I agree with you Max, but still think there is a large possibility that those claiming dream sharing will be lumped into a category with high school kids who pretend to have supernatural powers after watching the new 'Twilight' movie. We would have to test pairs of individuals, who had never met, in a lab. The individuals would be given pictures of the other, also maybe soundless video of the other interacting with people and vocal recording. This would rule out any possibility that the two had colluded to deceive the scientists and ruin the empirical data by sending some sort of message to the other, while giving them some way to recognize the other.

The dreams themselves only have to be the same in places. If the dream lasted 20 minutes, but in the midst of things the two sat down and shared pudding then viola, the two shared a moments dreaming. This would have to be tested farther than a longer and stronger dream share, because it will invariably be conceived as coincidence by the scientific community until the results were reproduced regularly. However, if the the details of the dreams were the same in more ways than they were different then this could not be chalked up to coincidence in my opinion. How many coincidences have to occur for someone to stop labeling them as coincidence?

For instance, If the two woke up in some place both claim never to have been watching the sunrise together on a park bench (but describe SLIGHTLY different scenes) when one offers the other pudding before the dreams begin to differ; Could one say this was a coincidence? Could one chalk it up to some genius level collusion to defraud the scientific process?