Species accumulation in small-large vs large-small order: more species but not all species?

Although groups of small habitat patches often support more species than large patches of equal total area, their biodiversity value remains controversial. An important line of evidence in this debate compares species accumulation curves, where patches are ordered from small-large and large-small (aka 'SLOSS analysis'). However, this method counts species equally and is unable to distinguish patch size dependence in species' occupancies. Moreover, because of the species-area relationship, richness differences typically only contribute to accumulation in small-large order, maximizing the probability of adding species in this direction. Using a null model to control for this, I tested 202 published datasets from archipelagos, habitat islands and fragments for patch size dependence in species accumulation and compared conclusions regarding relative species accumulation with SLOSS analysis. Relative to null model expectations, species accumulation was on average 2.7% higher in large-small than small-large order. The effect was strongest in archipelagos (5%), intermediate for fragments (1.5%) and smallest for habitat islands (1.1%). There was no difference in effect size among taxonomic groups, but each shared this same trend. Results suggest most meta-communities include species that either prefer, or depend upon, larger habitat patches. Relative to SLOSS analysis, null models found lower frequency of greater small-patch importance for species representation (e.g., for fragments: 69 vs 16% respectively) and increased frequency for large patches (fragments: 3 vs 25%). I suggest SLOSS analysis provides unreliable inference on species accumulation and the outcome largely depends on island species-area relationships, not the relative diversity value of small vs large patches.