We use household level panel data to look at the diffusion of seed technology among Ugandan farmers. We present a simple target-input model to conceptualize the adoption decisions and management of a new technology in which the best use of inputs under the new technology is unknown and stochastic. In this framework, there is path dependency in the adoption process, since use and adoption are important indicators of the superior efficiency of new technology. Our analysis suggests that the adoption of certified improved seed remains low in Uganda, despite the presence of at least two distinct legal systems regulating seed sales and plant breeding i) TRIPS 27(3)b and ii) national laws such as the Seed and Plant Varieties Regulations. Notably, we find evidence of unsuccessful experimentation patterns by the group of most likely adopters. We focus on the characteristics of the group who undertakes most of this experimentation, and find that this group of innovators is well-defined but too small to overcome the inertia within the system. In short, we find little evidence that seed policy reforms implemented in Uganda in the past 20 years have boosted agricultural productivity in the country, largely on account of inadequate use and diffusion of new seed varieties within the agricultural system.