The overall shape of an adoption curve depends on the particular mechanism that describes why individuals adopt. This paper examines the implications of four distinct models of adoption behavior: inertia, imitation, hearsay, and learning. It is shown that learning models lead to a novel class of adoption curves that have not been studied before except in special cases. A distinctive feature of learning models is that they frequently yield super-exponential growth rates in the early phases of adoption. This fact that can be used to distinguish statistically between learning models and those generated by alternative mechanisms.