PSYC 212 Lecture Notes - Lecture 3: Detection Theory, Psychophysics, Stanley Smith Stevens
6 Nov 2018
School
Department
Course
Professor
Continuation of psychophysics
Thresholds and the dawn of psychophysics
"each JND is perceptually equivalent"
○
Logarithmic shape: because of the Weber fractions principle, It
will always take more and more physical stimulus intensity to
produce one JND
○
Thresholding method: what is the best way to quickly obtain
the most accurate threshold?
○
Methods of constant stimuli:
Several intensities are systematically tested in a random
order
§
Most accurate, but takes a long time
§
○
Methods of limits - ascending/descending
Ascending/descending cycles; change direction when a
"yes" or a "no"
§
A bit less accurate
§
Some intensities still don't contribute much information
§
○
Staircase method
Go back as soon as there is change in response
§
Faster than the method of limits, and almost as accurate
§
○
Method of adjustment
Let the participant increase/decrease intensity in order
to identify the threshold
§
Really fast
§
Not very accurate
§
○
•
Magnitude rating
Stanley Smith Stevens (1906-1973)
○
In proportion, how much more/less intense are two stimuli of
different intensities perceived?
○
"modulus" or standard stimulus
○
Test stimulus
○
Steven's power law: S = aIb
S: sensation
§
I: intensity
§
B: controls the curvature of the function
§
A: corrects for the scaling of measurement units used for
I (not very important)
§
○
Can be logarithmic or linear
○
Absolute magnitude ratings: how intense is a percept in
relation with two "absolute" boundaries
○
Problem: is my 10/100 equal to your 10/100?
Solution: cross-modality matching
§
○
•
Signal detection theory
Detecting a square in a 'photo'
○
In general, there will be more instances when we will see a
square when there really isn't one
○
Sensitivity(d'): how well can you distinguish the stimulus from
the noise?
Your sensitivity to a stimulus is illustrated by the
separation between the distribution of your response to
noise alone and to signal plus noise
§
○
D'~= hits - false alarms
○
The criterion for saying yes vs no can change as a function of
several motivational factors related to the consequences of
making a false alarm or a miss
○
Bias reflects the general tendency to say more/less "yes" than
"no"
○
However, this doesn’t change the sensitivity
○
As you move your criterion from the left to the right, hits and
false alarms decrease in a proportional manner
○
That is why your sensitivity index (d') remains the same despite
changes in the response criterion
○
Receiver operating characteristic (ROC) curves: in studies of
signal detection, the graphical plot of the hit rate as a function
of the false alarm rate
Chance performance will fall along the diagonal
§
Good performance (high sensitivity) "bows out" towards
the upper left corner
§
○
Responses:
Miss: the square is present but you don't say you see it
§
False alarm: the square is not present but you say you
see it
§
Hit: the square is present and you say you see it
§
Correct rejection: the square is not present and you don't
say you see it
§
○
•
Lecture 3
Tuesday, January 16, 2018
1:01 PM
Document Summary
Logarithmic shape: because of the weber fractions principle, it will always take more and more physical stimulus intensity to produce one jnd. Several intensities are systematically tested in a random order. Go back as soon as there is change in response. Faster than the method of limits, and almost as accurate. Let the participant increase/decrease intensity in order to identify the threshold. A: corrects for the scaling of measurement units used for. Absolute magnitude ratings: how intense is a percept in relation with two absolute boundaries. In general, there will be more instances when we will see a square when there really isn"t one. Your sensitivity to a stimulus is illustrated by the separation between the distribution of your response to noise alone and to signal plus noise. The criterion for saying yes vs no can change as a function of several motivational factors related to the consequences of making a false alarm or a miss.
