Correlational research is carried out for one of two basic purposes—either
to help explain important human behaviors or to predict likely outcomes.
Explanatory Studies
A major purpose of correlational
research is to clarify our understanding of important phenomena by identifying relationships
among variables. Particularly in developmental psychology, where experimental
studies are especially difficult to design, much has been learned by analyzing
relationships among several variables. For example, correlations found between
variables such as complexity of parent speech and rate of language acquisition have
taught researchers much about how language is acquired. Similarly, the
discovery that—among variables related to reading skill—auditory memory shows a
substantial correlation with reading ability has expanded our understanding of
the complex phenomenon of reading. The current belief that smoking causes lung
cancer, although based in part on experimental studies of animals, rests
heavily on correlational evidence of the relationship between frequency of
smoking and incidence of lung cancer.
Researchers who conduct explanatory
studies often investigate a number of variables they believe are related to a
more complex variable, such as motivation or learning. Variables found not to
be related or only slightly related (i.e., when correlations below .20 are obtained)
are then dropped from further consideration, while those found to be more
highly related (i.e., when correlations beyond 0.40 or 0.40 are obtained) often
serve as the focus of additional research, using an experimental design, to see
whether the relationships are indeed causal.
Let us say a bit more here about
causation. Although the discovery of a correlational relationship does not establish
a causal connection, most researchers who engage in correlational research are
probably trying to gain some idea about cause and effect. A researcher who carried
out the fictitious study whose results are illustrated in Figure 15.2, for
example, would probably be inclined to conclude that a teacher’s expectation of
failure is a partial (or at least a contributing) cause of the amount of
disruptive behavior his or her students display in class.
It must be stressed, however, that
correlational studies do not, in and of themselves, establish cause and effect.
In the previous example, one could just as well argue that the amount of
disruptive behavior in a class causes a teacher’s expectation of failure, or
that both teacher expectation and disruptive behavior are caused by some third
factor—such as the ability level of the class.
The possibility of causation is
strengthened, however, if a time lapse occurs between measurements of the variables
being studied. If the teacher’s expectations of failure were measured before
assigning students to classes, for example, it would seem unreasonable to
assume that class behavior (or, likewise, the ability level of the class) would
cause the teacher’s failure expectations.
The reverse, in fact, would make
more sense. Certain other causal explanations, however, remain persuasive, such
as the socioeconomic level of the students involved. Teachers might have higher
expectations of failure for economically poor students. Such students also
might exhibit a greater amount of disruptive behavior in class regardless of
their teacher’s expectations.
The search for cause and effect in
correlational studies, therefore, is fraught with difficulty. Nonetheless, it
can be a fruitful step in the search for causes. We return to this matter later
in the discussion of threats to internal validity in correlational research.
Reference
Fraenkel, J. R. & Wallen, N. E.
(2009). How to design and evaluate research in education. McGraw Hill Inc.
Prediction Studies
A second purpose of correlational
research is prediction:
If a relationship of sufficient
magnitude exists between two variables, it becomes possible to predict a score
on one variable if a score on the other variable is known. Researchers have
found, for example, that high school grades are highly related to college
grades. Hence, high school grades can be used to predict college grades. We
would predict that a person with a high GPA in high school would be likely to
have a high GPA in college. The variable that is used to make the prediction is
called the predictor variable; the variable about which the prediction
is made is called the criterion variable. Hence, in the above example,
high school grades would be the predictor variable, and college grades would be
the criterion variable. As, prediction studies are also used to
determine the predictive validity of measuring instruments.
Using Scatterplots to Predict a
Score. Prediction can be illustrated through the use of scatterplots.
Suppose, for example, that we
obtain the data shown in Table 15.2 from a sample of 12 classes. Using these data,
we find a correlation of .71 between the variables teacher expectation of
failure and amount of disruptive behavior. Plotting the data in Table 15.2
produces the scatterplot shown in Figure 15.2. Once a scatterplot such as this
has been constructed, a straight line, known as a regression line, can
be calculated mathematically. The calculation of this line is beyond the scope
of this text, but a general understanding of its use can be obtained by looking
at Figure 15.2. The regression line comes the closest to all of the scores
depicted on the scatterplot of any straight line that could be drawn. A
researcher can then use the line as a basis for prediction. Thus, as you can
see, a teacher with a score of 10 on expectation of failure would be predicted
to have a class with a score of 9 on amount of disruptive behavior, and a
teacher with an expectation score of 6 would be predicted to have a class with
a disruptive behavior score of 6. Similarly, a second regression line can be
drawn to predict a score on teacher expectation of failure if we know his or
her class’s score on amount of disruptive behavior.
Being able to predict a score for
an individual (or group) on one variable based on the individual’s (or group’s)
score on another variable is extremely useful. A school administrator, for
example, could use Figure 15.2 (if it were based on real data) to (1) identify
and select teachers who are likely to have less disruptive classes; (2) provide
training to those teachers who are predicted to have a large amount of
disruptive behavior in their classes; or (3) plan for additional assistance for
such teachers. Both the teachers and students involved would benefit
accordingly.
Reference
Fraenkel, J. R. & Wallen, N. E. (2009). How to design
and evaluate research in education. McGraw Hill Inc.
