SL and at about 10 dB SL above 2 kHz (Henry and Meikle, 1999). Raising the level in 1-dB steps might therefore require an average of 10 to 20 tone presentations (and subject responses directing the computer to increase the tone's loudness) to reach a LM level at each frequency. Because the present efforts were directed toward clinical expediency, it was necessary to design algorithms to obtain LMs rapidly while main-taining response reliability and minimizing residual inhibition and loudness discomfort.

General Procedures for Loudness Matching. The computer algorithm selected was an interleaved design in which the subject was guided through threshold and subsequent LM testing at each frequency. Starting at the low-est frequency (1 kHz), a threshold was obtained, as described above, followed by the LM proce-dure . When a LM was obtained at 1 kHz, the computer stepped to 1 .26 kHz and obtained a threshold and subsequent LM. This interleaved protocol continued in ascending frequency order until thresholds and LMs had been obtained at all 13 test frequencies, or to a frequency where the subject's pure-tone threshold exceeded the maximum output of 100 dB SPL.

When the computer changed tasks from acquiring a threshold to acquiring a LM at the same frequency, subjects were first shown instructions for the loudness matching task (see Fig. 3a). When they understood the instruc-tions, they touched the "Go" button on the screen and the testing/response screen for loudness matching appeared (see Fig. 3b).

Tones for loudness matching were presented continuously for 4 seconds, and the subject made a decision regarding how to adjust the loudness of the tone to match the tinnitus loudness . More specifically, during the first 1 second of the tone presentation, only the "Tone On" message appeared on the screen (see Fig. 3b). After 1 second, the response choices appeared and were visible beyond termination of the tone or until the subject selected a response . Thus, a response could be made after the first 1 second of the tone presentation, and if the response occurred while the tone was on, the tone was immediately terminated .

Response choices for loudness matching included "Much Louder," "Louder," "Equal," "Softer," and "Much Softer" (see Fig. 3b). Fol-lowing the response choice, the computer adjusted the stimulus presentation level for the subsequent matching tone . Generally, a "Much Louder" or "Much Softer" response doubled the change in output level relative to a "Louder" or "Softer" response . At any point during LM test-ing, an "Equal" response choice directed the computer to record the previous tone's level as the LM for the active test frequency.

Both the fixed and random protocols used a progression of three bracketing series to grad-ually reduce the step sizes of tones presented for loudness matching. The initial series was designed to quickly determine the general loud-ness of the subject's tinnitus, with two subse-quent bracketing series to define the LM with increasing precision. Unlike threshold testing, LM test series were not repeated; the subject could determine a final LM following any tone presentation during any of the three series by selecting "Equal." This would immediately ter-minate the bracket series under test and direct the computer to move to the next frequency.

The fixed and random procedures differed only in the first bracketing series-specifically, in how the initial presentation level was deter-mined and in the step size adjustments follow-ing "Louder" and "Much Louder" responses . Following is a description of the fixed protocol and the variations that defined the random protocol .

Computer Algorithm for "Fixed" Loudness Matching. For Series 1, the initial presentation level was 10 dB SL (i.e ., 10 dB above the thresh-old determined immediately prior to the loudness matching task) (Table 2) . However, the initial level was reduced to 3 dB SL if the threshold was between 70 and 89 dB SPL and further reduced to 1 dB SL if the threshold was between 90 and 99 dB SPL. If the threshold was 100 dB SPL (the maximum output level), the initial LM tone was presented at 100 dB SPL and could not be pre-sented at a higher level. If the subject selected "Louder" or "Much Louder" as a response choice when 100 dB SPL was presented for loudness matching, the computer logged the LM for that frequency as ">100 dB SPL."

Table 2 shows the changes in presentation level provided by the computer based upon the various response choices for changing the level. For Series 1, response choices of "Softer" and "Much Softer" resulted in reductions in output

508

Tinnitus Loudness Matching/ fenry et al

Table 2 Presentation Levels of Pure Tones for Automated Tinnitus Loudness Matching

Bracket Series

Presentation Level Range (dB SPL)

Start Level (dB SL) ouder

Presentation Level Adjustment (dB) Based Upon Response Choice

Much Louder Softer

Much Softer

0-69 10 +5 +10 -4 -8 1 70-89 3 +3 +5 -4 -8 90-99* 1 +1 +2 -4 -8

0-69 N/A +3 +6 -2 -4 2 70-89 N/A +2 +4 -2 -4 90-99* N/A +1 +2 -2 -4

0-69 N/A +1 +1 -1 -2 3 70-89 N/A +1 +1 -1 -2 90-99* N/A +1 +1 -1 -2 *The maximum output was 100 dB SPL. If the threshold was 100 dB SPL, the start level for loudness matching was 100 dB SPL. If the subject indicated that a tone should be made louder than 100 dB SPL for a LM, the computer logged the LM for the active test frequency as ">100 dB SPL."

level of-4 and-8 dB, respectively, regardless of the previous presentation level . For increases in level, however, the step sizes were reduced if the previous presentation level was between 70 and 89 dB SPL and further reduced when the pre-vious level was between 90 and 99 dB SPL. Thus, when the previous presentation level was within 69 dB SPL, a choice of "Louder" resulted in an increase of 5 dB, whereas a choice of "Much Louder" resulted in an increase of 10 dB . If the previous presentation level was between 70 and 89 dB SPL, increases in level were reduced to +3 dB for a "Louder" response and to +5 dB for a "Much Louder" response . Level increases were further reduced to +1 dB and +2 dB for "Louder" and "Much Louder" responses, respectively, when the previous presentation level was between 90 and 99 dB SPL.

Progression of the test (ascending or descending intensity of consecutive tone pre-sentations) was determined based on the sub-ject's response choices . A subject's LM was bracketed using the concept of response rever-sal . A reversal was determined each time the sub-ject changed the direction of their responses on consecutive tone presentations (e.g ., going from a "Louder" response to a "Softer" response) . A response reversal defined either an upper or lower limit (depending on the direction of the reversal) for an expected range of where the final LM level may be found . By decreasing the ascending and descending step sizes following reversals, it was possible to narrow the inten-sity range to determine a final LM to within 1 dB . This narrowing is indicated by the three bracket

series shown in Table 2 . Each successive series provided smaller step sizes to enable progres-sively finer control for precise loudness match-ing . The main computer program shifted from one series to the next following a reversal from a response choice directing the computer to pre-sent a softer tone to a louder tone . The loudness matching continued in Series 3 until the subject chose "Equal" to indicate a LM with the tinnitus .

Computer Algorithm for "Random" Loud-ness Matching. The random loudness match-ing procedure was identical to the Fixed procedure described above except for the pre-sentation of tones during Series 1, as shown in Table 3. With the random protocol, the level of the first tone presentation for loudness match-ing ("Start Level") was selected randomly by the computer between 1 and 10 dB SL when the threshold was between 0 and 69 dB SPL. If the threshold was between 70 and 89 dB SPL, the randomized range was reduced to between 1 and 3 dB SL. If the threshold was 90 dB SPL or greater, the start level was 1 dB SL, as for the fixed protocol .

With the random protocol (Series 1 only), a selection of "Louder" or "Much Louder," follow-ing the start level presentation, resulted in randomized selection of increases in output level for loudness matching. If the previous presenta-tion level was <_ 69 dB SPL, a "Louder" selection resulted in an increase in level that was ran-domized between 1 and 5 dB, and "Much Louder" resulted in a level within a randomized increase between 6 and 10 dB (Table 3) . These ranges

509


Table 3 Comparison of Series 1 Level Step Sizes for Pure Tones: Fixed versus Random Loudness Matching Protocols

Presentation Level Adjustment (dB) Based Upon Response Choice

Presentation Level Range (dB SPL)

Start Lev

Fixed

el (dB SL)

Random

Louder

Fixed

Response

Random

Much Louder

Fixed

Response

Random

0-69 10 1-10 +5 1-5 +10 6-10

70-89 3 1-3 +3 1-3 +5 4-6

90-99* 1 1 +1 +1 +2 +2

*See note for Table 2.

were reduced when the previous presentation

level was between 70 and 89 dB SPL: a "Louder" response resulted in an increase randomized between 1 and 3 dB, and a "Much Louder"

response resulted in an increase randomized between 4 and 6 dB. When the previous pre-sentation level was between 90 and 99 dB SPL,

changes in levels were the same for both fixed and random protocols.

Data Analysis

Loudness Matches: dB SPL as dB SL

In the literature, it is equivocal whether to express tinnitus LMs in dB SPL or dB SL . LMs

in dB SPL reflect the absolute sound pressure level of a tone that is perceived as having the

same loudness as the tinnitus . LM measures in

dB SL are the difference between the dB SPL LM

and the hearing threshold at the same frequency.

LMs expressed in dB SL thus introduce a sec-ond source of within-subject variability. To allow

a comparison of results between the two metrics,

the data analyses are shown for both dB SPL and

dB SL.

compare the four across-subjects mean LMs at

each of the 13 test frequencies.

LM Differences

Descriptive statistics were also used for

LMs to provide another perspective of the

test-retest reliability of these responses. Dif-

ferences were calculated between measures obtained during the two sessions . Means of these

differences were determined across subjects and

were reported as actual differences (to determine

if there were trends toward higher or lower

responses during the second session) and

absolute values of differences (to determine the

absolute magnitude of the differences between sessions).

Pearson Product-Moment Correlations

To evaluate within-subject reliability of

responses, Pearson product-moment correla-

tions were calculated for hearing thresholds

and for LMs.

RESULTS

Repeated-Measures Analyses of Variance

Data analyses focused on evaluating test-retest reliability of LMs obtained with both

fixed and random protocols. Repeated measures

of LMs were obtained within (between protocols)

and between sessions . For each subject at each

frequency, there were therefore four measure-

ments for LMs (fixed and random during each

of two sessions). Repeated-measures analyses of

variance (ANOVAs) were used to evaluate how

consistently subjects as a group responded from

measure to measure, both within and between

test sessions . ANOVAs were therefore used to

Repeated Hearing Thresholds

Figure 4 shows the mean thresholds,

obtained using the automated technique, for

the group of 20 subjects . There are four means

at each frequency, reflecting the two thresholds

(for fixed and random methods) obtained during

each of the two sessions .

Across-Subjects Means of LMs

Across-subjects means of the LMs were cal-

culated, as shown in Tables 4 (dB SPL) and 5 (dB

SL) and Figures 5 (dB SPL) and 6 (dB SL).

510


100

90

60

~ 70 a N m 80 v

50 0 L

m 40

~ 30

20

10

0 _ -

r r n e ~n ~ m - °,d

Frequency, Hz

Figure 4 Mean hearing thresholds (as operationally defined for automated testing) for all subjects, test ear only (N = 20). S = Session.

Repeated-measures ANOVAs were done for the four means at each frequency, and the p values are shown in the last columns of Tables 4 and 5. Because of the large number of ANOVAs that were run, the alpha level was reduced to .01. The only significant difference was at 2 kHz (p = .0068) for the data expressed in dB SPL.

Means of Actual Differences in LMs: Within Subjects and Between Sessions

The above analysis shows that the group of

subjects as a whole provided reliable LMs between methods and between sessions . How-ever, it did not address the within-subject reli-

ability of responses . It was therefore determined how the two LM techniques compared for within-subject, between-session reliability. For each subject, the differences in corresponding LMs were calculated . That is, for each method (fixed

Frequency, Hz

Figure 5 Mean loudness matches expressed in dB SPL. S = Session.

and random) and each test frequency, the Ses-sion 2 LM was subtracted from the Session 1 LM. The across-subject means of those differences are shown in Tables 6 and 7 and Figures 7 and 8. These data reflect the actual differences between each pair of LMs, thus indicating the direction-ality of the responses between Sessions 1 and 2. It is evident that there was no systematic trend in the directionality of responses between ses-sions (i .e ., responses did not tend to get larger or smaller between sessions for the group of subjects). Thus, the variability associated with the LM responses was random in direction.

Means of Absolute Values of Differences in LMs: Within Subjects and Between Sessions

To determine the magnitude of the within-subject differences in responses between ses-sions, the absolute values of the LMs were

Table 4 Means of Loudness Matches Expressed in dB SPL and p Values for Repeated-Measures ANOVAs at Each Test Frequency

Frequency (Hz) N

Session

Fixed

1

Random Fixed

Session 2

Random p

1000 20 41 .85 39.00 44.15 40.70 0170

1260 20 44.75 41 .90 45.80 43.10 0636 1580 20 46.95 47.20 49.25 46.55 1773 2000 20 54.15 50.45 53.15 52.00 0068

2520 20 60.45 59.30 61 .00 59.35 4688 3180 20 66.05 64.70 66.60 64.15 1033

4000 20 68.10 66.75 69.25 66.65 0787 5040 20 69.80 67.75 70.20 68.85 0934 6340 19 68.32 67.47 68.05 64.05 0747 8000 19 67.05 67.32 68 .42 67.26 8966

10,080 15* 76.86 76.53 77 .27 75 .67 2579 12,700 11 80.91 81 .50 81 .09 81 .36 8337 16,000 8 91 .25 85.83 89.25 90 .63 3109

*N = 14 for Session 2 .

511


Table 5 Means of Loudness Matches Expressed in dB SL and p Values for Repeated-Measures ANOVAs at Each Test Frequency

Frequency (Hz) N Fixed

Session 1

Random Fixed

Session 2

Random p

1000 20 16.65 14 .00 18.25 14.70 0372 1260 20 16.20 13 .90 16.70 15.00 3147 1580 20 15.30 16 .25 16.80 14.00 .3698 2000 20 16.05 14 .10 14.95 13 .35 .3108 2520 20 14.90 13 .00 13.65 12 .35 2882 3180 20 12.05 10 .60 12.10 10 .30 3197 4000 20 12.00 11 .00 12.00 9.85 1452 5040 20 11 .10 9.25 10.50 10 .15 .2414 6340 19 11 .68 9.74 10.89 9.26 0507 8000 19 8.58 8.68 9.53 9.05 .9473

10,080 15" 9.36 6.93 7 .67 6.47 2544 12,700 11 7 .00 4.58 8 .45 7.82 1676 16,000 8 5 .37 4.83 4 .63 5.75 3435

`N = 14 for Session 2.

determined for each difference . The means of the absolute values of the Session 1 minus Session 2 differences are shown in Tables 8 and 9 and Figures 9 and 10 . These data reveal that there were no systematic trends in the magnitude of the LM differences between Session 1 and Ses-sion 2 for either the fixed or random methods. There was further no systematic variation in the magnitude of differences between the two methods.

evaluated further to determine within-subject, within-session reliability of responses. For each subject at each frequency, the LM from the ran-dom method was subtracted from the LM from the fixed method . This was done for each session, and the means of these differences were com-parable to the data shown above for evaluating within-subject, between-session reliability (tables and figures are not shown to conserve space) .

Means of Differences in LMs: Within Subjects and Within Sessions

Since there were essentially no significant differences between means at each test fre-quency, the LM data obtained for the fixed and random methods could be considered as demon-strating that the group of patients provided reli-able within-session LMs. The data were therefore

Frequency, Hz

Figure 6 Mean loudness matches expressed in dB SL. S = Session .

Pearson Product-Moment Correlations : Within- and Between-Session Reliability

For the final evaluation of LM reliability, Pearson is were calculated for all combinations of repeated LMs, both within and between ses-sions (Tables 10 and 11). As a standard of ref-

Table 6 Means of Loudness Match Differences across Subjects, Session 1 Minus Session 2, Expressed in dB SPL

Frequency (Hz) Fixed Random

1000 -2.30 -1 .70 1260 -1 .05 -1 .20 1580 -2.30 65 2000 1 .00 -1 .55 2520 -.55 -.05 3180 -.55 .55 4000 -1 .15 10 5040 -.40 -1 .10 6340 26 3.42 8000 -1 .37 .05

10,080 1 .00 .93 12,700 -.18 -1 .18 16,000 2.00 -1 .67

512


Table 7 Means of Loudness Match Differences across Subjects, Session 1 Minus

Session 2, Expressed in dB SL

Frequency (Hz) Fixed Random

1000 -1 .60 -.70 1260 -.50 -1 .10 1580 -1 .50 2 .25 2000 1 .10 75 2520 1 .25 65 3180 -.05 30 4000 0 .00 1 .15 5040 60 -.90 6340 79 47 8000 -.95 -.37

10,080 1 .08 50 12,700 -1 .45 -2.45 16.000 75 -2.50

erence, Pearson is were also calculated for the hearing thresholds obtained in each condition . Pearson is for all thresholds and LMs in dB SPL were significant at p <_ .0001 (indicating that the measures varied together systematically) . For LMs in dB SL, all Pearson is were significant at p < .01 except for those indicated in Tables 10 and 11 by the asterisks .

DISCUSSION

A s treatment protocols for tinnitus are becoming increasingly effective, the num-

ber of clinics offering treatment is correspond-ingly increasing . As part of the initial tinnitus evaluation, it is important to quantify the tin-nitus characteristics that a patient experiences as an acoustic sensation . However, a clinical method with documented reliability of responses has not emerged for this purpose . A series of ongoing projects, conducted in this laboratory,

Frequency, Hz

Figure 7 Means of individual differences in loudness matches (dB SPL), Session 1 minus Session 2.

Figure 8 Means of individual differences in loudness matches (dB SL), Session 1 minus Session 2.

is designed to systematically develop a clinical technique for matching an individual's tinnitus to acoustic stimuli.

A prototype was used to demonstrate the fea-sibility of the automated technique for tinnitus matching (Henry et al, 1996). For the present study, hardware was upgraded and computer programming rewritten to improve features, enable user control, and add flexibility to test-ing protocols. Despite these differences, for both studies, testing was repeated over two sessions and LMs were obtained at 12 of the same fre-quencies . Data analyses showed that LM response reliability was comparable between studies.

It is important to minimize redundancies in the testing protocol that might enable subject responses to be cued by the redundant features,

Table 8 Means of Loudness Match Differences (Absolute Values)

across Subjects, Session 1 Minus Session 2, Expressed in dB SPL*

Frequency (Hz) Fixed Random p

1000 4.60 5.80 1339 1260 4.25 3.80 6899 1580 4.90 3.65 4120 2000 3.70 3.15 .6065 2520 4.05 3.75 7748 3180 3.25 3.95 4672 4000 3.85 4.00 9110 5040 3.60 3.40 8112 6340 3.74 5.53 4373 8000 6.63 5.10 5024

10,080 5.00 2.93 2201 12,700 3 .09 3.18 9362 16,000 2 .25 3.33 8528

*p values are for t-tests between fixed and random means at each test frequency.

513


Table 9 Means of Loudness Match Differences (Absolute Values)

across Subjects, Session 1 Minus Session 2, Expressed in dB SL*

Frequency (Hz) Fixed Random P

1000 4.20 5.60 1569 1260 5.00 4.60 7013 1580 4.90 5.45 7911 2000 3.90 4.75 5293 2520 3.85 4.35 6618 3180 3.35 4.70 1932 4000 3.90 3.25 5778 5040 4.10 3.00 0469 6340 3.95 2 .58 1932 8000 5.89 4.89 6300

10,080 3.69 3.50 9999 12,700 4.00 3.36 6966 16,000 1 .75 4.17 1886

Frequency, Hz

Figure 10 Means of absolute values of individual dif-ferences in loudness matches (dB SL), Session 1 minus Session 2.

*p values are for t-tests between fixed and random means at each test frequency.

which can lead to spurious responses. This is a common concern for all audiologic testing, and sophisticated examiners will appropriately incor-porate randomization into their personal test-ing style. For threshold testing with the automated system, intertone intervals were ran-domized to reduce the possibility of subjects responding to fixed temporal patterns . During loudness matching, however, it was possible that the fixed level changes relative to thresh-old could lead subjects to respond on the basis of counting the number of tone presentations . A malingering patient could use such a strategy to provide spuriously consistent responses. It was therefore necessary to evaluate whether patients could respond with equal reliability when output-level step sizes were randomized .

Frequency, Hz

Figure 9 Means of absolute values of individual dif-ferences in loudness matches (dB SPL), Session 1 minus Session 2.

Two versions, fixed and random, of a com-puter-automated protocol for conducting tinni-tus loudness matching were evaluated. All data analyses were directed toward determining response reliability of the LMs. Results showed that the fixed and random methods both pro-vided reliable responses and that reliability was comparable between methods. Thus, either method could be used with equal confidence for tinnitus loudness matching . A further implica-tion of these results is that they support a method for loudness matching that has been used in a tinnitus clinic for many years (Vernon, 1982 ; Vernon and Fenwick, 1984 ; Vernon and Meikle, 1981, 1988).

The tinnitus loudness matching literature reveals that there is no standard procedure for obtaining LMs, and further that there is con-troversy regarding the need for conducting such tests. In the following discussion, some relevant points regarding these issues are discussed.

"Paradoxically Small" Tinnitus LMs

Early studies revealed a seemingly para-doxical relationship between patients' subjec-tively reported tinnitus loudness and the LMs that were typically less than 10 dB SL at the tin-nitus frequency (Fowler, 1943 ; Reed, 1960 ; Gra-ham and Newby, 1962). This was first noted by Fowler (1943), who concluded that tinnitus was a faint signal that should not be perceived as loud and therefore should not cause such distress . Numerous investigators suggested that these inordinately low-level LMs might be explained by the phenomenon of loudness recruitment (Vernon, 1976 ; Goodwin and Johnson, 1980; Tyler and Conrad-Armes, 1983 ; Meikle et al,

514


Table 10 Pearson is for Within-Session Thresholds and Loudness Matches

Session 1 Session 2

Frequency (Hz) Thresholds LMs dB SPL LMs dB SL Thresholds LMs dB SPL LMs dB SL

1000 992 946 832 982 964 930 1260 993 931 779 995 996 .966 1580 989 947 837 994 982 928 2000 958 977 740 992 962 896 2520 994 973 875 989 947 869 3180 995 .979 858 993 .974 919 4000 999 974 832 991 971 862 5040 984 985 903 995 988 921 6340 996 985 922 929 919 920 8000 990 940 748 992 992 811

10,080 985 950 565* 995 985 845 12,700 997 982 745 974 991 467* 16,000 995 996 366* 995 995 927

*p > 01 . All p's <_ .0001 except as indicated .

1984). This led to studies attempting to define more appropriate ways to measure tinnitus loud-ness . One way was to obtain LMs at a fre-quency(s) where hearing sensitivity was normal or near-normal, resulting in an increased dynamic range and larger LM relative to thresh-old (Goodwin and Johnson, 1980; Penner, 1984 ; ftisey et al, 1989 ; Tyler et al, 1992). Other stud-ies considered the premise that decibels are not a measure of loudness and used different approaches to transform tinnitus LMs into units of loudness (Hinchcliffe and Chambers, 1983 ; Tyler and Conrad-Armes, 1983 ; Hallam et al, 1985 ; Matsuhira et al, 1992). Penner (1986) pointed out the equivocal results of these efforts

and tested the idea independently of any math-ematical description of the loudness function . She concluded that the size of the LM bore a simple relationship with the slope of the loudness func-tion at any test frequency. Henry (1996) used direct loudness-growth measurements in patients with tinnitus in an attempt to correct the measurements, which was hypothesized to increase the correlation between the LMs and subjective tinnitus loudness ratings. Even after the corrections, the correlation remained small. The discrepancy between LMs and subjective judgments of tinnitus loudness remains unre-solved, and further work is needed to elucidate this relationship .

Table ll Pearson is for Between-Session Thresholds and Loudness Matches

Frequency (Hz) Thresholds

Fixed Method

LMs dB SPL LMs dB SL Thresholds

Random Method

LMs dB SPL LMs dB SL

1000 974 953 859 991 903 689 1260 990 954 842 994 912 719 1580 993 936 847 966 958 705 2000 991 967 904 959 975 754 2520 986 963 899 990 978 848 3180 990 977 896 985 969 764 4000 986 976 847 991 964 833 5040 987 978 883 984 974 873 6340 992 974 856 923 906 903 8000 980 889 648 982 958 619

10,080 990 965 809 990 984 792 12,700 986 919 371* 995 986 666* 16.000 995 993 897 982 997 358*

*p > 01 . All p's <_ .0001 except as indicated .

515


Implications for Assessment of Tinnitus Malingering

In the present study, obtaining LMs at the series of test frequencies effectively created LM functions (curves). Meikle et al (1996) reviewed data from 121 patients (from four separate stud-ies) and reported that consistent LM curves emerged despite different measurement tech-niques . Thus, when LMs were plotted as dB SL against frequency, as also shown in the present study (see Fig. 6), there was a majority of monot-onically decreasing functions ranging from large negative slopes to slopes near zero . An addi-tional group of patients had LM curves that could not be fit with a linear regression function. These cumulative data showed that LM curves can be used to separate the clinical population into several major categories according to the shapes of their LM curves . Such categories may have some yet undiscovered relationship to tin-nitus etiology. These curves may have further value in that their distinctiveness for each patient may serve as a tinnitus "fingerprint," which could be useful in assessing the validity of compensation claims.

The use of randomized step increases dur-ing loudness matching is particularly impor-tant in the design of a "tinnitus malingering" test that requires the patient to provide reliable responses. Such a test has not yet been formal-ized, but for medicolegal purposes, repeated LMs have been the basis for documenting the presence of a plaintiff's tinnitus (Vernon, 1996). The test is described as "five or six" repeated LMs at the tone that is best matched to the tinnitus pitch. The tone is presented in 1-dB ascending increments, and if the results of all repeated tests agree to within 2 dB of each other, that is con-sidered evidence for the presence of tinnitus . (These tests are not repeated consecutively but rather interspersed among other tinnitus and audiometric tests over a period of about 1 hour.)

Jacobson and Henderson (1999) recently conducted a study evaluating the ability of lis-teners without tinnitus to imagine a constant tonal tinnitus and attempt to match the tinni-tus loudness with the loudness of tones. These subjects were seen to provide LMs that were at a higher level than actual tinnitus patients, but within- and between-session reliability of the LMs was not significantly different. These results led the authors to conclude that validating a patient's tinnitus may depend more upon the overall LM levels than upon the reliability of the LMs.

For the tinnitus malingerer, it would be expected to present a greater challenge to pro-vide reliable LMs at multiple frequencies versus at a single . frequency. The preliminary results of Jacobson and Henderson (1999), however, sug-gest that loudness matching may be repeatable even without tinnitus . If such an ability exists for the average person, it might be surmised that normal listeners can produce equal loudness contours (Robinson and Dadson, 1956) without the use of a reference tone . There is some evi-dence for this in the saucer-shaped audiogram that some authors have suggested is typically seen with the patient who provides false or exag-gerated hearing thresholds (Doerfler, 1951; Carhart, 1958; Goetzinger and Proud, 1958). Such an audiogram appears similar to an equal loudness contour, which would suggest that patients are able to use some kind of internal ref-erence to produce responses at equal loudness levels across frequencies (Martin, 1994). Chaik-lin et al (1959), however, reported that saucer-shaped audiograms are seen infrequently in cases of functional hearing loss .

It thus remains to be determined if a tinni-tus malingering test can be based on repeating LMs at a series of frequencies . For the tinnitus malingerer, it may present a greater challenge to provide reliable LMs at multiple frequencies, especially if the output-level step sizes are ran-domized. Results of the present study will there-fore be used to design a similar test to specifically evaluate tinnitus malingering by establishing a tinnitus "fingerprint" that should be easily repeatable only if tinnitus is present.

Reliability of Tinnitus LMs

The literature concerning the repeatability of tinnitus LMs is sparse . Tinnitus clinicians have reported that within-session repeated LMs within ± 1 dB could be routinely obtained from their patients (Bailey, 1979 ; Vernon et al, 1980). Goodwin and Johnson (1980) reported "extremely consistent" within-session LM data for nine patients . Using traditional psychoacoustic pro-cedures, Penner (1983) and Burns (1984) each reported high variability in responses across sessions . Penner reported that the within-session responses had "substantially smaller" variabil-ity. Variability of responses would be expected to be smaller within than between sessions because of tinnitus fluctuations over time . Pen-ner postulated, however, that patients may be using auditory memory within a session to provide consistent matches.

516


Penner and Bilger (1992) used two psy-choacoustic procedures to match tinnitus loud-ness and pitch : a method of adjustment and a forced-choice double-staircase procedure . For obtaining LMs, the within-session variability of the two methods was comparable and simi-lar to the variability when using the same pro-cedures but with a pure tone that simulated the tinnitus . It was concluded that tinnitus could be matched reliably within a session but that measurements between sessions were not reliable .

Mitchell et al (1993) used a computer-assisted technique to obtain tinnitus loudness and pitch matches and tinnitus masking curves . The test-retest variability of LMs using this procedure was reported to agree with the small variability reported by some (Bailey, 1979 ; Ver-non et al, 1980), but not with the large variations reported by others (Penner, 1983 ; Burns, 1984) .

Henry and Meikle (1999) conducted a study in which tinnitus LMs were measured in 26 subjects with chronic tinnitus using both pulsed and continuous tones . LMs were determined at 11 frequencies between 0.5 to 10 kHz . No sig-nificant differences were found between pulsed versus continuous measures ; thus, the results could be used as a demonstration of within-ses-sion reliability. With an intertest interval of 10 to 15 minutes, these subjects were capable of reproducing their LMs to within ±2 to 3 dB .

The study evaluating the prototype auto-mated technique for reliability of tinnitus LMs was discussed above (Henry et al, 1996) . The pre-vious and current studies reported Pearson is to evaluate between-session reliability of LMs. The Pearson is were comparable between stud-ies, and almost all were statistically significant . The present study thus corroborates results of the preliminary study.

It is clear that there are few studies from which to draw conclusions about the test-retest reliability of tinnitus LMs . Since each study has used somewhat different procedures to obtain LM measurements, results between studies can-not be easily compared . A standardized procedure would facilitate the comparison of measure-ments between investigators.

Tinnitus LMs Expressed in dB SPL or dB SL

The present study reports all LM data in both dB SPL and dB SL . Tables 10 and 11 reveal a clear difference between the two metrics when evaluating test-retest reliability of LMs. LMs

expressed in dB SPL were consistently more reliable than their dB SL counterparts, both within and between sessions. The Pearson is for the dB SPL loudness matches were almost uni-versally > 0 .90 at each test frequency (all p's <- .0001), and correlations were comparable to those obtained for hearing thresholds . Reli-ability of LMs expressed in dB SL, however, was generally below 0.90, with correspondingly poorer p values .

In the interpretation of these apparent reli-ability differences, it is known that narrowing the range of scores can markedly reduce the correlation between two variables (Welkowitz et al, 1976) . When the LM measures are trans-formed to the dB SL metric, the variability of the scores is restricted for both variables, which effectively decreases the correlation between the two . A further factor that could reduce this correlation is that the dB SL metric introduces a second source of variability : hearing thresh-olds . Thus, determining the correlation of repeated LMs transformed to dB SL uses a sin-gle measure derived from two measures that each vary from test to test .

For purposes of obtaining reliable tinnitus LMs, the present data indicate that express-ing the measures in dB SPL will provide good response reliability. This seems to be true regardless of which frequency is used to obtain the measures . This point has relevance to any application in which tinnitus LMs must be obtained serially, such as determining the effect of treatment on tinnitus loudness or evaluating for malingering by obtaining repeated measures .

Clinical Value of Tinnitus Matching Tests

For clinical application, it is controversial whether tinnitus matching has any diagnostic value (Jastreboff, 1996). However, which side one takes in this issue may depend upon the type of treatment endorsed . If the treatment is mask-ing, which is still routinely used in many clin-ics, identifying the spectrum of the tinnitus may have relevance for determining the bandwidth of the masking noise. This enables the applica-tion of "minimal masking," which would tailor the masking noise to the individual's tinnitus spectrum and level (Penner, 1983 ; Vernon, 1997). The effect would be similar to the critical band-width concept in which a restricted bandwidth of noise, centered around a pure-tone frequency, is sufficient to effectively mask the pure tone (Fletcher, 1940 ; Hawkins and Stevens, 1950 ;

517


Bilger and Hirsh, 1956; Goldstein and New-man, 1994). The masking approach to tinnitus includes the use of partial masking to provide relief, and the minimal masking concept would apply to both complete and partial masking (Vernon, 1997). Minimal masking would reduce the amount of sound presented to the patient, which would assist in making the masking noise more acceptable to the patient.

Acknowledgment. Funding for this study was provided by Veterans Affairs Rehabilitation Research and Devel-opment (RR&D) Service.

REFERENCES

Bailey Q. (1979) . Audiological aspects of tinnitus . Aust J Audiol 1:19-23 .

Bilger RC, Hirsh LJ . (1956) . Masking of tones by bands of noise. JAcoust Soc Am 28:623-630.

Tinnitus Test Battery

General guidelines for the clinical evalua-tion of tinnitus were established at a symposium sponsored by the CIBA Foundation in London (Evered and Lawrenson, 1981). They recom-mended a battery of tests to include pitch match-ing, loudness matching, tinnitus maskability, and residual inhibition . Vernon and Meikle (1981) provided procedural details for conduct-ing these tests. These recommendations are still appropriate for today, but standardization of the procedures has not been accomplished .

There are at least two major concerns for development of a standardized clinical mea-surement technique. First, the method must be conducted rapidly so that all measurements can be obtained within a period of time that is appro-priate for a typical clinical schedule . For tinni-tus patients, a series of tinnitus measurements in one ear should not require over 30 minutes. Second, within-patient measurements must be shown to be consistent over repeated tests. Between-patient reliability is not currently a concern because standard measurement tech-niques have not been established; thus, popu-lation norms are unknown.

Future Work

The present study provides further docu-mentation of the reliability of an automated technique to measure tinnitus LMs and of the comparability of results using the fixed and ran-dom methods. Future work will be directed toward (1) reducing testing time while main-taining optimal test-retest reliability of responses; (2) using more complex stimuli for tin-nitus matching when the tinnitus is not tonal; (3) developing a "tinnitus malingering" test, possibly based on tinnitus LM curves and ran-domization of different testing parameters ; and (4) conducting clinical trials to document the utility of the technique in the clinical setting. Hopefully, this research will significantly impact the clinical management of tinnitus as well as tinnitus research involving humans .

Burns EM. (1984) . A comparison of variability among measurements of subjective tinnitus and objective stim-uli . Audiology 23:426-440 .

Carhart R . (1958) . Audiometry in diagnosis . Laryngo-scope 68:253-279 .

Carhart R, Jerger JF. (1959) . Preferred method for clin-ical determination of pure-tone thresholds. J Speech Hear Disord 24:330-345 .

Chaiklin JB, Ventry IM, Barrett LS, Skalbeck GA . (1959) . Pure-tone threshold patterns observed in functional hear-ing loss . Laryngoscope 69:1165-1179 .

Coles RRA, Baskill JL. (1996) . Absolute loudness of tin-nitus : tinnitus clinic data . In : Reich GE, Vernon JA, eds . Proceedings of the Fifth International Tlnnitus Seminar 1995 . Portland, OR: American Tinnitus Association, 135-141 .

Doerfler LG. (1951). Psychogenic deafness and its detec-tion . Ann Otol Rhinol Laryngol 60:1045-1048 .

Evered D, Lawrenson G (1981) . Appendix II. Guidelines for recommended procedures in tinnitus testing. In : Evered D, Lawrenson G, eds. CIBA Foundation Sympo-sium 85 . 7Ennitus. London : Pitman, 303-306.

Fausti SA, Frey RH, Henry JA, Knutsen JM, Olson DJ. (1990) . Reliability and validity of high-frequency (8-20 kHz) thresholds obtained on a computer-based audiome-ter as compared to a documented laboratory system . J Am Acad Audiol 1:162-170.

Fletcher H. (1940) . Auditory patterns . Rev Mod Phys 12:47-65 .

Fortune DS, Haynes DS, Hall JWI. (1999) . Tinnitus: cur-rent evaluation and management. Otolaryngol Internist 83:153-162 .

Fowler EP (1943) . Control of head noises . Their illusions of loudness and of timbre . Arch Otolaryngol 37:391-398 .

Goetzinger CP, Proud GO . (1958) . Deafness : examina-tion techniques for evaluating malingering and psychogenic disabilities . J Kans Med Soc 59:95-101.

Goldstein B. (1997) . Psychophysical and psychoacoustic correlates of tinnitus . In : ShulmanA, ed . TEnnitus Diag-nosis lTreatment. San Diego: Singular, 99-115 .

Goldstein BA, Newman CW (1994) . Clinical masking: a decision-making process. In : Katz J, ed . Handbook of Clinical Audiology. Baltimore: Williams and Wilkins, 109-131.

Goodwin PE, Johnson RM. (1980) . The loudness of tin-nitus. Acta Otolaryngol 90:353-359 .

518

Graham JT, Newby HA. (1962) . Acoustical characteris-tics of tinnitus : an analysis . Arch Otolaryngol 75 :82-87 .

Hallam RS, Jakes SC, Chambers C, Hinchcliffe R. (1985) . A comparison of different methods for assessing the ̀ inten-sity' of tinnitus . Acta Otolaryngol (Stockh) 99:501-508 .

Hawkins JEJ, Stevens SS . (1950) . The masking of pure tones and of speech by white noise . J Acoust Soc Am 22:6-13.

Hazell JW Wood SM, Cooper HR, Stephens SD, Corco-ran AL, Coles RR, Baskill JL, Sheldrake JB . (1985) . A clinical study of tinnitus maskers. Br JAudiol 19:65-146.

Henry JA. (1996) . Loudness recruitment only partially explains the small size of tinnitus loudness-matches . In : Reich GE, Vernon JA, eds. Proceedings of the Fifth Inter-national Tinnitus Seminar 1995. Portland, OR : American Tinnitus Association, 148-157.

Henry JA, Fausti SA, Mitchell CR, Flick CL, Helt WJ . (1996) . An automated technique for tinnitus evaluation . In : Reich GE, Vernon JA, eds. Proceedings of the Fifth International Tinnitus Seminar 1995. Portland, OR: Amer-ican Tinnitus Association, 325-326.

Henry JA, Meikle MB. (1999) . Techniques for measur-ing the loudness of tinnitus : pulsed versus continuous stimuli. J Am Acad Audiol 10:261-272 .

Hinchcliffe R, Chambers C. (1983) . Loudness of tinnitus : an approach to measurement. Adv Otorhinolaryngol 29 :163-173 .

Ishikawa T, Kusano H, Ogasawara M, Murai K, Tauiki T. (1991) . Investigation of tinnitus tests : comparison of ipsilateral and contralateral testing. In : Aran J-M, Dauman R, eds. Fourth International Tinnitus Seminar. Bordeaux, France : Kugler, 43-48.

Jacobson GP, Henderson JA . (1999) . Toward the devel-opment of a subjective measure of the presence of tinnitus : preliminary observations . In : Popelka GR, ed . Associa-tion for Research in Otolaryngology [abstract 682] . Mt. Royal, NJ : Association for Research in Otolaryngology.

Jastreboff PJ . (1996) . Usefulness of the psychoacousti-cal characterization of tinnitus . In : Reich GE, Vernon JA, eds. Proceedings of the Fifth International Unnitus Sem-inar 1995. Portland, OR: American Tinnitus Association, 158-166.

Kuk FK, Tyler RS, Rustad N, Harker LA, Tye-Murray N. (1989). Alternating current at the eardrum for tinni-tus reduction. J Speech Hear Res 32:393-400 .

Martin FN . (1994) . Pseudohypacusis . In : Katz J, ed . Handbook of Clinical Audiology. Baltimore: Williams and Wilkins, 553-567.

Matsuhira T, Yamashita K, Yasuda M. (1992) . Estima-tion of the loudness of tinnitus from matching tests. Br JAudiol 26:387-395 .

Meikle MB, Henry JA, Mitchell CM . (1996) . Methods for matching the loudness of tinnitus using external tones. In : Reich GE, Vernon JA, eds. Proceedings of the Fifth International Tinnitus Seminar 1995. Portland, OR: Amer-ican Tinnitus Association, 178-179.

Meikle MB, Johnson RM, Griest SE, Press LS, Charnell MG . (1995) . Oregon TFnnitus Data Archive. World Wide Web: http//www.ohsu.edu/ohrc-otda/> 95-01 .


Meikle MB, Vernon J, Johnson RM. (1984) . The perceived severity of tinnitus . Otolaryngol Head Neck Surg 92:689-696 .

Meikle M, Walsh ET. (1984) . Characteristics of tinnitus and related observations in over 1800 tinnitus patients . J Laryngol Otol 9(Suppl) :17-21 .

Mitchell CR, Vernon JA, Creedon TA . (1993) . Measuring tinnitus parameters : loudness, pitch, and maskability. J Am Acad Audiol 4:139-151 .

Murai K, Tyler RS, Harker LA, Stouffer JL. (1992) . Review of pharmacologic treatment of tinnitus . Am J Otol 13:454-464 .

Penner MJ . (1983) . Variability in matches to subjective tinnitus . J Speech Hear Res 26:263-267 .

Penner MJ . (1984) . Equal-loudness contours using sub-jective tinnitus as the standard . J Speech Hear Res 27:274-279 .

Penner MJ. (1986) . Magnitude estimation and the "para-doxical" loudness of tinnitus . J Speech Hear Res 29:407-412 .

Penner MJ, Bilger RC. (1992) . Consistent within-session measures of tinnitus . J Speech Hear Res 35:694-700 .

Reed GF. (1960) . An audiometric study of two hundred cases of subjective tinnitus . Arch Otolaryngol 71 :94-104.

Risey J, Briner W, Guth PS, Norris CH . (1989) . The supe-riority of the Goodwin procedure over the traditional procedure in measuring the loudness level of tinnitus . Ear Hear 10:318-322 .

Robinson DW Dadson RS . (1956). A redetermination of the equal loudness relations for puretones. Br J Appl Phys 7:166-181 .

Roeser RJ, Price DR . (1980) . Clinical experience with tin-nitus maskers. Ear Hear 1 :63-68 .

Scott B, Lindberg P, Lyttkens L, Melin L. (1985) . Psy-chological treatment of tinnitus . An experimental group study. Scand Audiol 14:223-230.

Tyler RS, Aran J-M, Dauman R. (1992) . Recent advances in tinnitus . Am JAudiol 1 :36-44 .

Tyler RS, Conrad-Armes D. (1983). The determination of tinnitus loudness considering the effects of recruit-ment. J Speech Hear Res 26:59-72 .

Vernon JA. (1976) . The loudness (?) of tinnitus . Hear Speech Action 44:17-19 .

Vernon JA . (1982) . Relief of tinnitus by masking treat-ment . In : English GM, ed . Otolaryngology. Philadelphia : Harper and Row, 1-21 .

Vernon JA . (1987) . Assessment of the tinnitus patient. In : Hazell JWP, ed . Tinnitus New York : Churchill Liv-ingstone, 71-87.

Vernon JA . (1996) . Is the claimed tinnitus real and is the claimed cause correct? In : Reich GE, Vernon JA eds. Proceedings of the Fifth International Unnitus Seminar

519


1995. Portland, OR : American Tinnitus Association, 395-396 .

Vernon JA. (1997) . Common errors in the use of mask-ing for relief o£ tinnitus . In : Shulman A, ed . Tinnitus : Diagnosis/ Treatment. San Diego: Singular, 50-66.

Vernon JA, Fenwick J. (1984) . Identification of tinnitus : a plea for standardization. J Laryngol Otol 9(Suppl) :45:45-53 .

Vernon JA, Meikle MB . (1981) . Tinnitus masking: unre-solved problems . In : Evered D, Lawrenson G, eds. CIBA Foundation Symposium 85 . Tinnitus . London : Pitman, 239-256.

Vernon JA, Meikle MB . (1988) . Measurement of tinni-tus : an update . In : Kitahara M, ed . Tinnitus, Pathophysiology and Management . Tokyo : Igaku-Shoin, 36-52 .

Vernon JA, Johnson RM, Schleuning AJ, Mitchell CR . (1980) . Masking and tinnitus . Audiol Hear Educ Summer:5-9 .

Welkowitz J, Ewen RB, Cohen J. (1976) . Introductory Statistics for the Behavioral Sciences. New York : Acade-mic Press.

Reliability of Tinnitus Loudness Matches under Procedural ...Reliability of Tinnitus Loudness Matches under Procedural Variation James A. Henry*t Christopher L. Flick* Alison Gilbert*

Documents

Reliability of Tinnitus Loudness Matches under Procedural ...Reliability of Tinnitus Loudness Matches under Procedural Variation James A. Henryt Christopher L. Flick Alison Gilbert*