Information Technology Reference
In-Depth Information
variable in terms of time. For example, the terms
pretest
and
posttest
have
been chosen because the temporal relationship of the measurement has
been conveniently built into the terminology.
A pretest-posttest design can have more than two levels. For example,
it might be important to have two or more pretests to establish a stable
performance baseline prior to the administration of some treatment con-
dition. It is also possible that researchers will want to have two or more
posttests, perhaps to establish the degree to which the treatment effect will
diminish (or exhibit itself) over time or perhaps to determine how long it
takes for the treatment to maximize its effect.
10.3.2 VARIABLES UNRELATED TO TIME
A within-subjects variable does not have to be time related to measure
participants under all of the research conditions. To illustrate this, consider
a simple hypothetical study in which we ask students to classify briefly
presented selections of music as quickly as they can. We use two types of
music: rock and classical. Students are asked to press one key if they think
that the music is rock and another key if they think that the selection
is classical. We measure reaction time as well as whether the student is
correct on each trial. We administer 100 trials to the students, 50 for
each music condition, such that the rock and classical music selections
are presented to each student in a different random order. At the end
of this data collection procedure in this simplified example, we average
the reaction times for rock and classical music for those trials where the
students answered correctly; thus, reaction time is our dependent variable.
This gives us two reaction time scores for each student, one for rock music
and one for classical music. These measurements represent a within-
subjects independent variable whose two levels are rock music and classical
music.
10.4 THE ISSUE OF CARRY-OVER EFFECTS
A carry-over effect is the change in a later measure resulting from exposure
toapreviouscondition.Thisisacommonoccurrenceinoureveryday
lives. Having learned to roller skate, we can more efficiently learn how to
ice skate; having taken our first statistics course, we are better prepared
for the next statistics course. In a research study, having experienced a
particular treatment condition may result in differences in performance
in some subsequent task compared to those not having experienced it.
These are all carry-over effects and, generally speaking, they are expected
and desirable.
Some carry-over effects in the context of research design are not desir-
able. As an extreme example, if we wished to evaluate the efficacy of com-
paring cognitive-behavioral therapy to psychoanalysis for a set of clients,
we would not first provide seven years of psychoanalysis to them and then
give them some cognitive-behavioral therapy (or the reverse). We would
