PROBLEM 3 + PROBLEM 4
Goldstein, Coren
Problem 3: PART 1 & 2
Signal detection experiment
- Studying the detection of tones
- This experiment differs from a classical psychophysical experiment in 2 ways:
o Only one stimulus intensity is presented: one single low-intensity tone that’s
difficult to hear
o On some of the trials no stimulus is presented: 100 trials with tone and 100 with no
tone, mixed
- Hit in signal detection: saying yes when there is a signal
- Miss: saying no when there is a signal
- False alarm: saying yes when there is no signal
- Correct rejection: saying no when there is no signal
- Changing people’s percentages of hits and false alarms without changing the intensity at all
manipulating each person’s motivation by payoffs
- E.g. a person who is hesitant to say “yes” (conservative), says “yes” more when they tell
her she will win rewards for making correct responses
- So, the percentage of hits and misses depends on the person’s criterion for saying “yes”
- Receiver Operating Characteristic (ROC) curve:
- C for conservative, L for liberal, N for neutral
Signal and Noise:
- Signal: stimulus presented to the person, the tone in the experiment
- Noise: all the other stimuli
- (S+N): signal and noise (green in graph a.2)
- (N): just noise, no signal (red in graph a.2)
- “perceptual effect” on the horizontal axis is what the person experiences on each trial
, Criterion:
Once a person adopts a criterion (L, N or C) s/he uses the following rule:
- if the perceptual effect is greater than (to the right)
the criterion the person says “yes”
- if the perceptual effect is less than (to the left) the
criterion the person says “no”.
Liberal criteria: (says “yes” a lot)
1. Present (N): probability of false alarm is high (say yes)
2. Present (S+N): probability of hit is high (say yes)
Neutral criteria:
1. Present (N): the person will say “yes” rarely false-alarm rate is low
2. Present (S+N): the person says “yes” frequently the hit rate will be high
Conservative: (hesitant to say “yes”)
1. Present (N): false alarms are low (say no)
2. Present (S+N): hits will be low (say no)
- Increasing the distance (d’) between (N) and (S+N) probability distributions changes the
shape of the ROC curve
- When the person’s sensitivity is (d’) is high, the ROC curve is more bowed
PART 3
Psychophysics: study of relationship between the physical stimuli in the world and the sensations
that we experience
3 problems:
- Detection: finding a way to measure the minimum intensity of a stimulus that we can
perceive
- Discrimination: devising a way to measure how different stimuli must be before they no
longer appear the same
- Scaling: describing the relationship between the intensity of the stimulus and the intensity
of our sensation
Goldstein, Coren
Problem 3: PART 1 & 2
Signal detection experiment
- Studying the detection of tones
- This experiment differs from a classical psychophysical experiment in 2 ways:
o Only one stimulus intensity is presented: one single low-intensity tone that’s
difficult to hear
o On some of the trials no stimulus is presented: 100 trials with tone and 100 with no
tone, mixed
- Hit in signal detection: saying yes when there is a signal
- Miss: saying no when there is a signal
- False alarm: saying yes when there is no signal
- Correct rejection: saying no when there is no signal
- Changing people’s percentages of hits and false alarms without changing the intensity at all
manipulating each person’s motivation by payoffs
- E.g. a person who is hesitant to say “yes” (conservative), says “yes” more when they tell
her she will win rewards for making correct responses
- So, the percentage of hits and misses depends on the person’s criterion for saying “yes”
- Receiver Operating Characteristic (ROC) curve:
- C for conservative, L for liberal, N for neutral
Signal and Noise:
- Signal: stimulus presented to the person, the tone in the experiment
- Noise: all the other stimuli
- (S+N): signal and noise (green in graph a.2)
- (N): just noise, no signal (red in graph a.2)
- “perceptual effect” on the horizontal axis is what the person experiences on each trial
, Criterion:
Once a person adopts a criterion (L, N or C) s/he uses the following rule:
- if the perceptual effect is greater than (to the right)
the criterion the person says “yes”
- if the perceptual effect is less than (to the left) the
criterion the person says “no”.
Liberal criteria: (says “yes” a lot)
1. Present (N): probability of false alarm is high (say yes)
2. Present (S+N): probability of hit is high (say yes)
Neutral criteria:
1. Present (N): the person will say “yes” rarely false-alarm rate is low
2. Present (S+N): the person says “yes” frequently the hit rate will be high
Conservative: (hesitant to say “yes”)
1. Present (N): false alarms are low (say no)
2. Present (S+N): hits will be low (say no)
- Increasing the distance (d’) between (N) and (S+N) probability distributions changes the
shape of the ROC curve
- When the person’s sensitivity is (d’) is high, the ROC curve is more bowed
PART 3
Psychophysics: study of relationship between the physical stimuli in the world and the sensations
that we experience
3 problems:
- Detection: finding a way to measure the minimum intensity of a stimulus that we can
perceive
- Discrimination: devising a way to measure how different stimuli must be before they no
longer appear the same
- Scaling: describing the relationship between the intensity of the stimulus and the intensity
of our sensation
