Pseudoreplication in Primate Communication Research : 10 Years On

Pseudoreplication is the statistical error of collecting numerous datapoints from a single unit (such as an individual), which are not independent, and applying statistical methods that assume independence of data. Importantly, pseudoreplication increases the chances of Type 1 errors (i.e., false positives), bringing findings and conclusions based on pseudoreplicated analyses into question. Ten years ago, Waller et al. (2013) published a paper highlighting the prevalence of statistical pseudoreplication throughout the nonhuman primate communication literature. In this current study, we examined the literature published since the original publication (between 2009 and 2020; 348 papers) to assess whether pseudoreplication is still as widespread as it was, if it has become more problematic, or if the field is beginning to overcome this issue. We find that there has been a significant decrease in pseudoreplication over the past ten years (38.6% then, compared with 23.0% now). This reduction in pseudoreplication appears to be associated with an increase in the use of multilevel models throughout primatology (which allow for nonindependent data to be nested appropriately). Pseudoreplication was historically more prevalent in research using observational (vs. experimental) methods and those working with wild (vs. captive) primates. However, these biases do not seem to exist in more recent literature with a more comparable likelihood of pseudoreplication seen across the field regardless of methods. Although these current findings relate specifically to primate communication research, we think they will translate broadly across nonhuman communication research, and throughout biology. We continue to emphasise the need to monitor these issues, as although now seen at much lower rates, pseudoreplication is still present and therefore potentially impacting the accuracy of findings.