Psychophysics and Signal Detection Theory
1) Psychophysical Methods
-Classical psychophysical methods first developed in 19th century by Gustav
Fechner
-Measuring Thresholds
-(absolut)Threshold = smallest amount of stimulus energy necessary for an
observer to detect a stimulus
-Three different methods of measuring the absolute threshold:
1) Method of limits
>presenting stimuli in ascending or descending order
>participant says if he or she detects the stimulus
>change from „yes“ to „no“ = crossover point
>repeating starting either above or below the threshold
>absolute threshold: average of all the crossover points
2) Method of adjustment
>observer or experimenter adjusts the stimulus intensity in a continuous
manner
>absolute threshold: point of barely detection
>least accurate but fastest method
3) Method of constant stimuli
>5 to 9 stimuli presented in random order
>differenciate between detected/never detected and how often detected
>absolute threshold: intensity that results in detection on half the trials
>most accurate method, but takes the longest time
-The Difference Threshold:
-also: „just noticeable difference“ (JND)
-smallest difference between two stimuli that a person can detect
-Size of the JND depends on size of the standard weight
-Weber‘s law: JND/S = K (S:value of standard stimulus;
K:constant [„Weber fraction])
-Magnitude Estimation
-Stevens‘ (1957): measuring perception above threshold
-relationship between perceived magnitude and stimulus intensity
-response compression (light): Doubling the intensity --> only small change
in the perceived brightness
-response expansion (electric shock): Doubling the strength --> more than
doubles the sensation
-Stevens‘ power law: P = KSn (P:perceived magnitude; K:constant;
S:stimulus intensity; n:Power)
2) Signal Detection: Procedure and Theory
-Is there an absolute threshold?
-hypothetical experiment: liberal vs. conservative responder
-each has a different response criterion
-this doubts the validity of the absolute threshold
--> new procedures based on Signal Detection Theory (SDT)
-different trials but only one stimulus intensity
-Low intensity tone vs. no tone at all
-“Laurie“ and „Chris“ are presented randomly a tone vs. no tone (each 100 times)
Stimulus present Stim. not present
„yes“ Hit False alarm
„no“ Miss Correct rejection
-Larie (liberal): high hit rate but also 40% false alarm
-Chris (conservative): lower hit rate but also lower false alarm rate
-Classical theory would say: no stimulus, no response!
-Receiver operating characteristic (ROC curve) shows that constant stimuli method
would mislead in this case
-if all data points of two observers fall on the ROC curve, they have got the same
sensitivity
-a change in the subject‘s motivation, say by a payoff, could influence the
person‘s responding (without changing the tones‘ intensity!)
-example: Hit : win 100$ Correct rejection : win 10$
False alarm : lose 10$ Miss : lose 10$
--> change the criterion and stop being so conservative
-Signal Detection Theory
-(1)Percentage of hits and false alarms depends on criterion
-(2)a subjects sensitivity to a stimulus is indicated by the shape of the ROC curve
-Differenciate between SIGNAL and NOISE
-Signal: stiumulus presented to subject
-Noise: all other stimuli in the environment
-Presenting Signal and Noise = (S+N) or Only Noise = (N)
-Decision depends on the subjects criterion, if the loudness is in the middle (20)
-Distance (d‘) between the peaks of (N) and (S+N) distributions
determine the shape of the ROC curve àSubjects sensitivity influences the shape
