Three different strains of the biotrophic endosymbiont, Neotyphodium, in perennial ryegrass were compared with endophyte-free ryegrass (Nil) for their effects on root herbivory, soil biota, root growth, root morphology and alkaloid and nutrient content in pot trials. Ryegrass infected with the endophyte strain, AR37, which produces janthitrem alkaloids, was toxic to a root aphid, Aploneura lentisci. Relative to Nil plants, ryegrass with Wild-type endophyte (producing the alkaloids peramine, ergovaline and lolitrem B) showed occasional resistance to A. lentisci whereas ryegrass with AR1 (producing peramine only) was highly susceptible. A high variability in root aphid populations among individual AR1 plants was related to plant genotype or a plant genotype/endophyte interaction. Differential effects of endophyte on A. lentisci were maintained under nutrient stress. Neither AR37 nor AR22 (a strain similar to AR1) in ryegrass affected feeding or survival of root-feeding larvae of the scarab Costelytra zealandica. Endophyte infection had no discernible adverse effects on populations of Collembola, oribatid mites or dorylaimid nematodes. Higher root aphid populations were associated with higher populations of Collembola and lower populations of nematodes. Rate of mycorrhizal colonisation of ryegrass infected with AR37 and Wild-type was slower than in AR1 and Nil but then proceeded rapidly to the extent that they became more heavily infected than AR1. After 2 years, infection of roots of AR1-infected plants was less than on AR37, Wild-type and Nil plants. Alkaloid content of roots was very low relative to that of leaf sheaths. Two janthitrem fractions were consistently found in roots of AR37 plants and may be the cause of toxicity to root aphid. Ergovaline and lolitrem B were only found in roots under certain conditions and adverse effects of Wild-type infection on A. lentisci could not be attributed conclusively to the presence of either alkaloid. Manipulation of herbivory by insecticide showed that root and foliar growth of AR1 and Nil plants were reduced by A. lentisci and also by an unidentified pseudococcid species infesting tillers. Root/shoot ratios were occasionally reduced in the presence of both species and in the presence of A. lentisci alone suggesting plant allocation to foliar growth in response to damage by these insects. Not all effects of endophyte infection were mediated by herbivory. AR1-infected ryegrass had a higher specific root length than Nil plants with AR37 and Wild-type intermediate between these. Insecticide treatment increased specific root length in all plants but relative differences between endophyte treatments remained the same. AR37 differed from other endophyte treatments in having a lower investment in root growth and increasing root nitrogen concentrations during summer followed by a relatively larger investment in root growth in autumn and early winter. Protection from herbivory meant that actual root biomass of AR37 plants was not less than in other treatments. Endophyte-infected plants had higher concentrations of potassium and phosphorus than Nil plants in roots but not in leaf blades. Nitrogen concentrations in roots were inversely related to root biomass resulting in AR1-infected plants having the highest concentrations of nitrogen due to the adverse effects of A. lentisci on root growth. The higher nutrient content of AR1-infected ryegrass compared with Nil may increase its susceptibility to A. lentisci. It is concluded that Neotyphodium infection has multiple but highly strain-specific effects on root ecology of perennial rye grass.