Terminal restriction fragment length polymorphism (T-RFLP) is an increasingly widely used technique in mycorrhizal ecology. In this paper, we review the technique as it is used to identify species of mycorrhizal fungi and distinguish two different versions of the technique: peak-profile T-RFLP (the original version) and database T-RFLP. We define database T-RFLP as the use of T-RFLP to identify individual species within samples by comparison of unknown data with a database of known T-RFLP patterns. This application of T-RFLP avoids some of the pitfalls of peak-profile T-RFLP and allows T-RFLP to be applied to polyphyletic functional groups such as ectomycorrhizal fungi. The identification of species using database T-RFLP is subject to several sources of potential error, including (1) random erroneous matches of peaks to species, (2) shared T-RFLP profiles across species, and (3) multiple T-RFLP profiles within a species. A mathematical approximation of the risk of the first type of error as a function of experimental parameters is discussed. Although potentially less accurate than some other methods such as clone libraries, the high throughput of database T-RFLP permits much greater replication and may, therefore, be preferable for many ecological questions, particularly when combined with other techniques such as cloning.