Interesting concept nicely executed. Only flaw was the inclusion of too many identical stories of the oft-debunked "I had a dream/premonition thing X would happen, and it did, so obviously there must be something supernatural going on!" form. I did a quick calculation, with the following figures: 10 million american soldiers participating in the war, each away for 3 years (~1000 nights), each with one worrying relative who has a 1% chance each night of seeing a dream in which the soldier is kileld, and each soldier with a 10% chance of dying in the war, distributed uniformly over the 1000 days. Using those statistics, the probability that not one of those 10 million relatives would see a dream and then have it come true the same day was 0%. I say 0%, because that's what my calculator said. When I ran the calculations with super-precision on my computer, the real answer is 0.0 followed by a couple of thousand zeroes, and then some real numbers. To put this into some perspective, using the above numbers, the most likely number of deaths preceded the previous night by a dream "predicting" it is 10000. Choose the 5-10 most startling from those, put them in a book, and stupid people are inclined to say "no way could that happen naturally".