Piecewise structural equation modelling in ecology: common pitfalls and best practices for causal inference

Structural equation modelling (SEM) has become a cornerstone tool in ecology for testing hypotheses about causal relationships and disentangling direct and indirect effects. However, its growing use has led to frequent conceptual and methodological misapplications. Here, I clarify the theoretical foundations of SEM and provide practical guidelines for its rigorous implementation in ecological research. SEMs assess whether data are consistent with hypothesised causal structures – they do not prove causation – and every path must be justified a priori by ecological reasoning. I outline common pitfalls, including overfitting, unjustified bidirectional paths, inconsistent data transformations, and theory‐free modelling. Special attention is given to the logic and diagnostics of piecewise SEMs, including Fisher's C test, d‐separation, and model selection. An accompanying R vignette demonstrates transparent, reproducible workflows that integrate model specification, diagnostics and interpretation. Properly implemented, SEMs offer a powerful framework for refining ecological theory and identifying mechanisms linking environmental and biological processes. By explicitly linking SEM specification to ecological theory (e.g. trophic cascades, environmental filtering, and indirect effects in community assembly), this framework facilitates theory‐driven causal inference and the formal testing of complex ecol‑ogical hypotheses.