Experimental Research Designs

Experimental research designs are familiar to most people as the classic science experiment, much like those performed in high school science class.  Imagine that half of a group of plants receive fertilizer and half do not, though all receive the same amount of water and sunlight.  If the fertilized plants have grown larger at the end of the experiment than those that did not receive fertilizer, we can conclude that fertilizer causes greater growth.  Furthermore, if we are certain that all plants in both groups were identical at the start of the experiment and that receiving fertilizer was the only way they differed, we can conclude that fertilizer will cause growth in all similar plants.

Experimental research designs are also widely used in psychology, education, and the social sciences, and are considered the gold standard for evaluating causal hypotheses.  Their logic as illustrated above is straightforward, though their actual execution can be quite challenging.

First, all experimental research designs are based on comparison between two or more groups.  These groups must be composed of subjects who are similar on all characteristics which might influence the outcome of interest; otherwise, it is impossible to rule out the possibility that any observed differences at the end of the experiment were due to baseline differences between the groups at the start of the experiment.  The ideal way to ensure comparable groups is through completely random assignment to either the treatment or control group, which ensures that any characteristics which might influence the outcome will be randomly distributed between the two groups.

The key characteristic of all experiments is manipulation by the researcher; common terms for this manipulation include treatment, stimulus, or intervention.  In short, this manipulation is the experiment – it is what the researcher believes to be the cause of the outcome of interest.  Familiar examples include using an innovative teaching technique with one group of students, showing a new advertisement to one group of consumers, or providing preventive health information to one group of at-risk individuals.  In each case, the researcher hypothesizes that the manipulation applied to one group will be the cause of a different outcome relative to the unexposed (or control) group.

The idea of “control” plays a crucial role in experimental research designs.  Researchers must be able to carefully control all aspects of the experiment in order for a claim about causality to be made.  Researchers must control any baseline differences between the groups (through random assignment, careful matching of group members, or statistical control of any differences).  Researchers must also ensure that the treatment is administered in the exact same way every time.  At the conclusion of the experiment, the researcher assesses both the control group and the treatment group on the outcome of interest.  If a difference exists – for example, if the group that received the intervention exhibits less frequent health risk behavior – then the researcher can reasonably conclude that the intervention was the cause of reduced risk behavior.

Though experimental research designs are the best available means for assessing causal hypotheses, they require strict adherence to rigorous procedures.  Ensuring that treatment and control groups are truly comparable can be difficult, especially when researchers are not able to randomly assign subjects to a group.  The types of manipulations which researchers can reasonably apply are also limited by ethical and practical considerations.  The generalizability of findings from experimental research designs can be uncertain, as subjects who volunteer to participate in an experiment may differ from those who are unwilling to do so.  Given the challenges associated with conducting research on complex processes using human subjects, many researchers use modified versions of experimental research designs, known as quasi-experimental designs, which relax many of the strict requirements of true experimental research designs but which also reduce the researcher’s ability to make causal statements.