On misinterpretations of pseudoreplication and related matters: a reply to Oksanen


  • Stuart H. Hurlbert

    1. Stuart H. Hurlbert, Dept of Biology and Center for Inland Waters, San Diego State Univ., San Diego, California 92182, USA. (shurlbert@sunstroke.sdsu.edu)
    Search for more papers by this author


Pseudoreplication has become a widely accepted label for a certain class of statistical error common in the literature of ecology as well as of other fields. A wide-ranging critique by L. Oksanen recently published in this journal criticizes the term and concept and concludes it to be a “pseudoissue,” one reflecting an intellectual disease, “a totally outdated epistemology” known as “inductionism.” The present article addresses some of Oksanen's complaints. His critique is based on a misconception of pseudoreplication, reflects unawareness of most of the literature on the topic, and mistakenly argues that the seriousness of the error is a function of whether an experiment is conducted in an inductive or deductive spirit. Oksanen's advocacy of using resources available for large scale ecology more for large numbers of experiments with unreplicated treatments than for fewer experiments with modest replication of treatments is unrealistic. It is based on an overly optimistic view of the ability of a meta-analysis to compensate for deficiencies, such as very noisy estimates of treatment effects, of the individual studies that are fed into it. A definition is offered of the term manipulative experiment, since adequate ones are lacking in the literature. Attention is called to the fact that for certain types of manipulative experiments lacking treatment replication, there are valid ways to test for treatment effects.