The most refined catalogue imaginable will, in spite of its inherent quality, be only as good as is permitted by the accessibility of the information it contains. A poorly indexed catalogue is, therefore, automatically a poor catalogue. This section considers the question of accessibility.

An archive using a card based catalogue will probably be restricted to a single copy catalogue, available in only one order. An , organisation using a loose-leaf (paper) catalogue, could fulfil several 'indexing' functions by producing extra copies of the catalogue sorted by different priorities. (This option is, of course, also available with card catalogues, but only if quantities of typing and clerical assistance, usually not available to new collections, are to hand). If restricted to a single copy catalogue, the archive should maintain that catalogue in accession number order. It is true that this removes the possibility of using the catalogue in any self-indexing capacity, but the accession number is the most available, logical and incontrovertable label identifying an interview for the purposes of all references to it. A file in alphabetical order of informants' names is the next most useful tool, but never as a sole course. Remember that ultimately references would be not just 'SEE SMITH' or 'SEE SMITH, JOHN' but potentially 'SEE SMITH, JOHN (Accession Number XXXXXX)' to achieve positive identification. A decision on the importance of other aspects of the informant or the interview circumstances for indexing purposes would have to be taken by the individual archive. If its work is built around interview projects, lists of the interviews relating to each project are obviously essential; a collection serving regional interests may also feel a need for an index by informant's home-town, and so on. The bulk of indexing work, however, will lie in providing access to the subjects covered in an interview.

If the circumstances of the archive permit, the work of subject-indexing should be carried out at the same time as the writing of the catalogue synopsis and by the same cataloguer. Not merely is the possibility of actual error or mistaken emphasis then reduced compared, say, to a practice of employing an independent indexer working only from the synopsis, but the cataloguer may use the index to supplement the information in the catalogue. If an informant, for example, has spoken of the qualities of Hurricane, Typhoon and Tempest aircraft, the cataloguer may for the sake of brevity be able only to write in the synopsis 'Good qualities of Hawker aircraft' while the index could refer to each machine individually.

The archive must decide how far its resources permit it to be helpful to its user. In the first place, it must decide how accurately to locate each reference; whether to the correct interview, to the correct reel of the interview, or to a precise, measured or timed location on that reel. If synopses have been written reflecting reasonably the order of items on each reel, the second suggestion is probably the best compromise. If interviews tend to be short in a given collection (say an hour or less) even a reference simply to the interview may suffice.

Next, a decision must be reached on how much information to convey in each index entry. That is, should a reference consist merely of a keyword or a UDC code linked to an interview number (as the index at the back of a book merely links topics to page number) or should it incorporate a brief, descriptive statement. In a strict sense, when the index exists as a guide to the catalogue, the former approach should be adequate, while the latter requires extra effort. On the other hand, the latter approach does reduce the labour involved in discovering appropriate references if the keyword or classification selected offers a multitude of possible citations. If a user were researching food supply in the trenches in the First World War in an archive with much material on army life throughout the twentieth century, consider the varying degrees of usefulness to him of the following three types of heading for a reference to a particular reel:

  1. 355.65 (this is the UDC code for 'Military Administration: catering, feeding, rations, messing')
    1917, Passchendaele Ridge: irregular supply and terrible condition; effects on men's health and morale.

If an archive has the resources, the last example is likely to be the most popular with researchers, although the extra effort required is far from negligible.

The most important decision to be taken in subject indexing is, of course, what kinds of information to select for indexing and how to label them. It is, to begin with, extremely desirable that the indexing staff should have a knowledge of the subject matter covered by the interview and an understanding of the intentions of the interviewer (and of the archive) in making the recording. Without such knowledge, a cataloguer may be liable to miss, or misinterpret, the significance of a point covered in the interview. Staff should also understand absolutely what constitutes useful coverage of a topic in oral history, and what does not. A brief or passing reference is only useful if the fact related is of particular interest; opinions or arguments unsubstantiated by evidence or example are similarly only of occasional value; second-hand information may be too unreliable to merit indexing, but if of special interest may be recorded with suitable qualification. It is above all the informant's own experiences, his own recollections of events, and his own attitudes and opinions that interest the oral historian and that should be reflected in indexing.

Once the subjects to be indexed have been chosen, the next step is the allocation of appropriate subject headings or keywords. If the archive is using an extant classification system, the problem must be one of detailed selection. For example, does an informant talking about government direction of agriculture in war time constitute, in UDC Terms, 351.778.2 ('Public Health, Food, Housing Etc/Food Supplies, Housing/Food Supplies, Household Goods') or 351.823.1 ('Economic Legislation and Control, Agriculture, Trade and Industry/Extractive Industries/Agriculture, Stockbreeding, Hun ting, Fishing') or 355.241 ('Forces (Service) Personnel, Mobilization etc/Mobilization of the whole Nation, Economic Mobilization/Agriculture, Industry, Manpower')? If the archive is using its own classification system the problem may be still more complex, for the appropriate classification may have to be devised and not merely selected.

Some types of indexing present few problems, and (as suggested earlier) it may well repay a new archive to concentrate its indexing activities at first on such areas. These types are, generally speaking, those where identification is positive and labelling amenable to broadly accepted rules or usages eg places, personalities, events (arranged by date) and equipment. Provided agreement is reached on which conventions or textbooks are to be regarded as authoritative, sections of a subject index covering areas such as these may be left to grow virtually unsupervised.

The classifier's real difficulties begin with those rare areas of subject matter not amenable to concrete definition; that is, areas of abstract or conceptual information which are, of course, bound to figure largely in the personal opinions and reminiscences which interviews elicit. The problems may be reduced if both index user and index compiler are encouraged to visualise even the conceptual in as concrete a manner as possible. If an enquiry can be turned from a hazy 'I am interested in medical advances' to a more solid 'I would like to find a description of an early blood transfusion' -and if indexing is arranged to cater for such enquiries - then everybody's life becomes considerably easier. The quantity of conceptual work, however, can only be reduced in this way, not eliminated, and will continue to present two main problems. The first is the question of the level of detail at which the index should seek to work: should, for exam pie, an expression of anti-semitic opinions by a Nazi sympathiser be indexed as 'Politics' or 'Fascist Politics' or 'National Socialism', or should the approach be 'Racialism' or 'Anti-Semitism '? The second major problem which is bound up with the first is that of establishing a controlled vocabulary for the conceptual index.

To ensure that a common vocabulary is shared by all its indexers (and by users) an archive must expect to produce, sooner or later, some kind of guide to its index. Part of the guide's function is purely linguistic - establishing 'preferred' terminology among synonyms (eg referring users from 'CALL UP' or 'DRAFT' to the preferred terms 'CONSCRIPTION'). The remaining part of the guide's job is to explain the classification hierarchies used by an index, so that a researcher may have explained to him the relationship between general and detailed terms (reminded, for example, that an interest in 'PREVENTIVE MEDICINE' may be furthered by exploring 'VACCINATION', 'MASS X-RAY PROGRAMMES', etc) and be guided between terms at the same level of detail (eg be informed on looking up 'BLACK MARKET' that the index also has a term covering 'SMUGGLING'). If an adequate guide exists, the question of selecting the appropriate level of detail when making an index entry becomes less severe as the intelligent user of the index will find his own way through the hierarchies.

The nature of the collection, in any case, will probably suggest an appropriate level of detail for indexing. If an ex-soldier interviewed by a military museum refers briefly to his pre-conscription work as a footman , or an ex-footman interviewed by a social history archive refers briefly to his military service, the two organisations would probably feel that, respectively, 'DOMESTIC SERVICE' and 'MILITARY SERVICE' were the most detail their users would expect of them. In their own areas of specialisation, however, they would naturally wish to go into much greater detail. Even in those areas that appear to require detailed attention, however, it may be possible to save time on indexing if the appropriate information is readily accessible elsewhere. If an archive carries out its interviews in subject groups or projects, and each project is well catalogued, a user with broad interests that lie within a particular project's terms is likely to find the items that interest him as quickly by reading the catalogue entries for that project as by hunting through an index. There are, of course, dangers for both user and archive in relying too heavily on this short-cut; a researcher interested in the miners' strike of 1974, for example, should not expect to find all his material in interviews with miners, although he would find a great deal of it there.

The archive must also decide what arrangement of its subject indexes will most help its users, choosing between a combined alphabetical list (on the model of a dictionary) -which should help the general browser, but will frustrate the specialist by forcing him to refer back and forth -and a structured classified list (on the model of a thesaurus) -which should help the specialist, but will frustrate the researcher whose interests cross the borders of the classifications used by the index. As both systems have their advantages, an archive may attempt to provide both to a degree, by adopting one for use and reflecting the other in its vocabulary guide. In weighing advantage against disadvantage, however -as so often in this chapter the ultimate decision must be taken by the individual archive, in the light of its own circumstances, needs and experience.