The purpose of this study is to develop a conceptual model for a collocating index that will provide a quick and easy one-source access to the Civil War photograph collection. These items are currently housed in different areas, and indexed with entirely different systems - some are card-based while others are electronic databases. There is simply not an effective method of determining the size or comparing the contents of these holdings at the Ohio Historical Society.

As a result of the literature review (Chapter III), the author's personal experience as an OHS employee researching photographic reference requests over the past seven months, and invaluable input from Jennifer Songster, the OHS Audio-Visual Curator, a three-step process was mapped out. Step one was to chart reference requests from records of the Ohio Historical Society. Secondly, a model of bibliographic records for each item was created based on the information gleaned from step one, and from examining other graphic bibliographic models, such as the Library of Congress records for graphic materials. Finally, once these records had been fully researched and fine-tuned, they were loaded into a full-text indexing program, Zyindex, the system already in place in this facility.

First, a total of 768 reference requests was analyzed and charted. In this process, Civil War requests were tallied against the balance of all audio-visual reference requests (see Appendix A). This study covered a twenty-two month period from July 1995 through May of 1997. It was determined that the Civil War requests totaled 23.8% of all reference requests, or 183 total requests over this period (see Chart I). Estimates set the OHS photograph collection total at around 1,000,000 images. While the real number of images in the Civil War Collection is not known precisely, estimates place it somewhere in the 3,000 to 4,500 image range. It is quite apparent that a collection that represents less than .5% of the overall collection, but garners almost 24% (see Chart I) of the reference requests, should be given very special attention in regard to its indexing and cataloging.

All reference requests were also broken down by subject. It is necessary to understand how photographs are requested in a given institution, in order to determine which facets to incorporate into each record in order to most effectively match each request. Generally, it was found that in the Civil War requests, indeed all reference requests, were very specific. The most common request (see Chart II) was for photographs of specific soldiers, which accounted for 38.8% of the Civil War category, or 70 requests. Further break-down of the specific soldier category indicates that 35 requests were for officers, 22 were for soldiers with no rank specified, and 13 were for enlisted men. The second most common request was by fighting unit, which accounted for 30.6% of the Civil War requests, or 56 requests. Examples of these requests include:

• anything from the 53rd Infantry Regiment
• photographs of anyone in Company C, 6th Ohio Cavalry
• the regimental flag of the 110th Regiment

The categories of specific soldiers and specific fighting units combined represents 69% of all Civil War related reference requests. The balance of the Civil War requests also contained some very specific requests, such as for images of specific abolitionists (11), Camp Chase (3), John Brown (3), Frederick Douglas (2), Johnson's Island (2), Camp Denison (1), and so on. In all, only 11 requests, or 6%, could be considered generic or general requests. The vast majority, 94%, were very specific. For a complete list of all non-Civil War reference requests, see Appendix Aa.

The effectiveness of a photographic index is dependant completely upon the bibliographic record. It is quite simply the key to success or failure. The process of developing the information fields, or access points, for the bibliographical record model most relevant to this collection, was the principle focus of this project. Other essential record elements such as the lay-out and sequencing of fields presented additional challenges. At all points, primary consideration was given to the intended format, the Web index. Patrons searching for specific images expect fast and likewise specific results. The amount of detail and level of specificity represented in any bibliographic record must match as closely as possible the level of both in each patron request.

First, field identifiers (attributes) and contents (attribute values) had to be selected. Models considered included records used in several graphic databases and guidelines for Internet documents, most notably:

• University of Michigan Digital Library Project Art Image Database
• International Standard Bibliographic Description (ISBD) record format
• Dublin Core Metadata Element Set
• Bancroft Library's Japanese-American Evacuation and Resettlement Photograph Collection
• Library of Congress Prints and Photographs Division bibliographic records

While some elements from all these models were incorporated, the Library of Congress documentation offers the best model for the OHS Collection. The LC collection is more similar to the OHS collection because it also has a large Civil War photograph collection, and as such has had to confront many of the same indexing challenges. However, in order to most efficiently represent OHS holdings, the records developed in this study (see Appendix C for model records) were modified in many ways from the LC Model. It was generally found, in the process of constructing these records, that "internal consistency and logic are much more compelling exigencies" (Benemann, 1994), than a strict adherence to Library of Congress practice. A field by field analysis follows which will reveal the similarities and departures from LC records (see Appendix C for Sample Records).

The first departure is the obvious difference in scope of the two databases. LC records have to be interfiled with records of many other formats, and this condition necessitates a higher level of specificity in subject headings, along with broader field tags. For example, the first field that appears in the LC record is labeled the author field, a term that is generalized to represent authors of books, photographers, painters, composers, lithographers, and so on. The OHS database only represents items that are created by either photographers, graphic artists, illustrators, or printers. The field term creator, the person or persons responsible for the DNA of the work, was a better fit for this field in this context, so it was chosen over the Library of Congress term, author. In terms of field sequencing, the author field was selected to be initial field in this model, as it is in the LC records. This was a difficult choice, in light of the fact that the vast majority of searchers search by subject, not by the person or persons responsibility for the creation of the work. The temptation was great to place the TITLE field first, as it is undoubtedly the most useful for culling images in graphic records. The practice of listing the CREATOR field first however, is in line with one of the most essential cataloguing principles stated in the Paris Principles of 1961, that of recognizing the primacy of the personal author.

The title field presents some special problems in the case of graphic records. The vast majority of the graphic images comprise this collection have no supplied title, that is, a title printed on the item. Additionally, few offer accompanying documentation. This necessitates the construction of a title by the cataloger. Consequently, a field named TITLE slightly misleading, but as other names proved even more misleading, the field name TITLE was retained. The Library of Congress catalogers generally include a line in the "notes" that reads, "title derived". The derived title is inclosed in square brackets, while a title from the item is not. Consequently, the "title" field is a combination of several elements: identifying the persons, units, locations, or battles depicted, and often a description of how they are arranged within the frame. For example, the derived title of a portrait of U.S. Grant is constructed the following way in an LC record:

[Ulysses S. Grant, full-length portrait, facing left, standing alongside his war horse, "Cincinnati"]

This field generally requires more research than any other, because the most important subjects simply have to be identified and verified. To identify the most important subjects for the anticipated users of an index, the reference request tables need to be analyzed. In the OHS index it is clear from studying the nature of past reference requests, that our research efforts should be concentrated on identifying all people and fighting units depicted in the in our Civil War images. Secondarily, the geographic locations and other physical details could be identified.

In this project one major issue was how to treat hand-written captions either on the face or the verso of the item. The majority of the photographs in the OHS collection come from private donors, so one would hope that the hand-written documentation found on the item reflects the primary-source perspective that provides the necessary authority for sound documentation. The problem often is that there is no way of determining just who wrote the captions, or when. Once again, the indexer needs to research the available sources (see Appendix D for a list of Civil War Ohio reference books) to verify the information contained on the item. In reality, this hand-written text is often the only clue to the identity of the persons or events depicted in the photograph, and it should be regarded as the point-of-departure for further research.

Although this field label is a little vague in terms of whether it represents a description of the contents of the image, or a physical description of the extent of the image, the field contents can quickly dispel the ambiguity. Quite simply, this field describes the physical details of the image. It corresponds to area 5 of the ISBD record-the extent of the item, and the standard physical description terms are well-represented in the Thesaurus for Graphics Materials II, the Genre and Physical Characteristic Terms. This format guide is simply the best available for uniform language to describe the various formats contained in a graphic collection. It is very detailed, offers syndetic references, cataloger and public notes, and was conceived with the stated purpose of effecting a terms match between the cataloger and the researcher. Ultimately, the OHS Web index would include a hyper-text link to the online version of the Thesaurus for Graphics Materials, to provide the researcher with both a subject and genre thesaurus of terms.

This field also contains data on dimensions of the item, constructed as specified in Anglo-American Cataloguing Rules, second edition, 1988 Revision, rule 8.5D. Essentially, the dimensions are listed in centimeters, height first, followed by the width (see Appendix C for record details). This field serves as an aid to the identification of a specific or known item, while also offering the researcher an idea of the level of image resolution, based on its size and format characteristics.

Another major departure from LC was the level of specificity of the subject headings used. Once again, all LC subject strings have to be interfiled with other formats, other wars, other topics, even works about other countries. The OHS index presents surrogates for items of a very narrow scope, that is, images of Ohio's part in the Civil War. An identical image would be represented quite differently in the LC record when compared with the OHS record. Consider the example of an image of unidentified Ohio soldier - in this example an enlisted man. It would be recorded in the Library of Congress record in this fashion:

United States -- History -- Civil War, 1861-1865 -- Soldiers -- Union -- Ohio.

Soldiers -- Ohio.

It was determined that the inclusion of the Library of Congress Name Authority Files (NAF) for all proper names would be very useful in an index that will be often queried for proper names. These files, now available online, provide the necessary level of detail for Ohio's long list of historical figures that figured prominently in the Civil War. These established headings include detail on the fullness of name (initials are spelled out), and birth and death dates (see Appendix C). Often the lesser known soldiers are not represented in the NAF, but whenever possible, it is beneficial for the indexer to construct their names in a similar fashion, that is, to provide a full name including birth and death dates. This can be accomplished with the help of the Official Roster of the Soldiers in the War of the Rebellion, and with any regimental history that may be available on the fighting unit that the subject was associated with. Because the Ohio Historical Society is the repository for materials on Ohio history, the collection provides tremendous reference capabilities. It would be difficult to imagine trying to accomplish the level of research necessary to represent this graphic collection in any other institution in the country.

A field indicating a collection name was included for several reasons. A researcher may have past experience with a specific collection, so a Boolean search for, "Higby Family Photograph Collection" AND "Captain" could provide a hit even though the Captain's name is not known to the researcher. Certain collections may also be known for a certain level of technical quality, authoritative documentation, or even for representing a specific geographical area of Ohio, or specific Ohio military unit. These are all filtering tools that can be effectively used by the researcher.

In consideration of the fact that most OHS collections come from private donors, it is good public relations to acknowledge the name of a given collection, as a tip-of-the-hat to the donor.

One interesting element of item identification in a graphic collection, is that no universal classification system exists for photographs, unlike the well-established systems in place for textual materials. There is simply no substitute for a LC or Dewey Classification numbers that have proven invaluable for cataloguing, indexing, and shelving text-based materials. This situation presents the challenge of uniquely identifying each item. The graphic items in the LC catalog are listed in a field labeled "LC Call No.", but the attribute values found in this field are really more in the nature of locators than classification or call numbers. They include a combination of elements such as file location, proper names, and descriptive elements. Examples of the "LC Call No." field include:

• Item in PRES FILE - Grant, Ulysses S.--in uniform--Photo--Three-quarter and full
• Illus. In AP2.H32
• Item in LOT 4192
• Item in BIOG FILE - Chase, Salmon Portland, 1808-1873
• DRWG/US - Lumley, no. 5

The field contents are collection-specific, and obviously do not present any sort of universal standard. It seems that the P&P Division has several very large files, the PRES FILE or the BIOG FILE for example, in which items are filed by their appropriate LC Name Authority. It doesn't appear that items are always retained in their original collection, but have been separated and grouped in subject files. Images in the OHS collection, however, are housed differently in the four major collections:

1. SC Files: grouped by subject
2. P-Collections: groped by provenance
3. AP-Collection: grouped by format
4. OVS-Collection: grouped by size

The most appropriate common method of providing a call number for each item was to further narrow the existing collection number. Thus, a P-Collection item already labeled P220/6/4, can be uniquely identified by simply adding another numeric character at the folder level: P220/6/4/#1. Thus, the item in question can be readily understood as being held in collection P-220, box #6, folder# 4, item #1. Instead of labeling this field LC Call No., which is more than a little misleading because no such call number exists, this field in the OHS record is simply labeled LOCATION.

The copyright field is included more as a disclaimer by the Ohio Historical Society. The blanket statement that is included in most of the images records reads:

Copyrights have not been dedicated to the public. Consideration of the requirements of copyrights is the responsibility of the author and publisher.

This field is a valuable catch-all for providing additional descriptive and subject information that doesn't fit well in any other field. It is less structured, and the cataloger is given discression to add information that is deemed likely to be helpful to identifying either the uniqueness of an item, or its relation to other items.

In this field, the caption on the face is transcribed in quotes, as well as the text from the verso (see Appendix C). This treatment may help to mitigate any unsound conclusions drawn by the indexer in the process of using these hand-written elements to construct a derived title. In other words, the researcher can examine the text exactly as it appears on the item, and draw his or her own conclusions.

Once the item's location has been precisely represented, the last field, the Control Number, completes the record. This number is useful only to the institution, as its purpose is to simply to identify the division of LC (in this case the Prints & Photographs Division), and list a unique control number for the item or file that is being cataloged. In the OHS database, this field was used to provide a link from the item to the electronic record, to facilitate future record updates. It could also be utilized in the Web page in the photo duplication ordering process. This control number corresponds directly to the original file number, and as each file is considered a "document" by the index, is regarded as the bibliographic unit in this index.

Once the bibliographic records had been fine-tuned, the next step was to generate an index and test it. It was not practical to post this index on the Web for public response due to restrictions from the Ohio Historical Society. All Web elements must be presented to a the OHS Web Committee. Decisions concerning which new elements to incorporate in its Web pages rest in its hands. As this process was beyond the scope and time frame of this project, it was not attempted. It was, however, possible to generate an in-house index, and invite members of the reference staff to test and comment on this product. The results from these tests are included in the Chapter V, Testing and Subjects for Future Research.