If a research is highly reproducible, it may be repeated over and over again and get the same results each time. External validity is vital because if a research cannot be reproduced, how can we know that the results are correct?
External validity refers to the ability of a study's findings to apply to people outside of the laboratory or classroom. Studies must be able to be replicated by others to have high external validity. Without this ability, we would not be able to say with any certainty what effects, if any, the findings of a study might have on people outside of the lab or classroom.
Studies need to be reproducible because without this ability, we would not be able to say with any certainty what effects, if any, the findings of a study might have on people outside of the lab or classroom.
For example, let's say that I conduct a study where I find that people who eat chocolate tend to like cats more than people who do not eat chocolate. This result seems reasonable given that most people who eat chocolate also like cats, so this would be an expected finding. However, I cannot predict or understand whether this result would still hold true if another researcher conducted the same study some time in the future. If my result was not reproducible, we would not be able to say with any certainty what effect, if any, eating chocolate might have on liking cats.
It is critical that study findings be reproduced since it allows other researchers to test the research's conclusions. Replicability keeps researchers honest and can instill trust in readers. If the research can be replicated, any incorrect findings may be disproved. Science would be impaired without replication because this is how we learn which theories are valid and which are not.
When research results cannot be reproduced, this calls their validity into question. For example, if one researcher claims to have found a link between coffee drinking and cancer risk but when another group of scientists tries to reproduce these findings they do not confirm this relationship, then we know there is a problem with the first set of results. In this case, the original finding was probably due to a statistical error or some other issue outside of caffeine's effect on cancer cells.
Replication is also important for establishing scientific standards. If one researcher claims to have proven something using data from studies performed by others, then others need to be able to verify this result before it can be accepted as truth. This means compiling documentation about the methods used in each study and having independent researchers perform these experiments or survey studies where possible. Only then can we be sure that certain findings are accurate and reliable.
Finally, replication is necessary to establish new standards. Scientists often compare their results to those of previous studies to see what other researchers have found.
Without replication, there is no way to know whether a particular result is accurate or not.
Replication provides evidence that the original finding was not a fluke but rather a reflection of some underlying truth. If similar results are obtained by others, then this increases the likelihood that there is something important going on in those studies. It also decreases the likelihood that any one study might be influenced by random error or other factors that could compromise its internal validity.
Even if original findings are not replicated, they often lead to new lines of investigation that may reveal further insights about the topic under review. For example, Drs. Robert Rosenthal and Lenore Jacobson conducted two independent studies using single-subject design to investigate the effectiveness of behavior therapy for treating obsessive compulsive disorder. The first study (1968) found modest support for the approach while the second study (1972) provided strong support for its efficacy.
Despite these successes, many therapists remained skeptical about behavioral treatments because they had been unable to reproduce their findings. But after learning about statistical power, they were able to conduct their second study with a sample size large enough to detect significant differences between treatment conditions should they exist.