Introduction

In the 18th century, in order for new ideas to reach people, authors had to be printed so that their work could be shared. Without being printed, it did not matter how smart a person was or whether the person had important things to say, as nobody outside of that person’s direct environment would notice. Thus, publishing and printing networks during this time had highly influential powers regarding whose ideas and opinions would be propagated, controlling which ideas and movements could transition into a big revolution (Grafton and Eisenstein, 1980).

The analysis of printing and publishing networks are therefore an important part of understanding these movements happening during the Enlightenment. One problem encountered by many 18th century scholars is that authorial bias of metadata collection resulted in missing metadata on publishers and printers (Sher, 2010). This is especially a challenging problem for works containing ideas and opinions against the established political structures, which often are highly relevant for scholars who want to examine the emergence of new movements.

Figures found on the printed documents, such as emblems or stylized lettering, are together known as printing ornaments, and they are considered to be a printer’s unique fingerprint. Because these ornaments were produced with hand-cut wooden blocks unique to the printer, historians in the last decades have been able to study these ornaments in order to learn about printers and their networks (Maslen, 2001; Maslen and Lancaster, 1991; Goulden, 1988; Maslen, 1973; Ross, 1990). This work, however, has only been done manually using the human eye. The novelty of our work is in the use of new image recognition software to undertake a first identification process via clustering of the same printing ornaments. In this project, we work to understand whether a common image clustering algorithm can cluster printing ornaments sufficiently enough to be used for the identification of printers.

It is important to highlight that statistical models require data that is good enough to produce satisfactory results. What “good” data depends on the task at hand; in this case, we require clear images from the books’ decorations, with documented labels identifying the ornament type, and possibly metadata that connects the ornament to additional information, such as the printer, the year or location. Because of these concerns, another question we explore in this project is if it is possible to produce better data than what is currently available through the use of machine learning methods, specifically for extracting the decorations under study.

Background

The Enlightenment was one of the most significant movements in Europe in the 18th century. It was based on intellectual development regarding new ideas, concepts and rights. For a movement to be considered a revolution, ideas needed to be spread within and also across countries, from those in the high to the low class. The dissemination of ideas in the 18th century was enabled through the increasing and industrialized publishing and printing industry and, consequently, a new availability of books and pamphlets (Shen et al., 2017).

From the perspective of scholars and authors, publishers and printers were gatekeepers to the broader audience. Printers had the power of whether ideas would get printed and spread or not, especially considering that reprinting works was often decided by the printers themselves. This gatekeeping function resulted in complex intellectual networks containing authors, publishers, printers and booksellers connected via shared endeavors and interests (Raven, 2007). To be able to fully retrace the flow of ideas and information these intellectual networks need to be reconstructed and analyzed. A network analysis can reveal important factors and influences for the upscaling of the Enlightenment movement and for the identification of relevant decision-makers within the publishing and printing industry. Big databases such as the English Short Title Catalogue (ESTC), which contains metadata records about a large number of works printed in the 18th century, enable the reconstructions of those networks using the bibliographic metadata (Hill et. al, 2019). One main problem of this approach is the lack of metadata beyond the basic metadata fields, such as author and title, exactly like just who printed a work. An enrichment of the existing metadata records regarding printers could help to extend the pictures of intellectual networks and their key figures.

Before the year 1800, printing was an artisan craft which needed time and effort in order to produce a single book or pamphlet compared to the latter mechanized processes. The establishment of the letter press in the 18th century resulted in an increased number of printing output and a decrease of the price and therefore an increase of the accessibility for those products for the middle class. The printing devices for letter printing, wood or metal-cut blocks, were designed and cut manually. Particularly, ornaments for the decoration of the book, like initials, head and tail pieces, etc., were one of their kind and usually owned and used by a single printer. Additionally, the printing devices would inherit signs of damage through the use over the years which is visible in the printed ornaments (Greenwood, 2015; Wilkinson et al., 2021). The resulting uniqueness of the ornaments can be used to identify printers who were conveyed for some printing outputs but not all.

Missing printer information is often the case for works which contained ideas and opinions opposing the established societal structure, because the printer did not want to be associated with the content of the books or pamphlets. One example is from the printer John Darby II, who likely printed over 1000 works although he only put his own name on 230 books in the time period between 1707 and 1732 (Wilkinson et al., 2021). But for analyses of big movements, those works are the most interesting ones precisely because they criticized the established structures.

As mentioned in the introduction, several historians (Maslen, 1973; Maslen, 2001; Goulden, 1988; Ross, 1990) identified works printed by single printers in laborious projects. The researcher collected ornaments known to be printed by this printer and manually compared those with ornaments in books or pamphlets of interest or thought to be printed by the printer.

The two assumptions, tested by Maslen in 1973 on the Bowyer ornament stock, on which the identification of printers via printing ornaments is ground are the following:

1. printer’s ornaments of that time and place were unique and are in consequence individually recognizable,
2. printers of that time possessed their own stock of cuts.

For our study, we assume that the algorithm is also able to recognize the differences between the unique ornaments and, thus, that it clusters the similar ornaments together.

Anyhow, the printing practices in the 18th century were not as clear cut as assumed by the theory. Printers and publishers are not necessarily easily distinguishable because the tasks of the one profession was sometimes partially overtaken by the other and the other way around. Shared printing practices of bigger volumes may mislead regarding who printed which parts. Additionally, printing devices were not necessarily only owned by the same printer. Printing devices were inherited to the next generation or if a printing house was restructured or in financial troubles sold to other printers. Therefore, single printing devices can be associated with several printers and may not clearly identify one printer (Maslen, 2001). Those aspects have to be considered, if printer can be automatically identified via ornament clustering.

Next section: Data >

University of Helsinki

Spring 2023: Printmark Classification

Table of Contents

Introduction

Background