Here’s my attempt at more skeptical assessment. I’ll rely mostly on some practical considerations.
What experts say
Experts don’t know exactly how it happened. Professor B.J. Rahn in The Oxford companion to crime and mystery writing mentions a couple of novels where the butler is a suspect and tells us
Thus, whether the butler did it or not, it is clear that it has become
customary to accuse him. Because he usually is not guilty, the phrase
has passed into lighthearted general parlance.
This seems reasonable, but this account lacks details (only two novels are mentioned), and the author uses the word “speculating” in the next paragraph. With more evidence, this could be more convincing. Some support for this I found in the book The Art Of The Mystery Story (cited in the Original Answer); Stepehen Leacock’s story:
Of course he didn’t really kill Sir Charles, but the local police
always arrest the butler.
Oxford dictionary of modern quotations says that the origin of the phrase is unclear:
Nigel Rees, in Sayings of the Century (1984), quotes a correspondent
who recalls hearing it at a cinema c.1916 but the origin of the phrase
has not been traced
Rees’s later book Mark my words (2002) repeats that the origins remain untraced.
The Guardian article blames Mary Rinehart’s The Door:
After the commercial success of The Door, the butler was now an easy
target for comedians and satirical writers alike who quickly pounced
on the butler-as-murderer archetype. Damon Runyon's 1933 short story
"What, No Butler?" is an obvious riff on the cliché and PG Wodehouse's
1957 comic novel "Something Fishy" was published in the US as "The
Butler Did It".
But was Runyon’s story a reaction to the book? The article doesn’t give citations, so it is hard to check. I searched for the word “butler” in two Rinehart’s biographies, but couldn’t find anything interesting. I skimmed through Rinehart’s fan mail related to The Door - to no avail. (It was published as a serial in The Saturday Evening Post, and readers sent letters even before they knew the murderer.)
There’s also a novel hypothesis in the Original Answer. Silent movies could influence the trope! BookRiot article explains:
the source is actually silent films, listing 16 that had butlers that
did it or were suspected of doing a crime between 1915 and 1922, which
could explain the public being familiar with, and subsequently sighing
over, the trope in later literature.
In books, the criminal’s identity is important because we invest our energy in solving the intriguing puzzle. Do movies work the same?
E.g. I enjoyed Paddington 2. I remember that Paddington was a really nice bear and was wrongly convicted. But the villain, the crime… I’m at loss. I certainly don’t remember if there was a butler in the movie!
I know that it’s hard to find a compelling whodunit movie mystery, and a classic example of one, The Last of Sheila, wasn’t a big hit.
I’m a little confused. I need to consult practitioners’ opinion.
Screenwriter Dennis Palumbo explains that viewers don’t remember the details of the plot, so writers don’t rely on intricate mysteries. Instead, they pay attention to the characters:
In the thriller Fractured, what was the mistake Anthony Hopkins made that proved he killed his wife?
You got me.
the best mysteries […] are also about the exploration and resolution of psychological tension. In other words, how do the characters interact? What do they want?
as much as I admire the plotting in the film The Last of Sheila, I don’t
love the movie because I don’t care about anyone in it.
Similar complaints in silent era (AFI page for The Voice of Destiny):
An anonymous reviewer in the 29 June 1918 Motion Picture News
complained, "Little Marie is not given much of a part in this picture.
Of the rest of the characters, none gets any sympathy because none
The mystery writer's handbook highlights the importance of sympathy for the character in movies (rather than action). The identity of the killer is not always that important:
In a prize-fight picture, two fighters merely battering each other
into senseless hulks to the screaming of the crowd will arouse
boredom. But if one of them is a Clifford Odets hero whose broken
hand will forever bar him from playing the violin, the suspense can
If a detective story depends merely on establishing the identity of
the murderer, or the means of murder, or the like, it does not have
true suspense. It has instead curiosity. This is why many simple
detective stories are destroyed when the criminal is revealed, but a
truly suspenseful story remains suspenseful each time.
The motion picture Dial M for Murder remains suspenseful even for
those who saw the stage production and who know precisely how the true
murderer will be foiled.
Vox.com recommends The Human Nature of Playwriting by American playwright and screenwriter Samson Raphaelson:
A creative piece of writing – play, story, poem rides on emotion.
Usually on the emotion of a central character. By emotion I mean a
hunger, a desire, something burning under that character, humming and
beating like a motor, sending him forward.
OK, so movies rely on characters driven by personal desire - but detectives like Hercule Poirot in whodunit books have mostly professional relation to the crime! How can movies make the audience interested? Experts say that writers connect the crime with the personal life of the protagonist.
It’s the most effective way to generate empathy with the audience
Here’s Girl in the Web (1920):
Esther's clever investigation seems suspicious to the detectives
assigned to the case, but in reality she is tracking down the culprit,
who turns out to be Ferguson's new man-servant[…] Esther's efforts
also bring the Prices back together, and she wins herself a mate in
The crime here is just theft, but it’s OK (nobody cares about murder either). What’s important is that the heroine fulfils her ambitions in proving her abilities, bringing order to the family, and finding a boyfriend. The crime and the criminal are less valuable.
So movies seem to have more diversity in the choice of the perpetrator. It shouldn’t surprise us that we find criminal butlers there! According to David Bordwell, the same is correct for the theatre:
The suspects are both high and low: businessmen, doctors, lawyers,
dowagers, flappers, playboys, ne’er-do-wells, gangsters, and servants
Let’s show the listed movies are not strictly about solving the puzzle.
The Silent Command:
Unable to pay for the operation that cured his daughter, a man
promises to surrender her to the doctor on her eighteenth birthday
[!!!]. Knowing that the physician wishes to make her the subject of
his experiments with hypnotism, the distressed father ultimately
reneges on his promise, whereupon the doctor hypnotizes the girl from
This is melodrama. The heroes have a problem with this doctor. To save the girl, her boyfriend must disclose the doctor’s evil deeds (a real villain of the story). The butler is not in the credits.
The Green Cloak is also about using hypnotism to acquit the girl (this time by her father). The butler seems to be the main villain (openly hinted before the murder: “When the butler appears, Duncan is visibly disturbed”).
The Mystic Hour. Guido is struggling with the irrational and unconscious. He suspects that he’s a murderer. Eventually, he outpours his frustrations into the painting of his dreams. Seeing the result, the horror-stricken butler confesses. So the murder is solved by solving a real problem (with the unconscious).
In Just for Tonight, Ted really wants to restore his finances and to win beautiful Betty Blake. Pursuing this, he accidentally solves the robbery. This is a romcom, so the crime doesn’t seem important.
How many silent mysteries are there? I got an e-copy of Silent Mystery and Detective Movies: a comprehensive filmography by Ken Wlaschin. Searching for “American” gave me 1091 matches. It’s a rude assessment and includes titles before 1915. I tried Advanced search on IMDB. It gave me 361 mystery titles (1915 to 1925) and 465 crime ones. “Crime + mystery” produced 39 matches. So, 361 + 465 – 39 = 787.
10+ or 20+ titles out of 800 – seems more like diversity than a cliché?
I conclude that movies have more diversity for practical reasons, the plot is not that important, and the audience for hypnotic doctors and 18-year old virgins doesn’t seem sophisticated enough to have much cultural influence or care about plot intricacy. Solutions in these 1-hour films are simple. They need not to be obvious (and for some films, like The Donovan affair, it’s more important), but they don’t inspire analysis. Some movies have the butler as the main villain, and sometimes he’s only an accessory. It’s not clear why viewers would pay attention.
What do experts think? David Bordwell only mentions, “At this point in history, having hired help commit the crime wasn’t yet out of bounds.” Wlaschin in an entry on The Other Half of the Note (1914) says:
of those rare mysteries where the butler actually did do it
Maybe we can find some evidence that the viewers were unhappy? I noticed that AFI pages have references for the press coverage.
Variety says “one of the best independent films ever” and “enough touches to keep it away from the sordid and conventional”.
Exhibitor's Trade Review: “usually pleases movie-goers, when they are kept guessing as to the real criminal's identity and this is accomplished neatly in "The Verdict."
The Film Daily gives more reviews: Evening World says “most everybody is kept guessing right up to the final fadeout”. Morning Telegraph mentions “flimsy dramatic finish”, but it probably relates to melodramatic issues of the finale, and the review is itself positive.
The Bromley Case.
The Billboard: “solving of the mystery comes at the very end, with a pleasant surprise”.
The Moving picture world: “Like the others of the series it is strong on suspense, and baffles the average person who would solve its secret before the last reel enters the game”. It also says the housekeeper shot the victim while he was struggling with the butler.
The Great Diamond Mystery
Exhibitor's Trade Review: “surprise twist neatly turned, for the identity of the real assassin is kept to the last.”
Even for 1929 The Donovan Affair we have East Liverpool Review with “the conclusion, which comes as a complete but logical surprise”.
What, no butler? We don’t see any complaints about the trope. The killer’s identity is only one element of a picture and works mostly because the audience cares about the characters.
Overall, it’s hard to tell things about the early audience. Was it interested in theories? Wasn’t it more naïve (the viewers were afraid of the train, or something?)?
I examined practical issues, experts’ opinion, and evidence. Right now, this seems like valuable research that helps to show that the trope wasn’t popular before 1930.
So, was the trope popular in 1920s? I’m not sure how to conduct this research, so I’ll reuse the evidence kindly provided by the OA.
… The Donovan Affair,” and if it did not have to stand comparison with
“The Ghost Train” it might seem more exciting. But we are getting to
the point now where, after fifteen or twenty guests have been grilled
and suspected of murder in turn, we not only don’t know who did it,
but don’t care. We have a system now whereby we automatically suspect
the butler right at the start and then pay no more attention. It may
turn out that the butler didn’t do it, but it’s a safe bet that none
of the suspected guests did, either.
Two problems with literal interpretation.
From the writer’s side: OK, everybody suspects the butler from the start. Only experienced playwright Owen Davis doesn’t know that. Somehow, his play is a hit despite everyone knowing how it ends. Then Columbia decides to adapt it into its first all-talking picture. Which is demonstrated with a special note asking not to reveal the ending. The movie is a success too.
From the critic’s side: If the play’s problem is its predictability, why it might look better if not the comparison with the other play? How will it save us from the spoiler? Why we don’t care after guests’ve been interrogated?
If there’s another explanation that doesn’t have this difficulties, it might be reasonable to accept it.
I think the critic means that we suspect someone seemingly innocent (for example, the butler) and stick with this intuition not to be overwhelmed with too many twists.
Susanna Calkins on CrimeReads site gives this example:
Burns Mantle, in his review of “The Bride” (1924), helps us see some
of these conventions: “You know what mystery comedies are. They’re
plays in which the butler is never what he pretends to be and the
smartest secret service agent on the force takes charge of the last
act and collects automatics from all the other members of the cast.”
(Chicago Daily Tribune, May 11, 1924).
It doesn’t mean that we figured it out: the butler trope comes from plays. Just this play had a criminal butler, and the convention is that the characters are not what they seem.
The play creates twists by moving suspicion from one guest to another. Benchley doesn’t want to invest his attention into this game: why should he think the current main suspect did it, if the next minute we have to suspect someone else? One might get irritated by constant twist; e.g., see this IMDB user review of a Netflix movie:
There are too many turns that cannot be tied together into a
consistent narrative. This may leave you almost betrayed by the story.
This is too common for mystery plays, The Ghost Train’s mechanism for creating suspense seems not so outdated to the critic. Too many twists – he doesn’t care – he says it might be OK play “provided you give much of a darn who killed Jack Donovan anyway”.
Robert Benchley was a famous humorist, and “automatically suspect the butler” is just a joke. He sticks to one suspect because there’s no consistency: anyone could’ve done it!
Academic support for this is in FORM AND FORMULA IN DETECTIVE DRAMA by Charles Bernard LaBorde (PDF):
The favorite choice for the identity of the murderer is the least
likely person. It was mystery writers' fondness for this approach
that prompted Robert Benchley's "The butler did it," since that often
unassuming, colorless type is as least likely as anyone.
Consistency with other reviews
Dana Skinner in The Commonweal (1926-09-15) guessed the culprit right from the start:
But perhaps I can say, with all propriety, that, as a somewhat
hardened viewer of mysteries, my first suspicion, which I discarded
for the simple reason that it seemed too obvious, turned out correct.
But what he means is just that he chose the least likely candidate:
The time has come when most people decide in the first act that the
seemingly most innocent of all parties must turn out to be the
culprit—not because of any semblance of logic, but because the
hackneyed method has been to direct suspicion everywhere except in the
Vogue (1926-11-01) is happy with the play:
In what it tries to do, “The Donovan Affair” succeeds admirably. It is
absorbing, thrilling, frightening.
Elizabeth Jordan in America; A Catholic Review of the Week (1926-09-25) doesn’t see a cliché problem:
certainly since “ The Bat” we have seen no better example of skilled
dramatic craftsmanship than it offers us
And, of course, according to the good old formula of mystery plays, in
the end the guilty wretch proves to be the person least likely to have
committed the crime.
There is great fascination in good mystery plays, but there is also
She jokes that in her play the murder will be committed by the baby. So, some critics like the play, some criticize the ending because the motive is weak, but the butler cliché is not there.
- A servant must not be chosen by the author as the culprit. This is begging a noble question. It is a too easy solution. The culprit must
be a decidedly worth-while person—one that wouldn’t ordinarily come
One can imagine Van Dine wanted to punish bad writers for using a bad solution too often. This doesn’t seem right. Van Dine writes this as a practitioner. Writers’ rules are popular on the Internet, and they are usually about what was learned through practice. Owen Davis in his autobiography says: “they are not words of wisdom at all, they are simply scars from the battlefield.” How to write is practical or tacit knowledge.
Oxford companion seems to agree:
their combined practice in forging new, longer narratives resulted in
a collection of distinct conventions
I think Van Dine generalized from his own practice (how to choose the culprit most effectively, what feels right) and the best works of colleagues. He says that there are no examples of a servant-perpetrator.
So the “bottom-up” explanation (via practice) seems like a default, and the burden of proof is on those who prefer more prescriptive take.
That said, there’s some evidence for this claim too.
Van Dine states explicitly:
Herewith, then, is a sort of Credo, based partly on the practice of
all the great writers of detective stories, and partly on the
promptings of the honest author's inner conscience.
Tropes used too often are placed separately under Rule 20:
I herewith list a few of the devices which no self-respecting
detective-story writer will now avail himself of. They have been
employed too often
This is consistent with the opinion of his colleague. Nigel Rees in Mark my words says:
Ronald Knox, compiling a list of rules for this kind of fiction in his
introductions to The Best Detective Stories of the Year, 1928, noted:
‘The only person who is really scratch on morals is the aged butler. I
cannot off-hand recall any lapse of virtue on the part of a man who
has been with the family for sixteen years. But I may be wrong; I have
not read all the detective stories.’
Some version of Occam’s razor might be applicable: we predict we’ll not find books with murderous servants – this is exactly what happened (according to the OA). To assume Van Dine meant another medium isn’t economical.
Was this rule even invented by Van Dine in 1928? Here’s The Technique of the Mystery Story (1913) by Carolyn Wells:
The next character to be chosen must be our criminal. Here again is
one, who, if he is to be convicted, must not be too deeply in the
reader's sympathy. And yet he must be a worth-while character; it is
old-fashioned, now, to have the crime committed by the butler or the
Maybe we’ll find something in The Door reviews?
The Atlanta Constitution is full of praise: “one of the best mystery novels of the age”, no butler mentioned.
Saturday Review doesn’t complain either:
The narrative is smooth and at all times entertaining, and the
solution, although not entirely satisfactory, is certainly well
New Outlook: “The denouement is logical and startling.” No butler again.
Now we have movie reviews, theatre reviews, and mystery books indicating the trope wasn’t popular in 1920s.
Judge magazine (1930):
“Oh, you have just started to read it? Isn’t it a swell book? It
fooled us right up to the last chapter. Of course the butler did it.
Mabel thought the old nurse did it. No, they kill her, too. Still I
don’t want to spoil it for you. It’s a good book tho’. So full of
And here’s 1933 Runyon’s story What, No Butler?
“But,” I say to Ambrose Hammer, “you do not pin the foul deed on any
of these parties, but on the butler, because this is the way these
things are done in all the murder-mystery movies and plays I ever see,
and also in all the murder-mystery books I ever read.”
One way to think about it: the writers are ridiculing the clichéd mysteries. Since there are no such books, we have to return to previous discussion of movies and theatre – which looks like a stalemate. Another way: the writers used the butler for some unclear reason. But it’s hard to figure out why the butler.
Returning to practical considerations, I listen sometimes to podcasts with comedians, and they say that they make jokes not because it’s true, but because it’s funny.
How jokes work? Writers like to say it’s important to use precise details. Suppose the author writes: “Of course Mr. Smith did it.” Or “Of course Jane Jones did it.” Something is wrong here: who is that Mr. Smith – it’s hard to have a vivid image. It’s not funny – the author has to be more precise. The same book The Art Of The Mystery Story cited in the OA has a good example:
He can say that the butler did the murder, or the chief of detectives
did it, or charming little Candace (“friends called her Candy”) did
it, or the quiet librarian did it, or Jim did it to shield Mary, or
Count Xerxes did it because Solange had got him under her skin…
So we can’t have “Jim did it”, it should be “Jim did it to shield Mary” – but the joke loses tightness now! We need some time to process the situation. If that’s correct, of course the butler did it – everybody else has a name!
Butlers are both authoritative and subordinate. Most people don’t have butlers, so there’s reason for envy. They are somewhat emblematic of the genre: as Adam Sandler puts it in Murder Mystery:
I don't even think there are butlers. That's just a word created for
those goofy books you're addicted to.
Some comic (and scapegoat) potential of the butler is discussed in a thesis Liminal Butlers by Katie Smith (PDF).
In Runyon’s story, the narrator states that the culprit must be the butler because this what he gets from all the plays, films, and books. When the perpetrator is revealed, by coincidence he happens to be a professional butler out of job. As a practitioner, Runyon could find it useful for his character to make a zany claim, which later will turn out to be true in the universe of “Guys and Dolls”. Runyon’s life and works were connected to Broadway. Now I remember something from Calkins’ research…
The detective and the butler were so prevalent that actor Reynolds
Denniston, the chief detective in “Whispering Wires,” had the wild
idea to create the “stage detectives association.” Not to be left out,
this venture was soon followed by the establishment of the “stage
butlers association” by Stanley Harrison, who plays a butler in the
same play. (Unfortunately, I could find no clues as to what may have
happened to either of these associations.” (Washington Post, April 29,
Alexander Woollcott joked about “one of the many butlers in which such dramas abound”.
Now I see my mistake. I thought the butler was just a random character interchangeable with everybody else. It's not correct.
Comedian Jack Gilford employed this line in his routine in 1938. But Gilford doesn’t say it was a reference to a popular cliché. He says he, Jack Gilford, was the first person to use it in a comedy way. A book 170 years of show business tell it this way:
“He went out on the stage and did his movie routine, in which he gave
quick capsule summaries of various kinds of movies, with a description
of a Grade B mystery where the first scene has a castle on a hill, the
moon scudding behind the clouds, and when the hero knocks on the door
the butler opens it. And, Jack said, "We all knew then that The Butler
Did It." The line has become part of the language and Jack invented
it. That opening-night audience laughed.
Cafe Society : the wrong place for the right people by Barney Josephson:
Everybody's copying his routines, his slow-motion golf ball, his movie
scenes. Jack Gilford became the byword in entertainment.
Maybe the phrase was used in comic context before, but Gilford can claim his routine was popular and was copied.
A mystery writer Chris Lansdown examined Runyon’s story in his blog and, being a practitioner, wasn’t convinced by satire claims:
Very clearly, in context, this was not a criticism of the butler as a
culprit, but playing with the audience’s expectations to set up a
He also speculates that the phrase could come from jokes (about the butler never being the culprit). It’s not far from B.J. Rahn’s suggestions we started with.
Experienced practitioner thinks the trope probably came from jokes.
Respected scholar doesn’t believe in “criminal butlers” examples influencing the trope.
The evidence gathered by the OA seems to support this too now.
I conclude that most plausibly there were no examples whatsoever. It happened through practice (bottom-up). It’s not clear were the jokes based on the butler being the suspect, “least likely to be guilty”, or practical needs. Anyway, this is beyond the scope of the question, which is about examples of criminal butlers.
But people with different worldviews have different intuitions of what is clear. Further research is needed to settle the question.