One of the advantages of dB nHL is that it takes into account the behavioral sensitivity (averaged ABR thresholds for a particular frequency should be designated as 0dbnHL). Strictly speaking the threshold/dBnHL should be calculated at the rate which would be used in the clinics (due to temporal integration). This paper might be helpful https://www.researchgate.net/publication/247830562_Behavioural_and_Electroacoustic_Calibration_of_Air-conducted_Click_and_Toneburst_Auditory_Brainstem_Response_Stimuli

Article Behavioural and Electroacoustic Calibration of Air-conducted...

The international standard ISO 389:6 does provide some important information on this issue.

According my understanding, the sole purpose of bringing dBnHL is to reduce the difference between behavioral threshold and physiological threshold estimated using ABR. By which I mean a normal hearing subject (BTH = 5dB) should get ABR threshold at or less than 10dBnHL. However, in regular clinics there is always correction factor of 15 dB above the dBnHL which now been termed as eHL (i.e ABR threshold -Behavioral Threshold).

From the above I understand that sole purpose of bringing of dBnHL is not been completely resolved.    

This is an interesting question. I have been "involved" with this topic since the 1908s, thus my answer is somewhat long, and certainly reflects my personal opinion.  However, any answer to this question is not without controversy.

As noted above, the primary reason for using "dBnHL" is to account for the differences between behavioural thresholds at each frequency, so that "0 dBnHL" represents the median or mean threshold for normal adults. This is the same concept as "dB HL", which uses (nearly) continuous tones. As also noted above, however, in contrast with behavioural HL, even with the "nHL" correction, one still has to apply "eHL" correction factors reflecting  differences between AEP (usually ABR) thresholds and behavioural thresholds; correction factors which are different for infants compared to adults.

Unfortunately, it is not so simple as dBHL, because "dBnHL" has had to be calculated for so many different stimuli (e.g., clicks for ABR, 5-cycle brief tones for ABR, chirp stimuli for ABR or ASSR, 40-60-ms tones for corticals). Complicating this issue further is that different presentation rates are used -- behavioural thresholds change with rate (even though AEP, especially ABR, thresholds do not show such changes with rate). Finally, and very importantly, the physical calibration of non-continuous tones is a little more complicated; most clinicians and many researchers, and, indeed, many audiometric calibration technicians, do not know how to calibrate these stimuli.

My colleagues and I have been pretty adamant that individual clinics should not typically determine their own "nHL" values, as this requires large numbers of normal subjects, careful threshold procedures, and quiet test rooms. Rather, assuming they are using stimuli essentially identical to others' studies which investigated nHLs (or a "standard" such as the ISO standard), they should have their stimuli carefully calibrated acoustically to the previously published nHL calibration standard.

In our previous studies, we determined nHL calibrations for all stimuli using a 10/s stimulus rate, regardless of the actual rate used to record the AEP in the study (we did this because the ABR and MLR do not show better thresholds with increasing rates, whereas behavioural thresholds do). The ISO standard mentioned above by Stig Arlinger states that a 20/s rate should be used when determining nHL calibrations -- I am not sure why they selected 20/s, but the difference in threshold would be very small (only 1-2 dB) (Stapells et al., JASA, 1982).

The above confusing situation is made worse by the fact that there is not yet agreement as to formal RETSPLs/RETFLs (i.e., "standards")  even for the most commonly used stimuli for ABR testing. As Stig Arlinger has noted above, there is an ISO standard for brief stimuli, specifically clicks and brief-tone stimuli. However, these "standards" are not fully accepted, especially for tonal stimuli (e.g., they are not commonly used in North America). They tended to ignore or discredit substantial preceding research into nHL thresholds and, in the case of tonal nHLs, they have some significant differences. (Oddly, the reference force levels for bone conduction in these "standards" were estimated from pure-tone standards, rather than directly determined.) 

The move towards  "standard" reference thresholds is important, but it must also take into account previous research and consider differences. My greatest concern is that the large majority of research into adult and, importantly,  infant tone-ABR thresholds has been carried-out using nHL values that are different from the ISO standard -- any possible move to a different standard must consider how this relates to the results (and interpretation) of the past and future studies.

Problems will still remain even if a standard becomes widely accepted worldwide. Every time a new or different stimulus comes out (e.g., chirp stimuli), no standard will exist, and some sort of nHL needed.

Lately, I have wondered if we should never have moved to developing "nHL". Rather, perhaps we should have just used well-established pure-tone "HL" calibrations, and then determined what are the normal ABR (or MLR or ASSR) offsets/corrections. (This would eliminate problems with calibration, but would still entail questions with determining the offsets/corrections.) This is essentially what most researchers carrying out threshold assessments using the ASSR or CAEP (cortical auditory evoked potential) have resorted to.

Given the current situation, although not all would agree with me, I recommend one use the nHL calibrations from the study(ies) one is trying to emulate. For example, if you are testing infant ABR thresholds using the stimuli and parameters I have recommended, then you would use the nHL calibrations I have published. On the other had, if you are using Michael Gorga's stimuli and parameters, then use his published calibrations. I expect, with time, there will be more movement towards use of standard reference thresholds as we understand/explain current differences.

Everyone performing AEP testing, particularly ABR for the purpose of estimating behavioral thresholds in infants and young children, should read Dr. Stapell's response.  As an audiologist working in the audiometric equipment and calibration business for almost 30 years, I have struggled with this issue most of that time. Different manufacturers use different calibration offsets which complicates the matter even further.  Using Dr. Stapell's procedures he mentions, we have developed a comprehensive calibration process for all of the AEP units we sell.  Unfortunately, the users are not very well informed on this topic and when we tell them what the cost is to calibrate their AEP device, most reject the need to do so as they either don't understand the value or simply don't want to spend the money necessary to ensure accurate calibration of their equipment.  Of course, having a published ANSI standard would help.

Dr. Stapell's point about using HL calibration values is quite interesting.  It probably would have made this issue less complicated in the long run.

Thank you very much for very good description of nHL. I definitely agree with on the statement that not every one clearly follows the standard correction factors.  Along with the dBnHL is there need for controlling other parameters (EEG noise etc) so that difference between dBHL and dBnHL would reduced.     

Thank you David! I will certainly be using this material too in my class as I do get asked the same and students often feel (mistakenly) that getting 10 "normal" subjects is so much easier than the acoustic calibration.. 

thank you..great question Vijay!

Thanks Mridula.

Of course, one of the preceding nHL calibration studies that seemed to be missed by the ISO "standard" was your 2003 study (noted below).

Article Behavioural and Electroacoustic Calibration of Air-conducted...

As a small correction,  I have been involved with this topic since the 1980s, not the "1908s" (I'm old,  but not that old!).

I believe calibration issues can explain much....

regarding the stimulation rate around 20Hz is great for a good wave form quality. Higher values save time but often apart from wave V other wave forms get lost.

Daniel: ~20/s rate is indeed a good rate for click-ABR for neurological/retrocochlear testing. Not at all the best rate for click or (more-relevant) tonal threshold testing -- much faster is better.

Nevertheless, "waveform quality" is not really relevant when determining behavioural thresholds for ABR stimuli (i.e., "nHL").

David: Correct, for threshold a faster rate is ok. But sometimes if you do not get any response, lower rate is needed to get at least a response at maximum stimulation. And when expecting a threshold almost a the maximum stimulation level quality of waveforms is very important.

So use fast rate as David wrote in general, but switch to low rate if you expect a very high to not measurable threshold (e.g. cochlear implant candidate). Here a retrocochlear setting will help to get responses near the EP system's limit.

How fast would you recommend to stimulate? And TB, pip? Or chirp (which one)? Which EP system do you use?

We use 2-1-2 (cycles) brief-tone stimuli (aka tonepips). We do not use chirps as they are less frequency specific and importantly, have few published data for infants with hearing loss.

The maximum rate for threshold ABR testing is determined by the latency of the response (which, in vast majority of testing, is the negative-positive-negative wave V-V' complex). Because wave V latency increases with decreasing stimulus frequency, longer analysis windows are required for lower frequencies, thus longer ISIs required. (Remember too that responses to near-threshold stimuli as well infants' responses are later in latency, so this needs to be accounted for.)

Thus, in practice, for infant threshold ABR testing, recording windows of 20-25 ms are required for 500 Hz, 20 ms for 1000 Hz, and 15 ms for 2000-8000 Hz. Using typical/normal averaging techniques, these recording windows determine the maximum stimulus rate that may be used:

500 Hz: 40-50/s

1000 Hz: 50/s

2000-8000 Hz: ~61-65/s

We do NOT find that these "faster" rates are problematic for ABR threshold testing in infants. It is a common "ABR myth" that "fast" rates degrade infant ABRs. (Longer latencies yes, but not much affect on wave V amplitude.) Faster rates result in faster testing.

The British Columbia Early Hearing Program currently uses ~39.1/s rate for all frequencies. Faster rates for mid/high frequencies are being considered, but it takes time to implement change in larger/estblished programs.

[NOTE: Not all manufacturers' ABR systems are equal. Different ABR systems add different processing times which slows the fastest possible rate. Usually, either a slightly slower rate (1-3/s less) or slightly shorter analysis time (1-3 ms shorter), or combination of both, is/are required so as to not "skip" stimuli. When setting up protocols, each system MUST be checked to ensure it is not skipping stimuli with the chosen rate/analysis time combination.]

Thank you for these details. Is this specific setting tested for several EP systems in Vancouver? I got interacoustics eclipse and Nihon Kohden Neuropack S1 here in Munich. Tested several other EP system the last years, too. But these two are our standard devices.

PS: Especially 2k to 8k Hz --> 61-65/s sounds really great, will try.