Soil ecosystem function and plant diversity is receiving much attention in karst ecological studies. Changes and correlation laws in plant diversity indices and soil physical and chemical indicators of 41 karst forest plots were analyzed on different functional areas in the Maolan National Nature Reserve of southern Guizhou Province, China. This research adopted the combination method of wild field plot investigation and laboratory analysis. Results show:(1) The importance values of the arbor species clustering method resulted in five karst forest types:Koelreuteria minor-Cyclobalanopsis glauca forest, Lindera communis-Liquidambar formosana forest, Toona sinensis-Lindera communis forest, Cornus controversa-Cornus parviflora forest and Loropetalum chinense-Pinus massoniana forest. The arbor plant diversity indices, bare rock coverage, soil water storage indicators, soil fertility indicators and nutrient indicators showed a decreasing trend from the core area, to the buffer area, to the experimental area and to the outer area. The Margalef index, Simpson index, Shannon-Wiener index and Pielou index of the arbor and herb layer species showed a significant difference between some forest types. However, the diversity indices differences of the shrub layer were not significant. Significant differences of some indicators between the different forest types confirmed the influence of human disturbance to the karst forest in the natural recovery process. (2) Factor analysis revealed different correlation trends between the plant diversity factors of different layers, and the physical and chemical factors of different soil horizons. The analysis conformed to the general regularity of plant growth in the development of the arbor, shrub and herb layer and the co-evolution between the vegetation and soil system of karst forest in the natural recovery process. The number rate of significant factor-pairs of the soil physical and chemical indicators related to the arbor, shrub and herb plant diversity indices were 36.38%, 27.27% and 18.19%, respectively. The arbor plant diversity factor had the strongest relationship with the soil physical and chemical indicators. (3) The significant correlations between the arbor plant diversity indices and the soil physical and chemical indicators were divided into three types:i) Linear correlation is a general relationship between the plant diversity indices and soil physical and chemical indicators; 39.84% of the indicator-pairs were in this category, where the plant diversity indices increased to raise or lower the soil physical and chemical indicators accordingly. ii) Curve correlation is a complex relationship between the plant diversity indices and soil physical and chemical indicators; 46.10% of indicator-pairs are accounted for in this way. Inflection point values of the plant diversity indices without ecological meaning accounted for 11.72% where the relationship could be explained using straight lines or other models. The ecological inflection points of the plant diversity indices, which periodically impacted on soil physical and chemical indicators between the upwards and downwards trends of the quadratic curves, accounted for 17.19% of the indicator-pairs. The values of these plant diversity indices can be adjusted on the basis of the soil management goal in forest planting. iii) No significant correlation means the plant diversity indices are independent of the soil physical and chemical indicators, and planting measures for plant diversity indices cannot effectively change the soil physical and chemical indicators; 54.69% of indicator-pairs corresponded to this model. The different response modes aid the interpretation of ecosystem function through the plant diversity hypothesis. (4) Inflection points are the top values of the upwards and downwards trends in the quadratic curve response of arbor plant diversity indices to soil physical and chemical indicators. They are in index values of Margalef for 0.71-2.70 and 4.80-9.70, Simpson for 0.57-0.66 and 0.72-0.86, Pielou for 0.50-0.76 and 0.83-0.95 and Shannon-Wiener for 1.35-1.43 and 2.57-3.47, respectively. These values are used for one of the selection references of species and number proportions in artificial forestation.