Sometimes research goes badly, and red herrings lead to a blind alley (qv), but sometimes there’s a smoking gun: conclusive evidence just a little bit short of being caught in the act. But note “just a little bit short.” This means that smoking gun shouldn’t be applied to properly conclusive evidence, and that it doesn’t make sense to ask if something is a smoking gun. A smoking gun is obvious, in-your-face evidence; the question is whether it’s valid, well illustrated by these opening sentences of a commentary article on the evidence that T cells cause rheumatoid arthritis. This article came to my attention some years ago for its title, “A fistful of T cells”, which, rather than metaphorical, is an allusive title. Of course, the important property of the smoking gun is the temporal one, and that is not so for most evidence in research, but the sentences capture the essence of the smoking gun.

“Outside the saloon lay the sheriff, wounded. In the middle of the gathering crowd stood a man holding a smoking gun. He claimed to be innocent. Nobody had witnessed the crime. Whodunit?”

In a similar, and similarly correct application of the metaphor, a particular peptide is suggested as the trigger for the evolutionary change in a clade of cone snails than enabled them to hunt fish instead of worms. But it is incorrect to write that “cigarettes [are] more of a potential… than a definitive smoking gun [for] reduced taxane myelotoxicity” because the whole point of a smoking gun, as above, is that it isn’t potential and it can’t be definitive. I guess that when looking for the agent in tobacco that is responsible at the cellular level for carcinogenicity, smoking gun is just too compelling a metaphor. It is correct but ill chosen to call the Chicxulub meteorite crater in Mexico a smoking gun for the Cretaceous-Tertiary boundary extinction: there would have been a lot more smoke than from a gun.

Smoking guns are obvious; needles in haystacks are not, and there are twice as many needles in haystacks as smoking guns in PubMed®. I don’t find it objectionable as a way of saying that it’s very difficult looking for bits of broken minimal access kit in a child’s abdomen, searching for gravitational waves with ground-based interferometers, or looking for circulating tumour cells as prognostic markers; but is it appropriate for identifying the miRNA-target pairs that may be important in disease mechanisms? The problem there is that there are too many needles and we don’t know which are the important ones, but what is the haystack? It is also sometimes applied to difficult diagnoses, but that again is inappropriate because the problem is not knowing whether it’s a needle or a nail.

If you want to use needle in a haystack, don’t use it if you’ve actually lost bits of surgical needle.

Neville Goodman is a retired consultant anaesthetist and a writer, and co-author of a book on medical English.

Competing interests: I have read and understood BMJ policy on declaration of interests and declare that my only competing interest is my co-authorship of a book about medical English.