The effects of fire on artificial secondary forests of Pinus massoniana with an average diameter at breast height (DBH) from 5.5958 cm to 19.3900 cm, average height from 4.11 m to 18.60 m, and average density from 500 clumps hm^-2 to 2400 clumps hm^-2, were investigated by comparison of burnt and non-burnt plots in a karst mountain area covered by Quaternary clay in central Guizhou province, China. The percentage damage to P. massoniana trees differed among plant organs and followed the order bark (95.51%) > branches (71.49%) > crown (62.95%). The percentage plant death in different forest layers was highest in the herbaceous layer (100%), intermediate in the arbuscular layer (30.43%) and lowest in the arboreous layer (29.09%). With increasing DBH of P. massoniana in the arboreous layer, the percentage damage to bark was unchanged, whereas damage to branches and the crown decreased, and the percentage death of plants also decreased. At P. massoniana DBH<5 cm, the percentage damage to bark, branches and the crown, percentage plant death and direct loss of biomass were all 100%, whereas at DBH≥25 cm the percentage damage to bark, branches and the crown were 85.00%, 25.00% and 15.00%, respectively, and percentage plant death and direct loss of biomass were zero. In the arbuscular layer, either all individuals of a plant species died, such as P. massoniana, Betula alnoides and Camellia oleifera, all plants survived, such as Prunus serrulata, Myrsine africana and Rhamnus leptophylla, or only some plants died, such as Cunninghamia lanceolata, Castanea seguinii and Aralia chinensis. Plant species in the herbaceous layer either died or germinated after death of aboveground branches and leaves because of fire. In burnt forest, the potential biomass loss (68.7755 t/hm²) was higher than direct biomass loss (12.1818 t/hm²; percentage direct loss 22.41%). Direct biomass loss differed among the layers and followed the order arboreous layer (6.9382 t/hm²) > litter layer (3.3441 t/hm²) > arbuscular layer (2.4964 t/hm²) > herbaceous layer (0.8861 t/hm²); in terms of the corresponding percentage direct biomass loss, the order was herbaceous layer (or litter layer, 100.00%) > arbuscular layer (33.36%) > arboreous layer (23.59%). Decreases in the Patrick, Gleason, Simpson and Hurlbert indices followed the order herbaceous layer > arbuscular layer > arboreous layer, and reflected potential losses were higher than direct losses of plant diversity in the different layers of burnt forests. The range and average values of direct diversity loss in the arboreous layer indicated by the Patrick, Gleason, Simpson and Hurlbert indices were 0-1 and 0.1429, 0-0.1669 and 0.0238, 0-1.0000 and 0.1586, and -0.1098-0 and -0.0166, respectively; the corresponding range and average values of percentage direct biodiversity loss were 0-100 and 14.29, 0-100 and 14.29, 0.66-100 and 17.85, and -18.61--3.98 and -11.29, respectively. The range and average values of direct loss of diversity in the arbuscular layer as measured by these indices were 0-5 and 1.8571, 0-0.6948 and 0.4301, 0-0.2987 and 0.1115, and -0.3477-0 and -0.1241, respectively; the corresponding range and average values of percentage direct biodiversity loss were 0-42.86 and 26.76, 0-42.85 and 26.76, 0-86.35 and 37.63, and -52.00-0 and -18.53, respectively. The range and average values of direct diversity loss in the herbaceous layer indicated by these indices were 2-6 and 3.8000, 1.8205-5.4614 and 3.4589, 0.2593-0.8313 and 0.5300, and 0.1728-0.7619 and -0.4863; the percentage direct loss was 100.00% for all of the indices. The average burn height of the arboreous layer was negatively correlated to DBH, positively correlated to density and percentage damage to branches, the crown and entire plants, and uncorrelated with percentage damage to bark. The arboreous layer showed higher burn heights than those of the arbuscular layer, and their correlation was negative, and was negatively correlated to potential or direct loss of biomass and positively correlated to percentage direct loss of biomass. Burn height was positively correlated to direct loss or percentage direct loss of biomass, and negatively correlated to potential loss of biomass in the arbuscular layer. Direct loss or percentage direct loss of biomass with the Patrick and Gleason indices was not obviously correlated to burn height in the arboreous layer and was positively correlated to burn height in the arbuscular layer. Burn height was positively correlated to direct loss, and negatively correlated to percentage direct loss, of biomass with the Simpson index in the arboreous layer, and positively correlated to direct loss and percentage direct loss of biomass with the Simpson index in the arbuscular layer. No obvious correlation between burn height of the arboreous or arbuscular layer and direct loss or percentage direct loss of biomass with the Hurlbert index was detected.