One of the good things about being a public amateur, negotiating my way through American history, is that I've been fortunate enough to attract the attention of people who do this for a living. The result is that I constantly get e-mails, and comments after posts, suggesting other lines of inquiry or clarifying my own. I am not always as gracious as I would like to be.
I wanted to make a comment from my seat as a historian on this:
4.) Primary documents are useful, but like all sources, limited. They are helpful and useful, but not God's word.
The way "professional" history is practiced, this is the nub of the question, not really a "settled" issue at all. Indeed, there's a way in which the practice of history is precisely that -- figuring out how to use primary sources to make conclusions. In that way, the [believability] of your explanation about how and why you use a particular source is also the credibility of your argument.
To use your specific example, the slave narratives, professional historians would immediately tag them as "oral history" documents and then note, among other things:
1) All oral histories are "memory" -- retrospective accounts, in contrast with documents like wills or memos that were produced when events were actually taking place. Over time, people forget and revise, as when many Holocaust survivors began to retell their stories to match up with the narrative of Schindler's List. It is not dishonesty -- it just makes oral histories a particular type of document, which we need to note has been produced AFTER the event.
In the case of the slave narratives, we would also want to do the math -- many of the subjects who were still around to be interviewed in the 20th century were very young when they were slaves. How "accurate" are your memories of childhood? How "accurate" will they be when you are 80? What assumptions do kids make that adults don't, and vice versa?
2) All oral histories are "co-produced" -- by the interviewer and interviewee. As your commenter notes, this means that we have to account for differences in background and power between interviewer and interviewee. At a recent conference, I talked to an oral historian who is employed by a big investment firm to record the stories of the CEO, founder, and other employees. When we hear the oral histories he produces, won't we want to note that he was paid by his interviewee?
This is particularly fraught for black-white interviews. In the 1930s and 1940s, when opinion polling was first used, polling firms like Gallup spent a lot of time figuring out how best to get "honest" answers from black interview subjects. Some early pollsters found that it helped to send black interviewers to talk to black subjects; and, during World War II, government pollsters found that it helped to tell the black interviewees that the government was running the poll. (In the Roosevelt years, black Americans placed a lot of trust in the government, certainly trusting it more than many other institutions).
As these conversations will continue on this blog