Task 2

Question 1

Classical Method / Theory

Answer 1

Presents several different intensities of the same stimulus in the attempt to determine a person’s sensitivity to the stimulus

Question 2

Criterion / Strategy (beta, C

Answer 2

The three possible strategies with which the subject chooses how easily to respond with “yes” or “no”: Liberal, Neutral, Conservative.

Question 3

Noise

Answer 3

All other stimuli in the environment, which might be mistaken for the signal

Question 4

ROC curve

Answer 4

Responder Operating-Characteristic: Its shape indicates the sensitivity of the subject / detection mechanism. Demonstrates that factors other than stimulus sensitivity can determine the subject’s response

The ROC curve plots P(H) against P(FA) for different settings of the response criterion. Points in the lower left of the curve represent conservative responding and points in the upper right liberal responding. On the left, more NO responses are given.

Question 5

SDT

Answer 5

Presents only one intensity of the same stimulus and the trials are either Noise (no stimulus) or Signal+Noise (signal)

Question 6

Sensitivity (d’)

Answer 6

Keenness or resolution of the detection mechanisms, d^’= Separation/Spread

Question 7

Separation

Answer 7

Difference between the means of the (N) and (S+N) probability distributions.
Large separation = high sensitivity
Small separation = low sensitivity

Question 8

Signal

Answer 8

The stimulus presented to the subject / Target that needs to be detected

Question 9

SNR

Answer 9

Signal-to-noise ratio, Low (=little difference between the signal power and noise power, noise masks gaps in tone) vs. High (=a lot of difference between the signal power and noise power)

Question 10

Spread

Answer 10

Standard deviations of the probability distributions

Question 11

Saying “yes” when the stimulus is present

Answer 11

hit

Question 12

Saying “no” when the stimulus is present

Answer 12

miss

Question 13

Saying “yes” when no stimulus is present =

Answer 13

fasle alarm

Question 14

Saying “no” when no stimulus is present =

Answer 14

correct rejection

Question 15

A liberal responder

Answer 15

will have more hits and a higher false alarm rate than a conservative responder.

Question 16

Receiver operating characteristic (ROC

Answer 16

The ROC curve indicates the subjects sensitivity. If two subjects have the same ROC curve, their sensitivities are equal (their L, N and C points do not have to be equal, they just have to fall on the same line).

When the subject’s sensitivity (d’) is high, the ROC curve is more bowed. ROC curves for stronger signals bow out further than ROC curves for weaker signals.

curved=> better senistivty
Streaught=> equal chance of hits and false alarms

Question 17

a participants criterion

Answer 17

after adopting a criterion, the participant decides how to respond according to the following rule: if the perceptual effect is greater than the criterion (on the right side), press “YES”. If it is less (on the left), press “NO”.

Question 18

a) Liberal Criterion

Answer 18

• Present N: most falls to the right. So, presenting N will likely result in a YES response when N is presented. Therefore, the probability of a false alarm is high.
• Present S+N: the entire (S+N) distribution falls to the right, so the chances are high that every (S+N) will be answered with “YES”. Therefore, the probability of a hit is high.

Question 19

b) Neutral Criterion

Answer 19

• Present N: only a small proportion of (N) is on the right. The participant will therefore only rarely respond YES when only N is presented and has therefore a fairly low false alarm rate
• Present S+N: most of the (S+N) distribution is on the right. Therefore, the hit rate will be fairly high (but not as high as for the L criterion)

Question 20

c) Conservative Criterion

Answer 20

• Present N: false alarms are very low, since none of the N curve falls on the right
• Present S+N: hits will also be low, since only a small portion of (S+N) falls on the right

Question 21

A subjects’ sensitivity to a stimulus (=d’) influences the shape of the prob distribution

Answer 21

For a participant with sensitive hearing (e.g. Alice), the (S+N) curve will shift towards the right side. Alice’s high sensitivity is indicated by the large separation (d’) between the (N) and the (S + N ) probability distributions.

Question 22

When the subject’s sensitivity (d’) is high,

Answer 22

the ROC curve is more bowed. ROC curves for stronger signals bow out further than ROC curves for weaker signals. Alice’s curves (distributions) are separated more.

Question 23

d´

Answer 23

Characterization of the detectability of the signal assuming that the noise follows a normal distribution with a fixed variance, independent of the signal strength

Thee discriminability of a signal depends both on the separation and the spread of the noise-alone and signal-plus-noise curves.

• Separation = difference between the means
• Spread = standard deviation of the probability densities

 Assumption of independent and identically distributed (IID)
 Values of d’ do not depend on the participant’s criterion. It is a true measure of the internal response

can be determined by calculating the proportion of hits and false alarms.

We need to know both the hit rate and the false alarm rate to get a measure of performance that is independent of the subject’s criterion.

Question 24

external noise

Answer 24

many possible sources of external noise (stimuli in the environment like light, …)

Question 25

internal noise

Answer 25

processes that can distort a response but cannot be controlled (i.e. neural processes, …)

Question 26

The spread of the probability curves determines how well the stimulus can be detected. The signal is more discriminable when there is

Answer 26

less spread (=less noise).

The overlap between (N) and (S+N) means that for some perceptual effects this judgment (was there a tone or not) will be difficult. In this situation (i.e. loudness of 20) the subject decides depending on the location of her criterion.

Question 27

BEta

Answer 27

ratio of neural activity produced by signal and noise at Xc.
 Beta > 1 = more conservative; Beta < 1 = more liberal; beta = 1 neutral (lines intersect)

SDT can prescribe exactly where the optimum beta should fall, given the likelihood of observing a signal (signal probability) and the cost and benefits of the four possible outcomes (payoffs).

Question 28

signal probability

Answer 28

If signal and noise are both equally likely, Xc should be placed in the middle and beta should be 1. If a signal is more likely than noise, Xc should be lowered (beta < 1).

Question 29

the participant decides how to respond according to the following rule:

Answer 29

if the perceptual effect is greater than the criterion (on the right side), press “YES”. If it is less (on the left), press “NO”.

Question 30

for a person who is extremely sensitive to the signal the probability distribution moves to the

Answer 30

right, and the noise stays the same
because the area in wich the signal+ noise is hard to distinguish from just noise is very very small

(Alice’s high sensitivity is indicated by the large separation (d’) between the (N) and the (S + N ) probability distributions)

Question 31

probability distrubiton

Answer 31

The spread of the probability curves determines how well the stimulus can be detected. The signal is more discriminable when there is less spread (=less noise).

A stronger signal (greater stimulus strength) will shift the (S+N) curve to the right

Question 32

beta

Answer 32

Beta = ratio of neural activity produced by signal and noise at Xc.

 Ratio of the height of the 2 curves (divides the height of the point of the 2 curves where your criterion is set)
 Beta > 1 = more conservative; Beta < 1 = more liberal; beta = 1 neutral (lines intersect)

If signal and noise are both equally likely, Xc should be placed in the middle and beta should be 1.
If a signal is more likely than noise, Xc should be lowered (beta < 1).

Question 33

The adjustment of the optimal beta in response to signal and noise probabilities is represented as follows:

Answer 33

Beta(optimal) = P(N)/P (S)

 Defines where beta should be set, determined by the ratio of the probability with which noise and signals occur in the environment
 Where (normal) beta is set is determined by the observer

Question 34

sluggish beta

Answer 34

Experiments have shown that beta is not adjusted as much as it should be, resulting in a sluggish Beta. (see figure: people are less conservative than they should be when Beta is high and less liberal when Beta is low).

Question 35

Statistical Prediction Rules (SPR)

Answer 35

= the mathematic algorithms that specify the best tests to include in a diagnostic workup, and that calculate the odds that a condition is present

Question 36

adaptor stimuli

Answer 36

Adaptor Stimuli = a tone that alternated with noise bursts

• Rated as “continuous” when signal-to-noise ration – SNR – was low (noise masked the gaps in the tone)
• Rated correctly as “discontinuous” when the SNR was high

Question 37

target stimuli

Answer 37

Target Stimuli = identical to adaptor stimuli except that the SNR was adjusted so that the continuity of the target stimulus was ambiguous

Question 38

RIECKE – COMMENTARY ON RIECKE (2011), RECALIBRATION OF THE AUDITORY CONTINUITY ILLUSION: SENSORY AND DECISIONAL EFFECTS

Answer 38

 This is an application to test if the presentation of a sound (adaptor) changed the sensitivity and the criterion and the findings suggest that both the sensitivity and the criterion were affected. If a soft adaptor was presented, interrupted targets were rated more often as interrupted.

 - Target was more likely to be rated as “continuous” when presented after adaptors rated as “discontinuous”
- Unknown whether this effect is sensory or decisional in nature
- After effect might alter:
Perception of continuity (sensory-perceptual bottom-up process)
 Criteria applied to rate continuity (cognitive top-down process)
 Psychophysical experiment to address this

Question 39

std assumptions

Answer 39

: only when SDT assumptions hold !
 Normality assumption: distribution and shape of the curves are normal
 Making no mistakes is impossible. You need to make mistakes to calculate the criterion, etc. You need to have uncertainty in your decision making. The signal intensity needs to be near to a threshold for these experiments to work (individually determined for each participant)
 We need to have signal and noise trials
 Listeners’ internal noise contributed equally to the sensory representations of the different stimuli

Question 40

RIEKE Statistical Analysis

Answer 40

• Outcomes
  o False Alarm: report interrupted target signal as “continuous”
  o Hits: report uninterrupted target signal as “continuous

Assumptions:
Listeteners’ sensory noise levels are randomly distributed across trials
o Noise contributed additively and independently to the sensory representations of the different target stimuli

Z-scores
Hits and FA rates transformed into z-scores
Z-scores used to compute d’ and C separately for each listener

Results
Loud adaptor (lower SNR) judged as continuous far more often than the soft adaptor
As expected, because this causes the continuity illusion
Sensitivity d’ varied significantly across the adaptor conditions
= the same ambiguous stimulus could elicit two very distinct percepts, depending on the preceding adaptor stimuli
Significantly smaller d’ values following soft adaptors that loud adaptors

Decision criterion C also varied significantly across the adaptor condition
After soft adaptors, listeners used more liberal decision criteria (more inclined to report target as continuous)

THUS: Adaptors induced a sensory aftereffect (varying d’) and a decisional aftereffect (varying C)