drawn cannot be easily generalized to other individu-
als.
Quantitative Methods of Child Study
There are three types of quantitative methods of
study: correlational, experimental, and quasi-
experimental. To some degree each of these designs
allows the researcher to identify relationships be-
tween different factors and to then specify the causes
of these relationships. The initial responsibility of the
researcher is to find the general design that tests the
hypothesis with the maximum amount of clarity.
The correlational design is used to determine
whether relationships exist between two or more vari-
ables. Ultimately the investigator wants to determine
whether a change in one variable coincides with a
change in a second variable. In a correlation design
no variable manipulation occurs. For example, a
child’s behaviors are measured as they naturally
occur, and a numerical index reflecting the relation-
ship between the measures’ outcomes is then comput-
ed. Usually, a correlation coefficient is used to
calculate the strength and type (positive or negative)
of relationship that exists. While the correlational de-
sign is extremely useful, it cannot be used to deter-
mine cause-effect relationships between variables.
The primary reason for this is that variables other
than the ones under study cannot be controlled, mea-
sured, or otherwise considered in the correlational
design; such variables could influence the relation-
ship between the variables under study. This kind of
control is afforded only in designs in which variables
can be manipulated and participants can be randomly
assigned to groups.
An experimental design does allow cause-effect
conclusions since variables can be manipulated and
participants can be randomly assigned. With respect
to manipulation of variables, the researcher must as-
sign independent and dependent variables to the ex-
periment. The independent variables are the various
treatments that the participants receive (and that are
manipulated by the researcher), while the dependent
variable represents the responses of the participants.
For example, an independent variable might repre-
sent the amount of direct reading instruction students
receive, and the dependent variable might be reading
achievement scores. In essence, the researcher wants
the dependent variable to reflect the effect of being
treated with the independent variable. Experimenter
control of the independent variable (e.g., amount,
duration, and type of treatment) partially affords the
necessary confidence to reach cause-effect conclu-
sions. In other words, differences in the dependent
variable may then be attributed to the various treat-
ments the participants received (in the example, dif-
ferences in reading achievement scores may be
attributed to the various amounts of direct reading in-
struction).
In order to say that the treatment has caused
some effect (on the dependent variable), it is impor-
tant that all the traits of the participants be about the
same, especially those that could confound the study.
One way to accomplish such group equality is to ran-
domly assign participants to groups. In essence, when
‘‘chance’’ is the force behind who gets the various
treatments, it is assumed that the groups contain par-
ticipants who were more or less alike prior to receiv-
ing treatments.
The ability to gain the necessary control that al-
lows for cause-effect conclusion is also an important
shortcoming of the experimental design. Conducting
research in a controlled setting may alter the natural
behavior patterns of participants and therefore de-
crease the ‘‘ecological validity’’ of the results. The re-
searcher must also stay within ethical bounds,
meaning that treatments that have adverse physical
or psychological effects on participants cannot be
used. Finally, the requirement of random assignment
may not be possible for ethical or practical reasons.
To counter the latter drawback to the experimen-
tal design, a researcher might turn to the quasi-
experimental design. The quasi-experimental meth-
od permits the researcher to compare groups that
have been manipulated but not randomly assigned.
For example, in the above example of a study of the
effect of direct reading instruction on reading
achievement, suppose that homerooms have already
been assigned in the school where the researcher in-
tends to conduct the study. It may still be possible to
treat different classes, but the researcher must take
into account that the participants were not randomly
assigned to the classes. In such a case, a quasi-
experimental design could be used, but the research-
er must temper any cause-effect conclusions because
of the possibility that uncontrolled variables ‘‘caused’’
the results.
Finally, it may not be possible to manipulate vari-
ables or randomly assign participants to groups. In
this case, a causal comparative design might be used.
In the causal comparative method, already existing
groups are studied (after the ‘‘independent variable’’
has already occurred) and group differences are stud-
ied on some dependent variable of interest. For ex-
ample, a researcher might choose to study the
intellectual development of children in orphanages
compared to that of children raised in a home setting
with their biological parents. The causal comparative
design is often used to study treatments that would be
unethical to impose on participants. Obviously, the270 METHODS OF STUDYING CHILDREN