Many creationists insist that the New Testament, in particular, is based on eyewitness accounts. There doesn’t seem to be any particular compelling evidence to support this claim, but it is a very common belief.
The problem is that eyewitness accounts are very unreliable. It is the common assumption by most people that eyewitness accounts are the single most reliable type of evidence. That is not the case at all.
Legal experts have known this for a long time. Here’s a news report:
Andrew Roberts, a lecturer in law at Leeds University specialising in evidence, said courts have recognised for a long time that eyewitness identification evidence is "inherently unreliable".
That same news report gives a number of examples.
One of the problems is that people’s ideas about what happened changes over time. Here is an interesting documented account of how these accounts vary.
Within one day of the space shuttle Challenger explosion, Ulric Neiser, a psychologist studying "flashbulb" memories (the recall of highly dramatic events), asked his class of 106 students to write down exactly how they'd heard about the explosion, where they were, what they had been doing, and how they felt. Two and a half years later they were again interviewed. Twenty-five percent of the students subsequent accounts were strikingly different than their original journal entried. More than half the people had lesser degrees of error, and less than half the people had all the details correct. (Prior to seeing their original journals, most students presumed that their memories were correct.)
So if the Gospels really are eyewitness accounts recorded well after the fact, as most creationists believe, then it appears that only about 10% of what we read can be expected to be accurate.
In fact, there are some variations in what we read in the Bible.
For example, who found the empty tomb on the morning of the first Easter? How many people found it?
According to Matthew 28:1, the first visitors to the tomb were Mary Magdalene and the other Mary. That is a total of two people.
According to Mark 16:1, both of the Marys were joined by Salome. Therefore by this account there were three people who found the empty tomb.
But, according to Luke 23:55 (as well as 24:1, 24:10), Mary Magdalene, Joanna, Mary the mother of James, and "other women" found the empty tomb. That means that there were at least five people.
Then we have John 20:1 who says that only one person, Mary Magdalene was there.
There are some other relatively minor variations such as whether things happened exactly at dawn.
Biblical apologists try to reconcile these accounts. In general they say that the maximum number of people was present (that would be at least five) and the other gospel stories simply neglected to mention the others.
By making such arguments these apologists actually argue against the integrity of the Gospel accounts!
Note that these accounts all agree on the major events: Jesus was crucified, buried and arose from the dead. The differences in the accounts vary only in the relatively minor details.
Police officers are familiar with eyewitness accounts. They listen to them all of the time. Imagine that you were to ask a police detective this hypothetical question:
There are two crime scenes.
Each crime scene, quite coincidentally, has four witnesses named Matthew, Mark, Luke and John.
At one scene, the four witnesses agree on every single detail. There is not a single contradiction.
At the second crime scene, the four witnesses agree on all of the major events, but have contradictions and inconsistencies in regard to the relatively minor events.
Then ask the detective, which set of witnesses is probably more reliable?
I can pretty much guarantee that the detective will select the set of witnesses who had minor disagreements. In the case of the witnesses who agreed on every detail it is more likely that there was only one witness and the others are only providing a copied account. Alternatively, and more menacingly, the identical accounts may indicate a conspiracy of some kind. In that case, the events may not have happened at all!
So it is actually the case that there are times where inconsistencies increase, rather than necessarily decrease, the reliability of accounts of an event.
 http://news.bbc.co.uk/1/hi/uk/4177082.stm, referenced on February 2, 2009
 Robert A. Burton, "ON BEING CERTAIN: Believing You Are Right Even When You're Not", pp. 10-11.