EXTERNAL VALIDITY
A common criticism of laboratory experiments is that it is hard to generalize their results to real-life settings - the so-called issue of ‘external validity'. Laboratory settings seem so simplified and rarified that there are doubts about whether people behave the same way in practical settings.
There is no simple answer to this criticism. Sometimes there is a good match between a laboratory setting and a practical setting and sometimes there is not. The critical issue is not whether the two settings look alike but whether the same processes intervene between the independentvariable(s) and the dependent variable(s) and hence whether the laboratory has captured the conditions that allow these processes to go forward.
In defense of the laboratory, a number of generality studies have found similar results in the field as in the laboratory. For example, the two culture-of-honor studies discussed earlier produced nearly identical results (aggression and dominance are two sides of the same coin), and the same mechanism was probably at work in both settings - Southerners, compared to Northerners, had a stronger feeling of being ‘put down' by an insult and hence a stronger need to take action to rescue their dignity.5 Similarly, Anderson and Bushman (1997) showed that eight well- known laboratory findings about aggression were also supported in field settings.
Yet there are also cases where field results do not match those obtained in the laboratory. For example, our laboratory findings about the impact of med-arb were a pale copy of those obtained in the field. As mentioned earlier, this is probably because med-arb produced stronger emotions in the field than in the laboratory.
It is important to note that collecting data in the field is no panacea for the problem of external validity, because variables often relate to each other differently in different field settings.
Consider, for example, Sherman and Berk's (1984) study of recidivism in wife abuse. Can the finding that a jail sentence is superior to counseling be generalized to other kinds of assault or to recidivism in juvenile delinquency? Again, the issue is whether similar mechanisms are at work in the research setting and the setting to which we wish to generalize, and hence whether the critical conditions are the same.To be absolutely certain that one can generalize from a research setting (laboratory or field) to a practical setting, one needs to do generality studies in that precise practical setting. But this is often not feasible, and it is important to be able to make educated guesses on the basis of the research findings on hand. The best way to make educated guesses is to identify the critical mechanisms and variables that produced our research findings and to reason clearly about whether they also obtain in the setting to which we wish to generalize. This means that we must generalize over the ‘bridge of theory'6...and laboratory experiments are a good way to develop and test theory.