Common Library 1.0: A Corpus of Victorian Novels Reflecting the Population in Terms of Publication Year and Author Gender

09/05/2019 ∙ by Allen Riddell, et al. ∙ 0

Research in 19th-century book history, sociology of literature, and quantitative literary history is blocked by the absence of a collection of novels which captures the diversity of literary production. We introduce a corpus of 75 Victorian novels sampled from a 15,322-record bibliography of novels published between 1837 and 1901 in the British Isles. This corpus, the Common Library, is distinctive in the following way: the shares of novels in the corpus associated with sociologically important subgroups match the shares in the broader population. For example, the proportion of novels written by women in 1880s in the corpus is approximately the same as in the population. Although we do not, in this particular paper, claim that the corpus is a representative sample in the familiar sense–a sample is representative if "characteristics of interest in the population can be estimated from the sample with a known degree of accuracy" (Lohr 2010, p. 3)–we are confident that the corpus will be useful to researchers. This is because existing corpora–frequently convenience samples–are conspicuously misaligned with the population of published novels. They tend to over-represent novels published in specific periods and novels by men. The Common Library may be used alongside or in place of these non-representative convenience corpora.



There are no comments yet.


page 11

This week in AI

Get the week's most popular data science and artificial intelligence research sent straight to your inbox every Saturday.

1 Previous Research and Related Data

Population growth, new technologies, and new financial institutions contributed to exponential growth in the rate at which previously unpublished novels (“new novels”) appeared between 1837 and 1901 in the British Isles. At the start of this period roughly 100 new novels appeared each year. By the end of the period, publishers produced over 1,000 new novels every year. In a period during which the population roughly doubled, we witness a tenfold increase in the rate of production of new novels and, as the size of print runs did not decrease, in novel copies generally [294]eliot2012few.111According to eliot2012few, 14,550 book titles (novels and non-novels) appeared in the decade starting with 1800 and 60,812 appeared during the 1890s. Print runs of the most successful books increased by a factor of roughly eight during the period. Given this, a ten fold increase in the number of book copies produced in 1837 versus 1901 seems possible. Much of this increase is likely due to the declining cost of paper and declining costs associated with printing. Steam-powered papermaking and steam-powered printing were widely adopted by mid-century [64]weedon2003victorian[224]raven2007business. Financial institutions also matured, especially during the 1830s. More mature financial institutions further lowered costs to publishers as they made raising money to pay for capital improvements and new publications less expensive [62]weedon2003victorian. An expanding population of readers able to afford access to novels also likely contributed to the growth in new novel production. One number makes the scale of the expansion clear: the median publication year for a Victorian novel is in the mid 1880s. As many new titles appeared during the 16 years after 1884 as appeared during the preceding 47 years.

Literary historians, book historians, and sociologists of print culture have aspired to a fuller view of the population of ca. 25,000 novels—their particular morphology, style, syntax, etc.—as well as of the novelists involved—their lives, social background, professional networks, etc. The scope for learning here, or, alternatively, the extent of literary historians’ ignorance is immense. John Sutherland, the doyen of Victorian literary history, laments the “sheer unavailability of necessary empirical knowledge” for research, adding that “one of the things that makes literary sociology so easy to do at the moment is that we don’t know enough to make it difficult” [558]sutherland1988publishing. Literary historians do not know, for example, how many people pursued careers as novelists in the British Isles during the 19th century [574-575]sutherland1988publishing[151-164]sutherland1995victorian. This ignorance persists despite decades of sustained scholarly activity on the Victorian novel at research universities across the world.

Many researchers find the present situation unsettling, particularly in light of the standard classroom practice of teaching the history of the Victorian novel using a small number of novels by canonical authors [207-210]moretti2000slaughterhouse[27]bode2018world[87]bode2017equivalence. This approach neglects discussion of the broader literary system. In particular, it ignores the range and variety of other novels which readers would have encountered at the same time as they encountered a canonical novel. The standard presentation of the Victorian novel also typically neglects discussion of the range of material and economic forces or the range of intermediaries —e.g., booksellers, reviewers, advertisers, circulating libraries, book clubs—operating in the literary market. Serious discussion of these would require more organized information about the period than literary historians currently have available.

Research of practical value to those interested in a more extensive and democratic account of the novel has tended to come, especially in recent decades, not from literary studies but rather from scholarship allied with book and publishing history. Work with a machine-readable version of the Nineteenth-Century Short Title Catalog (NSTC) in eliot1994patterns and eliot1997patterns is a notable example. eliot1997patterns delivers a time series which describes the total number of books published in London, Oxford, Cambridge, Edinburgh, and Dublin each year between 1801 and 1870. Combined with weekly records of publisher-reported editions in Publishers’ Circular it is possible, as weedon2003victorian shows, to estimate the annual number of books (novels and non-novels) published in the British Isles. Although this series does not directly tell us the number of (new) novels published during the period, it does bound from above the annual number of novels published. And if we are willing to assume the percentage of books which are novels does not change radically from year to year during the 19th century, the time series tells us a great deal about how the rate of novel publication changes over the period.

If the work of Eliot and others gives us a sense of the maximum number of previously unpublished novels appearing each year, bassett2018atcl provides confidence in the minimum number of new novels published each year. Started in 2007, the online database At the Circulating Library (ATCL) contains records of more than 17,000 titles published between 1837 and 1901. (In this paper we use a March 2018 snapshot of the database with 15,322 titles.) Although ATCL aspires to provide an exhaustive list of new novels, it is, as of this writing, incomplete. ATCL’s coverage is not uniform, some years are virtually complete (e.g., 1838) whereas others (e.g., 1900) are perhaps only 60% complete. Even in the incomplete years, ATCL bounds from below the number of new novels. The database tells us the minimum number of new novels which appeared in each year.

Another stream of research, intermittently connected with book and publishing history, has used convenience sampling of digitized novels in work which purports to provide information about the population of novels published during the 19th century jockers2013macroanalysis,jockers2013significant, algee-hewitt2016canon,underwood2018transformation,piper2018enumerations. For example, jockers2013macroanalysis uses a corpus of 3,346 novels and algee-hewitt2016canon uses a corpus of 1,117 novels. All these convenience corpora are gathered opportunistically from texts available from commercial providers or from digitizations of works found in North American and UK libraries. In jockers2013macroanalysis and algee-hewitt2016canon, the researchers admit that their corpora are, in fact, non-probability convenience samples [172]jockers2013macroanalysis[2-3]algee-hewitt2016canon.

2 Methods

The novels which are included in the Common Library are gathered using a probability-proportional-to-size sampling strategy. First, a publication year and author gender pair is sampled according to its share in the population. Previously published estimates of the number of new novels published by year and author gender provide the information we need for this first step riddell2018reassembling. Then a novel with a matching publication year and author gender is sampled uniformly at random from the At the Circulating Library (ATCL) database. If the novel does not have a publicly-available digital surrogate, the sampling is repeated until one with a surrogate is found. Having selected a specific novel, we then randomly sample a chapter and key-in the chapter text. These chapter texts comprise the Common Library. This sampling strategy does not, needless to say, yield a simple random sample from the population. It does not yield a representative sample either. (A sample is representative if “characteristics of interest in the population can be estimated from the sample with a known degree of accuracy” [3]lohr2010sampling.) This is because the sampling frame is restricted to novels with digital surrogates and the ATCL is not, at this point, exhaustive. The strategy does, however, ensure that, with respect to publication year and author gender, shares of novels in the Common Library are aligned with corresponding shares in the population.

The following example illustrates the two-stage sampling procedure. Of the ca. 25,000 novels published during the period, 288 (1.1%) were published in 1894 and written by women authors. The probability of sampling this group in the first stage is 1.1%. Assuming we sampled this group, we would then sample a specific novel at random from the list of novels in the ATCL database which have matching year and author gender until one is located which has a first edition digital surrogate. With a specific novel in hand, we randomly sample a chapter and manually key-in its text.

A novel associated with this particular group (1894, woman-authored) appears in the Common Library. The group was sampled on the 21st probability-proportional-to-size draw and the following novel was sampled from ATCL: Edna Lyall’s To Right the Wrong published in 1894 by Hurst and Blackett. This three-volume novel has 41 chapters and the 25th chapter was randomly sampled for encoding. For reproducibility, we use the novel’s ATCL database identifier as a random seed when sampling the chapter.222For example, using Python 2, random.seed(4705); random.randrange(41) + 1 yields 25 (for chapter 25). The addition of 1 to the sampled integer is required because random.randrange(41) samples an integer uniformly at random from the interval . Since the chapters we are interested in are numbered from 1 to 41 (inclusive), we add 1 to the result.

This section is organized as follows. First we describe in detail the two-stage sampling procedure: (1) the partitioning of the population into subpopulations defined by publication year and author gender and (2) the use of the ATCL database to select a novel for each sampled group.

New novels by year and author gender, 1837-1901.

We begin by sampling a group defined by publication year and author gender. As each novel has a distinct publication year and author gender, these groups partition the population. We sample a group with probability proportional to the number of novels associated with it. For the sizes of the groups we use medians of the estimated counts in riddell2018reassembling (Table 3). Publication year is the year indicated on the title page of the first edition. Author gender is the gender of the historical individual acknowledged as the novel’s author. For pseudonymous and anonymous novels which advertise an author gender on the title page (e.g., “By a Lady of Rank”), we use the advertised gender. Here we follow the convention established by garside2004british in assuming that the gender of the historical individual(s) who wrote the novel is the same as the gender of the advertised author. When no information is available from the title page about an anonymous or pseudonymous author, we record the author’s gender as “unknown”. In the exceedingly rare case of a title with more than one author, we use the gender of the author listed first on the title page of the first edition.

One technical detail concerning the definition of “novel” deserves to be mentioned. There are two definitions of the novel used in large bibliographies of the 19th-century English novel: the descriptive definition found in raven2000english and garside2000english (hereafter “RFGS”) and the definition used by ATCL (hereafter, “ATCL”). A novel according to RFGS is a book described as a “novel” by contemporaries. The definition used by ATCL is more permissive: prose fiction of at least 90 printed pages that is not addressed exclusively to children. These definitions are largely consonant. All books which are novels according to RFGS are novels according to ATCL. Some books considered novels by ATCL are not novels according to the more restrictive RFGS definition. Disagreements are predictable as they tend to concern novel-like books in well-known subgenres. ATCL’s definition permits novel-like works of (didactic) religious and juvenile fiction to be counted as novels; RFGS exclude these books. We discuss these two definitions and list conforming examples in Appendix Appendix: Novel definitions. In the sampling strategy used here, we make the assumption that during the 1837-1901 period any count of novels using the ATCL definition is equivalent to a count using the RFGS definition after increasing the latter by 12.5%. This figure is the midpoint of the estimate that between 10% and 15% of ATCL titles would be excluded from a RFGS bibliography, were RFGS to cover the 1837-1901 period. Future work might revisit this 12.5% figure if, say, men are much more likely to be authors of works which would be excluded by the RFGS standard (e.g., novel-like didactic religious fiction). Although riddell2018reassembling use the RFGS definition of the novel, the sampling probabilities do not change after translating the counts into ATCL terms.

At the Circulating Library Bibliography, 15,322 titles.

Once we have sampled a group of novels defined by publication year and author gender, we sample a title uniformly at random from titles with matching publication year and author gender in the At the Circulating Library (ATCL) database.333 For those interested in reproducing the sample, we provide the random seed and software version used. The SQL command which performs the sampling is the following: SELECT author_id, title_id FROM titles LEFT JOIN authors USING (author_id) WHERE publication_year=? AND gender="?" ORDER BY RAND(1);. MySQL version 5.7.21 was used. We use the March 7, 2018 snapshot of the database. This version includes 15,322 titles published between 1837 and 1901. If the title sampled has no publicly available first-edition page images, we sample another title until we find one which does. In order for a title to be counted as having a first-edition digital surrogate, page images of the first edition must have been available online on or before December 31, 2018 from the Internet Archive, Google Books, HathiTrust, or the British Library. We count other editions published by the first-edition publisher in the same year as the first edition as first editions. These editions include second printings (sometimes labeled “second edition”) as well as export editions of novels, which often feature a variant title page.444 For example, we count a Canadian edition of Swan’s Mrs. Keith Hamilton (1897) as a first edition as the book was printed by the London publisher in the same year as the London edition and appears to be identical, save for an alternate title page. We count a multivolume novel as having a first-edition digital surrogate if all volumes of the first edition have been digitized. For example, the novel mentioned earlier—Lyall’s To Right the Wrong—is a three-volume novel (“triple decker”) with first-edition digital surrogates of all three volumes. (The volumes were digitized in October 2008 by the Internet Archive from originals at the University of Illinois Urbana-Champaign.)

Once we have a title with an available first-edition digital surrogate, we sample a chapter uniformly at random and then manually key-in the sampled chapter. We encode a single chapter to limit the time and resources expended on keying-in texts. We welcome reports of discrepancies between observed characteristics in the chapters encoded and page images of the first edition. If the sampled novel is not divided into chapters, we sample a section at random. (Every novel sampled was explicitly divided into chapter-like sections of some kind.) Non-narrative sections of the text (e.g., backmatter, dedications, non-narrative authors’ prefaces) are not counted as sections. We further annotate the UTF8-encoded “plain text” version of each chapter using a small set of HTML5 tags. For example, we mark italics and paragraph boundaries. We label the resulting collection of 75 novel chapters as the “Common Library”. The version of the collection described here is 1.0. The dataset accompanies this paper.555The filename is The file contents have the following SHA256 hash: caa78440bbc237b1bfa3600dd3b2577263fed3cbde723dbc2adf250a0fae6b58.

Over-representation of multivolume novels and multiple-novel author novels in At the Circulating Library.

The ATCL database aims to be exhaustive but the March 7, 2018 snapshot we use does not contain a record for every published novel during the Victorian period. Although the database has grown primarily by the systematic addition of titles from annual lists of titles published in The English Catalogue of Books, titles also arrive in the bibliography via other routes, resulting in over-representation of certain kinds of novels: multivolume novels and novels by authors who published more than one novel. For example, whereas about 29% of Victorian novels are multivolume novels, 57% of the Common Library titles are multivolume novels. In a future article we will quantify the biases precisely.

Restriction to first-edition digital surrogates.

In addition to being restricted to ATCL titles, our sampling frame is further restricted to titles with first-edition digital surrogates created on or before December 31, 2018. This biases the novels in the Common Library towards novels which were targets for library collection during the 19th and 20th century. The reason for this is that first editions which survive in a greater number of libraries are more likely to be found at a library which digitized its holdings.

We do not know precisely what made novels targets for library collections. Nearly all novels from the period survive in libraries. For example, 98% of titles published in 1838 have a surviving copy. Although only some libraries participated in library digitization, two of the libraries which did, Oxford’s Bodleian and the British Library, tended to receive a copy of every novel published as a consequence of the library deposit requirement introduced by the Copyright Act of 1842. (The Copyright Act’s deposit requirement specified the British Museum, whose library department is now the British Library.) The library deposit requirement should allay concerns that Oxford or the British Library tended to exclude certain kinds of novels, at least after 1842.

Concern about the first-edition digital surrogate limitation introducing unaccounted-for bias into the sample is warranted. Oxford’s Bodleian did not digitize its entire collection and parts of the British Library’s collection were destroyed during World War II. So novels which were targets of collection by other libraries are more likely to have first-edition digital surrogates. This means that novels which were targets for collection are more likely to be in the Common Library than novels which were not targets.666Future work might approach this problem by treating books which have first-edition digital surrogates from the Bodleian or British Library—where legal deposit can be assumed to be the reason the book is present—differently than books which are not available from those two libraries.

For novels associated with a given year and author gender, we believe a promising account of which novels were targets for collection is available. We hypothesize that novels written by novelists who wrote at least one other novel—multiple-novel author titles—are more likely to be targets for collection than novels by single-novel authors. We make use of ATCL’s exhaustive coverage of novels published in 1838 to evaluate this hypothesis. Comparing 1838 novels by authors of the same gender with first-edition digital surrogates to novels without surrogates provides preliminary, non-decisive evidence in support of the hypothesis. Whereas 6 of 9 novels (66%) published in 1838 by single-novel women novelists have first-edition digital surrogates, 24 of 28 novels (86%) by multiple-novel women novelists have first-edition digital surrogates. We suspect that authors of a commercially or critically successful novel was encouraged to write—as they are today—additional novels by bookseller-publishers and other text industry intermediaries. Commercially and critically successful novels were, we believe, targets for collection. These novels would be more likely to be requested by patrons and many libraries do respond to patrons’ demands. With respect to each subpopulation defined by a given publication year and author gender, restricting the Common Library collection to novels with first-edition digital surrogates likely increases the tendency for the Common Library to over-represent multiple-novel authors.

3 Common Library vs. Reprint Canon

Our primary contribution is the 75-novel Common Library corpus, a sample from the population of Victorian novels. With respect to groups defined by publication year and author gender, the shares of novels in the Common Library associated with each group reflect the shares of novels associated with the group in the population.

In this section we offer a superficial comparison of the Common Library with a familiar collection of Victorian novels. This other corpus, which we label the “Reprint Canon”, consists of the 204 Victorian novels which are in print and available from Penguin, Oxford, or Broadview in 2017.777 We use the 2017 catalogs from Penguin (Classics), Oxford (World Classics), and Broadview. Although the set of titles in print does change occasionally, there is typically little or no change from year to year. We study this corpus through the Reprint Canon Sample, a simple random sample from the Reprint Canon. The Reprint Canon Sample contains 88 novels. We encode one chapter, chosen at random, from each Reprint Canon novel using the same procedure we used with the Common Library. We compare the Common Library to the Reprint Canon Sample using a simple feature, length in words. We approximate length by multiplying the number of chapters in each novel by the length in words of the encoded chapter.

The Common Library novels differ from the Reprint Canon Sample in terms of length. The mean length in the Common Library collection is conspicuously smaller than the mean length of Reprint Canon Sample novels (Figure 1). The reason for this is the prevalence of very long novels in the Reprint Canon Sample. Very long novels (> 250,000 words) are more common in the Reprint Canon Sample than in the Common Library. The Reprint Canon Sample has 12 (14%) very long titles. The Common Library, by contrast, has only 3 (4%).

Figure 1: Estimated Novel Lengths in the Common Library () and the Reprint Canon Sample (). Novel length calculated by multiplying the length (in words) of the encoded chapter by the number of chapters.

This difference is suggestive, provided we accept that the Common Library novels do better reflect the diversity of novels published during the Victorian period. The difference should call into question the claim that the novels in the Reprint Canon capture the diversity of prose fiction published between 1837 and 1901. This claim is explicitly or implicitly present in university courses on the Victorian novel. It remains standard classroom practice to teach the history of the Victorian novel using only novels which are part of the Reprint Canon [207-210]moretti2000slaughterhouse[27]bode2018world[87]bode2017equivalence. Although this approach is regarded as neglecting discussion of the broader literary system, it has been difficult to find specific evidence that backs up such a belief. The difference observed here—combined with the assumption that the Common Library novels do better reflect the diversity in the population—supplies this evidence.

4 Conclusion

This paper makes available a collection of 75 novels from the Victorian period which reflects the larger population of novels in terms of publication year and author gender. We label this corpus the Common Library. Saying that the collection reflects the population in terms of publication year and author gender means that, for example, the share of novels in the collection published in the year 1888 by women writers matches the share of such novels found in the population. In other respects, the collection reflects the population poorly. Multivolume novels and novels by authors of more than one novel are over-represented. (In a future publication we will describe how these biases can be adjusted for using post-stratification.) Other publicly available collections of novels from the period plainly do not reflect the population of published novels. One prominent and widely-used collection, Victorian novels in print today (the 204-novel Reprint Canon), over-represent novels by men writers and novels published during the 1860s. The Common Library, by contrast, better reflects the diversity of novels published during the Victorian period.


AR and TJB planned the research and wrote the paper. All individuals contributed encoded texts. AY and AR organized and reviewed the encoded texts.

Appendix: List of Novels in the Common Library Collection

Last Name Title Year Gender Single-Novel Author Multivolume
15732 Nisbet Comrades of the Black Cross: A… 1899 M no no
1237 Curtis Favourite of Fortune 1886 W no yes
7083 Gleig The Hussar 1837 M no yes
4847 Churchward Jem Peterkin’s Daughter: An… 1892 M no yes
11712 Brooks The Naggletons, and Miss Violet and… 1875 M no no
9982 Adams The White Brunswickers: or,… 1865 M no no
951 Gibbon The Braes of Yarrow: A Romance 1881 M no yes
15354 Frith In Search of Quiet: A Country Journal 1895 M no no
14740 Prevost Entanglements 1898 M no no
7173 Cummins El Fureidis: A Tale 1860 W no yes
8550 Boothby Pharos the Egyptian 1899 M no no
1053 Alexander The Admiral’s Ward 1883 W no no
1519 Lyon The Signora: A Novel 1883 M no yes
10236 Green The Heir of Hascombe Hall: A Tale of… 1900 W no no
7155 Thackeray Vanity Fair: A Novel without a Hero 1848 M no no
1120 Booth Fragoletta: A Novel 1881 W no yes
11919 Clifford In a Corner of Asia: Being Tales and… 1899 M no no
502 Murray John Alston’s Vow: A Tale 1865 W no yes
2534 Knight A Romance of Acadia Two Centuries Ago 1874 M no yes
1453 Jenkins Jobson’s Enemies 1882 M no yes
4705 Lyall To Right the Wrong 1894 W no yes
6459 Peregrinator A Flight to Florida: And All That… 1888 U yes yes
7900 Terrell The City of the Just 1892 M no no
12419 Marshall Constantia Carew: An Autobiography 1883 W no no
7741 Alexander Through Fire to Fortune 1900 W no no
12089 Croker Angel: A Sketch in Indian Ink 1901 W no no
14640 Hood A Disputed Inheritance: The Story of… 1863 M no no
3583 James The Fate: A Tale of Stirring Times 1851 M no yes
2360 Hill They Were Neighbours: A Novel 1878 M yes yes
13760 Conyers The Thorn Bit 1900 W no no
552 Platt Angelo Lyons 1866 M no yes
8440 Raymond Two Men o’ Mendip 1899 M no no
956 Gibbon The Golden Shaft 1882 M no yes
7143 Power Nelly Carew 1859 W no yes
1523 MacEwen Miss Beauchamp: A Philistine 1883 W no yes
4783 Sherer Alice of the Inn: A Tale of the Old… 1893 M no yes
1012 Marryat Peeress and Player: A Novel 1883 W no yes
8165 Swan Mrs. Keith Hamilton, M.B.: More… 1897 W no no
4228 Peake Cartouche, the Celebrated French Robber 1844 M yes yes
1359 Gray The Reproach of Annesley 1889 W no yes
15770 Chetwode The Knight of the Golden Chain 1898 U no no
1289 Fane The Story of Helen Devenant 1889 W no yes
4651 Gissing The Emancipated: A Novel 1890 M no yes
6565 Coleridge Waynflete 1893 W no yes
424 Harwood Lord Ulswater: A Novel 1867 M no yes
8557 Boothby Long Live the King 1900 M no no
4040 BTAO Sin and Sorrow Sin and Sorrow: A Story of a Man of… 1850 U no yes
4844 Castle Consequences: A Novel 1891 M no yes
6797 Marston James Vraille: The Story of a Life 1890 M yes yes
2623 Payn A Woman’s Vengeance: A Novel 1872 M no yes
4383 Henningsen The White Slave: or, The Russian… 1845 M no yes
1256 Diehl Eve Lester: A Novel 1882 W no yes
7951 James Tales of Three Cities 1884 M no no
1727 Russell Marooned: A Novel 1889 M no yes
7447 Brontë Agnes Grey: A Novel 1847 W no no
8031 Grant Frank Hilton: or, "The Queen’s Own" 1855 M no no
12062 Craik Cousin Trix and her Welcome Tales 1868 W no no
2551 Mackenna A Child of Fortune 1875 M no yes
14966 Hearn The Cathedral’s Shadow 1871 W no no
1278 Edwards Pharisees: A Novel 1884 W no yes
5683 Spicer Bound to Please 1867 M no yes
8532 Boldrewood A Romance of Canvas Town and Other… 1898 M no no
12957 Shipton Bearing the Yoke 1884 W no no
11917 Clifford Studies in Brown Humanity 1898 M no no
7430 Ainsworth The Star-Chamber: An Historical Romance 1854 M no yes
3949 Chapman Mary Bertrand 1860 W no yes
2231 Hardy Friend and Lover 1880 W no yes
4217 Pardoe Reginald Lyle 1854 W no yes
5152 Anonymous A Heart Well Won: or, The Life and… 1874 U yes yes
539 Parr Dorothy Fox 1870 W no yes
4026 Anonymous The Prince of Orange: A Tale of the… 1845 U yes yes
2664 Reade A Simpleton: A Story of the Day 1873 M no yes
7437 Yates Two By Tricks: A Novel 1874 M no yes
9550 Phillips The Birth of a Soul: A Psychological… 1894 W no no
8371 Fletcher Pasquinado 1898 M no no
Table 2: Novels in the Common Library 1.0 corpus. Note: “BTAO” stands for “By the Author of.”

Appendix: Novel definitions

Two definitions of the novel are used in large bibliographies of the 19th-century English novel:

  • “RFGS”: a descriptive definition found in raven2000english and garside2000english, and

  • “ATCL”: a more permissive definition than RFGS which includes all prose fiction of at least 90-printed pages that is not addressed exclusively to children.

RFGS adopts “a more rigorous definition” of the novel than previous scholars. Their bibliography “includes what contemporaries thought of as novels, incorporating works categorized as ‘novels’ in contemporary periodical reviews and under ‘novels’ headings in circulating library catalogues, but excludes religious tracts, chapbooks, literature written only for children and juveniles, and very short separately issued tales.” They add the qualification: “Collections of tales (including some mixed genre compilations) are included; separate verse novels are not” [4]raven2000english.

garside2004english also uses this definition of the novel.

ATCL uses a more inclusive definition of the novel. ATCL includes any prose fiction of at least 90-printed pages but excludes any work aimed exclusively at children (younger than 12 years of age). Hence, religious tracts or literature written for juveniles (12 years of age or older) is included if it consists of prose fiction and meets the minimum length requirements. These criteria are applied with considerable consistency. We believe that exceptions (e.g., an 89-page novel), if any do appear in the database, occur in fewer than 1 in 200 records.

The works of author Eliza Paget offer an illustrative example in the application of these two definitions. Paget wrote ten prose fiction works between 1829 and 1841, all of which are religious in theme and aimed at juvenile and adult readers. Of those published before 1836, the bibliographies using the RFGS definition exclude all seven titles, categorizing them as juvenile fiction. Of those published after 1837, ATCL includes all three titles since they are all prose fiction and meet the minimum length requirements.

Appendix: Probability-proportional-to-size group sizes

Man-authored Woman-authored Unknown
1837 49 (0.2%) 32 (0.1%) 14 (0.06%)
1838 49 (0.2%) 33 (0.1%) 14 (0.06%)
1839 52 (0.2%) 36 (0.2%) 15 (0.07%)
1840 51 (0.2%) 36 (0.2%) 15 (0.07%)
1841 52 (0.2%) 38 (0.2%) 16 (0.07%)
1842 54 (0.2%) 40 (0.2%) 17 (0.08%)
1843 54 (0.2%) 42 (0.2%) 17 (0.08%)
1844 56 (0.3%) 44 (0.2%) 18 (0.08%)
1845 56 (0.3%) 45 (0.2%) 19 (0.09%)
1846 59 (0.3%) 49 (0.2%) 20 (0.09%)
1847 66 (0.3%) 55 (0.2%) 23 (0.1%)
1848 69 (0.3%) 59 (0.3%) 25 (0.1%)
1849 71 (0.3%) 62 (0.3%) 26 (0.1%)
1850 77 (0.3%) 70 (0.3%) 29 (0.1%)
1851 81 (0.4%) 74 (0.3%) 31 (0.1%)
1852 82 (0.4%) 76 (0.3%) 32 (0.1%)
1853 83 (0.4%) 80 (0.4%) 33 (0.1%)
1854 87 (0.4%) 84 (0.4%) 35 (0.2%)
1855 85 (0.4%) 84 (0.4%) 34 (0.2%)
1856 82 (0.4%) 82 (0.4%) 33 (0.1%)
1857 84 (0.4%) 85 (0.4%) 34 (0.2%)
1858 84 (0.4%) 86 (0.4%) 34 (0.2%)
1859 87 (0.4%) 91 (0.4%) 35 (0.2%)
1860 90 (0.4%) 95 (0.4%) 37 (0.2%)
1861 91 (0.4%) 95 (0.4%) 35 (0.2%)
1862 92 (0.4%) 96 (0.4%) 35 (0.2%)
1863 93 (0.4%) 98 (0.4%) 34 (0.2%)
1864 98 (0.4%) 102 (0.5%) 35 (0.2%)
1865 105 (0.5%) 113 (0.5%) 38 (0.2%)
1866 105 (0.5%) 111 (0.5%) 35 (0.2%)
1867 108 (0.5%) 114 (0.5%) 34 (0.2%)
1868 114 (0.5%) 121 (0.5%) 35 (0.2%)
1869 115 (0.5%) 123 (0.6%) 34 (0.2%)
1870 118 (0.5%) 122 (0.5%) 31 (0.1%)
1871 124 (0.6%) 136 (0.6%) 34 (0.2%)
1872 124 (0.6%) 138 (0.6%) 33 (0.1%)
1873 130 (0.6%) 147 (0.7%) 33 (0.1%)
1874 127 (0.6%) 146.5 (0.7%) 32 (0.1%)
1875 133 (0.6%) 162 (0.7%) 33 (0.1%)
1876 142 (0.6%) 168 (0.8%) 33 (0.1%)
1877 150 (0.7%) 177 (0.8%) 34 (0.2%)
1878 158 (0.7%) 186 (0.8%) 35 (0.2%)
1879 171 (0.8%) 202 (0.9%) 36 (0.2%)
1880 175 (0.8%) 203 (0.9%) 37 (0.2%)
1881 180 (0.8%) 204 (0.9%) 35 (0.2%)
1882 182 (0.8%) 204 (0.9%) 34 (0.2%)
1883 207 (0.9%) 226 (1.0%) 38 (0.2%)
1884 220 (1.0%) 231 (1.0%) 38 (0.2%)
1885 209 (0.9%) 218 (1.0%) 34 (0.2%)
1886 219 (1.0%) 215 (1.0%) 36 (0.2%)
1887 240 (1.1%) 226 (1.0%) 38 (0.2%)
1888 269 (1.2%) 244 (1.1%) 42 (0.2%)
1889 269 (1.2%) 236 (1.1%) 41 (0.2%)
1890 290 (1.3%) 239 (1.1%) 41 (0.2%)
1891 281 (1.3%) 229 (1.0%) 41 (0.2%)
1892 311 (1.4%) 247 (1.1%) 45 (0.2%)
1893 328 (1.5%) 251 (1.1%) 48 (0.2%)
1894 343 (1.5%) 256 (1.1%) 50 (0.2%)
1895 373 (1.7%) 277 (1.2%) 55 (0.2%)
1896 370 (1.7%) 263 (1.2%) 55 (0.2%)
1897 418 (1.9%) 292 (1.3%) 63 (0.3%)
1898 422 (1.9%) 289 (1.3%) 64 (0.3%)
1899 438 (2.0%) 293 (1.3%) 67 (0.3%)
1900 434 (1.9%) 283 (1.3%) 69 (0.3%)
1901 423 (1.9%) 275 (1.2%) 67 (0.3%)
Total 10,559 (47%) 9,436.5 (42%) 2293 (10%)
Table 3: Counts of new novels published by year and author gender used in the probability-proportional-to-size sampling strategy. We use the medians of the estimated counts in riddell2018reassembling. Counts shown here use the RFGS definition of the novel, not the ATCL definition of the novel.