A Biased System

Human hearing is not linear – our ear canals and brains have evolved to give greater bias to the frequencies where speech intelligibility occurs. This is represented in the famous Fletcher-Munson equal-loudness curves, and it’s where the concept of A-weighting for measuring noise levels originated.

Our ears are most sensitive to frequencies in the upper midrange – around 3 kHz – and less sensitive at the frequency extremes. Figure 1 shows the threshold of human hearing over frequency. Notice that a 62 Hz tone has to be at least 30 dB SPL just to he perceived. However, if another sound of a similar frequency is also audible at the same time, we may experience the phenomenon known as auditory masking.

This can be illustrated by the experience of talking with a friend on a train station platform when a train speeds by. Because the noise of the train encompasses the same frequencies occupied by speech, suddenly we can no longer clearly hear what our friend is saying, and we must either shout to be heard or wait for the train to pass. The train noise is masking the signal of the speech (Figure 2).

The degree to which the masking effect is experienced is dependent on the individual – some people are still be able to make out what their friend is saying if they only slightly raise their voice, while others need them to shout loudly in order to carry on the conversation.

Masking also occurs in a more subtle way. When two sounds of different frequencies are played at the same time, as long as they are sufficiently far apart in frequency, two separate sounds can be heard. However, if the two sounds are close in frequency, they’re said to occupy the same critical bandwidth, and the louder of the two sounds will render the quieter one inaudible.

For example, if we play a 1 kHz tone so that we can easily hear it, and then add a second tone of 1.1 kHz at a few dB louder, the 1k tone seems to disappear. When we mute the second tone, we confirm that the original tone is still there and was there all along; it was simply masked. If we then re-add the 1.1k tone so the original tone vanishes again, and slowly sweep the 1.1k tone up the frequency spectrum, we will hear the 1k tone gradually “re-appear.”

The further away the second tone gets from the original one, the better we hear them as distinct sounds.