We report on the first results of experiments to measure the recombination rate of hydrogen on surfaces of astrophysical interest. Our measurements give lower values for the recombination efficiency (sticking probability S x probability of recombination upon H-H encounter $gamma$) than model-based estimates. We propose that our results can be reconciled with average estimates of the recombination rate (1/2 n(H) n(g) v(H)A S $gamma$) from astronomical observations, if the actual surface of an average grain is rougher, and its area bigger, than the one considered in models.