In his major study, Attributing Authorship: An Introduction (2002), Harold Love states that the field of authorship attribution arises from the fact of the uniqueness of the individual and the manner in which this uniqueness is enacted in writing. Although, in Love’s view, “this hypothesis has never been tested, let alone proven, it remains true to say that “the issue is not whether individual idiolects and grapholects are indeed different but how these differences are to be detected with a certainty that permits the confident ascription of works to authors” (12). In the same passage, however, Love rehearses one potential solution to this puzzle:
One determinant of uniqueness is biological: at the moment of conception a mingling of genetic information occurs which is unprecedented and unrepeatable. This mingling is partly rule-governed and partly a random process. The rule-governed part ensures a degree of resemblance between siblings and close relatives and of uniformity over the race and species: individuality is never absolute. But then neither is it ever absent: in the most inbred of populations there will still be immeasurable possibilities of variation. Nature’s poker machine never gives the same prize twice. (4)
Curiously enough, Love does not use then these biologically-based insights to suggest a more general theory of the individual idiolect. What is more, although most stylometricians appear to accept some version of Love’s “primary hypothesis,” most stylometric work has situated itself in relation to forensic science without believing orientation need involve the understanding of genetic inheritance (Mosteller & Wallace 1964/2007; Burrows, 1987; Holmes, 1994; Burrows 2003; Juola, 2006; Coulthard, 2014; Craig and Brett Greatley-Hirsch 2017). In spite of the regular invocation of the analogy of the authorial “fingerprint”, a concept that can be traced back to the work of Francis Galton, the cousin of Charles Darwin and the author of both Hereditary Genius (1869) and Fingerprints (1892), then, most stylometricians appear content to view the author as someone who invents an individual style out of the free play of language.
This lack of consensus about what author attribution studies is attempting to understand has caused problems. For example, in his article, “Visualization in Stylometry: Cluster Analysis Using Networks” (2017), Maciej Eder introduced for the first time a new method for enhancing the explanatory power of stylometric visualization. As Eder states, “since particular frequency strata are responsible, to some extent, for different signals hidden in a literary text, the dendrograms generated using longer or shorter MFW vectors presumably will also be heterogeneous. And they are” (53).
In order to create a more reliable visualization, Eder suggests the need to combine “the information revealed by numerous dendrograms into a single consensus plot” (55). Eder then describes how he used this tool on a corpus of 66 English novels in order to explore the resulting network visualization. Eder reproduces four examples of discrete dendrogram snapshots: the first offers the graphic representation for the 100 MFWs, the second, for 300; the third, for 969; and the fourth, for 970. As Eder notes: “a consensus tree of the corpus of 66 English novels has been shown (the ‘snapshots’ were computed for 100, 200, 300, etc. up to 1,000 MFWs). Some text groupings can be easily identified, including, among others, an expected cluster of the three Brontë sisters, and a branch of Kipling/Conrad, clearly subdivided into two distinct authorial voices” (56; my emphasis). I have reproduced the first two figures for reference.
The Importance of “the Noise”
In his discussion of consensus tree analysis, Eder notes a major issue: “it seems that the authorial signal is spread throughout the whole frequent and not-so-frequent words spectrum, but at the same time it may become obscured by additional and unpredictable signals, which are considered noise in classical approaches to attribution” (53). To Eder’s mind, this “noise” ought to be investigated: “Why are some authors misclassified? Which texts are wrongly attributed to a given author, and why are they linked to this very author and not to others?” (53). As Eder suggests:
The difficulties with separating one specific signal suggest that a text (written or spoken) is a multilayer phenomenon, in which particular layers are correlated. These layers include authorship, chronology, personality, gender, topic, education, literary quality, translation (if applicable), intertextuality, literary tradition (e.g. sources of inspiration), and probably many more. Arguably, literary quality somehow depends on education, genre depends on topic, authorial voice is affected by chronology, gender affects personality, and so on. (Eder 53).
Eder’s list of potential contributing factors appears oddly incomplete: nothing is said about the possibility of genetic inheritance.
What is more, the stylometric similarity observable among the Brontës is not an isolated case. Within the recent stylometric literature, there has appeared evidence for similarities among the stylometric profiles of many biologically-related writers, including the Fieldings, the Corneilles, the Amises, the Godwins, the Maughams (cf. Cafiero and Camps, 2019; O’Sullivan 2022). Indeed, the recent documenting of a larger number of such instances suggests the need for researchers to set out the postulates and corollaries of the two possible explanations, genetic and environmental, in order to discover the correct one.
The Genetic Explanation
Within the field of behavioral genetics, it has been established using multiple methods that cognitive abilities, including language ability, are among the most heritable of all behavioral traits (Abury & Plomin 2014; Plomin 2018). What this means is that behavioral geneticists expect to find similarities among siblings and other related family members in reading and writing ability.
The most important ongoing study of the behavioral traits of closely-related individuals is that of Robert Plomin and his colleagues at King’s College, London. In 1994, Plomin initiated the Twins Early Development Study, with its focus on the association of early development delays with behavioral problems and educational attainment. In G is for Genes: The Impact of Genetics on Education and Achievement (2014), Kathryn Asbury and Robert Plomin report on the work of the TEDS cohort, noting:
A study led by Dr. Bonamy Oliver in TEDS explored the genetics of writing achievement using UK National Curriculum levels awarded to the TEDS twins when they were seven years old (Oliver, Dale, Plomin, 2007). Dr. Oliver gathered teacher ratings of the children’s achievement and used the twin method to figure out the genes accounted for 2/3 of the differences between individual children, shared environment only 7%, and non-shared environmental influence the rest. The pattern is noticeably similar to that for reading achievement. She also explored whether the same pattern of nature and nurture was true for the lowest-performing children and found that it was. Low writing ability was no more or less heritable than average or high writing ability. (39)
Asbury and Plomin go on to suggest:
For both reading and writing then, we have seen heritability estimates of over 60%, evidence that the same genes operate throughout the ability spectrum (the abnormal is normal), and evidence that the same genes remain operational as children grow (continuity is genetic and change is environmental). (39)
What is more, it is not only the ability to read well that tends to run in families; it has been demonstrated that reading difficulties do also (DeFries, Vogler, and LeBuda, 1986) (35).
The Genetic Hypothesis and its Corollaries
The correlation between pairs of identical twins reared apart is a measure of a trait’s heritability. For these identical twin pairs, the heritability rate for both reading and writing ability is approximately 60%.
Genetic Hypothesis:
The genetic hypothesis states that the individual idiolect is substantially a genetic inheritance from that individual’s father and mother.
Corollaries:
- Within individual families, the individual idiolects of identical twins will be the most similar;
- The individual idiolects of two brothers or two sisters ought to show marked stylometric similarities;
- As a result of sexual differences, the individual idiolects of a brother and a sister may diverge, albeit not widely (cf. Rybicki, 2016);
- The idiolects of siblings will exhibit similarities whether or not those siblings were in close contact in childhood and adolescence;
- The idiolects of siblings will exhibit similarities whether or not there is a large age difference between them;
- An individual idiolect will bear a set of lesser affinities with more distant family members;
- The individual idiolect will remain consistent to itself, both before and after highly significant personal encounters, even including those encounters that the individual author regards as educationally or artistically formative;
- As a consequence, the individual idiolect of a particular author will even survive its translation into a foreign language, with the translated text bearing much greater similarity to the idiolect of the original author than to the idiolect of its foreign language translator (cf. Rybicki 2012).
A Broader Discussion
The closest possible kinship relationship is that between identical twins. In consequence, the first corollary is that the individual idiolects of identical twins ought to be the most similar. As Asbury and Plomin note, for both general reading and writing ability, the heritability estimate based on monozygotic twin correlations is over 60%. One possible consequence of this strong similarity may be that the number of total literary collaborations between monozygotic twins will be quite high. Two monozygotic twins will tend to collaborate more on literary projects than will two dizygotic twins or two ordinary brothers or sisters. This is because two monozygotic twins may recognize that their individual idiolects (and possibly their literary interests also) are highly compatible.
The second corollary is that the individual idiolects of two brothers or two sisters ought to show marked stylometric similarities. In the case of ordinary sisters, for example, the expected correlation would drop to around 30%, but this similarity is still sufficiently high to cause a noticeable clustering among two (or more) siblings in a generational cohort of regional, national, or international stylometric profiles.
The third corollary is that the individual idiolects of a brother and a sister may diverge, albeit not widely. If the researcher is shown two pairs of stylometric profiles, one for a pair of brothers or for a pair of sisters, and another for a brother and a sister, that researcher might be able to tell them apart. In the latter case, the researcher might likely see a convergence within this pair of individual idiolects, yet crossed with a divergent sexual signal.
The fourth corollary is that individual idiolects will exhibit similarities whether or not the siblings were in close contact in childhood and adolescence. The idiolects of a pair of brothers or sisters, say, who were adopted separately when they were very young, would still exhibit similarities. This is because the individual idiolect is substantially a genetic inheritance and forms part of the author’s biological personality; as a consequence, the individual idiolect will not be substantially affected by conversations with friends or companions either in childhood, in early adolescence, or in later life.
The fifth corollary is that sibling stylometric profiles should exhibit similarities in spite of large age differences. This corollary is of importance because beyond a certain age limit, it seems unlikely that siblings will communicate in a sufficiently prolonged and intimate manner for the older child to affect the individual idiolect of the younger child—or vice versa. A difference in age of more than five years would seem to be a reasonable estimate for the limits to the purported effect.
The sixth corollary is that an author’s stylometric profile may well bear a set of lesser affinities with more distant family members. These cases would include the second-degree relation between grandparent and grandchild and the third-degree relation between first cousins (cf. Nettle 53). One possibility here might be to study whether the growing distance between individual idiolects at one or two generation’s distance obeys probabilistic rules attendant upon greater genetic distance or not. The necessary proviso here is that language habits can change drastically over time, as shown by the temporal changes of word frequencies in large corpora. As a result, there may be limits to such connections. The obvious view would be that the more distant the degree of relatedness, the more likely that major genetic and epigenetic changes will intervene in perhaps unpredictable ways. However, the idea that certain biologically-inherited traits may run in lines of familial descent for many generations is one with which most people have an intellectual acquaintance (Galton 1892; Clark 2014). The only surefire way to test this corollary would be to create suitable dendrograms of biologically-related writers that span four or five generations of published achievement, where such extended lines of published achievement exist.
The seventh corollary is that the individual idiolect will remain consistent to itself, both before and after highly significant personal encounters, even including those encounters that the individual author regards as educationally or artistically formative (cf. Rybicki 2012; Seminck, Gambette, Legallois, & Poibeau 2022). From this viewpoint, genuine changes in an author’s idiolect might be best correlated with major changes in that author’s personality, the effects of basic biological development over the lifespan or, more problematically, the effects of significant mental deterioration, perhaps as a consequence of dementia (Holmes 1994; Le, Lancashire, Hirst & Jokel 2011; Hirst & Weng 2012).
The eighth corollary is that the individual idiolect of a particular author will even survive its translation into a foreign language, with the translated text bearing much greater similarity to the idiolect of the original author than to the idiolect of its foreign language translator. Among other things, what this means is that those literary texts which contain the works of translation of a poet-translator, such as Ezra Pound’s translated texts from the Chinese, will tend not to cluster together with the texts that contain his original, non-translated poetry. Instead, in a suitably constructed second corpus, those texts will cluster with other translated work by the original Chinese writer (Rybicki 2016).
The Environmental Explanation
Environmental Hypothesis:
The environmental hypothesis is that the individual idiolect is environmentally transmitted. It is the unique upshot of ordinary language learning in the presence of the mother and father; siblings, classmates and friends; neighbors, teachers and strangers; and of exposure to written texts from childhood on.
Corollaries:
- Any idiolectal similarities between an author’s idiolect and another member of that individual’s immediate family will be the upshot of the fact that the individual in question and that other family member lived and learned together;
- An author’s idiolect will not bear stylometric affinities with other members of that author’s immediate family if the age gap is more than about five years;
- An author’s idiolect will not bear stylometric affinities with members of his or her more distant family, unless that author was in close contact with these members of his or her more distant family when they were young;
- The individual idiolects of close family members will become more similar, if those family members live in the same region of the country and see each other regularly;
- The individual idiolects of close family members will diverge gradually, particularly if those family members rarely or never see each other or live in different regions or different countries;
In each of these five cases, it behoves the researcher to substantiate or refute the claim by specific biographical evidence—and to take note when that biographical evidence does not support the claim.
Do Monozygotic Twins Collaborate More Readily?
The first corollary of the biological hypothesis is that the closest possible relationship is the one between identical twins. An immediate upshot of this similarity might be a greater tendency for monozygotic twins to collaborate on literary compositions.
Within the last hundred years, there have been a number of significant examples of monozygotic twin literary collaborators. Perhaps the most famous example is that of the Hollywood screenwriters Julius and Philip Epstein. Together with Howard E. Koch, the Epsteins wrote the screenplay for Casablanca (1942) and were the co-authors of The Strawberry Blonde (1941), The Bride Came C.O.D. (1941), The Man Who Came to Dinner (1942), Arsenic and Old Lace (1944), Mr Skeffington (1944), and My Foolish Heart (1949). Although Philip tragically died of cancer in 1952, two more previously co-authored screenplays were subsequently made into films: The Last Time I Saw Paris (1954) and The Brothers Karamazov (1958).
A second example would be the identical twins Anthony and Peter Schaffer. The Schaffers were British dramatic playwrights and screenwriters who co-wrote three novels together using the penname “Peter Antony.” These novels were: The Woman in the Wardrobe: A Lighthearted Detective Story (1951), How Doth the Little Crocodile (1952), and Withered Murder (1955). Both of them later went on to greater success as independent authors: Anthony Schaffer wrote the film screenplays for Sleuth (1970), Frenzy (1972), and The Wicker Man (1973); Peter Schaffer wrote the film screenplays for Equus (1977) and Amadeus (1984).
More recently, the identical twins Matthew and Michael Dickman (b. 1975) have authored together two collections of poems, 50 American Plays: Poems (2012) and Brother (2016).
Perhaps the most spectacular case of all, however, involves the identical twins, Paula Bernstein and Elyse Schein. Born in New York City, these two young girls were separated as infants and adopted into separate families. They met again when they were 35. To their shock and horror, they discovered that they had been deliberately separated as part of a secret research project by the Louise Weiss Services, an adoption agency for Jewish families in New York. The project was overseen by Peter Neubauer, a child psychiatrist who also served on the board of the Freud Archives and Viola Bernard, a child psychologist. The impact of these revelations led to the twins penning a book, Identical Strangers (2007), about their separate lives and the discovery of the secret experiment. Interestingly, although the experiment concluded in 1980, its findings have been sealed at the request of Peter Neubauer in an archive at Yale University until 2066.
Identical Strangers, Bernstein and Schein’s book, takes the form of a long series of alternating entries, first by Paula, then by Elyse. As Paula suggests, “Twins really do force us to question what is it that makes each of us who we are. Since meeting Elyse, it is undeniable that genetics play a huge role—probably more than 50 percent” (Richman). “It's not just our taste in music or books; it goes beyond that. In her, I see the same basic personality. And yet, eventually we had to realize that we're different people with different life histories.” (Richman) The fact that these separated twins were together able to pen a book so quickly after their reunion offers a small piece of evidence for the genetic hypothesis.
One Brother and Three Sisters: The Brontës
In a recent paper entitled “Who Wrote Wuthering Heights?”, Rachel McCarthy and James O’Sullivan use the techniques of stylometry to disprove the thesis that Branwell Brontë, the dissolute elder brother of the three Brontë sisters, wrote Wuthering Heights, a novel most scholars have attributed to Emily. During the course of their discussion, however, the authors take note of a curious fact: “Conducting a further cluster analysis that includes a relatively comparative sample of British fiction from beyond the Brontë family reveals something that has been absent from the previous tests: in a broader context, all of the Brontës are stylometrically similar” (McCarthy and O’Sullivan 5). As the authors note: “One would not anticipate members of the same family having a stylistic fingerprint so similar that they form their own cluster when assessed in the context of their wider national canon. These findings prove that the Brontës had a significant influence on each other’s style, such that, at a macro level, they gather together as a unique cluster” (McCarthy and O’Sullivan 6). In their final footnote, the authors then make an intriguing suggestion: “A further interesting study would be to test the degree to which this trend is replicated across other literary families, or if there is, in fact, something very particular about the ways in which the Brontës developed something of a shared style (McCarthy and O’Sullivan 15n9).
The Commonness of Familial Clustering
In a more recent paper, “The Sociology of Style: Writing and Influence within Literary Families,” James O’Sullivan has extended his research and discovered basic stylometric similarities among the following close relatives: Kingsley and Martin Amis (father-son); Anne, Charlotte, and Emily Bronte (sisters); William Godwin, Mary Wollstonecraft, and Mary Shelley (father-mother-daughter); A. S. Byatt and Margaret Drabble (sisters); W. Somerset and Robin Maugham (uncle-nephew); John le Carré and Nick Harkaway (father-son). As O’Sullivan suggests, “It would appear from this initial analysis that there is in fact nothing at all exceptional about the similarities in style shared by the Brontë siblings, and that most literary families cluster together when tested with stylometry. The familial units consistently cluster in close proximity, which suggests—acknowledging the limitations in this dataset—that members of the same family typically tend to write in similar styles” (3).
Two Brothers: Pierre and Thomas Corneille
In their essay, “Why Molière Most Likely Did Write his Plays”, (2019), Florian Cafiero and Jean-Baptiste Camps use a suite of “state-of-the-art attribution methods” to study a corpus of comedies in verse by major authors at the time of the French seventeenth-century playwright Molière in order to try to settle the question whether or not Molière was the actual author of the plays attributed to him (1). The hypothesis that Molière, a poor man from the countryside, was not the author but rather that Pierre Corneille ghostwrote these plays was first popularized by Pierre Louys at the turn of the twentieth century. According to an analysis of “lexicon, rhymes, word forms, affixes, morphosyntactic sequences, and function words”, however, the authors conclude that there is no other likely author for these plays, other than Molière himself.
This selection process left us with 71 plays by 12 alleged authors: Edme Boursault (1638– 1701), Chevalier (?–1674), P. Corneille (1606–1684), T. Corneille (1625–1709), Gilet de La Tessonerie (c. 1620–c. 1660), Jean de La Fontaine (1621–1695), Molière (1622–1673), Antoine Le Métel d’Ouville (1589–1655), Philippe Quinault (1635–1688), Jean de Rotrou (1609–1650), Paul Scarron (1610–1660), and Jean Donneau de Visé (1638–1710). (Cafiero and Camps 6)
In the midst of making their argument, however, Cafiero and Camps present some striking evidence for the stylometric similarities in the work of the two brothers, Pierre and Thomas Corneille. As they state: “T. Corneille’s early plays and two plays by his brother (Le Menteur and La suite du Menteur) are, according to some criteria, mixed in the same cluster” (3). They suggest the following explanation: “Critics already noted that T. Corneille drew his inspiration from these two plays. This strong similarity does not come as much of a surprise. It could reflect the strong influence the already famous P. Corneille had on his much younger brother at early stages of his career, either punctual collaboration, corrections, or advice given by P. Corneille” (3).
The comment about the “strong influence” that Pierre Corneille had “on his much younger brother” is worth pondering. The age gap between the brothers was extremely wide: Pierre Corneille was 19 years older than Thomas Corneille. In 1629, when Pierre Corneille, at the age of 23, wrote Mélite, the first of his 36 complete plays, his brother Thomas Corneille was just 4 years old. In 1647, when Thomas Corneille, at the age of 22, wrote Les Engagements du hasard, the first of his 30 complete plays, Pierre would have been 41. Given this extreme difference in age, it might be suggested that the stylometric similarities between these two French brothers are more parsimoniously explained by means of shared genetics.
Brother and Sister: Henry and Sarah Fielding
In their paper “Anna Boleyn and the Authenticity of Fielding's Feminine Narratives,” John Burrows and Anthony J. Hassall consider the possibility that Sarah Fielding may have been responsible for the writing of a number of the female-character sections contained in some of her brother’s novels. The fact that the sibling pair did occasionally collaborate is not in dispute. As the authors suggest, “Since Fielding was known to have contributed a Preface and some editorial amendments to the second (1744) edition Sarah’s novel David Simple, and the Preface and the concluding letters (XL-XLIV) to the sequel Familiar Letters (1747), it seemed not unlikely that Sarah might have contributed a fragment to her brother’s Miscellanies” (428). In an interesting discussion of the method used in order to calculate the likelihood of such a collaboration in some of Henry Fielding’s later texts, Burrows writes:
Our calculations rest upon lengthy frequency profiles drawn from a set of different texts. The profiles are mutually dependent only in the sense that the texts are written in English; they are the work of one or other of two novelists; and the main set are all instances of the narrative mode couched in the first or third person and known as the “history.” Each of those stylistic determinants gives shape to the graphs that follow. With the proviso that (while the evidence of this short paper cannot show it) period-based and gender-based differences of style often contribute to authorial differences, I know of no other determinant that might make for the dependence of one profile upon another and so adulterate the analysis. There is no concealed reason, I believe, why the profiles drawn from Henry Fielding’s texts should correlate more closely with each other than any of them correlates with those drawings from Sarah Fielding’s texts. (439)
The astute reader will note that there is one possible concealed reason why the two stylometric profiles might be similar: the fact that Henry and Sarah were brother and sister, born of the same parents, Lt Gen Edmund Fielding and Sarah Gould. Burrows and Hassall entirely overlook this possibility of shared genetic inheritance, both in their attempt at explaining the earlier literary collaboration of the Fieldings and in explaining a significant set of stylometric results.
The Environmental Explanation: The Godwins and the Edgeworths
In a long footnote to his paper on sexual differences among British eighteenth-century female authors, Jan Rybicki notes the similarities among the idiolects of the father, mother and daughter set of William Godwin, Mary Wollstonecraft, and Mary Shelley. To deal with the similarities between father and daughter, Rybicki suggests the possibility of “parental meddling” with the texts of the daughter (761n2). To deal with the similarities of the texts of mother and daughter, he suggests the importance of tracing “the influence, on Mary Shelley, of her mother, Mary Wollstonecraft” (761n2). As is widely known, Mary Wollstonecraft died giving birth to Mary Shelley, so this influence, according to Rybicki’s implicit environmental hypothesis, must be the upshot of the daughter’s compulsive reading of her mother’s literary works during her childhood and adolescence. Rybicki offers no evidence for this compulsive reading, however.
A second possible instance noted by Rybicki involves the stylometric similarities between Maria Edgeworth and her father, Richard Lovell Edgeworth. To deal with this issue, Rybicki notes that Maria Edgeworth’s writing “tends to become very Protean around the time when she collaborated on her father’s autobiography” (761n2). The notion of a Protean writing style tends to fly in the face of the primary hypothesis of authorship attribution studies, even though it is consistent with the sixth corollary of the environmental approach.
In each of these cases, the possibility of genetic inheritance is overlooked; and no substantial evidence is offered for the environmental explanation.
The Consistency of the Idiolect Over Time
The seventh corollary of the genetic hypothesis is that the individual idiolect will remain consistent to itself over the lifetime. This consistency will not be affected by highly significant personal encounters, even those that the individual author regards as educationally or artistically formative. Scholars who adhere to the environmental hypothesis often reveal mixed views on the concept of the consistency of the idiolect over time. For example, scholars investigating literary collaborations implicitly accept that the idiolects of the authors remain consistent enough during the collaboration to be safely distinguished.
For example, in their paper “Collaborative Authorship: Conrad, Ford and Rolling Delta” (2014), Jan Rybicki, David Hoover and Mike Kestemont detail “the story of the notable literary collaboration between Joseph Conrad (1857–1924) and Ford Maddox Ford (1873-1939)” which “dates back at least to September 1898, the date of the first meeting between the two writers (Najder, 2007, p. 271)” (422). As Rybicki, Hoover and Kestemont state: “the duo managed to produce three acknowledged collaborative novels: The Inheritors (1901), Romance (1903), and The Nature of a Crime (1909, 1924). And the story is made even more interesting by various authorship claims by Ford, including those concerning a fragment of Nostromo (1904) and the dramatization of Conrad’s To-morrow (1901–2), One Day More (1904)” (423).
Nonetheless, a critical consensus is lacking regarding the importance of this extended literary collaboration on the subsequent careers of the two writers. As the authors state:
Jocelyn Baines calls the work with Ford ‘the most important event in Conrad’s literary career’ (quoted in Najder 2007, p. 273), and this is echoed by Meyers when he speaks of ‘the literary friendship that had the greatest impact on [Conrad’s] career’ (Meyers 2001, p. 177); Najder (2007, p. 273), however, calls such a statement ‘an exaggeration’ and cites other, greater influences. In general, Conradian and Fordian scholars usually side with the main object of their study in this dispute, which reached its critical mass around the date of publication of the third joint work and that was also when James’s ‘bad dream’ started to come true. (423)[1]
In an attempt to sum up the extent of the collaboration, the authors note:
The first collaborative work, The Inheritors, was, for Ford, his second published novel, coming out 9 years after The Shifting of the Fire. He did most of the writing himself, though he discussed it extensively with Conrad, whose role, he said, was to give each scene a final tap (Saunders 1996, pp. 135-36). For Romance, based on Ford’s earlier unfinished Seraphina, however, the consensus seems to be that it is about two-thirds Conrad and one-third Ford. According to the former, we collaborated right through, but it may be said that the middle part of the book is mainly mine with bits by FMH—while the first part is wholly out of Seraphina, the second part is almost wholly so. The last part is certainly three-quarters MS FMH with here and there a par. by me (Karl 1997, p. 147). (423)
The rest of the academic paper is given over to a description of the rolling delta, one of the stylo package tools, which is designed to tease out those portions of the manuscripts that were written by either Joseph Conrad or Ford Madox Ford; and the authors regard their results as robust.
However, what is striking here is the absence of any consideration that the individual idiolects might start to blend in with each other—in a way that Rybicki believes was possible with what he termed the “Protean” nature of the collaboration between Maria Edgeworth and her father, for example, or with the purported influence of Mary Shelley’s parents on her work. Instead, the underlying assumption is that each individual idiolect remains consistent to itself over the lifetime, resisting the attempt of the other idiolect to change it in any measurable way.
[1] “Henry James, on hearing of [the possible collaboration] exclaimed that it was ‘inconceivable’ and that the whole notion was ‘like a bad dream which one relates at breakfast’ (quoted in Demarest 1997).” Quoted in Rybicki, Hoover and Kestemont, p. 423.