In 1962 Jan Mycielski proposed a very general notion of interpretability [M1]. This led to the question whether a given theory could be interpreted in the disjoint union of two theories, without being interpretable in any of them. He argued that in such a case it would be presumably simpler to study each of these theories separately, and hence conjectured that this situation can never occur for any of the well-known theories of mathematics. This conjecture has now been verified for the following theories (see [MPS], [P], [S1, 2]): ELO (endless, i.e., without maximal element, linear order), Th(〈ℚ, ≤〉), Th(〈ω, ≤〉) and all sequential theories (those which can code finite sequences of elements of their models). The latter include PA, ZF, GB and Th(〈ω,+,·〉). In view of these confirmations it became ever more plausible that the conjecture is valid also for RCF (real closed fields), i.e., for Th(〈ℝ,≤,+,·,0,1〉). In the present paper we show that Mycielski's conjecture is valid for a class of theories which includes RCF and OF (ordered fields).

We consider only theories with equality and without function symbols. Interpretations will be meant local, multidimensional, and with parameters, as defined in [M1], [M2] and surveyed in [MPS] (for a recent definition see also [S2]). We shall write T0 ≪ T1 to say that T0 is interpretable in T1 (or that T1 interprets T0), and this will mean that for every theorem α of T0 there is an interpretation of α in T1.