[I would like to express appreciation to Lynn El Hoshy, Cataloging Policy and Support Office, Library of Congress, for her assistance in interpreting some of the data.]
I used a list of LCSH form subdivisions that is still in the process of being created. I determined which subdivisions would be considered form, then determined whether the item represented by that record was an example of the form or was about the form. If the term was a form term, the item was an example of the form, and the term was already at the end of the subject string, I coded it $v. The following summarizes the records that would be problems if the above were performed by a computer and thus could not make the judgment of about vs is:
2 headings would be wrong because a form subdivision at the end of the string would represent "about" rather than "is." With 95% confidence, this is 0.2% of all headings + or - 0.3%. It represents 0.7% of all headings with form subdivisions. The fact that form subdivisions seem so seldom to represent "about" (0.2% of all headings) is the reason that the Subject Analysis Committee has said that if retrospective conversion were done, the small inconsistency of having a few subdivisions coded $v when the item was really about the form could be lived with. From the time of implementation of the code, of course, all records would have the fact of "aboutness" or "isness" identified by the cataloger. One additional heading in the sample has form terms used in the sense of "about" but is followed by $x History and criticism-a sure clue that the form is being discussed in, not represented by, the item in question.
18 headings on 12 records would have form subdivisions left as $x's when they should be $v's because there are 2 form subdivisions at the end of the same string (e.g., Agriculture $z Indiana $x Statistics $v Periodicals [Statistics is also form]). At 95% confidence this is 1.8% of all headings + or - 0.8%. Many of these could be handled in an automated fashion. The following combinations occur in the sample:
6 headings - $x Dictionaries $x Chinese [or other language name]12 headings on 9 records have form headings that are now obsolete LC practice and so would not match a current list of form terms. Since these exist in most catalogs, they really should be corrected before an x to v conversion program (e.g., European Economic Community $x Collected works [LC doesn't use "Collected works" any more]). At 95% confidence this is 1.1% of all headings + or - 0.7%. A list of these could be created so that they also could be worked into an algorithm.
5 headings - $x Statistics $x Periodicals
3 headings - $x Bibliography $x Microform catalogs
2 headings - $x Biography $x Early works to 1800
1 heading - $x Formulae, receipts, prescriptions $x Early works to 1800
1 heading - $x Instruction and study $x Juvenile
These headings in this combination would virtually always signal that the penultimate $x is form. So a list of combinations could be worked into the algorithm.
A situation that did not appear in this sample, but that is known to exist is one in which a heading would have form subdivisions left coded $x that really should be $v because they are followed by a $z (e.g., $x Catalogs and collections $z United States). Such subdivisions are identified in LCSH as being able to be subdivided geographically and so could be identifed as such in a list for an algorithm to use.
So the total of the problems is 32 headings. The remaining 950 headings would be o.k. At 95% confidence the problems would constitute 3.2% of all headings + or - 1.1%. But since it appears that all but the "aboutness" problem could be solved automatically, only 0.2% would remain.
In response to several suggestions that field 655 be used instead of creating $v:
In my sample 59 of the 162 records with form subdivisions (36.4%) did not use the same form subdivision in all headings on those records. A total of 124 headings on those 59 records either did not have a form subdivision or had a different form subdivision from other headings on the same record (12.6% of all headings). For example:
The above four headings appear on one record. That is, the book is both a biography and description & travel. But it is not a biography of Australia or New Zealand, nor is it description and travel of Authors or Humorists. How would the 655 be made to recognize this distinction? In another example the following three headings appear on the same book:
The book is both poetry and a translation, but not a translation of Rome's Civil War or of the Battle of Pharsalus. If 655 were to be used in this case, the whole first heading should be in 655, because that heading is there because the book *is* Latin epic poetry that has been translated into German. So would the whole heading be moved to 655? Would it be placed in one 655 or in two 655s? If in one, how would the second part be coded? As a $x, or would it need a $v? If placed in two 655s, there would be the problem of "Translations into ..." not being appropriate with the headings for Rome and the Battle.
Here's another observation: At least 84 of the field 650 $a headings in my sample are form terms [see explanation of "at least" below]. Of these at least 39 (46.4%) are used because the item being cataloged is about that form. [In comparison the 3 subdivisions that mean the item is about the form is 1.1% of the 273 headings having form subdivisions.] Should the 53.6% of the field 650 $a form terms that mean the item is that form be moved to 655? If so, what will it take to move them to 655? Many of them are qualified by $x, $y, and/or $z subfields. Would the whole heading be moved? Some of the $x subfields qualifying these headings are also form subdivisions, but not all (e.g., $a Short stories, American $x Women authors).
Compared to these problems, the form subdivisions that represent "aboutness" seem relatively easy to deal with.
["At least" is used in relation to the 84 form main headings because this count was done on a quick pass-through, not with the more rigorous methods used in counting the form subdivisions.]
Back to Form Data Follow-Up page