If language processing and development is viewed as a dynamic process in which all subsystems interact over time, then some basic assumptions behind more traditional approaches to language analysis are problematic: new methods of analysis and modeling are needed to supplement and partly replace existing paradigms. This argument is illustrated with two examples from recent studies. After a brief history of the reasons for a paradigm shift and an explanation of the role of variability in development, the first example study presents a variability-based approach to reaction time measurements in which spectral analyses of variability found during repeated measures of the same experiment may indicate moments of behavioral change. Then the principles of dynamic modeling are explained, illustrated with vocabulary developmental data. The second recent study shows how the vocabulary development of three learners may be dynamically modeled using a logistic model.