The transition from classical to quantum mechanics rests on the recognition

that the structure of information is not what we thought it was: there are

operational, i.e., phenomenal, probabilistic correlations that lie outside the

polytope of local correlations. Such correlations cannot be simulated with

classical resources, which generate classical correlations represented by the

points in a simplex, where the vertices of the simplex represent joint

deterministic states that are the common causes of the correlations. The `no

go' hidden variable theorems tell us that we can't shoe-horn correlations

outside the local polytope into a classical simplex by supposing that something

has been left out of the story. The replacement of the classical simplex by the

quantum convex set as the structure representing probabilistic correlations is

the analogue for quantum mechanics of the replacement of Newton's Euclidean

space and time by Minkowski spacetime in special relativity. The nonclassical

features of quantum mechanics, including the irreducible information loss on

measurement, are generic features of correlations that lie outside the local

correlation polytope. This paper is an elaboration of these ideas, and its

consequences for the measurement problem of quantum mechanics. A large part of

the difficulty is removed by seeing that the inconsistency in reconciling the

entangled state at the end of a quantum measurement process with the

definiteness of the macroscopic pointer reading and the definiteness of the

correlated value of the measured micro-observable is only apparent and depends

on a stipulation that is not required by the structure of the quantum

possibility space. Replacing this stipulation by an alternative consistent

stipulation resolves the problem.