The most important reason for randomizing participants is to reduce bias as much as it is possible (Rosnow, 2013). But you may ask how is that achieved? Let us assume that you and a group of friends are on a camping trip and all food was stolen by some pesky raccoons happening to be ambling by. Everyone is sitting by the campfire thinking about what to do next when one person discovers 9 small pieces of caramel. While everyone celebrates this good fortune it doesn’t take long to realize there are 10 campers. Who decides which camper will not enjoy a piece of caramel? So instead of arm wrestling for one piece of candy it’s decided to draw names from a hat. So everyone writes their names on small pieces of paper, folds them into smaller pieces and then places them into a Yankees baseball cap. You can’t see the names as each piece of paper is pulled out of the hat, hence the bias that may have occurred by allowing one person decide who gets a piece of candy is virtually eliminated.
In psychology research randomization works similarly but for large populations it’s necessary to select a random sample because it isn’t feasible to collect data from every person in the larger group, or entire population (Rosnow, 2013). But to directly address your concerns related to obtaining valid data directly from large populations. While it may be tempting to forego random sampling the reality is that the risk of making even more mistakes is quite inherent when attempting to study large populations. Human error can lead to non-sampling errors, meaning the larger the sample size the chance that data errors increase as well, and detection of mistakes actually decrease as well (Rosnow, 2013). Bias is also a key consideration with large population samples, known as selection bias it prevents researchers from establishing that the chosen sample represents the target population. Selection bias occurs in convenience sampling where random sampling is not a key consideration, and in instances where key demographic characteristics are omitted or underrepresented (Rosnow, 2013). The question of whether the research conclusions are generalizable rears its head as well. With larger samples it is sometimes difficult to obtain responses from every participant. This happens for various reasons, but in order to reduce instances of bias and to increase the potential that a study has external validity (Anderson, Lindsay & Bushman, 1999), it is advisable to utilize random sampling.
Perhaps your concerns stem from confusion on how to randomly assign research participants after you’ve attained your random sample. Random assignments consist of assigning research subjects to a treatment by chance, meaning that all participants have the same opportunity of being assigned to a group. Let’s take the example of smoking cessation. If you were conducting research on the potential of a new treatment for smoking cessation you would assign subjects randomly to be included in one of two groups. The treatment group received the intervention, the newly developed treatment, while the second group was provided with a pamphlet providing information on how to quit smoking. In this particular study the researchers found that the treatment was equally as effective as reading a pamphlet, but the point is that participants in both groups were randomly assigned thus reducing chances that the results were due to any group characteristics (Rosnow, 2013).
Let’s look at another fictional study that illustrates the use of randomly assigning research participants. You’ve decided to conduct a study to see if listening to Beethoven while taking a test in statistics results in better grades. There a two groups, one will be assigned to take the test while listening to the music (group A), while the other will take the test under normal classroom conditions where no music is played (group B). Let us pretend there will be 10 participants in both groups who have been randomly chosen from students taking an introductory statistics course for the semester. Our participants are then assigned to either group by simply using a coin toss, or by drawing names out of hat, or assigning random numbers (Rosnow, 2013).
Let us now discuss the issue where studies use non-randomization, or forego the use of random sampling. Non-randomized studies are quasi-experimental because they meet the requirements for experimental research involving random assignment or control groups (Rosnow, 2013). Quasi-experimental designs occur, for example, when researchers wish to control the exposure to the intervention, or when it is either impractical or unethical to do random assignments. However, there are some inherent problems concerning internal validity that must be accounted for, and examples in which researchers might increase internal validity include: blinding participants so they are not aware of whether they are in the control group or the group chosen for the intervention; blind treatment providers so they are not aware of who is in the intervention group; and, blind the research evaluators so that they too are unaware of who received the intervention (Rosnow, 2013). Studies without randomization, for example quasi-experimental, are typically correlational, thus it is difficult to identify which variables represent cause and which are the effect. To assess causality it is important that researchers use designs that provide for the best approximation, meaning they must make up for the fact that a study is not randomized. This would include using matched comparison groupings, pretesting measures and assigning the same intervention to other groups at different intervals (Rosnow, 2013). While by no means definitive, it should be understood that quasi-experimentation is perhaps ideal when used in field research or in observational studies, or as a means of augmenting true experimental designs using a mixed-methods approach. However, there is no arguing that experimental research that includes random sampling and assignments is the “gold standard” actually less used in psychology today.
- Anderson, C. A., Lindsay, J. J., & Bushman, B. J. (1999). Research in the psychological laboratory truth or triviality?. Current directions in psychological science, 8(1), 3-9.
- Rosnow, R. L. (2013). Beginning behavioral research: A conceptual primer (7th ed.). Upper Saddle River, NJ: Pearson.
