Experimental design.

Plants and AMF were grown in microcosms with plant roots divided between two sides with a central plastic barrier. Mesh membranes (21μm pores) allowed AMF growth throughout each side of the microcosm. Plant roots were restricted to the innermost of the three compartments per side. The plant carbon (C) pool was labeled with ¹⁴C, and soil phosphorus (P) pools in the outermost compartments were labeled with either ³²P or ³³P. ¹⁴C transferred from plants to AMF was measured in the outer compartments of the microcosm. Radioactive P (*P) taken up by AMF from outer compartments and transferred to plants was measured in plant shoots. The space between inner and exterior compartments served as a buffer and prevented direct root uptake of applied *P from these outer compartments and direct root exudation of ¹⁴C into these outer compartments. Each plant species interacted with a pool of 3 AMF species: there were three choice microcosms with two AMF species and three no-choice microcosms with a single AMF species. Microcosms with 2 AMF species were replicated twice (with reciprocal ³²P/³³P labeling), resulting in 9 microcosms per host plant species.