Share this article on:

A few secrets about bone-conduction testing

Margolis, Robert H.

doi: 10.1097/01.HJ.0000368588.05083.17
Page 10

Since the earliest testing using tuning forks, assessment of bone-conducted hearing sensitivity has been an essential component of audiometric differential diagnosis. It was Bekesy, in 1932, who demonstrated that it was possible to cancel a bone-conducted tone by introducing an air-conducted tone of the same frequency, but with different phase. This supported the notion that the two signals had the same mechanical displacement patterns at the basilar membrane—a topic of some debate at the time. In the 1960s Tonndorf added to Bekesy's work by providing a detailed description of three different modes of bone-conduction transmission.

From a clinical standpoint, bone-conduction testing has pretty much always been a routine part of an audiologic evaluation (at least since these evaluations have been called “audiologic”). Today, the testing is so routine that it sometimes is conducted in a rather casual manner. When immittance findings are normal, maybe it's not conducted at all. But like all things that we do over and over, it's useful to step back and take a critical look at what we are actually doing. We've found an author to provide us that perspective on bone-conduction testing for this month's Page Ten.

Robert H. Margolis did his early audiologic training at Kent State University before earning his PhD from the University of Iowa. He worked at the UCLA Department of Otolaryngology and the Syracuse University Department of Communication Sciences and Disorders before joining the University of Minnesota Department of Otolaryngology as professor of audiology in 1988. His research has focused on developing improved methods for hearing assessment, including acoustic immittance, electrocochleography, and pure-tone and speech audiometry.

In 2000 he established Audiology Incorporated to develop and commercialize automated hearing tests. His work on automated audiometry has been supported by small business technology transfer grants from the National Institutes of Health. He is currently a collaborator on the NIH Toolbox project ( to provide a standard hearing assessment tool for large epidemiologic and clinical outcome studies.

When Bob isn't thinking about bone-conduction secrets or automated testing, you can find him on the tennis court (where he prefers to serve wide to the ad court) or carefully pouring a dark Belgian beer (which he prefers to serve in the middle of the backcourt). For now, Dr. Margolis is serving up some useful information about that old test of ours, bone conduction. He's provided a few things to think about that might alter your clinical practices.


Page Ten Editor

Back to Top | Article Outline

1 Aren't you the guy who published an article on bone conduction called “Audiology's Dirty Little Secret”? Why are you airing audiology's dirty laundry in public?

That was me, but I was not referring to bone-conduction in general, just one or two specific little secrets. Frequently during my short 35-year career in this field, I have heard complaints about air-bone gaps and bone-air gaps (bone-conduction threshold worse than air-conduction threshold) that don't fit the patient's audiometric picture and therefore must be wrong. That is, you see a significant air-bone gap when you're pretty convinced it's a completely sensorineural hearing loss. Or even more puzzling, bone scores that are 10-15 dB worse than air scores.

In the development of AMTAS®, an automated pure-tone test (see, we have noticed these “errors” even more frequently than we see them in manual audiometry. So I've been trying to get to the bottom of it.

Back to Top | Article Outline

2 But doesn't that simply mean that audiologists get more accurate thresholds than automated procedures?

Maybe–but maybe not. Let's refresh our memory on the variability associated with air-bone gaps. This is not a new topic. Studebaker addressed this very issue in an attempt to clarify that when there is variability associated with a measurement, the measured value doesn't always land on the mean.1 He modeled the air-bone gap as a normally distributed variable with a standard deviation of 5 dB. His model predicts that in the absence of a conductive hearing loss the air-bone gap is 0 dB only 38% of the time. If four frequencies are tested, air-bone gaps at all frequencies would be zero only 2% of the time and gaps of 10, 15, and 20 dB are expected to occur much more often than you might think.

Back to Top | Article Outline

3 Interesting. Do you have anything to contribute to the explanation?

I have a friendly amendment to the Studebaker model, one that was suggested to me by my friend and colleague Aaron Thornton. Aaron pointed out that the air-bone gap is a normally distributed variable, as Studebaker told us, but it is the difference between two normally distributed variables (the air- and bone-conduction thresholds) that have different variances. That produces a normally distributed variable with variability that is greater than either air conduction or bone conduction alone. If we assume standard deviations of 5 dB for air-conduction thresholds and 7 dB for bone-conduction thresholds, the standard deviation of the air-bone gap is 8.6 dB. The new model predicts air-bone gaps of 0 dB for only 21% of thresholds and gaps of 10 dB or more (air-bone gaps and bone-air gaps) almost half the time (48% of thresholds).

Let's look at the example in Figure 1. This audiogram was obtained from a patient who is an experienced listener, having participated in many research studies. The audiometer was calibrated to the ANSI standard (S3.6-2004)2 and bone conduction was tested with the vibrator on the forehead, using the appropriate mastoid-forehead corrections. Ignoring 4000-Hz for a moment, the air-bone gaps are well within the expected variability for patients with sensorineural hearing loss. But it looks a little sloppy, right? Note that AMCLASS®, our validated audiogram classification system, called the hearing loss in the right ear a mixed loss because of the air-bone gap at 4000 Hz.3-5 If this was a manual audiogram and I had any doubt about the bone-conduction thresholds, I would be tempted to nudge them toward smaller air-bone gaps, especially if I had already obtained normal immittance results.

Back to Top | Article Outline

4 I agree. It doesn't look quite right.

But wait, I wasn't finished telling the story. Now let's look at another example in Figure 2. What's wrong with this audiogram? Nothing, right? I know from previous audiograms, patient history, tympanometry, and otoscopy that the patient has a sensorineural hearing loss. But the likelihood, based on variability of air-conduction and bone-conduction testing that the patient has 0-dB air-bone gaps at four frequencies in each ear is 1 in 250,000. Let me repeat that: 1 in 250,000!

Most of us should not see an audiogram like this in our professional careers. The audiogram in Figure 1 is plausible. The one in Figure 2 is almost certainly fudged. But, if everything matches nicely, the ENT doctor wouldn't walk back to the booth and question it. (It's our dirty little secret.) I know, you're thinking that you've seen many audiograms conducted by competent audiologists where everything matches up nicely. What I'm saying is that audiologists don't always report air-bone gaps that occur in patients with sensorineural hearing loss.

Back to Top | Article Outline

5 So you're saying that audiologists are inherently dishonest?

I would never say that. I am saying that bone conduction is a biased experiment. When we are testing bone conduction we almost always have an idea of what the result is going to be. We get these premonitions from previous audiograms, thresholds at other frequencies, other test results like immittance and otoscopy, and patient history. If all these sources of information point toward sensorineural hearing loss, the audiologist is biased toward recording bone-conduction values that are equal to air-conduction thresholds.

Back to Top | Article Outline

6 But if audiologists are honest people, why would they report inaccurate bone-conduction thresholds?

Two reasons. First, it is the nature of bias that we don't always recognize that our behavior is affected by biasing factors. There is literature on effects of bias on human behavior that shows that performance is affected even when the person is aware of the potential for bias. See, for example, Messick and Sentis.6

Second, as an audiologist, I want to communicate the correct status of the patient's hearing. If I know in my heart that the patient has a sensorineural hearing loss, I am less likely to report air-bone gaps that are expected to occur as a result of the inherent variability of the measurements. I might be concerned that an otolaryngologist would order further testing and/or follow-up appointments because of this “apparent” conductive loss. And, of course, we all have been taught that bone-air gaps are theoretically impossible, so I am biased toward under-reporting those when they occur even though they are the expected result of the variability of the measurements.

There is a good reason that clinical trials are designed as double-blind experiments. When people have prior knowledge of the expected results, the outcomes are different from when they have no prior knowledge. And this is true with the most honest, ethical humans on earth–all of whom are audiologists!

Back to Top | Article Outline

7 You started out saying that air-bone gaps in patients with sensorineural hearing loss occur more frequently with automated tests. How do you account for that?

It's very simple. AMTAS and other automated tests are not biased. They don't care if there is an air-bone gap or a bone-air gap. They report the results from the patients' behavior uninfluenced by any expectations. They haven't read Studebaker's article and they don't have any dirty little secrets.

Back to Top | Article Outline

8 Okay, I'll buy that. I've heard about erroneous air-bone gaps at 4000 Hz. Is this just something related to the variability you've been talking about?

Well, not entirely. There's something else going on at 4000 Hz. Look at the audiogram in Figure 3. This patient has a sensorineural hearing loss. The air-bone gaps at 4000 Hz are unlikely to be related to variability. You can prove that by testing the patient repeatedly. If you get the same air-bone gap all the time it's not the result of variability. Note that AMCLASS wants to call the hearing loss mixed in both ears. AMCLASS was validated against the judgments of expert audiologists. Unless we ignore the air-bone gaps at 4000 Hz, the hearing loss is mixed in both ears. But it's not.

Back to Top | Article Outline

9 So where did that big air-bone gap come from?

My guess is that our bone-conduction, reference-equivalent threshold force level at 4000 Hz is wrong. But in the study that was the source of the standard bone-conduction thresholds, subjects with sensorineural hearing loss were tested at three locations to derive bone-conduction, reference-equivalent threshold force levels (RETFL) that would produce 0-dB air-bone gaps (on average).7

Then to verify that the values were correct, a new group of subjects with sensorineural hearing loss was tested and, sure enough, their air-bone gaps averaged 0 dB at all test frequencies, including 4000 Hz. The threshold levels were incorporated into the audiometer standard and we began calibrating bone conduction to those levels. But soon audiologists began noticing air-bone gaps at 4000 Hz in patients with sensorineural hearing loss.

Back to Top | Article Outline

10 Is it possible that the patient can hear air-conduction radiation from the bone vibrator at 4000 Hz and that causes bone conduction to be better than it should be?

That explanation continues to be kicked around, but it was debunked early on by Frank and Holmes (1981).8 They tested bone conduction in subjects with ears open and ears plugged and got no difference at 4000 Hz.

In an AMTAS validation study we found the same unexplained air-bone gap at 4000 Hz with the ears covered by circumaural earphones and the bone vibrator on the forehead.8 If you can block the ear with an earplug as Frank and Holmes did or with a sound-attenuating muff as we did and the air-bone gap remains, it is not due to acoustic radiation.

Back to Top | Article Outline

11 But a lot of my friends don't get air-bone gaps at 4000 Hz except when the patient has middle ear disease. Why isn't it always there?

Here's another dirty little secret. Rumor has it that some calibration services have become tired of hearing complaints about 4000-Hz air-bone gaps and calibrate bone levels at that frequency off standard. And they don't always tell us they are doing that.

It may be a reasonable solution because the source of the problem appears to be an incorrect RETFL at 4000 Hz. But if this really happens, they shouldn't do it without telling us.

Back to Top | Article Outline

12 Is that legal?

I'm glad you asked that question. Some state licensure laws require that testing be performed with a calibrated audiometer. That implies that levels are calibrated to the audiometer standard. If we have a good reason for doing it and we have data to support it, we are probably on safe ground if we use a different reference level. But we should do it with our eyes open.

Back to Top | Article Outline

13 Do we actually have the data?

We have some, but they are conflicting. The Dirks et al. data show no air-bone gap at 4000 Hz6 and our data show a 12-dB air-bone gap for manual testing and 22-dB air-bone gap for automated testing.9 I suspect the difference is due to bias in manual testing. We are testing a new group of sensorineural hearing loss subjects now to shed more light on it. We hope to report the results early this year.

Back to Top | Article Outline

14 If 4000-Hz bone conduction is such a problem, why don't we just skip it?

I don't think that would be a good idea. High-frequency air-bone gaps can be clinically important. Look at the audiogram in Figure 4. This patient came in with a complaint of aural fullness in the left ear. Her 226-Hz tympanogram was normal. High-frequency hearing losses like this are usually sensorineural and it would be easy to send this person away without further evaluation.

But her 1000-Hz tympanogram was flat and an otomicroscopic examination revealed middle ear effusion. The hearing loss and abnormal high-frequency tympanogram are consistent with mass loading of the middle ear. She was treated for otitis media and the hearing loss resolved. Middle ear effusion in adults can be a sign of serious conditions such as nasopharyngeal carcinoma. I wouldn't want to miss this case. I think we should do more high-frequency bone-conduction testing, not less.

Back to Top | Article Outline

15 How do you know that the 4000-Hz air-bone gap in this case is real rather than the same erroneous finding you've been talking about?

Good question, and it's not always an easy distinction. In this case, the 3000-Hz bone threshold, the high-frequency tympanogram, and the careful ear examination confirmed that there was a real high-frequency conductive hearing loss. The case illustrates the importance of getting our 4000-Hz bone thresholds right. We don't have a definitive answer yet, but we should all know how our audiometers are calibrated and use our diagnostic skills to interpret these cases appropriately.

Back to Top | Article Outline

16 If tympanograms and acoustic reflexes are normal, do we really have to test bone conduction at all?

Yes. Let's look at the case in Figure 5. This patient had fluctuating hearing loss and vertigo and, based on her symptoms, could easily be diagnosed with Ménière's disease. In spite of normal immittance findings, there is an air-bone gap.

My colleague Lisa Hunter followed a group of patients like this for several years before Rosowski and his colleagues at Harvard-MIT explained that patients with dehiscent superior semicircular canals behave just like this.10, 11 The enhanced bone-conduction sensitivity is explained by a “third window” effect in which the opening of the bony labyrinth into the subdural space results in greater cochlear stimulation by bone conduction.

In some cases with normal immittance findings and a documented history of sensorineural hearing loss, bone-conduction testing may be unnecessary. But more information is always better if you want to understand your patient's hearing. If you don't do it you won't see the surprising cases that may teach us something. And you may make the wrong diagnosis.

Back to Top | Article Outline

17 Let's back up a moment. Earlier you mentioned forehead placement of the bone-conduction vibrator. Is that a new way to test bone conduction? How does it compare to mastoid placement?

It was recognized very early in the development of hearing testing that it doesn't really matter where you place the bone vibrator. Forehead placement has been around for decades and the reference equivalent threshold force levels are in the standard. It takes roughly 10 dB more force to reach threshold with forehead placement than with mastoid placement. For many years that was a problem because it restricted the maximum hearing levels that could be produced. Current audiometers and bone vibrators can reach higher levels, so it's no longer a problem.

Back to Top | Article Outline

18 So why do most people continue to use mastoid placement?

The common use of mastoid placement is related more to the earphone used to provide masking to the non-test ear than to the bone vibrator. The supra-aural earphones that are most commonly used produce a large occlusion effect in the low frequencies that shifts the bone-conduction thresholds. That's why we uncover the test ear during masked bone-conduction testing. Then we have to turn the whole arrangement around to test the other ear. When you use occluding earphones there is no advantage to forehead bone because you still have to rearrange the transducers when you switch ears to leave the non-test ear uncovered.

But if we had an earphone that didn't produce an occlusion effect, we could place the bone-conduction vibrator on the forehead, earphones over both ears, and test air conduction and bone conduction without moving the transducers. This would make manual testing more efficient and make automated testing possible. That's how AMTAS works.

There is a tendency to think that when we place the vibrator on the right mastoid we are testing the right ear. Of course, we learn in basic audiology that there is no interaural attenuation for bone conduction, but inexperienced testers may forget that. With forehead bone conduction it is clear that you only know which ear heard the tone if you properly mask the non-test ear.

Back to Top | Article Outline

19 I'm liking the idea of using forehead placement. Is there a good non-occluding earphone that can be used for air-conduction testing and for masked bone?

There certainly is. The Sennheiser HDA200 earphones that many clinics use for extended high-frequency testing can be used for the conventional frequency range as well. The reference equivalent threshold sound pressure levels are in the standard so it can be calibrated and used for air-conduction testing. It is also a thousand times more comfortable than supra-aural earphones and provides good ambient noise attenuation so you can test in any reasonably quiet space–not just sound rooms.

Back to Top | Article Outline

20 Mastoid bone or forehead bone. Automated testing or manual testing. Supra-aural earphones or circumaural earphones. There are some choices in pure-tone testing, aren't there?

Yes there are, but there has been very little innovation, despite the fact that audiologists spend more time performing pure-tone audiometry than any other single activity. We have textbooks that have been teaching the same method for decades and we have an audiometer standard that stifles innovation. We can do better. We need to think outside the box.

Back to Top | Article Outline


1. Studebaker GA: Intertest variability and the air-bone gap. J Sp Hear Dis 1967;32:82–86.
2. American National Standards Institute: ANSI S3.6-2004, American National Standard Specification for Audiometers. New York: Acoustical Society of America, 2004.
3. Margolis RH, Saly GL: Toward a standard description of hearing loss. IJA 2007;46:746–758.
4. Margolis RH, Saly GL: Distribution of hearing loss characteristics in a clinical population. Ear Hear 2008a;29:524–532.
5. Margolis RH, Saly GL: Asymmetrical hearing loss: Definition, validation, prevalence. Otol Neurotol 2008b;29:422–431.
6. Messick DM, Sentis KP: Fairness, preference, and fairness biases. In Messick DM, Cook KS, eds., Equity Theory: Psychological and Sociological Perspectives. New York: Praeger, 1983.
7. Dirks DD, Lybarger SF, Olsen WO, Billings BL: Bone conduction calibration: Current status. J Sp Hear Dis 1979;44:143–155.
8. Frank T, Holmes A: Acoustic radiation from bone vibrators. Ear Hear 1981;2:59–63.
9. Margolis RH, Glasberg BR, Creeke S, Moore BCJ: AMTAS®-Automated Method for Testing Auditory Sensitivity: Validation studies. IJA 2009, in press.
10. Songer JE, Rosowski JJ: A mechano-acoustic model of the effect of superior canal dehiscence on hearing in chinchilla. J Acoust Soc Am 2007;122:943–951.
11. Merchant SN, Rosowski JJ: Conductive hearing loss caused by third-window lesions of the inner ear. Otol Neurotol 2008;29:282–289.
Copyright © 2010 Wolters Kluwer Health, Inc. All rights reserved.