Generic delineation within the Arctium group ( Compositae, Carduceae-Carduinae ), formed by the genera Arctium, Cousinia, Hypacanthium and Schamalhausenia , has proven a complicated task. In particular, the precise limits between Arctium and Cousinia are very difficult to establish. Therefore, we have carried out a molecular survey of DNA sequences of two regions, the chloroplast gene mat K and the nuclear-ribosomal spacers ITS 1 and 2, of a representation of all the genera of the group (in the case of Cousinia , centered in the species more obviously related to Arctiium ). Our results show a precise correlation between molecular phylogeny and two very important characters, pollen type and chromosome numbers: all the investigated species with the Arctiastrum pollen type and x= 18, characteristics of Arctium sensu stricto, form a monophyletic clade, sister of another monophyletic clade formed by all the investigated species of Cousinia sensu slricto. However, the resulting Arctioid clade cannot be defined on macroscopic morphologic characters, because the main trait for segregating Arctium and Cousinia , the spiny pinnatifid-pinnatisect leaves of Cousinia , is adaptative and of scarce systematic relevance. In fact, our results suggest that spines have appeared at least in two different lineages: the genera Hypacanthium and Schamalhausenia , spiny and thus morphologically closer to Cousinia , are unambiguously related to the unarmed genus Arctium . An hypothesis on the evolution of morphology, pollen and chromosome numbers in the group is formulated. The systematic implications of this incongruence between molecular, pollen and karyology, on the one hand, and morphology, on the other hand, are evaluated. Some possible solutions are proposed, but none of them is totally satisfactory: more studies are necessary with the inclusion of new species of Cousinia subg. Hypacanthodes .