About Localization

Our localization mechanism (including the ears, auditory nerves and auditory cortex) is multi-faceted and complex. It excels at identifying and localizing sounds sources in chaotic reverberant environments. It also does quite well at perceiving the environment as well.

The simple models we use for localization – time and amplitude differences between the two ears – don’t really describe how we hear in reverberant spaces. We have an extremely highly-evolved echo-location mechanism that makes use of a variety of sonic behaviors to create a remarkably seamless holistic perception of a sound source, even in very reverberant (i.e. chaotic) spaces. Not only that, but we actually prefer playing and listening to music in reverberant spaces, a preference strong enough that musicians will often refuse to play in dry or non-reverberant spaces. Further, we all unequivocally regard such spaces as “bad sounding” when listening to music.

Because of this, I believe we need to redefine the event we call “a sound.” In any normal reverberant space, we perceive the direct sound energy from the source as well as a volley of early reflections also from the sound source. We identify this group of sound artifacts as a “single sound” by their unique phase-locked relationship. We identify the timbre and location of the source via an integration of the spectra, times and angles of arrival of all such artifacts that arrive within 50 milliseconds of the direct sound.

This is a remarkable feat! Instead of being swamped and confused by the multipath array of 6-30 nearly identical sounds arriving from 6-30 different directions, we instead use this accumulation of sound artifacts to develop an extremely rich perception of the timbre of the sound (we have 6-30 versions of it, remember), its position in the room, our position in the room, and the physical nature of the room (including its dimensions, furnishings and surface materials).

Further, we use different parts of the audio spectrum for different parts of this task. We use high frequencies/short wavelengths to localize (at each ear) the angles of arrival of the various artifacts, and to specifically identify the distance from us in space of the direct sound. Meanwhile, we use low frequencies/long wavelengths (using both ears in combination) to learn about the room boundaries.

We do all this identification and learning preconsciously and through neurological feedback, iteration and cross-correlation. What we actually end up consciously perceiving is a fully integrated, highly processed, edited and developed construct that includes remarkably clear and unambiguous sensations of timbre, position and surroundings of a sound source.

In a general way, then, our perceptions of individual sounds are based on an accumulation of information that piles up over 50 milliseconds. We learn what the sound is, where it is in the room, and what the room is made of.