This note discusses the contradiction between β and σ- convergence tests of agricultural per capita GDP, which was recently presented by Spohr and Freitas (RESR, vol. 49, No. 02, p. 341-368, April / June 2011). While the first type of test supports the hypothesis of convergence, the σ- convergence test does not validate this hypothesis. It is argued that this contradiction probably arises from nonlinearities in the convergence process, which can characterize the formation of convergence clubs. The convergence test was done again in a more rigorous way, according to the methodology proposed by Caselli, Eesquivel and Lefort (1996) and, subsequently, the formation of convergence clubs is tested using the methodology proposed in Phillips and Sul (2007), which has the advantage of accommodating various types of heterogeneity. The formation of three convergence clubs is confirmed and these results try to clarify the contradiction previously reported.