It is commonly assumed that the observation of an interference pattern is incompatible with any information about the path taken by a quantum particle. Here we show that this assumption is contradicted by theoretically predicted experimental observations. A precise evaluation of the effects of sufficiently small polarization rotations in the slits of a double slit experiment indicates that the particles forming the interference pattern must have been physically delocalized, such that particles detected in the interference maxima have a presence of exactly one half in each slit while particles detected close to the minima appear to have a negative presence in one of the slits and a presence of more than one in the other.