Sir

Your Editorial “Let's replicate” and News Feature “The trouble with replication” ( Nature 442, 330; 2006 and Nature 442, 344–247; 2006) address the problem of how to obtain reliable knowledge in science. These articles usefully point to difficulties in conducting successful replications, and note the change in social processes associated with resolving contested experimental results. Taken together, though, they risk misleading the reader about the process of replication in science.

Your Editorial acknowledges the diversity of the practice of replication in science, but uses a rather narrow definition of replication. By writing of “direct”, “full” and “exact” replication, you seem to presuppose an ideal of literal replication — the idea that replication involves attempts to produce a 'carbon copy' of the original experiment. Of course, such an objective cannot be attained, if only because of the unavoidable changes in time, location and investigative personnel that replication brings with it.

Close replication is a more realistic aim in science than literal replication. A close replication uses the conditions and procedures of the initial study, perhaps correcting for its perceived inadequacies. It is a checking strategy that attempts to achieve similar results to the first study.

Close replication must, however, be distinguished from constructive replication. The latter process is undertaken to demonstrate the extent to which results hold across different experimental methods, treatments, populations and occasions. It is a triangulation strategy designed to reveal the extent to which the results identified by successful close replication can be generalized. Constructive replication is of vital importance to science because it is the main means by which empirical phenomena are detected. Your Editorial mentions the scientific aim of confirming and extending unpublished experimental conclusions, but in my view it errs in not stating that this is a form of replication.

To be sure, craft or tacit knowledge is often needed to establish the reliability associated with effectively carrying out experimental studies and successfully replicating their findings. Social studies of science have made clear that there is also an unavoidable social dimension to the successful practice of experimental science. However, it is the successful use of consistency tests in the practice of experimental replication that provides the evidential basis for its claims about empirical phenomena.