SINGLE SUBJECT DESIGN
John B Wasson
You might ask yourself if you should consider the use of colored overlays or colored transparencies to assist a student with a severe reading disability to perceive text more effectively. In a previous article (Wasson, 2003) we looked at the possible effectiveness of colored transparencies for improving the perception of text by disabled readers. These results do not help us very much in deciding if the use of colored transparencies will be effective with a particular pupil with a severe reading disability. We need to use an experimental strategy with our pupil in order to answer this question.
The strategy we need to use is called single subject research. The single subject research design strategy is the most scientifically valid method of making the appropriate decision for our student. Let us look then at single subject research designs so that we can ascertain what specific design we should use to answer our question. In the process we will also gain a better understanding of this valuable teacher tool, single subject research designs.
Single Subject Research Designs (also referred to as single-case experimental designs) are designs that can be applied when the sample size is one or when a number of individuals are considered as one group. These designs are typically used to study the behavioral change an individual exhibits as a result of some treatment. In single-subject designs, each participant serves as her or his own control, similar to a time-series design. Basically, the participant is exposed to a non-treatment and a treatment phase and performance is measured during each phase.² (Gay & Airasian, 2003, page 383)
James H. McMillan (2004, pages 227-228) has summarized five characteristics of single-subject research.
1. Reliable measurement: Since these designs involve multiple measures of behavior, it is important for the instrumentation to be reliable. Conditions for data collection, such as time of day and location, should be standardized, and observers need to be trained. Consistency in measurement is especially crucial in the transition before and after the treatment.
2. Repeated measurement: The same behavior is measured over and over again. This step is different from most experiments, in which the dependent variable is measured only once. Repeated measures are needed to obtain a clear pattern or consistency in the behavior over time. They control for the normal variation of behavior that is expected within short time intervals. This aspect of single-subject designs is similar to time series studies, which investigate groups rather than individuals and do not provide for a return to conditions that were present before the treatment was implemented.
3. Description of conditions: A clear, detailed description of the conditions of measurement and the nature of the treatment is needed to strengthen internal and external validity.
4. Baseline and treatment conditions: Each single-subject study involves at least one baseline and one treatment condition. The baseline refers to a period of time in which the target behavior (dependent variable) is observed and recorded as it occurs without a special or new intervention. The baseline behavior provides the frame of reference against which future behavior is compared. The term baseline can also refer to a period of time following a treatment in which conditions match what was present in the original baseline. The treatment condition is a period of time during which the experimental manipulation is introduced and the target behavior continues to be observed and recorded. Both the baseline and treatment phases of the study need to be long enough to achieve stability in the target behavior.
5. Single-variable rule: During a single-subject study, only one variable should be changed from baseline to treatment conditions. In some studies two variables are changed together during the same treatment condition. This is an interaction in single-subject research.
There are three commonly accepted types of single subject research designs.
1. A-B-A Withdrawal Designs
2. Multiple-Baseline Designs
3. Alternating treatments Design
Let us look at each of these designs types and see how each might be used to answer our question about the use of colored overlays. We should then be able to select the best strategy to use with our pupil.
A-B-A Withdrawal Designs
In the A-B-A withdrawal family of single subject design strategies, A refers to the non-treatment or control phase of the experiment while B refers to the treatment phase of the experiment.
The simplest variant of A-B-A withdrawal designs is the A-B design. In this design type a non-treatment phase is initiated until the behavior in question demonstrates stability. Once the behavior becomes stable, the treatment phase is initiated. The behavior in question, the dependent variable in the experiment, is measured during both phases and the results for the two phases of the experiment are compared.
How could we set our experiment up to find the effectiveness of colored overlays for a reading disabled student using the AB research design? Our first step would be to state our research problem in a manner that includes operational definitions of our experimental variables. We might state our research problem as for a 12 year old student name Billy, as follows: The effect of the use of self-selected colored transparencies on the number of words read correctly in a one minute time period by Billy.
We then do a match to sample with Billy to find which color of transparency he likes the best. That is which of the colors makes reading the easiest for him. We can do this as follows: Prepare two letter sized sheets of white paper with the sentence, A this is the letter A. on one and B this is the letter B on the other sheet. You are going to place colored transparencies over each of the sheets, two at a time and ask the student which one he or she can read best or which one he or she prefers. With the following six different transparencies, yellow, green, red, pink, blue, and clear, we could proceed as follows. Place the yellow and green transparencies on the white sheets before the student and say, ³Which one can you read best A or B? Which one can you read the easiest? Which one do you like the best?² Set the winning color aside (for example yellow) and then proceed in the same manner with the red and pink transparencies. Letıs say that the student preferred pink for our example. Then place the blue and clear transparencies on the sheets and again ask the student which one he prefers. Letıs say the winner was blue. We have now identified three potential preferred transparencies to use with our student, yellow, pink, and blue. So next cover the white sheets with the yellow and pink transparencies and ask the student which one he or she prefers. Letıs say our student prefers pink. Now for our final matching present the sheets covered with the pink and blue transparencies. Whichever color the student selects will be the preferred transparency color for that student. Colored transparencies to use with students in cases such as this can be purchased at a school or office supply business. They can also be purchased on the internet. For example, colored transparencies can be ordered at http://www.howtolearn.com/filters.html
Then we have Billy do an oral reading sample for one minute without the colored overlay on his reading selection (or use the clear transparency) until he has established a stable baseline rate (3-5 sessions). We count the number of words Billy reads correctly during the one minute daily timings. Then we use the self selected colored overlay (pink for our example) for Billyıs daily reading assignments for the same number of sessions as the baseline or control condition. In the spreadsheet below we have recorded the session number, the condition (A = baseline, B = colored overlay), and the number of words read correctly in a one period of time. The results are also recorded in a behavioral chart.
The results of a single subject experiment are typically interpreted by referring to the behavioral chart in which the data were graphed. We can see from Billyıs chart that his performance did increase with the use of the colored transparency. He read more words correctly while using the colored overlay than he did under the baseline conditions. Statistics are not generally used to interpret the results of single subject experiments, but we could find the average number of words read under baseline conditions (50.2) and compare that with the average number of words read correctly while using a colored transparency (56.6). It looks like on the average that Billy read 6.4 more words per minute under experimental conditions, an increase of 11%, which would certainly seem like a significant increase and would justify the use of pink transparencies for Billy to use in his reading.
In the ABA design strategy we add another control phase after the end of the experimental phase (using the colored overlays) of the project. So in Billyıs case we would do 5 sessions without the colored overlays, 5 sessions with the colored overlays, and then 5 sessions without the colored overlays. The results for Billy are show in the chart below.
As with the AB design, the ABA design suggests an improvement in the number of words read correctly as an adjunct to using the colored transparency. We can also see that the number of words read correctly is under experimental control. In other words there was not a sustained increase in the number of words read correctly after the use of the colored transparency was discontinued.
In some cases the investigator or teacher conducting a single subject research design, is reticent to conclude the experiment in the control condition. This is particularly true if the experiment results in an improvement of the target behavior. This would constitute a problem with the ABA design. To overcome this the investigator introduces an additional treatment phase at the end of the ABA design, which makes it an ABAB design, that is a control condition is followed by an experimental condition, which is followed by an additional control condition and experimental condition. If we subjected our subject Billy to this design we might get the following results.
As we might have expected if the dependent variable is under the control of the experimental condition, our use of the ABAB design shows a significant increase in words read correctly in a one minute time period with the use of the colored transparency. In both baseline conditions, during which the colored transparency was not used the number of words read correctly decreased below the level of the experimental condition.
In the typical single-subject designs we have considered a single subject, one behavior, and a single setting. In a multiple-baseline design, we systematically vary one of the three parameters (subject, behavior, or setting) while keeping the other two parameters constant. In the designs we have considered so far we have kept all three parameters constant. That is we have considered a single behavior for a single subject in a single setting. What if we looked at the number of words read correctly in one minute by three subjects while using their favorite color of transparency. In this design, we do a baseline line condition (that is reading words without using a colored filter) for all three subjects. After 3-5 trials, one of the subjects is subjected to the experimental condition (using a colored overlay) while the other two subjects continue with the baseline condition. After 3-5 more trials, the first subject is returned to the baseline condition while a second student goes into the experimental phase. The third student continues with the baseline condition. After 3-5 more days, the third student starts the experimental phase, while the other two students, return to or continue in the baseline or control situation.
Mrs. Smith is working with three boys with reading problems and as part of her reading program already does daily, one-minute reading sessions on all her students who are attempting to improve their reading skills. Here is a spreadsheet showing the number of words read correctly and experimental condition for a total of 12 sessions, conducted over 16 school days. Each cell shows the studentıs condition (A = baseline or control, B = experimental). This is followed by the words read correctly in a one minute period for each student. These data are followed by a behavioral chart of the data.
We can see that Billy, showed a slight improvement in number of words read correctly while using the colored overlay. Robert showed a dramatic improvement in number of words read correctly with the colored overlay and probably should continue to use colored overlays in reading. Archie on the other hand showed no improvement as a result of using colored overlays while reading. Archie also showed greater day by day fluctuation in the number of words read correctly. It is doubtful that the use of colored overlays would be at all useful with Archie.
In this multiple baseline study, we have looked at the performance of several subjects on the same behavior (number of words read correctly) and the same setting (in school during the reading instruction period). We could also do a multiple baseline study by looking at several behaviors for one subject in one setting. The third variant of a multiple baseline study could involve a single subject with a single behavior in several settings. For example we could take one minute samples for Billy reading words correctly in his regular classroom, in the resource room, and at home
Alternating Treatments Design
The final type of single research designs we will consider is the alternating treatments design. In this design we wish to compare the effect of two treatments on a subject. The subject in an alternating treatments design is given one of two treatments at each experimental session. Which treatment to use during a given session is determined randomly. This could be done by flipping a coin after assigning treatment 1 to heads and treatment 2 to tails. We could use this design with our colored transparencies in reading by considering the use of a clear transparency as treatment 1 and the use of a preferred colored transparency as treatment 2.
If we evaluated Jimmyıs performance in reading words correctly under the two treatment conditions over a 10 session period we might get results such as the following.
The chart for Jimmyıs data shows the number of word read correctly during a one minute time period while using a clear transparency (T1) and the number of words read correctly while using a colored transparency (T2). For Jimmy we do not see an advantage for his reading with a colored transparency. In fact during sessions 6 through 10 he seemed to be performing slightly better with Treatment 1, the use of clear transparencies. We could make a couple of other observations. 1) There is little overall difference in Jimmy reading behavior as a function of treatment type. 2) there does seem to be a general increase in the number of words read correctly over time, but this is not related the treatment choice.
In all of these studies we have used as the dependent variable the number of words read correctly during a one minute time period. It would do well for us to look at this variable in more detail, which we will do in the next issue of Practical Update as we discuss How to Improve Word Recognition Skills.
Gay, L. R., and Airasian, P. (2003). Educational Research: Competencies for Analysis and Applications. Merrill Prentice Hall: Columbus, OH.
McMillan, J. H. (2004). Educational Research: Fundamentals for the Consumer, 4th Edition. Allyn and Bacon: Boston.
Wasson, J. B. (2003). Using colored overlays with pupils with severe reading disabilities. Practical Update 7(3), 3-6.