Week 4 - Psychophysics and Signal Detection Theory
Thresholds concern what are Smallest changes in stimulus that can be detected
What are the magnitudes of stimuli for each sensory modality
Signal detection concerns how we detect a signal out of noise of input
Difference Threshold: smallest change in a stimulus that can be detected (JND - Just Noticeable Difference)
Absolute Threshold: minimum intensity of a stimulus that can be detected
Weber’s Law:
Size of Difference Threshold is a function of the magnitude of a reference stimulus
IE. If a weight has to be 41g before it can be discriminated from a 40g weight, the JND is 1g in this case, but 10g for a 400g weight
Constant ratio of JND and Intensity of a reference stimulus is called Weber’s fraction for that stimulus dimension. Constant ratios change based on modality of sense
Fechner’s law:
Built on Weber’s law finding that if a Weber fraction is a constant for a given stimulus dimension, then the mind might use the Weber fraction as a unit for perceiving that stimulus dimension
Fechner’s law relates to internal experience and physical environment → Psychophysics
Fechner’s law asserts that our psychological experience of the intensity of a stimulus tends to change less quickly than the actual change in stimulus intensity
It’s not easy to measure and determine thresholds as we are good perceivers, and surprisingly sensitive to changes in stimuli
Process developed to measure thresholds:
Method of constant stimuli:
Construct a set of stimuli with magnitudes ranging from above to below the presume threshold value
Present these stimuli a number of times in random order
Participants respond whether or not they detect the stimulus on each trial
Plot the proportion of detections occurring at each stimulus magnitude
The threshold is taken as the magnitude at which the stimulus is detected a criterion proportion of the time
Typically, a clear cut discontinuity between detectable and undetectable stimuli in this psychometric function
Advantages of this are that it allows the shape of the psychometric function to be established, and provides an accurate estimate of threshold, Disadvantages that pre-testing to roughly estimate the threshold is required, wastes a lot of trials which lie far from threshold (making method time-consuming) and is difficult to measure changes in threshold over brief periods with this method
Methods of limits:
Method measures the threshold without determining the shape of the psychometric function
Uses ascending and descending sets of trials
Descending series presents stimulus in small steps until participants can no longer detect
Ascending presents at subthreshold level until participants can detect stimulus
Threshold is then calculated as an average of the limits from each run
Advantages are that this method is quicker, an reasonably accurate many trials are still wasted, participants can habituate to overshoot thresholds, and overall shape of psychometric function cannot be derived
Staircase procedures:
Designed to overcome the problems with other two methods by
Involving a linked series of ascending and descending runs, but with each successive run being based on the outcome of the preceding run
Stimulus is presented either above or below threshold, and intensity is changed in small steps until a reversal (change in responses) ocurs
Direction of change is then reversed when another reversal response occurs
Procedure is terminated after a criterion number of reversals
Threshold is taken as the average of these reversal intensities
This method is even more efficient than other two, and can be modified to overcome other limitations
Standard Staircase Procedure yields 50% threshold, and by requiring two “yes” responses before the stimulus intensity is decreased, this procedure can estimate the 70% threshold
Disadvantages stem from estimation of thresholds requiring more complex calculations, particularly when modifications are required
Signal Detection:
We can never perceive stimuli under perfect conditions, and can never know whether we are perceiving the true stimuli or noise
‘On some signal trials the level of activity will be above the criterion, leading to a correct “yes” response
On other signal trials the level of activity may be below the criterion, leading to an incorrect “no” response
Some catch trials the level of activity may be above the criterion, leading to a false alarm yes response
Some other catch trials the level of activity will be below the criterion, leading to a correct “no” response (correct rejection)
The separation between the signal + noise and noise distributions tells us how sensitive an observer is to that stimulus
The measure of sensitivity is called the d-prime
We can estimate d by seeing the proportion of hits (or misses), as well as the proportion of false alarms (or correct rejections) tells us the location of the criterion relative to the signal+noise distribution
These proportions can be converted to Z scores, and then d is the sum of these distances
d is a measure of sensitivity that is independent of the response bias, in other words its possible to get the same d from a range of different response patterns
We can see this effect in a receiver operating characteristic curve (ROC curve) which shows us the range of hit and false alarm rates that yield to the same sensitivity
