Any analysis of locked room mysteries must begin with John Dickson Carr, the acknowledged master of the genre. He wrote more classic locked room mysteries, perhaps, than any other author and teased his readers with ingenious means by which seemingly impossible murders were pulled off in the confines of an apparently impregnable sealed room. The solutions to these puzzles – howdunnits – were required to be possible but need not necessarily be probable. Indeed, the murderer might frequently have carried out his crime more easily by other means. Means which might not have been investigated so closely as an apparently baffling and insoluble crime might be investigated. Who, after all, would call in the expert detective, when faced with an obvious, but wrong, solution?

In these stories, and frequently they were short stories rather than full length novels, where the tension of finding the solution to howdunnit questions is more difficult to maintain, greater emphasis was generally placed on the puzzle than on characterisation. Here, perhaps, even more than elsewhere in Detective Fiction of the Golden Age, solving the puzzle was the game being played between author and reader.

This analysis of the genre will examine trends in these mysteries over time and look at other factors which might influence the likelihood of one solution being favoured over another by the author. It may therefore assist you in directing your thinking to the more likely solutions for any given story but it will never give away a solution.

According to John Dickson Carr, there are no less than seven distinct types of locked room mystery. At least that is what his character Dr Fell tells the readers in chapter seventeen of his 1935 classic The Hollow Man (widely regarded as one of the masterpieces of that genre). To summarise Dr Fell’s categorisations as a reminder for those who have read the book some time ago or to enable those who have yet to do so to benefit from Dr Fell’s elucidation while skipping over the relevant chapter (as Dickson Carr invites them to do) so they can get on with the plot, the categories he proposes are:

1. The murder is not a murder but is, in fact, an accident. The circumstances are such that it appears there has been a murder but this is not the case. Instead there has been a fatal accident within the locked room. Hence no murderer was present or has made his escape without leaving any trace. Often the accident will involve a fall with the victim striking their head a blow on the fender. This method is becoming more difficult to carry off now with open fireplaces being replaced by central heating.

2. The murder is achieved by means of a poison gas which overcomes the victim (perhaps driving him into a frenzied paroxysm which causes damage to the furniture leading to investigators mistakenly believing a desperate fight has taken place between the victim and the murderer).

3. The murder is done by a mechanical trap planted in the room, which is set off by the victim while the murderer is safely elsewhere. The trick here is to make the method by which the trap is sprung undetectable and, if the trap is concealed, for the weapon to return to (or else remain in) its place of concealment after being triggered – such as a gun hidden in the workings of a clock which fires when the clock is being wound (a method which surely is falling out of fashion due to the inexorable rise of battery or mains electricity powered clocks).

4. It is suicide which is rigged up to look like murder, frequently with the intention of incriminating an innocent party against whom the suicide holds a grudge. The weapon might be an icicle with which the victim stabs himself. The icicle then melts, speeded no doubt by the fading body heat of the “victim”. In the absence of a weapon in the body, murder is presumed with the supposed murderer having made his escape with the weapon.

5. The murderer impersonates the victim after having first killed him. The murderer is later observed to enter the room disguised as the victim. He emerges immediately afterward having slipped out of the disguise thereby giving himself the alibi of having been seen to leave the room without having had sufficient time to commit the murder. The timing of the death is critical in this category; the elapsed time between actual and supposed later time of death must be sufficiently short for the body to be in an appropriate state of rigor mortis and at the correct post-mortem temperature.

6. The murderer manages to carry out the murder from outside the room in a manner which suggests that the murderer was inside the room to carry out the killing. Bullets made of ice, or even frozen blood, have been fired in through windows and subsequently melted leaving no trace to detect the method used.

7. This is the reverse of category 5. Here, the murderer has merely stunned or otherwise rendered the victim unconscious. They leave the room and after a suitable interval, during which no-one enters or leaves the now locked room, ensure they are on hand as efforts to break down the door are made. They ensure they are first into the room and in the initial confusion after entry is gained, they killed the unconscious victim swiftly – a stiletto is a favoured method – while misdirecting the others who have crashed into the room with them. This gives the impression that the victim has been lying dead in the hitherto locked room for some time.

If any of you are currently reading a book where it turns out that one of these tricks has been used, please don’t blame me for having outlined the method here. I’m only repeating Dr Fell’s 80 year old lecture which anyone can read for themselves.
Naturally, others have endeavoured to categorise such mysteries differently. Robert Adey in his book Locked Room Murders identifies twenty possible solutions:

1. Accident
2. Suicide
3. Remote control – the use of poison gas or the victim is impelled to kill themselves inadvertently
4. Mechanical or other devices
5. An animal carries out the murder
6. An outside intervention is made to appear as if the murder has taken place by a murderer inside the locked room, e.g. by throwing a dagger through a window at the victim from outside the room
7. The victim has been killed earlier but is made to appear as if they were alive at a later point
8. The victim is presumed to be dead but is in fact killed later than believed, e.g. by first person to enter the room
9. The victim is wounded outside the locked room but enters, locks the room and dies inside
10. The key, bolt or catch securing the door is manipulated from the outside, using pliers, string or some other device, to lock the door after the murderer has exited
11. The door or window of the room is unhinged and removed to gain entry and the murderer then replaces it after committing the crime
12. As above but confining the removal to a window pane
13. Entry to the room is gained by some acrobatic manoeuvre
14. The door is locked or wedged from the outside and the key is only replaced on the inside after entry to the room by those who find the body
15. As above but the key is returned into the room before everyone enters to discover the body
16. Other methods of gimmicking the door or windows
17. The murderer enters and exits through a secret panel or uses one to enable the weapon to be propelled at the victim
18. The murderer is in the room all the time
19. The murderer is provided with an alibi for the critical time when the murder is committed
20. Other impersonation stunts

It will be apparent that the more detailed breakdown suggested by Adey is a refinement of Dickson Carr’s approach. While it offers greater precision, for the purposes of statistical analysis it is more problematic. In order to draw meaningful conclusions from analysis of a sample, if there are twenty possible categories rather than seven, you will need a vastly larger sample to ensure that the resulting numbers are not spuriously accurate.

In fact, Adey’s twenty solutions can be allocated, or in some cases “shoehorned in”, as mere subsets to one or other of Dickson Carr’s categories.

As can be seen from the table, fully half of Adey’s twenty methods are variations on the theme which Dickson Carr pulls together under the catch-all heading of being achieved from outside.

Which brings me now to the vexed question of how to satisfy the ideal of selecting a random sample of locked room mysteries from which to draw statistically meaningful conclusions.
Out of the countless thousands of detective fiction stories I have neither the time, nor the boredom threshold to read them all and identify the “population” of locked room mysteries from which to derive a random sample for analysis. I must, therefore, fall back on a more practical method of selection. I will turn to a number of compilations, edited by recognised authorities on the genre, and trust that these will provide a suitably randomised sample for my purposes. I would contend that since the editors’ selections are based on aesthetic/literary grounds with an eye to commercial success, and not based on the criteria under which I am intending to make my analysis that their choices are suffieciently “random” for my purposes. It also, thankfully, means I get to read a selection of stories that the editors, whose views I respect on such matters, think are going to be amongst the best of the genre.

My sample for analysis is therefore the collected stories in the following compilations:

The Mammoth Book of Locked-Room Mysteries and Impossible Crimes edited by Mike Ashley

Miraculous Mysteries: Locked-Room Murders and Impossible Crimes edited by Martin Edwards

Ye Olde Book of Locked Room Conundrums edited by J.J.

The Black Lizard Big Book of Locked Room Mysteries edited by Otto Penzler

These give me a sample for analysis of 114 stories which should suffice to draw some meaningful conclusions about the writing, and reading, of locked room mysteries.

The first caveat I have to stress is that I am reading in English which gives a huge bias towards the UK and US markets for detective fiction. The “other category” includes Ireland, South Africa and Australia – all native English speaking – with only a couple of translations from the original French. This sample therefore doesn’t reflect the diversity of locked room story-writing globally but does reflect the paucity of translations of short-stories – sadly a reflection of commercial constraints, there being little profit in a short story sale compared to a sale of a novel which might therefore be worth translating for the English speaking markets. I would also add that the slight weighting in favour of UK over US mysteries is largely attributable to the selection of stories in the Martin Edwards compilation which is devoted to UK writers of the genre.

The period covered by the four compilations starts in 480BC (a single outlying example in the J.J. compilation) but really starts in the first half of the nineteenth century.

The chart shows a growing interest in the locked room mysteries from the beginnings of modern detective fiction in the nineteenth century through to the second world war. There is then a marked decline up to the current time. It is not too strong to describe interest in the genre as going off a cliff edge during the second world war. It is interesting to speculate why this might be. One possibility is the shift away from the puzzle element of the Golden Age of detective fiction to a greater emphasis on character. Certainly the locked room mystery places less emphasis on characters and motivation and more on methodology of the criminal. Indeed, it is often as much a howdunnit? as a whodunnit? with consequence precedence of plot over psychological motivation.
I should also mention that the above chart excludes fifteen of the stories included in the Mike Ashley compilation on the grounds that this modern compilation was intended to feature heavily works written (some even commissioned for the compilation) by contemporary writers of the genre. Their inclusion would have resulted in a massive spike in the final decade of the chart which would distort the picture and create the impression of a resurgence which, while this may be enjoyed by a small but dedicated audience for locked room mysteries, is not an accurate reflection of a wider popular interest.

We can now look at how these 114 locked room mysteries fall into the categories outlined by Dickson Carr. In order to avoid yet more shoe-horning, I have included, in addition to the seven categories of Dr Fell’s lecture, three additional categories which I think are sufficiently distinct as to merit recording separately. These are: Unbreakable Alibis, Impossible Crimes (i.e. in the spirit of the locked room but not involving a locked room or its equivalent) and Others, as a catch all for stories that defy categorisation under any of the other headings no matter how hard I try to get them to conform to one or other of the headings.

As you might expect, the most common form of locked room solution is that the murder was committed from outside – often by one of the sub-headings identified by Adey. I ought to emphasise that none of these murders from outside is so mundane as simply shooting the victim through an open window, which would not count as a true locked room mystery unless, for example, it was done with a pistol which was subsequently tossed into the room to give the impression that the victim had killed himself and then dropped the weapon. (This would qualify as a crude locked room mystery but might be improved if somehow the murderer contrived to have the victim to have scorch marks or powder burns apparently from the shot to increase the evidence suggesting it was suicide.)
I have a particular fondness for the next most common approach, the mechanical trap. These generally demand a high degree of ingenuity on the part of the murderer to set the trap in such a way that it will not be evident afterwards how the trap worked but, perhaps, even more than other locked room types tend toward a plot-bias over character-bias in the construction of the story. That the murderers could have achieved their goal more simply and at less risk of either failure or of detection by the experts called in to solve the crime is one of the ways in which I am happy for my credulity to be stretched a little.

The third most common solution is one used extensively by Dickson Carr himself in which the murder is seen entering and/or leaving the room which is subsequently found locked thereby “proving” that he could not have dunnit. This solution relies on accurate timing and an ability to misdirect both readers and the all important witnesses who testify to the murderer’s movements.

Poisoning – so easily achieved by the murderer in absentia – and the supposed murder turning out not to be anything of the kind are the other most frequently encountered solutions but, to my mind, they lack the sparkle of the more complicated solutions which, I think, have greater appeal as puzzles. That said, some of the reasons why it was not a murder are thoroughly ingenious.