Minimizing observer bias in animal behavior studies revisited: Improvement, but a long way to go

Abstract

For decades, texts on methods in animal behavior research have stressed the need for observers of behavior to work to minimize potential unconscious biases in their coding of data. Two major ways of minimizing these biases are to carry out data coding blind to the key comparisons being made in the study and to have high inter-observer reliability. Over 10 years ago, Burghardt et al. (2012, Ethology, 118, 511) reviewed five major journals in the field of animal behavior and coded randomly selected articles from five decadal volumes (1970 to 2010). That earlier article found poor rates of reporting these two common methods for minimizing potential biases. Here, we carried out similar coding for the 2020 volumes from those same five journals. We found that rates of reporting have increased in all five journals – some substantially. However, rates of reporting still lag behind the journal Infancy, which publishes research on human infant development and relies on many of the same behavioral observation and coding methods used by animal behavior researchers. Given increased calls for transparency and reproducibility in many different fields of scientific study, we argue that we – researchers, reviewers, and editors – can and need to do better at making sure we are actively conducting research in ways to minimize potential observer biases.