Deck 2: Signal Detection and Absolute Judgement

Question 1

What is the difference between a signal detection and identification (absolute judgement) task?

A) signal detection typically involves several stimulus states or categories, identification requires only two
B) there is no difference-the two terms are synonyms
C) identification typically involves several stimulus states or categories, signal detection requires only two
D) signal detection occurs at a later processing stage than identification

Answer 1

C

Question 2

When is signal detection theory (SDT) applicable?

A) when there are two discrete states of the world: referred to as signal and noise
B) when there are no discrete states of the world
C) when there are several discrete states of the world
D) when a signal is presented and the user assigns a score to its magnitude

Answer 2

A

Question 3

Which of the following is not one of the classes of joint events in signal detection theory (SDT)?

A) hit
B) false alarm
C) false miss
D) correct rejection

Answer 3

C

Question 4

The quantification of information is influenced by three variables. Which of the following is NOT one of them?

A) the number of possible events that could occur, N
B) their sequential constraints, or the context in which they occur
C) the probabilities of those events
D) redundancy

Question 5

If 20 signal trials and 10 noise trials were presented and there were 2 hits and 18 misses, which of the following is the correct hit rate?

A) 2/20 = .1
B) 18/20 = .9
C) 2/30 = .066
D) 18/30 = .60

Question 6

Can the value of evidence variable X ever exceed the criterion X_C?

A) only when a signal is presented
B) only when noise is presented
C) when either signal or noise is presented
D) no, X_C can never exceed evidence variable X

Question 7

In SDT we represent signal and noise as a pair of distributions. Which of the following is a true statement concerning these distributions:

A) there is always some overlap between the distributions and the distributions are normal
B) there is no overlap between the distributions and the distributions are normal
C) there is always some overlap between the distributions are the distributions are not normal
D) there is no overlap between the distributions and the distributions are normal

Question 8

In SDT we represent signal and noise as a pair of distributions. Which of the following describes conservative responding?

A) X_C is placed to the left of where the distributions meet
B) the observer says "no" (signal absent) much more often than she says "yes" (signal present)
C) the observer says "yes" (signal present) much more often than he says "no" (signal absent)
D) X_C is placed where the distributions meet

Question 9

The ______ the difference between signals and noise, the greater these error probabilities become because the amount of variation in X resulting from randomness increases relative to the amount of energy in the signal.

A) smaller
B) bigger
C) more improbable
D) none of the above

Question 10

Which of the following defines β_opt in response to changes in signal probability?

A) P(N) / P(S)
B) P(S) * P(N)
C) P(S) / P(N)
D) 2 * P(N) + ½

Question 11

Optimal beta can be defined in terms of payoffs (costs and values). Which of the following should increase optimal beta?

A) an increase in the value of a correct rejection
B) an increase in the value of a hit
C) an increase in the cost of a miss
D) a decrease in the cost of a false alarm

Question 12

Sluggish beta refers to:

A) the optimal value of beta does not change with payoffs
B) empirical beta values are affected by probabilities
C) as optimal beta is adjusted by probabilities or payoffs, there is a smaller shift in empirical beta values than is necessary
D) as optimal beta is adjusted by probabilities or payoffs, there is a larger shift in empirical beta values than is necessary

Question 13

Which of the following statements about sluggish beta is false?

A) sluggish beta is more evident for probabilities than payoffs
B) sluggish beta is not related to an observer's sensitivity
C) sluggish beta is a laboratory phenomenon and does not occur in the real world
D) sluggish beta means that people cannot adjust their criterion in an optimal manner

Question 14

A key contribution of signal detection theory is that it:

A) predicts the response criterion for a particular individual
B) predicts the sensitivity for a particular observer in a particular context
C) makes a clear distinction between response bias and sensitivity
D) shows that sensitivity is less important than response bias

Question 15

A Receiver Operating Characteristic (ROC) curve plots:

A) hits against misses
B) hits against misses plus false alarms
C) false alarms against correct rejections plus hits
D) hits against false alarms

Question 16

How many points on the ROC can be generated from a 2 x 2 response matrix?

A) one
B) two
C) four
D) eight

Question 17

Points along a single ROC curve have the same:

A) bias
B) sensitivity
C) sensitivity and bias
D) false alarm rate

Question 18

Confidence levels is a:

A) useful method in shifting the criterion to an optimal level.
B) way to demonstrate risky beta setting
C) method for determining which observers are more confident in their judgments
D) better way to measure sensitivity

Question 19

A point on the positive diagonal on an ROC curve indicates which of the following:

A) extremely high sensitivity
B) a low criterion setting
C) chance responding; i.e., zero sensitivity)
D) optimal beta

Question 20

What happens when an ROC curve is plotted using z-scores?

A) the ROC curve becomes a straight line
B) the ROC curve becomes more shallow
C) the ROC curve becomes straight and shallow
D) the ROC curve cannot be represented using z-scores

Question 21

What does d' represent?

A) response bias
B) a non-parametric measure of bias
C) sensitivity
D) accuracy

Question 22

What does A' represent?

A) a non-parametric measure of response bias
B) a parametric measure of sensitivity
C) bias divided by sensitivity
D) a non-parametric measure of sensitivity

Question 23

As more dimensions are added, more total information is transmitted, but less information is transmitted per dimension.

A) integral dimensions
B) orthogonal dimensions
C) configural dimensions
D) correlated dimensions

Question 24

As more dimensions are added, the security of the channel improves, but Hs limits the amount of information that can be transmitted.

A) integral dimensions
B) orthogonal dimensions
C) correlated dimensions
D) configural dimensions

Question 25

Some dimensional pairs, particularly those spatial dimensions like the height and width of a rectangle_______ when they are combined (Garner, 1976).

A) collied
B) integrate
C) configure
D) separate

Question 26

When human operators transmit information along all dimensions of a two-dimensional (or more) stimulus, they must ________ between the dimensions. When they are asked to process one dimension and ignore changes on the others, they are _______.

A) separate, dividing attention
B) divide attention, focusing attention
C) choose, focusing attention
D) divide attention, selecting

Question 27

A study described by Swets (1998) indicated that radiologists provided with a "reading aid" (checklist of features and a response scale) were better able to distinguish a cancerous tumor from a benign cyst than radiologists without the aid. In SDT terms, this means the radiologists _______.

A) with the reading aid were more conservative than those without
B) with the reading aid were more liberal than those without
C) without the reading aid were more influenced by disease prevalence than those who used the aid
D) with the reading aid showed greater sensitivity than those without

Question 28

A physician has to weigh the value of human life against financial and other costs when making medical decisions. Rather than consider the values of hits and correct rejections and the costs of misses and false alarms, Swets suggested that physicians:

A) ignore the costs
B) develop a ratio of benefits and costs in the form of a statement
C) weight the values twice as much as the costs
D) ignore the costs and values and go with a gut judgment

Question 29

Which of the following approaches to eyewitness identification tends to reduce false alarms?

A) a sequential lineup
B) a simultaneous lineup
C) information the witness that the suspect might not be in the lineup
D) post-identification suggestions

Question 30

What is the main reason that alarm and alert systems have a low beta setting?

A) to avoid the "cry-wolf effect"
B) The cost of misses is typically greater than the cost of false alarms
C) the base rate of dangerous events is high
D) it is an inherent characteristic of automated detection systems

Question 31

Which of the following is not a solution to the mistrust caused by alarm false alarms?

A) use multiple alarm levels
B) keep the human in the loop
C) improve operator understanding of alarm false alarm rates
D) lower automated beta slightly

Question 32

In the experiment by Mackworth (1948), a pointer moved in small jumps around a clock face. At intermittent intervals a target event occurred (a double jump of the pointer). What was the key result?

A) the target event was detected
B) the target event was not detected
C) it became easier to detect the target event over time
D) it became harder to detect the target event over time

Question 33

The vigilance decrement is a result of:

A) a decrease in sensitivity
B) a decrease in the response criterion (more liberal)
C) an increase in the response criterion (more conservative)
D) both a decrease in sensitivity and an increase in the response criterion (more conservative)

Question 34

Sensitivity in a vigilance task is higher when:

A) a target's signal strength is reduced
B) there is uncertainty about time or location of the signal
C) targets and non-targets are presented simultaneously (rather than successively)
D) the event rate is increased

Question 35

Which of the following theories accounts for the observed shift in the response criterion over the course of a vigil?

A) arousal theory
B) expectancy theory
C) sustained demand theory
D) arousal theory and expectancy theory

Question 36

Which of the following theories accounts for the sensitivity decrement over the course of a vigil?

A) arousal theory
B) expectancy theory
C) sustained demand theory
D) arousal theory and expectancy theory

Question 37

Sensitivity in a vigilance task can be improved by:

A) requiring the inspector to remember the target stimulus
B) decreasing target salience
C) increasing the event rate
D) decreasing the event rate

Question 38

Which of the following is NOT an effective method for shifting the response criterion downward towards an optimal level:

A) introducing false signals
B) providing feedback to the inspector (knowledge of results)
C) training the observer to produce automating processing of the target
D) instructions about the costs of misses

Question 39

Which of the following does NOT occur in an absolute judgment task?

A) multiple stimuli are presented
B) the subject gives its stimuli its name
C) the stimuli lie along some continuum (e.g., different brightnesses)
D) the stimuli are presented in systematic order (e.g., from dimmest to brightest)

Question 40

What is the limit of human performance in an absolute judgment task, in terms of information transmitted (HT)

A) 2-3 bits
B) 2-3 stimulus levels
C) 7 plus or minus 2 bits
D) there is no limit

Question 41

What are the implications of the bow effect for human absolute judgment?

A) stimuli in the middle of a range are more likely to be misclassified
B) stimuli at the ends of a range are more likely to be misclassified
C) stimuli at the high end of the range are more likely to be misclassified
D) stimuli at the low to middle range are most likely to be misclassified

Question 42

Which of the following pairs of dimensions are most likely to be orthogonal?

A) weight and hair color
B) weight and height
C) weight and gender
D) height and gender

Question 43

As dimensions are combined orthogonally, what happens to human information transmission (i.e., total H_T)?

A) it increases with the number of combined dimensions
B) it matches perfect performance
C) the number of bits per added dimension decreases
D) it increases with the number of combined dimensions and the number of bits per added dimension decreases

Question 44

As dimensions are combined redundantly (e.g., traffic light), what happens to human information transmission?

A) it increases but is limited by the total amount of information in the stimulus
B) it increases with each increase in the number of combined dimensions
C) it decreases
D) it decreases but H_loss increases

Question 45

When humans must identify or classify three or more stimuli at different levels along a sensory continuum, a task called absolute judgment, we find that performance is relatively ____ even when there may be substantial physical ????_______ between the levels

A) poor, differences
B) better, similarities
C) poor, similarities
D) better, differences

Question 46

In a graph demonstrating typical human performance in absolute judgment tasks, the level of the flat part or asymptote of the function indicates ____.

A) the maximum amount of information in the stimulus
B) the minimum of information transmitted.
C) the experimental results
D) the channel capacity of the operator

Question 47

SDT concepts are highly applicable to many engineering psychology problems (Fisher, Schweickert, & Drury, 2006). What are the two general benefits?

A) provides a diagnostic tool that implies different corrective actions
B) provides the operator with the ability to focus attention on one dimension
C) provides the ability to compare sensitivity and quality of performance
D) a and c
E) b and c
F) a and b

Question 48

Which of the following is an area of application of SDT?

A) medical diagnosis
B) eyewitness testimony
C) alarm design
D) all of the above

Question 49

Which of the following results describes configural dimensions?

A) the size of rectangles does not affect classification of their color
B) classification performance is affected by how the dimensions of the height and width of rectangles are paired
C) classification performance is not affected by how the dimensions of the height and width of rectangles are paired
D) the height of different sized rectangles is easily classified

Question 50

Which of the following does not affect the amount of information in a stimulus

A) the number of alternatives
B) its probability
C) its sequential context
D) its salience

Question 51

Discuss the difference between sensitivity and bias in conceptual terms.

Question 52

Describe what is meant by "sluggish beta".

Question 53

Draw an ROC curve and label the axes. Indicate sensitivity and bias. What happens to the curve when z-scores are used?

Question 54

Summarize the key differences between orthogonal dimensions and correlated dimensions.

Question 55

Summarize the methods in which are useful in shifting the criterion to an optimal level and briefly describe the four of them.

Question 56

Summarize the implications of multidimensional absolute judgement for sorting/classification tasks. Be sure to discuss configural dimensions and emergent features in your answer.

Question 57

Consider the application of signal detection theory to alarm systems. What is special about this situation, in a signal detection sense?

Question 58

What is a vigilance task and what is meant by the vigilance decrement?

Question 59

What are the key results in a vigilance task (consider both sensitivity and bias).

Question 60

Describe some techniques to combat the loss of vigilance (again consider both sensitivity and bias in your answer).

Question 61

Summarize the implication of vigilance inside and outside the laboratory and the differences between the two.

Question 62

In his classic 1956 paper, George Miller discussed various experiments that used the absolute judgment paradigm. He concluded that there appears to be a limit to classification in this task. Explain how he arrived at this conclusion.

Question 63

Describe how signal detection theory works and how it is relevant to many real-world tasks.

Question 64

Summarize the major theoretical accounts of vigilance.

Question 65

What is information, and how is it relevant to display design?