Spineless Books.

Mechanics, Kinematics, Constraint: Harry Stephen Keeler


William Gillespie

Keeler Who?

Keeler Graphical Bibliography

As tempting as it is to describe Harry Stephen Keeler in terms of Harry Mathews, Agatha Christie, Raymond Queneau, Ann Quin, Roland Barthes, Italo Calvino, Thomas Pynchon, or Stephen King, such comparisons are ultimately misleading—there's really nobody like Keeler. Keeler is an individual, an eccentric, a believer, driven by the click of a peculiar well-made box.

The impression of Keeler I get from reading his fiction is one of exuberant naïveté. He seemed to write transgressive detective fiction without a sense of subversion, irony, or radical literary experimentation, and it is perpetually unclear whether he meant his writing to be as wonderfully funny as it is. Francis Nevins, who, in 1970, was the first to publish an article declaring Keeler to be a writer worthy of study (in the Journal of Popular Culture), suspects that an unhappy childhood led to emotional withdrawal and the construction of a meticulous fantasy universe that his life's work attempted to express. Harry Stephen Keeler was born in 1890, the same year as Agatha Christie. He stayed in Chicago his entire life. His father died when he was eighteen months old. His mother's second husband gambled away the family money, mortgaged the family home, then committed suicide. To make ends meet, his mother turned their home into a boarding house. Her third husband died three years after they married; Harry was 13. In high school, Harry worked as a paperboy. He received a degree in electrical engineering from the Armour Institute of Technology, and worked both as an electrician and in a steel mill.

His writing life apparently began at age 20. He sold his first story in 1913. The following year his literary output increased, and he sold a story for $10, which, according to Nevins, allowed him to pay four weeks of back rent and take Hazel Goodwin, his future wife, out to dinner (with wine) and a show (great seats). From 1919-1940, he served as an editor of the magazine 10 Story Book. In 1919 he married Hazel.

In 1917 and 1926 he authored or co-authored a series of magazine articles explaining his elaborate methods. The main series in 1926 was called The Mechanics and Kinematics of Web-Work Plot Construction.

After accumulating success as a short story writer, he sold his first book in 1924: The Voice of the Seven Sparrows. While this novel is already almost unreasonably complicated, it is mild compared to what would come. Harry continued to publish books of increasing size, implausibility, and density. The Box From Japan (1932) is 762 pages, set in smaller type than Gravity's Rainbow. By 1935 his web-work constructions could no longer fit in books, and a 350,000 word trilogy was published as The Marceau Case (1936), X. Jones of Scotland Yard (1936), and The Wonderful Scheme of Mr. Christopher Thorne (1937). Perhaps Keeler's experiences crafting serial fiction for magazines was useful in composing serial novels. I suspect though that his plots were too bizarre and fast-paced to generate much actual suspense.

The volume of Keeler's output, especially during this period, was staggering, though offset somewhat by his tendency to recycle his earlier short fiction by including it in his novels, usually by having a character who was a writer present a story to another character, then including the entire text of the story. In Thieves' Nights (1929) he pulls a daring metafictional maneuver by having stories within a story within a story, in which a character from the innermost stories unexpectedly appears in the outer story.

From 1939-1942, Keeler wrote and published two simultaneous meganovels, one released as a trilogy and the other as a tetralogy. At the same time, he began to publish a third collection of six novels joined by only by an element that appears in each: a (fictional) book of Chinese aphorisms entitled The Way Out.

One year after the publication of The Peacock Fan (1941), with its bizarre and paranoid portrayal of publishers, he was dropped by his publishing house Dutton. This was the beginning of the end, professionally speaking. He moved to the much smaller Phoenix press, who published nine shorter novels until they dropped him in 1948. His English publisher Ward Lock continued for a few years to publish books now unavailable in North America. Then they dropped him. He continued to publish novels in Spain (unavailable in English, the language in which Keeler wrote them), as translated by the heroic Fernando Noriega Olea. When Instituto Editorial Reus of Madrid, his Spanish publisher, dropped him, Keeler's next two books were available only in Portugal. During this time his novels became less dense with events, less webby, less fraught with implausibly dense strings of coincidences, but the characters and situations continued to become stranger and arguably more tasteless. Many of these novels concern a recurring character: Angus MacWhorter, leader of the Greatest Little Circus on Earth, a character originally created by his wife Hazel Godwin Keeler (herself a writer) in a story published in 1930.

Not being published in any language didn't seem to slow down Keeler's output. It was the death of Hazel, his wife and collaborator, in 1960, that stopped him. The following three years must have been dark ones. But in 1963 Keeler remarried. With the support (and collaboration) of Thelma Rinaldo Keeler, Harry began to continue to finish new manuscripts. In 1965 he completed the triumphant, inscrutably weird masterpieces Strange Journey and The Scarlet Mummy. Keeler was devoted to both his wives and collaborated with them, including (and crediting to them) their writing in his own novels. To Keeler scholars, Hazel and Thelma's writing lacks the strangeness, charm, and relentless inappropriateness of Keeler's prose, but this scholar finds Harry's tendency toward collaboration with his better halves fun, radical, subversive, feminist, and touching.

Keeler died, unpublishable, prolific, and happily married, in 1967, leaving over 70 complete books, and twelve unfinished manuscripts. The original editions are very difficult to find, but all of Keeler's novels, including those never published in English, are now being brought back into print by the plucky Ramble House (ramblehouse.com). Where to start? For beginners: The Voice of the Seven Sparrows (1922). For the more intrepid, The Skull of the Waltzing Clown (1934). For the totally cocky, perhaps the Marceau Case trilogy (1936-1937) (now available from Ramble House as a six-book set) or Strange Journey (1965). Or, if you want to know more about Harry, Francis Nevins claims that The Mysterious Mr. I (1938)/The Chameleon (1940) (one novel in two volumes) is the most personal, The Man Who Changed His Skin (1959) is the best novel to understand Keeler's problematic obsession with race, and The Face of the Man From Saturn (1933) is the most political. As to which novel is the strangest, to my knowledge nobody has yet ventured a guess.

My theory about Keeler (theories abound) is that he developed a mechanical fiction-writing procedure, originally intended to produce marketable detective fiction with a "plot feel," and then discovered that his procedure, if pushed, could produce unique and unmarketable fiction with a truly peculiar feel.

Constraint and Fiction

My interest in Keeler grows out of a study of form and constraint, and can be traced back to a serial poetry technique based on 12-Tone Music called 20 Consonant Poetry I designed to challenge creative writing students. My investigation of form and constraint in language started with techniques that used as their formal basis the alphabet. The lipogram, the palindrome, the anagram are methods of constraining and filtering and shaping language based on the spellings of words.

From the alphabet I proceeded up the layers of language to structure words, phonetics (a poem with only one vowel sound), grammar, and on into the elements of story.

Moving from poetry into fiction, formal mechanics get slippery in that it becomes progressively harder to empirically verify a structural technique. How might one formally structure point of view, character development, or motivation, or relationships? How can a ludic constraint be also lucid?

Canonical works of constrained novels seem to fall into two catgories: constraints that are alphabetic and not narrative (A Void, Alphabetic Africa), or novels known to be constrained but the constraints themselves are unknown, as with many of the novels of Harry Mathews or Raymond Queneau. With Raymond Roussell, strange narratives are yielded by puns, so his most celebrated constraints are phonetic, not narrative. Gilbert Sorrentino's Gold Fools is written entirely in questions, a grammatic constraint. Is it possible to push the definition of constraint one more level into abstraction, and manipulate the building blocks of narrative?

In a story entitled Variations on a Plot Curve, I took Freitag's Plot Triangle,

and attempted to retell the same story with different dramatic arcs:

and even:

taking as a cue the music composition princliples of retrograde, inversion, and retrograde inversion. The results of this experiment are not empirically verifiable because the vertical axis on Freitag's chart represents dramtic tension, which exists in the reader, not on the page the way the letter e might be said to exist on the page.

In The Story That Teaches You How To Write It, I used the theory of metaphor as articulated by George Lakoff, Mark Johnson, and Mark Turner, to create characters by designing each character's private metaphor system.


In one of the chapters, I modified a diagram by Gerard Genette representing a passage by Proust to create a sequence of scenes in which the ratio of historical time (time as the characters experience it) to story time (number of pages) steadily decreases.

All fiction writers use structure and constraint in some ways, even if generally not in the spirit of the Oulipo. The purpose of form in realist fiction is to be invisible. Fiction writers also tend to speak of their writing process in intuitive, passive terms: they put the characters in the room together and see what they do. To speak this way about your writing is to surrender your authority or at least to deny it: characters don't do anything, the reader and writer do the doing. Is there a way to describe the interaction of characters that is mechnical and not psychological? When creating a character, is the psychology you construct that of the character or the reader? Where is the line between structuring the language and structuring the reader's interpretation of it?

Unknown in Chicago, Keeler made plot his topic of investigation. His idea of a beautiful novel resided in the complexity of its plot, and he followed highly specific rules to make his plots complex and yet, to his mind, structurally sound. Keeler believed in an infinity of possible stories and scoffs at those who say there are only a few basic plots. Keeler finds coincidence an acceptable, even indispensable, element of the novel. The plots are complex machines that render their characters cogs, turning deterministically to conserve and transmit energy. Keeler’s plots are not flat maps. They are dynamic, in motion, a diagram of a row of dominos falling, playing pool and knocking in all the balls, eight ball last, on the break, skull in the corner pocket. He has made the puzzle plot unsolvable.

The Mechanics and Kinematics of Web-Work Plot Construction is as lucid a model of narrative as any I learned in the course of getting three degrees in writing fiction. Keeler's stylistic indulgences tend to make his treatise unnnecessarily impenetrable and strangely organized. The passage that might best be considered the introduction occurs in section XXIII, when Keeler offers what appears to be the philosophy behind web-work:

Aside from normal interest in dramatic happenings, is it not true that in every human being is a longing--an instinctive hunger--to believe that life, in its great complexity and utter meaningless involvements, does move in a regulated manner; that it is not all incoherent, all mixed up and utterly without pattern, but that the whole thing is mathematically accurate in its causes and effects?


In How To Write A WebWork, I attempt to distill Keeler's principles to the basic methods.

What the rules and diagrams don't express is that these diagrams represent structured objects, and these objects are then enacted with characters and objects. Sometimes he describes these schematics as wiring diagrams, weavings, suspension bridges, eye muscles, or geodesic domes. Although Keeler changes metaphors, I am convinced there is something solid and tangible, something visceral he can feel about a well-constructed plot.

The web-work at times literally seems to quiver like a sea of jelly. Every tap of the hammer makes a dozen changes necessary. But gradually as both the story and plot develop together, it becomes more and more stable, till finally it is as near perfect as it can be made.

In other words, that the vertical axis of Keeler's schematics corresponds to nothing is not an oversight. He requires that space to draw curves to show the tension in the system. So one of the fundamental features of the diagram is that which appears incidental: the curves. This is why he does not simply connect the incidents with straight lines. The curves show how the trajectories of the characters change as a result of plot incidents.

Therefore, since our first space-and-time diagram has two disabilities, first, that objects occupying the same space—or practically the same space—at the same time will show as the same thread; and second, that there exist changes in people's affairs which are not merely the occupancy of a succession of new areas, we will adopt a new expedient of diagraming. We will retain plot-threads, and still continue to allow their left-right extension to represent passage of time; but we will utilize the up-and-down scope only for the purpose of showing by means of knots or intersections where these plot threads come into vital deviational relationships with each other.

And do not disregard that word "deviational."

For it is in that essential that a plot incident is a plot incident, and deviation is the keynote of plot itself.

Web-Work Diagrams

The Voice of the Seven Sparrows, by Harry Stephen Keeler
Included as an example, with a detailed key, in the Mechanics and Kinematics of Web-Work Plot Construction.
Nevermore, by Marie Redonnet
Drawn in response to reading the novel by William Gillespie, 2003
School for Performing and Creative Arts, Cincinnati
On April 2nd, 2004, I taught two creative writing classes the basics of web-work plot construction and enjoined them in the writing of two collaborative web-work novellas.
Hobo Story, by William Gillespie
This story was designed purely to serve as a test of Keeler's methods
Keyhole Factory, by William Gillespie
A more ambitious attempt to write web-work, in progress

Keeler & Narratology

I don’t like reading Keeler’s fiction. Yet. It’s got problems. To me. The arrhythmia. The relentlessly questionable style choices. The race stuff, the race stuff, the race stuff.

The Mechanics and Kinematics of Web-Work Plot Construction then, do not interest me as a method of writing like Keeler: I think they stand as a structural analysis of plot that can be applied to all fiction. I think any story or novel can be represented in a webwork schematic, even if it just looks like a line. In many cases this will involve guesswork. In the case of a complex time travel story or metafiction, you may have to do some serious head-scratching and invent new notation. And in the case of a large and complicated novel, a webwork diagram might reveal the author’s mistakes as well as their intentions. Either way, I believe such an exercise would be instructive and would reveal aspects of the novel’s structure otherwise difficult to pick up on. Especially if one were to diagram a number of different well-read books such that their structures could be compared at a glance.

A conventional detective novel, indeed all narrative by many definitions, is bound by sequence. The story is a sequence of events in story time, and the story telling is a sequence in which these events are revealed to the reader. These two sequences do not have to correspond, as the author has recourse to flashforward or foreshadowing, flashbacks, and any number of tricks. Narrative is a double helix of two interlocking sequences.

Keeler’s diagrams--encompassing chronology, causality, character, and motivation—provide a model to analyze one half of the double helix, that of the story independent of its telling. When you look down on one of those schematics, though the events it represents are bound by a sequence, you can study them as an object.

For example, in the The Voice of the Seven Sparrows, thanks to the surviving webwork diagram, we can see the story (the diagram) independently of its telling (the novel), and speculate how the storytelling could have been accomplished in any number of ways. Keeler chose to remain fixed on a particular character as he moved through a portion of the story (going forward in time), and the story was revealed through his encounters with other characters who had access to parts of the diagram neither the main character nor the reader otherwise had access to. But what if the story were told in a different way? From multiple points of view? Or in reverse chronology? Entirely in unattributed dialog? Without the letter E or entirely in interrogative sentences? Through visual descriptions with no narration? Or from the perspective of one of the other characters? Well, since this is Keeler, we can guess that the story would be written in excruciating racial gibberish if it were written from the perspective of one of its characters, or in unattributed dialog. Let me put it this way: what if the story were written by a different author?

To some extent, in the Voice, as in conventional detective fiction, the reader is drawn into enigmas—questions the story poses that the reader desires answered (closure),such that they feel compelled to continue reading (in case they fail to surrender to the pure beauty of the prose). But my sense is that Keeler’s enigmas, through their sheer density, are hard to keep track of. Many surprising details are revealed, some of the weirdest at the end--for example that four of the characters are two of the characters each in disguise--but these peculiar gestures toward clearing up confusion by resolving enigmas lack the satisfying clarity of “the butler did it." In a conventional mystery story (which I assume was the starting point for Keeler’s web-work genre), the reader is engaged because the enigma is simple enough to second-guess, and the author’s art sustains this guesswork while keeping the enigma suspended. But the satisfaction of guessing what happens is utterly denied to Keeler's readers. Instead of an “ah!” you get a “huh?” or a “ha” or a groan or a “&^%$()*^#%!”

One aspect of the “ah!” I call retroactive correction. When you discover the butler did it, you immediately want to reread the book because certain details that did not make sense or were forgettable are now significant and you want to reexamine them in the light of the new interpretation your knowledge of the ending can bring to them.

A Keeler schematic makes it possible to construct a story independent of its telling, in the manner a composer might compose a musical score that is fixed independently of its performances. The Voice is Keeler’s performance of his webwork diagram score. We could give that same score to Paul Auster and ask him to perform the novel. We might even allow the reader to be a part of the performance. In multisequential fiction (such as hypertext fiction), the reader has a hand in choosing the storytelling sequence. What happens to the “ah!” of the surprise ending if the reader reads it first? To my knowledge, nobody has properly interrogated this question from a literary (rather than a Computer Science) perspective.

Because Keeler gives us the criteria for composing a novel in the form of a score, in which every scene is essential to the structure of the whole, his “web-work” schematics (which resemble hypertext node-link flowcharts) provide insight into the possibility of writing a multisequential fiction in which every scene, no matter which order it is read in, provides unexpected information that might trigger a retroactive correction of scenes read previously, regardless of what order they happen in the original story.

Web-Work and Hypertext

Keeler's diagrams remind me of hypertext maps, for example those used by Storyspace. Keeler's novels, usually told through a single perspective in chronological order (although usually dense with, if not composed entirely of, digressions and reminiscences) are not, in my opinion, hypertextual. They are not only sequential but bound by sequence in order that they may utterly confuse the reader, such that in a novel that appears to be a whodunnit, the character who is the murderer might be introduced on the last page, or such that, in a complicated mystery, it might be revealed that four characters are actually two characters wearing clever disguises, in the closing passages. Keeler's novels seem to involve suprise endings, whether or not those endings are actually satisfying in a normal narrative sense.

However,in approaching the problem of writing multisequential fiction, a problem that is will continue to be of interest for a long time, the web-work diagrams suggest a solution.

Imagine that the threads are links and the intersections are texts, scenes, nodes,pages, chapters, or chunks: bits of story. Because the sturcture is so woven as to make every intersection integral to the composition of the whole, it might be possible to write a multisequential web-work that could sustain dramatic tension. The trick would be to write a hypertext novel with a surprise ending no matter which chapter you read last.

The difference between a web-work plot diagram as applied to a print novel versus one applied to multisequential fiction is that in a mutlisequential web-work diagram, the reader does not follow a single character from left to right, but can traverse the web in any direction following any thread.

By way of demonstration, I have attempted to take some of the simple forms Keeler calls Elemental Plot Constructions and found examples of them in the hypertext novel The Unknown.

The Unknown


An intersection between two plot threads causes a second intersection. In 'detailsofwhich.htm,' William meets Paul Auster at a party. They play shuffleboard, Auster cheats, William accuses him of cheating, Auster becomes enraged, forfeits the game, and William wins. Later, in 'denouement.htm,' the two meet again to play Trivial Pursuit on supposedly friendly terms. Paul Auster has rigged the deck of Trivial Pursuit cards. The first incident has motivated the second. Auster, who hates losing, has arranged a rematch for revenge. William, who respects Auster's work, wishes to reconcile their differences and agrees to the Trivial Pursuit.

Dirk, William, & Scott have a breakdown in 'boston.htm.' In rehab, Dirk is abducted by aliens, and taken into orbit ('inorbit.htm'). The aliens extract a writing sample from Dirk and take it to Thomas Pynchon on the mothership. Pynchon, threatened by Dirk's powers, arranges Dirk's assassination at the Y2K Party. Meanwhile, Scott realizes William is having an emotional collapse, and witnesses his nearly-fatal bungie-jumping accident in 'bungie.htm.'. This causes Scott to organize the Y2K Party, to try to knit the Unknown back together. Dirk's assassination at the stroke of the new millennium ('laparty.htm') thus results from both his abduction and William's bungie-jumping accident.


The Mechanics and Kinematics of Web-Work Plot Construction
     at Spineless Books