1 You mention “softness imperception” in your title. I'm not sure what that means.
Softness imperception—SI for short—is an inability to hear some low loudnesses that are audible to normal listeners.1 When a sound is at threshold, it will seem louder to a person with SI than to a person with normal hearing. Our new findings indicate that this phenomenon is common in hearing losses of primarily cochlear origin.
2 Can you give me an everyday example of SI from clinical practice?
Yes. The simple task of raising an index finger when a sound is heard provides a quick indicator. Have you ever noticed that when patients hear a loud sound, their index fingers pop up quickly, and when they hear a very soft sound their index fingers rise slowly or may not extend all the way up? When measuring absolute threshold of patients with noise-induced hearing loss, you might have noticed a quick and full-finger extension at 4000 Hz where threshold is elevated and a slow finger response at 1000 Hz where threshold is within normal limits.
The rate at which the index finger extends is likely to indicate the confidence of the response. So, the louder a sound is perceived, the quicker the response. The patient's quick response to the tone at the elevated threshold at 4000 Hz is likely to be an indicator of SI.
3 Why do you use the term “softness imperception”? Isn't what you are describing simply what most of us call “recruitment”?
The answer is both yes and no. You probably learned that most patients with cochlear hearing losses have recruitment. However, loudness does not grow abnormally fast near threshold in these patients, which is what most of us think of when we hear the term “recruitment.” Therefore, “recruitment” does not really describe what is happening. Recruitment never was a good term, and it hasn't become better with age!
“Softness imperception” describes the perception of most people with cochlear hearing losses. That is, sounds that are just audible may not be all that soft, because hearing loss causes an inability to perceive softness. Once hearing healthcare professionals get used to this new term, I believe they will actually prefer it to recruitment because it is easier to explain to patients with hearing losses.
4 I remember learning about the SISI (Short Increment Sensitivity Index) test way back in graduate school. It's been around since the 1960s. Isn't that an example of SI?
It is not clear exactly how the SISI relates to SI for at least two reasons: First, the SISI measures discrimination ability at levels somewhat above threshold—typically 20 dB SL. SI primarily describes perception at threshold. Second, the relationship between the growth of loudness and the SISI—which measures the ability to hear a short increment in the level of a tone—remains unclear. (For a review of the SISI, see Buus et al.2,3)
I tend to agree with James Jerger, who wrote, “SISI is not an indirect test for loudness recruitment; it is not an indirect test for anything. It is nothing more than a way of telling whether the patient can hear very small changes in sound intensity. There is only one reason for wanting to know this. Evidence exists that the ability to hear these very small changes is unique to disorders in the cochlea.”4
5 Okay, I think I understand. So, does the presence of SI mean that there is no point at which loudness grows abnormally rapidly in people with cochlear hearing losses?
In most people with cochlear hearing losses, there is a range of levels around 20 dB above threshold in which loudness grows more rapidly than in normal listeners. This rapid loudness growth is probably due to a loss of compression, which is thought to result from a reduction in outer hair cell function.5 Such growth is unlikely to be caused by an abnormal increase in the number of afferent fibers that respond to the sound, which is what was originally implied by the term “recruitment.”
Whatever its origin, this small range of rapid loudness growth is really a minor point in relation to SI. For example, the elevated loudness at threshold for a person with a 65-dB hearing loss may account for a factor close to 16 (on average) of the “catching up” of loudness in terms of sones (an additive scale of loudness), whereas the increased slope around 20 dB SL would contribute only another factor of 2.5. (This is because the normal loudness function is so compressive at moderate levels that the normal loudness changes by only a relatively small factor over the range of levels where compression is strong.) Therefore, loss of compression is not the major cause of loudness catching up and it has practically no effect on loudness growth near threshold. However, note that loss of compression may contribute significantly to other perceptual problems that result from hearing loss.6
6 I might need more explanation on that later. But first, are you telling me that we shouldn't be using the term “recruitment”?
That's right, because it leads to misunderstanding of what is occurring. The major cause of the alteration in loudness in people with cochlear hearing losses is softness imperception. The secondary effect is due to compression loss. The classic view of recruitment is incorrect. Continuing to use the term “recruitment” while trying to think in a new way will cause even greater confusion because people will not know how “recruitment” is being used (i.e., the old or the new definition).
7 How should I explain SI to my patients?
It is difficult to tell people that they have lost some of their hearing. As you know, it is best to describe what they have lost with empathy and as clearly as possible. At the same time, it is important to give them positive and realistic expectations about rehabilitation.
For example, you could say to a potential hearing aid user something like, “I am sorry that you have lost your ability to hear soft sounds. Although you have lost the perception of soft sounds, you do have a range of hearing in which you can still hear louder sounds. Hearing aids will make many low-level (that is, low-physical-intensity) sounds audible to you again. The problem is that your hearing loss comes with softness imperception. This means that if we amplify low-level sounds enough for you to hear them, they will probably sound louder than they did before you had a hearing loss. We will need to work together to find the right compromise between making the sounds you want to hear audible and making the sounds you don't want to hear (like fan noise) too loud.”
8 If amplified low-level sounds are not heard as soft, how does a hearing aid user experience them?
Hearing aid users will experience them differently, but there are many things working in the patient's favor. Most people can easily identify the natural loudness of a sound, even when it is amplified. For example, whispered speech can be distinguished from normal speech, which can be distinguished from shouted speech. Even if the loudness is the same, speech can be identified as soft from its frequency content/spectrum.
9 How does SI impact on my patient and family counseling?
Understanding SI can have an enormous, positive impact. Once patients and their families understand the kinds of problems they will experience, a lot of misunderstandings can be avoided.
I imagine that something like the following scenario occurs over and over in families around the world: Grandfather, who has a cochlear hearing loss, is reading in the living room. His grandson is asked to tell him that dinner is ready. The boy comes into the room and says in a normal voice, “Grandfather, dinner is ready.” The grandfather doesn't hear the boy. The boy raises his voice somewhat, “Dinner is ready.” Again, no response. Then the boy raises his voice even more: “TIME FOR DINNER!” Grandfather, now annoyed at being startled, turns to the boy and reprimands him, “Don't shout! I can hear you.” The boy's feelings are hurt and he doesn't understand why his grandfather can't hear sounds at a normal level, but complains when sounds are loud enough for him to hear.
An understanding of SI could have prevented this situation from occurring. If you had discussed SI with the grandfather and his family, the grandfather would know that there is something about his hearing that makes him easily startled as sounds suddenly become audible and the family would realize that when he finally does hear sounds they may not be all that soft.
10 How was SI discovered?
It was discovered in the same way as most discoveries. New data called into question prevailing assumptions. In the case of SI, it took several steps of logical inquiry and questioning of an old assumption.
It all began when new measurements in our laboratory showed that rates of loudness growth near the thresholds of listeners with hearing losses of primarily cochlear origin were within the range of those obtained in normal listeners.7,8 This discovery raised an obvious question: If loudness in listeners with cochlear hearing losses grows at a normal rate, then how does loudness “catch up” to become near normal at high levels?
11 Interesting question. How does loudness catch up?
It really doesn't catch up. It gets a head start! Let me explain.
Søren Buus and I started with two facts. First, loudness becomes near normal once a sound is well above an elevated threshold of a person with a cochlear hearing loss. Second, the data were unequi-vocal about the rate of loudness growth being normal in people with cochlear hearing losses. So, something had to give.
After some thought, we realized that the classic notion of recruitment as an abnormally rapid growth of loudness was derived from an assumption—usually unstated and never tested—that loudness at threshold is the same for all listeners and all frequencies.9 In other words, loudness at threshold is the same regardless of hearing status and test frequency. However, there was no evidence that this assumption was true.
Once we realized that we did not have to abide by this assumption, the solution to our problem was simple. If we abandoned the untested assumption and assumed instead that loudness at threshold was greater than normal when threshold was elevated by cochlear hearing loss (i.e., loudness gets a head start), then the loudness function for listeners with cochlear losses would approach that for normal listeners at high levels.
To check our logic, we modified our model of loudness10 to assume that loudness at threshold would increase as hearing loss increased.8 This modified model allowed us to fit the data from listeners with a variety of audiometric configurations. Indeed, the model showed that, in general, loudness at threshold should increase with the amount of hearing loss to achieve a good fit to the data.
12 So the modeling worked. But how did you test your new assumption? Assump-tions need to be tested, right?
We certainly think so. Incorrect assumptions can quickly lead one astray. However, measuring loudness at threshold is not easy. We had to find a way of assessing the loudness of a sound that is heard only half the time. To do this, we decided to use a reaction-time paradigm. We felt justified in doing this because reaction times are correlated with loudness: the louder the sound, the faster the reaction time.11
13 I'll agree with that. So, how did you do the testing?
We measured simple reaction times to sounds at various levels, starting at the carefully measured threshold. You can measure reaction times quite simply by asking listeners to press a key as soon as they hear a sound.
For example, we have measured reaction times to tones presented at and above threshold at two frequencies in listeners with sloping high-frequency hearing losses. One frequency had normal or near-normal threshold, the other had elevated threshold. The results show that reaction times are faster at and near the elevated thresholds than at and near the normal thresholds in most listeners with cochlear hearing losses. These findings support the presence of SI in cochlear hearing losses.12
14 Why don't you just conduct an ABLB (alternate binaural loudness balance) test in a listener with a unilateral cochlear hearing loss? Wouldn't that be an easier way to measure SI?
Actually, it would be more difficult. The reason that no one has tested loudness growth at threshold is that the psychoacoustic tools used to measure loudness (i.e., alternate binaural loudness balances) don't work at test levels very close to threshold.13 However, when the slope of loudness-growth functions of people with cochlear hearing losses are extrapolated toward their thresholds, the slopes do not differ from normal when compared at equal SLs.14
15 Does this mean that what I've read about loudness growth in the literature is incorrect?
Yes and no. The data in the literature are accurate, but our interpretation of them has been wrong because we were working with the wrong theoretical framework.
16 I typically fit wide dynamic range compression (WDRC) instruments, and my patients often complain that low-level sounds are too loud. Is this related to SI?
I think so. Once low-level sounds are amplified so patients can hear them, they will be louder than the unamplified sounds are for a normal listener, because loudness at threshold is greater than normal in people with cochlear hearing losses. Therefore, the subjective complaint that “refrigerator noise is too loud” just might be validated in the light of our new theoretical framework of softness imperception.
17 I don't usually use large compression ratios when I fit WDRC hearing aids. Does SI indicate that I'm doing the right thing?
Yes. Softness imperception teaches us that it is wrong to try and squeeze the normal range of audible SPLs into the much narrower range of SPLs that are audible to someone with a hearing loss. We have to accept that there is a range of low loudnesses that cannot be restored to a person with cochlear hearing loss. This means that the compression ratio needed to restore loudness perception to as normal as possible is lower than that indicated by the reduction in the dynamic range expressed in terms of dB SPL.
18 This seems to contradict the IHAFF/VIOLA method. How do you explain this?
Perhaps there is some opposition, but it may be more in terms of underlying assumptions than actual outcomes. These fitting procedures are based on the idea that hearing aids should restore loudness to be more or less normal, while avoiding uncomfortable loudness. If the loudness-restoration assumption is combined with the classic view of recruitment, the prescription will attempt to fit all sounds that are audible to a normal listener into the reduced range of SPLs that can be heard by the patient.
For low input levels, this requires a gain equal to the amount of hearing loss, which I think most clinicians would agree would result in overamplification. (Never mind that it is often unachievable, and note that VIOLA moderates the gain for soft inputs by setting it equal to the average across three loudness categories.)
Our findings about SI indicate that amplifying sounds at levels near the normal threshold by the gain required to make them audible to the patient would make them too loud. In other words, SI simply confirms the usual practice of always using a gain that is less than the hearing loss. Given that low-level sounds should be amplified less than the hearing loss, it also follows from SI that the ideal compression ratio must be less than that required to fit all of the normal dynamic range into the patient's reduced dynamic range. That said, almost all hearing aid fitting procedures provide target gains only for input sound levels between 40 or 50 dB SPL and about 90 dB SPL. Thus, they often avoid going to the extremes that might be indicated by their underlying assumptions and theoretical underpinnings.
We think that both research and common practice support these implications of SI. For example, few WDRC aids allow compression ratios greater than about 3 and few clinicians actually use compression ratios greater than about 2.5. In addition, Keidser et al. reported that 24 patients with a variety of audiometric configurations tended to prefer hearing aids fitted according to NAL-NL1, which prescribed average compression rates of 1.3 in the low band and 2.2 in the high band, over hearing aids fitted according to a procedure that prescribed average compression rates of 2.4 and 3.2.15 Likewise, Neuman et al. had 20 hearing aid users rate speech in various types of background noise for clarity, pleasantness, background noise, loudness, and overall impression. They found that all the average ratings became poorer when the compression ratio was increased from 1.5 to 3.16
19 What are the implications of SI for hearing aid design?
Two implications immediately come to mind. First, current technology cannot completely restore normal loudness to people with cochlear hearing losses. If low-level sounds are made audible, they are likely to sound louder than the unamplified sounds do for people with normal hearing. Therefore, it may be unwise to amplify very low-level sounds to audible levels. (Of course, feedback is another reason not to do this!) In fact, many compression hearing aids already use relatively high compression thresholds and expansion of low-level sounds to reduce the loudness of low-level sounds and to avoid feedback problems.
Second, listeners with cochlear hearing losses have reduced dynamic ranges in terms of both dB SPL and loudness. Therefore, low to moderate compression ratios (less than about 2.5) would be indicated. In fact, such compression ratios are generally found to be appropriate when fitting hearing aids with WDRC.17 Even hearing aids for severe and profound losses often limit the compression ratio to no more than 3.0, even when the usable dynamic ranges of the listeners are between one-fifth and one-tenth the normal dynamic range measured in terms of dB SPL.
In other words, hearing aid manufacturers and clinicians already knew from experience and hearing aid research that it is not appropriate to squeeze the entire dynamic range for normal listeners into the limited range of SPLs available to patients with hearing losses. Our finding of SI is consistent with many aspects of hearing aid design and clinical practice.
20 I'm getting to like the notion of softness imperception. Why has it taken us so long to discover it?
Sometime during our very early education in the study of hearing losses, we memorized the definition of recruitment as an abnormally rapid growth of loudness near an elevated threshold. The notion that loudness grows abnormally fast in listeners with cochlear hearing losses is a “truth” that has pervaded the fields of audiology and psychoacoustics. It is explained as a fundamental effect of cochlear hearing loss in every textbook on hearing loss and is cited in literally hundreds of scientific articles.
In fact, until now all the models of loudness growth in listeners with cochlear hearing losses have assumed that loudness at threshold is the same for normal and impaired listeners. As with most concepts learned at an early age, few of us questioned it. Only when we were faced with unequivocal data showing that rapid growth of loudness does not occur did we feel compelled to go back to investigate the foundation of the old “truth” about recruitment. This re-examination has caused considerable rethinking of basic ideas about auditory perception in people with cochlear hearing losses.
Recruitment: (1) A term that usually is associated with gathering new members or employees, popularized by the military. Known well by Army Captains Carhart and Northern. (2) A word of French origin, meaning “to grow.” In the audiologic literature, this term first appeared in 1937, when Edmund Fowler used “recruitment” to describe the rapid loudness growth that he observed in some of his patients using his ABLB test. Although not calling it recruitment, in another 1937 publication Scott Reger described similar loudness growth findings with the MLB.
At the height of MLB/ABLB popularity (circa 1965), audiologic lunchtime conversations were peppered with the terms “partial recruitment,” “complete recruitment,” and “hyper-recruitment.” If you were cool, you'd even toss in “decruitment.” Identifying people with recruitment seemed important back then.
Over the years we've learned that from a diagnostic standpoint, recruitment isn't really a big deal. It usually means that the patient doesn't need medical or surgical treatment. However, now and then, we still hear lunchtime comments like “My 10 o'clock patient didn't like the output of his hearing aids; he was really ‘recruiting’!”
So, after all these years, we now have a Page Ten author who is telling us that our interpretation of the classic data was off the mark, and what we have been observing really isn't recruitment—it's softness imperception. What?
Mary Florentine, PhD, is a Matthews Distinguished University Professor in the Department of Speech-Language Pathology and Audiology at Northeastern University, in Boston. She has received her university's Excellence in Teaching Award and she directs Northeastern's Institute for Hearing, Speech, and Language. Dr. Florentine is known worldwide for her research and publications related to experimental and theoretical psychological acoustics in normal and impaired hearing. Her areas of expertise range from basic research in hearing to wide dissemination of knowledge to the general public. Even if you've missed her articles in JASA, maybe you've seen her interviews in Time, Self, and Redbook, or listened to her on NPR's All Things Considered.
Speaking of “all things considered,” consider what Mary has to say about this thing she calls “softness imperception.” It's a new way of thinking about loudness growth that can have direct applications on how we select and fit hearing aids, as well as how we counsel patients. Maybe our patients aren't “recruiting” as much as we think!
Page Ten Editor
© 2003 Lippincott Williams & Wilkins, Inc.