12
6. Methodological Issues and the ABAB Design: This page describes issues associated with causal
inference and ethical issues related to the ABAB design.
7. Multiple-Baseline Design: This page introduces the multiple-baselines design.
8-9. Illustration of Multiple Baseline Design: These two pages illustrate an example of multiple baselines
across behaviors, and the second addresses frequently asked questions.
10. Problems with All Single-Case Designs: Problems associated with baselines and external validity are
described on this page.
13
Single-Case Research Designs
• Idiographic rather than nomothetic
• Often used in clinical psychology and neuropsychology
• Two major types
▪ Case study
▪ Single-case experimental designs
• Case study
▪ Intensive description and analysis of single individual
▪ Clinical observations, self-report, archival data (e.g., medical records)
o Typically report results of treatment
▪ Major problem: Lack of scientific control
o Simultaneous treatments, extraneous variables
• Advantages
▪ Source of ideas, hypotheses about behavior
▪ Opportunity for clinical innovation
▪ Method to study rare phenomena
▪ Challenge to theoretical assumptions
▪ Tentative support for a psychological theory
▪ Complement to nomothetic approach
• Disadvantages
▪ Difficult to draw cause-and-effect conclusions (limited internal validity)
▪ Possible biases when interpreting outcomes (e.g., observer bias, bias
in data collection)
▪ Problem of generalizing findings from a single individual (limited
external validity)
14
Case Study Method vs. Single-Case Experiments
• Points to remember about case studies
▪ Anecdotal evidence and “testimonials” are not scientific.
▪ Research reports in popular press and media are not necessarily
based on good science.
▪ People want to believe treatments described in testimonials will work
for them, but they often don’t.
▪ Better research: Single-case experiments
• Single-Case Experimental Designs
▪ Based on B.F. Skinner’s approach called applied behavioral analysis
o Greater scientific control
• Critical feature: Independent Variable
▪ Treatment
▪ No–treatment control (baseline stage)
• Compare treatment and baseline conditions
▪ Monitor behavior continuously
▪ Search for discontinuity in behavior following implementation of
treatment
15
Baseline Stage
• Describe behavior prior to treatment
▪ Predict future behavior with no treatment
o Average of baseline observations
• Suppose a treatment is implemented to decrease a behavior.
• This figure shows treatment seems effective.
Days
1
5
10
0
5
10
15
responses
Baseline Projected future
Days
Frequency of Responses
1 5 10 15 20
0
5
10
15
Baseline Treatment
16
▪ Some other factor at time of treatment might have changed behavior.
▪ Can’t make causal inference
▪
ABAB Design
• Contrast Baseline (A) and
Treatment (B) conditions
• Illustration of a treatment
effect
▪ Goal: decrease
behavior
▪ Frequency of behavior
decreases during
treatment (B)
▪ Reverses when
treatment is withdrawn (2nd A)
▪ Reverses again during treatment (2nd B)
▪ “Reversal design”
• Illustration of no reversal
▪ Behavior does not reverse when treatment is withdrawn (2nd A).
▪ Treatment looks very effective, but…
▪ It’s
difficult to
determine
whether
treatment
caused the
decrease in
behavior.
Days
Frequency of Responses
1 5 10 15 20 25 30 35 40
0
5
10
15
Days
Frequency of Responses
1 5 10 15 20 25 30 35 40
0
5
10
15
17
Methodological Issues and the ABAB Design
• Causal Inference
▪ If behavior does not reverse during 2nd baseline, we cannot conclude
treatment was effective.
▪ Alternative causal inferences
o A variable other than treatment caused behavior to change.
o Treatment may have promoted change, but some other variable
may have taken over to cause behavior change to persist (e.g.,
positive attention).
o Some behaviors may be expected not to change once improved
(e.g., new skills are learned, but not un-learned).
• Ethical issues
▪ Is it ethical to remove a treatment that appears to be beneficial?
▪ Are there times when 2nd baseline (A) should not be implemented?
o violent, self-injurious behaviors?
▪ Ethical dilemma: conflicting goals
o Goal of understanding causes of behavior change
o Goal of creating change and improving human condition
18
Multiple Baseline Designs
• Solve ethical dilemma of the ABAB design
▪ Treatment (B) not removed
• Procedure
▪ Establish several baselines
▪ Implement treatment in one baseline at a time
▪ Compare behavior in treatment and baseline stages
• Establish multiple baselines across
▪ Situations
▪ Behaviors
▪ Individuals
• Interpretation
If treatment is effective
▪ Behavior changes when treatment is implemented.
▪ Behavior does not change in remaining baselines.
▪ Behavior changes only when treatment is implemented.
▪ Behavior changes directly after the introduction of treatment.
19
Illustration of Multiple Baseline Design
A teacher asks your help in planning a behavioral intervention that will help
manage the behavior of a problem child in the classroom. The child does
not stay at her desk, blurts out questions without raising her hand, does not
remain quiet during “quiet time,” and exhibits other behaviors that disrupt
the class. The treatment is to administer a positive reinforcer, such as a
token that can be exchanged for candy or small toys, to reward desirable
behavior in a multiple-baseline across behaviors design.
• 1st behavior: Leaving desk
▪ 1st 2 days: Baseline observations
(count frequency of behavior)
▪ Day 3: Implement treatment
(reward)
▪ Graph displays improvement
(decrease in leaving desk)
• 2nd behavior: Blurts out questions
▪ 1st 5 days: Baseline observations
▪ Day 6: Implement treatment
▪ Graph displays improvement
(decrease in blurting out
questions).
0
2
4
6
8
10
1 3 5 7 9
Frequency
Days
0
2
4
6
8
10
1 3 5 7 9
Frequency
Days
20
Illustration, continued
• 3rd behavior: Not quiet during “quiet
time” (e.g., frequency of
verbalizations)
▪ 1st 7 days: Baseline observations
▪ Day 8: Implement treatment
▪ Graph displays improvement
(decrease in verbalizations)
• Frequently Asked Questions
▪ How many baselines are needed?
o Minimum of 2, recommend 3 or 4
▪ What if behavior changes before treatment?
o Hard to interpret cause for behavior change
o Analyze situation to see if treatment effects generalized
▪ What if treatment generalized to other situations, behaviors, or
individuals?
o Modify research design if generalization is possible
0
2
4
6
8
10
12
1 3 5 7 9
Frequency
Days
21
Problems with All Single-Case Designs
• Baseline records
▪ Increasing or decreasing trends
o Effect of treatment will be hard to interpret
o Trends: interpretation depends on desired direction of treatment
▪ Unstable baseline: Extreme variability in behavior
o Difficult to detect clear discontinuity in behavior following treatment
o Try to control variability or wait for baseline behavior to stabilize
• External Validity
▪ Will treatment effects observed for one individual generalize to other
individuals?
▪ Reasons why treatment effects may not be limited
o Treatments are often powerful.
o Multiple-baseline designs demonstrate generality of treatment
effects.
o Group treatments can be used to demonstrate effectiveness of
treatment across individuals.