So far, we have spent our time learning about fundamental concepts in research design such as measurement, reliability, validity, operational definitions, and types of variables. The remainder of the course will focus on design options, recognizing the elements of the designs, and learning about the uses of each.

Single-Case Research Designs 

 Most empirical investigations that evaluate treatment and intervention techniques employ between-group research designs. The difficulty with the between-group design is that it is not always well suited to the many applied situations in which treatment focuses on the individual participant. Many of the requirements of a group design (homogeneous groups of participants, random assignment to groups, standardization of treatment) may not be possible in applied settings. 

Single-case research designs allow experimental investigation with one or more participants. This is not to say that one is ideal, it is just possible. The designs that fall under the category of single-case research designs provide a methodological approach well suited to the investigation of individuals, single groups, or multiple groups of participants. Hence, the designs can be used in group analysis as well, replacing the traditional two group comparison if necessary. In short, understanding single-case designs is a powerful tool as you begin thinking about not only your research question, but what design is the best for answering that research question. 

To introduce single-case designs, we'll start with the most rudimentary design  which is actually not experimental at all and will not be enough for a research design in the course. All experimental single-case designs build from this most rudimentary design so it's important to begin here with the AB design or what your text will call a "time series" design in chapter 7. 

AB design: In the AB design, there is an independent variable that has two levels (A versus B).  In single-case designs, the independent variable is administered in phases. So you will have an A phase and a B phase. The A phase is typically referred to as the baseline phase or the control phase. It represents the control group or the comparison group in a group design. After establishing a baseline phase, the independent variable is manipulated and data are collected during the intervention, or B phase. 

Let's pretend that we want to know if a students "staying in seat" behavior changes when we implement a reinforcement schedule. We would collect data on or measure staying in seat behavior every day for a period of days until we established stable rates of behavior. This is our A phase or our baseline phase. It demonstrates the level of the dependent variable before we intervened. At some point, we intervene and begin the B phase. All the while, we continue to collect data on staying in seat behavior. We end up with several data points from this one individual in the A phase and in the B phase. We can summarize that data by calculating the mean for each phase and then compare the phases. If our intervention worked, the rate of behavior should be less in the B phase than is was in the A phase. 

Why this isn't enough: There are no controls for threats to internal validity in an AB design. We have to find a way to demonstrate that the change we measure in our dependent variable (staying in seat behavior) only and always changes when we choose to manipulate the independent variable. Thus demonstrating the internal validity of the study. That is, when we manipulate the IV, the DV changes.

ABAB designs: Chapter 8 in your text book is devoted to this form of design. One way to demonstrate internal validity is to simply repeat the AB design over and over so that we can show the data jump when we tell it to. These designs are called reversal designs in some cases and withdraw designs in others. These designs can become more complicated as you add variables. For instance, if you wanted to compare more than two levels of an independent variable you would add a C phase into the mix ABACABAC design with a baseline in-between each intervention phase.  You can see how quickly this will get complicated. 

What if you can't reverse or withdraw the intervention you used in your B phase? Think about this for a moment. You collect baseline data. You intervene and the behavior you are measuring changes. Then you go back to A phase and nothing changes. The dependent variable does not go back to it's baseline level. This makes demonstrating internal validity difficult because you can't show that you made the data change repeatedly. There is no reversing it. As an example, what happens when you collect data on self-management skills. In the baseline phase, the student knows nothing. Then you intervene and teach them self-management skills. Now the student has those skills. How will you remove them for the second A phase?

Solution: Use a multiple baseline design. 

Multiple Baseline Designs ABC: Chapter 9 of your text book is devoted to the multiple baseline design. Multiple baseline designs demonstrate internal validity by comparing across the baselines of several behaviors, participants, or settings. You must have multiple behaviors, participants, and/or settings to use these designs.  Baseline data are gathered on two or more behaviors, participants, or settings. The baseline data for each behavior, for instance, represent the current level of performance during the A phase prior to intervening. After performance is stable for all of the behaviors, the intervention is applied to only 1 behavior at a time. Data continue to be gathered for each behavior. If the intervention is effective, we would expect to see changes in the data for the behavior to which the intervention was applied. Meanwhile, we would expect to see no change in the data for those behaviors that are still in the baseline phase. When the first behavior changes and the others remain at their baseline levels, this suggests that the intervention caused the change. To be sure, we repeat the process with behavior #2, and then again with behavior #3, etc. In this way, we demonstrate that when we choose to intervene, we can make the dependent variable change while the other variables remain constant.

Within-Subject Designs

Within-subject designs are those designs that compare individual participants to themselves instead of comparing them to a comparison or control group (between-subject designs). Time series designs, repeated measures designs, reversal designs, withdrawal designs, and in some cases multiple baseline designs are all within-subject designs because they compare individuals to themselves using repeated measurement over time across phases.

Within-subject designs require that all participants participate in all conditions of the study. This makes it possible to have fewer participants in a study but also causes some problems. The conditions must be presented in a specific order to account for order effects. An order effect is what happens when the order of presentation of conditions effects the dependent variable. For instance, in the ABAB designs discussed above, if the first B phase effects the dependent variable by teaching a skill that the participant will not forget for the next A phase, then an order effect has occurred.  To fix this problem we counterbalance the presentation order of conditions so that order effects can be identified. To counterbalance the order of conditions, you run participants through the experiment using every possible order of conditions. Say you have 3 independent variable levels A, B, and C. You would then run participants through one of six different possible orders.

ABC, BAC, CBA, CAB, ACB, BCA,

The need to counterbalance a study, re-introduces the need for more participants to run through each order option.

You can see that with additional independent variable levels, the number of sequences grows exponentially.

The benefits of using repeated measures designs, a) fewer participants and b) the ability to see individual differences, are often outweighed by the trouble counterbalancing can cause. It is often unrealistic to implement counterbalancing in real world applied settings.

Research on Low-Incidence Populations and Investigating Individual Differences

When investigating issues with low-incidence populations it's important to choose a design that is realistic to the numbers of individuals you will have access to. This is not to say that it's okay to use these designs out of convenience, but rather out of necessity if the numbers of individuals who qualify for a particular study are low. If you are interested in studying reading problems in third grade students, a group design using a large sample size is very appropriate and expected because this is a high incidence population. If however, you want to study the effects of using a new teaching strategy on language acquisition in children with autism, a single group design might be appropriate. Single-case designs allow you to assess individual differences which is highly applicable to classroom situations in which individual learning gains are the priority.

When you look at the literature base on low-incidence populations, you will find that much of the research conducted uses single-case designs. One way to contribute to this literature is to put the effort in to recruit enough individuals to conduct a group design using the same or similar research questions that were addressed using single-case designs, or to compute a meta analysis of the research that is available.

Lesson Activities

• Go to the John C. Pace library (on-campus or online), and locate an article using one of the single-case designs outlined above.  
• Write the article reference and a summary including the research question including the dependent and independent variable and the design used.  Post your work in the "Week 6 Lesson 1 " dropbox in the eLearning system.

Need Help?: A sure bet to find some single case design research is to use search terms including "Autism" and social stories, social skills, language acquisition, joint attention... Journals that will be full of single case designs include the Journal of Applied Behavior Analysis, the Journal of Positive Behavioral Support, and Focus on Autism and Related Disabilities. A sure bet that you have found a single subject research design is the use of graphs to analyze data on each individual participant rather than statistics that summarize everyone's information at once. The results section will have graphs with some unit of time on the X-axis and the dependent variables on the Y-axis of the graph. The Independent varialbes are arranged as phases across time and are usually labeled at the top of the graph. As an example, check out Sample Paper #1 and Sample Paper #3. Scroll through and notice the design is indicated in the Method section. Also notice the results sections refer to "figures" which are graphs of the data. If you scroll to the very end of the manuscripts, you will see the graphs I'm referring to.

If you need additional help, seek assistance from the library. Call them, complete the remainder of the tutorials online, or go to the library and ask for assistance in finding one of these journals. You will know you found a study using a single case design, when you see their plotted graphs using an x and y axis with the dependent varialbes on the y axis and time across phases (independent variables) on the x axis.

Need Help?  Have a question but can't find the answer? Here are some options:

• Check the list of frequently-asked questions.
• Email Dr. Barry*.

*Remember to include the course name and your name in your message.

To Page 3

Updated on May 16, 2007