Ordination and classification techniques were used to analyze patterns of forest vegetation, species diversity and soil type in the Wichita Mountains Wildlife Refuge. Cluster analysis based on species composition of the tree size class produced 3 general community types: (1) stands codominated by Quercus stellata and Q. marilandica, (2) stands dominated by Q. stellata and (3) mesophytic forests. A polar-reciprocal averaging ordination (PO-RA) produced a gradient of vegetation that corresponded to a moisture gradient. Many of the high diversity forests were located on loamy drainageway soils or north facing slopes. Tree species diversity (H') was inversely related to the importance of Q. stellata. Cluster analysis based on species composition of the seedling size class produced 4 general seedling community types: (1) stands dominated by Q. marilandica, (2) stands codominated by Q. marilandica, Q. stellata and Juniperus virginiana, (3) high diversity stands dominated by Ulmus americana, Celtis reticulata and Bumelia lanuginosa and (4) stands dominated by Acer saccharum. The third seedling type occurred almost exclusively on loamy drainageway soils. There was no relationship between stand location on the first axis of the tree ordination and the first axis of the seedling ordination, suggesting a differential response to the moisture gradient.

Of forty three forest stands in the Wichita Mountains Wildlife Refuge that had been surveyed in 1961, twenty seven stands had the same relative species composition and nine had only slight changes twenty years later. Seven stands, mostly mesic, differed with respect to minor species. The stands that showed the least changes were the least diverse. On the whole, Refuge forests were less diverse in 1981. Quercus stellata, the most important tree in Refuge forests, ranged in age from 32 to 270 years with most individuals being between 60 and 80 years old. These ages coincide with the establishment of the Refuge and probably indicate the conversion of savannahs to forests. Size is a poor predictor of age for Q. stellata trees in Refuge forests, especially for small trees. Twelve stands had nonsignificant regressions of basal area on age. Size differences in these forests must result from differential growth rates.