Psychoacoustic Methods

Psychoacoustics is concerned with how we perceive the physical stimuli impinging on our senses; the perception of sound

Psychophysics - how we perceive they physical stimuli impinging on our senses
- the study of the relationship between the physiological and physical aspects of a stimulus
- how does the perception relate to the actual stimulus
Psychoacoustics - the branch of psychophysics that deals with the perception of sound
- the science concerned with the psychological correlates of the physical parameters of sound
  - aka auditory perception
- making a sharp distinction between the physical stimulus and the psychological response to it
- includes coding of sounds, processing sound, integrating sound and responding to sound
Pitch is the perception related to sound frequency; pitch gets higher as frequency increases

Psychoacoustics & Clinical Audiology

It is important to understand how the normal auditory system functions
- then we can begin to consider how auditory pathology may affect the system
- it also provides information to develop models of the normal system to predict how the system operates
- it provides information for the development of methods and devices for remediation of hearing loss
Many of the diagnostic tests used clinically are based on our understanding of how the ear functions
- e.g. clinical masking based on concept of the critical band
Audiologic equipment incorporates information about normal hearing auditory processing

Conducting Hearing Research

What is an experiment?
- It is a collection of research designs which use manipulation and controlled testing to understand causal processes
- Generally, one or more variables are manipulated to determine their effect on a dependent variable
- The experimental method is a systematic and scientific approach to research in which the researcher manipulates one or more variables and controls and measures any change in other variables
What is an independent variable?
- a variable (often denoted by x ) whose variation does not depend on that of another.
What is a dependent variable?
- a variable (often denoted by y ) whose value depends on that of another
What do you want to determine?
- DV = f(IV) the functional relationship between the two variables
- looking at how the DV changes when the IV is manipulated
  - e.g. change in frequency difference between 2 sounds (IV) and measure the ability to perceive a difference in pitch (DV) or change the intensity difference between 2 sounds (IV) and measure the ability to perceive a difference in loudness (DV)
- A problem is that in most cases there are other variables in addition to the IV that may affect the DV
- It is important to know about these variables especially if they covary with the IV
  - imagine you are interested in the relationship between food and fluid intake
    - you think that the more a person drinks the less they will eat
  - no suppose that the amount of physical activity is increased to increase the amount of fluid the person drank
    - then fluid intake and physical activity covary together
  - if you are really interested in just the relationship between fluid intake and food intake then the fact that fluid intake and physical activity covary confounds the results
  - so physical activity is a confounding variable (CV) in this experiment
  - to conduct a scientifically sound experiment, need to control the CVs
One of the most important aspects of any experiment is presenting the findings

Methods in Psychoacoustics

Two general approaches to studying the relationship between the psychological and physical aspects of a stimulus
- discrimination
  - are the stimuli the same or different
  - indirect means of determining what a subject can respond to
- scaling procedures
  - ask the subject directly about the stimuli
  - direct estimate of what subject responds to

Scales of Measurement

Our ability to properly analyze what we find and to arrive at valid interpretations depends on knowing the properties of the measurements made and the qualities of the resulting data
Nominal scales are the least restrictive in the sense that the observations are simply assigned to groups
- makes no assumptions about the order among the classes; thus it is the least restrictive and least informative of the levels of scaling
- example: male vs female; type of subcompact car (toyota, ford, chevrolet, etc)
- Allows us to express the mode of the data
  - which category contains more cases than any other
- lowest order scale
Ordinal scales imply that the observations have values, which can be rank-ordered so that one class is greater or lesser than another with respect to the parameter of interest
- it does not tell how far apart they are
- gives rank of the order of the categories but does not specify the distances between them
- a>b>c>d>e
- ordinal scales permit the use of the median
  - the value with the same number above and below it
- observations can be rank ordered
Interval scale specifies the order among categories and the fixed distances among them
- the distance between any two successive categories is equal to the distance between any other successive pair
- do not imply a true zero point
- fixed distance and specific order among categories
Ratio scales include all the properties of interval scales as well as an inherent zero point
- permits values to be expressed as ratios or in decibels and the use of mathematical operations
- give the most information about the data and their interrelationships
- like interval scale but with a true zero

Summary

In summary, nominal variables are used to “name,” or label a series of values.
Ordinal scales provide good information about the order of choices, such as in a customer satisfaction survey.
Intervalscales give us the order of values + the ability to quantify the difference between each one.

Finally, Ratio scales give us the ultimate–order, interval values, plus the ability to calculate ratios since a “true zero” can be defined.

Psychophysical Procedures

two general classes of methodology
- classical methods
- signal detection theory procedures
- responses are also affected by response proclivity or bias
- classical methods cannot entirely control or eliminate bias, so procedures incorporating signal-detection theory are often used
- the procedures we will discuss attempt to measure sensitivity without the effects of response bias

Terminology

Stimulus - an auditory presentation of a sound
Parameter - characteristics of the stimulus -- intensity, frequency, duration
Run - a number of individual presentations of the stimulus
- varies with the type of experiment and stimuli
Trial - each presentation of the stimulus or group of stimuli for which a response is required
Block of trials - in some experiments a number of trials, 1 or more, is considered together as a block for purposes of evaluating the results
Step Size - increment by which the stimulus is changed from trial to trial
Catch trial - aka blank trial - a trial in which either no stimulus is presented or stimuli that are exactly the same are presented
Reversal - in an adaptive method this refers to when the direction of change goes from increasing to decreasing or decreasing to increasing
Interval - aka token - part or all of a trial depending on the type of measurement
Threshold (nominal definitions)
- Absolute threshold - the lowest intensity at which a stimulus can be detected
- Discrimination threshold - The minimal amount of change in a stimulus that can be detected
  - often called JND (just noticeable difference) or DL (difference limen)

Measurement Methods

Establishing relationships between the sound presented and how the subject perceives it is a primary goal
The stimulus and response are clearly specified and then some aspect of the stimulus is manipulated
The subject's task is to respond in a predetermined manner so that the investigator can get an unambiguous idea of what was heard
- absolute sensitivity - lowest level at which the sound is heard
- differential sensitivity - the just noticeable difference
We must also distinguish between what the subject actually hears and the manner in which he responds
- sensory capability - former (for the most part what we are interested in)
- response proclivity - latter (reflects bias and criteria that affects response)

Classical Methods of Measurement

METHOD OF LIMITS!

Method of Limits

The stimulus is under the investigators control and the subject simply responds after each presentation
determine the stimulus intensity at which the observer can either no longer detect the stimulus as its intensity is decreased from a high level, or begin to detect the stimulus as its intensity is increased from a low level
experimenter controls the stimulus in discrete steps
stimuli are presented in ascending or descending order
subject says present/absent or same/different after each presentation
do a number of runs (individual estimates) and get an average - this is threshold or DL
method is also affected by HABITUATION; the subject doesn't change response till the threshold is exceeded
- descending runs - a stimulus is lowered until impersceptible
- ascending runs - a stimulus is raised until persceptible
- "threshold" for each series is taken as the halfway point between them
- "threshold" is obtained by averaging the threshold levels across runs
- it is customary to define a threshold at which the sound is heard 50% of the time
several response bias are associated; method is limited because it's affected by RESPONSE BIAS
- the subject can anticipate when the response should change
  - anticipation results in a lower (better) ascending threshold
- the subject can habituate
  - does not change their response until threshold is exceeded by a few trials
limited in terms of step size
- too large of a step size reduces accuracy
  - may place a level within 0% probability of a response
- a smaller step size permits a more precise estimate of threshold
psychometric function shows the probability of responses for different stimulus levels
the method of limits with an appropriate step size is still a popular
- particularly true in pilot experiments and in clinical evaluations
may be used to determine differential thresholds
- two stimuli presented in each trial
to minimize bias
- do both ascending and descending runs
- vary the starting level
- don't tell the listener if you are ascending or descending
- present a different number of stimuli in each series
Examples of Method of Limits
- two point discrimination test; observable distance of two objects on the skin
  - ascending trial; distance gets bigger
  - descending trial; distance gets smaller

Method of Adjustment

Differs from the method of limits in two ways
- the stimulus is controlled by the subject instead of by the investigator
- the level of the stimulus is varied instead of in discrete steps
The observer, not the experimenter adjusts the stimulus intensity until the observer can either no longer detect the stimulus or just begin to detect it
the subject is told to bracket the threshold, go below and then above, etc.
problems associated
- the subject can change criterion during a run

Method of Constant Stimuli

sort of a tabulation of the method of limits
the stimulus is presented a specified number of times at various levels in random order
equal number of trials at each level
trials are presented non-sequentially - not ascending or descending
subject indicates if the signal is present/absent or same/different on each trial
Involves presentation of various stimulus levels to the subject in random order
- it is a non-sequential procedure
- the stimuli are not presented in an ascending or descending manner
- a range of intensities is selected which, based upon previous experience or a pilot experiment, encompasses the threshold level
- a step size is selected and the stimuli are then presented to the subject in random order
- enables the investigator to include "catch" trials over the course of the experiment
  - these are intervals during which the subject is asked whether a tone was heard, when no tone was really presented
after all trials completed, plot a PSYCHOMETRIC FUNCTION
- a graph indicating the % correct as a function of the stimulus parameter being changed
- relates a dependent and independent variable
- if step size is too large, decreases accuracy, too small, waste trials
- Estimate of threshold of DL is taken from the psychometric function
  - experimenter decides apriori what level of performance they want to measure
  - after the psychometric function is plotted can read value off the curve
may end up with a lot of wasted trials at the top and bottom of the function where all correct or all incorrect
can estimate bias by including "catch" trials
- no stimulus is presented in some trials
- this will reflect subject bias
procedure is more precise than method of limits or adjustment but less efficient because so many trials are needed

Variation in Sensitivity

The absolute threshold is not a rigidly fixed value
sensitivity fluctuates irregularly so that a given stimulus level may trigger a response at one time but not another
the threshold is usually defined statisticall
sensitivity is also subject to systemic variation, either of the the permanent kind encountered in aging or in pathology of the sensory tissues or of a temporary kind observed
these and many other systemic changes in sensitivity are frequently expressed as alterations of the absolute threshold

Adaptive Procedures

Designed to take advantage of signal detection theory
all adaptive procedures converge on threshold and make most of the measurements near threshold
procedures maximize precision and efficiency by not wasting trials at ends of underlying psychometric function
many different adaptive procedures in use in psychophysical experiments
- number of intervals used varies as well as the number of trials in a block
- stimulus level depends on previous response
Beskesy tracking
- tracking is fast and fairly precise, can adjust rate at which threshold is tracked
- if attenuation rate is too fast, reaction time may become a factor and the precision of the measure may decrease
- up down staircase method
  - converges on 50% correct on psychometric function
Designed to control for bias in a response
- focus measurement on actual sensitivity
All adaptive procedures converge on threshold and make most of the measurements near threshold
procedures maximize precision and efficiency by not wasting trials at ends of underlying psychometric function
many different adaptive procedures in use in psychophysical experiements
number of intervals used varies as well as the number of trials in a block
stimulus level depends on previous response

Adaptive Procedures - Transformed - What if we want 70% correct instead of 50%?

modification of simple up-down procedure
the level of the stimulus changes after a certain sequence of responses
can target any point on psychometric fx you want
- target level can be changed by setting criteria for increasing the stimulus level after a negative response or a positive response followed by a negative one and lower the stimulus level after two successive positives.
- in other words we have established the following rules for changing stimulus level
  - up rule: - or +,-
  - down rule +,+
  - then, you have to use math using formula

the numbers indicate reversals; an example of simple up-down method; 50% of performance is being targeted; threshold is calculated by average of the midpoints of the runs; or as the average of the peaks and troughs

Psychometric Function

The plot of proportion of correct detections as a function of level is called the psychometric function
for each level, the proportion of times that the subject correctly heard the sound is determined
a psychometric function for frequency discrimination would show the proportion of times the listener could tell that the frequency of a tone had changed as a function of the size of the frequency change
there is no common sense threshold
there is no level above which the listener always hears the sound and below which he never hears the sound
- rather there is a gradual improvement in detection with level
we define the threshold as the level at which the listener achieves some arbitrary proportion of correct detections
THE METHOD OF CONSTANT STIMULI
- provides complete picture of sensitivity
- easy to administer
- have to know what the threshold is going to be (approximately) before you start
- if you just want a threshold, a lot of trials and time are wasted

Response Bias

several forms of response biases
- anticipation
- habituation
- persistance/perseveration
When the sound is near threshold, it is hard to tell the difference between "sound" and "no sound"
- people use the evidence their ears and brain provide but whether they say "yes" or "no" depends on
  - how much they like to say "yes"
  - other things they know about the stiuation

Two alternative forced choice 2AFC

if the listener listens to a signal and to a "no signal" and then chooses the one that was the signal, then bias applies equally to the two observations and is cancelled out

Conclusions

Psychophysical methods vary from inefficient and limited in scope to efficient and complete
Bias always affects the results of a psychophysical experiment; methods to control bias are available
The psychometric function provides information about attentiveness and memory
The most popular means of studying hearing sensitivity is an adaptive 2AFC method
- bias free thresholds can be estimated efficiently using this combination