Native hardwood trees are frequently planted to restore forest cover to abandoned agricultural lands in the Neotropics. Few studies compare rainfall interception in such plantations and the vegetation the trees are planted to replace. We compared throughfall (TF) and stemflow (SF) in a mature stand of wild sugar cane Saccharum spontaneum L. and six tropical hardwoods growing in a restoration trial in the Republic of Panama. Interception in trees was dominated by TF, with SF accounting for less than 3% of gross rainfall (Pg). By contrast, SF was a dominant means by which rainfall reached the ground in S. spontaneum, and this dominance became greater at higher Pg. Slopes of Pg/SF regressions were considerably steeper in S. spontaneum than among tree plots. We tentatively concluded that trees intercept more rainfall than S. spontaneum. However, median total I was negative in S. spontaneum when SF was estimated using the median and the median plus maximum absolute deviation of stem densities as a multiplier. Sources of potential bias in S. spontaneum measurements include overestimation of stem density, high inter-event variability of water volumes in individual SF collectors, and contact between stems diverting SFs. To our knowledge, ours is the first study to compare I in restoration plantings of tropical hardwoods and in S. spontaneum. It is also only the second study to quantify TF and SF in a sugar cane, likely due to the methodological difficulties involved in retrieving reliable estimates of these variables in very dense grass stands. Copyright © 2012 John Wiley & Sons, Ltd.