Lecture delivered at Oxford, March 5th, 1935[1]
Some twenty-five years ago, it was rather the fashion among commentators to deplore that Aristotle should have so much inclined to admire a kind of tragedy that was not, in their opinion, “the best.” All this stress laid upon the plot, all this hankering after melodrama and surprise—was it not rather unbecoming—rather inartistic? Psychology for its own sake was just then coming to the fore, and it seemed almost blasphemous to assert that “they do not act in order to portray the characters; they include the characters for the sake of the action.” Indeed, we are not yet free from the influence of that school of thought for which the best kind of play or story is that in which nothing particular happens from beginning to end.
Now, to anyone who reads the Poetics with an unbiased mind, it is evident that Aristotle was not so much a student of his own literature as a prophet of the future. He criticised the contemporary Greek theatre because it was, at that time, the most readily available, widespread and democratic form of popular entertainment presented for his attention. But what, in his heart of hearts, he desired was a good detective story; and it was not his fault, poor man, that he lived some twenty centuries too early to revel in the Peripeties of Trent’s Last Case or the Discoveries of The Hound of the Baskervilles . He had a stout appetite for the gruesome. “Though the objects themselves may be painful,” says he, “we delight to view the most realistic representations of them in art, the forms, for example, of the lowest animals and of dead bodies.” The crawling horror of The Speckled Band would, we infer, have pleased him no less than The Corpse in the Car , The Corpse in Cold Storage or The Body in the Silo . Yet he was no thriller fan. “Of simple plots and actions,” he rightly observes, “the episodic are the worst. I call a plot episodic when there is neither probability nor necessity in the sequence of the episodes.” He would not have approved of a certain recent book which includes among its incidents a machine-gun attack in Park Lane, an aeroplane dropping bombs on Barnes Common,[2] a gas attack by the C.I.D. on a West-End flat and a pitched battle with assorted artillery on a yacht in the Solent. He maintained that dreadful and alarming events produced their best effect when they occurred, “unexpectedly,” indeed, but also “in consequence of one another.” In one phrase he sums up the whole essence of the detective story proper. Speaking of the dénouement of the work, he says: “It is also possible to discover whether some one has done or not done something.” Yes, indeed.
Now, it is well known that a man of transcendent genius, though working under difficulties and with inadequate tools, will do more useful and inspiring work than a man of mediocre intellect with all the resources of the laboratory at his disposal. Thus Aristotle, with no better mysteries for his study than the sordid complications of the Agamemnon family, no more scientific murder-methods than the poisoned arrow of Philoctetes or the somewhat improbable medical properties of Medea’s cauldron; above all, with detective heroes so painfully stereotyped and unsympathetic as the inhuman array of gods from the machine, yet contrived to hammer out from these unpromising elements a theory of detective fiction so shrewd, all-embracing and practical that the Poetics remains the finest guide to the writing of such fiction that could be put, at this day, into the hands of an aspiring author.
In what, then, does this guidance consist? From the start Aristotle accepts the Detective Story as a worthy subject for serious treatment. “Tragedy,” he observes (tragedy being the literary form which the detective story took in his day), “also acquired magnitude”—that is, it became important both in form and substance. “Discarding short stories and a ludicrous diction, it assumed, though only at a late point in its progress, a tone of dignity.” I am afraid that “short stories and a ludicrous diction” have characterised some varieties of the genre up to a very late point indeed; it is true, however, that there have recently been great efforts at reform. Aristotle then goes on to define tragedy in terms excellently applicable to our subject; “The imitation” (or presentment, or representation—we will not quarrel over the word) “of an action that is serious”—it will be admitted that murder is an action of a tolerably serious nature—“and also complete in itself”—that is highly important, since a detective story that leaves any loose ends is no proper detective story at all—“with incidents arousing pity and fear, wherewith to accomplish its catharsis of such emotions.”
Too much has already been said and written on the vexed subject of the catharsis. Is it true, as magistrates sometimes assert, that little boys go to the bad through reading detective stories? Or is it, as detective writers prefer to think with Aristotle, that in a nerve-ridden age the study of crime stories provides a safety valve for the bloodthirsty passions that might otherwise lead us to murder our spouses? Of all forms of modern fiction, the detective story alone makes virtue ex hypothesi more interesting than vice, the detective more beloved than the criminal. But there is a dangerous error going about—namely that “if . . . detective fiction leads to an increase in crime, then the greater the literary merit, the greater will be the corresponding increase in crime.”[3] Now, this is simply not true: few people can have been inspired to murder their uncles by the literary merits of Hamlet . On the contrary, where there is no beauty there can be no catharsis; an ill-written book, like an ill-compounded drug, only irritates the system without purging. Let us then see to it that, if we excite evil passions, it is so done as to sublimate them at the same time by the contemplation of emotional or intellectual beauty. Thus far, then, concerning the catharsis.
Aristotle next discusses Plot and Character. “A detective story,” we gather, “is impossible without action, but there may be one without character.” A few years ago, the tendency was for all detective stories to be of the characterless or “draught-board” variety; to-day, we get many examples exhibiting a rather slender plot and a good deal of morbid psychology. Aristotle’s warning, however, still holds good:
“One may string together a series of characteristic speeches of the utmost finish as regards diction and thought, and yet fail to produce the true dramatic effect; but one will have much better success with a story which, however inferior in these respects, has a plot.”
And again:
“The first essential, the life and soul, so to speak, of the detective story, is the plot, and the characters come second.”
As regards the make-up of the plot, Aristotle is again very helpful. He says firmly that it should have a beginning, a middle and an end. Herein the detective story is sharply distinguished from the kind of modern novel which, beginning at the end, rambles backwards and forwards without particular direction and ends on an indeterminate note, and for no ascertainable reason except the publisher’s refusal to provide more printing and paper for seven-and-sixpence. The detective story commonly begins with the murder; the middle is occupied with the detection of the crime and the various peripeties or reversals of fortune arising out of this; the end is the discovery and execution of the murderer—than which nothing can very well be more final. Our critic adds that the work should be of a convenient length. If it is too short, he says, our perception of it becomes indistinct. (This is meiosis; he might have said that it will not be perceived at all, since the library subscriber will flatly refuse to take it out, on the ground that “there isn’t enough reading in it.”) He objects, still more strongly, to the work that is of vast size or “one thousand miles long.” “A story or plot,” he reminds us, “must be of some length, but of a length to be taken in by the memory.” A man might write a detective story of the length of Ulysses ,[4] but, if he did, the reader would not be able to bear all the scattered clues in mind from the first chapter to the last, and the effect of the final discovery would be lost. In practice, a length of from 80,000 to 120,000 words is desirable, if the book is to sell; and this is enough to allow, in Aristotle’s general formula, of “the hero’s passing by a series of probable or necessary stages from misfortune to happiness or from happiness to misfortune.” Later, however, he conveys a very necessary warning: “A writer often stretches out a plot beyond its capabilities, and is thus obliged to twist the sequence of incident.” It is unwise to “write-up” a short-story type of plot to novel length, even to fulfil a publisher’s contract.
The next section of the Poetics gives advice about the unity of the plot. It is not necessary to tell us everything that ever befel the hero. For example, says Aristotle, “in writing about Sherlock Holmes” (I have slightly adapted the instance he gives)—
“the author does not trouble to say where the hero was born, or whether he was educated at Oxford or Cambridge, nor does he enter into details about incidents which—though we know they occurred—are not relevant to the matter in hand, such as the cases of Vamberry the Wine Merchant, the Aluminium Crutch, Wilson the Notorious Canary-Trainer or Isadora Persano and the Remarkable Worm.”
The story, he says—
“must represent one action, a complete whole, with its several incidents so closely connected that the transposal or withdrawal of any one of them will disjoin and dislocate the whole.”
In other words, “murder your darlings”—or, if you must write a purple passage, take care to include in it some vital clue to the solution, which cannot be omitted or transposed to any other part of the story. Thus, in Trent’s Last Case , the description of Marlowe’s room conveys the necessary clue that he has been a member of the O.U.D.S. and is therefore to be presumed capable of acting a part; the poker-game in The Canary Murder Case throws needful light on the murderer’s character; the picture of the Shivering Sands in The Moonstone prepares us for the discovery of the paint-stained nightgown in that spot; and so forth.
But now comes the important question: What kind of plot are we to choose? And this raises the great central opposition of the Probable and the Possible. It is possible that two Negroes should co-exist, so much alike as not only to deceive the eye, but to possess the same Bertillon measurements; that they should both bear the same Christian and surnames, and that they should both be confined in the same prison at the same time: it is possible, since it actually occurred.[5] But if we are to found a plot upon such a series of coincidences it will have an improbable appearance.
It is open to us to contrive stories based upon such incidents in real life, either giving the characters their real names or otherwise calling upon the witness of history. Thus there have been books founded on the Bravo case, the Crippen murder, the Penge tragedy, the case of W. H. Wallace, and so on. When the facts are well known, the reader will accept the events as narrated. But it often turns out that the stories so written appear less convincing than those that are wholly invented; and it is frequently necessary to add inventions to the known facts, in order to make these true events appear probable. “So that,” says Aristotle, “one must not aim at a rigid adherence to the traditional stories,” particularly as “even the known stories are known only to a few.” Thus, even where the possibility cannot be challenged, probability should be studied.
But where both names and incidents are invented, then, “a likely impossibility is always preferable to an unconvincing possibility.” It may be impossible that the leaden bullet buried in a man’s body should be chemically recovered from his ashes after cremation; but, by skilful use of scientific language, Dr. Austin Freeman persuades us that it is probable, and indeed inevitable. Whereas, when an author seeks to persuade us that a pleasant young Cambridge man of gentle birth is affronted by being asked to take his place in a queue behind a taxi-driver or some such person, the incident, though physically possible, offends by its improbability, being contrary to the English character, whose eternal patience in arranging itself in orderly queues is well known to amount to genius. “The story,” says Aristotle, “should never be made up of improbable incidents; there should be nothing of the sort in it.” Lest this seem too severe, he suggests as a practical compromise that “if such incidents are unavoidable, they should be kept outside the action.” Thus, in the story of The Gloria Scott , while the previous history of Old Trevor is not merely improbable, but, according to the dates given, impossible, we do not notice this in reading, because the episode stands outside the action of the plot. Similarly, as regards the characters, the impossible-probable is better than the improbable-possible; for (says Aristotle again) “if a detective such as Conan Doyle described be impossible, the answer is that it is better he should be like that, since the artist ought to improve on his model.”
In the matter of scientific detail, Aristotle is all for accuracy. If, he says in effect, you cannot attain your artistic end without some impossible device (such as the instantaneously fatal and undiscoverable poison), then, at a pinch, you may be justified in using it:
“If, however, the poetic end might have been as well or better attained without sacrifice of technical correctness in such matters, the impossibility is not to be justified, since the description should be, if it can, entirely free from error.”
Thus, in Mr. John Rhode’s The Corpse in the Car , the emission of an undetectable gas from the wireless set is more justifiable, because scientifically feasible, than the same author’s release of hydrocyanic acid gas from a rubber hot-water bottle in Poison for One , a method which (I am told) would not be effective in practice.
Concerning the three necessary parts of a detective plot—peripety, or reversal of fortune, discovery, and suffering—Aristotle has many very just observations. On suffering, we need not dwell long. Aristotle defines it as “action of a destructive or painful nature, such as murders, tortures, woundings and the like.” These are common enough in the detective story, and the only remark to be made is that they ought always to help on the action in some way, and not be put in merely to harrow the feelings, still less to distract attention from a weakness in the plot.
A reversal of fortune may happen to all or any of the characters: the victim—who is frequently a man of vast wealth—may be reduced to the status of a mere dead body; or may, again, turn out not to be dead after all, as we had supposed. The wrongly suspected person, after undergoing great misfortunes, may be saved from the condemned cell and restored to the arms of his betrothed. The detective, after several errors of reasoning, may hit upon the right solution. Such peripeties keep the story moving and arouse alternating emotions of terror, compassion and so forth in the reader. These events are best brought about, not fortuitously, but by some hamartia or defect in the sufferer. The defect may be of various kinds. The victim may suffer on account of his unamiable character, or through the error of marrying a wicked person, or through foolishly engaging in dubious finance, or through the mistake of possessing too much money. The innocent suspect may have been fool enough to quarrel with the victim, or to bring suspicion on himself by suppressing evidence with intent to shield somebody. The detective suffers his worries and difficulties through some failure of observation or logic. All these kinds of defect are fruitful in the production of peripety.
Aristotle mentions many varieties of the discovery which forms the dénouement. This is usually the discovery, either of the identity of the murderer, or of the means by which the crime was committed.
(1) The worst kind are discoveries made by the author himself . These are, indeed, so inartistic as to be scarcely permissible in the true detective story: they belong to the thriller. It is, however, possible, where the villain’s identity is known, to make an agreeable story by showing the moves and counter-moves made successively by villain and detective (Wilkie Collins in No Name , Austin Freeman in The Singing Bone ).
(2) The discovery by material signs and tokens is very common: in The Trial of Mary Dugan the discovery that a person is left-handed leads to his conviction; in The Eye of Osiris the identity of the (supposed) Egyptian mummy with the missing corpse is proved by the discovery of identical tooth-stoppings and a Potts fracture in both.
(3) Discovery through memory is also used: thus, in Unnatural Death , the murder-method—the production of an air-lock in a main artery—is discovered to the detective by his memory of a similar air-lock in the petrol-feed of a motor-cycle.
(4) Discovery through reasoning is perhaps most common of all: the murderer was in the house at such a time, he is an electrician, he is tall and smokes Sobranie cigarettes; only X corresponds to all these indications, therefore X is the murderer.
(5) Aristotle’s fifth type of discovery is particularly interesting. He calls it discovery through bad reasoning by the other party . The instance he adduces is obscure, the text being apparently mutilated and referring to a play unknown. But I think he really means to describe the discovery by bluff . Thus, the detective shows the suspect a weapon saying, “If you are not the murderer, how do you come to be in possession of this weapon?” The suspect replies: “But that is not the weapon with which the crime was committed.” “Indeed?” says the detective, “ and how do you know? ”
This brings us to the very remarkable passage in which Aristotle, by one of those blinding flashes of insight which display to the critic of genius the very core and centre of the writer’s problem, puts the whole craft of the detective writer into one master-word: Paralogismos . That word should be written up in letters of gold on the walls of every mystery-monger’s study—at once the guiding star by which he sets his compass and the jack-o’-lantern by which he leads his readers into the bog; paralogism—the art of the false syllogism—for which Aristotle himself has a blunter and more candid phrase. Let us examine the whole paragraph, for it is of the utmost importance.
“Homer,” says he—if he had lived in our own day he might have chosen some more apposite example, such as Father Knox or Mrs. Agatha Christie, but, thinking no doubt of Odysseus , he says Homer—“Homer more than any other has taught the rest of us the art of framing lies in the right way .[6] I mean the use of paralogism. Whenever, if A is or happens, a consequent, B is or happens, men’s notion is that, if the B is, the A also is—but that is a false conclusion. Accordingly, if A is untrue, but there is something else, B, that on the assumption of its truth follows as its consequent, then the right thing is to present us[7] with the B. Just because we know the truth of the consequent, we are in our own minds led on to the erroneous inference of the truth of the antecedent.”
There you are, then; there is your recipe for detective fiction: the art of framing lies. From beginning to end of your book, it is your whole aim and object to lead the reader up the garden; to induce him to believe a lie. To believe the real murderer to be innocent, to believe some harmless person to be guilty; to believe the detective to be right where he is wrong and mistaken where he is right; to believe the false alibi to be sound, the present absent, the dead alive and the living dead; to believe, in short, anything and everything but the truth.
The art of framing lies—but mark! of framing lies in the right way (ὡς δεῑ). There is the crux. Any fool can tell a lie, and any fool can believe it; but the right method is to tell the truth in such a way that the intelligent reader is seduced into telling the lie for himself. That the writer himself should tell a flat lie is contrary to all the canons of detective art. Is it not amazing that Aristotle, twenty centuries ahead of his time, should thus have struck out at a blow the great modern theory of fair-play to the reader? A is falsehood; B is truth. The writer must not give us A upon his own authority, for what he says upon his own authority we must be able to believe. But he may tell us B—which is true—and leave us to draw the false conclusion that A is true also.
Thus, at the opening of a story, the servant Jones is heard to say to his master, Lord Smith, “Very good, my lord. I will attend to the matter at once.” The inference is that, if Jones was speaking to Smith, Smith was also speaking to Jones; and that, therefore, Smith was alive and present at the time. But that is a false conclusion; the author has made no such assertion. Lord Smith may be absent; he may be already dead; Jones may have been addressing the empty air, or some other person. Nor can we draw any safe conclusion about the attitude of Jones. If Jones is indeed present in the flesh, and not represented merely by his voice in the form of a gramophone record or similar device (as may well be the case), then he may be addressing some other party in the belief that he is addressing Smith; he may have murdered Smith and be establishing his own alibi; or Smith may be the murderer and Jones his accomplice engaged in establishing an alibi for Smith. Nor, on the other hand, is it safe to conclude (as some experienced readers will) that because Smith is not heard to reply he is not therefore present. For this may very well be the Double Bluff, in which the reader’s own cunning is exploited to his downfall. The reader may argue thus:
Jones spoke to Smith, but Smith did not speak to Jones. Many authors employ this device so as to establish the false inference that Smith was alive and present. I therefore conclude that Smith is absent or dead.
But this syllogism is as false as the other. “Many authors” is not the same thing as “all authors at all times.” It does not exclude the possibility that an author may at some time imply the truth in such a manner that it looks like a lie.
A fine example of this double bluff is found in Father Knox’s The Viaduct Murder . A man is found dead, with his face beaten into unrecognisable pulp. Circumstantial evidence suggests that the dead man was X. The detectives and the reader are invited to reason after the following manner:
The dead man is thought to be X. But he is unrecognisable. Therefore he is not X. Therefore he is someone else, namely Y. And, since X is undoubtedly missing, X is probably the murderer.
But the disfigured corpse turns out to be X after all; so that all the ingenious conclusions founded upon the false premise are false also.
Another variety of the paralogism is found in a syllogism built upon the following lines:
A is the obvious suspect. But in a detective story, the obvious suspect is always innocent. Therefore A is innocent.
But for the middle term of this proposition there is no warrant whatever. The statement is neither universally true nor logically necessary. The obvious suspect is innocent more frequently than not, but nothing compels the author to make him so.
Nothing in a detective story need be held to be true unless the author has vouched for it in his own person . Thus, if the author says—
Jones came home at 10 o’clock
then we are entitled to assume that Jones did indeed come home at that time and no other. But if the author says—
The grandfather clock was striking ten when Jones reached home
then we can feel no certainty as to the time of Jones’s arrival, for nothing compels us to accept the testimony of the clock. Nor need we believe the testimony of any character in the story, unless the author himself vouches for that character’s integrity.
Thus, let us suppose that the butler gives evidence that Jones returned at ten. The butler’s employer asserts that he has always found the butler scrupulously truthful. Are we therefore to believe the butler? By no means; for the employer may be deceived, or may have deceived the butler, or may be backing up the butler’s testimony for reasons of his own.
But if the author himself says: “No one could possibly doubt that the butler was speaking the truth”—then, I think, we must believe that the butler is a truthful witness, for the author himself has stated, on his own authority, that doubt was impossible.
Remember, however, that the person telling the story is not necessarily the author. Thus, in The Murder of Roger Ackroyd , the story is told by the detective’s fidus Achates or (to use the modern term) his Watson. Arguing from the particular to the general, we may be seduced into concluding that, because the original Dr. Watson was a good man, all Watsons are good in virtue of their Watsonity. But this is false reasoning, for moral worth and Watsonity are by no means inseparable. Thus, the first man sinned and laid the blame upon his wife; but it would be an error to conclude that all men, when they sin, blame their wives—though in fact they frequently do. There may be found rare men who, having wives, yet refrain from blaming them and are none the less men on that account. So, despite the existence of a first innocent Watson, we may yet admit the possibility of a guilty one; nor, when the Watson in Roger Ackroyd turns out to be the murderer, has the reader any right to feel aggrieved against the author—for she has vouched only for the man’s Watsonity and not for his moral worth.
This brings us, however, to the consideration of the characters, concerning whom Aristotle takes a very twentieth-century point of view. He says that they must be good . This, I suppose, must be taken relatively, to mean that they should, even the meanest and wickedest of them, be not merely monsters and caricatures, like the personages in a low farce, but endued with some sort of human dignity, so that we are enabled to take them seriously. They must also be appropriate : a female, he says, must not be represented as clever. This is a delicate point—would he, or would he not, have approved of Miss Gladys Mitchell’s diabolically clever Mrs. Bradley? We may take it, however, that the cleverness should only be such as is appropriate to the sex and circumstances of the character—it would be inappropriate that the elderly maiden sister of a country parson should carry out or detect a murder by means of an intricate and clever method knowable only to advanced chemical experts; and so with the other characters. Thirdly, the characters must be like the reality (τό ὅμοιον). Scholars differ about what Aristotle means by this word. Some think it means, “conformable to tradition”: that the villain should be easily recognisable as villainous by his green eyes, his moustache and his manner of ejaculating “Ha!” and the detective by his eccentricities, his pipe and his dressing-gown, after the more ancient models. But I do not agree with them, and believe that the word means, as we say to-day, “realistic,” i.e. with some moderate approximation in speech and behaviour to such men and women as we see about us. For elsewhere, Aristotle takes the modern, realistic view, as when he says, for instance, that the plot ought not to turn on the detection and punishment of a hopelessly bad man who is villainous in all directions at once—forger, murderer, adulterer, thief—like the bad baron in an Adelphi melodrama; but rather on that of an intermediate kind of person—a decent man with a bad kink in him—which is the kind of villain most approved by the best modern writers in this kind. For the more the villain resembles an ordinary man, the more shall we feel pity and horror at his crime and the greater will be our surprise at his detection. So, too, as regards the innocent suspects and the police; in treating all such characters, a certain resemblance to real life is on the whole to be desired. Lastly, and most important and difficult of all, the characters must be consistent from first to last. Even though at the end we are to feel surprise on discovering the identity of the criminal, we ought not to feel incredulity; we should rather be able to say to ourselves: “Yes, I can see now that from the beginning this man had it in him to commit murder, had I only had the wits to interpret the indications furnished by the author.” Thus, the villainy of the apparently amiable father in The Copper Beeches is betrayed by his participation in his offspring’s cruel enjoyment in the slaughter of flies, and the character is seen to be consistent. Inconsistency in the characters destroys the probability of the action, and, indeed, amounts to a breach of the rule of fair play, since we are entitled to believe that a character remains the same person from beginning to end of the story and nemo repente fuit turpissimus .
This discourse is already too long. Let me remind myself of Aristotle’s own warning: “There are many writers who, after a good complication, fail to bring off the dénouement.” This is painfully true of detective stories; it has also some application to lectures and speeches upon whatever occasion. But indeed, everything that Aristotle says about writing and composition is pregnant with a fundamental truth, an inner rightness, that makes it applicable to all forms of literary art, from the most trivial to the most exalted. He had, as we say, the root of the matter in him; and any writer who tries to make a detective story a work of art at all will do well if he writes it in such a way that Aristotle could have enjoyed and approved it.
[1] The translation of The Poetics used throughout this lecture is that of Professor Ingram Bywater, published by the Clarendon Press.
[2] It is perhaps necessary to remind readers that this kind of incident, though it has since become quite commonplace, was unusual at the date (1935) when this paper was first written.
[3] Editorial in The Author , spring, 1935.
[4] I refer, of course, to Mr. James Joyce’s novel; not to Homer’s poetical treatment of the subject.
[5] The case of the two Will Wests, U.S. Penitentiary, Leavenworth, Kansas, 1903.
[6] ψευδῆ λέγειν ῶϛ δεῑ.
[7] προσθεīναι [δεῑ] Bywater: “to add on the B.” Wharton: “it is natural to pre-suppose the first” (i.e. the A). Whichever translation is preferred, the general sense is clear: if the author provides the consequent, the reader may be trusted to infer (falsely) the antecedent for himself.