RESUMO
BACKGROUND AND AIMS: In 2019, Daou and Shipley produced an operational definition of 'generalized' soil fertility (FG) for plant community ecology and quantified FG using a structural equation model (SEM) invoking a single latent variable. We evaluate a critical assumption of this model: that FG is generalizable to any combination of plant species; i.e. that any combination of plant species will respond in the same direction to the soil 'fertility' gradient in terms of growth. METHODS: We grew nine widely different species singly in each of 25 soils from southern Quebec, Canada, whose FG value had been previously quantified. The original SEM was tested using every possible combination involving from four to nine species. KEY RESULTS: The assumption was rejected due to a subset of three species that responded to a second latent dimension. We then proposed an alternative model that includes FG plus a second latent variable that measures species' deviations from FG due to specific adaptations to soil pH. This alternative model was consistent with every combination of up to eight species. The predictions of FG when ignoring this second dimension and when using the new model were extremely correlated (r =0.98). CONCLUSIONS: The initial unidimensional model of Daou and Shipley was successful in non-acid soils but not in soils with extreme pH and when species specifically adapted to such extreme soils were included. The alternative two-dimensional model takes into account these exceptions and is consistent with the notion of shared physiological niche responses along a gradient of generalized soil fertility.