I agree with Brugman's general comment that the lexical/grammatical level is a primary determinant of the discourse level, and so must be a determinant of the input to the spatial representation system (SRS). Further, I agree with Brugman's argument about the importance of the lexical level in determining the discourse level. Spatial information carried in lexical items, especially spatial prepositions, can directly influence the formation of mental spatial models by the SRS. I examine this hypothesis in terms of the information necessary to build a spatial model in the SRS. Brugman also asks why it was assumed that discourse is represented propositionally by the language system. My response to Brugman's question is to admit that the original assumption is somewhat arbitrary. The basis for the assumption was an effort to address the longstanding debate in psychology over propositional and mental-model representations. If spatial terms can call up image schemas that describe geometric properties of the environment, the schemas can be used to fill in many of the details left out of discourse.
1.1 Brugman (1992) offers a compelling argument why we should attend to the lexical/grammatical level of language when attempting to explain spatial representation. I agree with Brugman's general comment that the lexical/grammatical level is a primary determinant of the discourse level, and so must be a determinant of the input to the spatial representation system (SRS). One of the important points Brugman makes is that the lexical level provides the basic spatial categories needed to construct a discourse representation. Every language has a relatively small set of spatial prepositions (Jackendoff & Landau, 1991) that embody a particular type of spatial relation. There is not much overlap in meaning between prepositions, and so there is little freedom in choosing lexical items to describe a particular layout of objects. And because the categories of spatial relations contained at the lexical level are a small subset of all possible relations, the lexical level constrains how one can describe space.
1.2 It is interesting to note that spatial prepositions generally do not specify a frame of reference (see Retz-Schmidt, 1988). One of the major arguments of the SRS account (Bryant, 1992) was that space is represented in up to three coordinate systems, the allocentric, egocentric, and external reference frames. A linguistic description of an environment must specify a particular frame of reference, which must be used consistently. However, most spatial prepositions can be used in all three frames of reference (exceptions seem to be words like "through" and "inside" that require an object-centered interpretation). For example, I can say "the computer is BEHIND me," and mean that the computer occupies a position to my rear. On the other hand, I can also say "the computer is BEHIND the filing cabinet," and mean that both the computer and filing cabinet are to my front, but the computer is further away. The word BEHIND does not specify whether I am using an egocentric or external frame of reference.
1.3 I entirely agree with Brugman's argument about the importance of the lexical level in determining the discourse level. Lexical items, especially spatial prepositions, carry the meaning necessary to construct propositions that state a spatial relation between two objects, or between an object and a location in a coordinate framework. This means that the lexical level does play an important role in ultimately determining the nature of the input to the SRS. However, Brugman further suggests that the lexical level serves directly as input to the SRS. Thus, spatial information carried in lexical items, especially spatial prepositions, can directly influence the formation of mental spatial models by the SRS. In the next section I will examine this hypothesis in terms of the information necessary to build a spatial model in the SRS.
2.1 The SRS must receive information concerning distance and relative direction to code the location of objects in a coordinate space. Perceptually, an individual estimates distance and angular displacement relative to the self and computes allocentric position. Verbal descriptions can provide information in allocentric as well as egocentric frames of reference. However, in both cases, a location must be related to a position in a frame of reference. Lexical entities on their own, however, don't provide this information.
2.2 Consider the following examples:
(1) The box is NEAR me (2) NEAR me. (3) NEAR.
Number 1 is a grammatical sentence that locates a box with respect to myself. This sentence contains sufficient information to build a simple spatial model in the SRS. It specifies a frame of reference, the egocentric frame, by locating the box with respect to me. It specifies an object that will occupy a location in that frame of reference, and it specifies a range of possible locations with the preposition NEAR. Number 2 might also provides sufficient information to form at least the basis of a spatial model. It specifies a frame of reference (egocentric) and a range of locations that an object might occupy, although it does not indicate what the object is. Even though number 2 is not a grammatical sentence, one can imagine a particular space and set of locations within that space. This could be termed a spatial model, albeit an extremely impoverished one. Number 3, on the other hand, cannot serve as the basis for a spatial model. It does not specify a frame of reference, so the word NEAR does not refer to any specific spatial situation. The purpose of the SRS is to create concrete spatial models of particular environments. The word NEAR, although it has meaning in all three frames of reference, does not contain enough information to build a spatial model of a specific environment.
2.3 These examples demonstrate the power as well as the limits of the lexical level. A single sentence or even a single nongrammatical sentence fragment can serve as input to the SRS and can begin a mental spatial model. Thus, if we think of discourse level as a representation only of connected sentences, my original assumption (Bryant, 1992) is incorrect. A spatial model in the sense defined by the SRS can be invoked by a single sentence. Moreover, complete grammatical sentences don't seem to be necessary to specify the two most important pieces of information for the SRS - the frame of reference and location in the coordinate space. It is clear from the examples that Brugman makes a valid point; it is too simplistic to talk about the discourse level being the only one that contacts the SRS. Example 2 above suggests that a small construction that is clearly not discourse can form the basis of a mental spatial model. This demonstrates the importance of the lexical level, and of spatial prepositions in particular, because most of the meaning of the phrase, "NEAR ME," is derived from the category embodied in the preposition NEAR. However, to form the basis of a spatial model, the preposition requires one other item (in this case, ME) to indicate the frame of reference and make the phrase refer to a specific spatial situation that can be realized in a mental model.
3.1 Brugman asks why it was assumed that discourse is represented propositionally by the language system. She cites a number of theorists who advance nonpropositional representation of discourse (e.g., Lakoff, 1987; Langacker, 1987). Other theorists, however, propose that discourse is analyzed into a propositional record (e.g., van Dijk & Kintsch, 1983), and the issue has not been settled. My response to Brugman's question is to admit that the original assumption is somewhat arbitrary - none of the studies reviewed in Bryant (1992) provides conclusive evidence that discourse must be represented in a propositional format. The basis for the assumption was an effort to address the longstanding debate in psychology over propositional and mental-model representations. I would also note that reflecting on Brugman's comments on this matter has made the merits of nonpropositional representation clearer. Schematic theories such as Lakoff's (1987) appear to offer a useful way of dealing with the problem of underspecification of spatial information in verbal descriptions. If spatial terms can call up image schemas that describe geometric properties of the environment, the schemas can be used to fill in many of the details left out of discourse.
Brugman, C. (1992). Spatial Cognition: The Perspective from Theoretical Semantics. PSYCOLOQUY 3(45) space.6.
Bryant, D. J. (1992). A Spatial Representation System in Humans. PSYCOLOQUY 3(16) space.1.
Jackendoff, R., & Landau, B. (1991). Spatial Language and Spatial Cognition. In D. J. Napoli & J. A. Kegl (Eds.), Bridges Between Psychology and Linguistics: A Swarthmore Festschrift for Lila Gleitman. Hillsdale, NJ: Lawrence Erlbaum Associates.
Lakoff, G. (1987). Women, Fire, and Dangerous Things. Chicago: University of Chicago Press.
Langacker, R. (1987). Foundations of Cognitive Grammar, vol. 1. Stanford, CA: Stanford University Press.
Retz-Schmidt, G. (1988). Various Views on Spatial Prepositions. AI Magazine, 9, 95-105.
Van Dijk, T. A., & Kintsch, W. (1983). Strategies of Discourse Comprehension. New York: Academic Press.