In order to scientifically guide the management of middle-aged black locust (Robinia pseudoacacia) plantations established in the early stage of the Grain for Green Program on the Loess Plateau, based on the understanding of the effects of stand density and spatial structure parameters on the diversity of understory shrubs and herbs, a study was conducted in the Guanshan Forest Farm, Jingchuan County, Gansu Province. 12 sample plots of middle-aged (18—25 years) black locust plantations with four density gradients were selected, all growing on similar terraced field on the top of residual plateau. Spatial structure parameters, such as neighborhood pattern, mingling, crowding, openness, and dominance, were investigated, along with the diversity indices of understory shrubs and herbs, including Simpson's index, Gini diversity index, Shannon-Wiener index, Pielou's evenness index, and Patrick's index. Person's correlation analysis and grey relational analysis were employed to explore the relationship between stand density, spatial structure parameters, and the plant diversity of the undergrowth shrub layer and herb layer in middle-aged black locust plantations, and to identify the main influencing factors. The results revealed the followings: (1) The total number of understory plant species firstly increased and then decreased with rising stand density, with the peaks of the total number of understory plant species and the diversity index of the herbaceous layer within the density range of 1300—1700 trees/hm2. (2) Stand density mainly affected the light transmittance, distribution pattern, and mixing degree of black locust plantations. When the density was less than 1300 trees/hm2, the light transmittance was optimal. Within the density range of 1300—1700 trees/hm2, 57% of the black locust trees exhibited a random distribution pattern, closest to a natural state. When the density exceeded 2100 trees/hm2, the mixing degree reached its highest level. (3) Person's correlation analysis showed that a very significant positive correlation (P<0.01) existed between the openness and the Simpson's index and species richness index of shrub layer; and a significant positive correlation (P<0.05) existed between the mixing degree and the species richness index and Gini index of herbaceous layer. The Grey relational analysis indicated that when the stand density was <1300, 1300—1700, and >1700 trees/hm2, the main factors influencing the diversity of understory shrubs and herbs were mixing degree, openness, and DBH (diameter at breast height) dominance, respectively. (4) For middle-aged black locust plantations established on the site conditions studied, a density range of 1300—1700 trees/hm2 is recommended as a rational range for managing the understory plant diversity. The findings of this study can be used to optimize the stand density and spatial structure of middle-aged black locust plantations to promote the development of the understory shrub and herb layers for enhancing the understory plant diversity and the multifunctionality of forest.