Most studies of seed survival over time are not repeated in different years because they are long-term efforts. Methodological research reported in this paper shows that the shape of seed survival curves can differ for the same species for seed produced in different years. Consequently, seed survival of wild mustard could not be described by a single type of seed survivorship curve. Wild mustard seed survival was a negative exponential function of time in one 4-yr-long trial [Y=a + b*exp(-X)], but decreased as a negative linear function of time in a second trial (Y=a + b*X), where Y is percent seed survival, X is time in years, and a, b, and c are coefficients. However, both data sets could be adequately described by a mixed negative exponential model [Y=a + b*X + c*exp(-X)]. The negative hyperbolic model [Y=a/(X + b)] did not describe both trials as well as did the mixed negative exponential model. Sample sizes of four groups of 100 seed each, harvested each year for four years, provided acceptable nonlinear regression models for seed survival data. It was not advantageous to increase sample size more than this because r2 was not greatly increased and the residual mean square error (RMSE) was not decreased.