Define external validity
How valid (applicable) the results of the study are externally (outside the study)
i.e. How generalisable are the results to other participants, settings, and times?
External validity vs. Confounds
External validity does not vary within each level/ condition of the IV
External validity applies to the whole study - does not cause results, but establishes results may only be caused in the settings/ participants/ time the study was taken in.
External validity examples
- Exercise program works when at the gym; did not test program at home
- Medication only effective for men; did not test meds on women and children
- Study technique works for uni students; did not test study techniques on primary/ secondary schoolkids
Relationship between external validity and sensitivity
Applying a process/ trait to all conditions limits confounds, and increases sensitivity (results easily seen when conditions are more similar).
At the same time, similar situations do not account for varied situations of other subpopulations/ everyday life, thus have lower EV.
Factors of external validity
[Spacial EV]
* Setting
* Tasks + Materials / Process
* Sampling
[Temporal EV]: Time
Setting as a factor of EV
Are the results replicable in situations other than the situation of the experiment.
Example of setting as a factor of EV
E.g. Is Milgrim’s obedience study replicable outside of a university (pretige, scientific) setting?
Setting checklist
- Research setting (scientist dressed professionally?)
- Physical setting (at a university or a bar?)
- Researcher attributes (big imposing males?)
- Researcher expectancy (communication)
Tasks/ Materials as a factor of EV
Are the results replicable using different tasks/ materials from the experiment?
Example of tasks/ materials as a factor of EV
E.g. Is Milgrim’s obedience study replicable if the participant is raising the temperature of a metal pad attached to the “learner” instead of electrically shocking them?
E.g. Is Milgrim’s obedience study replicable if the “learner” was younger, or a woman?
Tasks/ Materials checklist
- Artificial-ness of study (does it feel real?)
- Carryover effects (if Milgrim started with a lethal dose?)
- Pre-test effects (pre-screening questions of morality or obedience?)
- Demand characteristics (Encourage lethal use from beginning of experiment)
Sampling as a factor of EV
Are the results replicable with different populations?
Example of sampling as a factor of EV
E.g. Is Milgrim’s obedience study replicable if the participants were not university students?
E.g. Or if participants were of different cultures, different genders, etc.
Sampling checklist
- Convenience samples
- Uni (psych) student samples
- Volunteers
- Attrition
Sampling strategies list
- Convenience sampling: recruit target number of diverse participants, non-randomly
- Simple Random: random draw, may not reflect population
- Proportionate Stratified Random: randomly sample from sub-groups.
- Cluster: identify subcategories (e.g. diff schools when studying education), and randomly sample from those clusters
Temporal limitations of EV
