In September 2019, Prof. Elaine Hobby was invited to speak as part of Yale’s Symposium on ‘Scholarly Editing of Literary Texts from the Long Eighteenth Century’.
Now, through digital wizardry (led by Sue Walker at The Lewis Walpole Library and edited by Guy Ortoleva at the Yale Broadcast Studio), the talks from the day are available to view, incorporating speakers’ powerpoint slides. These can be seen on the Yale Library YouTube channel:
Talks from the morning session: including the sympoisium organiser, Stephen Clarke discussing The Yale Edition of Horace Walpole’s Correspondence; Robert DeMaris Jr outlining the editorial processes involved in The Yale Edition of the Works of Samuel Johnson (first volume published 1958, and the 23rd and final volume published in 2019); and Peter Sabor, whose talk is entitled ‘Hemlow and Beyond: Editing Frances Burney’s Journals and Letters, 1972-2019’.
Talks from the afternoon session: including Gordon Turnbull, general editor of the Yale Boswell Editions – a project that was initially conceived in the mid-twentieth century, with the final volume of The Life of Samuel Johnson appearing in 2019; Elaine Hobby on the editing processes and wider aspects of the Editing Aphra Behn in the Digital Age project, and Michael F. Suraez, whose paper asked ‘What Might it Mean to Editing a Book? Pope, Poesis, and the Possibilities of Bibliography’.
In the demanding world of the commercial
theatre, Restoration dramatists were adept in the art of adaptation, producing
re-workings of Elizabethan, Jacobean and Caroline plays. 60 Restoration
adaptations have been identified from the period 1661-1700, not including possible
unacknowledged adaptations of lost plays (Harbage 1940). With the exception of
Shakespeare’s and Fletcher’s work, which accounts for almost half of Restoration
adaptations, playwrights often re-packaged earlier plays without
acknowledgement. A notable example, of course, is Behn’s The Rover (1677),
adapted from Thomas Killigrew’s Thomaso (1654); accused of plagiarism (a
fairly novel concept in the period), she was compelled to defend herself in the
published play’s postscript: ‘the Plot and Bus’ness (not to boast on’t) is
my own: as for the Words and Characters, I leave the Reader to judge and
compare ‘em with Thomaso’ (1677, p. 85). Tellingly, Killigrew, and
many other male contemporaries, never had to account for their own
unacknowledged borrowings (see Hobby forthcoming). For the purposes of
attribution, these adaptations present a particular problem.
Computational techniques have been developed to identify authorship signals in contemporary, and roughly synchronous, collaborations within a single literary text (see e.g. Craig and Burrows, 2012), and studies have shown that genres tend to possess distinct linguistic profiles. Moreover, as Mel Evans has shown with Behn’s drama, language use evolves over time (see Evans, 2018), so that authorial profiles rarely remain stable. But what if different authors who have contributed to a single work are separated by decades? This blog post highlights our most recent work, exploring the issue of ‘precursory authorship’ (Love, 2002) in Behn’s Rover, and in The Counterfeit Bridegroom (1677), a play traditionally attributed to her, the results of which can be found in our article ‘Stylistic Palimpsests: Computational stylistic perspectives on precursory authorship in Aphra Behn’s drama’ (2020), published open access in Digital Scholarship in the Humanities.
What is an adaptation?
In the Restoration, forms of dramatic adaptation could include 1. straightforward reproductions of an earlier play 2. re-adaptations of a play already adapted 3. prose adaptations and 4. translations from classical sources or continental drama and prose. In its merging of the authorship question with temporal distance, defining the processes of adaptation is a complex undertaking. Julie Sanders notes that adaptation can be a ‘transpositional’ practice, in which one genre is transformed into another; it can be editorial in its minor revisions; or an ‘amplificatory procedure engaged in addition, expansion, accretion and interpolation’ (Sanders 2016, pp. 22-3). An adaptation might also be derived from multiple genres and, in principle, could import any number of re-workings into its current form. This variability poses a challenge for attribution; for example, an editor’s minor revisions within a text may not provide sufficient data for an accurate authorial profile. Our work on Behn’s drama, therefore, used a range of computational techniques to determine whether precursory authorship is discoverable within adapted plays.
There were enough plot, character and verbal echoes of Thomaso in The Rover for Behn’s contemporaries to note the similarities. Behn’s practices of adaptation involved the cutting and compression of large amounts of dialogue into smaller speeches and the expansion of the women’s parts (see Hobby forthcoming). We would expect, then, that the computational analysis would identify the linguistic closeness of the source-text with the adaptation, compared to additional works by Killigrew, or any other author (excepting Behn). Our approach involved measuring the most frequent words across works by Behn, Killigrew and a sample of Restoration authors, building authorial profiles to compare to The Rover and Thomaso. Using a combination of exploratory statistical measures (Principal Components and Cluster Analysis (as introduced in previous blog posts)) and machine learning, (Delta and Rolling Delta), we found that Rolling Delta was most consistently sensitive to precursory authorial style in the adapted texts. Rolling Delta is designed to identify stylistic variation in collaborations at a finer-grained level than the broader patterns elicited by PCA. The test conducts a linear examination of the collaborative text, and play texts can be divided into Acts, word segments of varying lengths (e.g. 2500), or overlapping segments to assess degrees of similarity and difference. Since a collaborating author is more likely to have contributed to a play in parts (e.g. their style clusters in one area, or is dispersed throughout a text), such segmentation methods allow us to examine different sections of the text in more detail.
Figure 1 shows a Rolling Delta analysis (calculated using Hoover’s excel spreadsheets) comparing plays by Behn, Thomas D’Urfey and Killigrew (including Thomaso) to The Rover, which is divided into 2500-word segments. The test measures the 600 most frequent words (MFW) across the dataset; it also removes pronouns and uses a process of culling (at a rate of 60%) to remove words that re-occur at a disproportionally high frequency in one text only, such as character names or place names. As expected, Behn’s plays have the greatest likeness to The Rover, indicated by the fact that they have the lowest values across the majority of the test. The plays of Behn’s contemporary playwright, D’Urfey, on the other hand, have consistently higher scores than either Behn or Killigrew. Of Killigrew’s plays, Thomaso has the lowest values overall, but particularly so in segments 1 and 3. Furthermore, the low scores of Killigrew’s Parson’s Wedding suggests that the similar profile is authorial and not just play-specific. By contrast, Bellamira’s constant high scores are likely a result of Rolling Delta picking up on the fact that the play is the only tragedy in the test; studies have shown that Early Modern plays have distinctive genre-aligned lexical profiles. In general, Killigrew’s authorial likeness is strongest in the first half of the play. Whilst the retention of Killigrew’s style in The Rover is evident, it is likely diffused throughout Behn’s dialogue, containing a mixture of function and content words that are captured here by quantitative measures (see Hobby forthcoming). We might then consider Behn vindicated by computational analysis in her challenge to readers to judge the originality of her ‘words’ – the majority of which, the test suggests, are hers.
The Counterfeit Bridegroom
Middleton’s 1638 comedy, No Wit No Help Like a Woman’s, is the primary source-text for The Counterfeit Bridegroom. Traditionally attributed to Behn on some shaky speculative evidence (by John Genest in 1832), our tests were unable to attribute the text to a single Restoration candidate, perhaps because of multiple hands in the play, or because the real author was missing from our corpus (see Evans and Hogarth, 2020). However, our tests were able to identify Middleton’s role as precursory author.
The Rolling Delta test of figure 2 was calculated using the same parameters as the test on The Rover discussed above (600 MFW, 60% culling, no pronouns); but this time the test-text is segmented by act to account for particular critical observations about the retention of Middleton’s language in Act 2, and parts of Act 3 (see Challinor forthcoming). As the chart shows, the source-text No Wit has the lowest value of -2.19 (thus greatest similarity) in Act 2 and a low score of -1.12 in Act 3. Behn’s low scores in each act, however, are not sufficient evidence to claim that she adapted the text single-handedly. Not only do Edward Ravenscroft’s plays have negative values in Acts 1, 2, 3 and 5, but in other versions of this test with different candidate authors, similar patterns emerged that precluded a positive attribution.
The commonplace that all texts are a patchwork of
acknowledged, unacknowledged, conscious and unconscious borrowings from earlier
works is a crucial consideration for attribution studies – the purpose of which
is to disentangle the authorial from stylistic factors such as genre and
chronology. Our work on precursory authorship suggests that to fully understand
the stylistic properties of early modern texts, such chronological layers need
to be accounted for – especially in Restoration drama. It also opens up new
avenues of enquiry – for example, are there common patterns/conventions of
adaptation across texts in this period that are identifiable at the
quantitative level? Or, is it possible to determine precursory prose-texts in
drama using these methods?
In our next article (forthcoming in a special Behn-focussed issue of Women’s Writing), we tackle questions of genre, lack of data and the possibility of tracking minor editorial interference by a hack writer in The Younger Brother (1696).
Challinor, J. (forthcoming). Headnote to The Counterfeit Bridegroom. In Bowditch, C., Evans, M., Hobby, E. and Wright, G. (eds), The Cambridge Edition of the Works of Aphra Behn Vol II. Cambridge: Cambridge University Press.
Craig, H. and Burrows, J. (2012). A collaboration about a collaboration: the authorship of King Henry VI, Part Three. In Deegan, M. and McCarty, W. (eds), Collaborative Research in the Digital Humanities. Farnham: Ashgate, pp. 27–65.
Evans, M. (2018). Style and Chronology: A Stylometric Investigation of Aphra Behn’s Dramatic Style and the Dating of The Young King. Language and Literature, May. https://doi.org/10.1177/0963947018772505.
Evans, M. and Hogarth, A. (2020). Stylistic palimpsests: Computational stylistic perspectives
on precursory authorship in Aphra Behn’s drama. Digital Scholarship in the Humanities.
Genest, J. (1832). Some Account of the English Stage, from the Restoration in 1660 to 1830. Vol. I. Bath: H. E. Carrington.
Harbage, A. (1940). Elizabethan: restoration palimpsest. The Modern Language Review, 35(3): 287.
Hobby, E. (forthcoming). Headnote to Aphra Behn’s The Rover. In Bowditch, C., Evans, M., Hobby, E. and Wright, G. (eds), The Cambridge Edition of the Words of Aphra Behn Vol II. Cambridge: Cambridge University Press.
Love, H. (2002). Attributing Authorship: An Introduction. New York: Cambridge University Press.
Sanders, J. (2016). Adaptation and Appropriation. 2nd edition. New York: Routledge.
Registration for our conference ‘How to do things with early modern words’ is still open!
More information about the three-day conference (23rd-25th April) at Loughborough University can be found on our dedicated conference page, including the draft programme, advice on accommodation, and the schedule of social events.