Signal Detection Theory Flashcards

1
Q

Signal Detection Theory (Goldstein)

A

Response Criterion :

  • Criterion by which the Responder decides whether a signal was detected or not.

-

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
2
Q

ROC Curve

A

Receiver operating characteristics :

  • A curve which does plot the sensitivity against the specificity ( rate of false alarms / rate of hits)
  • In this curve it can be determined whether a person is rather liberal or conservative in their signal detection criterion.
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
3
Q

Probability Distribution

A
  • does represent the probability of a response - given a signal actually occurred.
  • entails 2 curves

N -> the probability that a given answer will be due to noise

where the curves overlaps, it is hard to decide whether one listens to noise or a Signal.

S+N –> probability that a given answer will be due to Noise and the Signal.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
4
Q

Criterion

A

Participants have their individual criterion to differentiate between noise and signal.

Pp can either be conservative, neutral or liberal.

Liberal criterion
N: most of the probability distribution to the right: good changes of “yes”: good chance of a false alarm
S+N: entire probability distribution to the right: very high chance of “yes”: good chance of a hit

Neutral criterion
N: only small proportion to the right: small probability of “yes”: low rate of false alarms
S+N: most of the distribution to the right: frequent “yes”: fairly high hit rate

Conservative criterion
N: none/ almost none of the probability distribution to the right: small probability of “yes”: low rate of false alarms
S+N: only small proportion to the right: small probability of “yes”: small amount of hits

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
5
Q

Beta

A

Ratio of neural activity produced by signal and noise at XC (correlated with XC): β= P(XIS)P(XIN)

Optimal beta: minimal probability of error, depends on likelihood of observing a signal.

β optimal= P(N)P(S)
β =1: signal occurs as often as it doesn’t
β >1: signal occurs more often than it doesn’t: criterion should be lowered
β<1: signal doesn’t occur more often than it does: should be increased

–> the criterion should be lowered or hightened depending on whether the costs are higher for false rejections or false alarms.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
6
Q

d’ sensitivity

A

Sensitivity to detect a signal.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
7
Q

Riecke Study

A

Participants were more likely to judge a discontinuous tone as continuous is they previously listened to a tone containing a clear gap (soft adaptor), rather than a tone not containing a clear gap (loud adaptor).
If the gap was clear before, ambiguity after would trip them up.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly