Engineering Psychology PSY 378S

1
Engineering PsychologyPSY 378S

• University of Toronto
• Spring 2004
• L12 Action Selection and RT

2
Outline Lecture 1

• Response Time
• Hick-Hyman Law
• Speed-Accuracy Tradeoff
• Speed-Accuracy Operating Characteristic
• Processing Stages
• ltBREAKgt

3
Outline Lecture 2

• Stimulus-Response Compatibility
• Static and Dynamic Compatibilities
• Modality Compatibility
• ltGO HOME!gt

4
Simple and Choice Reaction Time

• In a simple reaction time (RT) situation, there
  is no uncertainty what the signal is, and there
  is no uncertainty how to respond
• Like the sprinter in the starting blocks
• In a choice reaction time task, there can be more
  than one signal, and more than one type of
  response
• Each response corresponds to a signal

5
Factors Affecting Simple Reaction Time

• 1) Stimulus Modality
• RT(aud) lt RT(vis)
• 2) Stimulus Intensity
• More intense stimuli lead to shorter RTs
• Can be modeled using SDT (aggregation of neural
  evidence over time)
• Can raise or lower criterion (e.g., false start
  for sprinter)

6
Factors Affecting SimpleReaction Time

• Temporal Uncertainty
• Greater uncertainty increases RT
• Greater warning interval increases RT
  (uncertainty in internal timing mechanism)
• But not too short
• Van der Horst (1988) traffic lights found that
  short yellow light led to hi red light violation
• Increasing yellow light duration by 1 s led to
  reduced red light violation
• Expectancy
• Lowers RT

7
Factors Affecting Choice RT

• Factors affecting simple RT also affect choice
• In a choice response time situation, the subject
  is transmitting information from stimulus to
  response in the information theory sense
• Hymans (1953) experiment S observes set of
  lights responds with particular response when
  particular light flashes
• Hick-Hyman Law (H-H Law)
• Choice RT increases linearly with stimulus
  information
• RT a bHs

8
Hick-Hyman Law
RT a bHs
N Number of Alternatives
9
Hick-Hyman Law

• Slope b is about 170 ms/bitamount of extra time
  resulting from each added bit of stimulus
  information to be processed
• Can derive information transmission rate
  (bandwidth) by 1/b 0.00588 bits/ms 5.88
  bits/s
• Intercept a (around 180 ms) represents time to
  encode the stimulus and execute the response)
  factors unrelated to the stimulus information

RT a bHs
10
Hick-Hyman Law

• Doesnt matter how the amount of information in
  the stimuli is varied
• What are three factors affecting the amount of
  information?
• Number of alternatives, probability, context
• Higher probability events lower total amount of
  stimulus information
• Mean RTs (averaging across lower and higher
  probability events) shortened
•  When probability or sequential constraints
  (context) varied to affect amount of information,
  H-H Law still holds

11
Hick-Hyman Law
E1 Number of Alternatives E2 Probabilities E3
Payoffs
12
Hick-Hyman Law

• Hick-Hyman Law tested many times with different
  kinds of stimuli and responses, and is generally
  accurate
• However, there are many other factors that affect
  RTs

13
Problems for Hick-Hyman Law

• RT not just a function of number of bits
• Six variables that affect RT but not easily
  quantified using information theory
• Stimulus discriminability
• Repetition effect
• Response factors
• Practice
• Executive control
• S-R compatibility
• See text for details

14
Speed-Accuracy Tradeoff

• Speed-Accuracy Tradeoff People tend to make more
  errors when they respond more rapidly if they
  take longer they tend to be more accurate
• A reciprocity between speed and error

15
Speed-Accuracy Tradeoff

• Constant bandwidth not quite accurate
• Rather there is one level of the S-A tradeoff
  that produces optimal performance
• Information transmission tends to be optimal at
  moderate speed-accuracy sets

16
Pushing People along the Tradeoff

• When you push people to be extremely accurate,
  reaction time increases a lot for little increase
  in accuracydiminishing returns
• Effects of instructional manipulations (Howell
  Kreidler, 1963, 1964) depend on task difficulty
• To get most efficient performance
• For easy task, best to emphasize bandwidth
• For hard task, best to emphasize speed

17
SAOC (Speed-Accuracy Operating Characteristic)

• Typically plotted as logP(correct)/P(error) vs.
  RT
• Accuracy scale typically plotted in log units
• Tends to linearize the functions

18
SAOC (Speed-Accuracy Operating Characteristic)
19
SAOC (Speed-Accuracy Operating Characteristic)

• Northwest is best southeast is least good vs.
  poor performance
• analogous to d in SDT

20
SAOC (Speed-Accuracy Operating Characteristic)

• Going from southwest to northeastmoving along
  the SAOCrepresents different speed-accuracy
  tradeoff settings
• analogous to ? in SDT

21
SAOC (Speed-Accuracy Operating Characteristic)

• How could we change peoples settingscould use
  payoffs, instructional strategies
• A and B might represent two pieces of
  equipmenttwo kinds of keyboards (Dvorak vs.
  QWERTY for example)

http//www.mwbrooks.com/dvorak/
22
SAOC (Speed-Accuracy Operating Characteristic)

• Here A is better than B across the range of
  speed-accuracy settings

23
SAOC (Speed-Accuracy Operating Characteristic)

• If however, slopes are different then must
  consider relative importance of speed vs.
  accuracy

A
24
Auditory vs. Visual

• Other factors shift performance along the SAOC
• Auditory processing tends to lead to more rapid,
  error-prone performance (quick and dirty) than
  does visual processing

Visual
Auditory
25
Stress

• High stress situations tend to move us to quick
  and dirty responding
• Regulations in the nuclear industry require
  workers to wait a certain amount of time before
  responding as a result

Lo Stress
Hi Stress
26
Stages in Reaction Time

• Most information processing model generally
  assume that total RT equals sum of duration of
  number of component stages
• Do particular variables affect particular stages?
• Two general approaches
• Subtractive Method
• Additive Factors Technique

27
Subtractive Method

• Donders (1869)
• Delete operation from one condition
• Compared simple vs. choice RT (assumed former has
  no response selection stage)
• So difference in RT between two conditions should
  represent time for response selection stage
• Problem How do you know that changing task
  doesnt affect duration of first stage(s)?
• In choice RT, may perceive stimulus differently
• Think of the way youd encode the stimulus when
  you only have to respond to one

28
Additive Factors Technique

• Sternberg (1969, 1975)
• Manipulate two variables in factorial design
• If the two affect a common processing stage, will
  produce statistical interaction
• If they affect different processing stages, will
  produce two main effects

29
Additive Factors Technique
Target Letter X
Target Letter X
X
O
X
O
R
T
Disc Lo
N4
X
Y
X
O
R
T
Mask
O
X


Switch responses L/R
Task press appropriate button when see target
letter among distractors
30
Additive Factors Technique
31
Problems with Additive Factors

• Assumption that stages proceed strictly in series
  is wrong (McClelland, 1979 Meyer Kieras,
  1997a, 1997b)
• Coles et al. (1988) showed that process of
  preparation for response can proceed will
  stimulus still being perceived
• Can produce underadditive effects (delay caused
  by increasing difficulty of one factor actually
  smaller at the more difficult level of the other
  factor)

32
Additive Factors Apps

• Nonetheless, additive factors useful
• How is speed of information processing influenced
  by different environmental and individual factors
• E.g., aging, drug and poison effects, mental
  workload
• Workers exposed to mercury in industrial
  environments (Smith Langolf, 1981)
• Interaction between memory load and amount of
  mercury poisoning in bloodstream
• Implied that toxins effect was localized at
  memory retrieval stage

33
Break
34
Stimulus-Response Compatibility

• Compatibility between displayed information and
  method of response or control
• Static sense Compatibility between a display
  location and the location of the response
• Dynamic sense Compatibility between display
  movement and movement involved in the response

35
Locational Compatibility

• We have natural tendency to move or orient
  towards source of stimulation in
  environmentinfants will orient to new pictures,
  new faces
• So why not put the control and the display in the
  same location? colocation principle
• A touch screen takes this idea to the limit
• Elevator buttons
• Cant always do that so, put controls right next
  to displays (as close as possible)

36
Stovetops Revisited

• More compatibile mappings between stimulus
  display and response means fewer mental
  operations, transformations from display to
  response
• Norman called natural mappings

37
Cheating Colocation

• Can make controls and burners a single object,
  and colocate

38
Principle of Congruence

• When cant follow the colocation principle, can
  get away with congruence
• Spatial array of controls is congruent with the
  spatial array of objects being controlled

39
Fitts and Seeger Expt

• Tested predictions of congruence principle
• Best performance for each stimulus array obtained
  from the spatially congruent response array

40
Compatibility Advantage Holds Up With Training
41
Congruence and Stovetops

• The spatial array of controls is congruent with
  the spatial array of burners

42
Principle of Congruence

• Stimulus display garage doors (left and right)
• Response buttons (left and right)

Door 2
Door 1
Door 2
Door 1
To House
43
Rules

• If congruence cannot be achieved, can use simple
  rule to map stimuli and responses
• Fitts and Deininger (1954) Used a circular
  array of 8 lights and 8 controls
• Used 3 mappings congruent, L/R reversed, and
  random
• Congruent better than reversed but reversed
  better than random
• Simple rule could be used to do the reversal!

44
Design Solution Cant

• Put a slight cant or angling of one array in a
  direction congruent with the other
• If the cant is as great as 45 degrees RT can be
  just as fast as a parallel arrangement

45
Movement Compatibility

• Compatibility in the dynamic sense
• Compatibility between display movement and
  movement involved in the response
• Typically movement of the control should
  correspond to the movement in the display

46
Movement Compatibility

• Sometimes this cant be done for practical
  reasons, however
• There are common ways to show an increase move a
  control up, to the right, forward, or clockwise
• These types of common conventions are called
  population stereotypes

47
Movement Proximity

• Place moving control close to moving display
• Principle of movement proximity

Better than
48
Movement Proximity

• Can run into problem when moving control is
  placed close to the moving display
• In (b) movement proximity violates movement
  compatibility

49
Movement Proximity

• But we can arrange things so that movement
  proximity corresponds to movement compatibility
  (c)

50
Organizing S-R Compatibility
S-R Compatibility
Dynamic
Static
Colocation (Locational Compatibility)
Movement Proximity
Movement Compatibility
Congruence
51
Modality Compatibility

• S-R compatibility can be affected by stimulus and
  response modality as well as by spatial
  correspondence
• If stimulus is a light, faster choice RT for a
  manual response than for a voice response
• If stimulus is a heard digit, faster with naming
  response than with a spatial pointing response

52
Modality Compatibility
Stimulus
Light (Visual)
Heard Digit (Auditory)
?
Manual (Spatial)
Response
?
Voice (Verbal)
53
Break
54
General Summary

• Stimulus-Response Compatibility
• Static and Dynamic Compatibilities
• Modality Compatibility Switches Grouping and
  Mapping

• Response Time
• Hick-Hyman Law
• Speed-Accuracy Tradeoff
• Speed-Accuracy Operating Characteristic
• Processing Stages