Forest litter has great potential for reclaiming degraded lands, being a natural seed source for herbaceous, shrub and forest species. Using litter for reclamation revegetation is a new technique and needs to be properly studied. One problem is lack of knowledge about seasonal variation in litter fall - not knowing when litter should be harvested and applied on degraded lands for best field results. The objective of this study is to provide such information about seasonal variation when harvesting forest litter seed banks from a Semideciduous Submountain Seasonal Forest fragment. The specific litter source was the "Mata da Garagem", forest fragment at the Universidade Federal de Viçosa campus, located in Municipality of Viçosa, (42o53’W, 20o45’S) southeastern Minas Gerais State, Brazil. Three sites representing different stages of secondary succession were selected within this fragment, and litter including a small portion of topsoil was collected to a depth of 5 cm from each site. This material was lifted from within a square wooden frame (0.25 x 0.25 m = 0.0625m2) placed on the forest floor at randomly located points. The collected material was immediately transported in labeled plastic sacks to the campus experimental nursery and deposited in wooden boxes (50 x 50 x 15 cm). Six repetitions were distributed within three treatments: 1) greenhouse, 2) field (nursery grounds close to greenhouse) and 3) field plus protection (same as treatment 2, but with screen protection against contamination by ambient propagules in the nursery). The following plants were identified and measured: 682 tree seedlings comprising 24 species and 17 families; 2262 herbaceous plants comprising 98 species and 35 families. As hypothesized, seasonal effects were observed for both tree and herbaceous species germination. Results indicate end of rainy season (with moderately abrupt transition to dry season in this region of Brazil) as optimal moment for maximizing number of forest species and forest species richness. Seasonal effects were also observed for herbaceous species, and results suggest that litter harvest and application during rainy season will also maximize number of species and species richness. Site 3 provided the greatest number of individual tree seedlings. There was no difference among the sites for the herbaceous species. From the 2946 individuals in the seed bank 76.78% were herbaceous species and 23.22% were forest species. As expected, the greenhouse treatment produced more tree and herbaceous individuals than either of the two field treatments. The field and field plus protection treatments presented low germination, especially for the forest species. This was probably due to hydric stress. The forest species embaúba and trema presented the highest number of germinated individuals. The harvesting technique proved efficient in providing considerable variety of pioneer, initial secondary, late secondary and climax species for potential field application in reclamation projects.