The genus Trisetum Pers. (Poaceae, Poeae, Aveninae) comprises about 70 species of mainly perennial grasses, inhabiting cold and temperate zones in both hemispheres or the mountains in tropical areas. The variability of the diagnostic characters and the overlapping of their ranges of variation make the delimitation of the species extremely difficult and often highly subjective. Multivariate methods are appropriate to establish the species boundaries in Trisetum, since most of the characters commonly used are quantitative. A data matrix comprising 247 specimens (OTUs) and 39 vegetative and reproductive characters (variables) was analyzed with principal component analysis (PCA), principal coordinates analysis (PCoA), cluster analysis (CA) and discriminant analysis (DA), to examine the morphological variation and to establish the discriminant characters in American species of Trisetum sect. Trisetaera Asch. & Graebn. The analysis of the data matrix including the 247 OTUs by means of CA resolved three main clusters which were analyzed independently using ordination techniques (PCA, PCoA) and discriminant analysis (DA). The first cluster includes South American species allied to T. barbinode. The second cluster includes   species related to T. oreophilum ranging from Mexico to northern Chile and Argentina. The   third cluster includes species related to T. spicatum. A total of 13 species, eight subspecies and two varieties were recognized. A new subspecies of T. oreophilum from Colombia is described, and new combinations (T. johnstonii and subsp. mattheii, T. oreophilum subsp. barbatipaleum and subsp. rosei) and new synonyms are proposed.