Information Technology Reference
In-Depth Information
obviously not be able to observe the “pure” effects of cognitive-behavioral
therapy under these circumstances - what we would see is the effects
of cognitive-behavioral therapy after clients have experienced extensive
psychoanalysis. Having received one type of therapy would presumably
carry over to the next research condition. The adverse consequence of
this carry-over would be to confound our measurements of cognitive-
behavioral therapy success, rendering this particular design untenable.
This potential problem of carry-over effects is something that re-
searchers considering a within-subjects study must directly confront.
If experiencing one level of the independent variable - the repeated
measure - will adversely carry over to the experience of another level of the
independent variable, then a within-subjects study should not be run. For
example, if we were studying the effectiveness of applying a memorization
strategy in trying to remember sequences of numbers (digit span), having
taught participants to group the numbers into sets of twos would change
their performance if we wished to test them when they did not use the
grouping strategy. On the other hand, if we tested them first without any
instruction of strategy, we might be able to use that as a baseline in testing
their performance with the grouping strategy being used. Even then, we
would want to have a control group given a second set of trials without a
strategy in place because having had some “practice” with the digit span
test might affect subsequent performance.
The issue of carry-over is usually less of a concern when we work with
time-related within-subjects variables because it is precisely the carry-
over effects that we intend to measure when we design such a study.
In a simple pretest-posttest design, for example, the treatment effect is
carried across the time between the measurements. As long as our actual
pretest measurement did not produce the changes in the values of the
dependent measure that we see in the posttest, and as long as some other
event occurring at the same time as the treatment or any natural changes
in the participants over the time period did not cause the changes, and
assuming that a control condition cannot be measured, then a pretest-
posttest design may be better than not running the study at all.
A within-subjects variable that is unrelated to time is more suscep-
tible to adverse carry-over effects. Because there is no natural temporal
sequence in which the levels of the independent variable must be experi-
enced, the sequence is usually under the control of the researchers. Here
it is vital that the researchers are able to reasonably eliminate carry-over
effects as potential causes of differences between the means of the inde-
pendent variable. In the music identification study described earlier, for
example, it can be argued that if the students received fifty successive trials
in identifying rock music before they received their fifty trials for classical
music, a number of confounds would present themselves (e.g., they are
well trained by the time they are exposed to classical music; they may
recognize that the trials are “blocked,” assume that the next trial will be
the same type of music, and respond on the basis of their guess of music
