As for selective memory, assume that 1 in 10,000 dreams is a dream about death and disaster and that people only remember a dream each second night. Then any randomly chosen person will have a dream about death and disaster very roughly every 60 years, that is about once per life. For the sake of simplicity assume that there are 200,000,000 US-Americans at an age in wich they can report dreams. Then 10,000 US-Americans must have dreamt about death and disaster in the night before September 11th by chance alone.

You can bet they all tell their friends about the 'precognitive' dream for the rest of their lifes. They and all persons they tell about are unlikely to ever forget that dream. The many more dreams of the same character that did not come true are usually not told ("listen, I had a dream about death and nothing happened") and if yes, they are forgotten quickly.

Would you have guessed that there are about 10,000 dreams about disasters each night in the USA even if any US-American only dreams such a dream once per life? I think you wouldn't have even been close to that figure.

That was one of the reasons why the more than 100,000 case reports about premonitions and precognitive dreams that have been collected and are filed worldwide are not taken as a serious proof: It is close to impossible to make a good guess what the chance baseline is.

Another reason is retrospective fitting. The more often the stories are told the closer the story tends to fit a premonition hypothesis. This is difficult to assess except by in depth studies of single cases. In many of those cases in which this could be done, blatant errors of recollection were abundant. A death was recollected as have occured at night to fit the time of the dream when in fact it was at daytime and so on. Vital facts tend to be forgotten if they distract from the supernatural hypothesis.

Robin's story reminds me of an in depth case study in which a mother had dreamt of her son's car accident when in fact he had a car accident the same day. Closer scrutiny showed that her son had promised to call the day of his accident which he couldn't due to the accident. No wonder, she was worried and dreamt about an accident.

Another reason is the unconscious use of weak cues (Jeri has mentioned that already). For example, a woman had dreamt about a disaster after it had happened but before she thought she had heard of it. In depth analysis could show that two men had been talking about that disaster the evening before her dream within earshot. She might have heard them (difficult to tell after the fact) without the message reaching her consciousness.

As for Sharon's (pre)monitions, the academic parapsychologists are using these terms:
Precognition if your 'feeling' comes before the fact (there are abundant case reports about that as well)
Clairvoyance if your 'feeling' comes concomitant to the fact and your feeling refers to the fact and not to another person's thoughts or feelings.
Telepathy if your 'feeling' is about thoughts and/or feelings of another person.

Wolfgang