in other geographical and ecological settings. Despite this reservation, the southeastern U.S. populations represent a crude series of experiments that trace to introductions within the last several hundred to one thousand years, and the ability to place temporal limits on the history of these populations is a strength of the morning glory program. Common patterns among populations are indicative of common initial conditions or common selective forces. Thus, the lack of spatial autocorrelation for the W/w locus phenotypes over populations is strongly indicative of a disadvantage that is independent of location. Similarly, the case for an isolation-by-distance model of population structure for the P/p locus is strongly supported by the consistency of this result over local populations.

A third consideration is the problem of statistical power. Because the structure of statistical hypothesis testing is deliberately biased against accepting the alternative hypothesis (power is usually much less than the size of the critical region, α, under the null hypothesis), very large experiments must be carried out to detect moderately large effects. Put another way, magnitudes of selection that can be quite effective from an evolutionary standpoint cannot be detected in experiments of reasonable size. This may explain the failure to detect pollen discounting and the failure to detect significant inbreeding depression in experimental populations of I. purpurea. Limited statistical power can in part be overcome by experiments that run over many generations in which the effects are cumulative. This is the strength of the geographical and population studies in the southeastern U.S., which represents the cumulative outcome of many generations rather than the result of single generation experiments.

Despite the limitations noted above, the study of model evolutionary systems is just as important as the study of other model systems. Model systems should reveal the important questions and provide a basis for inventing novel approaches that can then be extended to more refractory systems. As one example, Stebbins pioneered the analysis of plant development as an essential prerequisite to understanding the determination of phenotype and hence to the understanding of adaptation. This approach is just as important today as it was more than 40 years ago, when Stebbins began his classic work with barley awn development as his model system. Today we have a rich suite of molecular tools that assist in unraveling the determination of phenotype and in penetrating the complexities associated with the coordinated action of many genes. The determination of floral color development is an area of special promise because the translation between genes and phenotype is tractable. Similarly, the translation between environment and phenotype is more transparent for flower color than in most other cases, and there is still much to be learned from this research strategy.