According to the book Programming Clojure refs manage coordinated, synchronous changes to shared state and atoms manage uncoordinated, synchronous changes to shared state.

If I understood correctly "coordinated" implies multiple changes are encapsulated as one atomic operation. If that is the case then it seems to me that coordination only requires using a dosync call.

For example what is the difference between:

(def i (atom 0))
(def j (atom 0))

(dosync
  (swap! i inc)
  (swap! j dec))

and:

(def i (ref 0))
(def j (ref 0))

(dosync
  (alter i inc)
  (alter j dec))