Spatial Frames

Final Project, by Zach Tomaszewski

for LING 640G, Fall 2002, taught by Dr. Ben Bergen


A number of conceptual metaphors are based on space. Many of these are very basic: HAPPY IS UP, LONG PURPOSEFUL EVENTS ARE JOURNEYS, and the Location dual of the Event Structure metaphor, in which time and events are understood in spatial terms. Motion, which is distance traveled over time (both of which are understood in terms of space), is also part of a number of metaphors and frames, such as the source-path-goal frame. Obviously, our conception of space has a broad and profound impact on our understanding of many different realms. Yet what is this notion of "space" that is being mapped? What are the elements that structure so much of our thought?

It seems that space is a frame. A frame is a system of concepts in which you have to understand the whole to understand one of the concepts (Bergen, Sept 2002). Space invokes many vague notions--position, distance, direction, extension, substance, void, etc. Webster's states that it is "distance extending in all directions; the continuous expanse extending in all directions... within which all things exist; the distance, expanse, or area between, over, within, etc. things" (Guralnik 1980).

To look closer at this frame, I chose to analyze spatial prepositions. This is a closed class of linguistic forms, meaning that it very rarely receives new members. (Regier 1996, p.19). Spatial prepositions have been used by other investigators (Manning, Sera, & Pick 2002; Regier 1996; Lakoff 1987) as a way to study spatial conceptions.

Categorization of Spatial Prepositions

From a number of websites (;;, I compiled a long list of over 130 English "prepositions" (some of them could only barely be considered prepositions). The following 42 of them were primarily spatial in nature:

across from
ahead of
away from
close to
forward of
near (to)
on top of/atop
out of
outside (of)

Each of these small words could be the subject of a paper this length, and indeed many have been. Yet, even with only a superficial investigation, it seems that these spatial prepositions lend themselves to further organization.

Upon reflection, I found that some prepositions were primarily related to motion--such as to and from--while others are primarily used to describe position--such as at or above. Additionally, some are primarily used in relation to points, spaces, and objects.

This categorization exhibits strong prototypical structure. That is, the categories display varying degrees of membership, have fuzzy boundaries, and have central members that best represent each category (Lakoff 1987, p.56). Yet, even if the boundaries are fuzzy, I believe the basic groupings are correct. I tried to restrict myself to literal uses of the words. Their metaphorical and metonymic use--which further blurs the lines of this categorization--is discussed more later.

Here follows the categorization of these 42 spatial prepositions.



Some spatial prepositions seem most suited to describing position relative to a point. By point I mean a definite and precise position, with no size, shape or extension (Guralnik 1980).

At is the most central /prototypical of these. Beyond is the fuzziest member.

Bounded Space

The following propositions apply to some particular space. The space does not need to be surrounded by a solid or objective border, but it does have to have some sense of boundary that defines it. A desert does not have a clear border, though there is still an "in" the desert and an "out" of the desert. Lawns and fields usually have clearer borders, even if they do not have a fence around them and neighbor only another lawn or field.

Qualified bounded space

The qualifier here is that the space must have a surface. Something can be located on or off a horizontal surface, such as a train platform. Or something can be on or off a vertical surface, such as a picture on a wall. (We rarely use off in the position sense; more frequent is off's motion-oriented counter part, off of. See below). There is only a slight difference between a bounded space with a surface and an object, except perhaps that surfaces are primarily two-dimensional while objects are fully three-dimensional.


An object is a material thing, something that can be seen or touched and occupies space (Guralnik 1980). It usually has a clear, three-dimensional shape. Many of these prepositions require such a solid object.

The last half of the list is a little less clear-cut than the first. "I have a sprinkler system installed underneath my lawn." "From Waikiki Beach, Diamond Head can be seen behind Kapiolani Park." It could be said that lawns and parks are more like spaces than objects. While this is true, I also think these sentences are very slightly strange, that they imply a slight object nature to the lawn and the park. I think finding things underneath stones and behind buildings is more typical.

Complex positions

The following require multiple objects or spaces.

Between generally takes two objects. Among takes a collection of objects. Amid is a little hazier than among, since you can be amid-ship, which only takes one object. Across from takes a space (though usually an implied, unstated one) and an object or point.


So far we've looked primarily at static position. Yet about half of spatial prepositions imply motion.


These are the most general motion prepositions, describing motion relative to some point.

Bounded Space

These imply a generic, bounded space.

Qualified bounded space

These imply motion relative to a bounded space with certain characteristics.

Along is like moving through a space by traveling down its longer longitudinal axis. Across means traveling through the space via the shorter latitudinal axis. A good example is the difference between walking along a corridor and walking across it.

The qualifier here is that the space must have/be a surface, usually vertical. You can climb up or down a vertical surface. You can also walk up or down the road.


Onto and off of are frequently used with surfaces. They possibly belong in that category, though, as I've mentioned, the difference between space with a material surface and an object is frequently unclear. Over, under, and around can be used with spaces--over the field, under the lawn, or around the park--but I think they are more frequently used with objects.

This chart summarizes the above categorizations. In general, the prepositions on the left, dealing with points and spaces, can apply to some objects. Less frequently can the surface and object prepositions on the right be used to describe points and spaces.

Graphic depicition of spatial preposition categorization.

Discussion: Is it really that simple?

Probably by this point you've raised a number of your own objections to this categorization.

Only Captures Superficial Meanings

For one, this classification is based on superficial meanings. These prepositions encode much more detail than shown here. For example, on the surface, to and towards are the same--motion in a direction. Yet towards is just "in that direction," while to implies reaching the destination (Bergen, Dec 2002). Similarly, through and across both deal with crossing a space. Yet through signifies a more immersive journey, perhaps slowed down by the medium of the space. Across is simply passing from one side to the other (Bergen, Dec 2002). Within and throughout also describe the position of a certain state of affairs as being in a space: within means it is entirely contained by the borders of the space, while throughout focuses on the fact that it is present through the whole space. Forward of and behind are not absolute. If a the ball is behind the car, it may be that the ball is positioned nearest the rear, trunk side of the car. This is behind from the car point-of-view. Or it may be that the ball is positioned such that the car is between the ball and the observer. This is the observer point-of-view. (Bergen, Dec 2002; Regier 1996 p.29)

Graphical depiction of the two meanings of behind

These differences are important to note in order to understand the meaning of these prepositions. Yet I don't think recognizing these finer distinctions changes their position in the classification. To and towards still deal with motion in the direction of a location. Through, across, within, and throughout are still more relevant to spaces than points or objects. Forward of and behind are still stating the position of one object (ball) in relation to another (car).

Only Relevant to English Prepositions

I have looked here at only English prepositions. Guugu Yimithirr, an Australian language, relies largely on an absolute coordinate system. The Mexican language Mixtec spatially maps things in body-part terms. Thus a stone under a table would be said to be "belly-table". German differentiates between the two senses of English's on: when something is on a horizontal surface verses a vertical surface. Russian has iz-pod, meaning "out-from-underneath" (Regier 1996 p.21-26). I cannot say how much use the divisions of static verses dynamic and point verses space verses object would be in each of these languages. I suspect it would be useful in most cases, but probably not all.

Support of Other Studies

There are many other empirical studies of spatial prepositions. One that I know of tested over 300 pairs of prepositions, asking subject to rank the similarity of the two words. The rankings were then analyzed using the SPSS Scaling MDS module, producing a 4 dimensional solution. The four dimensions found were: Verticality, Containment, Distance, and Contact. Verticality and Containment were the most reliable dimensions, closely followed by Distance. A Front/Back dimension was relevant to some of the experiments. It is claimed that these dimensions are evidence of the organizational framework underlying English prepositions (Manning, Sera, & Pick, 2002).

I have already discussed how many prepositions encode extra information beyond what is captured by the categories posited here. Certainly Verticality and Distance are part of this. Containment is much like the category of Bounded Space. Yet, beyond this one slight similarity, the categories created here do not specifically coincide the findings of Manning, Sera & Pick.

Very Fuzzy Boundaries

The fuzzy, prototypical nature of the classification is also a little worrisome. There is an exceptional use for nearly every preposition, which suggests that it should be classified elsewhere. For example, beside implies something with a side, such as an object. But it could also be used with a surface or even a space. Over is usually over an object, but could also be over a space. When a preposition is centered in one category, but can still apply to nearly any other, the categorization is not terribly useful. Also, determining which is the "core", "central", or "primary" meaning for each of these words is exceedingly difficult. This is largely due to the multiple meanings supplied by context and metonymy.


A number of studies have looked at the effects of context on the meaning of prepositions.

Coventry and Mather have taken another look at over (2002). In a number of experiments, they asked subjects to determine when one object was over a second. Without a context, most subjects chose the graphic with the first object directly above the second. However, within the context of a plane dropping a bomb on a target, a greater number of subjects chose the graphic in which the plane was above and to left of the target.

Jording and Wachsmuth have demonstrated the difficult of communicating a relative position to another person (2002). This is much like the "behind the car" confusion discussed above. In one of their examples, a bowl is sitting on a desk in front of another person, who is facing the speaker. The desk is at an angle to both people. If the speaker says, "Move the bowl to the left," it can be interpreted three ways: to the speaker's left, to the other person's left (speaker's right), or left along the desk. These findings further illustrate how direction and position are relative to the context.

This categorization effort does not consider these effects of context, which may or may not influence certain prepositions' classification.

Metonymy, Metaphor, Construal: The Problem with Objects

Probably the biggest difficulty of classifying prepositions is that each has so many uses. For example, at seems, at first, to be the perfect central member of those prepositions that designate position at a point. Yet at can imply motion, as in "I threw the ball at John." Not only does this imply motion, but the prepositional object of at is John, who is more object than point. We could salvage things a little by saying that the object John is being used to metonymically refer to the point at which John is located. This initially seems suspect.

But we should not ignore metonymy in prepositional objects so quickly. Metonymy is exceedingly prevalent, especially in describing space. A good example is that of the word corner (Tomaszewski 2002). Corner exhibits a radial category of different meanings:

literal point
the point where two sides or edges of a shape or object meet to form a single point. The two sides generally form an internal angle of 90 degrees. However, we usually focus on the outside of the shape (the complementary 270 degree angle).
bounded internal space
In this sense, we are referring only to the internal space, near the corner, bounded by the edges. We are not referring to the point or the edges per se.
around the corner
This sense is used when an object or goal is obscured by a large physical corner. The corner is often formed by the walls of a building, either indoors or outdoors.

Now we can see how prepositions from nearly any spatial preposition category can be used with corner as their object. Literally, a corner is a point. Yet it is also part of an object. It also has a space bounded by the internal 90 degree of the corner, and another area around the external angle. Confusion results when we think a corner is an example of a point, but are actually metonymically referring to a space or object.

Corner has a number of metaphorical uses, such as "He's always in my corner when things get rough," "The cat cornered the mouse," and "Good times are surely just around the corner" (Tomaszewski 2002). Like metonymy, metaphor can also confuse the sense of a prepositional object.

Construal, in the form of fictive motion, is also confusing: "The wires run underneath the carpet." This is a case of position, not of motion. The wires don't actually move.

Creation Verses Application

Related to corner's rife metonymy is the distinction between creating the spatial categories and using them. During the creation process, seemingly similar subjects don't use the same prepositions. For example, deserts and oceans are pretty similar--broad, horizontal, possibly inhospitable areas of the Earth's surface. Yet we are in a desert and on an ocean. Unless we're swimming; then we're in an ocean. Yet we're under water as soon as we duck our heads below the surface, and not under the ocean until we (very nearly) reach the rocky bottom of the ocean floor.

This is confusing when classifying prepositions. However, once we have them classified, the revealed structure can be useful in determining the source of this initial confusion. With water, you can be on, in, or under it. This is like a surface or qualified bounded space. Oceans, if anything, are even more object-oriented. You can be on, in, or under one, as well as have a bungalow beside it. Deserts don't have these characteristics. You can't be on a desert. Being under it is rare. Going through a desert is more common than going through an ocean. A desert is more like a simple bounded space, with few object properties.

The graph of the prepositions is especially helpful in this sort of exercise. Looking broadly, one can determine where along the x-axis a certain object best fits.

Prepositions or Their Objects?

This raises the question: are we categorizing prepositions or prepositional objects here? We are categorizing prepositions based on whether they encode motion or position, and whether they are primarily used with points, spaces, or objects as their prepositional objects. The category of the prepositions is determined by the nature of their acceptable objects.


So what have we achieved through this process of categorizing spatial prepositions? We have created a structure that sheds some light on the nature of these prepositions, and what sorts of prepositional objects can be used with each. However, upon deeper examination, it is found that this structure is rife with problems. The categories tell us only a couple aspects of English spatial prepositions. The application of the findings is restricted by the extremely fuzzy boundaries between categories, by the context of any particular application, and by the metonymy of the prepositional objects. Some of these deficiencies could be minimized through further research.

The goal of this study, from which we wandered a bit, was to gain a better understanding the space frame itself. It is now rather unclear whether some prior, relatively unconscious notion of the frame has prompted the categories generated here, or whether the categories have inductively revealed, through prepositions, the real elements of the frame.

Further empirical study would reveal whether these spatial preposition categories are widely shared. This would support the notion that they tell us much about the space frame. Also, there may be a way to study the frame more directly. However, because it is so basic, abstract, and widespread, I imagine any study of it will be plagued by similar ambiguities and contexts.

Here are some of the elements of the space frame revealed by this endeavor:

Works Cited

Bergen, Ben. Class lecture. Linguistics 640G. University of Hawaii-Manoa, Honolulu, HI. 24 Sept 2002.

Bergen, Ben. Conversation with Zach Tomaszewski. Moore Hall 581 (Office), University of Hawaii-Manoa, Honolulu, HI. 12 December 2002.

"Cats Family - Grammar - English - Prepositions." <> Accessed: 4 Dec 2002.

Coventry, Kenny R. and Gayna Mather. "The Real Story of 'Over'?" Spatial Language. Eds. Kenny R. Coventry and Patrick Olivier. Boston: Kluwer Academic Publishers, 2002.

Guralnik, David B., ed. Webster's New World Dictionary of the American Language. Second College Edition. USA: William Collins Publishers, 1980.

Jörding, Tanja, and Ipke Wachsmuth. "An Anthropomorphic Agent for the Use of Spatial Language." Spatial Language. Eds. Kenny R. Coventry and Patrick Olivier. Boston: Kluwer Academic Publishers, 2002.

Lakoff, George. Women, Fire, and Dangerous Things: What Categories Reveal about the Mind. Chicago: University of Chicago Press, 1987.

"A list of the most important prepositions in English." <> Accessed: 4 Dec 2002.

Manning, Christina, Maria D. Sera, & Herbert L. Pick, Jr. "Understanding How We Think about Space." Spatial Language. Eds. Kenny R. Coventry and Patrick Olivier. Boston: Kluwer Academic Publishers, 2002.

"Preposition List." <> Accessed: 4 Dec 2002.

"Prepositions." < Grammar/preposition/Mainpreposition.htm> Accessed: 4 Dec 2002.

Regier, Terry. The Human Semantic Potential: Spatial Language and Constrained Connectionism. Cambridge, MA: MIT Press, 1996.

Tomaszewski, Zach. "Assignment 2: Radial Categories and Frames." <> 04 Oct 2002.