Auditory Time Constants: A Paradox?

Abstract

Three types of experiments on temporal discrimination are considered. The first is threshold detection of tone bursts with the duration of the burst as the independent variable. Results can be described by a model incorporating a “leaky integrator” which operates on the intensity of the signal. The appropriate time constant is 200 msec. Next comes the detection of amplitude modulation. The threshold found as a function of modulation frequency (for low modulation frequencies) can again be described by a “leaky integrator” model, this time equipped with a time constant of 20 msec. It is argued why these two models are incompatible: the auditory system appears to be able to detect modulations far easier than the 200 msec time constant model would allow.

More directly aimed at temporal phenomena are experiments on forward masking and the detection of gaps in a continuous signal. Interpretation of the findings is rather difficult because there is a peculiar nonlinearity in the temporal persistence of masking and because the “ringing” of the auditory filters is difficult to separate from the persistence of masking. It is argued why, on the basis of experimental results, one should come to the conclusion that the auditory system is capable of detecting a 2–4 dB decrement of intensity in approx. 10 ms. This is compatible with what we can infer about the ringing of the filters. However, it is incompatible with the 200 msec time constant model of temporal integration: the auditory system is far less capable of detecting increments than decrements. The conclusion is that each type of experiment leads to a model describing only the results of that experiment. The three models described should be regarded as “ad hoc” models because they cannot be united into one model.