Psychophysics and SDT - Center for Neural Sciencemsl/courses/0022X/slides/sl... · 2012. 9. 20. ·...
Transcript of Psychophysics and SDT - Center for Neural Sciencemsl/courses/0022X/slides/sl... · 2012. 9. 20. ·...
-
9/20/12
1
Psychophysics and SDT
• Psychophysical techniques – Magnitude estimation, matching, detection/
discrimination – Forced-choice procedure
• Weber’s Law, Fechner’s interpretation • Signal detection theory
– Information, criterion, internal response and noise, ROC curve, d-prime
1. Magnitude estimation 2. Matching 3. Detection/discrimination
Three methods for measuring perception
Magnitude estimation Have subject rate (e.g., 1-10) some aspect of a stimulus (e.g., how bright it appears or how loud it sounds)..
Steven’s power law
P = k Sn
P : perceived magnitude
S : stimulus intensity
k : constant
Relationship between intensity of stimulus and perception of magnitude follows the same general equation in all senses
Matching
Demo: brightness matching
In a matching experiment, the subject's task is to adjust one of two stimuli so that they look/sound the same in some respect.
Detection / discrimination
In a detection experiment, the subject's task is to detect small differences in the stimuli.
Psychophysical procedures for detection experiments
• Method of adjustment.
• Yes-No/method of constant stimuli.
• Forced choice.
-
9/20/12
2
Concerns of the Psychophysicist
• Bias • Criterion • Attentiveness • Strategy • Artifactual cues • History of stimulation • Who controls stimulation
The method of adjustment
Ask observer to adjust the intensity of the light until they judge it to be just barely detectable
Example: you get fitted for a new prescription for eye glasses. Typically the doctor drops in different lenses and asks you if this lens is better than that one.
Do these data indicate that Laurie’s threshold is lower than Chris’s threshold?
Yes/no method of constant stimuli
Per
cent
“ye
s” re
spon
ses
Tone intensity
Forced choice – Yes/no or signal-detection task
² Present signal on some trials, no signal on other trials (catch trials).
² Subject is forced to respond on every trial either “Yes” the thing was presented” or “No it wasn’t”. If they're not sure then they must guess.
² Advantage: With the forced-choice method, we have both types of trials so we can count both the number of hits and the number of false alarms to get an estimate of discriminability independent on the criterion.
² SDT EXPERIMENT DEMO
2-interval forced-choice task ² Two intervals (spatial side-by-side or temporal)
² Observer can’t be biased to say “yes”, because there is always a stimulus.
² Observer can have an interval bias, although that might be ameliorated with trial-by-trial feedback.
-
9/20/12
3
Tumor present
Tumor absent
Doctor responds “yes”
Doctor responds “no”
Hit Miss
False alarm
Correct reject
Forced choice: four possible outcomes
Tumor present
Tumor absent
Doctor responds “yes”
Doctor responds “no”
Hit Miss
False alarm
Correct reject
Information acquisition
Tumor present
Tumor absent
Doctor responds “yes”
Doctor responds “no”
Hit Miss
False alarm
Correct reject
Criterion shift Information and criterion Two components to the decision-making: information and criterion.
² Information: Acquiring more information is good. The effect of information is to increase the likelihood of getting either a hit or a correct rejection, while reducing the likelihood of an outcome in the two error boxes.
² Criterion: Different people may have different bias/criterion. Some may may choose to err toward “yes” decisions. Others may choose to be more conservative and say “no” more often.
Internal response: probability of occurrence curves
N S+N
The criterion
Criterion
Internal response
Prob
abili
ty d
ensi
ty
Say “yes” Say “no”
-
9/20/12
4
N S+N
Hits (response “yes” on signal trial)
Criterion
Internal response
Prob
abili
ty d
ensi
ty
Say “yes” Say “no”
N S+N
Correct rejects (response “no” on no-signal trial)
Criterion
Internal response
Prob
abili
ty d
ensi
ty
Say “yes” Say “no”
N S+N
Misses (response “no” on signal trial)
Criterion
Internal response
Prob
abili
ty d
ensi
ty
Say “yes” Say “no”
N S+N
False Alarms (response “yes” on no-signal
trial) Criterion
Internal response
Prob
abili
ty d
ensi
ty
Say “yes” Say “no”
Criterion shift
N S+N
Internal response
Prob
abili
ty d
ensi
ty
Discriminability (d-prime, d’)
d-prime is the distance between the N and S+N curves
d’!
-
9/20/12
5
Discriminability (d’) Discriminability (d’) and overlap
d’ = separation
spread
SDT review ² Your ability to perform a detection/discrimination task is limited by internal noise.
² Information (e.g., signal strength) and criterion (bias) are the 2 components that affect your decisions, and they each have a different kind of effect.
² Because there are 2 components (information & criterion), we need to make 2 measurements to characterize the difficulty of the task. By measuring both hits & false alarms we get a measure of discriminability (d’) that is independent of criterion.
1
2
3
d’
Signal intensity
Low intensity High intensity
Measuring thresholds
Absolute and relative thresholds
Background intensity
Diff
eren
ce in
inte
nsity
Weber’s law
Abs
olut
e th
resh
old
Relative thresholds
-
9/20/12
6
Ernst Weber, c1850 Gustav Fechner, c1850
Background intensity
Inte
rnal
resp
onse
In
crea
se in
inte
nsity
Weber’s law
Fechner’s analysis
Weber’s law: To perceive a difference between a background level x and the background plus some stimulation x+Δx the size of the difference must be proportional to the background, that is, Δx = k x where k is a constant. Fechner’s interpretation: The relationship between the stimulation level x and the perceived sensation s(x) is logarithmic, s(x) = log(x). Main difference: Fechner’s is an interpretation of Weber’s law, a hypothesis.