“…he would tell a story of mediaeval sorcery and, at the end, abruptly explain all the puzzles in the fashion of a detective story.”
“This case he was going to talk about was rather special and sensational. … Crime and the occult! These were the only hobbies for a man of taste!”

[CAVEAT LECTOR: there are going to be major spoilers for John Dickson Carr’s The Hollow Man/The Three Coffins in this post. If you haven’t read that, I advise that you do so—it is a seminal work in detective fiction—and then come back to this post. The point of the post (now that sounds Chestertonian…) is to come up with an alternate solution to the book. OK, don’t say I didn’t warn ya… I should also note that this is going to be a text-heavy post, as it is a lot of explanation. Sorry about that.]
This post has been long-promised, and it began like this:
JJ, from The Invisible Event, reviewed an atrocious book called The Third Gunman (2016), by self-published author Raymond Knight Read, for his amusing (for us, though not often for him!) “adventures in self-publishing” series. JJ wrote: “Essentially it’s a reimagining of John Dickson Carr’s The Hollow Man/The Three Coffins which someone who has read both Carr and Paul Halter found commendable — throwing around words like ‘baffling’ and ‘satisfying’…”

…………………………………………….. (both in the spirit of Read’s lack of knowledge about ellipses and of warning the reader that the spoilers begin here)
Back to JJ:
Imagine my disappointment, then — nay, my frank, bewildered, aghast astonishment — when this turns out to simply be The Hollow Man. The plot of The Third Gunman effectively runs thus: man is confronted by someone from his past, goes to kill that person while giving the impression he is in his office, gets shot in the act of killing them, returns to his office under the guise of the person he has just killed, sets off a gunshot to make it appear that he’s just been shot after locking himself in his office, and dies of his wounds as people break in to discover him. This is the plot of The Hollow Man, without even so much as a meaningful deviation to make it surprisingly clever.
Now, if I had read that book (and now I’ve read a sample), I would have had to resist the very persuasive temptation to “… lynch the publisher, shoot the bookseller, and strangle the author,” to put it in Carrian terms. In this case, all three seem to be more or less the same person, which would seem to infuse in the reader the desire to emulate the murderers of Rasputin. Our JJ, thankfully, is more sanguine than I, so the lucky Mr. Read got off with only a stringent review of his book. Such is, however, only Part I of the preamble. (*Here, the reader groans.*)
In the review, JJ gave us this very intriguing little nugget of information:
Read does at least try to work in a false solution, but this is dismissed for no reason other than he doesn’t want that to be the answer.
to which I responded:
I suppose my question is, is it possible to find another explanation to Carr’s set-up in The Three Coffins? … Perhaps Read’s false solution has a kernel of a good idea there? Or no?
JJ couldn’t quite recall, but he did write that Read’s false solution “…does something interesting…” Now to me: “Perhaps I shall try to take on the challenge of plotting another solution to Carr’s problem as set forth.”
Thus endeth Part II of the Preamble. (*“Thank God!” says the reader.*) That was my challenge, and… Well, let’s see. I shall try to tackle the first of the problems Carr sets forth in The Three Coffins—the murder of Dr. Grimaud and the vanishing of the “hollow man” from Grimaud’s study—as that was the one Read replicated [i.e., plagiarized]. (I should note that I did not, and do not, intend my solution to be half as satisfying as Carr’s—I knew from the get-go that it may well sound like a false solution, which is more or less what it is. Still, it’s an interesting thought-experiment.)

And now, as they say, “ze plot zickens.”
I should note that, in trying for “another solution to Carr’s problem as set forth,” I tried to eliminate a few minor possibilities. For example, one could change the Grimaud problem as to make it that Grimaud shot himself and then did everything in the room—a solution both disappointing and so similar as not really to warrant a discussion. (It was used in an early Simenon tale that might have inspired Carr’s book.)
I’m not particularly interested, then, in minor variations on Carr’s solution; I’m interested in a wholly different solution to the same problem—which is how it became increasingly clear to me that I had to have two people actually in that room.
The ingenuity of Carr’s solution is that, where it seems that two people (Grimaud and the murderer) were in Grimaud’s study, in reality there was only one. It is thus that the problem becomes impossible; there really is a “hollow man,” one who does not exist. The mirror-trick is only a secondary factor to the primary principle that there is only one man in the room; this latter point is, in reality, the trick on which the study-problem (and, in the same way, the street-problem) turns.
If we blast that to smithereens, then, and say that there really were two men in that room…where does that leave us? Well, Carr gives us the locked-room lecture (hereinafter LRL) for a few reasons, but one is misdirection: it gets us back to thinking about a solution in which there were two men in Grimaud’s study. If we take that point as a given, then we are bound, I think, by the LRL itself—specifically the part that may superficially seem the least important, near the end:
“Then we come back in a circle to the window, don’t we?” demanded Hadley. “You’ve gone on and on about ways that obviously couldn’t have been used. [This is actually a fair criticism on Hadley’s part—but, obviously, it helps JDC’s misdirection technique.] But in this catalogue of sensationalism you’ve omitted all mention of the only means of exit the murderer could have used…”
I think (with one exception, which I’ll go into in a moment) that Hadley is correct here. If we operate on two premises—(1) that there really were two men in that room and (2) that the murderer left the room—then “…we come back in a circle to the window…” Yes, indeed.
Let me note the one exception here, which I’m sure you’ve already seen: a denial of Premise 2, or the argument that the murderer never actually left the room at all. It is this oversight in Carr’s lecture that Clayton Rawson [through the Great Merlini] noted in Death from a Top Hat, in his own LRL. It can be used quite well indeed—there’s an episode of Murder, She Wrote, of all things, that uses it excellently—or it can be rather goofy. But I suppose it’s that way with most locked-room solutions. In this case, I was trying, more or less fruitlessly, to work out such a solution. With all of Carr’s pronouncements that Fell, Hadley, et al., searched the room to make sure that this isn’t the solution, it was fairly difficult, but there is one, rather unsatisfying way in which it could have worked, and that involves the sofa.

(This map is going to become very useful for us.)
As you may recall, “Just before the fire, a long brown-leather sofa had been knocked awry and a leather chair rolled back in a twisted-up hearthrug. There was blood on the sofa.” I wondered if the murderer could have hidden in the sofa, and I suppose it’s technically possible if it has cushions, if exceedingly unlikely. Thus: the mysterious visitor enters the room, shoots Grimaud, and goes over to the sofa. He uses the knife (the one used to slice up the painting—to suggest a struggle, I suppose) to rip out the inner stuffing of the sofa, which he then burns in the fireplace (or throws out the window, or even keeps with him). He then hides in the hollowed-out area, puts the cushions over him, and waits until the police have left the room to make his escape. An utterly ridiculous solution? Absolutely, and with a huge number of flaws. Granted, and it’s shot all to blazes if the sofa is (e.g.) a loveseat. Back to the window.

Carr analyzes in some depth the possible ways in which the murderer could have used the window to effect his escape—and rules them out, confounding our expectations (ch. 3 and Mills). This is vital for our solution, so I can’t help quoting in full:
There was a bright moon in the west, and every detail stood out sharp as a woodcut. It was a good fifty feet to the ground; the wall fell away in a drop of smooth, wet stone. Just below there was a backyard, like that of all the houses in this row, surrounded by a low wall. The snow lay unbroken in this courtyard, or any other as far as they could look, and along the tops of the walls. Below in the whole side of the house there were no windows whatever. The only windows were on this top floor; and the nearest one to this room was in the hallway to the left, a good thirty feet away. To the right, the nearest window would have been in the adjoining house, an equal distance away. Ahead there lay a vast chessboard of adjoining backyards from houses lining the square, so that the nearest house was several hundred yards away. Finally, there stretched above this window a smooth upward run of stone for some fifteen feet to the roof—whose slope afforded neither hold for the fingers nor for the attaching of a rope.
Now that would seem to do it, especially as Hadley’s first solution (the killer has tied a rope to the chimney and swings out to the roof after killing Grimaud) is thrown out because there are no prints on the roof. But the important point is this: the killer didn’t have to get to the roof. He only had to get to the nearest window—“a good thirty feet away.” We know fair well that he can’t climb that distance, as there are neither finger- nor foot-holds. But he can swing.
And therein lies the rub. Thanks to the website “Measure of Things,” I know that thirty feet is “about one-and-one-tenth times as long as a London bus” or “four-fifths as tall as a telephone pole.” (Or see here.) With a leap from the window, a sturdy chimney, and an especially strong rope, that’s not impossible. And we know that points 2 and 3 are at least likely: in Ch. 8, Sgt. Betts ties ropes to the chimneys to check the gutters. But how did the killer get the rope around the chimney, and how did he have a (literal) ledge to stand on once he swung over?

Now there are two ways to do this: the killer either swung left, to the windows on the diagram, or right, to the house next door (both 30 ft. away). As we don’t know exactly where the chimney is, however, Option 1, though less helpful for our purpose, seems more reasonable than Option 2. If one chimney were close enough to the house next door, the killer could have used that house as a base of operations, opened that window, and swung over there. If not, it makes our job a little more complicated. Nevertheless, it’s possible that he could have gotten over to one window, having purposely left it open, and entered the house in that way. It is stipulated that the windows in the hall are all locked, but we don’t know exactly when they were locked. He certainly would have had to tie the ropes to the chimney[s] before the snow started (having gotten to the roof by the trap door before Mills got there), or there would have been tracks.
Could he have done this without Mills and/or Mme. Dumont seeing? Hypothetically, we could explain this away with Mme. Dumont as the accomplice, or we could go with Option 2 (there are other ways to do it, too, I believe, but I’ve gone on too long already).
Now, as I said, this solution is rather disappointing, especially when compared with Carr’s, but I believe that it works. (Now, I’m sure, there’s something major I missed.) The only two others I can think up—Mills being fooled as to footprints on the roof and a play with the time the “hollow man” walked in—are slightly less workable, I think.
All right, well… As I said, this post has been long-promised, so I’m happy that I have a solution down on paper (well—you know what I mean) at long last. I’m more than willing to explain (or rip the whole thing up because of) any possible criticisms. Just a thought-experiment, but a fun one.
