The nearest neighbors (k-NN) method is a simple, easy to motivate procedure for supervised classification with functional data. We first consider a recent result by Cerou and Guyader (2006) which provides a sufi- cient condition to ensure the consistency of the k-NN method. We give some concrete examples in which such condition is fulfilled. Secondly, we show the results of a comparative study, performed via simulations and some real-data examples, involving the k-NN procedure (as a “benchmark choice”) together with other some recently proposed methods for functional classification.