How repeatable a process is evolution? Comparative studies of multicellular eukaryotes and experimental studies with unicellular prokaryotes document the repeated evolution of adaptive phenotypes during similar adaptive radiations, suggesting that the outcome of adaptive radiation is broadly reproducible. The goal of this study was to test this hypothesis by using phenotypic traits to infer the genetic basis of adaptation to simple carbon–limited environments in an extensive adaptive radiation. We used a clone of the bacterium Pseudomonas fluorescens to found two sets of experimental lines. The first set of lines was allowed to adapt to one of 23 novel environments for 1100 generations while the second set of lines was allowed to accumulate mutations by drift for 2000 generations. All lines were then assayed in the 95 environments provided by Biolog microplates to determine the phenotypic consequences of selection and drift. Replicate selection lines propagated in a common environment evolved similar adaptive components of their phenotype but showed extensive variation in non–adaptive phenotypic traits. This variation in non–adaptive phenotypic traits primarily resulted from the ascendance of different beneficial mutations in different lines. We argue that these results reconcile experimental and comparative approaches to studying adaptation by demonstrating that the convergent phenotypic evolution that occurs during adaptive radiation may be associated with radically different sets of beneficial mutations.