Sound Threshold Experiment

Introduction

One of the most basic measurements in perception is the measurement of threshold. The threshold measures
how strong a stimulus must be before it can be detected or how different two stimuli must be before they can be
discriminated from one another. Thresholds are of interest for several reasons. Threshold measurements are
often used to reason about the nature of sensory processing.
The goal of our third experiment is to measure thresholds for detecting tones. We will use the method of
adjustment and/or a forced choiced procedure. In the method of adjustment, the subject adjusts the strength of
the test stimulus until he or she is just able to detect it. In this sense, the method is subjective. In the forced
choice procedure, the subject must identify which of two intervals contains the test stimulus on each trial.
We will go through the material in this handout in class, but the information below should serve as a reminder.

Setting Up

You will need to get the materials onto your hard disk. You do so in the usual way, copying the SoundThresh
folder from the 0044 folder to your desktop.
Before running the experiment, check to be sure the computer’s sound volume is set properly. To do this, click
on the speaker icon in the lower right part of the screen. The following window (or one something like it) will
appear:

The left-most slider controls the overall system volume, adjust it about 1/3 of the way up from the bottom (i.e.
lower than shown above) and make sure the mute box at the bottom is not checked.
Psychology 0044 Sound Threshold Page 2

Volume and Safety

Once the computer’s sound volume control has been set, there are still two ways to control the volume level:
while the program is running, you can adjust the volume of the test tone by using the up and down arrows; if the
masker is too loud, you can adjust its volume by exiting the program (type q), then changing the
“maskerAmpDB” parameter in the text of the program; now re-run the program by typing selecting the RUN
option from the TOOLS menu.
I do not think it is possible to produce unsafe sound levels using our program and the earphones we provide.
Still, it is always wise to check the volume before putting the earphones on. Simply put them near your ear
(about 2” away) and play a few tones. If it seems too loud, turn the volume down. The experiment will work
just as well or better with moderate sound volumes.

Running the experiment

1) Once you have copied the materials and set the path (using the path browser), you can run the experiment by
typing SoundThresh in the command window. The following message should appear.

Creating masker
Creating test
Ready to go, type ? for help

If you type ?, you will see a menu of other options:

<space>: play a trial

<down arrow>: decrease test intensity
<up arrow>: increase test intensity
<right arrow>: coarser adjustments
<left arrow>: finer adjustments
f: forced choice experiment
d: guess first interval
k: guess second interval
c: see current intensity
t: play test alone
m: play masker alone
s: set test intensity
q: quit
?: display this menu

Each option is available to you by typing the appropriate key.

Note: This program must be run using the MATLAB –nojvm icon, which means you’ll have a command
window but no other nice user interface windows. It also means that you’ll be editing your programs by double-
clicking on them. This editor runs separately from Matlab. Thus, you will not be able to start the experiment by
using the “Tools:Run” item from the editor window, as you have done for other programs in this class. Also,
Psychology 0044 Sound Threshold Page 3

Matlab won’t know whether you’ve done a “Save” of any changes you make, so make sure to save your
changes before re-running the experiment. SoundThresh is fragile and can cause the computer to hang in
strange and annoying ways. Once you have started the program, you must quit out of it before you leave the
MATLAB command window. (Type ‘q’ to quit, and verify that the “>>“ prompt is showing in the command
window.) If you switch to the editor or use the mouse to click on any other application while the SoundThresh
program is running, the computer may hang.
2) For setting thresholds by adjustment, you will be playing trials and deciding whether to increase or decrease
the intensity of the test tone. You play a trial by typing the space bar. The program then plays two separate
sounds in two separate intervals. One interval contains the test tone plus some noise. The other interval
contains just the noise. Which interval contains the target is chosen randomly on each trial. Sometimes it is the
first and sometimes it is the second. We call the noise the masker. The masker is played in both intervals.
If you can tell the two intervals apart, then you are able to detect the test tone and we say that it is above
threshold. If you cannot tell the two intervals apart, then you are not able to detect the test tone and we say that
it is below threshold. Your task is to adjust the intensity of the target tone until it is just at threshold.
3) Sometimes, it is difficult to be sure whether you correctly identified the interval with the target. You can use
the program to double check. If you think the target was in the first interval on the previous trial, type d. The
program will print Correct or Wrong accordingly. Similarly, if you think the target was in the second interval,
type k.
4) You use the up and down arrow keys to adjust the intensity of the target tone. The up key increases the
intensity of the target on the next trial, the down key decreases the intensity. You may play multiple trials
before adjusting the intensity.
You may adjust the initial intensity by changing the parameter “initialTestAmpDB” in the SoundThresh.m
program and then re-running the script. If the target tone is clearly audible over the masking noise at the initial
intensity, you may find that decreasing the test intensity by typing the down arrow a few times doesn’t seem to
change the intensity much each time you hit the key. You can adjust the step size of the decrease by using the
left and right arrows. Use the right arrow to make larger decreases, so that you get in the vicinity of your
threshold faster, and then when the target tone starts being hard to hear over the masking noise, use the left
arrow to decrease the step size again to help you find your threshold more precisely.
5) If you wish to hear the target or masker alone, you may type t or m to do so.
6) When you have adjusted the intensity of the target to the threshold, type q and the program will print your
threshold intensity in dB.
7) You can use the f key to run a set of forced choice trials. These will find your 82% correct threshold by
using an adaptive method, which we will talk about in class. The forced choice method usually gives more
reliable thresholds than the adjustment method, but it takes a little getting used to.

Saving your data

To get your threshold after adjustments, just type c. It will print the current test intensity. The forced choice
procedure types out your threshold at the end of a run. Type it into Excel. Or just write it down on a piece of
paper. It is not copied to the clipboard.
Psychology 0044 Sound Threshold Page 4

Finding the threshold

Adjustment
You will want to take the average of several adjustment runs to find the threshold. Compute the standard error
of the mean in the usual manner.
The method of adjustment turns out to be quite sensitive to whether or not the observer obtains feedback about
error during the experiment (in our case, by using the d and k keys to find out whether your guesses are right or
wrong). For example, you might return to a particular condition for which you have already determined your
threshold, and find that you can now near the target tone at a lower dB setting than last time. You might also
find that you can’t hear the target at all at the threshold sound level you determined last time. This might be due
to small, moment-to-moment fluctuations in the state of your nervous system, but it’s more likely that you have
simply shifted your criterion for when you are willing to say that you detected the tone. Feedback about error
often has the effect of decreasing the value of the threshold, because it can induce the observer to try harder.
In order to obtain stable estimates of your threshold, then, it is quite important that you maintain a stable
criterion for when you are willing to say you heard the target tone. Be forewarned, however! You may be
surprised how hard this is to do. Because we don’t care too much about obtaining the lowest possible
thresholds, you may want to emphasize consistency instead, and not make use of the d and k keys to see
whether your guesses are right or wrong. Set a criterion for yourself something like: “I’m about 90% positive I
heard the tone”, and stick with it.

Forced choice

The key to the forced choice procedure is concentration. If you make a careless mistake, it will produce some
inaccuracy in the data. The other thing you need to concentrate on is which of the two tones is the target.
Sometimes you can tell them apart but not identify the target. But if you pay attention, you will find you get
good at this.
It is a good idea to start the forced choice procedure at a level that is a bit above threshold. This will give you a
few easy trials before things get hard.

The parameters

On the next page are the experimental parameters you may adjust. We went over most of these in class. I will
use the words “test” and “target” interchangeably.
In the parameter set, amplitudes (both masker and test) are measurements of total noise power. In other words,
if you double the bandwidth, and leave maskerAmpDB unchanged, the amplitude per frequency will be halved
(in technical terms: the noise spectral density will be halved). On a graph of frequency versus noise intensity,
the parameter maskerAmpDB describes the area of the noise, not the height, in a log scale. Thus, for example, if
you want to keep the height constant and double the bandwidth, you'll need to increase maskerAmpDB by 6 dB.
Also note that both the test and mask are constructed by adding a bunch of discrete frequency components
together. These frequencies are integer multiples of 2 Hz (given the setting of nBaseSamples in
CreateWaveFormST.m). Thus, if you ask for a frequency of 100.5 Hz, you'll actually get 100 Hz. And, if you
ask for a 100 Hz noise with a bandwidth of 7 Hz. you'll only get components with frequencies of 98, 100 and
102 Hz.
Psychology 0044 Sound Threshold Page 5

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
% EXPERIMENTAL PARAMETERS
%
% Masker parameters. The masker is defined by its
% center frequency, bandwidth, intensity, and duration.
maskerFreq = 1000; % Masker frequency (Hz)
maskerBandWidth = 100; % Masker bandwidth (Hz)
maskerAmpDB = 72; % Masker intensity in dB
maskerDuration = 1000; % Masker duration (millisecs)

% Test parameters. The test is defined by its center

% frequency, bandwidth, and duration.
testFreq = 1000; % Test frequency (Hz)
testBandWidth = 1; % Test bandwidth (Hz)
initialTestAmpDB = 76; % Initial intensity for test in dB
testDuration = 1000; % Test duration (millisecs)

% Experimental parameters.
ISI = 250; % Interstimulus interval (millisecs)
nTrials = 40; % Number of forced choice trials per run
adjFixedOrder = 1; % Set to 1 to always play masker first
% for adjustment trials.
questPlot = 1; % Set to 0 to eliminate Quest plotting.
reverseRole = 0; % Set to 1 to reverse role of test and masker.
beepindic = 1; % Set to 1 to begin each trial with a beep
beeptime = 100; % During in msec of beep and following silence
beepDB = 76; % Beep amplitude
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

Note: the last few parameters are used to insert a beep before each of the two intervals. This is useful if you plan
on running an unmasked detection experiment (with maskerAmpDB set to zero). In such an experiment, one
interval has the test and the other has silence, so it’s hard to figure out which interval is which without a signal
to tell you when the two intervals happened!

Your assignment

Your assignment is to measure auditory thresholds as a function of something. The two obvious experiments
are to check Weber’s Law for tone discrimination or to measure the critical bandwidth auditory masking. I
outline both experiments below. I personally think the critical bandwidth experiment is more interesting, but
you may exercise your own judgment. A third possibility is to compare thresholds measured with adjustment
and forced choice.
Weber’s Law

Weber’s law relates the threshold to detect one tone in the presence of a masker to the intensity of the
masker. Strictly speaking, the test and masker should have the same frequency and bandwidth, and I
recommend you set these parameters to match each other if you decide to test Weber’s Law. Weber’s Law
states that the threshold !I of the test is related to the intensity I of the masker by the equation !I = kI . To
test Weber’s law, you would measure the threshold for a series of masker intensities and plot the results.
Psychology 0044 Sound Threshold Page 6

The results should fall along a straight line. It is easy to use Excel to fit a line to the data—see the
“Graphing in Excel” handout for details. For a good report, you should check Weber’s Law at a number of
frequencies. Also be sure to see the calibration section below.
Critical Bandwidths

The idea behind critical bands is that sounds at different frequencies are processed by different auditory
channels or mechanisms. If this is so, then a masker at one frequency will not mask a test at another
frequency that is sufficiently different, since if the two differ in frequency they will be processed by
different channels. To test this, you would measure the threshold to detect a tone in the presence of a
masker at different frequencies. For this experiment, it is a good idea to set the test bandwidth to zero and
the masker bandwidth to about 200 or so. You then measure the threshold T as a function of the masker
frequency F. There should be a lot of masking (a high threshold) when masker has the same frequency as
the test. But as the frequency of the masker increases or decreases from the test, we expect the threshold for
the test to get smaller, producing an inverted U-shaped curve when T is plotted against F. The width of the
U-shape curve tells us the range of frequencies that the mechanism detecting the test is sensitive to. This
critical bandwidth is thought to increase with test frequency, and a really nice experiment would be to try to
measure this increase in bandwidth.
Compare adjustment and forced choice

The idea here is to compare thresholds measured by adjustment and forced choice. Which is lower? Which
is more reliable? You’d want to think carefully about counterbalancing the order of your measurements to
avoid confounding methodological differences with practice effects.