Do You Double Check Your Research Data? More on RIC
In a previous blog, I reported on some real-ear data obtained when measuring various insertion depths of a RIC closed system into the ear canal (Figure 1).
Figure 1. RIC (receiver-in-the-canal) REAR results for various insertion depths of the speaker into the ear canal. Results showed that gain increased in the high frequencies with deeper insertion.
In response, I received a nice note from a friend and colleague, Dr. Mead Killion, who mentioned that while he believes the results of the measurement, that it does not tell the entire story – my comments did not go far enough, and he proceeded to explain, as captured in the following paragraphs.
More Gain = More Acoustic Feedback?
He mentioned that based on work he and Don Wilson had conducted 27 years ago (1985) {{1}}[[1]] Killion, M. and Wilson, D. Response-modifying earhooks for special fitting problems, Audecibel, Fall, 1985 [[1]]when measuring tubing insertion depth with open fittings (tubing), that as the insertion depth increased, acoustic feedback increased rapidly as well. In other words, the maximum gain before feedback decreased with deeper insertion, and that a depth close to 8 mm produced the best results once the “horn action” back-out from the eardrum pressure increased (Figure 2).
Figure 2. Vent response ( ) and feedback SPL at hearing aid microphone (—) for three tube insertion depths (Killion & Wilson, 1985). This shows that the more deeply the tubing is inserted for these open mold fittings, the amount of gain before feedback is reduced (difference between the dotted line and the solid line).
Note the “suck out” at the half-wave resonance of the ear with 20 mm insertion depth (upper curves). This is caused because at a half wavelength the tube disappears (more or less, figuratively speaking) and the “outside air shorts out the signal.” This explains both the high transmission back-out and the drop in pressure at 2 X 2.8 (5.6) kHz.
I asked if this work could be republished on this blog site, especially since their original article had limited distribution at the time of publication. The issue of a potential reduction in the usable gain available resulting from coupling devices that inserted more deeply into the ear canal would be of considerable interest currently, especially since so many of the devices now fit more deeply into the ear canal. And, even though the hearing instruments and couplings were different (RIC in Figure 1 vs. OTE (over-the-ear) with tubing in Figure 2), the potential of acoustic feedback interaction with increased insertion depth of the coupling device is a significant issue.
An Interesting Development
It was at this point that an interesting, and important turn in the discussion occurred. While Dr. Killion was fine with the reprinting, he cautioned: “But, in light of the data that are shown in your figure, we should double check our results.” He wanted to make certain that what would be published was accurate. He stated that in 1985, they used an ER-2 earphone (flat-eardrum-pressure frequency response) and an ER-7 probe microphone at the eardrum-position and OTE mic position on KEMAR.
He suggested that as a double check, they could compare the SPL at the ER-11 Zwislocki-coupler microphone to the SPL at the ER-7C probe at the eardrum position. That would be one double check. Also, for interest, they could keep track of phase with the ER-7 at both positions, to see where feedback might occur with either receiver-wiring polarity. Simply repeating the experiment would be the primary double check, of course. Dr. Killion did not think they could have screwed up such straightforward measurements, but…. he felt also that the new data would add a nice P.S. to the original paper, and besides, he and his group should know these things for their own development work. But, why the double check on their previous measurements?
Even Brilliant People Can Make Serious Errors
The story that follows points out what may be serious, but often overlooked, concerns relative to published data.
Dr. Killion’s Story: When I was working on my Ph.D. at Northwestern University (NU), Dr. Raymond Carhart, my advisor, died. Another faculty member took over his NIH grant to study “Perceptual Masking” in children. To explain: In adults, several NU experiments had shown that actual speech as a masker was 8 dB more effective than speech spectrum noise modulated by the envelope of the actual speech, even though the spectrum shapes were carefully matched. Further experiments reported that speech played backwards, or non-English speech, was just as effective – a surprising result for “perceptual” masking.
Dr. Carhart’s replacement on the faculty asked me to check the experimental recordings, which had been used for years in data collection and subsequent article publications. I listened and immediately said, “The speech is louder!” I sealed the 1″ microphone of a Sound Level Meter up against the TDH-39 cushion and found the average SPL of the speech read 8 dB higher than the SPL of the noise!
It turned out that the calibration of the speech and modulated noise had not been left to chance. It had been done with extreme care, using a wave analyzer with 50 Hz bandwidth. At each center frequency, the analysis of the token was plotted out on a long strip of paper. The graph showed big jumps when a vowel formant happened to fall in the band being analyzed, and was fairly low otherwise. If they had converted dB to amplitude and used rms averaging, everything would have been OK, but they “averaged dB.” By repeating their “averaging of the dB”, I was able to reconstruct the 8 dB difference. It also turned out that the calibration had been assigned to the lab manager, who had re-assigned it to a Master’s student.
Moral: Even brilliant people can make serious errors.
Story Post Script: At the meeting when I reported my findings, the project faculty member said in a dramatic voice, “We must be very careful how we treat this information!” Dr. Tom Tillman, my own advisor, said calmly, “We should double check these results and then publish them so no one else repeats the same mistake.” Both of these faculty members are now deceased, but it is easy to guess which one I admired and which one I held in low regard. It is possible that there may be a few people teaching psychoacoustics that still think perceptual masking is real in the sense of those Northwestern University papers based on incorrect calibration data.
So, to publish in accordance with the direction of Dr. Tillman, “We should double check this result, and then publish it so no one else makes the same mistake.”
Irony
Dr. Killion suggested that he recheck his data, but did not suggest that I repeat mine, but I definitely now will. And, what about the results for both our re-checks? Stay tuned.
