Revising Archival Descriptions: part 2 – the results

WARNING: In describing work about offensive terms in archive catalogues it is necessary to include some of the terms that are considered to be offensive. The intent is not to shock but contribute to the ongoing debate about this work.

In part 1 we described the project methodology using AntConc linguistic software to search through a large quantity of archive finding aids for potentially offensive terminology and to export the results for reviewing in MS Excel. The opportunity arose to collaborate with the Archives Hub which resulted in a significant quantity of additional data from Welsh archive services.

Reviewing the matches

AntConc is able to search the archive catalogues to find matches, what AntConc calls a ‘hit’, against the terms list. In saving the hits AntConc also captures the ten words that precede and the ten words that follow the term. A manual review process considers the specific context in which the term has been used, and in doing so distinguish between:

1. terms which are not offensive and appear due to a common core with a term on the list; ‘association’ and ‘assignment’ both match the search for ‘ass*’
2. terms which appear in the list but appear in a different context; ‘Cissy’ and ‘Gay’ both feature on the terms list but also appear in finding aids as an individual’s forename 
3. terms which were used historically, but are no-longer considered appropriate or acceptable; ‘cripple’, ‘dumb’, ‘lunatic’ and ‘mental’ often appear in institutional names
4. terms which are offensive, for example ‘coon’, ‘negro’, ‘midget’ and ‘slave’

Terms that fall within the first two scenarios are considered to be ‘false hits’, the latter two are marked as ‘true hits’. This process is about discovery with details of the ‘true’ hits returned to the archive service and a suggestion it reviews the use of that term.

The headline results

Through collaboration with colleagues at 12 archive services across Wales and the Archives Hub we have successfully applied the methodology to over 29,000 archive findings aids totalling over 58 million words. From this body of data AntConc registered 195,000 hits and the manual review process refined this to just over 5000 instances of true hits arising from 71 distinct words.

With so much data a few patterns and trends emerged;

• 26 of the true terms (36.6%) were only found once amongst the 12 services [but not always the same service]

• 4 terms (‘lunatic’, ‘mental’, ‘mentally’ and ‘slaves’) were found to be present in ten or more of the 12 archive services

• a clear distinction between collection-level and detailed finding aids with the former far less likely to contain either false or true hits

Context is everything as the same word can be both ‘false’ and a ‘true’ hit in the same sentence. Some services have begun to replace references to slave to ‘enslaved’ but simply applying this change to all instances would not be appropriate. A considered approach needs to be applied to addressing this issue.

The project partners have also looked at the application of content warnings in archival finding aids and hope to implement this in the near future.

Next steps?

Archives and Records Council Wales hopes to collaborate with similar work being undertaken in the museum sector. It is hoped that a next phase of this work can engage with stakeholder groups from specific communities and with academic researchers about the terms. There is also a need to create a similar terms list in the Welsh language.

It is critical that we look to reduce the potential for our catalogues to shock or offend our users without impinging on the discovery and use of our collections. The full project report describing the methodology and analysis of the results can be found in our report.