This study conducts an investigation on the application of classical unit-root tests using parametric tests (the augmented Dickey-Fuller, 1979 – ADF), and nonparametric tests (Phillips and Perron, 1988—PP) to corn and soybean yields in the Delta states using county-level data from 1961 to 2009. The main concern of the paper is to assess what would be drawn about nonstationarity in crop yields using these tests versus using modified versions of these tests (Ng and Perron, 2001) that are assumed to solve size and power problems associated with the ADF and PP tests. The investigation focuses on methodological aspects of the classical tests, uncovers the nature of filtered yields often needed prior to density estimation, sheds light on the effect of lag truncation and provides guidance for future work. It is found that the complexity in yield behavior is such that at very small samples (40 observations or less), results can lead to ambiguous findings in larger samples (49 observations). This sample size analysis with county-level yield data uncovers that negative moving average effects are present and help explain certain ambiguous findings. As a by-product, this paper contains a condensed review of literature on unit-roots that may prove useful in applied research.