Laboratoryvs. field EXPERIMENTS

Laboratory experiments take place in artificial settings constructed by the investigator, while field experiments (see Pruitt, 2005a) take place in naturally occurring settings. Cohen's study of insult and dominance behavior is an example of a laboratory experiment, while Baron's horn-honking study is an example of a field experiment.

Other types of laboratory experiments

Some laboratory experiments involve sim­ulations, in which the participant is asked to play a role that exists in real life. For example, the author and one of his students (Ben Yoav and Pruitt, 1984) performed a study in which male undergraduates were asked to play the roles of buyer and seller in a wholesale market, setting the prices of three appliances. Each party knew the profit to his firm associated with each of the possible prices, and some combinations of prices were worth far more than others to the two parties collectively. Two variables were manipulated in a 2 x 2 factorial design: (a) accountability to the owner of one's company, a role played by another participant. At the end of the experiment, under high accountability, the owner would write an evaluation of the negotiator and decide how to divide a sum of money between himself and the negotiator. Under low accountability, neither of these events would take place; (b) expectation of cooperative future interaction (ECFI) with the opposing negotiator. Participants were told there would be a second study after this one was finished. In the high ECFI condition, they were told that the two negotiators were going to work together on a common goal.

The main dependent variable was joint benefit - the amount of money made by the two parties collectively. Joint benefit was con­siderably higher in the high accountability- high ECFI condition than in the other conditions. This may have been because the participants in that condition were under the cross-pressures of trying to serve the interests of the owner and maintaining the good will of the other negotiator. If so, it should be possible to generalize this result to other situations where negotiators are under such cross-pressures.

The experiment just described involved active simulation, in which participants were required to take actions in response to conditions they were facing. Much more elaborate active simulations are sometimes used, involving many more features of a real situation, for example, simulations of international situations (Wilkenfeld, 2005). Other researchers use passive simulation, where participants fill out a questionnaire as if they were in a particular role. For example, Rothbart and Hallmark (1988) had participants read about a conflict between two hypothetical countries, their own and a rival. They then rated a series of tactics as to how effective they would be if used against the rival and against their own country. Most participants thought that coercive tactics would be most effective when used against the rival and conciliatory tactics would be most effective when used against their own country.

Whether elaborate or uncomplicated, simu­lations always simplify reality by eliminating many features of the settings they are trying to portray. Even more simplified are the experimental games used by some experi­menters, in which participants are given a set of choices and a list of outcomes that depend on the decisions they and the other participant(s) make. The outcomes are usually monetary and, in research by experimental economists, may be quite sizable (Croson, 2005), so as to focus the participants' attention on the incentives and away from ‘extraneous' issues.

Some of the most interesting experimental games involve resource dilemmas where individuals in a group must make decisions that affect their own and the group's welfare. The dilemma inheres in the fact that decisions that help oneself tend to hurt the group as a whole, which means that one's own welfare may eventually be hurt. Onevariety of resource dilemma is the commons dilemma in which the group members withdraw resources from a common pool, which is slowly replenished. If people take too much, the pool disappears and everyone is hurt. In a study of this kind, where the resources withdrawn from a computer-administered pool were converted into money, Allison and Messick (1985) found that larger groups withdrew resources at a faster rate than smaller groups and ended up with less money. Another variety of social dilemma is the public goods dilemma, where individuals must decide how much to contribute to a common pool of resources that benefits all of them. In these situations, there is a temptation to become a ‘free rider' and retain one's own resources while letting others contribute to the common pool. But if enough people do so, the pool disappears, hurting everyone. Again, members of larger groups contribute less than members of smaller groups, resulting in smaller long-run benefits (Yamagishi, 1992).

Other types of field experiments

Examples have already been given of two types of field experiments: elaborate hoaxes in natural settings (like the horn-honking study) and parallel questions in surveys (like the study of how long a friendship would be disrupted by an insult). A third type involves trying out a new government policy in a pilot project or a new kind of treatment in a clinic or mediation agency.

An example of the latter kind of experi­ment was performed by the author and his students at a community mediation center (McGillicuddy, et al. 1987). Our aim was to assess the impact of med-arb on the process of mediation. Cases that came from City Court were randomly assigned to two conditions: med-arb, in which the mediator became an arbitrator if agreement was not reached, and straight mediation, in which it was not clear what would happen if agreement was not reached.

Advantages of doing laboratory experiments

Laboratory experiments are much more com­mon than field experiments in research on social conflict. There are two main reasons for this: greater control and a wider range of available manipulations and measures (Aronson, et al. 1998).

Control

Laboratory settings usually allow more con­trol over the elements of research than do field settings. This makes it easier to conduct good experiments, for example, to randomize assignment of participants to conditions. Randomization is easy to do in the laboratory but hard to do in public service agencies, because the agency usually insists on deciding who will go into what condition. In the med- arb study, we were lucky that the mediation center allowed us to randomize assignment to the three conditions. But at the end of the study, the director of this center told us that she would never again allow a randomization study because it was too disruptive for her staff.

Greater control also makes it possible to create more precise manipulations in the laboratory, holding more variables constant between conditions and thus reducing the number of confounds and alternative inter­pretations of the results. Holding variables constant also reduces random error, making it easier to reach statistical significance.

Range Ofmanipulations and measures

Another reason for doing laboratory exper­iments is that they allow a wider range of manipulations and measures than do field experiments.

There are many more constraints in field settings. Constraints are especially severe when field experimenters stage elaborate hoaxes in which they try to disguise the fact that participants are in a study so as to encourage normal behavior. This produces three kinds of constraints: (1) In order to avoid detection, investigators must impose conditions that would plausibly be encoun­tered in everyday life. (2) Because they cannot solicit informed consent, they are limited to producing innocuous conditions that do not stress the participants. (3) They are limited to measurements that are available in natural settings. We saw the latter limitation in the horn-honking study, where horn honking was the only available measure of aggression. This measure is ambiguous, because horn honking can be a way to communicate with the driver of the next car as well as a form of aggression.

Experimenters who use social service agencies are also constrained because they must stick to conditions that can be justified as part of the agency's mission. For example, the conditions in our med-arb experiment (McGillicuddy, et al. 1987) were part of the mediation center's ordinary program. Most of the other procedures we might have studied would have bent their program too far out of shape. Another constraint was that we could not tape-record the mediation sessions for fear that the recordings could be subpoenaed by the court. This meant that we could not check the accuracy of our content analysis or go back with new measures after the fact.

Advantages of doing field experiments

If laboratory experiments provide more con­trol and flexibility, why should anyone do a field experiment? One reason is that some con­ditions and variables can only be realistically produced in the field. Thus, one must move to the field in order to study whether arrest or counseling is more effective at stopping men from abusing their wives.

It is also easier to study long-term effects in the field, because laboratory effects are usually too weak to persist very long. The Emery study is a good example. They found that, ‘in comparison with families who litigated custody, nonresidential parents who mediated were more involved in multiple areas of their children's lives, maintained more contact with their children, and had a greater influence in co-parenting 12 years after the resolution of their custody disputes' (Emery, et al. 2001: p. 323).

Motivational and emotional impact

Disputants usually experience stronger pas­sions - such as frustration, anger, and the desire for revenge - in the field than in the laboratory, because they are dealing with issues that are more important to them (Barry, et al. 2004). This means that effects that are produced by such passions are likely to be strengthened as we move from the laboratory to the field, an advantage in research on conflict.

Our med-arb study (McGillicuddy, et al. 1987) is an example of an effect that was strengthened by moving into the field. Before doing the field experiment, we ran some participants in a laboratory negotiation task. A mediator, who could become an arbitrator in the med-arb condition but not in the straight mediation condition, helped them with the negotiation. This laboratory experiment yielded only weak trends in the direction later taken by the results of our field experiment. This is probably because the issues faced in the laboratory were less motivating and emotion-producing than those faced in the field. Hence, the participants in the med-arb condition were less anxious about losing control over the decision making and hence less motivated to reach agreement during mediation.