Hovland, C. I., Reconciling Conflicting Results Derived from Experimental
and Survey Studies of Attitude Change, American Psycholgist, 1959, 14, 8-17.
Attitude change through communication is measured by experiments
and sample surveys. The panel method is also used where the same subjects
are repeatedly interviewed. Studying the literature shows that these two
methods often generate conflicting results.
Divergence
Most research on the effects of mass communication on attitude change has
shown little correlation. Yet experimental research has shown large effects
(1/3 to 1/2 of subjects changed). One reason for the difference is that
in surveys the people most likely to actually seek out and expose themselves
to mass communication are ones that are already in favor of the opinion
already. People usually seek out that communication outlet with which they
usually agree with. Thus the people who claimed they were exposed were
a biased group to begin with. In experiments everyone (opposed and in favor)
are exposed.
Also surveys usually evaluate many communication methods at once (TV and
Radio and newspapers, for instance), where experiments usually focus only
on one. The time span for surveys are usually much longer than experiments.
The method of communication is also very different. Often communication
in an experiment may be in a classroom (vs the real media). Communicator
credibility may be very different. Also in natural settings where people
hear and discuss information in group mode, normative influences may cause
people not take a counter-norm view. Experiments are often with students
(due to accessbility) and surveys with random samples.
Finally in experiments we often look for variables in which we can get a
big effect, while in surveys we often look for more socially important (and
thus more complex and difficult for attitude change).
Another issue is that the amount of attitude change is also based on the
initial distance between the subjects initial position and the position
of the communicator. In experiments attitude change is usually proportional
to the initial opinion distance, while in surveys often the communicator
is discredited instead when the distance is too great.
People also tend recall a communicator's info quite accurately if they both
have the same opinions. If they are a little different, people usually
see the communicator's opinion as being more similar to their own than it
really is, and more extreme than their view than it really is when their
opinions are farther apart.
In surveys they found that the greater the opinion difference the more resistance
to change (when there is ambiguity about the credibility of the communicator
and people have a strong opinion). But with highly respected communicators,
the greater the difference the greater the effect.
Thus in most experiments people test rather innocuous issues in which opinions
are easy to influence by experts. In surveys the subjects are usually more
basic and people already have deep-seated opinions.
Also there is an effect of order of presentation. People tend to adhere
to the first information they receive, and then feel they don't need more
information on the topic. In experiments people get exposed to both sides
of an issue.
Integration
It seems that most of the discrepancy between survey and experimental data
is due to methodological differences. In reality both methods are useful.
While the survey method has correlational difficulties, it probably best
simulates the real situation. The experiment allows better control over
the variables, but is less generalizable. Of course, one could develop
one's experiments to be more realisitic.