Originally Posted by CharlesJ
Well, don't have to since Villchur's paper referenced the Florentine paper and Viemeister, N.E. "Audio Intensity Discrimination at High Frequency in the Presence of Noise" Science, vol 221, pp12-6-1208(1983).
I was also asking if you had the Florentine paper. If you read that, you see that is hard to draw strict conclusions from it. This is an example:"Not only do our listeners differ in overall sensitivity, but the effect of level and the interaction of level and frequency differ across listeners. However, the main effect of frequency is generally the same for all listeners as shown by the nonsignificant interaction of frequency and listener. "
As you see, it clearly states that individuals vary with respect to their level sensitivity. What they say is similar is what happens in a relative manner from one frequency to the other. And even that has some non-monotonicity that they can't explain.
And, it seems that I made a small mistake where that 3 dB happens. Villchure explains the Florentine paper and states that at 10 kHz the average was 2.12 dB, minimum was 0.47 dB; 16 kHz the average JND was 3.05 dB and the smallest was 1.31 dB.
So, that 3 dB is at 16 kHz not at 20 kHz.
I am glad we agree it didn't include 20 Khz.
On the other numbers, again, Villchure has no data of his own. It all comes from Florentine paper. And there, there is a breakdown of everyone's result. We see that the listener "RP" achieved a level difference of 1.03 db and that was the mean for all the trials. So some were even smaller than 1 db. Once one averages all the listeners you get the numbers you list. We can't bank on that being everyone's threshold.
Since the JND was increasing upwards from the most sensitive region between 1 kHz and 4 kHz one can clearly see that at 20 kHz it will be a lot more than 3 dB.
I don't know. Many people can't hear 20 Khz. Once we find people who readily hear it, they may have have sensitivities that are lower. After all, only 6 listeners were used in Florentine's study. That is too small of a sample size to determine the floor at 16 Khz let alone interpolate to 20 Khz.