Note: a comprehensive review and edit of this document is underway. Reviews of Chapters 1 to 5 are complete. Some serious errors where found and corrected. The review is currently up to here.
Using the test model outlined in chapter 5, an experiment was devised and executed. Subjects were drawn from the student population at Florida Technological University. Most, but not all, of these subjects were fulfilling communication department research requirements. This requirement is imposed on all students taking Speech 101, a course required of all students, regardless of major. The resulting sample population should be considered representative of the entire student body, not just of communication department majors. The experiment can be divided into three parts. The first part was a pre-study involving about 25 upper level communication department students. It was performed in July, 1976. The second part was a control study involving 151 students. Most of these students came from Speech 101 classes. About 20 came from an upper level course in communication. Five subjects from this sample were dropped because of incomplete questionnaires. The last part of the manipulation, the actual experiment, involved 56 volunteer subjects drawn from a second group of about 150 Speech 101 students. One subject was dropped because an incomplete post-test precluded evaluation. Both the second and third segments of the experiment were conducted in October, 1976.
The objective of the pre-study was to determine a message topic for use in the experiment. Several current issues were examined from the perspective of two selection criteria. The first criteria was that opinions on the issue were well mixed. The second criteria was that opinions on the issue be predictable from the Rokeach Value Scales (Rokeach, 1973). Both criteria had to be met in order for the issue to be considered. To test the issues on these criteria, subjects were administered two questionnaires, one a survey of opinions on current issues and the second the Rokeach Value Scales. Divergence of opinion was measured by computing means and standard deviations. Predictability was evaluated through forward (stepwise) multiple regression (Nie et at, 1975) of the Rokeach values on those issues which appeared to meet the criteria of mixed attitudes.
Several topics met both criteria. Since the topic, "God and the Bible are man's attempt to explain the world" most successfully satisfied the criteria, it was used in a first attempt at the experiment. The topic, how ever, proved too salient. A high refusal rate (only I of 16 subjects actually encoded a message) finally forced cancellation of the experiment. The second best topic, "the military budget should be cut to allow greater spending on social welfare programs," was then selected for use in the experiment.
The control sample was collected in order to develop a regression equation for use in predicting initial attitude in the experiment. Subjects were therefore asked to complete the pre-test prediction instrument, the Rokeach Value Scales, and the post-test instruments, a political survey containing the measure of post-test attitude, Brock's (1967) measure of counter-argument production, and a list of arguments used during the experiment. In order to test for possible order effects, that is, persuasion due to the order in which the questionnaires were administered, the questionnaires were distributed in six different orders. The positions of the Rokeach scale, the political survey, the arguments, and an unrelated scale administered during the experiment were manipulated. The Brock measure of counter-argument was always at the end of the questionnaire book let. Its order was also manipulated, however, as subjects were asked to complete the scale either at the very beginning of the survey period or after 5, 10, or 12 minutes.
At the beginning of the survey period subjects were informed that the data was being collected in conjunction with an experiment. The survey booklets (in the six orders) were distributed randomly in each class. If the class was among those asked to complete the Brock measure first, they were instructed to turn to the back of the survey booklet and complete the measure of counter-argument. At the end of three minutes they were instructed to stop and complete the rest of the questionnaire. Other classes were told to begin the questionnaire, but that they would be asked to stop at some point to do something else, After the specified period of time these classes were also given their three minutes to complete the measure of counter-argument.
Analysis of the data involved two methodologies. A composite post-test measure was developed using principal component analysis (Nie et al., 1975) with the component criteria set at I. The two factor matrix was obliquely rotated to obtain the direct oblimin loadings with delta set at zero. Component scores were computed for each subject. A regression equation for use in re building the first component scores from responses to the post-test political survey was computed using forward (stepwise) multiple regression (Nie et al., 1975). A second regression equation was computed for use in predicting first component scores from the Rokeach Value Scales. The effects of the varied orders on attitude were assessed via forward (stepwise) multiple regression.
Before arrival at their appointments, subjects were aware that they would be participating in a series of experiments. This was, to an extent, true, but most of their activity centered around the manipulation. Upon arrival they were given the Rokeach Value Scales and told that the experimenter would return in a few minutes. The value scales (Rokeach, 1973) are two ordinal scales of 18 values. Subjects are asked, on each instrument, to rank the values according to importance. Research involving the scales indicates that they are powerful predictors of personality, demographic factors, and even product preferences. Those interested in specific results should see Rokeach's (1973) own excellent review of the research.
In this experiment the scales were used as a pre test predictor of attitude. A particular strength of the scales are their topic neutrality. It would be difficult for a subject to draw any conclusions about the nature of the experiment from the value scales. This strength is the basis of this experiment, for the use of any pre test which might have indicated what was expected of the subjects would have seriously compromised the results. After completing the value scales, subjects were brought to a second room. It was here that the fundamental manipulation occurred. Subjects were seated in front of a desk on which a tape recorder rested and told the following. "This part of the experiment really isn't an experiment at all, but rather a preparation for an experiment we are planning on doing in a few weeks. In that experiment we are going to try to persuade people to another point of view. Usually, in this kind of experiment they use actors. That's great for those experiments because, as you can imagine, actors, with their training, can be very persuasive. What it doesn't tell us is how persuasive real people in the real world are. That's what we want to find out in this experiment. So what we're asking people to do is record a short persuasive message on a topic. We're asking everybody to record the same message on the same topic. "The topic is that 'military budgets be reduced in order to allow greater spending on social programs. Now I don't know how you feel about this issue and quite frankly, I don't care. However, we can offer you $.5O ($1.5O) to record a message on this topic." At this point the experimenter told subjects in the low reward ($.50) condition that: "I realize that $.50 isn't a lot of money, but that it's more money than you'll have if you don't encode the message." This was done to make the low reward appear salient. The experimenter then reminded the subjects that they didn't have to encode the message and asked if they would be willing to en code a persuasive message for the amount of money offered.
Subjects who declined were brought to a third room where they were administered the post-test. This post- test will be described later. Those who decided to en code the message were given a list of five arguments which could be used in encoding the message. They were then given an opportunity to prepare their message. When they were ready, subjects recorded their message on the tape recorder. The messages were anonymous and the medium (tape) provided low channel visibility. The messages were timed so that any effect of message duration could be analyzed. At the completion of the message subjects were paid and post-tested.
The post-test included four measurement instruments, two of which, collected in conjunction with other studies, were irrelevant to the experiment. The purpose of these irrelevant measures was the reinforcement of the concept that subjects were participating in several experiments. In the third measure, a political survey, only five of the nineteen questions related to the message. Indeed, the criterion measure of attitude used only three of these five questions. The questions were posed in a likert scale format. In the final instrument the measure of counter-argument, subjects were given three minutes to record the possible effects of reducing military spending and increasing social spending.
In a very real sense, the measure of counter-argument was the only measure which could have violated the experiment's internal validity. First, as it involved the topic of the persuasive message, it was possible that the measure could have given the subject an idea of what was being looked for. Second, it could have served as the equivalent of encoding a second persuasive message, with a consequent effect in either inducing or reinforcing persuasion. For this reason, the administration of the instrument was manipulated by giving the instrument either before or after the criterion political survey. Here, it was hypothesized that if the instrument violated the experiment's validity significantly, the effect could be measured in the post-test attitude scores.
In recording perceived 'effects,' as was done in the measure of counter-argument, subjects are not limited to writing counter-arguments. This is the basis of the measure, because subjects, when left free to record any possible 'effects,' whether message counter and consonant arguments, should be expected to write down those arguments which are most salient, which have been recently considered. As such, the measure, when used soon after a persuasive appeal, can be expected to measure those arguments, both counter and consonant, which the subject considered as a result of that message. The instrument should, however, be considered as a measure not only of counter-argument, but of consonant argument. It has been used, in this experiment, to measure both. Being an open ended scale, the argument measure required judging. An upper level communication class was recruited for this purpose. The judges (18 of them) were told that they were engaged in judging the results of an experiment. They were not told anything about the nature of the experiment. Written instructions were handed out. The instructions stated that a consonant argument was one which might be used in supporting the position(s) that the military budget should be reduced and/or that social spending should be increased. counter-arguments were those that might be used in supporting the position that the military budget should be increased and/or social spending should be decreased. The judges were asked to tally the counter, consonant, and neutral arguments on each measure. On the average, each measure was judged independently by four judges. Inter-rater reliability was high, over .90 in every case. Considered as a whole, the procedures conform to the test model depicted in chapter five's Figure 8. Subjects were manipulated on the basis of pre-test attitude, reward (using nth number randomization techniques) and post-test order effects. It was predicted that reward and pre-test attitude would affect persuasion. No prediction was made on the effect of post-test order, and it was hoped that no effect would be found. Message duration was measured to determine its possible effect on persuasion. Message counter and consonant argument were measured as endogenous variables, variables which were both independent in predicting persuasion and dependent in their predictability from the manipulated variables. Finally, post-test attitude was measured. An index of persuasion was constructed by subtracting the pre-test attitude prediction from the post-test attitude.
Analysis of the data involved a variety of techniques. Path analysis (via multiple
regression) was used as the principle test for the model, but an analysis of
variance cell structure was utilized to clarify, through the calculation of
means and standard deviations, the actual effect of the variables on persuasion.
One variable, the interaction of reward and pre-test attitude, was added to
the path model during the course of analysis because an analysis of variance
indicated that the effect of reward and attitude on persuasion was best indicated
in the interaction.
Foulger, Davis A. (2005). Methodology. From the Hypermedia Edition of Foulger, Davis A. (1977). Distraction and Dissonance: A model of the persuasive process. Master's Thesis. University of Central Florida. Retrieved from http://foulger.info/davis/mastersThesis/chapter6.htm.