Corpus Linguistics, Public Meaning, and the Second Amendment
Conservative jurists claim to focus on the text and nothing but the text as they seek to discover the original public meaning of the Second Amendment. But it’s not clear that the amendment ever had a single shared meaning, or if it did, whether that meaning is recoverable. That’s true of any text, not just legal ones. The best we can hope for when investigating any older text is to examine how it was discussed around the time it was written and to use historical sources to glean the meaning of any difficult or ambiguous words or phrases. And even so, there’s no guarantee that a reasonable reader in 1791 would interpret the Second Amendment the same way as their equally-reasonable neighbor. When the Supreme Court said in District of Columbia v. Heller (2008) that it had determined the original public meaning of the Second Amendment, its reading, like any linguistic interpretation, involved both an examination of the text and a certain amount of guesswork.
The modern concepts of textualism and originalism would have seemed strange to earlier legal scholars like Edward Saunders, who in 1571 called a statute’s words nothing but “verberation of the air.” Saunders taught that the significance of a law exists not in its words but “in the minds of the expositors of the words,” both the legislators who draft the statutes and judges like Saunders, who interpret them. Blackstone similarly observed that legal interpretation may require going beyond the text.
Still, all interpretation starts with text, and when a word or phrase in that text is problematic, judges have frequently turned to dictionaries for help. But Learned Hand warned in 1945 not to make a fortress of the dictionary. Perhaps he was thinking of Nix v. Hedden (1893), where the Supreme Court rejected the dictionary definition of tomato as a fruit because ordinary people consider tomatoes vegetables. Still, dictionaries remain popular legal tools. In Taniguchi v. Kan Pacific Saipan (2012), Justices Alito and Ginsburg cited a total of fourteen dictionaries as they backed competing interpretations of the word interpreter in the statute in question.
Corpus linguistics, which some hail as better than dictionaries for legal interpretation, allows us to access and analyze vast swaths of digitized text going back to the fourteenth century. A search of these corpora permits us to better understand some key elements of the Second Amendment’s meaning, though it’s not clear that the insights of corpus analysis will change the legal landscape.
An early corpus search of the verb ‘to harbor’ played a role in U.S. v. Costello (2012), where Costello appealed her conviction for harboring a fugitive. Opposing the appeal, the government argued that dictionaries defined harboring in the 1917 Immigration Act as ‘sheltering.’ But Judge Richard Posner wasn’t convinced. Googling harbor confirmed his intuition that the word usually involves, not sheltering someone, but ‘hiding’ them from the authorities. A more-rigorous corpus search by Gries and Slocum found that harbor more often means ‘to shelter’ than ‘to hide.’ But in the context of the Immigration Act’s penalties for anyone who “conceals, harbors or shields from detection” an alien, ‘hiding’ seems correct. Costello made no effort to conceal her fugitive boyfriend, who lived openly in her home and went about in public both alone and in her company. Nor did she shelter him. As Posner put it, “‘Sheltering’ doesn’t seem the right word for letting your boyfriend live with you.”
Since Posner’s foray into database searches, corpus linguistics has acquired more courtroom buzz. In 2019, Judge Amul Thapar suggested in a pension case that corpus linguistics should become part of the judge’s tool belt. And in his concurrence in Facebook v. Duguid (2021), Justice Alito put in a plug for corpus linguistics as well. Alito challenged the judicial reliance on canons of interpretation because they are guidelines rather than categorical linguistic rules, adding, “Perhaps someday it will be possible to evaluate these canons by conducting what is called a corpus linguistics analysis, that is, an analysis of how particular combinations of words are used in a vast database of English prose.”
Most recently, a Ninth Circuit panel asked both sides in Jones v. Becerra (now Jones v. Bonta), a challenge to a California gun law, if corpus linguistics could determine the original public meaning of three phrases in the Second Amendment: “a well regulated militia,” “the right of the people,” and “shall not be infringed.” Responding to the court’s question, appellants argued that corpus linguistics is a flawed tool for legal interpretation, proving their point with a flawed corpus analysis of their own. The appellees took a more positive view of the emerging field of corpus linguistics, yet their own database searches revealed that corpus linguistics was of “limited utility” in answering the court’s questions. The appellees added an important caveat. Corpus analysis, while a useful tool to guide legal interpretation, is not something easily done by an untrained judge or lawyer, and is best left to experts. Indeed, corpus data has become an important tool for professional lexicographers at all the major dictionaries.
Corpus linguistics didn’t come up during oral arguments in Jones, and it remains to be seen whether the Ninth Circuit will incorporate the new methodology in its deliberations. But corpus data has already provided important information about one of the Second Amendment’s most problematic phrases, “to keep and bear arms.”
In his opinion in Heller, Justice Scalia argued that the plain meaning of bear arms is simply, ‘carry a weapon’:
Although [bear arms] implies that the carrying of the weapon is for the purpose of “offensive or defensive action,” it in no way connotes participation in a structured military organization. From our review of founding-era sources, we conclude that this natural meaning was also the meaning that “bear arms” had in the 18th century. In numerous instances, “bear arms” was unambiguously used to refer to the carrying of weapons outside of an organized militia.
That interpretation contradicts the long-held understanding that bear arms has always been a military term, as we see when Judge Nathan Green wrote in Aymette v. State (1840), an early concealed-carry case in Tennessee,
A man in the pursuit of deer, elk and buffaloes, might carry his rifle every day for forty years, and yet it would never be said of him, that he had borne arms; much less could it be said that a private citizen bears arms because he has a dirk or pistol concealed under his clothes, or a spear in a cane.
Green was not alone in that opinion. A 2007 search by historian Saul Cornell found over 100 examples of bear arms in founding-era texts, with ninety-six percent having a military context, corpus evidence that the Heller majority shrugged off.
The following exchange during oral arguments in Heller also demonstrates that bear arms is a military idiom, not a synonym for ‘carry a gun.’ Solicitor General Paul Clement claimed that bear arms means “to carry them outside the home.” Justice Souter asked him, “But wait a minute. You’re not saying that if somebody goes hunting deer he is bearing arms, or are you?” Clement replied, “I would say that and so would Madison and so would Jefferson.” But Souter wasn’t convinced: “In the eighteenth century, someone going out to hunt a deer would have thought of themselves as bearing arms? I mean, is that the way they talk?” Clement finally conceded, no, that is not the way they talk: “I will grant you this, that ‘bear arms’ in its unmodified form is most naturally understood to have a military context.” Souter didn’t need to point out that bear arms appears in its unmodified form in the Second Amendment.
Two large digitized corpora, COFEA, the Corpus of Founding Era American English, and COEME, the Corpus of Early Modern English, were not available to the Heller Court. But searching them, I found that almost every one of some 900 distinct occurrences of bear arms before and during the founding era refers to war, soldiering, or some other form of coordinated armed action by a group rather than an individual. Only seven hits for bear arms were either ambiguous or carried no military connotation. Other researchers have found similar results using different search parameters. To be sure, frequency is not the sole determinant of meaning, but the overwhelming frequency of the military sense of bear arms cannot be ignored when we interpret the Second Amendment.
James Madison undoubtedly thought of bear arms as military when he used the idiom both in the familiar main clause of the Second Amendment and in a second clause exempting “those scrupulous of bearing arms” from militia service. That conscientious objector clause appeared in the first four drafts of the amendment. It was dropped from the fifth version, but its presence reinforces the military nature of bear arms. Judge Green thought bear arms was military in 1840 when he scoffed that hunters and malefactors don’t bear arms. And bear arms was still a military idiom in 1995, when the historian Garry Wills wryly observed, “One does not bear arms against a rabbit.”
Keep arms, rarer than bear arms, occurs only thirty-eight times in COFEA and COEME. Twenty-five of the twenty-six relevant occurrences of keep arms refer to weapons for use in the military or the militia, and one is ambiguous. The military sense of keep arms reinforces the military connotation of bear arms in the Second Amendment.
Searching the full phrase keep and bear arms in COFEA and COEME shows that it only appears when writers quote or allude to the Second Amendment. But that doesn’t tell us what keep and bear arms means. The two sides in Heller offered competing explanations. In his dissent, Justice Stevens argued that keep and bear are practically synonymous and refer to a single right. But Justice Scalia countered that adding keep to bear arms destroys any military sense that bear arms may have had, since keep means ‘own’ or ‘possess.’ Neither argument is satisfying. It’s true that keep and bear differ in meaning. But both keep arms and bear arms evoke military contexts, and it’s reasonable to conclude that combining them into the phrase keep and bear arms does the same.
The phrase keep and bear arms does not occur outside a Second Amendment context in the databases, so a corpus search of the phrase cannot resolve the matter. But given the other military references in the amendment, together with its revisions and the Congressional discussion surrounding its adoption, it’s fair to consider the amendment’s focus to be military. In addition, a search of the corpora for the conjoint verbs keep and bear without specifying an object yields just 20 hits, all followed by arms, confirming that, in the founding era, one does not keep and bear other things, only arms.
Appellants in Jones v. Bonta also argued that a corpus search is fatally flawed because the databases “favor elite usage over common usage.” They suggest that something akin to cancel culture is at work in the corpora: “The vast majority of planters, tradesmen, and frontiersmen who used American English—to say nothing of enslaved peoples, indentured servants, or Native Americans—did not write and publish any books, pamphlets, or broadsides.” Like Justice Scalia, the appellants want us to believe that the common people in the eighteenth century would have understood that bear arms does not entail a military context. They also want us to believe that the elite American writers in the founding era were preoccupied with the recent revolution, and so any references to bear arms in their writing reflect that military history. For everybody else, bear arms simply meant ‘tote a gun’—though these “ordinary” Americans had also lived through and borne arms in the same revolution. But a search of COEME, a database which stretches from 1475 – 1800, confirms that nonmilitary uses of bear arms are almost nonexistent in any sort of text, whether American or British, military, political, historical, religious, literary, philosophical, or belletristic. With almost no exceptions, bear arms evokes a military image.
It’s true that ordinary people didn’t write as much as the framers. But there’s no proof that ordinary people in the federal period said they were bearing arms when they hunted deer, elk, buffaloes, or rabbits. Nor is there any evidence that elite writers like Madison and the members of Congress who carefully edited and revised the Second Amendment baked a non-elite, non-military sense of bear arms into the amendment as a concession to an unattested “ordinary” usage.
And here’s something else to ponder. Everytown for Gun Safety filed an amicus brief in New York State Rifle v. Beach (now New York State Rifle v. Corlett as it heads to the Supreme Court), which traces weapons laws from the English Statute of Northampton in 1328 through colonial, federal, and territorial laws, as well as nineteenth and early twentieth century state laws. These statutes show that England and America were never places where people carried weapons freely and routinely, particularly in urban areas. A quick read of these criminal statutes, some prohibiting open carry and others, concealed carry, shows that the verbs associated with the possession of banned weapons, whether long guns, pistols, swords, swords in canes, knives, Arkansas toothpicks, or metallic knuckles, are go (about) armed, ride armed, carry arms, and have arms. In none of these statutes are offenders said to bear arms.
That brings us back to the question of original public meaning. To repeat: it’s not clear that any text has one single, original meaning that everyone would have shared. Not in the past. Not today. As Akhil Reed Amar, Jonathan Gienapp, and others have shown, immediately after the Constitution was ratified, the framers who drafted, debated, and carefully revised its clauses, openly disagreed about the meaning of what they had just approved. And the split in the Heller Court offers a compelling modern example of dueling public meanings: nine highly-educated jurists, who spent their entire careers dissecting legal meaning, arrived at two competing interpretations of the twenty-seven word sentence that is the Second Amendment.
To make things worse, as Posner and others have observed, the Heller majority, committed to the text-driven originalism, boldly read into the Second Amendment ideas that it did not explicitly contain. Justice Scalia insisted that the framers “undoubtedly” were thinking of self-defense when they drafted the amendment, though self-defense does not appear in the amendment’s text or in any other part of the Constitution, its emanations, or its penumbras. On the other hand, the militia appears explicitly both in Article I and in the Second Amendment.
Corpus linguistics is an important tool, and it can direct us toward a clearer understanding of the right to keep and bear arms. But it’s not a magic bullet. Today’s Supreme Court majority may cling to the myth that bear arms has nothing to do with soldiering. After all, to paraphrase the notorious NRA slogan, words don’t make meanings, people do. But to claim that Heller reads the Second Amendment the way it was understood in 1791 is as fatuous as insisting that “planters, tradesmen, and frontiersmen,” along with enslaved and indentured people and Indians in the early republic, spoke a non-elite version of English where soldiers, hunters, and criminals routinely bore arms. That’s just not how they talked. Such a claim ignores, not just the many statutes in England and America regulating who may have weapons and for what purpose, but also those statutes disarming slaves to prevent revolts, and the fact that the southern states saw militias as essential for controlling enslaved people. Yes, the corpus data helps us interpret the Second Amendment, but we shouldn’t need evidence from a corpus or the imagined utterances of an eighteenth-century proletariat to know that soldiers bear arms, hunters carry guns, and villains pack heat.[*]
[*] To better illustrate how they talked, the 18th- and 19th-century equivalents of modern expressions for going armed, like heavy, packing, strapped, and tooled up, include boden, burled, fixed, and geared.