Through an editing listserv I was directed to this posting about journal editing and repositories. Rather interesting, especially the editing statistics. In my experience, no errors should be ever be "introduced," as (a) queries are directed to the author regarding changes in context or data (if the CE is even halfway decent), and (b) the author reads over a proof set prior to final publication, thus any printed changes are "approved." Further, most articles are peer reviewed and assessed by the journal editor, so there should be little need for substantial editing. Of course, I have never worked for Blackwell, so I cannot speak for them. But I doubt their freelancers and production process are (or were) terribly different from most.

I also smiled at the statistic that about one-third of editing was "minor syntactical or grammatical changes." As I have a hunch that the label "minor" is subjective, I would guess that any CE would consider the grammar changes they make to be necessary (e.g., I must change "data is" to "data are" twice a day; that might be minor to some, but it's major to me and if retained would make the author look careless at very best).

Does this mean that another version is not appropriate for a repository? That's not for me to say, perhaps. But I'd like to think that one would want to put forth the best (and often more correct) version wherever it is placed. (Especially after references and citations are corrected. It is a research article, after all, for others to use for their own research. Why put a version out there with four citations without references and two references without citations??) I'll always say that's the version after the experienced journal CE got to it.

Although not scholarly, I will never forget the one journal article that insisted that the Yankees had not won the World Series since the 1970s. Whew ... this Yankee fan had a good time making that correction!