[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Infrapitch

Dear List,

This thread started about Infrasound, but now some interesting issues have been discussed about Infrapitch.

I agree with most of what Richard Warren said. My only trouble is with the sentence:

"Infrapitch and pitch form a seamless continuum of perceptual iterance."

I don't think it is quite that seamless. Personally I try to avoid the term "infrapitch" because it implies that periodicity detection below 10 Hz has something in common with pitch. In my view, it has more the characteristics of memory than of pitch perception.

In our own experiments we could extend the range down to 0.05 Hz (up to a period length of 20 s). Quite a number of participants could successfully detect periodicities of this length. The introspective sensation when doing so is definitively that of "remembering", not of "perception". Also, many other characteristics point to a close relation of this performance to classical short-term memory performance.

It is true that periodicities around 2 Hz (0.5 s) evoke a "perception" that is different from that of non-repeating noise. I believe that it is memory that helps to detect some statistical fluctuations in that noise that otherwise would pass unnoticed. In other words, memory is altering perception.

Now lets go to the extreme: a cycle of 0.1 s (10 Hz) sounds quite like a machine gun or some other fast acting machine. One could think of a kind of "timbre" of that sound. That might lead us to speak of infrapitch. However, this timbre perception breaks down if we go to very small numbers of repetitions. And it is still doable: a well trained participant can detect a single repetition of 0.1 s of white noise (embedded in some more white noise, say: 0.4 s noniterating noise, 0.2 s iterating noise = two times the same 0.1-s segment, 0.4 s noniterating noise, versus 1 s noniterating noise). A small click or something gets audible in the trials with the one-time iteration of 0.1 s of white noise. Now if we imagine this "click" repeated at 10 Hz this gives a kind of timbre percept. Nonetheless the mechanism at the base of this performance seems to be sensory memory, emphasizing statistical fluctuations so that they get audible. In order for this to work, strict periodicity is not required. Jittered repeating frozen noise will do as well.

Best regards,
Christian Kaernbach

Richard M. Warren wrote:

The term infrasound can be misleading;  while the frequency
threshold for hearing sinusoidal tones (at a reasonable amplitude)
is roughly 20 Hz, holistic frequency perception of complex
waveforms (e.g., iterated noise segments) continues for another
five octaves of infrapitch below 20 Hz.  Guttman and Julesz
(1963) found that a percept they called whooshing occurred from
roughly 0.5 to 4 Hz, motorboating from 4 Hz to 20 Hz.  A noisy
pitch is heard from 20 to 100 Hz, and a pure noiseless pitch with
interesting timbres from 100 Hz up to 16 or 20 kHz.  Thus,
infrapitch and pitch form a seamless continuum of perceptual
iterance extending for 15 octaves for stochastic waveforms,
subserved by partially overlapping neural mechanisms of periodicity
detection and place detection.

Studies of infrapitch, since the pioneering study of Guttman and
Julesz, have been conducted by Irwin Pollack, Christian Kaernbach,
and myself.  Infrapitch detection has also been studied in nonhuman

Infrapitch periodicity is quite different from broadband ventilation
noise that is amplitude modulated by a low frequency.  Acoustically,
a Repeated Frozen Noise segment (RFN) has a line spectrum consisting
of all integral multiples of the repetition frequency, each harmonic
having a randomly determined amplitude and phase.  Only harmonics
within the range of roughly 20 Hz to 20 kHz can contribute to
audibility.  However, RFNs having repetition frequencies from 0.5 Hz
to 20 Hz can be heard as patterns based on unresolved harmonics.  For
example, a 2 Hz RFN has thousands of harmonics in the audible range
(20, 22, 24, 26, ...20,000 Hz) which cannot be resolved because of their
close spacing.  But they do create different periodic patterns within
each critical band.  Each pattern has the identical repetition
frequency of 2 Hz (as can be heard when a 1/3-octave is swept through
the spectrum).  RFNs above 20 Hz can be considered as stochastic
complex tones without the special amplitude and phase characteristics
of specific complex tones such as pulse trains, voices, and musical
instruments.  For example, RFNs with a repetition period of 5 ms are
200 Hz complex tones having no hint of noise; each of these individual

RFNs has a unique exceptionally rich timbre based upon its particular
amplitude and phase spectrum.

Jim Bashford and I have used RFNs to study pitch and infrapitch
perception starting in 1981 (Perception of acoustic iterance:  Pitch
and Infrapitch, Perception & Psychophysics, 29, 395-402).

Richard M. Warren
Research Professor
   and Distinguished Professor Emeritus
Department of Psychology
University of Wisconsin-Milwaukee
PO Box 413
Milwaukee, WI  53201