Field tracer experiments in Borden aquifer in Canada and the Macrodispersion Experiment (MADE) Site, Mississippi, are reviewed. Both experiments injected tracers in aquifers and monitored their movements over fields at ten to hundred meters under natural gradient conditions. The behaviors of the tracer plumes at these two sites are distinctly different because the Borden aquifer is statistically homogeneous, and the aquifer of MADE is statistically heterogeneous. As a result, the validity of the classical ADE and non-Fickian dual-domain models becomes a contentious debate and deserves articulation of the differences between the ensemble mean nature of the models and the observations in one realization. The two experiments provided opportunities for understanding the limitations of applying solute transport theories and mathematical models based on soil-column experiments to real-world scenarios where heterogeneity is multi-scales, and groundwater flow varies spatiotemporally. Ignorance of the differences in scale of dominant heterogeneity and the observation, model, and interest scales is to blame. We explore and discuss the strengths and weaknesses of the theories and models.