March 24, 2005

Spatial gender

Last year, I heard an interesting talk by Len Talmy entitled " How Spoken and Signed Language Structure Space Differently". He started out by talking about cross-linguistic properties of closed-class forms (like prepositions or verbal affixes) that specify spatial structure, and what he said got me thinking about how complicated the interaction of prepositions and their head nouns can be, even in one language. Yesterday I cleaned out my briefcase and found the handout from his talk, on which I'd jotted down some notes about in and on in English, which I've expanded into the rest of this post.

For anyone who knows what the words mean, it's not surprising that you put something in the oven but on the stove.

However, it's somewhat less clear why you put something in the picture but on the screen, why the people who are in your heart are often on your mind, or why a file is in a folder but on the desktop. There's an obvious story about the relationship between preposition choice and the shades of meaning in such expressions, but it's not obvious whether a basic feature of the expression's meaning causes a certain preposition to be used, or alternatively an arbitrary choice of preposition in the expression generates an difference of interpretation. And when we get to the difference between being in town and being on campus, or for that matter the difference between being in time and being on time, we're pretty clearly in the realm of idiomatic phrasal patterns.

We get in a car, but most forms of transport are things we get on: a train, a ship, a plane, a bicycle. Is that because we basically think of trains, ships and planes as platforms that carry us, even though we generally ride inside them, while we view cars as spaces that we inhabit? Or is it because the passenger-carrying parts of trains, ships and planes are generally high up? Does the relatively small size of cars matter? In any case, the choice between in and on is not simply an arbitrary property of words for vehicles -- it depends on the meaning of the words. If you tell someone to get on the jeep, the van or the Buick, you mean for them to climb on the hood or the roof or to do something else that treats the vehicle as a platform. However, if you tell someone to get on the downtown local, or next flight to Tibet, or the Queen Elizabeth, or the 8:17, you just mean for them to get into the vehicle's passenger-carrying space in the usual way.

In this web example of getting on a car, car turns out to mean streetcar:

To get on the car, take hold of the front railing with the hand toward the front of the car, raise the corresponding foot to the step, and you are safe.

There are other examples where the distinction between in and on is predictable according to semantic class, even if assignment of preposition to class seems to be arbitrary. Thus someone appears in a play (in Macbeth, in Tom Stoppard's latest, in a student production), but on a recurrent show (on Letterman, on Saturday Night Live, on 20/20) or a media outlet (on CBS, on cable, on pirate radio, on the stage, on the runway -- but in the movies). Is a production of a play in some sense a container in which people are placed, while a named show or a communications medium is in some sense a surface on which people are displayed? Perhaps so, but I'm not sure whether this conceptual distinction precedes or follows the linguistically-determined choice of preposition.

Sometimes a class of semantically-similar words seems to split between in and on in lexically-arbitrary ways. An item is in the index, the table of contents, or the specifications; but on the agenda, the menu, the docket or the feature list. Is that because an index or a table of contents is a kind of container, while an agenda or a menu is a surface? You might think that this follows somehow from the idea that an index or table of contents is a fixed collection of items, while an agenda or menu involves different items on different occasions. This might have something to do with the origin of the difference. But stories are listed in the table of contents of a weekly magazine, while dishes are listed on the menu of a restaurant that hasn't changed its selections in years.

The combinations of bare singular nouns with in or on are all clearly idioms, and in some cases there's no semantic basis in current usage for the distinction: the fact that I'm in town but on campus is now as arbitrary as the fact that I can't be either in or on village, city or park (unless "park" is the same for an automatic-transmission state, in which case I can be metonymically in it...) In some of these [P N] idioms, though, the choice of preposition makes sense. Trouble is clearly something you're going to be in, rather than on, while you're on message or on target in the same sense that you can be on the rails or on the path to a destination. It's less obvious why you go on vacation but stand in line. It makes sense that something under control is in hand, but it's less obvious why something is on hand when it's in stock.

These patterns of preposition usage are reminiscent of the distribution of grammatical gender, or the related case of noun-classifier systems with multiple categories. These are partly determined by meaning in a simple way (e.g. by the sex of the referent), partly by meaning in a more complicated way (e.g. by extension to semantic categories that aren't connected directly to sex), and partly by arbitrary lexical assignment.

Lera Boroditsky's work (e.g. her paper on Sex, Syntax and Semantics) suggests that a language's quasi-regular patterns of grammatical gender can affect the way that speakers think about the referents of the words involved. It seems reasonable that the similarly quasi-regular patterns of preposition choice -- which notoriously differ from language to language -- might also affect people's associations with the concepts involved.

Postscript: it's interesting to look at what's involved in getting web-corpus support for the claim that cars take in while other modes of transportation take on. Raw Google counts go in the right direction, but not very strongly:

the car
79% in
a car
86% in
the bus
90% on
a bus
78% on
the plane
52% on
a plane
57% on
the train
82% on
a train
81% on

Looking at the minority examples makes it clear that nearly all of them are not relevant, for structural or semantic reasons:

A US armored vehicle in Iraq fired on the car carrying the freed Italian hostage...
...the numerous cameras positioned on the car allow you to share your driving experience with others...
Refuse to pay for it, even if it's on the car- they only paid $15 for it...

This page contains quick descriptions of the four types of symmetry in the plane.
Since the heliport control tower is strangely invisible in the "plane" photo...
The line in the plane with i = 0 is the real line.

If we add some context, such as "get in X" vs. "get on X", the numbers are much more nearly categorical:

get in
get on
the car
98% in
a car
98% in
the bus
99% on
a bus
98% on
the plane
96% on
a plane
99% on
the train
98% on
a train
98% on

Again, most of the (now few) contrary examples are not relevant:

If no moisture is present, and dust and dirt cannot get on the car, the car will be removed in the same condition as when stored.
This ensures that no resin will get on the car and gives you a surface to work on
That means the chances of seeing another in the valet parking lot outside Saks Fifth Ave. are about as slim as the dealer discounts you'll get on the car.

The officers are sometimes collecting a blue form, which you might get in the plane.
How would Anthrax get in the plane?
One comedian once pointed out that an airport loudspeaker might announce. "Last call for Flight 104. Time for all passengers to get ON the plane" "No thanks", said the comedian, "I'll get IN the plane."

However, especially with airplanes, there are some genuine contrary examples (i.e. with in). Many of these seem to be cases where the plane is a small one -- people seem to treat small aircraft like cars from this prepositional point of view:

(link) Time to get in the plane. You have to get yourself in first, then retrieve your stuff. [about a glider]
(link) Here, you get to fly within the second week of class. You get to get in the plane and get going right away! [about a flight school]
(link) Most people get in the plane and then consider it a victory if they crash into something that doesn't belong to their own team. [about a video game]
(link) As long as you get in the plane (which you will bc it doesn't get scary till youre in the plane) but once youre in the plane you don't have much of a choice! [about skydiving]
(link) Now I can just get in the plane and go fishing anywhere in the Western US, and be there in just a few hours.

Others seem to involve a special perspective on the process:

(link) Are you sure can get in the plane? [about fear of flying]
(link) We thought we would get in the plane first considering the amount of kids we have, but we waited like the rest of the pack.
(link) At the very last moment, one member of the crew came racing out of an automobile to the plane, carrying on all of his spare uniforms and clothing, running as hard as he could to get in the plane.

As usual, it would be very helpful to be able to get a random sample (i.e. not biased by page rank or similar things) of the hits for a given pattern.


[Update: Nicholas Sanders sent in some relevant information about Danish:

If it be assumed (not unreasonably) that (paa) stands for on and i for in, the matter of why one is på posthuset (on the post office) but i banken(in the bank) is a puzzle, and not only for learners.

Perhaps a clue may be found in that one is also på biblioteket (on the library), på hospitalet (on the hospital), and på arbejde (on work - on the job?), but i skolen (in the school) or i butikken (in the shop).

I suggest a connection with the number of the institutions concerned, at least in the past when a human settlement of a certain size would have had just one each of the group, but a few of each of the other category - work might have been included in the former because every adult male would have had a defined job, so that his occupation was likened to the official status of the other examples.

If his analysis of the /i distinction for building-like locations is correct, then Danish has crystallized a locative subregularity that (I believe) is missing in English. What seems to happen in such cases is that a local association spreads, to some greater or lesser degree, along a dimension that is salient in the initial exemplars. The spread of such patterns is limited as they bump up against regions in which other patterns hold, somewhat in the manner of grain boundaries in polycrystals. ]

[Update #2: John Cowan wrote:

New Yorkers (and immigrants like me) "stand on line", as is well known. What seems not to be so well known is the semantic difference between "get in line" (form a line) and "get on line" (enqueue yourself to an existing line). I'm told that this distinction has spread to the Twin Cities as well, supposedly through the medium of New Yorker Catholic teaching nuns.

"Get on line" is under some pressure from its homonym "get online" (connect to the Internet), though.

I've never lived in New York, nor been taught by nuns, but I have the same "get in line"/"get on line" distinction. ]


Posted by Mark Liberman at March 24, 2005 08:24 AM