Multiple Treatment Interference

Multiple treatment interference is a threat to the internal validity of a group design. A problem occurs when participants in one group have received all or some of a treatment in addition to the one assigned as part of an experimental or quasi-experimental design. In these [Page 980]situations, the researcher cannot determine what, if any, influence on the outcome is associated with the nominal treatment and what variance is associated with some other treatment or condition. In terms of independent and dependent variable designations, multiple treatment interference occurs when participants were meant to be assigned to one level of the independent variable (e.g., a certain group with a researcher assigned condition) but were functionally at a different level of the variable (e.g., they received some of the treatment meant for a comparison group). Consequently, valid conclusions about cause and effect are difficult to make.

There are several situations that can result in multiple treatment interference, and they can occur in either experimental designs (which have random assignment of participants to groups or levels of the independent variable) or quasi-experimental designs (which do not have random assignment to groups). One situation might find one or more participants in one group receiving accidentally, in addition to their designated treatment, the treatment meant for a second group. This can happen administratively in medicine studies, for example, if subjects receive both the drug they are meant to receive and, accidentally, are also given the drug meant for a comparison group. If benefits are found in both groups or in the group meant to receive a placebo (for example), it is unclear whether effects are due to the experimental drug, the placebo, or a combination of the two. The ability to isolate the effects of the experimental drug or (more generally in research design) the independent variable on the outcome variable is the strength of a good research design, and consequently, strong research designs attempt to avoid the threat of multiple treatment interference. A second situation involving multiple treatment interference is more common, especially in the social sciences. Imagine an educational researcher interested in the effects of a new method of reading instruction. The researcher has arranged for one elementary teacher in a school building to use the experimental approach and another elementary teacher to use the traditional method. As is typically the case in educational research, random assignment to the two classrooms is not possible. Scores on a reading test are collected from both classrooms as part of a pre-post test design. The design looks likes this:

If the study were conducted as planned, comparisons of posttest means for the two groups, perhaps after controlling for initial differences found at the time of the pretest, would provide fairly valid evidence of the comparative effectiveness of the new method. The conclusion is predicated, though, on the assumption that students in the traditional classroom were not exposed to the experimental method. Often, in the real world, it is difficult to keep participants in the control or comparison group free from the “contamination” of the experimental treatment. In the case of this example, the teacher in the comparison group may have used some of the techniques or strategies included in the experimental approach. He may have done this inadvertently, or intentionally, deciding that ethically he should use the best methods he knows. Contamination might also have been caused by the students’ partial exposure to the new instructional approach in some circumstance outside of the classroom—a family member in the other classroom may have brought homework home and shared it with a student, for example.

...