True experiment, field experiment, quasi-experiment or natural experiment? The answer is often a wild look in the eyes and a shrug of the shoulders. It is not always easy to be certain! See below for an explanation of the differences. All sources used are referenced at the bottom of the page, and linked study summaries are, of course, from Psychology Sorted.
The easiest one to define is the true experiment.
Often called a ‘laboratory/lab’ experiment, this does not have to take place in a lab, but can be conducted in a classroom, office, waiting room, or even outside, providing it meets the criteria. These are that allocation of participants to the two or more experimental (or experimental and control) groups or conditions is random and that the independent variable (IV) is manipulated by the researcher in order to measure the effect on the dependent variable (DV). Other variables are carefully controlled, such as location, temperature, time of day, time taken for experiment, materials used, etc. This should result in a cause and effect relationship between the IV and the DV. Examples are randomised controlled drug trials or many of the cognitive experiments into memory, such as Glanzer and Cunitz_1966.
A field experiment is similar, in that individuals are usually randomly assigned to groups, where this is possible, and the IV is manipulated by the researcher. However, as this takes place in the participants’ natural surroundings, the extraneous variables that could confound the findings of the research are somewhat more difficult to control. The implications for causation depend on how well these variables are controlled, and on the random allocation of participants. Examples are bystander effect studies, and also research into the effect of digital technology on learning, such as that conducted by Hembrooke and Gay_2003.
A quasi-experiment is similar to either or both of the above, but the participants are not randomly allocated to groups. Instead they are allocated on the basis of self-selection as male/female; left or right-handed; preference for coffee or tea; young/old, etc. or researcher selection as scoring above or below and certain level on a pre-test; measured socio-economic status; psychology student or biology student, etc. These are therefore, non-equivalent groups. The IV is often manipulated and the DV measured as before, but the nature of the groups is a potential confounding variable. If testing the effect of a new reading scheme on the reading ages of 11 year olds, a quasi-experimental design would allocate one class of 11 year olds to read using the scheme, and another to continue with the old scheme (control group), and then measure reading ages after a set period of time. But there may have been other differences between the groups that mean a cause and effect relationship cannot be reliably established: those in the first class may also have already been better readers, or several months older, than those in the control group. Baseline pre-testing is one way around this, in which the students’ improvement is measured against their own earlier reading age, in a pre-test/post-test design. In some quasi-experiments, the allocation to groups by certain criteria itself forms the IV, and the effects of gender, age or handedness on memory, for example, are measured. Examples are research into the efficacy of anti-depressants, when some participants are taking one anti-depressant and some another, or Caspi et al._2003, who investigated whether a polymorphism on the serotonin transporter gene is linked to a higher or lower risk of individual depression in the face of different levels of perceived stress.
Finally, natural experiments are those in which there is no manipulation of the IV, because it is a naturally-occurring variable. It may be an earthquake (IV) and measurement of people’s fear levels (DV) at living on a fault line before and after the event, or an increase in unemployment as a large factory closes (IV) and measurement of depression levels amongst adults of working age before and after the factory closure (DV). As with field experiments, many of the extraneous variables are difficult to control as the research takes place in people’s natural environment. A good example of a natural experiment is Charlton (1975) research into the effect of the introduction of television to the remote island of St. Helena.
The differences between quasi experiments and correlational research, and between natural experiments and case studies are sometimes hard to determine, so I would always encourage students to explain exactly why they are designating something as one or the other. We can’t always trust the original article either – Bartlett was happy to describe his studies as experiments, which they were not! Here’s hoping these examples have helped. The following texts are super-useful, and were referred to while writing this post.:
Campbell, D.T. & Stanley J.C. (1963). Experimental and Quasi-Experimental Designs for Research. Boston: Houghton Mifflin (ISBN 9780528614002)
Coolican, H. (2009, 5th ed.). Research Methods and Statistics in Psychology. UK: Hodder (ISBN 9780340983447)
Shadish, W.R., Cook, T.D. & Campbell, D.T. (2001, 2nd ed.). Experimental and Quasi-experimental Designs for Generalized Causal Inference. UK: Wadsworth (ISBN 9780395615560)