The semantics of sleeping in railway stations


“I really, really like the work in Congress, I really do, but I love my family more. People may try to make it more than that, but it’s really that simple,” [Jason] Chaffetz said on MSNBC. “I just turned 50. I’m sleeping on a cot in my office.”

Chaffetz on No 2018 Run: ‘I Just Turned 50, I’m Sleeping on a Cot in My Office,’ Talking Points Memo xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxx

Everyone familiar with the academic literature on statutory interpretation is aware of the no-vehicles-in-the-park hypothetical. It was formulated by the legal philosopher H.L.A. Hart to illustrate the argument that the words in which a law is written must have “a core of settled meaning”—a set of “standard instance[s] in which no doubts are felt about [the law’s] application”—but will also have “a penumbra of debatable cases in which words are neither obviously applicable nor obviously ruled out.”  But Harvard law professor Lon Fuller denied the existence of any core area in which the law’s applicability was clear; for Fuller, the law’s applicability turned not on linguistic semantics but on the law’s purpose. Thus, he asked whether, under the hypothesized prohibition against vehicles in the park, “mount[ing] on a pedestal in the park a truck used in World War II…in perfect working order” would fall within the law’s core or its periphery.

Less well known than Hart’s thought experiment is a separate hypothetical offered by Fuller to support his challenge to  Hart. Fuller posits a law making it a misdemeanor “to sleep in any railway station.” He then supposes that two people have been arrested for violating this law: one who dozed off while waiting for a train, and another “who had brought a blanket and pillow to the station and had obviously settled himself down for the night[,]” but who had been arrested before he fell asleep. “Which of these cases,” Fuller asked, “presents the ‘standard instance’ of the word ‘sleep’?” And would it be faithful to the law to say that the law had been violated by the second person but not the first?

The hypothetical is thought-provoking because applying what is assumed to be the literal meaning of the law—that it prohibits being asleep in a railway station—would yield a conclusion that seems nonsensical: that the law was violated by the dozing passenger but not by the person who was bedded down but still awake. The hypothetical has been discussed by some very smart legal scholars and philosophers over the years, including Kent Greenawalt, Fred Schauer, John Manning, Scott Soames, and Andrei Marmor, and with few exceptions (mainly the linguist Robyn Carston) they have accepted that assumption. Schauer put it as well as anyone: “Sleep is a physiological state, and as a matter of physiology Fuller’s businessman was sleeping. Period.”

But in fact (you can guess where this is going, can’t you?), the assumption’s validity is doubtful at best. It is entirely consistent with actual usage to use sleep in a railway station to mean ‘use a railway station as a place to sleep’ rather than ‘be asleep in a railway station.’

My path to this conclusion began with thinking about the expression sleep over at [someone’s] house. That expression doesn’t mean simply to be asleep at someone’s else’s house. In fact, if you ever went on a sleepover as a kid, or hosted one for your kids, you know that sleeping over at someone’s house involves a lot more than sleeping. (In fact, it may not involve much sleeping at all.) What sleeping over is about is spending the night—precisely what Fuller’s hypothetical homeless person wants to use the railway station for.

“But that’s an idiom,” you say, “just like sleeping with [someone] is an idiom that means having sex with them. That’s not a basis for generalizing about what sleep can mean in a different context.” And you’re right. That point occurred to me. But then I thought about what you might say when you have out-of-town guests staying with you and you want to give them their room assignments. You might say something like, You’ll sleep in the guest room, by which you mean ‘You can use the guest room as the place where you will sleep (and keep your stuff, and get dressed, and hang out when you want some privacy, and…).’ Unlike sleep over at [someone’s house], that doesn’t strike me as being clearly idiomatic.

And that got me wondering whether maybe the ‘use as a place to sleep’ sense of sleep is commonly associated with constructions consisting of sleep followed by a locative prepositional phrase (meaning a prepositional phrase that denotes a location; PPloc for short). If so, then the question to ask about Fuller’s ordinance is not “What does sleep mean” but “What does sleep in any railway station mean,” and the answer to the first question doesn’t necessarily determine the answer to the second.

This way of looking at the issue is not something I thought up on my own (though I’d love to take credit for it). Rather, it is based on insights that have come out of corpus linguistics and corpus-based lexicography. I’ve written about those insights in my paper A Lawyer’s Introduction to Meaning in the Framework of Corpus Linguistics, which you should read if this interests you. The basic idea is that it’s wrong to think about the basic unit of meaning as being the word. Considered in isolation, words can have a wide range of potential meanings, but as used in sentences, in most cases only one of those potential meanings is viable. And the relationship between the word’s meaning in context and the context itself is not random. Rather, it is often systematic to a greater or lesser degree. Particular meanings are often associated with particular grammatical environments. In other words, a word’s meaning when it is used in a given construction arises not from the word itself, but from the use of the word in that construction.

Which brings us back to the question whether the construction sleep+PPloc is commonly used to mean ‘use [location] as a place to sleep.’ To try to find out, I ran corpus searches seeking corpus lines that included the expressions sleep in, sleeps in, slept in, or sleeping in. I ran the searches in the Corpus of Contemporary American English (COCA), which includes texts from the period 1990–2015, and in the Corpus of Historical American English (COHA), with the search limited to the 1950s and 1960s. (For an overview of these corpora, go here.) The COHA search was done because Fuller’s article was published in 1958, and I thought it made sense to check roughly contemporaneous usage, in case there had been a relevant change.

The search results supported the conclusion that sleep in [location] is (and was) in fact commonly used in the ‘use as a place to sleep’ sense. In fact, that was the dominant sense, appearing in the concordance lines that I reviewed (200 lines in COHA and 100 in COCA) between 3.5 and 4.5 times as often as the ‘be asleep in [location]’ sense.

For some of the concordance lines, I thought it was clear which of the two senses was conveyed. The lines below, for example, seem to me to evoke the ‘use as a place to sleep’ sense; they are concerned more with living conditions and lodging than with being in a state of sleep. In some of them, in fact, the ‘be asleep in [location]’ sense would be implausible:

‘use as a place to sleep’ (from COHA)

My son, who is now in medical school, slept in a Rinso carton.” After a few more chuckles, Dr. Adair continued

John Ashley was laying plans without being aware of it. He spent eight days sleeping in the woods. Each evening he awoke with a project formed in his mind

When I was sixteen I left home. For weeks I slept in parked cars. Then I started “borrowing” cars, taking rides in

…making any big purses yet, I was living good for a boy who’d slept in cars not too long ago. I bought sport jackets and fancy slacks.

Lodging was of less importance. In summer a man can sleep in parks and under bridges….

Seaside towns are packed all the way around the island. People are sleeping in bathtubs and autos, on beaches, and on moors….

The lines below, in contrast, are very much about being asleep.

‘be asleep in [location]’ (COHA)

or seven hours of sleep daily after the first few crucial days. When they slept in orbit, their heartbeats dropped to the high 30s….

Jamesie, leaned against his mother’s knee. The baby, Theresa, slept in her mother’s arms….

…she carried with her as she crept quietly into the room where Butch still slept in his crib. With a safety pin from the dresser she affixed it securely

The flight was slightly rough, and Nell slept in the reclining seat, lulled by the steady powerful roar of the motors,

For some of the other lines, the appropriate classification was perhaps a little less obvious, but nevertheless I’m reasonably confident in my judgments:

‘use as a place to sleep’ (COHA)

The relationship between the two is suggested by the fact that Moyers slept in a large closet in the Johnson hotel suite during the 1960 Los Angeles convention

…in boot camp; in the bed, chaste and narrow, that he had slept in as a boy. And in a way, he was still, at

of Senator John. He eventually moved into the house with the old man, slept in the trundle bed, and padded in and out with slop jars, water

…the general would not sleep in comfort while his men were still sleeping in the open. We have his specifications for the huts that were to be

” There’s the operating control for your bed; I’ll sleep in back, where I did last night. First thing tomorrow we’ll get

Mr. Scaddon and I slept in separate rooms whenever this was possible. We always slept in separate beds. ” …

And in some instances, I couldn’t decide either way, either because aspects of both senses seemed to be present, or the context didn’t provide enough information, or the line was otherwise ambiguous:

uncertain (COHA)

…was bedded down next door in a white room with white woodwork. He slept in the same white wicker bassinette that was used by his mother and his sister

he lay among the whole unfamiliar Moore clan, both women and men, all sleeping in the unpartitioned half-story attic they called “upstairs. ” Walls had been left

skirtings raised, the furniture pillaged and ripped and smashed — no one could have slept in the house except on a bed of broken plaster. T. gave his orders

…Finding a place to sleep in the lunar module may be the’ toughest problem. The module wasn’t

…All told, 17 persons? were sleeping in the Tate apartment at the time of the firei-

Some thoughts about these results:

First, and most obviously, the near-unanimous conventional wisdom about the literal meaning of Fuller’s hypothetical statute is pretty clearly mistaken. Even if you ultimately disagree with me on that score, it’s hard to deny that there is a strong argument the other way. And what’s more important, the fact that the boat has so frequently been missed is a reflection of the impoverished way in which most people think about issues of word meaning.

Second, this is a cautionary tale about the use of dictionaries. The ‘use as a place to sleep’ sense doesn’t appear in any of the 12 dictionaries that I checked ((American Heritage, Cambridge, Collins, Longman, MacMillan, Merriam Webster’s 3rd (unabridged), Merriam Webster’s Collegiate, Merriam Webster’s Advanced Learner’s Dictionary, Oxford Advanced Learner’s Dictionary, Oxford Dictionary of English, Oxford English Dictionary, and Random House). That ought to give pause to anyone who thinks that dictionaries can answer all questions about word meaning.

Third, corpus linguistics is an important tool for legal interpretation—but there’s a sense in which this is a cautionary tale not only about dictionaries but also about corpus linguistics. Just knowing how to do a corpus search is not enough; that knowledge has to be informed by the insights about word meaning that corpus linguistics and corpus lexicography have generated over the past 30 years. (See my paper on this issue.) If I hadn’t been familiar with that literature, and had done only a simple corpus search for the verb forms of sleep, without having formulated a hypothesis about the meaning of sleep+PPloc, I might not have recognized the pattern seen in the concordance lines I’ve quoted. It certainly wouldn’t have jumped out at me. In both COCA and COHA (1950s–1960s), tokens of the construction sleep in account for less than ten percent of the tokens of sleep, and not all of them were relevant. But because I’m familiar with the literature of corpus lexicography—particularly Patrick Hanks’s work on Corpus Pattern Analysis—I was able to ask the right question and formulate the search query that produced the answer.

The moral of the story is that while judges and lawyers should be encouraged to learn about and use corpus linguistics, that should ideally happen in the context of learning more generally how to think like a linguist.

Leave a Reply