I felt I had to add this...
Several people do not believe DID exists. Why? Several use it as a legal defense to achieve a not guilty by reason of insanity plea. Others use it as a excuse ("I didn't steal your money, my other personality did!").
Some data shows that if the "alters" (the other personalities other than the main one), are ignored, they tend to appear less. Other data shows that in psychiatry units, if the patient is held responsible for breaking the rules, even if it's done by an "alter", the behavior goes down.
As for DID, several try to gauge it using dissociation assessment tools. Several criticize this because while we all agree that dissociation likely is real (hey, I dissociate everytime I get into an argument with my wife), it's hard to scientifically peg this down as a phenomenon.
Here's a recent article from someone arguing against the existence of DID
http://www.currentpsychiatry.com/article_pages.asp?AID=7830&UID=13981
Me? Personally I think it may exist. I have encountered people who believed they had it, but an attending of mine that I highly respect told me he believed the person was made to believe she had it by her previous therapist, and because she enjoyed playing the victim role, she played the part. She may have even believed she truly had it but did not.
He likened it to the phenomenon of using hypnotism to bring forth submerged memories. This practice, it turned out, was not scientifically valid. Several therapists up until the 90s (I'm sure there are others who still do it) got their patients to believe they were molested, kidnapped, what have you based on the results of using hypnotism. People were even convicted based on the testimony of mental health professionals claiming that this was a valid method--which it was not.
Yet despite the data proving that hypnotism to bring forth submerged memories was declared not reliable, several of the same people clung onto the beliefs they formed after the hypnosis.
Every time I encounter a DID patient, and it's extremely rare, I still feel like I'm walking on unestablished territory. I did consider that on a research level, we should start performing specific types of psychometric testing such as possibly using an f-MRI during an interview, incorporation of a SIRS among other tests to see how reliable this diagnosis truly is. Problem is this is such a rare phenomenon, it's hard to do this type of study.