Journey to the Center of the Ear: An exploration

of the principles of neural coding and auditory

psychophysics
Author: Mackenzie Andrews

ABSTRACT

Human audition depends on the ability to encode soundwave characteristics such as amplitude

and frequency into information processable by the central nervous system. The primary auditory

sensory cell is the hair cell located in the cochlea of the inner ear. Hair cells respond to sound via

mechanical disturbances of their cilia caused by pressure changes in a soundwave. The response

characteristics of hair cells encode the frequency and amplitude of a soundwave which is used for

higher-level auditory processing such as sound localization. To explore the mecanotansduction

properties of hair cells, the stimulus response of hair cells on the cockroach leg was observed as

an analogous system to the human auditory hair cell. To investigate the implication of hair cell

transduction on sound localization, psychophysics experiments were performed on a human test

subject. Our results show the capabilities and limitations of hair cell signal transduction and

subsequent effects on sound localization pathways of human hearing.

INTRODUCTION useful stimulus. This discriminative signal

The ability to perceive sound has been an encoding begins in the cochlea where the

evolutionary feat, over 300 million years in mechanics of the basilar membrane and

the making.1 The human ear has evolved to properties of the hair cells along the length of

sense the pressure changes produced by the cochlea are varied in a way such that the

soundwaves. When sound enters the ear, the hair cells at the base of the cochlea respond

pressure changes cause a vibration in the best to high frequency soundwaves and the

tympanic membrane which is attached to hair cells at the apex respond best to low

three small bones called the ossicles. The frequencies. This tonotopic arrangement of

ossicles serve to transmit the pressure hair cells contributes to the ‘place code’

changes from the air of the outer ear to the principle of frequency encoding and is

fluid-filled cochlea. As the pressure wave conserved throughout the entire auditory

travels through the fluid of the inner ear, the system. In addition to the place code, the

hair cell embedded basilar membrane is gating kinetics and ion channel properties of

displaced. Hair cells are the sensory receptor hair cells are also variable such that certain

cells for auditory transduction. The cells respond best to specific frequencies and

displacement of the basilar membrane causes increase their firing frequency in response to

a mechanical disturbance to the hair cells increased stimulus amplitude. This ‘temporal

causing an opening of ion channels and code’ depends on a hair cell’s ability to phase

subsequent depolarization of the cell. lock to a specific frequency; to depolarize

Humans use a number of mechanisms to and cause an afferent spike at the same phase

encode sound vibrations and process the angle in each cycle of the sound wave.

encoded information into a perceivable and

Higher level auditory processing such as about the strength, frequency, and timing of

sound localization happens at downstream an auditory stimulus. We hypothesized that

structures such as the Superior Olivary the cockroach hair cells would be able to

Complex (SOC). The SOC has two circuits phase lock to specific vibration frequencies

for localizing sound in the horizontal plane: and would increase their firing rate in

Interaural Time Difference (ITD) in the response to greater magnitude stimuli. Our

Medial Superior Olive (MSO) and Interaural results support the hypothesis that the hair

Level Difference (ILD) in the Lateral cells phase lock best to certain frequencies

Superior Olive (LSO). ITD localization is while suggesting that hair cells are subject to

based on the difference between sound harmonic frequency resonance. In addition,

arrival time to the left and right ears whereas our results show a sigmoidal relationship

ILD localization is based on the difference in between stimulus amplitude and firing rate.

sound loudness between the left and right In order to investigate the capabilities and

ears. limitations of sound localization processing

In order to investigate the auditory encoding in humans, we conducted a psychophysics

properties of hair cells, we used the hair cells experiment where we explored how either

on a cockroach leg as an analogous system to amplitude or time disparities in a headphone

those in the cochlea. Since the hair cells of delivered sound produced a perceived sense

the cochlea use a similar of off-center sound localization. We

mechanotransduction mechanism as the hypothesized that the perception of localized

cockroach mechanoreceptors, we used the sound is dependent on the size of the disparity

cockroach as our model organism to study the as well as the frequency of the waveform.

ways in which hair cells encode information The results from this experiment support the
hypothesis a human’s ability to localize A 1 second long, 0.1 V, 20 Hz stimulus

sound via the ITD and ILD circuits is repeated 10 times was delivered to the

dependent on the frequency of the sound as attached spine. This stimulus was repeated

well as the magnitude of the disparity. for 10 recording blocks. The data was

METHODS analyzed with the Per-Stimulus Time

Animal Preparation Histogram (PSTH) function of LabChart as

A cockroach leg was mounted such that one explained in the course manual (p.16-17). A

recording electrode pierced the femoral noise threshold of 70 mV was used for all

region of the leg while the other recording frequency response experiments. This was

electrode pierced the tibial region. One of the repeated for a 50 Hz stimulus.

spine’s on the cockroach leg was attached to Frequency Dependence of Firing Rate

the speaker probe using a ball of dental wax Using the set-up for the cockroach

as described in the course manual (p.16).2 mechanosensory lab in the course manual

Spontaneous Firing Control (p.16), the speaker probe was attached to a

A 1 second long, 0 amplitude stimulus spine on the femoral region of the cockroach

repeated 10 times was delivered to the leg. The frequency of the stimulus was

attached spine as a control for the frequency increased from 15 Hz to 60 Hz in steps of 5

experiments. The data was analyzed with the Hz (excluding 30 Hz). 10 blocks of a 1

Per-Stimulus Time Histogram (PSTH) second stimulus repeated 10 times were

function of LabChart as explained in the performed at each frequency. The averaged

course manual (p.16-17). total number of spikes per trial were

Measuring Frequency Response determined using the PSTH function and

setting the bin size to 10 seconds. The total

number of spikes for a frequency was then experiments outlined in the course manual

divided by the total number of waveform (p.21-23), the waveform was set to a 2 ms

cycles (frequency multiplied by duration of click. The waveform was played to the

stimulus) to get a ratio that represents the subject with time disparities decreasing in

phase locking fidelity of the response. intervals of 25 µs from 500 µs. The subject

Amplitude Dependence of Firing Rate stated whether the sound was perceived to

Using the set-up for the cockroach come from the left or the right. This was

mechanosensory lab in the course manual repeated with the white noise waveform.

(p.16), the speaker probe was attached to a Time Disparities

spine on the femoral region of the cockroach The waveform was set to a 1 kHz sine wave.

leg. The amplitude of the stimulus was The time disparities for 10 trials were set

increased from 0.1 mV to 0.5 mV in steps of from 100 µs to 900 µs in intervals of 100 µs.

0.1 mV. 5 blocks of a 1 second stimulus The sounds were delivered to the subject and

repeated 10 times were performed at each the subject stated whether the sound was

amplitude. The averaged total number of perceived to come from the left or the right.

spikes per trial were determined using the To determine the threshold for time disparity

PSTH function and setting the bin size to 10 lateralization, sounds with a 100 µs time

seconds. The total number of spikes for a disparity and increasing frequency (starting

given amplitude were recorded and plotted at 500 Hz, intervals of 100 Hz) were played

against stimulus amplitude. to the subject until the subject could no longer

Clicks/White Noise and Time Disparities accurately perceive lateralization.

Using the MATLAB program and

headphones for the psychophysics

Amplitude Disparities a Two-Alternative Forced-Choice manor (the

To determine the threshold for amplitude subject must choose left or right). The results

disparity lateralization, sounds with a 10% were recorded in a pre-generated table in the

amplitude disparity and decreasing frequency course manual.

(starting at 3000 Hz, intervals of 200 Hz)

RESULTS
were played to the subject until the subject
Cockroach Recording Set Up
could no longer accurately perceive
In order to begin recording, the cockroach leg
lateralization.
was placed on two electrodes, held in place
Amplitude Disparity Effect on Perceived
by a cork mounted to the microscope
Lateralization
platform (Figure 1).
Using the MATLAB program and

headphones for the psychophysics

experiments outlined in the course manual

(p.21-23), a series of 10 trials with amplitude

disparities of 2% to 20% (in intervals of 2)

were entered into the MATLAB program

Figure 1: Cockroach Recording Set Up - The cockroach
(Time Disparity = 0). The trial number and leg was mounted on two electrodes punctured
through the femoral and tibial regions of the
exoskeleton.
side of the disparity (left or right ear) were
The electrodes punctured the exoskeleton in
randomized for 20 blocks of the 10 trials. The
both the femoral and tibial portions of the
blindfolded subject listened to the sound
limb in order to record the extracellular
delivered to the headphones and told the
voltages within the leg. A probe attached to a
experimenter whether the sound was
loud speaker was positioned next to a spine
perceived as coming from the left or right in
proximal to the femoral-tibial joint on the
femoral side of the joint. The probe and A PSTH of the averaged activity over 10

attached to the spine with dental wax. blocks of 10 seconds of 20 Hz stimulus

20 Hz Stimulation Response response is shown in Figure 2.d (Noise

Before recording any stimulation response Threshold = 70 mV, bin size = 9 ms). The

from the leg, a recording of the spontaneous raster plot (shown above the histogram bars)

activity of the neurons within the leg was shows that the neuron fires consistently at the

taken. Figure 2.a shows 200 ms of same point in each trial which is also

spontaneous activity. A block of data represented by the peaks in the spike

consists of 10 sweeps of 1 second responses. histogram bins. There are 20 spike bursts in

A PSTH of the activity over 10 blocks each 1 second train which is consistent with

spontaneous activity is shown in Figure 2.b the 20 Hz stimulus. This response pattern is

(Noise Threshold = 70 mV, bin size = 9 ms). consistent with phase-locking behavior.

We then stimulated the spine with a 0.18 V, To illustrate the effect of performing many

20 Hz frequency stimulus (Figure 1.c). The trials to observe an averaged behavior, a

spine responded with a rhythmic spiking PSTH of a single trial block (Figure 3.a) and

pattern, phase locked to the stimulus of 5 trial blocks (Figure 3.b) is shown. As the

waveform at 150 degrees (peak = 90 number of trials is increased, spiking patterns

degrees). The spikes were between 91.72 mV can be observed through a data average

and 150.955 mV in amplitude and between which can be unclear when observing a

0.186 ms and 0.426 ms in duration (Figure smaller number of trials.

1.e).
Figure 2: 20 Hz Stimulation Response – a) Spontaneous activity in the leg without stimulation. b) Peri-Stimulus Time
Histogram of 10 blocks of spontaneous activity. c) Response to a 20 Hz sine wave stimulus showing phase-locking
behavior in the firing pattern. d) Peri-Stimulus Time Histogram and raster plot of 10 blocks of 20 Hz response showing
consistency in spike timing. e) Spike discriminator showing noise threshold and population boundaries.
Figure 3: Effect of Increasing Trial Number on Peri-Stimulus Time Histogram Analysis – a) PSTH of a single trial block
(10 – 1 second runs) of a 20 Hz stimulus response. b) PSTH of 5 trial blocks of a 20 Hz stimulus response.

50 Hz Stimulation Response increased, but the response was not as regular

The stimulation frequency was then as the response observed at 20 Hz. However,

increased to 50 Hz, all other stimulus and a PSTH analysis of 10 trial blocks (Noise

recording parameters were kept consistent. Threshold = 70 mV, bin size = 3.6 ms) shows

Figure 4.a shows the firing response to the 50 that the neuron’s averaged behavior still has

Hz stimulus. The neuron’s firing rate phase-locking patterns with lower fidelity.

Figure 4: 50 Hz Stimulation Response - a) Response to a 50 Hz sine wave stimulus showing increased firing rate with
decreased firing consistency. b) Peri-Stimulus Time Histogram and raster plot of 10 blocks of 50 Hz response showing
a larger distribution of firing times.
Effect of Varying Stimulus Frequency on cycles (frequency multiplied by duration of
Firing Rate
stimulus) to get a ratio that represents the
To observe the effect of varying stimulus
phase locking fidelity of the response. An
frequency on firing rate, the frequency of the
iso-intensity curve of the neurons response to
stimulus was increased from 15 Hz to 60 Hz
the various frequencies was generated
in steps of 5 Hz (excluding 30 Hz). 10 blocks
(Figure 5).
of a 1 second stimulus repeated 10 times were
The cell had the best response to the 20 Hz
performed at each frequency. The averaged
stimulus, showing a near 1: 1 spike to period
total number of spikes per trial were
ratio. The cell had a second, smaller response
determined using the PSTH function and
peak at 40 Hz. This may be explained by
setting the bin size to 10 seconds. The total
harmonic frequency resonance and the
number of spikes for a frequency was then
membrane protein properties of the cell.
divided by the total number of waveform

Figure 5: Effect of Varying Stimulus Frequency on Phase Locking Fidelity - Spike to period ratio response of the cell to various
stimulation frequencies. The ratio was calculated by dividing the total average response by the number of periods during the 10
second trial (20Hz * 100 sec = 2000 periods).
Effect of Varying Stimulus Amplitude on each amplitude. The averaged total number
Firing Rate
of spikes per trial were determined using the
A new spine was chosen to observe the effect
PSTH function (Noise Threshold = 100 mV).
of varying stimulus amplitude on firing rate.
The total number of spikes for a given
This spine was farther up the femoral part of
amplitude were recorded and plotted against
the limb. To observe the stimulus amplitude
stimulus amplitude (Figure 6).
dependence, the stimulus frequency was kept
The absolute threshold for response was 0.1
constant at 25 Hz while the amplitude of the
mV. The dynamic range of the cell was
stimulus was increased from 0.1 mV to 0.5
between 0.1 and 0.4 mV. At 0.4 mV, the
mV in steps of 0.1 mV. 5 blocks of a 1 second
response saturated (firing rate stopped
stimulus repeated 10 times were performed at
increasing).

Figure 6: Effect of Varying Stimulus Amplitude on Firing Rate – Total number of responses of the cell to various stimulation
amplitudes. The cell followed the typical sigmoidal response profile with a threshold of 0.1 mV and saturation point of 0.4 mV.
Psychophysics Experiment Set Up disparities as low as 25 µs. The fact that

smaller disparities could be perceived with

A MATLAB program was used to allow for
white noise as opposed to a click can be
dynamic simulations of localized sound
explained by the fact that white noise is the
delivered through headphones. The program
summation of many frequencies of the same
allowed for manipulation of the type of
amplitude. This more complex waveform
waveform, time disparity between the left
gives the auditory system more temporal cues
and right ears, and sound level disparity
to localize the sound.
between the ears.

Beats and Chords

Waveform Dependence of Sound

Localization Next, we observed how sine waves can add

together to produce more complex sounds.

To examine how the nature of a waveform
By generating two sine waves with relative
effects the ability to localize sound, a 2 ms
amplitudes of 1 and frequencies of 200 Hz
click was played to the subject with time
and 210 Hz, a beat frequency of 10 Hz was
disparities decreasing in intervals of 25 µs
generated. This beat is generated from the
from 500 µs. The subject was unable to
offset constructive and destructive
perceive the click as being from one side at
interference between the two sine waves with
time disparities below 100 µs. The subject
a frequency equal to the difference between
was also better at localizing sound coming
the two pure tones.
from the right rather than the left.

A C major chord was generated with three

This was repeated with the white noise
sine waves of relative amplitudes of 1 and
waveform. The subject was able to perceive
frequencies of 523.3 Hz, 659.3 Hz and 784.0
white noise as being from one side at time
Hz. The sound produced was different than a
C major chord played on an instrument such at 500 Hz, intervals of 100 Hz) were played

as a piano, because pianos are unable to to the subject until the subject could no longer

produce pure tones. The waveform of a piano accurately perceive lateralization. The

would likely include resonant frequencies, subject was unable to accurately perceive

dissipating tone amplitudes and unequal lateralization of sound at frequencies greater

relative amplitudes. than 2500 Hz. The period of the starting

frequency (500 Hz) was 2000 µs whereas the

Time Disparities
period of the threshold frequency was 400 µs.
To investigate the perception of time
As the period of the wave approaches the
disparities, the waveform was set to a 1 kHz
time disparity, it becomes impossible to
sine wave. The time disparities for 10 trials
distinguish between two offset waves and a
were set from 100 µs to 900 µs in intervals of
single cycle of one waveform.
100 µs. The sounds were delivered to the
Amplitude Disparities
subject and the subject stated whether the
To investigate the effect of amplitude
sound was perceived to come from the left or
disparities on sound localization, sounds with
the right. The subject was able to accurately
a 10% amplitude disparity and decreasing
localize the sound at all time disparities
frequency (starting at 3000 Hz, intervals of
delivered. However, as the time disparity
200 Hz) were played to the subject until the
approaches the wavelength of the sound,
subject could no longer accurately perceive
sound localization my become more difficult
lateralization. The subject was unable to
if not impossible in most cases.
perceive lateralization at lower frequencies
To determine the threshold for time disparity
(50 Hz).
lateralization, sounds with a 100 µs time

disparity and increasing frequency (starting

Effect of Amplitude Disparity on Perceived whether the sound was perceived as coming

Lateralization from the left or right. The percent of correctly

To further investigate the effect of amplitude identified lateralizations were plotted against

disparity on perceived lateralization, a series the amplitude disparities tested (Figure 7).

of 10 trials with amplitude disparities of 2% At lower amplitude disparities (2-4%) the

to 20% (in intervals of 2) were entered into subject correctly localized the sound between

the MATLAB program (Time Disparity = 0). 40% and 60% of the time, consistent with

The trial number and side of the disparity (left random guessing. At amplitude disparities

or right ear) were randomized for 20 blocks over 6%, the ability to localize the sound

of the 10 trials. The blindfolded subject increases roughly linearly until the subject

listened to the sound delivered to the was able to localize the sound 100% of the

headphones and told the experimenter time at 18-20% amplitude disparity.

Figure 7: Effect of Amplitude Disparity on Perceived Lateralization – Percent correct lateralization during a Two-Alternative
Forced-Choice experiment vs amplitude disparity of delivered sound.
DISCUSSION motor response to perhaps escape from a

The Temporal Code threat whereas in humans, this information

The human ear’s auditory sense cell is the would be sent to the auditory processing

hair cell. By studying an analogous system circuits for the ability to perceive sound.

such as the cockroach hair cells, insights into Harmonic Frequency Responses

the capabilities and limitations of human The hair cell observed for Figure 5 shows a

hearing can be discovered. The response of second response peak at 40 Hz, a first

the cockroach hair cell to the two different harmonic of the 20 Hz ‘best frequency.’

frequencies investigated in Figures 2 and 4 Harmonic resonance in hair cells is an

(20 Hz and 50 Hz), shows the ability of hair important yet underdiscussed quality of

cells to phase lock to a stimulation frequency. human hearing. Harmonic resonance of hair

Phase locking is responsible for the temporal cells contributes to harmonic distortion of

code of auditory transduction in humans. perceived sound via the outer hair cell

Figure 5 shows an example of an iso-intensity amplification pathway.3

curve for a single cockroach hair cell. The Limitations of the Temporal Code

hair cell being investigated was best tuned to As observed in Figure 4 and 5, as the

stimulus frequencies of 20 Hz. The gating frequency of stimulus increases, the ability of

kinetics of proteins in the hair cell membrane a cell to phase lock diminishes. This is due to

tune a hair cell to be able to respond best at a the fact that cells have a maximal firing rate

certain frequency. The frequency of the of approximately 1000 Hz. As the stimulus

sound is thus encoded as a firing rate that is frequency exceeds 1000 Hz, hair cells are

transmitted to the CNS. In the cockroach, this unable to efficaciously phase lock to the

information would likely be used to drive a waveform. This phase-locking drop off
partially explains why the subject was unable evoke a response in the cell and stimuli over

to localize frequencies greater than 2500 Hz the saturation amplitude cannot evoke an

using the ITD pathway in the psychophysics increase in firing rate as observed in Figure 6.

experiments. Implications on Sound Localization

The Rate Code The dynamic range of an auditory sense cell

In addition to the phase-locking behavior of determines the capabilities of higher level

frequency encoding, hair cells also use firing structures to process auditory stimuli. As

rate to encode the amplitude of a stimulus. As observed in Figure 7, the subject had a

shown in Figure 6, an individual hair cell has dynamic range of amplitude disparities in

a dynamic range in which it’s firing rate which he was able to accurately localize a

increases roughly linearly with an increase in perceived sound. The inability to localize

stimulus amplitude. This type of amplitude sounds with small amplitude disparities is

encoding the primary sound level encoding likely due to the fact that the amplitude

mechanism in the human auditory system. difference was sub-threshold of the hair cells

Like the cockroach leg, human hair cells have ability to encode amplitude.

a dynamic range of amplitudes in which the Combining the effects and limitations of

cell is able to distinguish between a relatively temporal (frequency) and rate (amplitude)

loud and relatively quiet sound. encoding of hair cells, offers insights into the

Limitations of the Rate Code mechanisms of sound localization in humans.

The dynamic range of a cell’s ability to In the psychophysics experiments, the

encode stimulus amplitude is dependent on subject was better at localizing high

the threshold and saturation points of the cell. frequency sounds with amplitude disparities

Stimuli below the threshold amplitude so not and low frequency sounds with time
disparities. This is partially due to the fact dependent on the ability of neurons to phase-

that the ILD localization pathway is lock to a frequency. The ability of neurons to

dependent on head shadowing. High phase lock to a sound decreases as the

frequency sounds are easily attenuated as frequency exceeds ~3 kHz which explains the

they pass through the head, however low reduced ability to localize sound via the ITD

frequency sounds are more robust to pathway.

attenuation. In addition, the ITD pathway is

REFERENCES

1. Cowen, Richard (2000). History of life. Oxford: Blackwell Science. p. 432.

2. Neurobiology 302, University of Washington, Autumn Quarter, 2017, “Introduction to

Systems and Behavioral Neurobiology.”

3. Olson ES. Harmonic distortion in intracochlear pressure and its analysis to explore the

cochlear amplifier. J Acoust Soc Am. 2004;115:1230–1241.