Modes of Composition in Henry James: Dictation, Style, and What Maisie Knew

A hundred things hummed at the back of her head,
but two of these were simple enough. Mrs. Beale
was by the way, after all, just her stepmother and
her relative. She was just—and partly for that very
reason—Sir Claude’s greatest intimate (‘ladyintimate’
was Maisie’s term) so that what together
they were on Mrs. Wix’s prescription to give up and
break short off with was for one of them his particular
favourite and for the other her father’s wife.
Henry James, What Maisie Knew (1897)
The difference between the styles of Henry James’s
early and later novels is well-known. No author’s
style is likely to remain constant over a time-span as long
as that between James’s first novel in 1871 and his last
complete novel in 1911, but the change in James’s style
is so extreme that many readers and critics consider the
later novels too obscure and enjoy only the early novels.
Others champion the later novels and consider the early
novels immature. One critic notes that James’s late ‘distortions’
often ‘obliterate the normal elements of connection
and cohesion. When he has undone the usual
ties, his meanings float untethered, grammatically speaking,
like particles in colloidal suspension’ (Short 1946:
73-4). Although most attention has been paid to syntax,
changes in James’s vocabulary are equally interesting,
and the frequencies of the most frequent words alone
very effectively identify when each novel was written
(Hoover 2007).
One often-mentioned reason for the change in James’s
style seems important from a literary perspective and
is especially amenable to and appropriate for computational
investigation: In 1896, while writing What Maisie
Knew, James began dictating because of persistent pain
in his wrist. His most famous typist believed that ‘The
different note [of the later novels] was possibly due more
to the substitution of dictation for pen and ink than to any
profound change of heart’ (Bosanquet 1924: 254). Edel’s
biography of James links the elaborate and convoluted
late sentences with the change in mode of composition
and reports that some of James’s friends claimed to be
able to tell which chapter in Maisie was the first one he
dictated (1969: 176-77). Recent interest in media and
technology, especially at the turn of the twentieth century,
has given new significance to the claim (Cappello
2007; Thurschwell 2001; Seltzer 1992). The fact that it is
mentioned by Marshall McLuhan (McLuhan & Zingrone
1995), and that it appears in Barron’s Notebooks (a study
guide for students) (1986), the online guide to the Dragon
Naturally Speaking dictation-to-text program (Newman
2000), an article on voice recognition in composition
(Honeycutt 2003), Wikipedia (2008), The Ivanhoe
Game (Bethany 2002), and the fiction of Cynthia Ozick
(2008), shows its ubiquity, appeal, and relevance to digital
media and digital humanities.
The fact that James’s later style is generally dated to the
late 1890s, just when he begins dictating, has undoubtedly
encouraged this idea. Despite of the widespread
currency of the belief that dictation significantly affected
James’s style, I know of no attempt to present any
solid evidence for it. Yet there is strong evidence for
the gradual and unidirectional development of James’s
style over his career that casts doubt on any significant
effect of the change to dictation. Consider the extraordinary
pattern of adoption of characteristically late words
and abandonment of characteristically early words over
time shown in Fig. 1. To produce this figure, I created
combined word frequency lists for the seven earliest
and eight latest novels, then selected all words that are
at least three times as frequent in the early novels as in
the late and vice versa. I calculated the percentage of
tokens represented by early and late words in a selection
of novels and novellas throughout James’s career. Note
how regularly the frequency of early words declines and
late words increases. These trends continue through the
middle period and include novels and novellas that had
no part in the creation of the word lists. The later style
is associated more with syntax than with vocabulary, but
the regularity of the vocabulary changes certainly offers
no support for an abrupt shift between the handwritten
novels that precede and the dictated novels that follow
Maisie.
Now let us examine Maisie itself for evidence of a significant
local shift in style caused by the switch to dictation.
Evidence from the most frequent words is shown
in Fig. 2. To produce this figure, I created separate files
for the chapters and analyzed them with cluster analysis;
to make the graph more readable, I labeled the chapters
‘pre’ (preliminary)-‘ten,’ ‘11’-‘20,’ and ‘twenty-1’-
‘thirty-1’. The two main clusters in Fig. 2 do not strongly
reflect the structure of the novel: most early chapters are
in the left cluster and most middle and later ones in the
right cluster, but at least three chapters from each group
appear in each. Among the dozens of novels I have analyzed,
Maisie is unusual in how weakly its narrative structure is reflected in the relationships among the chapters
(plot is not James’s major interest). Analyses based
on the 100, 200, 300, 400, 500, 600, 700, 800, 900, and
990 MFW all fail to suggest any change-point associated
with James’s adoption of dictation. Consider some other statistics by chapter (see Fig.3).
Sentence-length is an obvious choice, given the standard
view of James’s style, and, predictably, it is closely related
to the percentage of narrative. Some variables sharply
increase and decrease, but without any obvious changepoint.
Furthermore, average sentence length trends irregularly
lower throughout the novel, as does the percentage
of narrative, suggesting, unsurprisingly, that James
uses shorter sentences in dialogue, but conflicting with
the idea that dictation caused long, rambling sentences.
As Fig. 4 shows, average word length and the frequency
of long words also trend lower, but there is nothing to
suggest a change caused by dictation. The next two figures show the distribution of punctuation
across the chapters. There are obvious correlations
between some punctuation marks and the percentage
of narrative, and some surprising correlations (why are
there more commas in narrative?). Nothing, however,
suggests any shift caused by a change in the mode of
composition. Finally, Fig. 7 displays vocabulary richness, sentencelength,
and the percentage of short words in James’s
novels over time, and Fig. 8 shows just how regularly
the frequency of frequent words characterizes James’s
chronological development. The patterns shown in these
graphs are difficult to reconcile with any significant stylistic
change caused by James’s adoption of dictation in
1896. (The slightly anomalous position of The Outcry in
Fig. 8 may be related to the fact that it is a novelization
of a play.) In conclusion, the gradual, unidirectional nature of the
changes in James’s style does not provide support for
any radical alteration caused by dictation. As reasonable
as this idea has seemed, and as productive as it has
been for speculation about media, machines, and literary
production, such speculation may have to be revised or
abandoned in the light of new kinds of evidence–evidence
only available by computational examination of
a corpus of James’s work. If further research confirms
these preliminary results, computational stylistics may
have an important contribution to make to literary studies
in this case. References
Barron’s Notebooks. The Turn of the Screw. (1986). Barron’s
Educational Series, Inc. Online. Available: http://
www.pinkmonkey.com/booknotes/barrons/turnscr2.asp.
Accessed 11/6/08.
Bosanquet, Theodora. (1924). Henry James at Work.
London: Hogarth Press.
Cappello, Mary. (2007). Awkward: A Detour. New York:
Bellevue Literary Press.
Wikipedia. (2008). Henry James. Online. Available:
http://en.wikipedia.org/wiki/Henry_James.Accessed
11/4/08.
Edel, Leon. (1969). Henry James: The Treacherous
Years, 1895-1901. Philadelphia, Lippincott.
Honeycutt, Lee. (2003). Researching the use of voice
recognition writing software, Computers and Composition
20(1): 77-95.
Hoover, David L. (2007). Corpus Stylistics, Stylometry,
and the Styles of Henry James, Style 41: 174-203.
Bethany [Nowviskie]. (2002). The Ivanhoe Game,
‘The Turn of the Screw.’ Online. Available: http://
speculativecomputing.org/greymatter/ivanhoe/roles/archives/
00000019.htm Accessed 11/4/08.
James, Henry. (1897). What Maisie Knew. New York:
Stone. Online. Available:
http://ia331316.us.archive.org/0/items/whatmaisieknew00jamerich/
whatmaisieknew00jamerich_
djvu.txt McLuhan, Eric, and Frank Zingrone. (1995). Essential
McLuhan. Concord, Ont.: House of Anansi.
Newman, Dan. (2000). The Dragon Naturally Speaking
Guide, 2nd ed. Berkeley: WavesidePublishing. Online.
Available: http://lib.store.yaoo.net/lib/sayican/onlinebook.
html. Accessed 11/6/08.
Ozick, Cynthia. (2008). Dictation: A Quartet. New York:
Houghton Mifflin.
Seltzer, Mark. (1992). Bodies and Machines. New York:
Routledge.
Short, R. W. (1946) The sentence structure of Henry
James, American Literature, 18(2): 71-88.
Thurschwell, Pamela. (2001). Literature, Technology
and Magical Thinking, 1880-1920. Cambridge: Cambridge
University Press.