13.07.2015 Views

Contents

Contents

Contents

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

214 PART III: Experimental MethodsThe external validity of experimental findings also can be established throughpartial replication. Partial replications are commonly done as a routine part ofthe process of investigating the conditions under which a phenomenon reliablyoccurs. A partial replication can help to establish external validity by showingthat a similar experimental result occurs when slightly different experimentalprocedures are used. Consider the same basic experiment done in both a largemetropolitan private university and in a small rural community college; theparticipants and the settings in the experiments are very different. If the sameresults are obtained even with these different participants and settings, we cansay the findings can be generalized across these two populations and settings.Notice that neither experiment alone has external validity; it is the findings thatoccur in both experiments that have external validity.Researchers can also establish the external validity of their findings by doingconceptual replications. What we wish to generalize from any one study are conceptualrelationships among variables, not the specific conditions, manipulations,settings, or samples (see Banaji & Crowder, 1989; Mook, 1983). Andersonand Bushman (1997) provide an example illustrating the logic of a conceptualreplication. Consider a study with 5-year-old children to determine if a specificinsult (“pooh-pooh-head”) induces anger and aggression. We could then do areplication to see if the same insult produces the same result with 35-year-oldadults. As Anderson and Bushman state, the findings for 5-year-olds probablywouldn’t be replicated with the 35-year-olds because “ ‘pooh-pooh-head’ justdoesn’t pack the same ‘punch’ for 5- and 35-year-old people” (p. 21). However,if we wish to establish the external validity of the idea that “insults increase aggressivebehavior,” we can use different words that are meaningful insults foreach population.When Anderson and Bushman (1997) examined variables related to aggressionat the conceptual level, they found that findings from experimentsconducted in laboratory settings and findings from correlational studies inreal-world settings were very similar. They concluded that “artificial” laboratoryexperiments do provide meaningful information about aggression becausethey demonstrate the same conceptual relationships that are observed in realworldaggression. Furthermore, laboratory experiments allow researchers toisolate the potential causes of aggression and to investigate boundary conditionsfor when aggression will or will not occur.What about when results in the lab and the real world disagree? Andersonand Bushman (1997) argue that these discrepancies, rather than evidence for theweakness of either method, should be used to help us refine our theories aboutaggression. That is, the discrepancies should make us recognize that differentpsychological processes may be at work in each setting. When we increase ourunderstanding of these discrepancies, we will increase our understanding ofaggression.Establishing the external validity of each finding in psychology by performingpartial replications or conceptual replications would be virtually impossible.But if we take arguments like those of Dawes (1991) and Sue (1999) seriously, asindeed we should, it would appear that we are facing an impossible task. How,for instance, could we show that an experimental finding obtained with a group

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!