Frank Musiek, Ph.D., University of Arizona
Many years ago, in one of my first job interviews, I was asked what the difference and related implications was between electronic calibration and acoustic calibration. That question is one I never forgot because at the time, I was a little surprised. But, after thinking about it, I realized its importance. Lately, I have also been a little surprised by how many audiologists depend on “online” measurements without measuring the acoustic signal emitted by the earphone or ear insert. This, in turn, triggered my thinking about calibrations for central auditory test materials and related issues. So, I thought I would comment on some of the more critical calibrations that might be overlooked for some fairly complex stimuli — like those utilized in assessing the central auditory nervous system.
Though others could have as easily been selected and are equally worth discussing, I have selected dichotic listening and filtered speech tests to review, as they are commonly used in a central auditory test battery and are both susceptible to errors in acoustic representation.
One of the first necessary calibrations is for dichotic speech stimuli. The calibration signal for most speech stimuli is a 1000 Hz pure tone at the beginning of the test. This is true not only for central test materials, but in general for many speech materials. A 1000 Hz tone is used for calibration as most speech signals are highly variable and difficult to measure in sound pressure level (SPL). Therefore, “the SPL of a speech signal at the earphone is defined as the rms SPL of a 1000 Hz signal adjusted so that the VU meter deflection produced by a 1000 Hz signal is equal to the average peak VU deflection produced by the speech signal” (ANSI, 1969). It should be noted, however, that there are some speech materials that use the voice fundamental of the speaker as the calibration signal rather than the 1000 Hz tone. However, this method, at least to my knowledge, does not have standard approval. In dichotic stimuli, it is important that both channels be calibrated to be equal in intensity. These calibration signals should be checked at the VU meter as well as at the earphone via an artificial ear and sound level meter. In addition, the onset and offset of dichotic words and or numbers should be monitored (the closer the temporal alignment of words/digits, the more truly “dichotic” the test is and the more difficult the task). Onsets and offsets of speech stimuli are usually determined as a deflection from baseline that is a percentage of maximum amplitude of the signal.
There are many “online” programs that allow one to check temporal alignment, or one can use 2-channel oscilloscopes. These intensity and temporal measurements are key to dichotic test accuracy. That being said, there are additional factors in dichotic listening that need to be considered for accuracy to be sure each ear is receiving highly similar stimuli. Of importance are phonetic considerations; for instance, voicing of phonemes is a factor because unvoiced phonemes generally will yield better performance than voiced. The answer to this (and many other problems) is to counter balance test items. Some available dichotic tests do this, but some may not. By not attending to the aforementioned acoustic characteristics of dichotic stimuli, biases can be introduced that can lead to misinformation and then misinterpretation.
A recent experience with low pass filtered speech (LPFS) tests is my motivation to discuss this test in the calibration context. An audiologist called me complaining that too many patients were “failing” their relatively newly acquired LPFS test. The key to checking your test is to first test yourself. In this case, they did, but the test seemed okay. The test this audiologist had was supposed to have a band pass cut off at 500 Hz and then “roll-off” at a rate of 18 dB per octave (Figure 1). I asked the audiologist to send their LPFS CD recording to our lab. When it was checked, we found that the roll-off was not 18 dB per octave. Rather, it was much sharper, resulting in the test being much more difficult. More patients were failing the test artificially because the norms were based on a test that had an 18 dB per octave roll-off. Checking the accuracy of filters is a more involved procedure than those earlier mentioned for dichotic, as use of a spectrum analyzer is usually required to check on filtering accuracy, which is equipment beyond what most audiologists would have. Therefore, if there is suspicion that LPFS test accuracy is of concern, it may be reasonable to contact the company from which the test was obtained or a university that has appropriate equipment.
As a result of filtering of a speech signal, the total energy of the signal is reduced as well as its natural recognition. It is prudent to administer LPFS test at sufficient sensation levels to be sure that maximum intelligibility is achieved.
The experience with the LPFS teaches another important lesson: do not assume that commercial test materials and equipment are always accurate in what they portray. Industry does the best they can, but mistakes are made. It is important that inaccurate test materials are corrected to avoid misdiagnoses. If there is a suspicion that test materials and/or equipment are not doing what they is supposed to do, then it is the responsibility of the audiologist to troubleshoot, and if possible, to define the problem and inform the commercial entity about it. Otherwise, the problem with the test or equipment promotes other problems that lead to failures serving our patients, and our profession suffers.
- American National Standards Institute (1969) American National Standards specification for audiometers, New York.