NTERNAL AND XTERNAL ALIDITY LINICAL RESEARCH

1 Advances in clinical psychology criticallydepend on methods researchers use forinvestigating the causal relationshipsamong variables. RESEARCH questions commonlyinclude issues about the putative mechanismsassociated with various forms of psychopathol-ogy ( , do dysfunctional beliefs cause depres-sion?) and questions about the effects oftreatments ( , does cognitive therapy cause agreater reduction in eating disorder symptomsthan does placebo?). How do we go about draw-ing objective conclusions about causality?

2 Thisis a question about whether the manipulation ofone variable ( , the independent variable) haseffects on another variable ( , the dependentvariable). The answer to this question is morecomplex than it might seem. Consider the fol-lowing common 1:Dr. Smith is a well-known local advo-cate of a controversial form of psychotherapy. Heclaims that it works faster and more powerfullythan all other treatments. For many years he haspracticed this treatment and trained other cliniciansthrough workshops, even though there were no sci-entific data on its challenged on thispoint, Dr.

3 Smith retorted that he had seen all theproof he needed with his own eyes. That is, heclaimed his patients almost invariably benefitedfrom his therapy. When a treatment outcome studywas conducted and published by an independentgroup of investigators, it was found that the psy-chotherapy used by Dr. Smith was no more effectivethan giving patients a placebo. Dr. Smith s responsewas there must have been something wrong withthe RESEARCH ; the researchers didn t include patientsseen in the real world.

4 Scenario 2:The interns and their clinical supervi-sors gathered in the seminar room for the weeklyjournal club. This week s article, which recentlyappeared in a leading journal, described an experi-mental investigation of the effects of cognitive fac-tors (expectations of disapproval) on social participants led to expect high disap-proval (from an experimental confederate) experi-enced more social anxiety than participants whowere led to expect no investigatorsconcluded that expectations of disapproval cancause social anxiety and likely play a role in clinicallysevere conditions such as social anxiety the many methodological strengths of thestudy, the participants of the journal club quicklysearched out the methodological weaknesses.

5 Oneof the interns raised a particularly important233 internal ANDEXTERNALVALIDITYINCLINICALRESEARCHSTE VENTAY LO RGORDONJ. G. 11/17/2007 5:40 PM Page 23question about the generalizability of the study. Herobservation was met with nods of approval fromher supervisors. By the time the journal club had fin-ished, all the attendees had convinced themselvesthat the study was fatally flawed. Some attendeeseven left the meeting with the impression that psy-chological RESEARCH , even RESEARCH published inleading journals, is largely a waste of chapter is written largely in response tothese two types of scenarios, which we haveencountered time and again.

6 The reactionsillustrated in these scenarios retard the scien-tific progress of clinical psychology. The firstscenario raises many issues regarding internaland external validity . Dr. Smith claims that allhis patients benefit from his treatment. Yet, hisseries of case observations have many problemsof internal validity . His dismissal of a recentstudy of his psychotherapy raises the issue ofexternal validity . The journal club scenarioraises the question of what we can concludefrom RESEARCH that does not have perfect internal and external important issues raised in these scenariosare the focus of the remainder of this chapter.

7 Wewill begin by defining internal validity and illus-trating the various threats to it. Some studies,such as those using quasi-experimental designs,are widely used in clinical RESEARCH , despite theirimperfect internal validity . After reading thischapter, you should have a good understandingof why such studies are used and why they areuseful. After discussing internal validity , we thenexamine the concept of external validity (gener-alizability) and consider the relationshipbetween internal and external validity .

8 As youwill see, there are some RESEARCH situations inwhich high external validity is vital and othersituations in which it is not a priority. Finally, wewill conclude with some comments about howscientific knowledge can be advanced eventhough most RESEARCH studies have imperfectinternal or external validity . Throughout thisdiscussion we will consider a number of com-monly used experimental designs that have beendeveloped to deal with issues of internal andexternal validity . Our discussion of these designswill be illustrative rather than discussions of experimental and quasi-experimental designs are available elsewhere( , Asmundson, Norton, & Stein, 2002;Barlow & Hersen, 1984; Campbell & Stanley,1970; Cook & Campbell, 1979; Onghena &Edgington, 2005).

9 INTERNALVALIDITYI NTERNAL validity is the degree to which observedchanges in a dependent variable can be attrib-uted to changes in an independent , internal validity is a matter of degree ( ,high, medium, low) rather than one of presenceor absence. The researcher s confidence in his orher findings is proportionate to the strength ofinternal validity of the RESEARCH design (Finger &Rand, 2003). True experiments are designs thathave strong internal validity ; that is, participantsare randomized to experimental conditions, andother means are used to ensure that changes inthe dependent variable can be attributed to theexperimental manipulation of the independentvariable.

10 Quasi-experimental designs haveweaker internal validity , as we will illustrate are several types of threat to internal valid-ity (Cook & Campbell, 1979; Finger & Rand,2003; Rosenthal, 2002), including history, maturation, testing, instrumentation, statistical regression, attrition, selection, interactions with selection, diffusion or imitation of treatments, compensatory equalization of treatments, and experimenter of these threats to internal validity aredefined and illustrated in the following changes in the dependentvariable are due to some extraneous event thattakes place between pre- and posttest.