Abstract: A sample survey is an important method for investigating plant resources. Plot area and shape underlie sampling surveys and influence survey efficiency and accuracy. To determine optimum plot area and shape for investigating yield of Trollius chinensis, three plot shapes (square, rectangle and circle) with a range of side lengths or radii were used to fit the CV (coefficient of variation)-area curve by the Generalized Mitscherlich equation, Richard equation and Logistic equation. Two methods were put forward for determining plot minimum area, one was based on CV change rate, the other on minimum investigating cost. The optimum plot shape was obtained by also comparing plot shape. The three CV-area models were all pretty good and the correlation coefficient was above 0.94, with Logistic equation the most stable. The minimum plot area estimated by change rate was in the order Richard equation>Logistic equation>Generalized Mitscherlich equation for the same plot shape, and was circle>square>rectangle for the same regression equation: the areas of squares and rectangles were similar. The study also showed that the minimum investigation cost was close for the same plot shape and different regression models, but the minimum plot area varied. For the same regression equation, the minimum plot area based on minimum plot cost was circle>square>rectangle while the square and the rectangle were also similar. Finally and comprehensively, the minimum plot area of square, rectangle, and circle were determined as 36 m² (6 m×6 m), 32 m² (8 m×4 m), 78.5 m² (radius 5 m) respectively. The optimum plot shape was obtained from the minimum CV of plots of different shapes, cost of the same area plots and comparisons among regression models. The rectangular plot was the best followed by the square with little difference between them. The circle was the worst shape.