Range Effects in Experiments on People E. C. Poulton

1
Range Effects in Experiments on PeopleE. C.
Poulton
Medical Research Council Applied Psychology Unit
Cambridge, England
Presented by Janeen Sharma PSYC 562
2
Articles Claim

• No psychologist should use a within-subjects
  design, except for a special purpose, without
  combining it with a separate-groups design.

3
Latin Square Design
4
Asymmetrical Transfer Example

• Two groups of people, two tasks
• Tasks are performed in the order of AB for one
  group and BA for the other group
• Transfer of training may be asymmetrical
• May cause a range effect
• Example of pursuit vs. compensatory tracking

5
A Model Range Effect Example

• Participants choose an optimal value based on
  range of available values, not the true optimal
  value
• Example of block stacking at varying table heights

6
Scanning Alphanumeric Example

• Finding a target within a matrix of 50 symbols
• Repetitive tests of searching for targets
• Participants learned which symbols were targets
• Scanned for all symbols at once
• Range Effect due to parallel processing
• Must use separate groups to eliminate range effect

7
Short-term Memory Scanning Example

• Digit recognition
• Varying length list of digits up to six
• List length of no more than four or length of
  list was constant
• Serial-position effect occurred
• Varying length lists of 5 or 6 digits
• No serial-position effect
• No consistency in memorization
• Last digits not favored

8
Time Order Errors

• First stimulus
• Underestimated if it is large
• Overestimated if it is small
• Central tendency of judgment
• Estimates tend to the middle magnitude of a series

9
Ratings of Sensory Magnitude Example

• Observers were asked to rate the noisiness of
  vehicles
• Six categories
• Middle two acceptable noisy
• Middle ratings depended on the range of vehicle
  noisiness observed

10
Stimulus Generalization

• Three categories of stimuli
• easily distinguished from each other
• Increase in the number of categories
• Middle values tend to be confused
• Adding categories
• Increases confusion between stimuli in the middle
  of the range

11
Stimulus and Response Probability

• Intentional range effect
• Stimulus probability varied
• Participants responses measured

12
Probability Learning

• Prediction is based on learned range of previous
  stimuli

13
Range Effect Theories

• Parallel Processing for a Number of Targets
  (Neisser)
• Exhaustive Serial Scanning in Short-Term Memory
  (Sternberg)
• Adaptation Level (Helen)
• Response Bias
• Response Strategies

14
Avoiding Range Effects

• Separate groups experimental design
• Range unknown to participant
• Problems
• Differences between groups can cause confounding
• Large sample size (Could eliminate confounding
  caused by differences between groups)
• Does not eliminate all range effects
• Balancing the biases
• Reversing stimuli and responses between groups

15
Avoiding Range Effects

• Selection
• To avoid biases caused by past experiences
• Cannot retrain
• May cause asymmetrical transfer

16
Implications

• Within-subjects design will always cause range
  effects
• Unintentional range effects can cause false
  theoretical formulations
• Intentional range effects can be beneficial to a
  study
• If within-subjects design and separate groups
  design conflict, use separate groups design

17
Article Analysis

• Variety of range effects possible in experiments
  on people
• Unintentional range effects can cause inaccurate
  analysis for within-subjects design

18
Article Analysis

• Strong Statement
• No psychologist should use a within-subjects
  design, except for a special purpose, without
  combining it with a separate-groups design.
• Convinced that a well designed psychology
  experiment should never rely solely on within
  subjects design?