Cutting through the rhetoric about subject headings

I’ve returned from a vacation to see discussion on AUTOCAT of the utility of precoordinated vs. postcoordinated subject strings. I’m not all the way through my email list messages yet, but this discussion has prompted me to finally put into words something I’ve been stewing on for a while.

It seems to me that a great many of our disagreements in the library realm have at their root people talking past each other, each side meaning something different by a given term or two, but not cognizant of that fact. I see LCSH as a prime example of this phenomenon. A great deal of debate occurs over whether precoordinated subject strings or postcoordinated subject strings are more useful. But I see a fundamental difference in the way various participants in these discussions define “postcoordinated.”

One definition is that postcoordinated headings have no subdivisions at all; in LCSH-speak, have no --. The other definition is that postcoordinated headings are “faceted” (to introduce another term that complicates the issue); that each heading reflects only one characteristic of the work, such as “topic,” “place,” or “date.” The difference here is the difference between “subdivisions” and “facets.” These two concepts are not identical. A common criticism of postcoordinated headings is that they would not represent the essential distinction between concepts like “History--Philosophy” and “Philosophy--History.” While (ignoring the syntax; whether or not the double dashes are used is a style issue) this would be true according to the first definition, it’s not necessarily true according to the second—a “topical” facet may very well represent a complex concept. I’ve never seen a discussion on this issue in which this distinction is made clear to both sides. It’s unclear to me whether the “traditional” definition of postcoordinate allows the faceted interpretation or requires the subdivision interpretation, but I think what’s needed here is clarification of current definitions rather than historical ones.

I don’t have all the answers in this debate, nor does anyone else at this point. My inclination is toward the postcoordinate side, although I do very much want to keep an open mind on the issue. I’d like to see a well-reasoned argument for a postcoordinate system presented according to the facet definition (something I’ve long been wanting to write but find this is one of the many issues that have trouble finding their way from my brain to a shareable form). I personally read arguments for precoordinate indexing and think to myself, “We can do all of this with postcoordinate headings if we had systems that operated reasonably.” (Big IF there, considering our current state of affairs!) We need to have more room to experiment with these options to see if my interpretation is a good one. The Endeca use of precoordinated strings shows powerful promise; we need more large-scale implementations of systems working off of postcoordinated data to allow us to compare both user functionality and cataloging time (a much-forgotten but essential factor) of the two approaches. I want data, darn it! We can only go so far with the philosophical argument; to get beyond our current roadblock we need to see what will happen if we follow the various paths available to us.