External Validity Generalization—Applying experimental results to a different situations or populations than those found in the study. The extent to which we can generalize the results of the study to other populations, settings, times, measures, etc. The goal of good research is to make a connection between results of “this study” and the situations/ people in the real world . Population Generalization The degree to which a study’s results can be assumed true of people not included in the study. (from sample to pop) Would the same results be obtained with a different sample from the pop? The Replication Fix Replication—conducting a study in exactly the same way it was conducted before. Goal: More confidence in results consistently found upon multiple replications Replication with extension—replicating a previous study but with different participants or under different conditions Often used in cross-cultural and cross-group research. Results apply to Native Americans? Older persons? People of the WWII generation? Etc. Environmental Generalization The degree to which similar results would be obtained in environments/situations different than that of the experiment itself. If results occur only in the lab, then their practical application is limited (i.e., the study is pointless) Temporal Generalization Applying results to a time different from the time during which the study was conducted. As times change, old knowledge/findings become less relevant Occasionally discover "timeless” findings: findings replicated across time, eras, and seasons. Example: fight or flight mechanisms, tendency of men and women to process information differently Tend to be more biological findings, but even those are often subject to changes over time because the brain is plastic and fluid (not a stone, more like clay) Threats to External Validity Any characteristic of the study that limits the generalizability of the results May come from Experimental Methods Participants Method’s-Based Threats to External Validity Interaction of testing and treatment—taking a pretest may influence participants’ responses to a post-test. Responses to the treatment may be different due to the pretest (serves as a primer that may confound results) R O1 R O3 O2 (control group) X O4 (experimental group) Particularly troublesome for studies investigating attitude change Example: Measure attitudes toward teachers, watch Mr. Holland’s Opus, measure attitudes again A way to control this threat is to have non-pretesting group(s) Method’s-Based Threats to External Validity Interaction of selection and treatment—when results (or the treatment effect) are found only in a specific sample of participants. A particular threat for studies that require hard-to-find participants. It is more likely these participants will be uniquely different from the population (not representative) Example: Intentional sampling: A sample of the uber-rich, a sample of hospice nurses Unintentional sampling: summer school students, English teachers Method’s-Based Threats to External Validity Multiple-Treatment Interference—occurs when a set of findings is due to participants receiving multiple treatments in the same experiment Example: results on a memory task (learning a list of words) for participants who had already learned several word lists. Their previous treatment is impacting (likely heightening) their ability to learn a new task. Thus, results are NOT generalizable to persons without these previous treatments. Example: Test-taking skills of seniors vs. freshman. Results from seniors are not applicable to freshman. Method’s-Based Threats to External Validity Reactive arrangements—when the experimental situation alters participants’ behavior. Due to need for control, experimental situations are often highly contrived /artificial Since experimental settings do not exist in the real world, we cannot say that similar results would be found in the real world. “The play-acting, outguessing, up-for-inspection, I’m-a-guinea-pig, or whatever attitudes so generated are unrepresentative of the [real world], and seem to be qualifiers of the effect of X [the treatment], seriously hampering generalization.” (Campbell & Stanley, 1966, p. 20). Participant-Based Threats to External Validity Good Subject Attempt to corroborate experimenter’s hypothesis Negativistic Subject Attempt to refute experimenter’s hypothesis Faithful Subject Follow instructions and behaves in socially appropriate ways more readily than they may in real-world situations Apprehensive Subject Overly concerned with evaluations of their performance Participant-Based Threats to External Validity The infamous white rat The Norway white rat represents .001% of all living creatures that could be studied. Between 1993 and 1995, 75% of studies published in the Journal of Experimental Psychology: Animal Behavior Processes reported results obtained from rats and pigeons. Participant-Based Threats to External Validity College students are the human researcher’s lab rat and extra credit is the cheese. Sample of Convenience—including certain participants simply because they are readily available, often does not involve random sampling. Can college students represent an entire population. Absolutely NOT! Example: attitude change—traditionally aged college students have less crystallized attitudes than persons older than they. Example: learning/memory differences due to academic demands Example: some psychological disorders disproportionately impact younger (schizophrenic break early 20s) or older persons (Dementia, personality disorders calm down) Participant-Based Threats to External Validity The Gender bias Three quarters of all psychology majors are women (problem for academic researchers) The Caucasian bias Cannot automatically generalize results based on Caucasian persons to non-Caucasian persons. The American bias Cannot automatically generalize results based on American participants— even American participants of minority status (Beware acculturation).