Monday 12 March 2018

Genre Labels and the Rise of the Novel

I have recently been reading Leah Orr’s Novel Ventures: Fiction and Print Culture in England, 1690-1730 (2017). In her book, Orr frequently references an earlier article of hers, “Genre Labels on the Title Pages of English Fiction, 1660-1800” (Philological Quarterly, 90, No. 1 (2011): 67–95), so I thought I would read it too.

The title of Orr’s article is admirably clear: she sets out to examine and enumerate the different labels that appear on prose fiction published over a 140–year period, a subject that allows her to discuss various definitional problems relating to the “Rise of the Novel,” previous and present understanding of the key terms and so forth.

Although Orr sometimes understates the achievement of her predecessors in relation to her study, the project and her summary of the data collected is very useful, as is her clear methodology, and analysis. Orr provides a detailed summary of her data, which is also admirable, but she does not provide access to her full data set—which is a little frustrating.

While reading her article I was struck by the fact that Orr’s analysis is based exclusively (it seems) on the predominance of certain key-terms in the imprints of the type of prose fiction (often referred to as novels or proto-novels) in each decade, and the pattern of decade-to-decade changes.

While this sort of big-picture simplification is absolutely necessary to summarise results, and to identify broad trends over a 140-year period, Orr does not use her full data set to examine any of the key moments she has identified in detail. So, for instance, her analysis treats, the fiction market of 1731 and 1739, or 1741 and 1749, the same—both being subsumed under a rubric of the “1730s” or “1740s.”

The loss of year-on-year perspective is important because it conceals the fact that readers, writers, publishers, critics and so forth, in 1739 or 1741, would have conceptualised what they were reading in genre terms—the “labels” that Orr is investigating—in ways that were neither uniform across a decade, or determined solely by the labels used within that decade. Rather than being labels of the moment/decade, these labels were likely understood by individuals based on their accumulated prior knowledge of the genre label (however defined, and however comprehensive that might be).

Accumulated prior knowledge is obviously highly individualised. For younger readers, “prior knowledge” may be limited to works only recently published; the prior knowledge of older readers of contemporary fiction may stretch back through five or more decades. Then again, a younger reader may prefer older texts, just as a selective reader may have read only a very narrow range of fiction from a long period. And most readers, certainly once we get to 1800, would have only read a small percentage of the fiction published, however that fiction was chosen.

Orr uses a series of averages, represented as snapshots of decade-long “moments.” Other approaches could have been used to Orr’s data in an attempt to recover a more-or-less simplified and generalised short- or a long-(backwards)-view. There are various methods that might be employed to represent the personal habits and perspectives of individual readers—the easiest of these would probably have been to calculate rolling-averages, whether three, five, ten or more years, depending on assumptions made about different types of novice and experienced readers. Since Orr does not publish her full data-set, this sort of analysis can not be done by any of her readers.

Looking at Orr’s summary of her data, it occurred to me that I could, however, use it to generate cumulative averages, which would allow for a, decade-by-decade, long backwards view. That is, by adding up the totals from all previous decades, for each term in each decade, I could capture all the instances of every key genre term that had appeared on the title-page of any work of fiction since 1660. So, for instance, in 1670, this would be a decade-long backward view, in 1680 a two-decade-long view, in 1690 a three-decade-long view, and so forth.

By 1740, Orr’s imaginary (and near-comprehensive) library of fiction would have included 731 works, published in the eighty years since 1660. In 1740, even the most diligent reader, writer, publisher, or critic was unlikely to have read all 731 of these works. Then again, such an imaginary reader would also be very unlikely to have limited their reading only to works printed in the 1730s alone (as Orr’s analysis implies). Consequently, neither the comprehensive totals (which I discuss below), nor the decade-by-decade totals, offer a satisfactory substitute for the accumulated prior knowledge of an individual reader, writer, publisher or critics.

Individual readers, writers, publishers, and critics in 1740, would have used and understood genre labels in highly individualised ways, the result of exposure to an uneven mix of works from a number of decades. Depending on the assumptions one is prepared to make, there may be ways of representing some of this accumulated prior knowledge of such individuals in a generalised form.

So, for instance, if we assume that the most recent fiction looms largest in the minds of readers, writers, publishers, and critics, and that no reader reads every new work of fiction, we might use a linear regression over three- or four decades to represent the fading importance of older works in explaining how genre labels were used and understood by an average adult reader. Taking 1740 as an example, this sort of approach we might add 20% of the figures for the 1690s to 40% for the 1700s to 60% for the 1710s and 80% for the 1730s (i.e., a twenty percent decline per decade, over four decades). An exponential regression might better capture this fading-recollection or -significance approach; we might add something like 20% of the figures for 1700s to 40% for the 1710s to 80% for the 1730s (i.e., a doubling of the twenty percent per decade for three decades).

Yet another—more nuanced, but somewhat fanciful—possibility, in an attempt to better capture the average reader, would be to attempt to establish the relative significance of important and influential periods of fiction in the minds of readers, writers, publishers, and critics. This might be done by weighting decades according to what was actually most often read in the perios, based on reading records.

n my article on the New York Society Library, based on reader records from ca. 1790, I established that there was an uneven distribution of popularity by decade (see article here). Once the borrowings-per-volume in the earliest reading ledger were totaled and distributed by decade, it is clear just how much recent titles dominate borrowings, with almost sixty percent of all volumes borrowed between July 1789 to April 1792 being from the 1780s, far ahead of works from the 1750s, 60s and 70s, at approximately ten percent each.

If we adjust for multi-volume works, however, the number of borrowings-per-work is actually highest (narrowly) for the 1750s, followed by the 1780s (mostly due to borrowings of works from the second half of the 1780s). Although works from the 1740s make up less than six percent of all borrowings, the average borrowings-per-work for titles from this decade is high—higher than titles from the 1760s and 70s. Titles from the 1750s were slightly more popular than the most recent novels on offer, showing that many older British novels retained their appeal ca. 1790 in New York.

What this suggests is that, one way to attempt to better understand how New York readers, writers, publishers, and critics used and understood genre labels, ca. 1790, would be to give greatest weight to the terms used in the 1750s, followed by the 80s, 40s, 70s, 60s, 30s, 10s, and so forth, in that order. Of course, the New York Society Library ledger is an extremely unusual survival, and it would be hard to replicate it to produce the sort of nuanced representation of the varying significance of works of fiction from different decades post-1660.

I mention these alternatives, not to criticise Orr for not attempting such either a linear, exponential or a nuanced and idiosyncratic approach, but to suggest the limitations of only examining the pattern of decade-to-decade totals for genre labels.

* * * * *

   Since my only option for an alternative analysis to that offered by Orr was to calculate cumulative averages based on an imagined comprehensive-accumulation of fiction, this is what I did. Having extracted the data from Orr’s table, I used Excel to calculate the cumulative averages for each term, organising the terms based on the final 140-year view from 1800 (left-to-right being highest-to-lowest). To make it a little easier to read the table, I used highlighter to emphasise the top two terms within the decade and in the cumulative averages. The full table is below; the most important three columns are at the end of this post.

What emerges from the cumulative averages is a strengthening of Orr’s argument, that “The Rise of the Novel” is a particularly anachronistic (as well as critically vague) phrase. As you can see above, although both “History” and “Novel” are the most-frequently used terms in an equal number of decades, and “Novel” is the most frequently-used term across the 140-year period she examines (as Orr states), “History” is the top (or co-top) term in twelve of fourteen decades, while “Novel” is only the top term in three of the fourteen. Moreover, “History” is the top term across the first 130-years of the 140-year period; it is only the massive rise in the number of novels, and the predominance of the term “Novel” in the 1790s, that displaced “History” as the top label.

The massive rise in the number of novels in the 1790s is indicated by another set of figures I generated. When expressed as a percentage of all novels printed across the entire 140-year period, those of the 1790s amount to almost a quarter (23%) of the total—almost the same figure for the first eight decades (24%). The near-exponential rise in the number of fiction titles published is indicated by the (roughly) one-quarter markers (in blue): 1660–1739 (first 25%; 80yrs), 1740–69 (second 25%; 30yrs), 1770–89 (third 25%; 20yrs), 1790–99 (fourth 25%; 10yrs).

(Interestingly, these figures suggest that there is an exponential-regression is built into the un-adjusted, cumulative figures for the use of genre labels, since any backward view that includes all previous fiction, will necessarily give the greatest numerical significance to the most recent works.)

So, while the near-exponential rise in the number of fiction titles published, paired with a late-rise in the use of the term “Novel,” make it the most frequently-used term across the 140-year period. “History” dominated from 1700 to 1790, is the top term for all but one decade of the period 1660 to 1790, and in that one decade, missed out by only one percentage point from being the top term for the first 130-years of “The Rise of the Novel History.”