More seeds were removed from Colosso (8.8 ± 0.4; n = 48 stations) than the other two ranches (Porto Alegre, 5.4 ± 0.7, n = 45 stations; and Dimona, 5.0 ± 0.8, n = 42 stations), indicating an interaction between forest size and site (Table 2).

F-statistics are summarized in Table 2 .

The percentage of seeds found did not differ significantly by forest size, site or month (1-ha fragments, 38 ± 7%, n = 18; 10-ha fragments, 55 ± 4%, n = 35; continuous forest, 52 ± 4%, n = 31; Colosso, 54 ± 3%, n = 42; Porto Alegre, 54 ± 7%, n = 24; Dimona, 38 ± 8%, n = 15; April, 51 ± 6%, n = 27; June, 52 ± 5%, n = 27; September, 50 ± 6%, n = 27; Table 2).

The percentages of seeds hoarded, predated by a rodent and with another fate were similar between the three sites (Table 2)..

In 10-ha fragments, a high percentage of seeds were hoarded in April (90 ± 4%; n = 9 stations), with a decrease in June (44 ± 10%; n = 9 stations), and no further change in September (56 ± 11%; n = 9 stations; Table 2).

F-statistics are summarized in Table 2 .