After decades of theories built upon painfully abstract argumentation, here is an approach to language that fits how we actually use it--and with how we learn, master, and play with much of what we do in life (vision, music, work, sports, etc.). Chomskian structuralist linguistics treats language as if it were some mysterious coding machine, completely unlike other cognitive, purpose-driven human activities. In go concepts and noises; deep inside the black machine, its gears and rules and constraints turn and whir; then out come the utterances of English or Asmat or Vietnamese (depends on which "machine" we have). And once we get our machines, we're stuck with them--they don't evolve. Each machine/language has its own set of rules, and only elite linguists understand them.
Like cognitive linguistics, Bybee's usage-based approach treats our use of language as very much like our uses of other skills. As with other skills, fluency comes through frequent practice. Many other parts of the UB approach also seem obvious or "common sense," and Bybee excels at making their explanatory force clear. This approach is still new, and still developing, but it looks like it has a LOT of potential for future research--not just in phonology, but also in morphology, syntax, and in grammaticalization (and other studies of diachronic language change).