Letting the (real or complex) Banach space $F$ be linearly homeomorphic to the (strong) dual of some Banach space, one sees that there is a closed topological linear subspace $E$ in $F'_\beta$ such that the map $\iota:y\mapsto{\rm ev}_y|E$ with ${\rm ev}_y|E(u)=u(y)$ becomes a linear homeomorphism $F\to E'_\beta$. Using Hahn−Banach, one easily sees that there cannot exist two different such subspaces with one a subspace of the other. However, I can't see why there might not exist two such subspaces with no one a subspace of the other. So I ask
Question. Is the space $E$ unique?
A counterexample, proof or reference will be appreciated. Note that making $E$ bigger tends to destroy surjectivity of $\iota$ whereas making $E$ smaller tends to destroy its injectivity. Related questions are here and here. I didn't quickly see whether a counterexample could be constructed from the ideas given there.


$\ell_\infty = \ell_1^*$corresponding to the various $C(K)$ spaces. – Bill Johnson Jan 17 2012 at 22:51