Research
The Material Culture of Wills, England 1540-1790
This is a four-year research project funded by the Leverhulme Trust that seeks to understand how people's attitudes towards material objects changed between the sixteenth and eighteenth centuries. This period saw extensive social and economic change as England and Wales transformed from a nation of peasant farmers into an industrial nation fully integrated into global trade. We know much about how the ownership of goods changed in this period, but relatively little about people's relationship to their possessions, how they understood them and the meaning they attributed to goods. This project uses an unprecedentedly large sample of wills to investigate this issue.
Wills are the most common personal document that survive from the period before 1800. Millions of wills were proved at various ecclesiastical courts throughout the country. Although it was not necessary to do so, many people described in detail some of their possessions, what they were, their condition, who they were left to and, in some cases, why that bequest had been made. For example, William Radcliffe, a London cloth worker, left his sister 'a gold ring with a picture of death's head, for all her unkindness'. Descriptions like this one, therefore, offer an opportunity to examine not just the ownership of goods in the past, but also to investigate the meaning of those goods.
Handwritten Text Recognition
Scholars have used wills to examine all manner of topics, such as changes in the nature of religious belief and other effects of the Reformation, or familial relationships as signalled by the distribution of property. However, all previous studies have been restricted by the substantial amount of time it takes to transcribe a will by hand. Recent improvements in the automatic transcription of handwritten text, however, mean that it is now feasible to produce a large number of transcriptions of wills in a relatively short amount of time. We are transcribing 25,000 wills from five periods across the sixteenth, seventeenth and eighteenth centuries, 1538-52, 1604-08, 1664-6, 1725-6 and 1785-6. We are doing this this using a platform called Transkribus. Transkribus allows us to perform all the stages required to generate automatic transcriptions of handwritten text. First, we have worked with a set of expert volunteers to transcribe around 500 pages of ground truth. Second, this ground truth is used to train and validate a Handwritten Text Recognition (HTR) model. This process allows the AI model to learn how to deal with the different handwriting styles found across our period. Third, we can then run that model on the 25,000 wills that constitute our sample. This will produce automatically generated transcriptions of those wills, and it is that text which we need to then check. The model will be run through Transkribus, but will hopefully be made publicly available once it is fully developed to allow others to use it on any similar documents they may wish to digitise.
Wills in this period tend to be one to two pages long, and each page contains about 50 lines of text. The 25,000 wills, therefore, will produce something in the region of 2 million lines of text. This is too many for us to check by hand as a team, so we are using two approaches to ensure the outputs from the HTR are as accurate as possible. Wills in this period are formulaic, with particular phrases appearing time and again. This aspect of wills means that we can perform some automatic checks, both to correct obvious mistakes, and to identify lines of text that need closer inspection. This is where Zooniverse comes in, we will identify a subsample of lines that need to be checked by hand through two separate Zooniverse workflows.
The first workflow, ‘Checking Transcriptions’, involves you comparing a line of text outputted by the HTR model with the original document, and correcting any part that line which is inaccurate. The second workflow, ‘Transcribing Wills’, requires you to transcribe a single line of text from scratch based on an image of that line. This second workflow is designed to allow us to maintain quality control, we can compare these transcriptions from scratch to the checked outputs from the first workflow and ensure that confirmation bias does not distort responses to the first workflow.
The process will be an iterative one, as transcriptions are checked those corrections will be fed back into the HTR model to improve the automatic transcriptions, so your work will be invaluable not only in ensuring our work is accurate, but also in improving this technology for further work. You can read more about our methods in a number of blog posts here.
The images we have of the wills are scans made in the early 2000s of mid-twentieth-century microfilms of the original will registers held at the UK National Archives. Some of these images are, as a consequence, low quality and difficult to read. Our project will allow people to access these wills without having to decipher difficult-to-read scans, and will provide a model for future work opening up other document collections that are currently hard to read or use.
Content Warning
These wills cover much of the period in which the slave trade was active. Consequently, some of the wills included in our sample involve bequests of enslaved people or other mentions of objects related to human enslavement. It can be shocking to see such issues described in mundane ways and the language used can be upsetting. Please be aware of this possibility when working on our material, and either skip onto a different image by refreshing your browser, or take a break entirely.
Outputs and Further Information
These transcriptions will allow us to investigate how new goods arising from technological advances and developments in trade were integrated into English society, and how the role of objects in familial and financial relationships changed over time. Furthermore, the methodology developed here to transcribe these documents will aid in opening up the vast trove of valuable handwritten manuscripts held at The National Archives. We will be reporting our findings here and on our website, as well as holding in-person and virtual events. If you want to keep up to date with the project, our progress and outputs please sign up to our mailing list
The full transcriptions and a database created from their contents will be made publicly available. Several articles and an edited collection on the history of material culture in this period will be produced. These works will be open access to allow as many people as possible to benefit from this work, and all contributors to the transcriptions will be acknowledged.
