On empty islands and the small-island effect

AimThe small-island effect (SIE) has become a widely accepted part of the theoretical framework of island biogeography. A major criticism of SIE studies is the exclusion of empty islands from analyses. However, the generality and underlying factors determining the role of empty islands in generating SIEs remain obscure because few published datasets include islands with no species. The aim of this study was thus to evaluate the prevalence and underlying factors determining the role of empty islands in generating SIEs. LocationGlobal. MethodsWe compiled 278 datasets that included empty islands. For each dataset, we compared the fit of a logarithmic model with two breakpoint models separately for all the islands and for datasets excluding empty islands to determine the role of empty islands in generating SIEs. We then employed multinomial logistic regressions and an information-theoretic approach to determine which combination of island characteristics was important in determining the role of empty islands in generating SIEs. ResultsAmong 211 datasets with adequate fits, the exclusion of empty islands changed the evidence for an SIE in 68 cases (32.2%). SIEs were quite prevalent, both for all the islands (104 cases, 49.3%) and for datasets excluding empty islands (73 cases, 34.6%). Our results were not consistent with the hypothesis that excluding empty islands would increase the evidence for an SIE. Model selection and relative variable importance indicated that the number of empty islands, the minimum area of empty islands and area ratio were important variables that determined the role of empty islands in generating SIEs. Main conclusionsOur study demonstrates that the effect of empty islands in generating SIEs is quite prevalent. The exclusion of empty islands is thus an important methodological shortcoming for the detection of SIEs. We conclude that, for the robust detection of SIEs, empty islands should not be excluded in future studies.