Date: Mon, 26 Apr 2004 19:47:14 +0200 From: Thora Tenbrink Subject: Representing Direction in Language and Space
EDITOR: van der Zee, Emile; Slack, Jon TITLE: Representing Direction in Language and Space PUBLISHER: Oxford University Press YEAR: 2003 ANNOUNCED IN: http://linguistlist.org/issues/14/14-2433.html
Thora Tenbrink, University of Bremen, Germany
OVERVIEW This book contains a collection of thirteen carefully revised and extended papers originating at a workshop on Language and Space at the University of Lincoln in July 2000.
To begin with, the editors present a comprehensive overview and summary of the issues and contents of the book under the heading of "The Representation of Direction in Language and Space". They start by motivating the research issue as such, placing it in the wider research area on representations of space and time. Direction, in this book, comprises orientation (of an entity, i.e., object or person), location in a specific area with regard to another entity, and direction of movement. The introductory chapter sums up various insights gained from the different contributions, such as the range of variety in linguistic directional encoding, spatial distinctions underlying linguistic directional representations (vectors, half-axes, half-lines, or topological distinctions) and the ensuing different models that are presented by the authors throughout the book. Among the broad variety of insightful generalisations proposed in this introductory chapter is a comprehensive account of the notion of reference frame, reconciling a number of (seemingly) contradictory approaches in order to provide a clear basis for the analyses undertaken in the book.
In Chapter 2, Barbara Landau deals with "Axes and Direction in Spatial Language and Spatial Cognition". Building on empirical evidence from previous work, she considers how axes are used in language and cognition (e.g., memory) independent of, but often together with, direction information. Locations of objects are represented by imposing an axial structure on a reference object. Depending on the experimental task, in some cases axial information is sufficient, while in others, direction plays an additional role. In language, direction information is always present when using projective terms (left/right etc.). But directions can be confused while preserving the basic axial structure. In language, distance seems less important than in cognition.
Chapter 3, "Vectors across Spatial Domains: From Place to Size, Orientation, Shape, and Parts" by Joost Zwarts, builds upon earlier work by the author (and others) representing the semantics of place terms (used for locating one object's position relative to another) by means of vectors. Zwarts claims that axes play a role not only in the application of place terms but also of terms of size, orientation, shape, and parts. Thus, the earlier work is extended from the domain of place to other domains. Zwarts shows that vectors can be used in a consistent (or unified) way to represent different kinds of spatial relations. He suggests that this representation in fact mirrors concepts of spatial relations underlying spatial language, but provides no empirical evidence (other than illustrative linguistic examples) for this claim.
John O'Keefe deals with "Vector Grammar, Places, and the Functional Role of the Spatial Prepositions in English" (Chapter 4). He compares spatial prepositions with the firing fields of the spatially coded place cells in the rat hippocampus, updating earlier work with Lynn Nadel (1978) in which the involvement of the hippocampus in the construction of a "Cognitive Map" is proposed. O'Keefe presents a computational model employing direction vectors which is based on his research group's empirical work with rats, and relates it to intuitions on English prepositions which are partly supported by other empirical work.
In chapter 5, "The Unique Vector Constraint: The Impact of Direction Changes on the Linguistic Segmentation of Motion Events", Jürgen Bohnemeyer also employs vectors as primitives for spatial representation. He proposes the interesting idea that languages may be universally constrained in that all direction specifications in a single simple clause must denote the same direction vector. Thus, "Sally walked north away from her house" implies that "north" is the same direction as "away from her house", while "Sally walked away from her house and the north" indicates two distinct directions. This hypothesized language universal concerns neither language nor cognition per se, but rather, the coding possibilities at the syntax-semantics interface.
Laura Carlson, Terry Regier, and Eric Covey (Chapter 6), in "Defining Spatial Relations: Reconciling Axis and Vector Representations", address reference frames and spatial templates as underlying representations for mapping spatial terms onto spatial regions. They show that the Attentional Vector Sum model (Regier and Carlson, 2001) is suitable for accounting for various factors influencing spatial mappings (orientation, location with respect to the shape of the reference object, distance). The authors propose that axial structures underlying reference frames can be coordinated with vector representations in a spatial template. This approach captures both the directions indicated by projective terms (spatial relations) in terms of axes in a reference system, and the influence of various geometric and functional factors on the alignment of reference systems and the applicability of the relations.
In Chapter 7, "Places: Points, Planes, Paths, and Portions", Barbara Tversky sums up research dealing with information conveyance in directional descriptions. She provides evidence that, while people often refer to landmarks, explicit specifications of direction and distance are dispreferred and usually remain approximate, both in production and comprehension. Landmarks provide an easy and straightforward way of communicating beginning and end points of a path based on the interactants' shared knowledge about the spatial environment.
Chapter 8 by Pierre Gambarotto and Philippe Muller, "Ontological Problems for the Semantics of Spatial Expressions in Natural Language", again presents a model for the representation of direction, this time however not based on vectors but on regions of space as ontological primitives. The motivation for this approach is the insight that, from a cognitive point of view, topology is the most basic aspect of spatial knowledge. In their model, problems associated with using points, lines or intervals as primitives are avoided. Mereotopological relations can be represented in a straightforward way. The authors also show how the representation of orientation, distance, and motion works in their approach, and how functional information associated with the semantics of prepositions is integrated.
Hedda Schmidtke, Ladina Tschander, Carola Eschenbach, and Christopher Habel (Chapter 9) deal with "Change of Orientation". Their axiomatic approach uses half-lines in a geometric model of direction that is neutral with respect to distance. Using this approach, they propose an analysis for the semantics of verbs of change of orientation such as "abbiegen" (turn off), comparing them to related verbs of motion and verbs of orientation and determining patterns of geometric constraints on the applicability of the terms.
In Chapter 10, Urpo Nikanne describes "How Finnish Postpositions See the Axis System". Finnish makes more distinctions in the application of spatial terms than English does: thus, static and dynamic situations are differentiated grammatically. Furthermore, Nikanne finds that, in Finnish, vertical terms (above/below) can indeed only be used for vertical relationships, while expressions like in front of / behind (which basically express horizontal relationships) can also be required in certain vertical (or other) situations, namely, when motion into a vertical (or other) direction is involved. Motion induces its own reference frame (or rather, a front/back orientation) that is independent of horizontal or vertical positions (relative to the earth); the spatial terms that are used in such a situation are those that also express relative position on a horizontal plane. From this fact, Nikanne deduces a feature hierarchy for reference frames.
Emile van der Zee and Rik Eshuis (Ch. 11) address "Directions from Shape: How Spatial Features Determine Reference Axis Categorization". Based on three experiments on Dutch directional nouns employed in intrinsic reference systems, they present the "Spatial Feature Categorization model" which accounts for spatial features in the application of reference axes with different kinds of objects. The objects in the experiments are varied with respect to relative axis length, relative curvature of the main plane of symmetry, and contour expansion along a reference object's main axis.
In Chapter 12, "Memory for Locations Relative to Objects: Axes and the Categorization of Regions", Rik Eshuis presents empirical evidence suggesting that axes are used to carve up space into regions that are not - as previously suggested - quadrants, but rather, half-planes. Thus, an object that is located to the left, but not exactly on the left half-axis of a reference object, is also located on either the front or the back half-plane with regard to that reference object. In other words, space is not divided into four mutually exclusive regions, but into overlapping half-planes, resulting in the fact that different spatial relations (e.g., linguistic expressions) may be suitable for representation (though not equally likely). This interesting and intuitively appealing idea is developed in relation to previous findings on spatial templates and prototype effects of spatial expressions. Eshuis' account is based on his own non-linguistic experiments and is entirely compatible with previous evidence from linguistic tasks.
Finally, Kenny Coventry (Ch.13) addresses "Spatial prepositions, Spatial Templates, and 'Semantic' vs. 'Pragmatic' Visual Representations". He discusses previous work showing that the employment of spatial prepositions depends - apart from geometric factors - on a variety of associated factors that differ across terms, such as 'control' or 'support'. These two kinds of constraints are related to findings from visual imagery, where 'semantic' representations account for general visual information while 'pragmatic' representations concern the objects' affordances.
CRITICAL EVALUATION The book is carefully edited; readability is enhanced by the clear and concise layout as well as the availability of abstracts for each chapter. However, the usage of one comprehensive list of literature references at the end of the book rather than separate lists at the end of each chapter is a disadvantage for readers of single chapters who may not have access to the book as a whole.
Altogether, the book offers a wide range of useful and interesting insights into the research area of spatial language and spatial cognition. In spite of the fact that the title seems to suggest a focus on only a specific sub-part of spatial representation, many articles in the book offer a fairly broad view on related research issues. Several formal models for spatial representation (with a focus on direction) are presented that differ in scope and focus but are not incompatible. Furthermore, the book comprises several review articles on empirical work that together provide valuable insights into the field.
The title of the book contains an expression that has a potential for polysemic interpretation that could have been more directly addressed and explored. As indicated above, "direction" can be understood to denote (at least) orientation, relative location, and movement, opening up a wide range of different interpretations that surely carry various important distinct implications with regard to the spatial relationships involved. However, in the book the different senses are not always clearly differentiated (or else it is not clear how they are viewed to be related); therefore, the scope of the claims and insights is not in all cases clear, and important aspects might have been lost. For example, static and dynamic scenarios may well differ in the application of spatial relations. Distances are implicitly present in all static situations involving relative location, since objects in the real world are located at a specific distance from each other even though this may not always be specified in a representation. In contrast, movement in a certain direction may theoretically take place indefinitely, even though in practice usually a specific (though often unspecified) distance is covered. Thus, distance could play a different role in the representation of spatial relations in static versus dynamic situations.
The articles comprise a variety of different languages. As is often the case in this research field, English is predominantly used as the starting point for analysis, especially in approaches dealing with the interplay of language and cognition or with formalizations (Landau, Zwarts, O'Keefe, Carlson et al., Tversky, Coventry). However, several articles deal with the semantics of spatial expressions in other languages (Gambarotto & Muller: French; Schmidtke et al.: German; Nikanne: Finnish; van der Zee & Eshuis: Dutch; Eshuis: German, Bohnemeyer: various languages). Since the relationship between language and spatial cognition is often addressed, it is astonishing that the cross-linguistic achievements of the Space group of the MPIP in Nijmegen do not receive much attention throughout the book (excepting Bohnemeyer who is himself part of a neighboring group of the MPIP, namely, Event Representation). There, linguistic and non-linguistic tasks are systematically compared and the underlying relationships highlighted. A good overview can be found in Levinson (2003).
In spite of the fair number of psychological/psycholinguistic contributions in the book, only few articles (e.g., Eshuis, who presents impressively detailed analyses) describe original empirical work. Several others review earlier findings, in some cases reproducing depictions but (naturally) not providing detailed information regarding experiments described elsewhere (e.g., Landau, O'Keefe, Carlson et al., Tversky, Coventry). Articles of this kind risk that some conclusions cannot easily be comprehended without access to the earlier publications they draw upon, and it is not in all cases clear what exactly is new about the present contribution, as compared to the previous ones that serve as sources. However, they provide good overviews especially for newcomers to the field.With the more formal approaches, a different (and equally widespread) problem is that they sometimes do, and sometimes do not draw on empirical research to motivate the intuitions on spatial language that are formalized. For instance, whether or not a spatial term is considered as applicable in a specific spatial situation very much depends on contextual factors, not all of which can be captured by linguistic intuitions on the part of the authors.
Altogether, the book provides rich and varied insights into the field of spatial language and cognition and is highly recommended to all researchers in the field, including those that aim at integrating findings from psychology and linguistics into interdisciplinary work or other fields in Cognitive Science, and also including interested readers who are yet unfamiliar with current issues in Spatial Cognition.
REFERENCES Levinson, Stephen C. 2003. Space in Language and Cognition. Cambridge University Press.
O'Keefe, J. and Nadel, L. 1978. The hippocampus as a cognitive map. Oxford: Clarendon Press.
Regier, T. and Carlson, L. 2001. Grounding spatial language in perception: An empirical and computational investigation. Journal of Experimental Psychology: General, 130(2), 273-98.
ABOUT THE REVIEWER:
ABOUT THE REVIEWER Thora Tenbrink is a research assistant in the DFG Collaborative Research Center SFB/TR8 "Spatial Cognition: Reasoning, Action, Interaction" (Bremen & Freiburg, Germany). Her dissertation project deals with the question how objects and events are localised relative to other objects and events using spatial and temporal expressions in natural discourse. Previous work has dealt with discourse relations and information structure, presuppositions and non-temporal implications of temporal connectives, especially 'before' and 'after'. Her current focus is on empirical research on spatial reference systems in human- robot interaction.