In publications on tropical forest fragmentation regrowing secondary forests(SF) on abandoned agricultural land, which are highly fragmented components inthe landscape matrix (generally less than 100 ha in size), are often overlooked.This forest type is found on privately owned land. These forest fragments arecontributing to on-farm income by logging and/or use of non-timber products.Hence, defining pure conservation goals for this forest type seems to beunrealistic. In this paper, we used the process-based model FORMIX3-Q tosimulate successional processes and logging scenarios in SF exposed todifferent degrees of fragmentation (either facing non-forest land on onlyone side or totally surrounded by agricultural land) in the subtropicaleastern part of Paraguay. We compared results with primary forest embeddedin a similar matrix. Under light fragmentation, bole volume of SF approachedto primary forest values after about 50 years of succession. Speciescomposition, however, was clearly distinct from primary forest over thefirst 200 years of succession. The development of bole volume and speciescomposition in severely fragmented SF was similar to the less fragmented SFover the initial 50 years. However, limited seed input, largely confined toon-site seed sources in the severely fragmented SF, led to a decline in bolevolume by about one-third compared to the reference value over the simulationperiod of 400 years. By applying a minimum felling diameter (MFD) of 35 cm inthe lightly fragmented SF, first logging was only possible after 30 years,harvesting a mere 3.7 m/ha. Highest timber yields were obtained with 10-yrcutting cycles and a maximum removal of 20 stems, though only providingsustainable yields by lowering the MFD to 30 cm. Logging in severelyfragmented SF accelerated the decline of bole volume. Overall, resultssuggest that timber production in SF without silvicultural treatmentsis not a viable option to diversify farm income. Only managed SF mayserve this function. Research focusing on the elaboration of managementconcepts for SF, integrating as much as possible of the originalvegetation, is needed.