WHAT does it mean to “bear arms”? The Second Amendment to America’s constitution reads: “A well regulated Militia, being necessary to the security of a free State, the right of the people to keep and bear Arms, shall not be infringed.” Concerned by the number of firearms in America, and the epidemic of gun violence they cause, many commentators (including Johnson) have in the past examined the first half of the amendment. It seems obvious to some that the first clause qualifies the second: the right to bear arms is tied to militia service.

But gun-rights advocates think the second clause stands alone. Among them was the late Antonin Scalia, who in 2008 wrote a Supreme Court opinion, DC v Heller, holding that the amendment guarantees an individual right to guns, no militia service required. He went on to explain “bear arms”. For him, “to bear” was simple enough, meaning “to carry”. And “arms” were just weapons. He conceded that there was an idiom, “to bear arms”, which meant to belong to an organised military force. But this was only a possible import of the phrase, not its core meaning.

Scalia was an originalist—ie, he believed the constitution must be interpreted in the light of the meaning of its constituent words in the late 18th century. He bolstered his argument by citing an edition of Samuel Johnson’s dictionary from 1773, plus selected prose from the period in which the constitution was written.

He was mistaken. Selective quotations can prove anything, if you have clever researchers looking for them. But there is a far more robust way to find out what people meant by this or that word in the 18th century. That is to gather a large number of texts into a “corpus”, a searchable body of material, and then look for patterns in thousands of uses of a word or phrase. A corpus can be general, like Google Books, which has around 500bn words of English text. But it can also be specialised. Two newcomers are the Corpus of Founding Era American English, with 139m words across 95,000 documents from 1760 to 1799, and the Corpus of Early Modern English, with 1.3bn words from 1475 to 1800.

Dennis Baron, a linguist at the University of Illinois Urbana-Champaign, searched for “bear arms” in these databases, and found about 1,500 instances. Of these, he says, only a handful did not refer to organised armed action. It is true that several state constitutions guaranteed the right to “bear arms” and explicitly mentioned self-defence. So Mr Baron’s digging does not completely close the case. But it has shown that the default meaning of “bear arms” in the founding era was, indeed, military.

This research ought to prompt the justices to revisit Heller—though given the weight of precedent and the court’s make-up, that is unlikely. Still, the dispute has several other interesting lessons. One is that phrases are more than the sum of their dictionary definitions. Context isn’t just helpful; it is often crucial. The verb “bear” has 44 definitions in the Oxford English Dictionary (OED), not counting the ursine noun. Which “bear” is meant can only be grasped in context. Bearing interest does not mean literally carrying interest around, nor does bearing a grudge involve physical activity.

Second, there are phrases, sometimes called “phrasal verbs”, that cannot be understood by knowing the component words: consider bear down or bear up. Good dictionaries define these phrases separately. The OED defines “bear arms” in an entry under “arms”: “To serve as a soldier; to fight (for a country, cause, etc).” But it also takes note of the contested meaning in America’s constitution.

In any event, real-world usage matters more than dictionaries. Judges often hunt through dictionaries to support their rulings, but these can miss nuances or make mistakes. Instead judges should go directly to digital corpora. Nor are selected quotes enough. In any other field, this would be called cherry-picking. Instead, with the powerful, free resources now available, anyone—including readers of this column—can look at a huge body of usages and draw firmer conclusions about meaning. (Neal Goldfarb, a lawyer, has made the “bear arms” data available on Language Log, a blog.)

Originalists like Scalia can find out what words really meant in the 18th century. But their opponents—who believe laws should evolve with the meanings of the underlying concepts—get a powerful tool, too. Lexicographers have revolutionised their work using such data. Time for lawyers to do the same.