AbstractAimThe island species–area relationship (ISAR) assumes that the area of islands is homogeneous, or scales with environmental heterogeneity across an archipelago, which is not always the case. We compare the performance of models that adjust or substitute for island area with measures of habitat diversity, island age and resource availability for two taxonomic groups.LocationFive hotspot archipelagos (Azores, Galapagos, Hawaii, Cape Verde, Canary Islands).TaxaVascular plants, birds.MethodsWe used the mathematical framework of the power law to compare relevant models, treating the one containing only area as a null model against which others were compared. Data were collated from databases and the literature. Models were compared using linear regression within archipelagos and via mixed effect models with archipelago as a random effect.ResultsWeighting of island area by habitat diversity and resource availability systematically improved statistical significance and model fits versus the area only power law. Models including island age did not show the same systematic improvement in model fits. For vascular plants, weighting islands by resource availability (energy and water) performed better than weighting by habitat diversity, although for birds these weightings performed equally well.Main ConclusionsGiven that islands within archipelagos are fairly heterogeneous in climate, topography and geology, it is worth accounting for this in ISARs. Our results suggest that, for islands in volcanic hotspot archipelagos, this is best done by using direct measures of habitat diversity and resource availability rather than using island age as a proxy. We, therefore, recommend using direct measures, rather than proxies, when investigating the drivers of biodiversity patterns on islands.