Model systems can facilitate and focus research efforts but ill-chosen or inapt ones can distract or impede scientific progress. In this Opinion article, we pose the question: how can the literature provide appropriate general conclusions if the model systems upon which the literature is based are unrepresentative of the relevant biological diversity? A good example of this problem is the endophyte-grass symbiosis, which is considered to be a classic example of mutualistic interactions. Meta-analysis of the primary literature demonstrates that the conceptual framework for endophyte-grass interactions has largely been based on endophyte-plant-herbivore studies of two agricultural grass species, tall fescue and perennial ryegrass. Consistent with conventional wisdom, the meta-analysis indicates that endophytes slightly increase grass resistance to herbivores. By contrast, endophytes appear not to affect plant performance or competitive ability. The positive effects of endophytes appear to be dependent on genetic variation in the host and endophyte, and on nutrient availability in soils. Thus, the agronomic grass model systems fail to capture the breadth of variability inherent in wild grass-endophyte populations and communities.