The primary goal of this research was to determine whether crop damage from herbicides measured soon after treatment could estimate relative crop yield loss. Two to 4 wk after spraying soybeans with an unregisrered mixture of rhifensulfuron plus sethoxydim at various rates plus crop oil concentrate, percent stunting was visually rated and percent projected ground cover of soybeans was determined from photographs. In each of 3 yr, relative percent soybean yield was a negative linear function of relative herbicide rate from 0.25× to 2× the registered (1×) rate. The 1× rate of thifensulfuron and sethoxydim was 17.5 and 420 g ai ha−1, respectively. Relative soybean yield was also negatively related linearly to stunting but positively related linearly to soybean ground cover over 3 yr. Linear regression equation models of relative soybean yield versus percent soybean ground cover explained more model variability, more consistently (R 2 ≥ 0.60 in two of 3 yr) than did either stunting (R 2 ≥ 0.60 in one of 3 yr) or relative herbicide rate (R 2 = 0.37 to 0.48 over 3 yr). However, linear regression models for each independent variable differed from year to year and were related to differences in rainfall following treatment. Ways are suggested to increase regression model precision and between-year reproducibility.