EXAMPLES OF EXPERIMENTS
‘Manipulation’ means actively varying the states of a variable. There are two methods of manipulation: (a) assigning different states of the variable to different participants and (b) assigning different states to the same participant at different times.
Assignment to different participants is illustrated by an experiment that was designed to test the hypothesis that emotions that compete with anger will inhibit aggression by people who have been provoked (Baron, 1976). The study took place at a busy intersection. Periodically, a car driven by a male confederate (assistant of the experimenter) drove up to the intersection just as the traffic light was turning red and remained stopped for a period of time after the light turned green. If a car driven by a male driver pulled up behind the stopped car, the driver of that car (the participant) was unknowingly assigned at random to one of the four conditions (states of the independent variable) shown in the columns of Table 6.1.
There were three competing-emotion conditions. In the empathy condition, an attractive female confederate on crutches crossed the street in front of the participant’s car. In the amusement condition, the same confederate was dressed as a clown. In the mild sexual arousal condition, she wore skimpy clothes. To determine how these conditions affected aggression, they were compared with a fourth control condition in which no competing emotion was aroused. The female confederate crossed the street in ordinary conservative clothing.
In experiments, the dependent variable(s) are free to vary and are measured to see how they are affected by the independent variable(s). In the experiment just described, aggression was assumed to have occurred if the participant honked his horn. Two measures of aggression were employed: (a)Aqualitative measure: whether the horn was honked or not.
The first row in Table 6.1 shows the proportion of participants who honked their horns in each of the four conditions; (b) A quantitative measure: the latency of honking (the time elapsed between the light turning green and the honk if it occurred). A slower honk was assumed to be less aggressive. The second row shows the mean (average) latency in the four conditions.Comparison of the proportions in the first row shows that there were fewer aggressive participants in the competing-emotion conditions than in the control condition, supporting the hypothesis. Comparison of the means in the second row shows that honking came later in the competing-emotion conditions, again supporting the hypothesis. Both results were statistically significant, which means that the probability of their occurring by chance was less than.05.1
This study, like most experiments, involved replication, that is, the assignment of several participants to each condition. Replication is necessary because participants differ from one another in their circumstances and personal qualities, causing variation in the dependent variable(s) over and above that produced by the manipulation of the independent variable(s). Because of replication, we look at proportions and means rather than the scores of individual participants.
The other type of manipulation, assigning the states of a variable to the same participant at different times, is illustrated by a national opinion survey study in which participants were asked two questions: ‘how long their friendship would be disrupted if a friend of theirs got into a fist fight with them’ and how long it would be disrupted if their friend ‘called them a liar and a coward’ (Nisbett and Cohen, 1996: 31). Southern white Americans reported that the insult (the liar-coward statement) would disrupt their friendship for a considerably longer time than a fist fight; but Midwestern white Americans did not differ in their answers to the two questions.
The authors interpreted these findings, in conjunction with results of other studies, as evidence that there is a culture of honor in the South in which insults elicit an unusually high level of aggression.Table 6.1 Aggression as a function of emotions that compete with anger.
| Experimental Condition | ||||
| Empathy | Amusement | Mild Sexual Arousal | Control | |
| Proportion of drivers honking | 0.57 | 0.50 | 0.47 | 0.89 |
| Latency of honking (in seconds) | 10.73 | 11.94 | 12.16 | 7.99 |
Source. Adapted from Baron (1976)
The reader may have noted that there were two independent variables in the latter experiment: the fist-fight versus insult variable, which was manipulated, and the Southern versus Midwestern variable which was measured by asking respondents where they lived. The dependent variable was the degree of reported disruption in the friendship. Experiments in which there are two or more independent variables are said to employ factorial designs. This one employed a 2 ? 2 factorial design, with two types of annoyance from the friend and two regions of the country. Larger factorial designs are also found - for instance, 2 ? 3, 3 ? 4, and 2 ? 2 ? 2.
Factorial designs allow us to study how independent variables interact in their effect on dependent variables. Two independent variables interact if one of these variables has a different effect depending on the level of the other variable. For example, in the study just described, the impact of the insult variable on the disruption in the relationship was greater for the Southerners than for the Northerners.