“…Thus: A recurrent theme in studies of the relationships between classes of models of fragments (in one sense or another) of arithmetic has been cofinality of embeddings, it having been found that cofinality is often effective in either lifting properties from below or pulling them down from above. Theorem 4.1 of [4], the proof of which is rather routine, entails that any recursive ultrapower can be embedded properly, cofinally, and in A~ fashion in some other recursive ultrapower. In the A ~ case, the A~ of the embedding is automatic by virtue of the "DPRM" (or, as [1] prefers, "MRDP") theorem on diophantine representation of S ~ predicates; in the general case of A o ultrapowers, n__> 1, the A~ character of the embedding asserted in [4, Theorem 4.1] can be replaced by the stronger condition of n-elementariness, on account of a result of Dimitracopoulos and Gaifman [1 ] (Lemma 2.1 below).…”