Single-Subject Research Designs
Single Subject Research Designs
Study Aids and Important terms and definitions
Many of you have been inquiring about how to better prepare for the quizzes.
One way is to use the online resources that accompany the textbook.
The book website is located at:
Select the chapter you want to view.
Any item with a symbol of a lock next to it can only be viewed by instructors.
Other, non-locked, items can be viewed by students and for each chapter there is a glossary, flash cards that you can set to view either a word or its definition first, a crossword puzzle, and a practice quiz.
Chapter 14 vocabulary words and terms you should know the definition of include:
single-subject designs or single-case designs
phase
baseline observations
baseline phase
treatment observations
treatment phase
level
trend
stability
phase change
ABAB design or reversal design
multiple-baseline design
multiple-baseline across subjects
multiple-baseline across behaviors
multiple-baseline across situations
dismantling design or component-analysis design
changing-criterion design
alternating-treatments design or discrete-trials design
statistical significance, or statistically significant
practical significance or clinical significance
What is a single subject design?
As the name implies, a single subject design is a research study with only ONE participant.
Even with only a single particpant the design allows us to conduct true experiemnts.
Single subject designs are often used in clincial settings to answer questions such as:
- Does the noise level affect head baning in a child with autism?
- Does therapist self-disclosure increase client self-disclosure?
- Will praising the dog (or child, or spouse...!) when he fetches the newspaper increase fetching behavior?
- Does teacher eye contact (or proximity, or threats of punishment...) decrease disruptive behavior of a problem student?
Evaluating the results of a single-subject design (p. 397)
Since single subject designs have only one particpant there is not much statistical analysis that can be done.
Instead single subject designs are evaluated by observing results on a graph that illustrates changes in behavior observed before and after a treatment is implemented or treatment conditons change.
For instance, the graph below illustrated a high rate of disruptive beahvior during time A, and that disruptive behavior decreases after a child is praised for non-disruptive beahvior (time B).
(image of graph showing decrease in desruptive behavior after a treatment is implemented)
Here is a website that provides a nice overview of single subject designs, including graphs that illustrate different kinds of designs and labels showing hwo to read the graphs
If you think this appears to be a rather crude way to evaluate a reserch outcome, then you are correct!
We can only say, "it looks liek the beahvior decreased significantly" - there is no statistical significance test.
And we can't be 100% certain that praise, rathe rthan some other factor, caused the observed change in disruptive behavior.
Leter in the chapter true experimental single subject designs will be consdiered.
Phases and Phase Changes (p. 398)
A phase is a series of observations of the same individual under the same conditions.
A baseline phase is a series of observations where no treatment is administered
A treatment phase is a series of obserations made when tretment is being administered
by convention, a baseline phase is always labelled "A" and the treatment phase is "B", and if there were a second treamtent it would be "C", a third would be "D" and so on.
A phase change occurs when we switch from a baseline phase ot a treatment phase, from treatment to baeline, or from one treatment to another
THe simplest design is an AB design; a baseline phase followed by a tretment phase
The graph of the baseline phase should include enough data points to show a clear pattern to the behavior.
We can look at the pattern of behavior by three aspects of the graph, the level of the behavior, trend of the behavior, and stability of the beahvior.
- Level - what is the magnitude (frequency, number, or intensity) of the particpants responses
- Trend - is the differene from one observation tot he next consistently in the same direction and magnitide
- Stability - do the observtaions have a consistent pattern with respect to level or trend
Behavior/data must be stable during baseline, or we can't be sure if the tretment or something else is responsible for changes in the the beahvior
At minimum, three data points or observations are needed to determine the level, trend, and stability of the behavior being observed
Dealing with unstable data (p. 401)
When data are unstable you cannot begin the treatment phase. Afew things you might do to deal with the instability are:
- Wait - maybe it will be come stable after the person or animal being observed gets used to you watching them
- Average two or more observations - thsi may seem like 'cheating' however, what looks unstable over a period of one observation per hour might look more stable over a period of a few hours or a day. For instance, if we looked at eating behavior it might look unstable if we measured it every two hours, but over a day it would be more stable at 3 or 4 times per day.
- Look for patterns within the instability - for instance maybe the dogs only misbehave at 8:00 am and 5:30 pm because that is when people are coming and going to work and school.
changing phases
There are three types of phase changes
- baseline to tretament
- treatment to baseline
- one treatment to a second tretament
Generally a phase change can be implemented once the behavior of interest shows a stable trend.
Also, it is appropriate to change from intervention to baseline if the treatment isn't working or is having an unexpected detrimental effect on behavior!
Visual inspection techniques
As was mentioned above, data in a single-subject design are analyzed by visual inspection of graphed resutls rather than through statistical procedures.
Four things to look for in determining whether the phase change produces a meaningful change in behavior are:
- A change in the average level of the behavior between phases, that is it's averag elevel increases or decreases
- An immediate chang ein level, for instance a decrease or oncrease in behavior as soon as the treatment is stopped
- A change in trend, such as the behavior gradually increasing when it had been level or becoming level when it had been decreasing
- A delayed change in level or trend - it could be the case that the treatment takes time to have an effect, and if the change in observed behavior is delayed after the phase change this is BAD NEWS as it means you cannot be sure etha tthe change is due to the treamtment or withdrawal of treatment.
Your textbook provides some nice examples and illustrations of each of the four things to look for when visually inspecting a graph.
The ABAB Reversal Design - a true experiment with a single particpant!
An AB design has the big disadvantage of not allowing us to 'prove' that the treatment caused the observed change in behavior.
However, an ABAB reversal design does allow us to more conclusively demonstrate that observed changes in beahvior are indeed caused by the intervention.
In an ABA reversal design first baseline observations are made (A), then there is a phase change and treatment is introduced (B), then there is a second phase change and treatment is withdrawn with a return to baseline (B) - the reversal, and there is a second phase change and treatment is reintroduced (A).
If the graph looks like the one below, where there is a stable level and trend at baseline, followed by a noticeable change in level with the phase change, and then the behavior returns to baseline with the reversal, and chnages again with the second introduction of treatment, then this provides convincing evidence tha tthe observed change in beahvoior is caused by the intervention
(image of graph illustrating ideal ABAB outcome iwth distinct change in behavior observed with each phase change.
Limitations of the ABA design
Like all treatment designs the ABAB design has its limitations.
If tretment ahs a long-lasting effect there may not be a change in behavior even after it is withdrwn from B to A. For instance, after therapy for depression of anxiety ends most therapy clients continue to have lower levels of depressiona nd anxiety.
Sometimes it is unethical to withdraw tretment. If a person with distressing psychological problems or harmful behavioral problems improves with treatment it is hard to justify withdrawing treatment.
However, in many cases the ABA is a useful way to demonstrate a cuase and effect relationship between a treatmetn and corresponding change in beahvior.
Fancy Designs: going beyond the ABAB design by creating more complex designs
The AB and ABAB are fairly simple designs, there are variations on them that are somewhat more complex.
For instance, there may be two treatments added one afte rthe other, an ABC design in the simplest case, where A is the baseline phase and B & C are treatments.
The more complex variations are ABCAC, to show that change is caused by C, or
A B C BC A BC to show that the combination of B and C is mroe effective than either treatment alone.
Multiple baseline designs
Multiple baseline desings have two or mroe baselines that begin at the same time across two or more particpants, or behaviors, or situations.
The baselines begin at the same time and do not end at the same time.
For example...
In multiple baseline across subjects design, two baselines begin at the same time for each particpant, we can call them A1 and A2.
Then treatment B1 is begun for one particpant before it is introduced to the second participant.
There are 'multiple baselines' in that there is one baseline for each particpant.
If we can show that each particpants beahvior changes after treatment is introduced and NO MATTER WHEN THE BASELINE OCCURS, then we can be confident that observed changes in behavior are due to tretment and not to some random event...if it were due to a random event then both articpants behavior would change at the same time, and not only after treatment is introduced at different times.
In a multiple baseline design there is no need for a 'reversal' as in the ABAB design.
Below is an image of a sample chart for a multiple baseline design
Note that the baseline for each particpant is different, adn that each particpant's behavior changes as soon as treatment is introduced to them.
(image of graph for multiple baseline across pariticpants deisgn)
Other multiple baseline designs are across behaviors and across situations
Multiple-baseline across behaviors:
In this design there is one particpant and two or more behaviors that may be targeted. For instance, a child where praise is used to change 'out of seat beahvior' after a first baseline behavior and then we later treat 'yelling' in addition to out of seat behavior.
Multiple Baseline across situations
In this design we have one particpant and might work on changing a problem behavior tha toccurs at home and at school. For instance, we might collect baseline measurements for self-injury (a problem for many children with developmental disabilities), and introduce a treatment at home and then later introduce the treatment at school and see if there is a change in behavior in each. settings after treatmetn is introduced in that setting.
Note that the criteria for a multiple baseline design are the same as for an ABAB design, that is, beavhir changes with a phase change from baseline to treatment, and there are two demonstrations of this. In the ABAB design the two demonstartions are the first and second time the phase changes from baseline to treatment and in the multiple-baseline design it is after the first participants/behavior/setting and then again for the second particpant/behavior/setting.
Many researchers and clinicians who want to publish their studies use a multiple baseline design across subjects becaus eit is often more convincing to journal editors to have more than one particpant.
Other uses of a single subject designs (p. 419)
Dismantling and component analysis
Dismantling studies are used to study therapy with multiple components (say the 'cognitive' and 'behavioral parts of congitive beahvioral therapy.
At each phase change one component of treatment si added or subtracted and the aim of the design is to show which components of tretament are most effective or essential.
The changing criterion design
Your textbook uses the excellent example of smoking cessation to illustrate a changing criterion design.
In this example the 'criterion' is number of cigarettes smoked with the aim of reducing it each week until the person quits.
There is only one tretament, and the criterion for success changes at various intervals of time.
This design is useful for beahviro changes that may be gradual, such as decreasing cigarettes smoked, or increasing minutes or miles of running in an exercise program, or pages written per day in treating wirter's block.
The alternating treatments design
in this design two or more treatments are administered without a baseline in between, for instance ABCBCBC.
Often these designs have variable intervals between tretments and the aim is to show that one treatment is mroe effective or that they are both equally effective. For instnace,. if we wanted to know if praise or eye contact was more helpful in reducing disruptive beahvior we could swith from one to the other every hour or day and see which ahs a greater impact on behavior.
This design can also be treatment Vs. no treament, an ABABABABAB design...
Summary: advantages and disadvantages of single subject designs
strengths
- you only need one particpant!
- very flexible, it can be modified mroe easily than a desingwith many aprticpants
- frequent assessment - the particpant is observed regularly instead of only once or twice.
- the researcher can show practical/clincial significance even if they can't show statistical significance
Weaknesses
- Notice that the weakneeses are almost the same as the strengths!
- there is only one particpant - yes, even while an advantage, it leads to ppor external validity (using a multiple baseline design across subjects can help)
- continuous observation is needed, so the researcher has to keep an eye on the particpant
- can't show statistical significance (and can show clincial significance, which some regard as mor eimportant - see chapter 15 for more on practical/clinical significance)