Open access and accuracy: author‐archived manuscripts vs. published articles

Some approaches to open access (OA) use authors' manuscript copies for the OA version, in the form accepted after peer review but prior to full editing. Advocates of such approaches are certain that these versions differ only trivially from the publishers' versions; many of those who oppose them are equally certain that there can be major discrepancies. In a pilot study, we have examined the actual differences in a small number of such article pairs in the social sciences and in biology. Using an operational classification of the extent of error, we have determined that neither pronouncement is likely to be correct. We found numerous small differences that affect readability; we also found a low frequency of potentially confusing errors, but sometimes it was the publisher's and sometimes the manuscript version that was more accurate. In two cases errors introduced by the publisher omit technical details that are necessary to evaluate the validity of the conclusions. However, we found no error that actually affected the validity of the data or results. Interestingly, we did find problems with the stability of the document locations on authors' sites, and, in some cases, with their disappearance from PubMed Central after initial placement there.