Categoricity over P for first order T or categoricity for \(\phi\) \(\in {\mathcal L}_{\omega_ 1\omega}\) can stop at \(\aleph_ k\) while holding for \(\aleph_ 0,\dots ,\aleph_{k-1}\) (Q757348)

It is well-known that any countable first-order theory which is categorical in some uncountable power, is categorical in every such power. Several attempts were made to generalize this theorem to the case of non-classical logics as \({\mathcal L}_{\omega_ 1\omega}\), or some generalizations of the categoricity notion. One of those is relative categoricity. Let T be a complete first-order theory in a language containing the unary predicate P. T is relatively \(\lambda\)-categorical if whenever \({\mathfrak M}\), \({\mathfrak N}\) are models of T having identical relativisations to P of cardinality \(\lambda\), then there is an isomorphism of \({\mathfrak M}\) and \({\mathfrak N}\), being the identity on P. In the present paper the authors determine the limitation of the validity of the mentioned theorem: it does neither work for relative categoricity nor for the \({\mathcal L}_{\omega_ 1\omega}\) language. They give an example of a first-order theory T and an \({\mathcal L}_{\omega_ 1\omega}\) sentence \(\phi\) such that T is relatively \(\lambda\)-categorical and \(\phi\) is \(\lambda\)-categorical for \(\lambda =\aleph_ 0,...,\aleph_{k-1}\), but T is not relatively categorical and \(\phi\) is not categorical.