Darwin's model of evolution by natural selection was based on his observations of change in discrete organisms in which individuals are easy to define. Many of the most abundant functional groups in ecosystems, such as fungi and bacteria, do not fit this paradigm. In this review, we seek to understand how the elegant logic of Darwinian natural selection can be applied to distributed clonal organisms. The arbuscular mycorrhizal (AM) fungi are one such group. Globally, they are ubiquitous in terrestrial ecosystems, are locally distributed among many host plant species, and are significant drivers of nutrient cycling in ecosystems. The AM fungi are intractable to study, as the few taxa that can be cultured cannot be grown in the absence of plant roots. Research has focused on the plant-fungus interface, and thus on the symbiotic phenotype. A model is discussed for the interchange of materials at the interface that throws the emphasis of research onto the behaviour of the individual organisms and removes the need to test for phenomena such as selectivity, co-evolution, and cheating. The AM fungi are distributed organisms with an extensive external mycelium that is likely to be under strong environmental selection. AM fungi show sufficient phenotypic variation and fitness differentials for selection to occur, and developments in genetic analyses suggest that a better understanding of heritability in these organisms is not far away. It is argued that direct selection on fungal traits related to their survival and performance in the soil independent of the host is likely to be the major driver of differentiation in the AM fungi, and the evidence for direct fungal responses to soil conditions such as pH, hypoxia, and temperature is reviewed.