5D Library › Turning the Hearing Inward · 中文
Chapter 2. How the Ear Hears Itself
The sutra speaks of turning the hearing back. In auditory science the closest question is how we hear the sound we are making ourselves. This chapter covers only published research, and marks the author’s hypothesis at the end.
Two routes: air and bone conduction
Evidence
When we listen to someone else, sound reaches the ear canal only through the air. When we speak or sing, sound also travels through vibration of the skull and soft tissue straight to the inner ear; this is bone conduction. In 1949 Georg von Békésy analysed the structure of the middle ear and the part bone conduction plays in hearing one’s own voice (Békésy 1949). Stefan Stenfelt and Richard Goode later reviewed how bone-conducted sound reaches the cochlea by several paths at once: vibration of the skull, inertia of the ossicles, sound radiated into the ear canal, and pressure in the fluids of the inner ear (Stenfelt and Goode 2005).
Evidence
How much of our own voice arrives by each route? Sabine Reinfeldt and colleagues built an ear muff that blocked air-conducted sound without trapping sound in the ear canal, so they could measure the bone-conducted part separately. For most of the ten phonemes they tested, the bone-conducted part of one’s own voice dominated between about 1 and 2 kHz, and the balance between the two routes varied from one speech sound to another (Reinfeldt et al. 2010).
This is why we sound strange on a recording: the recording captures only the air-conducted part, and the voice we are used to hearing is a blend of both. For a singer it means that what you hear of yourself and what the audience hears were never the same. Judging tone and pitch only by how you sound to yourself is biased from the start, and the bias changes from vowel to vowel.
Calculation
The two routes arrive at almost the same time. From the mouth to the opening of the ear canal is roughly ; at the speed of sound in air, , the air route takes The difference between the routes is therefore not an audible echo. It is a difference of filtering: each route favours different frequencies, and at the cochlea the two add together. Even a fraction of a millisecond matters when they add, because what counts is the phase difference , which grows with frequency:
c = 343.0 # speed of sound in air, m/s
d = 0.15 # mouth to ear canal, m
dt = d / c # seconds
print(f"air-route delay: {dt*1000:.2f} ms")
for f in (220, 440, 1000, 2000):
print(f"{f:5d} Hz: one period {1000/f:5.2f} ms, 0.44 ms = {360*f*dt:6.1f} degrees")
# air-route delay: 0.44 ms
# 220 Hz: one period 4.55 ms, 0.44 ms = 34.6 degrees
# 440 Hz: one period 2.27 ms, 0.44 ms = 69.3 degrees
# 1000 Hz: one period 1.00 ms, 0.44 ms = 157.4 degrees
# 2000 Hz: one period 0.50 ms, 0.44 ms = 314.9 degreesAt the fundamental of a low voice the shift is small; in the range where the bone route is strongest it is large enough for the two routes to reinforce or partly cancel each other. That is one reason why the inner sound of one’s own voice has a colour no microphone reproduces.
The occlusion effect: a test anyone can do
Evidence
Close your ear canals gently with your fingertips and hum. The hum becomes louder and darker. This is the occlusion effect: with the canal blocked, low-frequency sound that bone conduction radiates into the canal cannot escape, and it is reflected back to the eardrum (Stenfelt and Goode 2005). It is the simplest way to hear the bone-conducted part of your own voice more or less on its own, and it is the first exercise in Chapter 3.
The ear turns your own voice down
Evidence
The middle ear contains a tiny muscle, the stapedius, which contracts in response to loud sound and reduces the energy reaching the inner ear. Borg and Zakrisson concluded that the stapedius reflex extends the dynamic range of the ear by about 15–20 dB (Borg and Zakrisson 1973). The muscle is also active when we vocalise: Borg and Zakrisson recorded stapedius activity in people speaking and singing (Borg and Zakrisson 1975), and standard accounts of the reflex note that it is triggered in anticipation of the voice, before the sound itself could have set it off (Møller 2000).
Evidence
The brain turns your own voice down too. John Houde and colleagues used magnetoencephalography to compare two conditions: participants speaking, and the same recording played back to them. The auditory cortex responded markedly less while they were speaking, an effect called speaking-induced suppression (Houde et al. 2002). The usual explanation is that when the brain sends the command to speak it also predicts what it will hear, and the part that matches the prediction does not need full processing.
The brain keeps correcting pitch
Evidence
If equipment quietly raises or lowers the pitch of the voice a person hears in their headphones, they shift the pitch they actually produce in the opposite direction without meaning to. Theresa Burnett and colleagues measured this pitch-shift reflex systematically in 1998 (Burnett et al. 1998). When the change is kept up over many trials, speakers adapt: John Houde and Michael Jordan showed that people who hear their vowels systematically altered gradually change the way they produce them, and the change persists for a while after normal feedback returns (Houde and Jordan 1998).
Evidence
Models of speech production describe this as two kinds of control working together. In the DIVA model of Jason Tourville and Frank Guenther, a learned command drives the voice forward (feedforward control), while an auditory and a somatosensory comparison check the result and correct errors (feedback control) (Tourville and Guenther 2011). Feedback control is slow compared with the command itself, but it keeps the command honest; over time, corrections are folded back into the command.
Calculation
Pitch error is usually expressed in cents; a semitone is 100 cents: If the target is and the singer produces , the error is cents, about a quarter of a semitone; gives about cents.
import math
def cents(f, ref):
return 1200 * math.log2(f / ref)
for f in (446, 436, 441):
print(f, round(cents(f, 440), 1))
# 446 23.4
# 436 -15.8
# 441 3.9Chapter 8 uses this quantity as its primary outcome.
Singers also steer by feel
Evidence
When noise masks their own voice, singers’ pitch becomes less accurate, but it does not collapse. Dirk Mürbe and colleagues found that masking reduced the pitch accuracy of voice students but did not remove it, and concluded that kinaesthetic feedback (the feel of the voice) carries part of the control (Mürbe et al. 2002). Boris Kleber and colleagues found that singing experience changes how the brain combines what it hears with what it feels: when auditory feedback was masked, the right anterior insula, a region that integrates bodily signals, was engaged more in trained singers (Kleber et al. 2013).
Imagining a note helps produce it
Evidence
Peter Pfordresher and Andrea Halpern asked 120 people to imitate pitches and to rate how vivid their auditory imagery was. Those who sang more of the trials in tune reported more vivid auditory imagery; the correlation was modest () and the study cannot show cause and effect (Pfordresher and Halpern 2013). Earlier work had found that most “poor-pitch” singers can tell pitches apart perfectly well: the problem lies in mapping what they hear onto what they do, not in hearing (Pfordresher and Brown 2007).
This finding matters for the third stage of the method, which asks the singer to hear the note internally before singing it (Chapter 5).
The inner ear’s amplifier answers to the brain
Evidence
In 1978 David Kemp discovered that the ear itself emits faint sounds, otoacoustic emissions, showing that the inner ear contains an active amplifier (Kemp 1978). Through the medial olivocochlear bundle, efferent nerves running from the brainstem to the inner ear, the brain can turn down the gain of this amplifier; John Guinan’s review covers its anatomy, function and measurement in humans (Guinan 2006). Sandrine Maison and colleagues found that cochlear activity changes when people focus their attention on a sound (Maison et al. 2001).
In other words, how we listen is not only a matter for the cortex. A change of attention can reach all the way to the inner ear.
Meditation and the default mode network
Evidence
Judson Brewer and colleagues compared experienced meditators with novices. During meditation, experienced meditators showed lower activity in the main regions of the brain’s default mode network, and different connectivity (Brewer et al. 2011). The default mode network is often linked with mind-wandering and self-related thinking. The study concerns meditation in general, not the Ergen Yuantong method.
Hypothesis: what turning the hearing back might correspond to
Hypothesis · not established · The physiology of turning back
The author’s hypothesis: turning the hearing back, that is, moving attention from the objects of sound to listening itself, may affect singing by three routes.
- Steadier monitoring. When attention no longer rises and falls with judgements of whether the sound is good, the brain’s prediction and comparison of its own voice (the machinery behind speaking-induced suppression and the pitch-shift reflex) is disturbed less, and pitch becomes steadier.
- Less compensating effort. A singer who stops trying to make the inner sound match an imagined outer sound has less reason to push. Effort that is not needed for the note shows up as tension in the muscles around the larynx (Chapter 3).
- Regulation of inner-ear gain. The shift of attention changes the activity of the medial olivocochlear bundle, and with it the balance between how much the inner ear amplifies one’s own voice and the surrounding sound.
None of these has evidence yet. The first can be tested directly by measuring pitch accuracy; the second by self-rated effort and, with equipment, by surface electromyography of the neck; the third needs equipment for otoacoustic emissions. Chapter 8 sets out how. If the experiment shows that the practice group’s pitch did not improve more than the control group’s, the first point is refuted.
Summary
- We hear ourselves through air and bone conduction; the bone route dominates between about 1 and 2 kHz for most speech sounds, so what we hear of ourselves differs from what others hear.
- The two routes arrive within a millisecond of each other; they differ in filtering, not in timing.
- The stapedius muscle, speaking-induced suppression and the pitch-shift reflex show that the body is constantly processing its own voice; singers also steer by feel.
- Vivid auditory imagery goes together with more accurate pitch imitation.
- The amplifier of the inner ear is under the brain’s control, and attention can affect it.
- Whether the Ergen Yuantong method improves pitch through these mechanisms is the author’s hypothesis, to be tested in Chapter 8.
References
Fan, Y. (2026). Turning the Hearing Inward: A Bioacoustic Study of the Ergen Yuantong Listening Method [反闻闻自性] (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23666557 · Licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0).