There are many types of clues in a detective story which can be left at the scene of the crime. They are often looked at from the perspective of the detective, or really of the reader, since he is on the detective’s side. I think it might be profitable to look at them from another perspective: from that of the murderer.
We can first divide clues found at the scene of the crime by whether they help the murderer or help to catch him. After that, we can divide them based upon whether they were intentional, unavoidable, or accidental.
There are not many sorts of clues which help a murderer. Aside from clues which lead to disprovable theories, by which plot a murderer can be found innocent at trial and thus protect himself by the legal prohibition on double jeopardy, all clues which help the murderer must lead the detective to suspect someone else. We can divide these up by whether they mislead to another person, or only to a trait which the murder doesn’t have.
A clue which misleads to a specific person can be desirable or undesirable from the murderer’s perspective. Sometimes a murderer hates the person toward whom the clue leads; other times the murderer may accidentally framed his fiancé. There is wide latitude here—any place on this spectrum is workable. Clues which accidentally frame the murderer’s fiancé will probably need to be accidental clues, such as the fiancé having been in the room for some other purpose before the murder took place and dropping an earring or a cigarette. On the flip side, clues which implicate someone the murderer dislikes or even hates can be either purposeful or accidental. Both can be made to work; it is not so distasteful to have luck on the side of the murderer as it is on the side of the detective. That said, it will probably be more satisfying for such clues to be intentional. Stories in which all of the complications turned out to be an extraordinary run of (bad) luck can be interesting, but they almost need to be titled Much Ado About Nothing. (Despite there being no coincidences in that play and the misunderstanding being the work of the villains.) Where such clues are by design, this tends to require quite a lot of planning on the part of the murderer, since he has to ensure that the person he’s framing has no alibi. This will almost certainly involve some risk on his part; it’s not easy to know what someone else is doing without being unobserved. Such clues will, except by very good luck on the murderer’s part, only work in a highly premeditated murder.
Clues which lead, not to specific people, but only to traits that the murderer doesn’t have give more leeway in how they are done, but are also more hard to pull off. A good example of this would be Chesterton’s story The Hammer of God. The victim was killed by a tremendous blow with a hammer, which points to an enormously strong man (which the real murderer was not). Another example which comes to mind is the Lord Peter Wimsey story Busman’s Honeymoon. In that story, the blow to the head points to an extremely tall man, which, again, the murderer was not. Come to think of it, it’s curious that both of my examples involve a blow to the head. It’s not necessary, though. It’s quite possible to shoot from a higher place than the natural standing point of the murderer, suggesting a taller person—bullets even have the advantage, if they pass through the body, of giving a second point in space to line up, showing the height more clearly.
The downside to this is that the number of traits one can indicate via evidence is limited. Height, strength, possibly in some circumstances weight—or more likely not being above a certain weight—are about all that come to mind. It would be possible to have evidence which seems to eliminate certain disabilities, though. A gunshot which requires sight, or something done when one hears a sound, or distinguishes who someone is by their voice. Those get quite special purpose, though, since the field of suspects has to get small if a blind man is the murderer and the evidence seems to rule out a blind mind. I suppose one could set a murder in a conference for the blind, but otherwise there just aren’t enough blind men around to fill up a suspect list, once the main issue of the evidence not actually ruling out a blind man is found. It’s the sort of thing which would work pretty well in a short story, but I doubt it could sustain a novel.
The other major classification of clues are clues that help to catch the murderer. These are the meat of a detective story. Without these sorts of clues, detective stories must be in vain. Clues which help the murderer are optional, but clues which hinder the murderer are mandatory. Very well, then. What can we say about clues which help the detective to catch the murderer?
I think that the most important thing to consider with such clues is that from the perspective of the murderer, they are mistakes. In designing these clues, we are choosing what mistakes the murderer will make. This will, of course, be a function of the murderer, whether this was planned, and the circumstances that may intervene.
The first category encompasses the murderer’s intelligence and imagination. There can be a pretty big variance here, though if the murderer is too lacking in either there won’t be much to investigate. The two do not even necessarily go in sync; a young murderer might be quite intelligent but not very imaginative through lack of experience. The reverse can be true as well—a murderer with wide experience of the world might be quite imaginative, though not highly intelligent. The experience would have to be relevant, though, which I think would mostly limit us to murderers who are or have been detectives. That’s a very specialized murder mystery. I suspect one could broaden it out to a person who is very familiar with the workings of a particular place, such as intimate knowledge of how a hotel or other business functions, to know who is where, when. The downside to that specialized knowledge taking the place of intelligence is that it will be harder to hide the murderer, since not many people will have that knowledge.
An interesting sub-category of intelligence and imagination would be the murderer trying to disguise a clue as something harmless and not realizing how it will look. An old school example might be a murderer flicking a burnt-out cigarette end onto the ground where there are others, and assuming that it will be taken to just be one among many. In that case the detective usually finds it either by its residual warmth or by a lack of dirt on top of it or some other sign which the murderer didn’t think would be present to indicate what it is. I’ve also seen cases where a murder thought something would be taken as belonging to the victim when a better knowledge of the victim would show that it didn’t. A good example of this would be the razor in the Lord Peter Wimsey story Have His Carcase. It would be plausible for most men that they owned a cutthroat razor (back in England in the 1930s), but minor investigation of the victim showed that he was extremely unlikely to have one. The pursuit of this clue helped to catch the murderers.
The second category is almost self-explanatory. If the murder was planned out, all else being equal there will be fewer clues in the crime scene since any quarter-way decent plan will have the avoidance of such clues as a primary consideration the forming the plan. A more impromptu murder will lend itself to the presence of more clues. In both cases, however, any clues left will have to pass the stage where the murder is leaving the scene of the crime. Unless fleeing in haste, the murderer will, presumably, look over the scene for clues to remove.
This is where circumstances can intervene to preserve clues for the detective. It can take the form of introducing something that makes the murderer flee in haste, of course. It can also take the form of something which conceals the clue from the murderer during his investigation. Something falling over in the death throws of the victim, for example, might conceal a clue beneath it. Poor lighting can also make a murderer overlook a clue during the cleanup stage. The field for intervening circumstances is very wide. Even pet animals that steal clues have been used successfully. Books that were put back when they should not have been can be a clue, or contain a clue. And of course there are the environmental clues that were so popular back in the early days of detective fiction. Things found that would not be noticed except that they were damp among dry things, or dry among damp things, or clean among dirty, or recent among moldy; the list can go on and on. Changing conditions also work well here; an unpredictable rain can show that the victim was killed before the ground was wet. Temperatures plummeting to below freezing can preserve a clue that would normally have melted away. Even the reverse is possible; hot temperatures can melt away a clue that was meant to make the murder look like suicide.
This very anemic overview of types of clues is meant only as a starting point. I’m not sure that I’m going to make it a series, but as time permits I think I’d like to go through each of these sorts of clues, one at a time, to consider them more closely. Until then, I hope that this systematization was of some interest. When writing a murder mystery imagination is key, but a little bit of order can help to pick among the vast array of possibilities.