On the usage of single measurements in behavioural ecology research on individual differences

Abstract

Since the early 21st century, behavioural ecologists have increasingly focused on the ecology and evolution of repeatable individual differences in behaviour (aka ‘animal personality’). More and more studies have investigated correlations between behaviour(s) and other phenotypic traits, thereby seeking to understand among-individual variation in behaviour from either a proximate or an ultimate perspective. Statistically, such studies require the estimation of among-individual correlations, necessitating study designs where suites of labile traits are repeatedly assayed. Most published studies, by contrast, instead assume that among-individual correlations can be approximated using data of suites of phenotypic traits measured only once. Such studies take an ‘individual gambit’ as they assume that phenotypic correlations between two traits measured once match among-individual correlations. A literature survey shows that this assumption was made in 62% of empirical studies; this is a worrying trend as a mismatch between the research question and the study (or statistical) design can lead to biased conclusions about biological patterns. In this paper we use a visual approach to illustrate these concerns for a broad audience, thereby complementing previous papers using a jargon more suitable for the statistically oriented. We thereby seek to reiterate the notion that reliable answers to any scientific question require specific types of data. While this commentary underlines the importance of spelling out key assumptions in discussions of (suboptimal) data in scientific papers on individuality, our main message is that researchers should typically be able to avoid publishing studies with conclusions heavily hinging on unverifiable assumptions. This is because researchers normally have the freedom to decide a priori to focus their research towards questions for which appropriate data can arguably be collected. Doing so would greatly facilitate our ability to robustly address key questions of biological interest, such as the ecology and evolution of individuality.