Replication

Any research conclusion is suspect if it is based on results that are obtained by chance. An indication of happenstance is a failure to obtain the same results when the original study is repeated. Replication refers to obtaining the same results when a previous study is conducted again in as similar a way as possible. The replication issue is discussed here in the context of choosing experimentally two theories.

The decay theory says that learning dissipates with time if one does not rehearse what has been learned. It implies that longer delays give rise to more forgetting. Experimenter D manipulates delay of recall to test it and obtains data consistent with the theory.

The interference theory says that forgetting occurs when subsequent learning interferes with earlier learning. It is expected that one forgets more if one has more additional learning. Experimenter E finds support for it by manipulating amount of the additional learning.

The onus is on both D and E to render it possible for themselves or any interested researcher to repeat their respective study by describing succinctly its experimental objective, specific to-be-tested hypothesis, method and materials used, and data collection, as well as analysis procedures.

A failure to replicate either study does not necessarily mean that there is no support for either theory. First, the description of the original study may not be as succinct as required. Second, there may be subtle variations in execution when the original procedure is repeated (e.g., the tone used in instructing the subjects). Third, the repeat study cannot (or may not) have all the features of the original study. For example, it might be impossible (or inadvisable) to employ the same group of participants in the repeat study. When another random sample is selected from the original population for the repeat study, an atypical sample may be selected by chance. Even if it were possible to employ the original participants, the participants themselves might have changed in ways that are relevant to the study.

Be that as it may, replicating Ds (or Es) results proves neither the decay nor the interference theory. The variables, delay of recall and amount of additional learning, are confounded with each other in the sense that longer delays necessarily mean more additional learning; being given more additional learning may mean longer delays in testing the original learning. Hence, every successful replication of findings consistent with the decay theory is also consistent with the interference theory, and perhaps vice versa.

In short, repeating an original study may duplicate a confounding variable. That is, replication does not disambiguate the confounding between two variables. This difficulty necessitates the adoption of conceptual replication, which is better categorized as converging operations, as a means of substantiating theories.