The online survey was disseminated through our professional network and advertised on the Ecolog‐l mailing list and Twitter. There was much interest in the topic, and 103 responses were collected which revealed the following key findings:
Fifty‐eight percent of respondents had faced a research question where they felt pseudoreplication was an unavoidable issue (Table
Categories of pseudoreplication problem identified in the questionnaire and the frequency with which respondents described themLandscape‐scale treatments/monitoring (including manipulations of forest stand structure) 10 Nested designs with insufficient replication at site level 9 Wildlife behavior/physiology (including repeated measures on a small number of individuals) 9 Confounded site/stand and treatment (including multisite vegetation chronosequences) 8 Demography and disease and – what is the appropriate analysis level site, plot, or individual? 8 Exclosures at a single site (including grazing and irrigation studies) 6 Aquatic ecology + hydrology ‐ unreplicated ponds/lakes/watersheds 5 Fire behavior and effects (including studies of individual wildfires) 5 Single‐site case studies or phenomena limited to one location 5 Spatial autocorrelation 3 Repeated measures of vegetation change (including studies on a single relevé) 2
Of those who'd faced the problem, 85% were aware of the concept before they started their research although most (89%) were not discouraged by it.
Nearly 70% of respondents had read Hurlbert (
Two‐thirds of respondents tried to deal with the issue during their statistical analysis or by acknowledging the limits of statistical inference possible given their design. A third dealt with the issue during hypothesis formation by clearly defining their population, and a fifth framed their conclusions as new hypotheses. Only four respondents admitted they just hoped no one would notice (which is honest but naughty!).
Half of the respondents admitted they'd had difficulties getting their research published, and 17% were never able to get their studies published at all. Of those who experienced publication difficulties, 41% received major corrections but 55% had their paper rejected (with less than half of those being given the option to resubmit). A quarter of the respondents had ended up in prolonged arguments with reviewers and/or editors.
When completing peer reviews, reviewers who had not encountered pseudoreplication issues in their own research, though a relatively small proportion of all reviewers (25%), appeared to be considerably more likely to reject or ask for resubmission of papers with pseudoreplication (59%) than those who'd had to deal with the issue themselves (36%).
pseudoreplication was inevitable in many types of research due to cost, scale, and other “real‐world” environmental issues such as a wildfire, drought, or flood only occurring once;
many kinds of pseudoreplication can be dealt with statistically using appropriate nesting or random effects.