Threats to Internal Validity

Maturation
Changes over time may result not only from specific events but also from processes within the subjects. Maturation refers to processes changing over time and included growing older, stronger, wiser, and more tired or bored. Maturation is only a problem if the design cannot separate the effects of maturational changes from the intervention.

Testing
This threat to internal validity refers to the effects that taking a test one time may have on subsequent performance on the test. In an experiment, pre- and post-intervention tests might be given to evaluate how much an individual improves or deteriorates over time on a particular measure. Performance at the second testing may be influenced by practice or familiarity with the test because of the first testing. Changes at the second testing might not be due to an experimental manipulation or intervention but to the effects of repeated testing.

Instrumentation
This refers to changes in the measuring instrument or measurement procedures over time. For example, an experimenter may alter a survey or the behavioral observations used to measure change in the population of interest. Consequently, changes in the dependent variable over the course of an intervention may result from changes in scoring criteria, rather than changes in actual behavior. This problem may also extend to scoring procedures. It is conceivable that casual remarks by the experimenter at the time of the test administration might affect the subject's response and effectively alter the nature of the test and how the responses are obtained.

Statistical Regression
Regression refers to the tendency for extreme scores on any measure to revert (or regress) toward the mean of a distribution when the measurement device is readministered. If individuals are selected for an investigation because they are extreme on a given measure, one can predict on statistical grounds that at a second testing the scores will tend to revert toward the mean. That is, the scores will tend to be less extreme at the second testing. Therefore, regression is a threat to internal validity if the change due to the intervention cannot be distinguished from the effect of scores reverting toward the mean.

Selection Biases
A selection bias refers to systematic differences in groups based upon the selection or assignment of subjects to experimental conditions. Obviously, the effects of an independent variable between groups can be unambiguously inferred only if there is some assurance that groups do not systematically differ before the variable was applied. Generally, this threat to internal validity often arises when intact groups are selected. Because groups are preformed, they cannot be rearranged for research purposes. Even when groups are not already formed, practical demands may interfere with randomly assigning subjects to groups.

Attrition or Experimental Mortality
The loss of subjects or attrition in an experiment may serve as a threat to internal validity. Loss of subjects does not refer to their veritable demise in most psychological experiments, although this too would present the same threat to validity. Attrition usually refers to subjects dropping out of the experiment over time. If there is a single group of subjects tested at different points in time, changes in overall group performance might be due to the gradual loss of subjects who scored in a particular direction, rather than to improvements in the scores as a function of some intervention. If the experiment is conducted over a long period or includes a follow-up period to assess the long-term effects of treatment, attrition is almost inevitable and may obscure the conclusions that can be drawn.

Attrition is a threat to internal validity if there is a differential loss of subjects between groups. Differential attrition is likely, for instance, in investigations in which experimental conditions are differentially attractive or effective. Subjects are more likely to remain available and cooperative during and after treatment if they are receiving a treatment that is interesting, has little or no cost or adverse side effects, seems plausible, and is effective than if they are receiving a condition that is less desirable on these and related dimensions.

Diffusion or Imitation of Treatment
It is possible that the intervention given to one group may be provided accidentally to all or some subjects in a control group as well. The administration of treatment to the control group may be inadvertent and, of course, opposite from what the investigator has planned. Nevertheless, the effects will be to diffuse what the investigator concludes about the efficacy of treatment. Rather than comparing treatment and no-treatment conditions or two or more distinct treatments, the investigator actually is comparing conditions that are more similar than intended. As a threat to internal validity, the effect of a diffusion of treatment will be to equalize performance of treatment and control groups in the study.

Special Treatment or Reactions of Controls
In an investigation where the intervention, treatment, or program is administered to the experimental group, the no-treatment control group may also be accorded special attention. Awareness of participating in an experiment can influence both intervention and control groups. For example, one group may receive the special program that is viewed as generally desirable whereas the control group receives some compensatory treatment to overcome the inequality. When subjects are aware that they are serving as a control group, they may react in ways that obscure the differences between treatment and no treatment. From the standpoint of internal validity, a problem arises when this awareness differentially affects groups so that the effects of the intervention are obscured.