Comment to Oksanen (2001): reconciling Oksanen (2001) and Hurlbert (1984)

17 years and more than 2000 citations later, the publication of the seminal paper of Hurlbert (1984) still manages to rile ecologists.Especially ecologists working with and experimenting on large systems struggle with the implications of pseudoreplication on how to conduct valid experiments.The mere publication of Oksanen's (2001) paper is a proof of this struggle.Oksanen's analysis adds some valuable ideas and suggestions to the debate, but, as we will show, in the end strengthens Hurlbert's suggestions for statisticians, for editors, and for experimentalists.This paper will try to reconcile the viewpoint of both authors by dividing experimental ecologists into those who recognize the importance of replication and take whatever measures necessary to achieve a good dose of it, those who understand that appropriate scale in the end must always have priority over replication (and recognize the associated consequences of limits in scope and the need for goodwill of the reader), and those who, by mere lack of knowledge or insight, do not see any need for replication (modified from both Hurlbert 1984 and Oksanen 2001).We acknowledge that the first two types can and should actually be united in one and the same person, but we merely want to stress that there is very often an underlying conflict between the perfect experimental design and relevancy of the question addressed.The problem treated in extentio by Hurlbert (1984, Table 1), Sources of confusion) is basically very simple: without replication, a treatment effect is indistinguishable from chance effects.This holds irrespective of the system that is manipulated, and is the basic reason to replicate and disperse treatments.An important problem that led to Oksanen's reply is that creating the perfect experimental design often involves compromises on the scale of the research, which may in the end interfere with the core of the questions addressed.