Terry Eastwood does articulate a comprehensive archival theory. However, Eastwood’s theory also has a flaw, namely his rejection of selection (appraisal and reappraisal, weeding, deaccessioning). Eastwood’s own concept of “uniqueness,” in contrast to that of Roberts, defines archives as documentation of relationships, and as such, even a record duplicated, tripled, etc. in one collection deserves preservation to manifest each instance, since each gives the record a different context, and thus a larger importance in the fonds as a whole. Eastwood’s definition is problematic for archival theory, and this paper will demonstrate that selection, far from being an invasive action, is a necessary component of archives, and that Eastwood’s uniqueness concept contradicts the other principles in his theory.
Eastwood’s archival theory identifies five characteristics of archives: impartiality (capturing the creator’ biases), authenticity, naturalness, interrelatedness, and uniqueness (Eastwood, 1992). Uniqueness, however, negates the other qualities. Eastwood states:
Each document has a unique place in the structure of an archives. Copies of the document may exist in the same archives or in others. Each one is unique in its place. Being there signifies its relationship to activity and to the other documents accumulated in the course of that activity. So every archival document, whether existing in more then copy or not, is unique. (Ibid, 128)
Rather this fifth quality should be termed redundancy. One can demonstrate the flaw in Eastwood’s model by imagining the natural extreme of his logic—for example, an archives constructed from one original document which is photocopied. There are now two documents, both enjoying a unique relationship with the other. The photocopy is photocopied; now there are at least three unique relationships (actually nine total). Continue to photocopy each new photocopy in succession. Would we have an archives? According to Eastwood’s theory, yes—it would be the impartial, authentic, natural, and interrelated evidence of an action (such as it is), with each document enjoying a unique relationship to all others! Yet our “archives” would also contribute no more information than the original document. Weeding by the creator would produce no more impartiality than not weeding; even disrupting original order would make little difference. Eastwood’s theory is thus incomplete in that relies upon human intuition to define an archives, and therefore Eastwood has not adequately answered the claim by Roberts that archives are a practice that one learns “on the job” (Roberts, 1990, 111), utilizing training and intuition, needing no theory. What is missing here? Selection.
Eastwood could still argue that pruning admittedly redundant information still erases a unique relationship within the archives that the creator included, thus removing a sort of information. However, given the backlog crisis in archives, how likely is it that creators prune their own collections to preserve multiple duplicates? Rather, Eastwood may have succumbed, and would consequently lead researchers, to a fallacious conclusion via the “file-drawer effect”—the creation of a relationship where one does not exist, a false correlation arising from mere coincidence.
Archival redundancy could lead a researcher to go on what is known in research as a “fishing expedition” (Hoofnagle, 2000). That is, the researcher, having found what he or she thinks is a significant outcome (the same document keeps showing up in different places—it must have all that much more importance!), concludes a meaningful finding and gathers further evidence of that, when in fact there will be, in any collection, a certain amount of false positives due to mere chance. This is articulated in John P.A. Ionanidis’s essay, “Why most published research findings are false,”  and reiterated by the experience of the editors of the Public Library of Science (PLoS):
[I]f you study anything, anything with statistics you’re going to find statistically significant correlations. In fact if you found nothing that would be strange as you would predict that if you study enough variables, roughly 5% should be statistically significant just by chance, and the nature of the scientific literature and the “file-drawer effect” is that this number goes up as researchers are more likely to publish big effects and “file away” the papers showing no effect even though the results are true (Hoofnagle, 2009).
Hoofnagle’s piece refers especially to false statistical correlation, but there is no reason to deny that if we study anything, anything with relationships (and statistics reveal relations) we would find false positives, qualitative correlations that Eastwood terms uniqueness. It is not true that only proximity preserves relationships in an archives. Each record enjoys an intellectual relationship with every other record in the archives, and therefore, any documents repeated in the archives bear their most significant relationships to each other. This relationship mirrors the “uniqueness” of that archives of photocopies in our thought-experiment, and is mere redundancy which should, and must, be purged to honor Eastwood’s theory.
Until we have more data to contradict these conclusions, our thought-experiment of photocopies suggests that selection, not uniqueness, is necessary to archives. Far from detracting from the authenticity, naturalness, and interrelatedness, careful pruning of a collection enhances these qualities. An archives is selection incarnate, because like an ecosystem it is collaborative and organic.
 However, this is a reductio ad absurdum and thus a playful thought-experiment rather than an argument: no archivist would really draw such a conclusion from Eastwood’s theory, and it is not the author’s intent to suggest that Eastwood demands it.
 Available at http://medicine.plosjournals.org/