The human sciences, like all sciences, should seek generalisation where generalisations may be found. For this reason, and for ethical reasons, it is desirable to sample more broadly than 'Western, Educated, Industrialised, Rich, and Democratic' (WEIRD) societies. However, restriction of the target population is sometimes necessary. For example, we would not recruit young children into studies on elderly care. Under which conditions is unrestricted sampling desirable and undesirable? Here, we use causal directed acyclic graphs (causal DAGs) to clarify structural features of bias that recur in measurement error bias, target population restriction bias at the start of a study, and target population restriction bias at the end of a study. We define any study that exhibits any one of these biases, or standard confounding bias, as **weird** (**w**rongly **e**stimated inferences due to **i**nappropriate **r**estriction and **d**istortion), clarifying why the first step in comparative study design is to mitigate 'weirdness.' We discuss the challenges in avoiding weirdness and explain how workflows for causal inference provide the preflight checklists needed for ambitious, effective, and safe comparative cultural research.