Reconciling Rigor and Range: Observations, Experiments, and Quasi-experiments in Field Primatology
- First Online:
- Cite this article as:
- Janson, C.H. Int J Primatol (2012) 33: 520. doi:10.1007/s10764-011-9550-7
I summarize here my experiences studying biological mechanisms of primate behavior across more than 3 decades of observing behaviors and performing ecological field experiments on a wild primate population. I begin with a general overview of the purpose of controlling variation in predictor variables, then describe how observational and experimental studies control such variation. After a description of the practical problems of using observational data to infer mechanisms or causes of a particular behavior, I describe the success and challenges of conducting ecological field experiments on animals, including some examples of (my own) failed experiments. Finally, I propose a middle ground between observations and experiments, what I term quasi-experiments. Quasi-experiments use systematic observation of predictor variables to measure their effects on the occurrence of a behavior, rather than measuring the values of the predictor variables only when the behavior occurs. Regardless of the study methodology, the use of general linear mixed models (GLMMs) for statistical analysis permits stronger inference about the effects of predictor variables than is provided by older statistical methods. GLMMs allow researchers to disentangle from the error term the variation that is due to consistent differences between focal animals, groups, years, food species, or other uncontrolled but repeatedly sampled categorical variables that affect our observations. Reducing the error term then allows more powerful inferential tests of the effects of the hypothesized causal variables. All field studies will benefit from a better quantitative theoretical framework within which to interpret the match between expected and observed outcomes.