The simplest way to explain the biochemical basis of the gene-for-gene concept is by direct interaction between a pathogen-derived avirulence (Avr) gene product and a receptor protein, which is encoded by the matching resistance (R) gene of the host plant. The number of R genes for which the matching Avr gene has been cloned is increasing. The number of host-pathogen relationships, however, for which a direct interaction between R and Avr gene products could be proven is still very limited. This observation suggests that in various host-pathogen relationships no physical interaction between R and Avr proteins occurs, and that perception of AVR proteins by their matching R gene products is indirect. Indirect perception implies that at least a third component is required. The 'Guard hypothesis' proposes that this third component could be the virulence target of an AVR protein. Binding of the AVR protein to its virulence target is perceived by the matching R protein, which is 'guarding' the virulence target. An intriguing aspect of the 'Guard hypothesis' is that the Avr gene product causes avirulence of the pathogen through interaction with its virulence target in the plant. This would mean that, although AVR proteins are generally thought to be bifunctional (avirulence as well as virulence factors), this dual function might be based on a single biochemical event. This review focuses on the way AVR proteins are perceived by their matching R gene products. The various components that determine the outcome of the interaction will be discussed, with an emphasis on the dual function of AVR proteins.