Testing the complex hypotheses reflecting the effect of stressors in natural conditions needs advanced multivariate analyses. This study presents a promising multivariate analysis (structural equation model: SEM) able to model networks of causal relationships and conceptual (latent) variables reflected by several observed variables. The aim is to show the interest of SEM for analysing monitoring data based on the example of the bioavailability of metals to earthworm. The concept of bioavailability perfectly illustrates the key features of SEM relevant for field data analysis (e.g. causal relationships and numerous variables supposedly reflecting bioavailability). A SEM reflecting the causal assumptions: the more metals are available in the soil the more they enter the organism, and an effect can only appear if metals have entered the organism, was tested. This definition involves 3 sub-concepts: 1) environmental availability: available metals in soil, measured by e.g. chemical metal extractions, 2) environmental bioavailability: metal absorbtion by the organism, measured by internal metal concents, and 3) toxicological bioavailability: internal metals leading to effects, measured by biomarkers. In the SEM each sub-concept was a latent variable reflected by several observed variables. The SEM was tested based on a large dataset for a relevant earthworm species: Aporrectodea caliginosa exposed to 31 field soils. The first interest of SEM was the possibility to test for causal relationships. For Cd, the SEM was supported by the data and suggested that earthworms may not be exposed only to readily available metals (e.g. dissolved Cd content in soil), but possibly also to metals bound to soil particles ingested by earthworms. Another key feature of SEM was to model unmeasured concepts reflected by several indicators. The observed variables of a latent are correlated in SEM. This is an important advantage of SEM, valuable when analysing monitoring data and dealing with the issue of inter-correlations between e.g. physico-chemical variables. This study shows that SEM can elucidate the causal links involved in the exposureeffects of chemical stressors. The potentialities of this modelling framework applied on large datatsets such as monitoring data are tremendous. The ability to model networks of interacting components will help refining our causal understanding of the effects of stressors on biodiversity and ecosystem functioning in natural environments.