Most of our research for the last decade has followed a linear model as
shown in Fig. 1, that tests species, interactions and prototypes for agroforestry
technologies first on-station, then later evaluates and refines the
"winners" on-farm [Shaner et al., 1982]. We often ask the "basic" questions
on-station and the "practical" questions on-farm, though it might make more
sense to address both types of questions in each place.