This is essentially a follow-up to my previous post, with a more practical focus, but it shouldn’t be necessary to read the earlier post to understand this one.

The pro-gun activists in the United States have a slogan: “guns don’t kill people, people kill people” (parodied by Welsh rap act Goldie Lookin Chain in the song Guns Don’t Kill People, Rappers Do.) The basic idea, presumably, is that guns, being inanimate objects, clearly cannot take responsibility for killing: rather, the responsibility for killing lies with people who use guns to that end. (And therefore we should, the argument goes, focus our attentions on stopping people from using guns to kill people, not on getting rid of guns themselves.) Even if we disagree with the sentiments behind this, we have no trouble understanding what is meant.

This is an interesting use of language because, from a strictly literal viewpoint, it’s undeniable that guns do kill people. Not as animate, volitional “agents”, of course, but nevertheless Guns kill people is a perfectly acceptable English sentence. And indeed, it’s quite normal for inanimate, non-volitional “instruments” to be used as subjects: there’s nothing syntactically or semantically wrong with Scissors cut paper or The knife sliced easily through the soft, white cheese.

Perhaps, we might argue, kill is different from cut or slice – it requires an animate agent as its subject. (Maybe it’s a bit like eat, which can’t take an instrument as its subject: as I pointed out in my last post, we can’t usually say The fork ate the peas to mean “someone ate the peas with the fork”.) But this is clearly false: surely nobody has any problem with The avalanche killed the skier or Trains kill people who ignore red lights at level crossings.  

No, Guns kill people is fine (strictly speaking, at any rate). But the aforementioned slogan does highlight something interesting about attitudes to language: although there’s nothing ungrammatical or unmeaningful about a sentence with an instrument as its subject, there is nevertheless a feeling that volitional agents make better subjects, and perhaps that it may even be in some sense incorrect to use an instrument as a subject when an agent would be available instead.

We see something similar in the arguments by cycling campaigners (e.g. this article) regarding the use of language in journalism relating to road collisions. Often, newspapers phrase things along the lines of A car collided with a cyclist or A lorry ran over a pedestrian. This, the cycling lobby claims, is undesirable because it appears to remove responsibility from the drivers of motor vehicles: cars and lorries do not generally run into things of their own accord, but because of actions taken by their drivers. In other words, given that in such incidents there is an agent (the driver), it is infelicitous to promote an instrument (the vehicle) to the status of subject.

Of course, in parallel with the gun case, an inanimate thing like a car or lorry is a perfectly acceptable subject of a verb like collide or run over as far as grammar or literal meaning is concerned. But the cyclists’ arguments nevertheless highlight, and indeed rest upon, an intuition that volitional agents are once again “better” subjects than instruments. Ordinary users of English have an impression that some types of construction are preferable to others, even when both are technically acceptable: an impression which links closely to what linguists have described as “thematic roles” like agent and instrument. This intuition may seem to support the linguistic analysis that agents are subjects by default, and instruments are only promoted to subject status when an agent is absent.

(In other cases the line between what is merely inappropriate and what is grammatically/semantically unacceptable becomes blurred. The article I linked to gives the example of [the cyclist] collided with a van, referring to an incident where the van was driven into the cyclist from behind. We would probably think of the cyclist here in terms of the thematic role of “patient”: he was not the principle cause of the action, didn’t bring it about on purpose and was the participant most affected by it. Is the use of a patient as a subject syntactically acceptable (as the journalist would appear to think), even if it is an undesirable phrasing, or is it just wrong in every way?)

So: even though things like thematic roles may seem like quite abstract linguistic concepts, it appears that they do have a role to play in the ways in which even non-linguists think about language – and in what is deemed advisable not merely semantically and syntactically, but socially as well.