Maintenance of variation in mutualism by screening

Models of partner choice leading to mutualism raise a conceptual problem: directional selection for high‐quality partners should ultimately erode variation in partner quality. How do we explain the persistence of variation in partner quality observed in nature? The problem arises in all models of partner choice, including screening models, in which a host induces potential symbionts of different quality to screen themselves by assigning them different costs and rewards. Using a screening model in which costs and rewards are sometimes assigned incorrectly, I show that a stable polymorphism can arise because rewards are higher when partners vary in quality than when there is only one type of partner. Partner quality, therefore, undergoes negative frequency‐dependent selection even though there is a preference for high‐quality partners. This also shows that partner choice by screening does not need to be totally accurate to be effective—inaccuracies enable both effective screening and the maintenance of variation.