Space, Time, and the Use of Language: An Investigation of Relationships 9783110198829, 9783110195200

Citation preview

Space, Time, and the Use of Language

≥

Cognitive Linguistics Research 36

Editors Dirk Geeraerts Rene´ Dirven John R. Taylor Honorary editor Ronald W. Langacker

Mouton de Gruyter Berlin · New York

Space, Time, and the Use of Language An Investigation of Relationships by Thora Tenbrink

Mouton de Gruyter Berlin · New York

Mouton de Gruyter (formerly Mouton, The Hague) is a Division of Walter de Gruyter GmbH & Co. KG, Berlin

앝 Printed on acid-free paper 앪

which falls within the guidelines of the ANSI to ensure permanence and durability.

Library of Congress Cataloging-in-Publication Data Tenbrink, Thora, 1969⫺ Space, time, and the use of language : an investigation of relationships / by Thora Tenbrink. p. cm. ⫺ (Cognitive linguistics research ; 36) Includes bibliographical references and index. ISBN 978-3-11-019520-0 (hardcover : alk. paper) 1. Space and time in language. 2. Grammar, Comparative and general. I. Title. P299.S53T46 2007 415⫺dc22 2007013918

Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available in the Internet at http://dnb.d-nb.de.

ISBN 978-3-11-019520-0 ISSN 1861-4132 쑔 Copyright 2007 by Walter de Gruyter GmbH & Co. KG, D-10785 Berlin All rights reserved, including those of translation into foreign languages. No part of this book may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher. Printed in Germany

Acknowledgements

This book is an abridged version of my doctoral thesis (Tenbrink 2006), written between 2003 and 2005 within the context of the SFB/TR 8 Spatial Cognition. The task of differentiating the most important from the less central ideas was rendered manageable by the very constructive and incredibly detailed comments by Eric Pederson, who kindly acted as a reviewer for the publisher. I also thank the publisher for their patience and cooperation during the revision process. Also, I wish to express my deepest thanks to my two capable, efficient, and very constructive thesis supervisors: John Bateman and Christian Freksa, who both intensely supported the work on the original dissertation project by their critical feedback and positive encouragement. Further thanks go to my colleagues of the SFB/TR 8, a project which proved to be a perfect environment for the present endeavour. Parts of the work described here have been presented in conferences and workshops, where I profited much from reviews, feedback, and discussions with participants. Other parts result from cooperation with colleagues who, by co-authoring earlier publications, have influenced my ways of thinking and researching to a high degree. Additionally, I have talked to and been inspired by a lot of people throughout the years; listing all names would certainly lead to space problems: but I do hope that each of them knows how much I honour their ideas. I wish to thank the student assistants who energetically supported data collection and transcription, the Linguistic Data Consortium who allowed free access to the Switchboard corpus data, and Prof. Erhard Mergenthaler of the University of Ulm who kindly provided me with data of the Ulmer Textbank. Special thanks go to Jannis and Lukas and to Alexander Spohr, whose support has been vital to the present achievement.

Table of contents

Acknowledgements ...................................................................................... v Chapter 1 1.1. 1.2. 1.3. 1.4.

Introduction ............................................................................ 1 The conceptual relationship of time and space....................... 2 Spatial and temporal language................................................ 3 Materials and methods............................................................ 5 Structure of the book .............................................................. 8

Chapter 2 2.1. 2.2. 2.3. 2.4. 2.5.

Time and space ..................................................................... 12 Are conceptions of time based on space? ............................. 12 Shared issues in time and space............................................ 19 Conceptual differences ......................................................... 24 Linguistic means of expressing time and space.................... 30 Conclusion............................................................................ 37

Chapter 3 3.1. 3.2. 3.3. 3.4. 3.5. 3.6. 3.7. 3.8. 3.9. 3.10.

Temporal dimensions: Linguistic repertory and issues of application............................................................................. 38 Language and temporal structure ......................................... 38 Figure and Ground in temporal language ............................. 40 Syntactic range of temporal dimensional terms.................... 42 Details of the temporal relationship...................................... 45 Causal interpretations of temporal connectives.................... 47 Presuppositions..................................................................... 51 Non-veridical interpretations................................................ 54 Non-temporal interpretations of temporal connectives ........ 57 Related terms ........................................................................ 57 Conclusion............................................................................ 64

Chapter 4 4.1. 4.2. 4.3. 4.4.

Temporal dimensions: Empirical analysis........................... 66 Relations between events: Analysis of English data ............ 68 Presupposition processes ...................................................... 86 Relations between events: Analysis of German data.......... 100 General Discussion............................................................. 113

viii

Contents

Chapter 5

5.2. 5.3. 5.4.

Spatial dimensions: Linguistic repertory and issues of application ........................................................................... 118 Linguistic repertory: Syntax and semantics of spatial terms .................................................................................... 118 Discourse-external factors involved in the application........ 130 Discourse-related factors influencing application ............... 183 Conclusion ........................................................................... 198

Chapter 6 6.1. 6.2. 6.3. 6.4. 6.5. 6.6. 6.7.

Spatial dimensions: Empirical analysis ............................... 200 The web study...................................................................... 201 Paradigmatic vs. syntagmatic perspectives.......................... 208 German web data analysis ................................................... 220 English web data analysis.................................................... 239 Comparison of English and German.................................... 244 General Discussion .............................................................. 257 Conclusion ........................................................................... 266

Chapter 7 7.1. 7.2.

Conclusion: Comparing the applicability conditions .......... 267 Summarising the results....................................................... 267 Treatment of spatial and temporal terms in natural discourse .............................................................................. 270 Comparing the applicability of spatial and temporal dimensional terms................................................................ 274 Outlook ................................................................................ 285

5.1.

7.3. 7.4.

Appendix .................................................................................................. 288 App. 1. Corpus examples from the Switchboard corpus ................. 288 App. 2. Web study (Chapter 6): Configurations ............................. 294 App. 3. Web study (Chapter 6): Coding schemes ........................... 296 App. 4. Relative frequencies: Paradigmatic / German .................... 299 App. 5. Relative frequencies: Syntagmatic / German...................... 303 App. 6. Relative frequencies: Paradigmatic / English ..................... 311 App. 7. Relative frequencies: Syntagmatic / English ...................... 315 References ................................................................................................ 323 Subject index............................................................................................ 343

Chapter 1 Introduction

Does temporal language depend on spatial language? This widespread assumption is intuitively appealing: Spatial and temporal expressions are often similar or identical. Time is generally treated as a “fourth dimension” in relation to space. There is a fair amount of metaphors that consistently express temporal phenomena in terms of spatial language. All this points to a close semantic and conceptual interdependency. But what about the application of the two kinds of linguistic expressions in natural discourse? Does the spontaneous usage of (non-metaphorical) spatial and temporal terms reflect the relationship between the two domains in any systematic way? Most research has focussed on the repertory of language, not the patterns of its usage. This book systematically explores findings on how speakers use genuinely spatial and temporal terms (in front / behind, before / after, etc.) to describe the relation of objects or events to each other. The investigation highlights the relationship between cognition and language usage. Using the method of cognitively motivated discourse analysis, novel empirical results on spontaneous usage by English and German native speakers (based on detailed analyses of various, predominantly web-based, corpora) are presented to complement earlier findings. The detailed investigation of a selected range of terms that appear to be parallel in space and time highlights both similarities and fundamental differences in their application. As a result, a new picture emerges: The concepts of space and time are represented in language usage in various systematic ways, reflecting how we understand the world – and at the same time reflecting how our concepts of space and time differ fundamentally. This book contributes to a debate that has been of interest for cognitive linguists for several decades, concerning the understanding of transfer processes between two conceptually intertwined domains. It addresses the novel question of how such processes come into play in the actual application of relevant expressions in natural discourse. By adopting established approaches from discourse analysis for issues deeply rooted in interdisciplinary research in cognitive science, insights are drawn together from two hitherto largely unrelated fields of research to approach the topic from an original perspective, leading to a deeper understanding of the relationship between the domains of space and time and their expression in language.

2 1.1.

Introduction

The conceptual relationship of time and space

In this book, cognitive science serves as the background for accounting for principles and preferences in the application of spatiotemporal expressions in context. The starting point is the investigation of general issues pertaining to spatial and temporal cognition. The focus then narrows down to the usage of English and German dimensional terms, i.e., terms expressing the dimensions of space and time, such as (in) front / vor(ne), behind / back / hinter, left / links, and right / rechts, and their temporal counterparts before / bevor, and after / nachdem, including their various syntactic variants. This selection allows for a direct comparison and intricate qualitative analysis of language usage, concerning a subset of conceptually interesting expressions, in two closely related languages within a largely shared cultural context. Both of these languages have been subject to linguistic study for a long time; crucially, there is a broad variety of earlier findings to draw from, specifically in the research area at stake, the investigation of spatial and temporal terms. This provides the necessary context for a well informed systematic comparison of applicability conditions independent of experimental tasks and across natural discourse settings. These earlier findings are complemented by targetted empirical research addressing a number of research gaps identified by the systematic account. The results spell out in more detail a range of principles that, in general, had already been identified as crucial for the usage of these expressions. The book addresses the popular hypothesis that temporal terms are closely related to, and conceptually (as well as historically) based on, spatial terms (e.g., Haspelmath 1997). While this hypothesis relies mainly on research findings concerning the morphosyntax and semantics of these terms (often in various languages), the present analysis specifically draws on research in the recently blooming area of discourse analysis, applied and accounted for in the light of insights from cognitive science. This allows for a cognitively motivated focus on the semantics-pragmatics interface that is new and unique to the discussion of spatial and temporal terms and their apparent interdependency, based on a fine-grained analysis of their actual usage in natural discourse in two closely related, well-researched and familiar languages. It deals with ways in which spatial terms differ systematically from temporal ones with regard to their particular discoursal applicability conditions. Thus, a central aim of the present study is the investigation of whether evidence for the assumed conceptual dependency can be identified with respect to application. The complications involved in

Spatial and temporal language

3

the separation of semantic and pragmatic levels in the analysis of dimensional terms are addressed and discussed in the course of the present work. One of the reasons for the presumption that temporal expressions depend on spatial ones is the idea that the spatial domain is concrete (i.e., perceptually accessible) while the temporal domain is abstract (i.e., less easy to perceive and to grasp). This is reflected by the notion that the entities associated with space are (concrete) objects, while the entities associated with time are (abstract) events. Now, if concrete entities are easier to handle than abstract ones, it seems to be natural to extrapolate from the experience gained in the concrete domain in order to cope with abstract experience. Thus, events are treated in some respects in a similar way as objects, and the underlying domain of time is understood in terms of experience gained from the more accessible domain of space. But if time is understood on the basis of concepts of space, this does not necessarily mean that patterns of usage of the linguistic expressions are similar for both domains. Spatial terms are employed in concrete settings, while temporal terms apply to the representation of abstract events. The capacity for transfer should be limited as far as application is concerned, since events are ontologically fundamentally different from objects. Therefore, the conditions for usage could be distinct in some basic and systematic respects, although parallels may well be identifiable. The present work aims to shed light on the patterns of discourse applicability for superficially (i.e., in terms of morphosyntax) similar, and semantically related, spatiotemporal expressions. It will identify the ways in which their usage patterns systematically reflect both conceptual similarities and ontological differences. This is achieved by contrasting the principles and concepts underlying the application of spatial terms with those of temporal terms, and by relating the resulting differences and similarities to the conceptualisations of the underlying domains, as well as to general principles of linguistic communication operating in discourse. 1.2.

Spatial and temporal language

During the past few decades, spatial terms have proved challenging enough to justify various approaches to their meaning and underlying concepts. Systematic research includes psycholinguistic experiments, formal analyses and specifications, and considerations of a sizeable range of variation in reference systems, i.e., ways of referring to entities in spatial surroundings

4

Introduction

by using qualitative descriptions such as to the left. However, a number of issues still remain unresolved. For instance, a systematic account of all kinds of reference systems that may potentially underlie a spatial expression is still lacking. Furthermore, the principles and preferences that lead speakers to rely on one kind of conceptualisation in favour of another have only been addressed with respect to a subarea of the available variability: most prominently, research has focused on the perspective chosen, typically either the speaker’s or the addressee’s. But already with respect to the choice between two basic kinds of reference systems available to speakers, namely, relative versus intrinsic, there is much controversy in the literature. In the present approach, these issues are addressed in a number of ways. A thorough literature review in Chapter 5 results in a systematic overview of the kinds and sub-kinds of reference systems that have been identified so far. Chapter 6 provides further empirical findings supporting and carrying further earlier hypotheses with respect to speakers’ preferences and conceptualisation processes. It also highlights in detail two major reasons why the identification of underlying concepts is specifically complex: On the one hand, there are many ways in which linguistic utterances are underspecified and incomplete in expressing underlying concepts. On the other hand, specific kinds of linguistic forms do not necessarily systematically coincide with specific kinds of underlying reference systems. Therefore, specific care needs to be taken to relate the current spatial scenario, and other information available to the interlocutors, to the linguistic utterance. For these aims, a web-based study was carried out. This approach is suited for restricting the setting in such a way as to allow for a mapping of utterances and spatial configurations. It allows for introducing fine-grained differences between spatial settings and at the same time for the collection of a great number of native speaker contributions in both English and German. Furthermore, a very common way of identifying one of several present objects is to simply point at the goal object using gestures. Since the target of the present analysis is language, not gestures, and since there is no corresponding device in the temporal domain, a setting is needed in which pointing gestures are ruled out. In a web-based study, pointing gestures (which might, if one wishes, be realised by mouse movements) can easily and naturally be ruled out by the design. As far as the related domain of temporal terms is concerned, before and after have been analysed with regard to the formal properties of the temporal phenomena involved, such as topological restrictions on the application of a term. Also, the semantics of these terms have been subject to a broad

Materials and methods

5

range of research, and some findings have been obtained concerning their presuppositional properties. But their actual occurrence in natural dialogues has seldom been in focus, creating the misleading impression that, generally, any two events can be juxtaposed verbally by combining them with the connectors before and after. However, speakers need a reason for mentioning two events together and expressing the temporal dimension verbally. Temporal order itself is generally sufficiently expressed by syntactic marking of the verb and by textual features, such as juxtaposing sentences in a way that suggests that the events happen in their natural temporal order. Thus, in general it is not necessary to add explicit markers specifying the temporal dimension; obviously, speakers do not employ before and after in every sentence in which temporal succession is implied. Therefore, it is interesting to look at the semantic and discoursal patterns in which these terms do occur. This argument is not mirrored by the spatial expressions since language does not, in a way comparable to temporal aspects, provide means of expressing spatial relationships without using explicit spatial markers. For both kinds of terms, thus, the question arises whether there are systematic differences in the application contexts in which these expressions are spontaneously applied by speakers of German and English. Further questions concern how they are used in terms of syntactic and underlying conceptual features, and in terms of other linguistic items that further contribute to conveying the intended relationship. 1.3.

Materials and methods

This book draws on a variety of data sources. About half of the book is dedicated to a systematic account of earlier results in the literature concerning how spatial and temporal dimensional terms are employed in natural discourse. These overviews already provide a fairly clear idea of the kinds of contextual and conceptual factors that come into play in each case. To complement these insights and to further spell out how the identified factors work in actual discourse, further information is drawn from original empirical analyses based on natural language corpora. For spatial expressions, the analysis uses original data collected in webbased online experiments on spatial terms for both German and English. For temporal expressions, online available corpora were used to gather English data, taken from the CHILDES and the Switchboard databases, and

6

Introduction

the CSPA sample corpus. For German, a dialogue corpus was obtained from the University of Ulm, since German spoken language corpora are not easily accessible (Schmidt 2005). These data sources offer access to a range of fundamentally different kinds of text types in which temporal dimensional terms may occur. Also, as motivated in more detail in the empirical chapters 4 and 6, the specific corpora used for each part of the analysis are specifically suitable for the approach adopted in the present work, which involves a number of additional requirements on the data sources for specific questions that are addressed in the analysis. The apparent discrepancy between the kinds of data sources used for the spatial versus temporal domains has practical as well as theoretical reasons. For practical reasons, it is unclear how a sufficient quantity of spatial expressions could be obtained in collections of naturally occurring discourse alone. As will become apparent in the present work, spatial expressions depend heavily on the spatial scenario, whether it is actually present or represented internally in a mental model shared to some degree by the interlocutors. Because of this, any analysis of spontaneously occurring dimensional terms is limited by the degree of information available with regard to the spatial knowledge that is necessary for the interpretation of these terms. A simple example illustrates this. Here is an extract from the Switchboard corpus (which was used for the analysis of temporal terms) that contains the phrase on the left: But he finds that by going straight down the middle he usually wins about a quarter a hole because they’ve been in the rough on the right and then in the rough on the left.

To understand the spatial concept behind this utterance, a number of contextual factors need to be taken into account, including the nature of the game that the speaker is talking about. But even then it remains unclear whether “on the left” refers to the person in question, or to intrinsic sides of the field in which the game takes place. A further, general problem concerns the fact that most natural occurrences of left and right are not spatial at all, rendering the analysis of syntactic diversity highly problematic. For these reasons, it is more feasible to collect data in specifically designed scenarios in which the spatial setting can be controlled, than to rely on previously collected natural language corpora in which relevant information may not be available. Web-based experimental studies are suitable in this regard: the relevant discourse factors can be controlled to a much higher degree than in natural discourse, and the spatial scenario can be reduced to a minimum and manipulated in useful ways.

Materials and methods

7

In contrast, temporal expressions do not depend on any entities present in the actual scenario, but are employed in order to juxtapose and locate events (or, more seldom, states) in time. Of course, concrete entities involved in the events might be present in a given scenario, and naturally they can also be referred to, depending on the setting, when talking about the events. But since time itself and the relationships involved in the domain are not perceptible, the conceptualisation of abstract temporal patterns as coherent events involving concrete entities must be captured in the discourse. Thus, the information necessary for interpretation must be contained in – or inferable from – the discourse itself, which makes it feasible to analyse natural language corpora. Furthermore, one of the most prominent research questions in the application of temporal relational terms concerns the ways in which they are spontaneously employed, since temporal order can also be conveyed by other means in language. This phenomenon can only be investigated using data that were not influenced by the analyst. For illustration, compare the following examples from the data used in the present work. (1.1) (1.2)

go to the cube to the left of the sphere Just one thing going back to the issue of time before you speak, Eunice.

The spatial example (1.1) requires analysis with respect to a number of conceptual factors such as the underlying perspective and the position on an axis with respect to the relatum. This kind of information is not available in the textual context, but only via knowledge of the spatial setting. The temporal example (1.2), on the other hand, does not rely on information that is not present in or inferable from the dialogue itself. The utterance invokes the idea that an event (addressing “just one thing”) is to be inserted between the present moment and another already expected event (Eunice will speak). This reflects an interesting underlying generalisable concept which may have been a motivation for using a temporal term. Throughout this book, it will gradually become clear in how far these examples may be representative of the ways in which spatial and temporal dimensional terms are spontaneously employed by speakers of English and German. Conclusions in this regard are drawn not only on the basis of the empirical results, which, after all, stem from very different kinds of corpora, but predominantly on the basis of a thorough review of the available literature in both areas. The main emphasis of the analysis in the spatial domain concerns the choice and application of spatial reference systems which underlie all us-

8

Introduction

ages of spatial dimensional terms, investigating the choice and explicitness with regard to all ingredients involved in using a reference system: an origin and a relatum, a spatial axis, and the possibility of modifying the spatial term by further linguistic means. For the temporal domain, a range of different discourse situations is investigated in which speakers naturally employ temporal expressions; here, their reasons for juxtaposing events linguistically become apparent through the discourse context. The correspondence between spatial and temporal expressions only occurs with some syntactic variants but not with others. For instance, spatial expressions cannot occur as conjunctions (example (1.5) mirroring (1.8)), and temporal dimensional terms (such as before) do not have an adjectival form (example (1.7) mirroring (1.4)). (1.3) (1.4) (1.5) (1.6) (1.7) (1.8)

The box is to the left of the sphere. It is the left box. *The box is situated left the sphere is placed. The accident was before Christmas. *It was the before event. The accident occurred before the Christmas party started.

The two languages targeted in the present work, English and German, though closely related, also differ with respect to syntactic variability: for instance, English left can appear as a superlative (leftmost); this is not mirrored in German, as linkest– is ungrammatical. Therefore, for both temporal and spatial expressions, one important part of the analysis concerns the conditions of their occurrence in different syntactic contexts. For example, the employment of before as conjunction, preposition, conjunctive preposition, or sentence adverbial may involve different semantic or pragmatic conditions and implications. For spatial expressions, it has been claimed that specific syntactic forms can only be used for specific kinds of reference systems, a claim which has not been addressed sufficiently in the literature so far. Therefore, it is a specific aim in the present work to investigate in how far different kinds of syntactic forms can be mapped to specific kinds of underlying concepts. 1.4.

Structure of the book

The main argument is reflected in the book’s structure as follows. In Chapter 2, the domains of time and space are introduced from the broader per-

Structure of the book

9

spectives of cognitive science and linguistics. Chapters 3 through 6 deal in detail with spatial and temporal language, most specifically with the dimensional terms in both domains. For each domain, there is one chapter providing a theoretical account of earlier findings (Chapter 3 for time, and Chapter 5 for space), and one chapter presenting original empirical results (Chapter 4 for time and Chapter 6 for space). Chapter 7 brings the two domains together again for a comparison, discussion, and assessment of the results in light of the general question addressed in this work, which is: Do the applicability structures of temporal and spatial dimensional terms reflect an underlying conceptual dependency, as indicated by their semantic and morphosyntactic similarity, or are these terms used in independent ways? Chapter 2 starts out from a general cognitive science approach, addressing the relationship between the conceptual domains of space and time. A widespread view is that temporal concepts are ultimately derived from spatial ones, based on the fact that time seems to be much harder to grasp than space, since space is perceptually accessible in contrast to time. Chapter 2 presents and discusses this view and then moves on to a more neutral approach, working out similarities and differences in cognitive issues pertaining to space and time. In addition, the range of linguistic means to express spatiotemporal relations is identified, and some fundamental differences are worked out that become apparent in the comparison of spatial and temporal linguistic means. Chapter 3 starts by addressing the question of how speakers represent temporal relationships through language. There are a number of methods available to speakers, such as simply presenting the events talked about in the order in which they occurred. This motivates the question of in what kinds of contexts those terms occur that explicitly specify temporal order, namely, before and after. The range of linguistic options available for temporal dimensional reference is specified along with their semantic scope. Factors influencing the application of temporal dimensional terms are specifically addressed, along with their ability to trigger associations and inferences that go beyond purely temporal information, and their syntactic variability is tested for in this regard. While Chapter 3 is based exclusively on previous findings in the literature along with discussions and considerations with respect to the present approach, which includes syntactic reformulation tests that lead to intuitive judgments of applicability, Chapter 4 turns to the empirical analysis of naturally occurring language. Here, syntactic reformulation tests are not

10

Introduction

adopted, but instances of temporal dimensional terms that are produced spontaneously by native speakers of English and German are analysed with respect to the situational and discourse context in which they occur. The analysis does not apply statistical measures, but is entirely qualitative and rather fine-grained, taking the discourse context into account. In this way, conceptual categories underlying the application of the temporal dimensional terms can be worked out which have not been systematically accounted for in the literature before. A further aim of the analysis concerns presupposition processes accompanying the application of the terms. These are identified by investigating in how far speakers account for their interlocutor’s prior knowledge, what linguistic markers are used to signal whether the information is taken as given or not, and whether there are systematic contextual situations in which temporal terms mark presuppositions. This kind of analysis is also exceptional because it accounts for the wider discourse context in a way that is seldom encountered in the literature. Analyses are carried out first for English, then for German expressions. The depth of analysis is slightly enhanced for the English data due to the available data sources. Thus, the analysis on English is used as a starting point, building the basis for comparison between the application of the terms in the two languages. Chapter 5 is specifically concerned with spatial language. First, the spatial dimensional terms are categorised and contrasted from other kinds of spatial expressions. Their syntactic range is identified. Furthermore, their basic semantics is discussed. Here, a non-trivial question is in how far it is possible to identify a core meaning that is flexible enough to account for all usages and discourse contexts where the terms are naturally applied, and the ways in which they can be interpreted. The major part of the chapter then deals with the factors that influence the application of spatial dimensional terms: the kinds of underlying reference system (relative or intrinsic as suggested by Levinson’s approach of 2003; external or internal, groupbased or landmark-based, etc.) and the perspective used (that of speaker, interlocutor, or a third entity), the function, distance and accessibility, and spatial extension of the objects that are linguistically related by a spatial dimensional term, as well as diverse kinds of discourse contexts in which these terms may occur. The notion of spatial templates (applicability areas of spatial expressions) is discussed, and the flexibility in conceptualising a spatial situation is accounted for. Thus, this chapter presents a thorough and systematic overview of the syntax, semantics, and application of spatial dimensional terms that in several respects goes beyond previous accounts

Structure of the book

11

in the literature, in spite of the vast amount of work that has been dedicated to the investigation of these terms during the past few decades. Some confusions in the literature that may be based on this lack of systematicity are identified and clarified along the way. Chapter 6 then turns to the empirical investigation of spontaneously produced spatial language. It starts out by discussing the complexities involved in interpreting such language, specifically with respect to the identification of underlying reference systems. To a high degree, complications arise because of underspecification or omission of crucial conceptual factors, such as the relatum and the underlying perspective. A method is presented in detail that sets the stage for a thorough and systematic qualitative analysis of spatial language, supported by the relative frequencies of occurrences of linguistic features. The data source is a web study carried out for German and English, designed for identifying basic underlying principles and preferences for the application of spatial dimensional terms. The final chapter re-opens the issues concerning the assumption that temporal language is conceptually based on spatial language, which were introduced in Chapter 2. Now, the findings on the semantics and application of temporal versus spatial dimensional terms are brought together for a well-founded comparison of the systematic principles underlying spontaneous usage. Furthermore, the different methodologies used in the empirical work are discussed against the background of the present findings, and directions of future research are addressed. The results of the literaturebased discussions in Chapters 3 and 5 as well as the empirical investigations in Chapters 4 and 6 both lead to the identification of basic principles of applicability, which are in Chapter 7 related to the differences pertaining to the conceptualisations of the two domains of space and time, as addressed in Chapter 2, and to general principles of communication active in discourse, as addressed at several places throughout the book. As an outcome, no conceptual dependency relation between the applicability conditions of spatial and temporal expressions can be identified, in contrast to previous findings relating to the morphosyntax and semantics of the terms. Instead, the comparison identifies a number of interesting similarities and parallels, but also fundamental differences in application. This allows for the conclusion that the applicability structures of spatial and temporal dimensional terms in German and English are essentially independent of (though in some ways parallel to) each other, in spite of their high degree of similarity and in spite of the intricate relationship between the conceptual domains of space and time.

Chapter 2 Time and space

According to a widespread conception in the literature of linguistics and philosophy, space and time are deeply intertwined with each other, both in language and in underlying concepts. Moreover, one of the domains – space – is often viewed as basic, while conceptualisations of the other – time – allegedly depend on the basic domain of space by metaphorical transfer. This relationship is claimed to be specifically transparent in language. According to Haspelmath (1997: 20), this “spatialist conception” is “apparently uncontroversial nowadays”. In this chapter, I will take a closer look at these ideas, narrow down the areas in which metaphorical concepts take their rightful place, and disentangle similarities and differences with regard to basic conceptions of space and time. In subsequent chapters, linguistic phenomena will be considered separately for each domain, to be brought together again for comparison in the final chapter. 2.1.

Are conceptions of time based on space?

The idea that the human concept of time is based on the concept of space is not new; indeed I will not attempt here to trace its origins. More than a hundred years ago, for instance, the philosopher Jean-Marie Guyau (cf. 1988: 124f.) stated: It is primarily by means of space that we determine and measure time (…). The present moment is clearly the point of origin for any representation of time. We can only conceive of time from a present perspective, in which we represent the past behind us and the future in front of us. But this perspective is always a spatial scene, some event that occurred in a material and extended context. The form of our representation of time, the way we imagine it, is essentially spatial. The space we perceive is in front of us; the space we simply represent without perceiving it is behind us. In fact, we can only represent the space behind our backs by imagining that we are frontally facing it. So it is with time; we can envision the past only as a perspective behind us, and the future emerging from the present as a perspective in front of us.

Are conceptions of time based on space?

13

This early citation highlights several basic elements of the general notion that human time concepts are based on space concepts: x It is not time itself but conceptions and representations of time that depend on space. x Time is measured by means of space. x Time is conceived of in terms of events (that take place in space). x The past is conceived of as behind us / the present moment, while the future is in front. Half a century later, Piaget’s research showed that young children have difficulties in separating temporal and spatial phenomena (Piaget 1946). They mix up spatial and temporal concepts especially in cases where both are involved, as in the filling up of containers with liquid, or in tasks involving movement. Even today, Piaget’s findings are often viewed as proving the conceptual intertwinement of space and time. Furthermore, space is used frequently for representations of time; Friedman (1990: 50) observes: (…) the spatial arrangement of symbols is so frequent a format for representing temporal relations that one wonders whether there is something psychologically natural about it.

Because of this intricate relationship between space and time, the view emerges that the relationship between temporal and spatial concepts is a metaphorical one. The argument goes as follows: Because we conceive of time in terms of space, we extend our concrete spatial experience metaphorically to the abstract conception of time. In the field of linguistics, probably the most influential proponent of this view is Clark (1973), who claims that temporal language is based metaphorically on spatial language, and that many relational temporal prepositions in English, such as before, after, ahead, behind, etc., are historically derived from front and back.1 Clark bases his claims on the assumption that the now primarily temporal expressions originally had spatial meanings. However, this claim2 can1. 2.

Clark does not provide any motivation for this claim. Tyler and Evans (2003:157) show that before is derived from be + Old English fore, meaning “located at the front of”. The OED entry of 1961 states that early occurrences of “spatial” before and after, from which purely temporal interpretations may derive, denote sequence in space; they are overwhelmingly presented as occurring only together with verbs of motion, which obviously are not independent of time. Moreover, there seems to be no distinct time lag between such spatialsequential early occurrences and clearly temporal usage.

14

Time and space

not be validated in light of the present state of knowledge, although many crucial similarities between spatial and temporal expressions have been pointed out in cross-linguistic research (e.g., Svorou 1994). In fact, a widespread assumption is that there is a far-reaching identity between spatial and temporal terms – if not synchronically, then at least diachronically. However, Traugott (1978: 373) points out that “synchronically an exact correlation between spatial and temporal terms is infrequent, if not unknown.” Furthermore, “at no synchronic point in a language is the relation sufficient to justify the claim that there is underlying identity”. Almost twenty years later, Haspelmath (1997) addressed this issue with regard to a large number of the world’s languages. His explicit aim (Haspelmath 1997: 4) was “to assemble cross-linguistic evidence for the hypothesis that temporal notions are conceptualised in terms of spatial notions.” Haspelmath’s basic approach is to identify linguistic surface similarities between spatial and temporal terms, to determine – as far as possible – whether one of the domains is etymologically prior, i.e., whether temporal terms are historically derived from spatial ones, and to identify cross-linguistic patterns and regularities that point to a conceptual dependency of temporal on spatial phenomena. He argues that, while it is not possible to prove such a dependency on the basis of linguistic data alone, it is highly probable that such a relationship exists if there is a general tendency for temporal expressions to be based on spatial ones. The impressive amount of data he presents, based on 53 languages, is interpreted to support this view because of the many similarities found between temporal and spatial expressions. However, in no language does there seem to be an exact identity between spatial and temporal terms, either synchronically or diachronically, making Traugott’s above observation still valid. Haspelmath further claims that the similarities can be explained by the assumption that time is based on space, but not – or at least not as consistently – vice versa. According to his analysis, time is ubiquitously secondary to space, whether one wants to interpret the relationship as metaphorical or not. In Haspelmath’s view, such interpretational variation can largely be ascribed to broader vs. narrower views of the term ‘metaphor’. A closer look at Haspelmath’s data and line of argumentation shows that, overwhelmingly, he simply points to existing similarities between spatial and temporal linguistic expressions. The arguments for the conceptual priority of space are mostly not (purely) linguistic: it is an intuitively appealing statement that space is prior because it is more concrete and can therefore be used for abstract concepts. It is also intuitively plausible that –

Are conceptions of time based on space?

15

as already Clark (1973) claimed – the spatial frontal axis is a good candidate for a mapping of temporal relationships using similar linguistic expressions. But the (etymological) dependency of temporal on spatial terms based on independent linguistic evidence seems fairly hard to prove. What Haspelmath claims to be diachronic evidence mostly amounts to the observation that, in some cases, the identity of spatial and temporal terms occurs in the present language, while in other cases, it is traceable in previous usage but today there are separate expressions for each domain. Thus, while Haspelmath’s research impressively underscores the fact that there are many similarities between spatial and temporal concepts as represented across the world’s languages, suggesting that there must certainly be some kind of conceptual relationship, the analysis of similarities as dependencies or even metaphors remains a matter of interpretation. To describe the spatial metaphor that, according to Clark (1973), can explain all temporal concepts, Clark puts forward two opposing but compatible notions: in the “moving time” metaphor, time is viewed as moving past us from future to past; and in the “moving ego” metaphor, we as humans are viewed as moving through time from past to future (see also Miller and Johnson-Laird 1976: 463). Terms like before (= in front of) and after (= in back of) derive from the “moving time” metaphor: Clark (1973: 50) accounts for the meanings of these terms by claiming that events are ascribed a front and a back, so that before noon means that the front face of noon is the one that leads and is directed pastward, and the back face is the one that follows and is directed futureward.

In the “moving ego” metaphor, in contrast, it is not the event’s ascribed front or back that is taken as a reference point but the ego’s, as in “I look forward to Monday”. However, the presence of metaphorical expressions in language does not prove the existence of an underlying pervasive conceptual metaphor that is responsible for all representations of one domain. Metaphors are fairly frequent in language, and they are often impressively consistent (Lakoff and Johnson 1980), but this does not imply that the concepts that are represented in metaphorical language could not also be represented in an independent way. The argument holds only in the opposite direction: Metaphorical usage suggests compatibility with a different domain. Metaphors are used for highlighting similarities, but they do not embrace the whole target concept, nor does their usage imply that all aspects of the source concept can be translated to the target concept.

16

Time and space

Radden (1997), for instance, explores what he calls the conceptual metaphor “Time is Space”. According to his research, the conceptualisation of time as space as evidenced by metaphorical expressions takes on many different forms and variants, between languages as well as within one single language. For instance, he points out that – apart from Clark’s two compatible metaphors of “moving time” vs. “moving ego” – there is also a widespread conception that time itself is moving, not from future to past as in the “moving time” metaphor, but from past to future. This becomes obvious, for example, in the natural interpretation of from February to January, where we conceive of a time span of twelve months (going forward) rather than one of only one month (going backward). Furthermore, Radden points out that there are many more metaphors through which temporal concepts may be expressed, such as time on a vertical axis as in Chinese culture, time as an object as in having a hard time or personified time, as in Time will tell. Thus, far from being subject to a consistent transfer from a single spatial domain (or – even more narrowly – from the front-back axis of the spatial domain) to the temporal one, as Clark (1973) claims, temporal and spatial concepts undergo similar complex metaphorical processes as many other concepts as described by Lakoff and Johnson (1980). A different approach to the semantic relationships between superficially similar terms used in various domains has recently been developed in detail by Tyler and Evans (2003), based in part on previous insights by Grady (1997), Talmy (1983) and Langacker (1987, 1991). They propose that, while it is their conviction that all other senses are derived from an originally and primarily spatial sense, the relationship is nevertheless not metaphorical as claimed in previous approaches, but experiential. In their view, experiencing regular and motivated correlations of events and scenes in the real world motivates the integration of such regularities into language. In this way, up became associated with more, since it is a regular experience in spatial scenarios that larger amounts reach higher up (e.g., liquid in a glass). Together with pragmatic strengthening it may then happen that concepts that were originally simply associated with other concepts in the usage of a certain preposition, come to be expressed by that preposition even without the presence of the original concept. Thus, in The prices have gone up there is no longer a spatial sense of vertical direction involved. This view is developed in terms of a generalised principled polysemy approach for a number of spatial prepositions. Basically, Tyler and Evans’ approach is in line with previous analyses in the cognitive linguistic tradition, distinguished mainly in their methodology that is sufficiently detailed to steer

Are conceptions of time based on space?

17

away from the traditional, rather vague conception of underlying metaphorical relationships, towards a more thorough investigation of how such cognitive dependencies may be explained. However, an underlying pervasive primacy of the spatial domain in the sense that all other experiences expressed by the prepositions are derived from spatial experience, is still postulated, which is apparent in the authors’ proposal of spatial protoscenes for the semantics of before, after, and even of (a term which they themselves acknowledge has little spatial meaning in current usage). An alternative conception giving less weight to the spatial domain has been put forward by Jackendoff (1983), who claims that the similarity between spatial and temporal conceptions is based on an underlying thematic structure that organises all conceptual domains in a similar (quasi-spatial) fashion (Jackendoff 1983: 209). Although Jackendoff rejects the idea of a direct dependency of other domains on spatial conceptions, he acknowledges that space is somehow specific (1983: 210): I am inclined to think of thematic structure not as spatial metaphor but as an abstract organization that can be applied with suitable specialization to any field. If there is any primacy to the spatial field, it is because this field is so strongly supported by nonlinguistic cognition; it is the common ground for the essential faculties of vision, touch, and action.

The general idea of an abstract – not exactly spatial – underlying structure that serves as basis for a variety of conceptual domains can also be found in Habel and Eschenbach (1997). Their approach offers an explanation why some terms are only applicable to one domain while still reflecting a structure that is similar to structures in other domains (1997: 374f.): As alternative explanation in the approach of abstract spatial structures we propose that different domains exhibit common structures (...). Accordingly, linguistic expressions that are applicable in diverse domains specify restrictions regarding the structure of domains in which they are applicable without explicit reference to these domains. Thus, their meaning is abstract in the following sense: meaning has not to be based on specific properties of individual domains but on more general properties which can be common for several domains. Although space and time are independent cognitive dimensions they have some ordering principles in common.

This approach allows for time to be an independent dimension that must be clearly differentiated from space, although there are some common principles. Habel and Eschenbach point out that there are some fundamental differences (1997: 376):

18

Time and space The cognitively most relevant difference between space and time is the inherent direction of (cognitive) time. (…) Straight lines in space are neutral with respect to direction. (…) To sum up, the domains of space and time have ordered geometry as a common structure, but they are fundamentally different with respect to the inherence of direction. The constraints for using prepositions as between, before, in front of, behind and after reflect the structural communities and differences of the domains. The common use of between in the spatial as well as in the temporal domain can be explained by assuming between to code a general, non-directed ordering relation (cf. Habel 1990). In contrast to this, the system of directional prepositions does not allow a canonical transfer from the spatial to the temporal domain: in front of – behind of the spatial case opposes temporal before – after. This difference can be related to whether the domain in question has an inherently distinguished direction.

In this view, space is not seen as a concrete source domain from which the more abstract concepts of time are consistently derived. Instead, space and time share a range of representational structures, which are systematically reflected in language. But the two domains are also sufficiently different to allow for an independent conceptual representation for each of them. Summing up, the similarities between spatial and temporal concepts can be explained by an underlying metaphorical relationship only to a certain degree. Metaphorical conceptions are certainly responsible for consistent transfers between domains such as those described extensively for many areas by Lakoff and Johnson (1980), including the conceptual metaphor “time is space” which is evidenced by many uncontroversially metaphorical expressions and idioms in language (Radden 1997). But there are limits to this phenomenon which need to be acknowledged. Certainly, the processes underlying the usage of otherwise purely spatial terms for temporal relationships as is the case in phrases such as “Difficult times lie ahead” need to be differentiated from those underlying the usage of a single linguistic expression for both spatial and temporal concepts; for instance, German vor has both spatial (in front of) and temporal (before) interpretations, both of which are in current usage. The difference can perhaps best be demonstrated by examining the German distinction between nach (after) and hinter (behind). Nach and after correspond conceptually to the spatial forms hinter and behind in that they can be employed nearly synonymously for spatiotemporal relations involving movement (“He is running behind / after me”). Usually, in German, temporal concepts (not involving spatial aspects) are expressed via the temporal expression nach (or other terms like earlier, later, etc. in their

Shared issues in time and space

19

German equivalents). But there is a metaphoric alternative conceptualisation that directly employs a spatial preposition, as in: das liegt hinter mir (“that lies behind me”). Interestingly, the orientation in the metaphorical concept is opposite to the non-metaphorical one, in that events that lie behind me are events that happened before (not nach or after) future events (which are conceptualised as in front of me in das liegt noch vor mir – “that lies in front of me”). In this way, metaphorical and non-metaphorical usages of spatial and temporal expressions can be usefully disentangled. While the metaphorical approach highlights many important aspects of the relationship between space and time, it neglects and obscures others. Most importantly, it does not account for the existing differences between temporal and spatial representations, both in language and other modes of representation. But in addition, it imposes a primacy on one of the two domains in order to account for the pervasive similarities. In the following, I abandon the metaphorical approach, at least for the moment, and leave uncontroversially metaphorical expressions such as “Difficult times lie ahead” aside. Instead, I adopt a more neutral view upon both similarities and differences with regard to spatial and temporal concepts, avoiding a bias towards a presupposed superiority of the spatial domain, or a presupposed necessity to explain all human temporal conceptions on the basis of spatial ones. The next sections serve to identify commonalities and differences between concepts in time and space. It can be assumed that the shared issues are sufficient to explain the many similarities between linguistic expressions in both domains, while the differences are also reflected in language (as well as other modes of representation) to some degree. 2.2.

Shared issues in time and space

2.2.1. Abstraction In both time and space, it is equally conceivable to consider the abstract domains as independent of entities that fill them, such as objects and events, or to view them as constituted by objects and events (Freksa 1997). Clearly, the domains serve as a means to locate either situations/events or material objects in them; therefore, abstract and more concrete notions are interrelated to a high degree. Differentiating between the abstract notion of the domains and the entities in them is however necessary in order to determine the properties of any of the concepts.

20

Time and space

In both domains, the relation between (concrete) entities and (abstract) locations is asymmetric: while all entities have a location in time and space that is implicitly assumed (while not always specified explicitly) when dealing with them, the locations (times, places) can be referred to without assuming entities (events, objects) that occupy them. But there are also some important differences between the entities in the two domains with regard to abstraction. Objects are material, concrete, and may have clear (coherent) shapes, while events are abstract entities that are perceived and represented as coherent, but it is difficult to determine objective criteria for such coherence. Miller and Johnson-Laird note that the conceptualisation of changes as events depends not only on perception but also on the nature of the perceiver’s conceptual representation of the world. This has consequences on language (Miller and Johnson-Laird 1976: 90): Because the perception of a change or pattern of changes as an event is highly subjective, our talk about events is even more dependent on conceptualization than our talk about objects. That is to say, some pattern of change may be regarded as an event simply because that is how we talk about it, not because we perceive it that way.

It will be seen in subsequent chapters that the conceptual representations of changes as events interrelates with notions of causality: we perceive events as connected because they are causally related, rather than because of any kind of perceptual coherence, as Miller and Johnson-Laird point out. Such conceptual phenomena may be reasons why time is so often described as an abstract domain, in contrast to space which is viewed as concrete – in spite of the fact that both domains as such are abstract notions. 2.2.2. Choice of representation All entities have different kinds of features and properties and can therefore be characterised in various ways. For example, depending on context, relevance, informativity, and other factors, Mr Smith may variously be introduced as Jim, my uncle, your brother, her teacher, the famous golf player, etc. While this may appear as a matter of linguistic choice, it is more than that: it is a reflection of underlying processes that are influenced by salience in a given context, by perceptual aspects, by focus of attention, by the underlying conceptualised level of granularity (Hobbs 1985, Habel et al. 1993), and many other variables. In non-linguistic representation, basically the same processes become apparent in the choice of what exactly and how

Shared issues in time and space

21

it is represented, as can be seen, for instance, in maps that show streets and bus stations, but not houses or trees, etc. (cf. Barkowsky and Freksa 1997). Here, a choice is made that pertains to the given task of informing the map reader about the relevant properties of a city in their spatial relationship to each other. Likewise, linguistic descriptions single out the properties of entities, or relationships between entities, that are relevant in a given context. In spatial localisation, a box can be described as being located on earth, near the church, to the left of another object, or simply as here, depending on requirements, though all of these descriptions (and theoretically indefinitely many more) might be true. Similarly, theoretically there is no limit to the kinds of temporal relationships that may be expressed; there is no law that forbids saying World War II began after I had brushed my teeth. It is one of the aims of the present work to identify the reasons for the choices speakers actually make in discourse: what kinds of objects and events are related to each other, what determines the preference of certain kinds of choices to others, and what influences linguistic principles such as systematic allocations of figure vs. ground (e.g., Talmy 1983), by which one entity is focused on while the other serves as background. 2.2.3. Deixis When talking about relations between entities it is possible (and very common) to use the speaker’s current context for reference. In linguistics, this phenomenon has been dealt with under the heading of deixis: Both temporal and spatial relationships can be expressed linguistically by implicitly or explicitly relying on the present position of the speaker and/or the moment of speaking, using expressions such as here and now. But deixis in the spatial domain can also be expressed non-linguistically using gestures. It is hard to see what could be a (non-linguistic) equivalent to gestures in the temporal domain.3 The term ‘deixis’ is also used frequently in the literature to distinguish various kinds of reference systems. This usage is motivated by the fact that spatial terms are sometimes used relying on the speaker’s vantage point, and sometimes in relation to (the features of) other entities. However, as Levinson (2003) notes, it is not sufficient to differentiate between intrinsic 3.

In sign language, temporal relationships can be expressed by specific hand signals. These are regarded as language here, not non-linguistic gestures.

22

Time and space

versus deictic reference systems, since this approach leads to diverse kinds of confusions (see Section 5.2.2 below). In the present work, therefore, the term ‘deixis’ is used independent of frames of reference; it relates solely to utterances that directly rely on the speaker’s temporal or spatial position, not orientation, as is the case in the employment of reference systems. For example, a term like today can only be interpreted in relation to the moment of speaking, just as this place refers to the speaker’s location. In contrast, non-deictic utterances are independent of the speaker’s actual position in space and time, as is the case in the mention of dates like July 1st, 2004, and place names, such as Bibliothekstraße 1, Bremen, Germany. 2.2.4. Proximity In both domains, entities are perceived to be particularly relevant to one another if they are in close vicinity to each other. This idea is captured by the notion of proximity, which is important for the interpretation of both temporal (Herweg 1991, Schilder 2002) and spatial relational expressions (Langacker 1986, Landau and Jackendoff 1993). With deictic terms such as here and there, then and now – and next, a term which can be used with both temporal and spatial interpretations – the issue of proximity is specifically obvious and equally relevant in both domains (Traugott 1978). 2.2.5. Dimensionality The abstract domains of space and time differ with respect to their dimensions (see Section 2.3.1 below). Nevertheless, many kinds of entities, in both time and space, can be conceptualised as extending to zero or more dimensions (see e.g., Habel et al. 1993). In space, objects and locations can often be represented in more than one way. Landau and Jackendoff (1993: 222) distinguish between surface-type and volume-type objects. Surfacetype objects are conceived of as two-dimensional, while volume-type objects are treated as three-dimensional. If a surface-type object, such as a (gramophone) record, is described by an attribute like big, the description applies only to the horizontal extensions, but not to the vertical dimension, since records are conceived of as two-dimensional. In contrast, a “big house” is usually conceived of as “big” in all three dimensions. Whether we conceive of an object as one–, two–, or three-dimensional may be a matter

Shared issues in time and space

23

of purpose. Thus, whether something (such as dust) is referred to as being in the bowl (a three-dimensional conception) or on the bowl (a surfacerelated, two-dimensional conception) may depend on context-related preferences (Coventry and Garrod 2004: 26). In time, events can be conceptualised as either (zero-dimensional) points or (one-dimensional extended) time spans. In example (2.1) the writing event is first represented as a point and then as a time period, as evidenced by the use of the progressive form. (2.1)

Michael wrote his section yesterday. While he was writing it, his daughters arrived.

“Dimensionality” as described here should not be confused with the notion of dimensional terms as employed, for example, by Wunderlich and Herweg (1991), and in the present work. This notion points to a different aspect of the spatial and temporal dimensions, namely, the relation of two (or more) entities on the dimensions rather than the extension of a single entity. 2.2.6. Movement In movement, both space and time are relevant. Obviously, the domains are both present in all matters of the real world: objects are located at a specific place at a specific time – the location may thus differ between times. Events take place at a specific location; similar events may take place at different locations. However, in representations of object locations or event times there is usually a clear focus on either of the domains, neglecting the other (e.g., by not encoding it in language at all), presupposing it, or presenting it as background information. With respect to time, however, a certain amount of information is necessarily provided by the verb’s tense in every utterance (cf. Section 2.4.3 below). Concerning movement, Miller and Johnson-Laird (1976: 462) note that there is such a deep interconnection between the two domains involved that one may be linguistically expressed by the other: [W]hen a person is in motion through space, there is a reciprocity between temporal and spatial designations. He can refer to distances by using temporal expressions (‘The house is about five minutes away’), and he can refer to time by using spatial expressions (‘He started to feel ill about 5 miles ago’; this usage is appropriate only when he is in motion).

24

Time and space

Since space and time are simultaneously involved in creating a trajectory of movement, it seems to be sufficient in some cases to express only one of these aspects in linguistic descriptions. The discourse context will typically provide the reasons why one is chosen in favour of the other. The implications with regard to the other domain may be inferred on the basis of the information available. Thus, an expression like five minutes away implies that the target is at a distance that can be reached within five minutes. Note, however, that this mapping of space and time is underspecified because it does not provide information about the manner of movement, including speed. Such information must also be inferred from the discourse context if it is not provided explicitly. In verbal route directions, movement is verbalised in order to instruct a person to move in a certain fashion, from one location via others to a goal location in a specific temporal sequence. Here, spatial (rather than temporal) information typically dominates because the discourse aim is to reach a specific spatial location. In Section 5.3.1 I will briefly address this kind of discourse task. 2.2.7. Summing up: Shared conceptual issues Many conceptual issues cross over the domains of space and time. Since they are relevant in both domains to similar degrees, there is no reason – as far as the issues summed up in this section are concerned – to assume an underlying metaphorical relationship or a dependency of time concepts on spatial concepts. On the contrary, in some cases (e.g., dimensionality), similar issues are dealt with in considerably different ways. 2.3.

Conceptual differences

As Freksa (1997) notes, several issues pertain to space but not to time, and the other way round: In the spatial domain, concepts such as location, shape, and size (height, width, length, etc.) are relevant. Issues pertaining solely to the temporal domain are duration, precedence, simultaneity, consequence, etc. Not all of these issues will be discussed here, since language (which is the main concern in this work) often abstracts from them. In the following, I will focus on concepts that are clearly relevant in linguistic representations of the domains.

Conceptual differences

25

2.3.1. Dimensions and directionality One of the most obvious differences between space and time is that space (as an abstract domain) is conceived as three-dimensional but time as onedimensional (linear). Phenomena of linearity, sequentiality, and directionality can be found both in time and space, but they play very different roles in the two domains. For time, these notions are much more central than for space. Linearity is crucial via the fact that any event (or time span) that is not simultaneous or overlapping with another event, takes place either before or after the other event. There are no alternatives to this. In space, in contrast, there are several possible dimensions and therefore many different options of ordering entities. In time, a theoretical alternative to linearity is cyclicity; in that case, time would repeat itself periodically or irregularly. According to some accounts there are cultures that actually perceive time as cyclic; however, such accounts may be due to a failure to differentiate consistently between (abstract) time and (concrete) events (cf. Klein 1994). Events are repeated periodically (or “cyclically”) in many cultures (including our own), considering the days of the week, Christmas and Easter celebrations, and the like. Directionality comes into play in the domain of time, for instance, in that we only have direct access to the present (Klein 1994: 62); in relation to an origo (a given position in time), the past time is accessible by memory, but the future only by imagination. It is not possible to move forward or backward in time. A common conception is that time itself is moving in the direction of the future, and we together with it, without being able to influence this directed movement. But we use the directionality to follow our own plans for the future, as reflected in the scheduling of actions and events. Often enough, such schedules do not only define time spans for specific unconnected actions that need to be carried out independently of one another, but they account for a certain underlying logical order: some actions need to be done before other actions, constituting preconditions for the latter ones; some actions are more important than others and should be carried out first for reasons of priority; some actions regularly occur in a certain order, following a habit or tradition. Space, in contrast, has no intrinsic directionality (Habel et al. 1993). Directionality in space occurs only with motion or with functional connotations, as, for instance, with people waiting in a queue. Linear aspects are relevant for ordering and sequencing phenomena. But generally, terms like left and in front of are conceptually different from before and after in that

26

Time and space

there is no a priori direction involved in spatial settings (except for a certain priority of the frontal axis as such, cf. Section 5.2.1). However, descriptions of spatial relationships are regularly, almost necessarily, linear; thus, a sense of sequentiality is imposed on spatial settings via the linguistic representation, mediated by the temporal aspect of representing locations in a comprehensible way. One area in which linearity and directionality play a specifically central role is route knowledge, the knowledge of how to reach a target from a starting point, possibly via other subtargets (landmarks). Here, space is essentially experienced as linear and oriented because the temporal order involved in the wayfinding process carries over to the spatial domain. Static knowledge about routes blends out the temporality, while actually following a route involves (linear aspects of) both domains. However, already static route knowledge incorporates a sense of ordering that goes beyond mere knowledge about the position of landmarks (Siegel and White 1975). When describing linear spatial and temporal relationships, the spatial domain is much more flexible than the temporal. Garnham (1999: 47f.) therefore finds that language is not a medium ideally suited to encoding spatial information. Relations that are simultaneously present in spatial arrays, have to be presented sequentially in linguistic descriptions of them.

Thus, linguistic descriptions of spatial situations always depend on the speaker’s flexible choice of how to order the entities described (unless they are themselves ordered on some imposed axis), since space offers three dimensions but language is inherently sequential (see Section 5.3.2). In the temporal domain, linguistic descriptions often reflect the natural order of events on the temporal axis, which may render representations easier (Herrmann and Grabowski 1994: 113). Nevertheless, it is by no means obligatory to stick to the natural order of events (see Chapter 3). Because of the greater complexity and flexibility in space, spatial relations need to be expressed through the use of reference systems and perspectives, while temporal relationships (such as those expressed by before/after) are always unambiguous and cannot be expressed or interpreted from different points of view. Non-deictic temporal relationships thus denote the order between two events on the time line independent of a speaker/hearer’s position, while deictic temporal relationships (such as those expressed by now and then) always depend implicitly on a specific point in time (usually the moment of speaking) which can be determined by the context and thus be integrated into a temporal sequence (either before

Conceptual differences

27

or after the time span mentioned in the respective utterance) without presupposing an underlying reference system. In space, in contrast, it is additionally possible to adopt the perspective of the relatum or reference object (in case the relatum has intrinsic properties), which gives rise to contradictions and misunderstandings; furthermore, in relative reference systems still other perspectives may be chosen, such as the interaction partner’s or a third entity’s. These phenomena will be dealt with systematically in Section 5.2.2 below. 2.3.2. Accessibility, simultaneity, and succession Time and space differ in their accessibility: it is possible to move around freely in space, but not in time, which cannot be explored actively. This allows for much more varied interaction with the spatial domain in contrast to the temporal. Simultaneous events and spatial arrangements can be accessed via visual perception in a fashion that is not comparable to any of the other senses, and that is not available with respect to successive events in time. The fact that spatial entities can be accessed perceptually is associated with the fact that concepts in the domain of space are viewed as more concrete than (abstract and perceptually non-accessible) concepts of time (e.g., Wunderlich 1985: 67). Space is associated with perception, while time is associated with memory (e.g., Dunlea 1989: 10). Due to the lack of coexistence of temporal instances, and the fact that time is not perceptually accessible, it is not possible to point to temporal instances using gestures, as is regularly done in the spatial domain. Thus, information conveyance with respect to the temporal domain is limited to other modalities that ‘freeze’ time, for instance, via spatial arrangements of symbols (Friedman 1990: 50), or that represent it in temporal order, as language can do. Another aspect following directly from the observations on accessibility and concreteness is that entities in the spatial domain are manipulable, e.g., they can be moved from one position to another, as in (2.2). With temporal entities, such operations are not possible: past events are conceptualised as having occurred in a particular sequence, which cannot be altered, as exemplified by (2.3). Future events are conceptualised as not yet real and can therefore not be manipulated, as (2.4) shows. (2.2) (2.3)

Move the box to the left (of the other box). * Move World War II before World War I.

28 (2.4)

Time and space * Move the events that will happen tomorrow night to the time before lunch.

Only the scheduling of future (i.e., not yet real) events can be altered because it is not the events themselves, but their assignment to a specific point in time that is manipulated, not in the real world but with regard to planning. Therefore, it is possible to (metaphorically) move forward a meeting. With respect to the past, it is also possible to change the assignment of an event to a specific point in time – again, not with respect to the event itself, but in this case with respect to the assumed time at which the event took place. Thus, it is possible to talk about misdating or dating back an event. Note that the morphemes used in these cases are usually spatial, not temporal: schedules are changed by moving the events forward and back, and in German nach vorne and nach hinten. In German, it is also possible to talk about vorverlegen (move forward), but not *nachverlegen. Since vor is both temporal and spatial in German, the expression can be used unproblematically. For a movement of a scheduled event in the opposite direction, spatial nach hinten or a different expression (such as auf später verschieben) must be used. These complexities and involvements of metaphorical processes reflect the fact that temporal entities cannot be manipulated in the same way as spatial entities can. Since objects can be co-present and perceptually accessible in a scenario, it is possible to point to objects, and to identify one out of several possible objects using language, as in answering to a question like Which object do you mean?. To reach this aim, a (qualitative) contrast to competing objects is established, i.e., sufficient information must be provided to single out the goal object from other candidates. This happens in relation to the current referential domain (e.g., Brown-Schmidt and Tanenhaus 2003), which is restricted, among other aspects, by the field of vision. Identifying one event out of several possible candidates is also conceivable (At which Christmas did that happen?) but rare; usually the question would be posed in a different way, e.g., as What year was that? Since events take place in space, simultaneous events may be perceptually accessible and contrastable using spatial reference: Do you mean the right or the left running event? But identifying one out of several candidates of events on the basis of their temporal features is only possible if they are not co-present. In that case, it is difficult to restrict the domain of reference in a sensible way similar to that naturally given by a perceptually restricted present scenario in which a contrast to competing candidates can be established.

Conceptual differences

29

While “Where” contexts have been examined in great detail in the work of the “Space” project of the MPIP Nijmegen (Levinson 2003) as well as in other work in spatial cognition, “Which” (identifying) contexts have largely been neglected. However, because of the conceptual differences due to accessibility, they seem to be specifically crucial with respect to the relationship between spatial and temporal concepts. Therefore, they receive specific attention in the present work (Chapter 6) in order to identify underlying systematic patterns of applicability, in relation to what is known about application in “Where” contexts (Chapter 5). 2.3.3. Semantic and functional relationships between entities To understand and represent relationships between entities in both domains, additional phenomena come into play. For example, there is a well-known interrelation between the concepts of causality and time (Reichenbach 1956, Piaget 1946); it seems to be the case that time can be much better understood and experienced through notions of causality. In fact, time is frequently associated with causal relationships, sometimes by way of the mere fact that time has passed (Evans 2004). Overwhelmingly, the world is viewed as a complex network of causes and effects, rather than a series of unconnected, random events (Carston 2002). Causality determines the possible order of events: if event A has caused event B, A must have happened before (or overlapping with the beginning of) B. Other possible relationships between A and B could be that A is a precondition (if not a cause) for B, which also entails that A happens before B; and that B is a consequence of A. Such relationships are frequently expressed in language, sometimes overtly, sometimes by presupposition and implication. They will be dealt with in some detail in Chapters 3 and 4. Entities in space also interrelate with each other. As has been recognised for some time and is presently subject to intensive research (e.g., Carlson and van der Zee, eds, 2005; Grady 1997; Tyler and Evans 2003, Coventry and Garrod 2004), objects are frequently experienced in certain specific correlations that can be characterised as functional. Thus, the functional concept of CONTAINMENT does not only imply a certain spatial relationship but also a functional one in which one object CONTROLS the other, such as a pear in a bowl, where the pear may even be located on top of other pears instead of inside the bowl’s shape (Garrod et al. 1999).

30

Time and space

2.3.4. Summing up: Conceptual differences In spite of their kinship, the domains of time and space exhibit a number of crucial conceptual differences. Time has one oriented dimension, while space has three, all of which are not conceived of as inherently oriented. Space is perceptually accessible, time is not; the same is true for the entities in the domains. Entities in time are frequently conceived of as related through notions of causality, while entities in space may be functionally related. All of these conceptual differences may well have some impact on the use of spatial and temporal expressions in natural discourse. 2.4.

Linguistic means of expressing time and space

As Miller and Johnson-Laird (1976: 411) note, the expression of spatial shapes and relations is optional in English, while the expression of temporal relations is obligatory since every sentence has to carry a tense marker. Obviously, therefore, the linguistic means and requirements for the two domains differ considerably. While temporal information is encoded in some way in every utterance, it has been claimed that the most important features of spatial information can be captured by examining a limited repertory of spatial prepositions (Landau and Jackendoff 1993). For temporal phenomena, language provides a variety of different means to express temporal relations (see, e.g., Vater 1994), not all of which are mirrored in the domain of space. However, not all of the linguistic means of expressing spatial relations are mirrored in the domain of time either, as will shortly be seen. In the following, I will provide a brief overview of the linguistic surface means of expressing relations of time and space. Most of them will not be pursued further since they represent research areas in their own right, and many have already been treated extensively in the literature. Those that will be relevant in subsequent chapters will then be dealt with in some more detail. The following overview contrasts the linguistic means for expressing temporal versus spatial relations. 2.4.1. Prepositions, adjectives, and other related grammatical categories Prepositions can serve to express both temporal and spatial relations. As stated above they are often seen as the standard means for spatial phenom-

Linguistic means of expressing time and space

31

ena. But prepositions, sometimes identical ones, are equally well suited to capture temporal relationships, especially those pertaining to the dimensions of time and space used in localising entities with respect to other entities. Such expressions are the main focus of attention in the present work. However, the terms that express spatial relations do not always appear as a preposition. Some relations are expressed by preposition groups (complex prepositions) rather than simple forms, and some related terms appear as adjectives, adverbs, or even nouns as well (Levinson 2003). Wunderlich and Herweg (1991), for instance, analyse the expression of spatial relations under the heading of ‘Lokale und Direktionale’ (locals and directionals), identifying nouns, prepositions, adverbs and particles as possible syntactic categories. Furthermore, English and German differ in the word classes associated with certain related expressions, as illustrated by the examples given in Table 2.1 below. Thus, even these two closely related IndoEuropean languages already exhibit a wide range of syntactic options. Cross-linguistic research would naturally reflect a still greater variability. Because of the high degree of semantic uniformity between the various syntactic categories (Miller and Johnson-Laird 1976: 412), it is for many purposes a good strategy to refer to the families of expressions with similar morphological appearance by a term that is neutral with respect to grammatical class, such as spatial and temporal marker, discourse marker, or simply term or expression. However, using preposition as a cover term is problematic as it suggests a specific grammatical subclass. In spite of the obvious parallels between spatial and temporal relational markers there can be no doubt that the range of syntactic variety in this broad (semantically defined) category is much greater in the spatial domain than in the temporal. For instance, while the spatial expression left can appear as adjective, adverb, noun, etc., the temporal term after is typically either a preposition or a conjunction (see Section 3.3 for more details). Spatial relational markers cannot appear as conjunctions (see next subsection), but their syntactic variety as markers is generally broader. Furthermore, clearly the repertory of spatial markers (left, right, above, below, front, back…) is greater than that of temporal markers, since there are far more possible relations with regard to space than with time.

32

Time and space

Table 2.1. Diversity of syntactic forms in English and German

English “The ball is in front of the house” (complex preposition) “The ball is at the front of the house” (noun in prepositional phrase) “The ball is on the left” (noun in prepositional phrase) “The ball is left of the house” (adverb modifying prep. phrase) “The ball is to the left of the house” (noun in prep. phrase modifying another prep. phrase) “The ball is behind the house” (simple preposition) “the left ball” (adjective) “the ball on the left” (noun in prepositional phrase)

German “Der Ball ist vor dem Haus” (simple preposition) “Der Ball ist bei/an der Vorderseite des Hauses” (noun in prepositional phrase) “Der Ball ist links” (adverb) “Der Ball ist links vom Haus” (adverb modifying prep. phrase) “Der Ball ist links des Hauses” (simple preposition; archaic) “Der Ball ist hinter dem Haus” (simple preposition) “der linke Ball” (adjective) “der Ball zur Linken” (noun in prep. phrase; marked4)

Differences concerning grammatical categories may reflect underlying conceptual differences. Eschenbach (2005) suggests that German spatial adjectives (as in der linke Karton – “the left box”) can only be used in a context where there are several objects of the same class present. Therefore the adjective is a suitable means to identify one object out of several possible candidates, for instance, when a “Which” question is asked. Other syntactic forms can also be used in such a context (der Karton links, the box on the left), but there is one crucial difference here: Such forms can easily be transformed (by adding a relational process, such as is) into a full sentence, such as those used in “Where” contexts: Der Karton ist links, the box is on the left. This is not possible with adjectives, which seem to be restricted to identifying (rather than localising) contexts (see Section 5.3.1). Interestingly, the temporal dimensional terms before and after cannot appear as adjectives at all. Instead, early can be used, for example, to contrast the 4.

The expression zur Linken is restricted in application to a limited range of contexts and may also be subject to dialectal variation as far as usage is concerned. At least by some speakers of German it is perceived as archaic.

Linguistic means of expressing time and space

33

early event with the late event. But here, the adjectives do not correspond morphologically to the dimensional terms used for indicating temporal relations (before and after),5 as is the case in spatial markers like left in the left box and on the left. This linguistic difference between spatial and temporal dimensional terms may well reflect the cognitive difference with respect to availability of entities (for example, on the grounds of their being co-present in a scenario) to be contrasted from each other (cf. Section 2.3.2). Note, in addition, that both spatial and temporal adjectives generally seem to be pragmatically and syntactically restrictive (for example, left and right cannot appear as comparatives), in contrast to other kinds of adjectives describing properties of objects (such as size as in large or material as in hard) rather than situational features. 2.4.2. Conjunctions Conjunctions relate a clause to a preceding one in the same clause complex. While there is a considerable variety of temporal conjunctions, such as while, when, after, before, when, since, until, etc., there are so few spatial conjunctions (as far as, where, wherever) (cf. Halliday 1994: 237) that they are often neglected as a category. Spatial dimensional terms cannot be used as conjunctions in the same way as temporal dimensional terms can, as the following examples show: (2.5) (2.6) (2.7) (2.8)

Jack arrived before Jill. Jack arrived before Jill arrived. Jack stands behind Jill. *Jack stands behind Jill stands.

In (2.5), before is used as a preposition; (2.6) is nearly identical except that before appears in the syntactic form of a conjunction, combining two different full clauses. In contrast, (2.7) cannot be transformed to (2.8); the spatial dimensional term cannot introduce a clause. Notably, the spatial conjunctions that occur at all are only capable of expressing (different kinds of) spatial identity, while temporal conjunctions cover a greater range of semantic variety, such as precedence, simultaneity, etc. While this may seem surprising considering the above observation that space offers a far greater variety of possible relations than time does, a pos5.

See Heinämäki (1974: 39ff.) for a detailed comparison of before/after vs. earlier than/later than, indicating crucial differences in applicability.

34

Time and space

sible explanation is the following. The typical way of referring to an object is by using a noun, while events (or processes) are usually referred to as verbs. But while there are usually no difficulties in turning verbs into nouns by nominalisation, the opposite is rare: While it is sometimes possible to use a noun like butter in verbal guise as a short form for the process of putting butter on bread (as in buttering the toast), the objects themselves cannot be expressed in verbal form. Conjunctions are used to connect clauses, in which events, not objects, are expressed by verbs. This may be a reason for the fact that spatial conjunctions are rare; there is less need to connect events spatially. Typically, objects are related spatially, and events are related temporally. In contrast, in direct correspondence to the fact that events can be expressed by verbs as well as nouns, temporal relational expressions can occur with clauses (comprising a process denoting an event) as well as with noun phrases (as long as the nouns refer to nominalised events, or else, specific metric time spans).6 In English, the surface form remains the same; with clauses, before and after occur as conjunctions, while with nouns, they are prepositions. In German, the prepositional forms (vor, nach) are distinct from (though morphologically related to) the conjunctive ones (bevor, nachdem). 2.4.3. Verbs, tense, and aspect The entities in the different domains are typically represented by different linguistic means: Events (or processes) are typically expressed by verbs, objects by nouns. This observation resembles the practice of calling the verb’s ideational contribution to clauses ‘processes’ (Halliday 1994). However, in spite of this general tendency for verbs to express temporal processes, a number of verbs are clearly spatial in nature, such as surround or contain. Apart from the semantic spatiotemporal information conveyed by the verb, verbs (in contrast to other grammatical classes) also reflect temporal relations in any instantiation of occurrence. Such information is traditionally divided into the categories of tense and aspect. As Miller and Johnson-Laird (1976: 412) note, tense is “the major grammatical mechanism for expressing temporal relations” to which other kinds of temporal informa6.

Furthermore, they can sometimes relate to non-temporal sequences, as in “The letter A comes before B”, and in some other derived usages. But at this point of the argument I am concerned with temporal interpretations.

Linguistic means of expressing time and space

35

tion must be compatible. In broad terms, the grammatical tense of a verb is used to express the time (past, present, future) of the proposition in relation to the speech event. As Traugott (1978: 374) observes: “Since the moment of speaking shifts reference, (…) tense is a deictic structure.” As such, tense can express relations of proximity and distance. Such relations can also be expressed in other ways that are equivalent in both domains, for instance by using expressions like this and that, now and then, here and there. But the capability of verbs to express deictic relationships in their tense system does not have any equivalent in spatial language. This very general description only touches the intricacies of the tense systems of English and German on the surface. For instance, the grammatical tense of a verb does not always coincide with the actual time the verb refers to: the simple present tense in both English and German can refer to events of the future, and sometimes also of the past. Furthermore, various combinations of tenses yield complex temporal relationships that will not be dealt with in detail in the present work (but see, for instance, Halliday 1994 for a detailed account). Although the notion of aspect is much disputed in linguistics,7 it should be noted that a distinction is sometimes drawn between representations of completed action (‘perfective aspect’, sometimes also called ‘perfective tense’) versus action in progress (‘progressive’ aspect or tense). In English, the perfective aspect is regularly expressed by have+en (as in I have written), while the progressive takes the +ing form as verb endings (as in I am writing). No equally simple distinction can be made for German (Krause 2002). This phenomenon, just as the tense system in general, is not mirrored directly in the spatial domain. 2.4.4. Discourse relations / Rhetorical structure In a coherent text, all clauses are related via their rhetorical structure. That is, they are not simply listed as singular entities in random order, but belong together in certain structured ways that have been subject to extensive research, for instance, in the areas of text linguistics and discourse analysis. Often, the semantic relations between clauses are marked overtly by lin7.

Halliday (1994: 241), for instance, uses the term for a different distinction occurring in the secondary verbal group, roughly to be sketched as that between realis (or Imperfective, as in “likes painting”) vs. irrealis (or Perfective, as in “would like to paint”).

36

Time and space

guistic means (sometimes called discourse markers) such as conjunctions (see above). But this is not necessarily the case. A very common way of expressing temporal relationships – especially in narrative discourse – is to simply add one situation (expressed linguistically by a clause) to another (a new clause) in the order in which they occur. This phenomenon has been formalised, for instance, by Lascarides and Asher (1993), who call the underlying discourse relationship NARRATION. In the large variety of approaches to rhetorical structure, using different terminology and different categorisations of possible relationships between sentences, the option of a basic temporal (sequential) relation has been recognised by most authors (e.g., Halliday and Hasan 1976, Martin 1992, Bateman and Rondhuis 1997); however, equivalent spatial relationships have not been identified. Discourse markers are not restricted to usage only in cases where their prototypical sense is conveyed. For instance, it has been noted (e.g., Lascarides and Oberlander 1993, Blühdorn 2002) that the prototypical temporal discourse markers after and nachdem are capable of conveying causal relationships: (2.9)

They were pacified after Major launched a charm offensive. (Lascarides and Oberlander 1993: 260)

Thus, overt discourse markers can help the analyst identify the underlying discourse relation, but they are not regarded as restricting the range of possible relationships between clauses to the specific semantics contributed by the conjunction itself. For example, in Segmented Discourse Representation Theory (SDRT, Lascarides and Asher 1993: 438f.; Asher and Lascarides 2003), linguistic realisations are considered, but are not essential for the analysis. In fact, usually implicit relations are assumed, but the analyses are supported by linguistic surface evidence such as the use of tense and constraints on anaphora resolution. An important underlying principle is that of Maximising Discourse Coherence, according to which a maximum of underlying discourse relations is assumed – independent of explicit cues in the text – whenever such an interpretation is consistent with compositional semantics, constraints on anaphora resolution, and so on. In a similar vein, Sperber and Wilson’s (1986) “Relevance Theory” incorporates principles that point to a maximum of inferences to be drawn from a given discourse, e.g., the Cognitive Principle of Relevance: Since a maximisation of relevance is central to human cognition, “the hearer should look for an interpretation which is relevant enough to be worth his attention.” (Wilson and Sperber

Conclusion

37

1998: 10). Generally, it is by now widely acknowledged that much freedom and flexibility is involved in interpretation; misconstruing the interrelation intended by the speaker may be a systematic cause for misunderstanding. Often, however, contextual clues resolve the issue. 2.5.

Conclusion

In this chapter, I have reviewed insights and concepts with regard to the domains of space and time. I have examined the widespread hypothesis of a metaphorical relationship between human spatial and temporal concepts, and reached the conclusion that although metaphors – specifically, a “Time is Space” metaphor – are pervasive in language, this is not sufficient to explain all human representations of the temporal domain. I have then pointed to similarities and differences between conceptualisations of time and space, and extracted relevant issues that have an impact on language. Finally, I have demonstrated how spatial and temporal relations can be expressed through language, and pointed to a number of crucial differences in the syntactic repertory between the domains, which already reflect the conceptual differences in systematic ways. Such effects are mostly neglected by earlier accounts, which typically focus on either temporal expressions in whatever form, without much (semantic) differentiation between different syntactic occurrences (e.g., Clark 1973, Tyler and Evans 2003), or on conjunctions connecting clauses – maybe concentrating on specific relations such as temporal ones – disregarding other syntactic forms (e.g., Lascarides and Asher 1993). Haspelmath (1997), for instance, restricts his cross-linguistic analysis explicitly to modifiers of noun phrases. This a priori decision is a reason for neglecting the crucial fact that temporal connectives have no equivalent in the spatial domain. If temporal prepositions derive metaphorically from spatial terms, as Haspelmath claims, temporal conjunctions must derive from their prepositional counterparts, implying that temporal prepositions must be conceptually as well as historically prior to temporal conjunctions. This has, in fact, been claimed (rather casually) by Clark (1973), but is not pursued by Haspelmath. However, this question is out of the scope of the present analysis which does not make any claims concerning the primacy of either spatial or temporal terms, or any kinds of syntactic form.

Chapter 3 Temporal dimensions: Linguistic repertory and issues of application

Humans are nearly incapable of representing arbitrary temporal information; they deal with time systematically, based on meaningful relationships between events. Michon (1988: 175) writes: Temporal information cannot be separated in a meaningful way from the structure of events as such. Events (meaningful transitions from one definite state of the world to the next) and episodes (meaningful series of events) carry an intrinsic temporal structure. This structure imposes constraints on the possible representations of time. Meaningless sequences of stimuli – such as they regularly occur in the psychological laboratory, but hardly anywhere else in the known universe – do not qualify as events or episodes, and consequently they are unable to impose constraints on the ways people will represent time. If a series of events has no inherent structure that can be comprehended by the observer, and moreover, if the observer does not succeed in imposing some invented regularity on the input, no temporal representation of that series of events will ensue.

This idea, formulated in a general way, can be straightforwardly applied to linguistic representations of temporal information. Language can be viewed as a central means of capturing and communicating abstract features of time. In this respect, the linguistic means available to speakers, as well as the way language is used in natural communication, are interesting as a research field. They reflect speakers’ underlying conceptions of the relations between events, which are generally not purely temporal but are also perceived as connected in some more or less direct way. After two sections addressing basic aspects concerned with temporal representation in language, this chapter will deal with features and concepts of time reflected by temporal dimensional terms. 3.1.

Language and temporal structure

How do speakers represent the natural order of events in linguistic representations? A natural assumption, which has received much attention in the

Language and temporal structure 39

literature (e.g., Levelt 1989) and is supported by a number of observations with respect to the usage of language, is that events are preferentially represented in their natural order, corresponding to Grice’s maxim ‘Be orderly’ (Grice 1975). This ‘default’ interpretation can be overridden by explicit information, such as that provided by temporal adverbials, and further grammatical aspects such as aspectual classes (Dowty 1986). Also, common sense or world knowledge can lead to other interpretations, for example, a clause may provide an explanation for, or an elaboration of, a previous clause (Asher and Lascarides 2003). Nevertheless, there seems to be a certain preference for an “orderly” linguistic representation of events. Because of this principle, it does not seem to be necessary to provide explicit information about the ordering of events in time, as long as they are represented in their natural order. Therefore, the usage of before and after should be best motivated in cases where the linguistic representation for some reason departs from the natural order of events. But there is no indication in the literature that these terms are preferentially used with reversed event order; in fact, the opposite seems to be the case, in accord with the general tendency of linguistically presenting events in their original order. Therefore, there must be some additional reasons for the usage of the terms, apart from providing information about the order of events. One reason may be that linguistic representations of temporal events are more complex than they seem, in spite of the fact that the onedimensionality of time is mirrored by the linearity of language. In fact, what language represents is not abstract (linear) time, but rather, multidimensional experiences in time (Habel and Tappe 1999). Although many events can happen simultaneously, language (unlike, for example, graphical representations) can only represent them in linear order. Furthermore, Habel and Tappe point out that linear descriptions of complex events involve segmentation processes resulting in hierarchical structures. Thus, speakers represent certain events or subevents and ignore others; usually events are not described at a very fine level of detail, but only concerning relevant aspects of the situation. These insights cast doubt on the assumption of a straightforward one-to-one correspondence between time and language, and furthermore, they underscore the necessity to identify the reasons why speakers pick out particular temporal aspects that they express explicitly.

40 Temporal dimensions: Linguistic repertory and issues of application 3.2.

Figure and Ground in temporal language

According to Talmy (2000), language usually represents spatial as well as temporal relations in a way that one of the entities is conceptualised as the Figure and the other as Ground. In the domain of time, these notions are related to the notions of assertion and presupposition. Talmy recognises a number of principles that promise to hold for a broad range of languages, with respect to the assignment of Figure and Ground to the events represented. For example, the Sequence Principle accounts for cases in which a temporal relationship of succession is explicitly represented (Talmy 2000: 327): The unmarked (or only possible) linguistic expression for any particular relation between two events in temporal sequence treats the earlier event as reference point, or Ground, and the later event as requiring referencing – that is, as the Figure. Where the complete syntactic form is a full complex sentence, the two events are in the subordinate and the main clause, respectively.

Thus, the unmarked case is that the earlier event is linguistically treated as Given and assigned a backgrounded syntactic form, while the later event is presented as based on the known event, via their temporal relation to each other. Similarly, for linguistically represented cases of temporal inclusion, another principle states that (Talmy 2000: 328): A larger, temporally containing event acts as Ground (in the subordinate clause) with respect to a contained event as Figure (in the main clause).

An example is given in Talmy (2000: 379) as follows. The usual case is exemplified by (3.1), which, according to Talmy, represents a temporal relationship that is not easy to reverse, as is attempted in (3.2). (3.1) (3.2) (3.3)

He had two affairs during his marriage. *He was married through-a-period-containing his having two affairs. He was married when he had his two affairs.

However, in my view example (3.3) counters this hypothesis: here, the shorter event (in this case, a summarised event consisting of two items) is presented as Given, or Ground, in a subordinate clause, while the Figure event is clearly the containing one, presented in the main clause. The information conveyed in (3.3) is that the already known event (of his having two affairs) happened in a period in which he was married, which may not have been known to the listener.

Figure and Ground in temporal language 41

Furthermore, events can be related in that one event is expected but does not occur, while another substitutes it. Such a relationship is regularly expressed, according to Talmy (2000: 329), in that the nonoccurring event appears as Ground, and the unexpected one as Figure, as in example (3.4). (3.4)

He’s playing instead of / rather than working.

These principles are explained in terms of the cognitive effects involved. Thus, earlier events are particularly good candidates for serving as the Ground (or basis) for the representation of later events because they are already in memory when the later events occur. In other words, it is a general cognitive process to assess newly occurring events on the basis of already established cognitive reference points that build on earlier events. In a similar vein, Talmy (2000: 328) proposes a general principle for causally related events: The unmarked (or only possible) linguistic expression for a causal relation between two events treats the causing event as Ground and the resulting event as Figure. Where the complete syntactic form is a full complex sentence, the two events are in the subordinate and the main clause, respectively.

This may chiefly be the case because causing events necessarily occur earlier than the caused ones, and according to the Sequence Principle, the earlier event in the unmarked case serves as Ground in the subordinate clause. Therefore, in most linguistic constructions expressing cause, the result is foregrounded rather than the cause. This is the case, for example, in clauses introduced by the explicit causal marker because as in (3.5), where the cause is presented in a subordinated clause and serves as Ground. (3.5)

He left because he was tired.

It should be noted, however, that causes can indeed appear in a main clause at least in English. In that case, however, the result is presented in a coordinated, rather than subordinated, clause, as with therefore, as a result, etc. The case seems to be similar in German, where the coordinating conjunctions darum and deshalb are used regularly for presenting results: (3.6)

Ich bin müde, darum gehe ich nach Hause. ‘I am tired, therefore I go home.’

A related principle dealing with preconditions rather than causes is called the Contingency principle (Talmy 2000: 329):

42 Temporal dimensions: Linguistic repertory and issues of application An event that is necessary for or determinative of a second event acts as Ground (in the subordinate clause) with respect to the second event that is contingent or dependent on it, which acts as Figure (in the main clause).

In (3.7), for example (Talmy 2000: 326), sleeping is a precondition for dreaming: (3.7)

He dreamt during his sleep.

At the same time, the temporal marker during expresses an inclusion relationship. This example shows how several relationships can be simultaneously active in one linguistic representation, even if only a temporal relationship is given explicitly. Similarly, causal inferences can be drawn from the mere juxtaposition of sentences or clauses which may or may not be connected by explicit temporal markers. This observation will be taken up again in Section 3.5 below. 3.3.

Syntactic range of temporal dimensional terms

Before and after often appear as markers of discourse relations (see Section 2.4.4); this discourse function is usually only analysed with respect to their appearance as conjunctions. However, following Talmy’s insights concerning Figure and Ground across syntactic constructions, there are reasons to assume shared features between different appearances of related terms. This section addresses the range of syntactic variability in temporal dimensional terms along with a number of features associated with the particular forms in German and English. As a preposition the temporal markers appear together with a noun or a nominal group: before/after noon; before/after the meal; vor/nach dem Essen. In English, there is the related option of a conjunctive preposition (Halliday 1994: 238f.), in which case the markers take a non-finite clause: 1 (3.8)

…, before / after assessing the damage.

As a conjunction they take a finite clause, which may be either telic or atelic. Here, the aspectual form is overwhelmingly the perfective (expressing completed action), as in the following examples: (3.9) 1.

after / before she had written a letter, … The examples in this and the following sections are all mine if not marked otherwise.

Syntactic range of temporal dimensional terms 43 (3.10)

nachdem / bevor sie den Brief geschrieben hatte,…

The imperfective aspect seems to be infrequent though acceptable, as in: (3.11)

The firemen arrived just a few minutes before the fire was getting fierce.

As a sentence adverbial, they take no further elements, and thus do not temporally relate two explicitly mentioned events as the other constructions do: (3.12) (3.13)

We arrived shortly after. The firemen arrived just a few minutes before.

This usage points to a period in time prior to or following another time that must be inferred from the context. This could be viewed as an abbreviated form of one of the other constructs, as shown by the following expanded version (see also example (3.11) above): (3.14)

We arrived shortly after Tom.

In German, this notion can be translated using danach and davor, in which case the deictic marker da points to the existence of a contextually inferable relatum: (3.15) (3.16)

Wir kamen kurz danach / nach Tom an. Die Feuerwehrleute kamen einige Minuten davor an.

Schilder (2002) notes that the absence of an explicit temporal relatum in clauses containing davor leads to a higher potential for ambiguity, and accordingly to a higher degree of uncertainty with respect to the temporal relation involved. Additionally, the English adverbial before often appears with fundamentally different interpretations, as exemplified by (3.17), (3.19), and (3.21). Crucially, they do not refer to a contextually accessible event that is omitted syntactically. These usages are not mirrored in German by the term (be)vor, but are translated in a different way, e.g. using noch nie zuvor as in (3.18), schon mal as in (3.20), or früher as in (3.22). (3.17) (3.18) (3.19) (3.20) (3.21) (3.22)

I have never done this before. Ich habe dies noch nie zuvor gemacht. I have done this before. Ich habe dies schon mal gemacht. He knew her from before. Er kannte sie von früher.

44 Temporal dimensions: Linguistic repertory and issues of application This cross-linguistic fact serves as a further indication that these variants of before differ semantically from the other options, although they are clearly related in that they, as well, denote temporal anteriority. However, to my knowledge these differences have not been systematically accounted for in the literature. Accounts of before regularly deal with the other syntactic forms only (preferably the appearance as a conjunction). The issue will be taken up again in the empirical investigation (Chapter 4). Altogether, the literature does not offer much information about the applicability, speakers’ preferences, or differences in the associated semantics of any of these syntactic forms. Often, they are treated as equivalent (e.g., Petkov 1979). Jekat-Rommel (1994) postulates a difference between temporal prepositions and conjunctions: while prepositions can be applied for any kind of temporal extension, temporal conjunctions may only be used to refer to extended events. With regard to the semantically related area of causal connectives versus prepositions (in Dutch), Degand (2000) investigates systematic differences that may serve as reasons for choosing one form rather than another. According to her analysis, different forms express different aspects of the (extra-linguistic) causative situation. In her view, prepositions are not markers of discourse relations, but involve a high lexical and grammatical density in expressing information that is treated as Given. In contrast, causal clauses are more elaborate and explicit. Unfortunately, so far there are apparently no similar studies available for other languages or other discourse relations (other than causal). However, Degand’s results may be applicable for German and English temporal connectives versus prepositions. In subsequent sections, I will address some features of temporal dimensional terms that were typically identified on the basis of one particular syntactic form (usually conjunctions). Along with presenting these features, I will take a further step in the direction of comparing different syntactic forms by examining whether these findings are (intuitively) also applicable to other constructions. Although this approach does not tell us anything about speakers’ actual preferences for packaging specific kinds of information into specific kinds of syntactic structures, it can at least highlight apparent limitations for applying structures in specific situations.

Details of the temporal relationship 45

3.4.

Details of the temporal relationship

The most uncontroversial aspect of the semantics of temporal dimensional terms, both in German and in English, is that they express a relation in time, i.e., either temporal anteriority or posteriority. This section deals with the ways in which additional temporal factors come into play in the interpretation of these terms. Intuitively, after and before (as well as bevor and nachdem) could be regarded as simple converses. However, already from a temporal perspective this does not always work, as has been recognised time and again since Anscombe (1964). This argument is linked up with other phenomena such as temporal overlap, time points vs. intervals, and (non–)veridicality (cf. Beaver and Condoravdi, 2003). Furthermore, Klein (1994: 156) notes that expressions such as after the war do not have a boundary in the future, and conversely, before the war none in the past. Even if these time spans do have boundaries in the real world, which may or may not be known to the speaker and hearer, they are not provided by the lexical content. Temporal dimensional terms often do not simply express a general temporal precedence relation, but a certain association of sequentiality or proximity comes into play in many contexts. Then, a contextually dependent time frame is necessary to account for the semantics of such terms. Herweg (1991: 64)2 claims that previous and subsequent proximal times can be identified for each event in question, determined by coherent events, processes, and states. These are introduced by the event in question or lead to it. The context plays a major role in the identification of the proximal area of an event in a concrete situation. Thus, a proximal time frame induced by a falling event can in one context be a few minutes during which the person gets up again, as in: (3.23)

After he fell, he got up again.

In another context, it could span a few weeks during which the person recovers from breaking a leg during the falling event, as represented by: (3.24)

After he fell, he was in hospital for three weeks.

Furthermore, the event in the main clause may also be conceptualised as occurring not within the proximal time frame following the event in the 2.

Herweg’s analysis is based on the German terms nachdem and bevor; he notes however that related terms such as German nach and English after are subject to similar phenomena.

46 Temporal dimensions: Linguistic repertory and issues of application temporal clause as in the previous examples, but already within the time frame of the event itself (Herweg 1991: 75): (3.25)

Bevor Peter Hut und Mantel abgelegt hatte, klingelte das Telefon. ‘Before Peter had put down his hat and coat, the telephone rang.’

In other words, the two events may overlap temporally. That temporal overlap is possible with the temporal dimensional terms, which are usually viewed as typical markers of temporal succession, is also claimed by Schilder (2002), who provides the following examples: (3.26)

(3.27)

Es regnete, bevor ich aus dem Haus ging. Es regnete sogar noch, als ich draußen vor der Tür stand. ‘It rained before I left the house. It still rained even when I stood outside of the door.’ Nachdem Peter Maria unterwegs getroffen hatte, fuhr er mit dem Fahrrad nach Hause. ‘After Peter had met Maria on the way, he cycled home.’

In both examples, the situation described in the main clause holds longer than could be expected by a sequential interpretation: in (3.26), the rain continues after leaving the house; and in (3.27), the lexical item “unterwegs” (on the way) indicates that Peter had been cycling home even before he met Maria. Also in the corpus work presented in the next chapter, we will encounter several examples of temporal overlap, showing that the widespread interpretation of the temporal dimensional terms as indicating succession may be too narrow. A related notion playing a role in inferring proximal time frames is that of granularity (Habel et al. 1993). The interpretation of terms directly expressing proximity in time or space, such as soon or near, depends on the level of granularity involved. Similarly, the sentence (3.28)

After lunch, they continued their work.

will be understood on a different level of granularity than (3.29)

After college, she started as a teacher.

In the former case, the proximal time frame of after is determined by the contextual cue lunch which suggests a scale of minutes or, at best, hours. In contrast, in the latter case the inferred proximal time frame involves weeks or months. Apart from the contribution of the context to assessing the extent of the time frame, there seems to be a tendential correlation between the extension of a proximal time frame and the extension of the event itself,

Causal interpretations of temporal connectives 47

i.e. – in general – the longer the duration of an event, the greater the extent of the proximal time frame. Thus, if the described event contains a period of several years, as in going to college, the proximal time frame is likely to encompass at least a few weeks in any context. However, the converse inference is not necessarily the case: in (3.24) above, the falling event can induce an extended time frame because of its potential for causing injury. Furthermore, events may also happen immediately after an extended event even if the proximal time frame encompasses a longer period. These phenomena seem to be largely independent of the particular syntactic construction of the temporal marker. 3.5.

Causal interpretations of temporal connectives

The events described in two adjacent clauses are often conceptualised as causally related even in the absence of explicit causal markers. This phenomenon is explained in part by the intricate relationship between the two abstract concepts of time and causality (cf. Section 2.3.3). However, this does not imply that causality is ubiquitous, which would mean that a causal relationship can be inferred between all events related in sequence in a text. The present section deals with the ways in which causal interpretations can be inferred, and the limitations of such inference processes. Causality is often (e.g., Heinämäki 1974) associated with usage of after: (3.30)

After he stumbled over a stone, he fell.

Here, the inference that the falling event was caused by the stumbling event is a natural conclusion to draw, even though it can easily be cancelled, for instance, by adding: (3.31)

But he did not fall because he had stumbled.

Indeed, the naturalness of the adversative marker but could not be explained if the conclusion were not obvious to the hearer (cf. Asher and Lascarides 2003: 181). Intuitively, the causal connection can also be induced without the temporal marker. But while it is true that semantic relationships can be associated in text comprehension even without explicit markers, this does not imply that the markers themselves are neutral in this respect. The same inference would indeed be weakened by using before instead of after: (3.32)

He stumbled over a stone before he fell.

48 Temporal dimensions: Linguistic repertory and issues of application Here, a natural interpretation is that the falling event is in temporal succession yet causally independent of the stumbling event. Schilder (2001: 93) describes the difference as follows: The causal relation does not have to be present for every after sentence. However, if such a relation is inferable due to our world knowledge, the after sentence invites the reader to draw this inference. The usage of a before sentence, on the other hand, prevents the reader from doing this.

However, according to the intuitions of some speakers, it seems to be possible to assume a – perhaps less obvious – causal relationship as well in cases such as (3.32). Generally, causality as a concept is non-absolute (Mackie 1965) and subject to interpretation (Davidson 1967); thus, in cases where world knowledge does not directly contradict a causal relationship, it may well be inferred. In (3.32), it seems that world knowledge supports a causal relationship, although it seems unclear whether it is intended by the speaker. Since this is so, tests for causality (e.g., via replacement by a causal marker) are not informative with respect to the speaker’s intentions: for example, while (3.32) seems to be related to the explicit causal relation expressed in (3.33), there is no objective way of determining whether the same causal relationship could indeed be inferred by (3.32). Possibly, in some cases a causal relationship is not intended by the speaker but the hearer may infer it nonetheless (Rossdeutscher and Stutterheim 2004). (3.33)

Because he stumbled over a stone, he fell.

One motivation for a possible difference between before and after with respect to the inferability of an intended causal relation comes from the linguistic structure. When two causally related events are linguistically connected using after, the causing event always appears in the backgrounded and presupposed temporal clause, while the result is focused upon in the main clause. If before were used to express causal connections, the cause rather than the result would have to be foregrounded, which is pragmatically unusual, according to Talmy’s (2000) principles outlined in Section 3.2 above. Discourse markers used for overtly expressing causality, such as because, since, as, etc. generally package the cause, not the result, into a backgrounded clause, or else the clauses are co-ordinated rather than sub-ordinated, as with therefore. With regard to the German terms, Blühdorn (2002) claims that not only nachdem but also bevor is capable of expressing causality. He gives the following example:

Causal interpretations of temporal connectives 49 (3.34)

Bevor er das Auto in den Graben lenkte, hatte er ausgiebig mit seinen Freunden gezecht.

An English equivalent could be the following: (3.35)

Before he drove the car into the ditch he had boozed extensively with his friends.

While an indirect causal relationship can certainly be pragmatically inferred here, it is not directly induced by the temporal connective. The disastrous driving event is not a direct effect of the drinking event: the drinking causes the driver to be drunk, and driving drunk can cause accidents. These causal chains cannot be expressed using bevor: (3.36)

? Bevor er betrunken war, hatte er ausgiebig mit seinen Freunden gezecht. ‘? Before he was drunk, he had boozed extensively with his friends.’

and (3.37)

? Bevor er das Auto in den Graben lenkte, war er betrunken. ? ‘Before he drove the car into the ditch he had been drunk.’

The analysis of naturally occurring instances of bevor (Section 4.3) will shed more light on the question of whether direct causal relationships can be conveyed. For the moment, it will be assumed that after and nachdem have a greater potential for supporting causal inferences than before and bevor. This seems, however, to be a matter of degree. After can also convey a causal interpretation in other syntactic constructions: (3.38) (3.39) (3.40)

After stumbling over a stone, he fell. After the explosion he was numb for a while. He was numb for a while after.

The same effects can be identified in German: (3.41) (3.42) (3.43) (3.44)

Nachdem er über einen Stein gestolpert war, fiel er hin. Nach dem Stolpern über einen Stein fiel er hin. Nach der Explosion war er eine Weile betäubt. Er war danach eine Weile betäubt.

In all these cases, a replacement of the temporal term by a causal one like because or weil does not seem to lead to an obvious difference in interpretation, apart from a higher degree of interpretative flexibility associated with temporal terms than with explicit causal markers.

50 Temporal dimensions: Linguistic repertory and issues of application This observation supports Talmy’s assumptions regarding the similarity of different syntactic constructions with respect to their associated relationships between Figure and Ground, rather than the rigorous approach of attributing underlying semantic relationships to discourse relations between full clauses only, which is often implicitly assumed when addressing such relations solely with respect to the occurrence as conjunctions. Causality may occur on different levels. A related, near-causal relationship that has been recognised uncontroversially to be conveyed by the German and English temporal dimensional terms (e.g., Schilder 2001) is ENABLEMENT (or, in different terms, CONDITIONALITY). Here, the earlier event is a precondition for the later event, as in: (3.45) (3.46) (3.47) (3.48)

She had to unlock the door before she could open it. Sie musste die Tür aufschließen, bevor sie sie öffnen konnte. She could open the door after she had unlocked it. Sie konnte die Tür öffnen, nachdem sie sie aufgeschlossen hatte.

This inference is often supported, as in these examples, by lexical markers such as had to, could, konnte, etc., which explicitly point to the conditional relationship. It is possible with all syntactic variations. Blühdorn (2002) notes that, in some cases, it is open to the recipient’s interpretation whether plain temporal sequence, conditionality, or even causality should be inferred. Furthermore, he observes that the temporal sequential factor loses its impact in conditional and causal interpretations. In a similar vein, Herweg (1991) describes causal (and related) inferences of temporal markers as re-interpretations of the originally temporal semantics of the terms. One different kind of causality conveyed by before as well as after has been described by Schilder (2001: 88) in terms of TERMINATION as follows: The situation expressed by the temporal clause sets an end to the situation described by the main clause.

For instance: (3.49)

Peter waited for Sue for hours before she finally arrived. (Schilder 2001: 88)

This inference may be interpreted as causal in the sense that the event in the temporal clause is responsible for terminating the situation in the main clause. In other words, while before does not seem to convey direct causal inferences between two temporally related events, it is capable of invoking a reversed causal relationship. Like causality, TERMINATION can be identi-

Presuppositions 51

fied with other syntactic constructions and also in German, though not (or not as straightforwardly) as an adverbial, which seems intuitively obvious since the event responsible for terminating the situation in the main clause is not mentioned explicitly in these cases (examples (3.52) and (3.55)). (3.50) (3.51) (3.52) (3.53) (3.54) (3.55)

Peter waited for Sue for hours before her arrival. Peter waited for hours before leaving. * Peter waited for hours before. Peter wartete stundenlang auf Sue, bevor sie endlich eintraf. Peter wartete stundenlang vor dem Aufbrechen / vor dem Aufbruch. ? Peter wartete stundenlang davor.

The same inference can also be drawn for after and nachdem, with the difference that the terminated event is the one in the temporal clause, which is ended by the situation in the main clause: (3.56) (3.57)

After Peter had waited for Sue for hours, she finally arrived. Nachdem Peter stundenlang auf Sue gewartet hatte, traf sie endlich ein.

In all these cases, a reformulation using until or bis (in German) seems to convey the intended meaning equivalently (although this is, of course, a matter of interpretation). Schilder (2001) notices that a TERMINATION relation can also be inferred between a sentence starting with after and the previous discourse: 3 (3.58)

John was looking for his glasses. After he found them, he left. (Schilder 2001: 97)

This inference is equally possible in German: (3.59)

3.6.

John suchte seine Brille. Nachdem er sie gefunden hatte, ging er.

Presuppositions

In her famous dissertation work, Heinämäki (1974) lists a range of peculiarities associated with the syntactic environment of before and after, only a few of which will be considered here. For example, she claims that clauses

3.

Another possible relation proposed by Schilder (2001) between sentenceinitial after and the previous discourse is a sub-part relation, which will not be discussed here in detail, but is taken up in Chapter 4.

52 Temporal dimensions: Linguistic repertory and issues of application starting with these connectives cannot take epistemic modals, as exemplified by:4 (3.60)

*We met John before he must have tampered with the tapes.

Furthermore, according to Heinämäki before and after cannot have an explicit negative: (3.61)

*John left before / after Bill didn’t leave.5

But only after cannot take negative polarity items: (3.62)

John left before / *after anybody else.

Finally, a different ordering of clauses seems to entail different implicatures: (3.63) (3.64)

Jan lived in London after the war was over. After the war was over, Jan lived in London.

Example (3.63) seems to entail that Jan did not live there at other times, while in example (3.64) there is no such entailment. These differences can be explained by thematic and information structure: In the former case, the topic of the discourse (the Given part) is the fact that Jan lived in London; the temporal clause (the New part) explains when this happened. Therefore, it would not be expected that Jan lived there at other times also. In the latter case, in contrast, the discourse accounts for what happened after the war, in which case what happened at other times is irrelevant. Miller and Johnson-Laird (1976: 426) point out that the choice of connectives in logically equivalent sentences, such as (3.65) and (3.66), has an impact on the presuppositions associated with the temporal clauses. (3.65) (3.66)

He finished before he left. He left after he finished.

The authors argue that the main clause expresses an assertion, and the temporal clause a presupposition. Therefore, the sentences lead to different assumptions about the associated discourse context. The temporal clauses are presuppositional in the sense that they remain true if the main proposi4. 5.

Examples (3.60) through (3.64) are Heinämäki’s. Note that the discourse context can provide a situation in which the explicit negative is indeed acceptable, for example, if John left because (after) Bill refused to leave. Thus, Heinämäki’s observations should not be viewed as hard constraints on the employment of temporal dimensional terms.

Presuppositions 53

tion is negated.6 Thus, in (3.65) the leaving event is supposed to be true no matter whether or not the finishing event occurred (before leaving), while in (3.66), the stable (or factual) assumption is the finishing event. The notion of presupposition can partly account for the differences reported by Heinämäki (1974) as well. Example (3.60) is not compatible with a presupposition because an epistemic modal indicates uncertainty as to whether or not the event in the temporal clause occurred. In example (3.61) it is denied that the presupposed event in the temporal clause occurs, which does not make sense (at least not without an appropriate context). In examples (3.63) and (3.64) it is in both cases presupposed that the war was over; however, the preposed construction puts the temporal clause in thematic position, which leads to different entailments. Example (3.62) illustrates a fact discussed in the next section, namely, that before-clauses are not in all cases presuppositional in a strict sense: with a non-veridical interpretation, it is only presupposed that the event described in the temporal clause was previously expected to occur. With regard to the German temporal connectives, Blühdorn (2002) claims that nachdem standardly needs to be anchored in the discourse, since it contains a definiteness component; furthermore, in certain discourse situations the proposition in the temporal clause may also be new, yet still factual. While this description is compatible with the findings concerning the English terms, it additionally introduces a default assumption for nachdem, thereby ascribing a secondary status to nachdem-constructions in which both propositions are discourse-new. Bevor, according to Blühdorn, has the opposite structure: it does not contain a definiteness component and therefore, by default, is not anchored in the discourse. However, it has the potential to increase complexity by discourse anchoring. These different ascriptions of default status – a distinction that I have not encountered in the literature on English – are based on the etymology of the two German terms, which is transparent in the sense that nachdem contains dem as a definite element, which bevor does not. However, it often happens in the literature on related issues that the reasons for determining a discourse anchor are not made explicit. Blühdorn (2002), for example, does not provide a wider discourse context for his examples in which the discourse anchor could be identified; instead, he claims that nachdem introduces the proposi6.

This view is based on the so-called semantic approach to presupposition (cf. Strawson 1950). In the pragmatic approach, “presuppositions are what is taken by the speaker to be the common ground of the participants in the conversation” (Stalnaker 1978: 320).

54 Temporal dimensions: Linguistic repertory and issues of application tion as known and therefore necessarily discourse-anchored. In a different example, he then shows that propositions of a different category introduced by nachdem may indeed be new to the discourse. It is unclear how the difference between the two categories should be determined without analysing the relations to the previous discourse context, or else relying on specific linguistic markers. In Chapter 4, I will present an analysis of how speakers use presuppositional clauses in relation to the previous discourse context and in light of the listener’s knowledge. Lascarides and Oberlander (1993) claim that temporal connectives trigger presuppositions with regard to underlying discourse relations that need to be inferred on the basis of background knowledge. Such relations may well be causal, as indicated above. Following and extending Lascarides and Oberlander’s approach into some more detail, Schilder (2001) accounts for the inferences associated with the usage of before and after in terms of presupposition binding. His proposal contains the interpretation of temporal clauses within DRT (Discourse Representation Theory, Kamp and Reyle 1993), covering a number of sentential and discourse linking relations that are possible with or blocked by the usage of the temporal connectives, such as CAUSAL and PRECONDITIONAL, NON-FACTUAL, SUB-PART, and TERMINATION relations. The main idea is that the variability of discourse relations holding between the clauses connected by the temporal markers, and to the previous discourse, is restricted in several ways. Thus, the existence of the markers in a discourse presupposes the inference of one of a number of possible discourse relations, leading to the formulation of precise conditions for application. This approach will be pursued further in Chapter 4. 3.7.

Non-veridical interpretations

It is well-known that before allows for so-called non-veridical readings (reinforced by negative polarity items such as anyone), as in example (3.67) where it can be assumed that the punching event did not occur (although different interpretations seem always to be available, depending on context). However, after is not capable of supporting such an inference, as example (3.68) illustrates. (3.67) (3.68)

Mary left the party before she punched anyone. *After Mary punched anyone, she left the party.

Non-veridical interpretations 55

Example (3.67) has a direct correspondence in German, so that it can be assumed that the languages do not differ in that respect: (3.69)

Mary verließ die Party, bevor sie irgendjemanden schlug.

However, according to some informants this is not perceived as natural in German; a better version would be to include a modal such as could (no doubt this would also improve (3.67)): (3.70)

Mary verließ die Party, bevor sie irgendjemanden schlagen konnte. ‘Mary left the party before she could punch anyone.’

Non-veridical readings seem to be more problematic with before and vor as a preposition rather than a conjunction. Together with an indefinite article, the non-veridical reading seems to be available at least in English (example (3.71)); however, examples (3.73) and (3.74) seem to suggest that the event in question indeed occurred. Example (3.72) is rather unclear without a wider discourse context. (3.71) (3.72) (3.73) (3.74)

Mary left the party before a confrontation. Mary verließ die Party vor einer Konfrontation. Mary left the party before the punching. Mary verließ die Party vor dem Schlagen.

The usage of adverbials yields the same result. In contrast, conjunctive prepositions, in English, again yield non-veridical readings (enhanced by the possibility of using a negative polarity item): (3.75)

Mary left the party before punching anyone.

While some earlier approaches accounted for this peculiarity of the usage of before by complicating its logical formulation, Miller and Johnson-Laird (1976: 426) propose looking for information in the context that “enables the listener to infer a preventive relation in addition to the temporal relation”. In other words, it can be assumed that the second event would have occurred, i.e., it was expected to occur, but the first prevents it. This observation goes back to Heinämäki (1974: 53f.), who formulates the following requirement: The main clause, A, expresses an event that prevents the before-clause, B, from coming true. A is both a sufficient and a necessary condition to prevent B, that is, to bring about not-B.

Heinämäki further notices that non-factual readings are only possible if the main clause expresses an accomplishment or achievement (cf. Vendler

56 Temporal dimensions: Linguistic repertory and issues of application 1957). This is because only a bounded event, rather than an (unbounded) state or activity, seems capable of preventing another event from happening. This idea can be exemplified as follows: (3.76)

* Mary was tired before punching anyone.

Heinämäki proposes reformulating the sentence in question using would in order to render the interpretation of non-factual before explicit, as in: (3.77) (3.78)

The bomb exploded before hitting the target. The bomb exploded before the time at which it would have hit the target.

However, not all usages of before render the temporal clause either clearly true or clearly not true. A further possibility, which can be regarded as another sub-case of the more general notion of non-veridicality, is that B is possibly true as in the following example (Heinämäki 1974: 62): (3.79)

I left the country before anything happened.

In this case, reformulating the sentence using would yields a clear semantic difference: it invokes the reading that leaving the country prevents an expected event of “happening”, as in the non-factual case. However, this is not necessarily what is intended by (3.79); a more likely interpretation is that the speaker does not know (or does not consider relevant in the current discourse) whether anything happened after leaving. Heinämäki (1974) claims that such non-committal before-clauses require a negative polarity item. She points out that there are not different kinds of before, but rather different kinds of contexts in the two connected clauses (1974: 64): The differences in truth value seem to follow from the different contexts: in the non-factual case, the speaker believes that the before-clause – which he believes not to be the case – would have been true if the main clause had not prevented it. In the non-committal case, the speaker believes that the before-clause was probable, although he does not have evidence as to whether it actually was the case. In the factual cases, the speaker believes the beforeclause to be true. That the conditional, would, is restricted to non-factual before-clauses is natural, since the conditional ordinarily appears in contexts that do not represent the real world, such as counter-factual conditionals.

She reasons (1974: 77) that there cannot be any non-factual or noncommittal after-clauses because

Non-temporal interpretations of temporal connectives 57 the after-clause expresses the earlier state of affairs, the main clause gives the later one. Therefore, it is impossible for the main clause to have any control over what happened earlier.

Neither is there any uncertainty regarding the truth value: all after-clauses are presuppositional (in the “semantic” sense). Therefore, no negative polarity items can occur together with after, as already pointed out in Section 3.6. The same holds for the German equivalent nachdem. 3.8.

Non-temporal interpretations of temporal connectives

Some temporal connectives can also express priority in a non-temporal interpretation, namely, on preference and likelihood scales (e.g., Heinämäki 1974: 66ff., Herweg 1991). These alternative interpretations can be classified as epistemic and deontic, since they reflect either a degree of certainty or a degree of obligation or preferability. Herweg (1991: 75) proposes assuming a non-temporal conceptual interpretation of the abstract relation “қ ”, which is part of the semantic representation of temporal bevor, where it is interpreted as a temporal precedence relation. A re-interpretation as non-temporal occurs, for example, when the temporal reading seems implausible. Some occurrences render a non-temporal interpretation more likely because of background information available to the recipient, as is the case in Blühdorn’s (2002) example for an epistemic reading (which seems problematic in its verbatim English translation): (3.80)

Bevor Schalke Deutscher Meister wird, spielt der FC St. Pauli in der Champions-Liga. ‘Before Schalke becomes German Champion, the FC St. Pauli plays in the Champions League.’

The option of deontic and epistemic readings basically parallels non-spatial usages of spatial terms. It will not be pursued further in the present work, since the focus is on the applicability of spatial and temporal terms with their core meanings. 3.9.

Related terms

In this section, a number of other temporal terms are considered that can be viewed as semantically close to the temporal dimensional terms under

58 Temporal dimensions: Linguistic repertory and issues of application analysis (i.e., before, after and their German equivalents), in order to point to shared features as well as (semantic and pragmatic) differences. 3.9.1. Comparative terms The comparative terms earlier/later (than) and their German counterparts früher/später (als) are obvious candidates for being treated as nearsynonyms of the temporal dimensional terms under analysis. They are similar in that they, as well, denote temporal relationships between two events, (at least) one of which is presupposed. However, their principal sense is one of comparing rather than relating. This has several effects on application. In certain cases, the terms can be exchanged without much apparent difference in interpretation: (3.81) (3.82) (3.83) (3.84)

He finished before he left. He finished earlier than he left. He left after he finished. He left later than he finished.

However, translating into German requires some syntactic restructuring, and still seems awkward, as example (3.86) shows: (3.85) (3.86)

Er wurde fertig, bevor / nachdem er losging. Er wurde früher / später fertig, als (dass) er losging.

It seems that this usage is in German for some reason dispreferred (in favour of using bevor / nachdem). Früher / später would probably be preferred when comparing the agents of events (rather than the events, as in the previous examples): (3.87)

Er wurde früher / später fertig als Peter. ‘He finished earlier / later than Peter.’

Further research would be needed to confirm speakers’ preferences in such cases, and to identify the reasons behind them. Preposing the temporal clause is not possible with these terms in either language: (3.88) (3.89) (3.90) (3.91)

Before he left, he finished. * Earlier than he left, he finished. Bevor er losging, wurde er fertig. * Früher als er losging, wurde er fertig.

Related terms 59

Further syntactic comparisons of the English terms can be found in Heinämäki (1974: 39ff.). She notices, for example, that before and after can be pronominalised, but earlier than and later than cannot: (3.92)

John left before/ after/ *earlier than/ *later than Bill, and George left then, too.

This can be explained by assuming that before and after refer to a definite time interval which is anchored in time by its relation to another time interval, while earlier and later remain indefinite, lacking the association of succession and proximity that has been identified for before and after (cf. Section 3.4 above). Compare examples (3.23) and (3.24) (illustrating effects of granularity in assessing proximity with temporal dimensional terms) with the following: (3.93) (3.94)

He got up again later than he fell. He was in hospital for three weeks later than he fell.

Although these examples are perhaps not ungrammatical, they do not convey the same sense of relatedness as the sentences with after do. Thus, proximal time frames do not seem to play a role in the application of the comparative terms. Heinämäki (1974: 45) further notices that earlier/later can be interpreted in a relative sense, which is lacking for before and after: (3.95)

This year, spring came earlier than/ *before (it came) last year.

This example, again, points to a more definite sense of before in contrast to earlier than. With respect to the other aspects associated with the temporal dimensional terms, it seems that earlier/later are much less rich in their interpretation. For instance, causal inferences are not likely, as the following example shows (cf. example (3.30) above): (3.96)

He fell later than he stumbled over a stone.

This description seems to inform about the relative temporal relationship of two (causally) unrelated events. Disclaiming a causal relationship seems strange in this context (as opposed to (3.31) above): (3.97)

But he did not fall because he had stumbled.

The conceptually related PRECONDITION relation, which is available for both before and after, is also not conveyed by the comparative terms (cf. example (3.45)):

60 Temporal dimensions: Linguistic repertory and issues of application (3.98)

She had to unlock the door earlier than she could open it.

Instead, this sentence seems to imply that (for some separate reason) there was a requirement to do the unlocking earlier than the opening. Also a TERMINATION relation does not seem available for the comparative terms (cf. example (3.49) above): (3.99)

Peter waited for Sue for hours earlier than she finally arrived.

This sentence does not lead to the inference that the waiting ended when Sue arrived (as is the case with before), but that he happened to wait at an earlier time, which ended long before she arrived. Furthermore, the non-veridical readings identified for before and bevor are not conveyed by earlier or früher (compare ex. (3.77)): (3.100) The bomb exploded earlier than hitting the target.

Here, a valid inference seems to be that at least parts of the bomb indeed reach the target, which is not necessarily the case in (3.77). Taken together, these findings indicate that, although earlier than, later than, and their German counterparts seem at first sight quite similar to the temporal dimensional terms, they are in fact applied in a radically different way, for different purposes and with different associations. 3.9.2. When / als Another expression that can come close to the temporal dimensional terms is when (and its German near-equivalent als). Miller and Johnson-Laird (1976: 429) note that when can denote a temporal relationship of succession (rather than the usually associated sense of simultaneity) specifically when both clauses describe momentary events, as the following example shows: (3.101) It exploded when he arrived.

They notice further that accomplishments, being similar to momentary events, also admit a sequential interpretation: (3.102) When John wrecked the car, Bill fixed it.

Possibly, the sequential relationship is simply not in focus here. Substituting after for when does not seem to yield much difference in interpretation: (3.103) After John wrecked the car, Bill fixed it.

Related terms 61

Notice that a causal relationship can be inferred in both cases. The case is similar in German, where both the causal and the temporal implications seem to be rather similar to the English versions in (3.102) and (3.103): (3.104) Bill reparierte das Auto, als / nachdem John es zu Schrott gefahren hatte.

Additionally, Miller and Johnson-Laird (1976: 429) show that when, used together with suitable tenses (e.g., the ‘past perfect’ in the main clause), can also express a reverse temporal succession, similar to before: (3.105) It had just exploded when he arrived

They suggest that [s]uch [occurrences] (…) suggest a relatively short interval between the two events, whereas ‘after’ and ‘before’ suggest nothing about the length of the intervening interval.

However, as described at length in Section 3.4 above, after and before do indeed suggest proximity between the two events involved. Furthermore, since removing just in example (3.105) eliminates the impression of a short temporal distance, this interpretation is obviously due to the contribution of the adverb, not the conjunction. In (3.101), however, the temporal succession indeed seems to be immediate due to the employment of when. In line with the observation that when can only be substituted by before in certain specific cases, i.e., together with the employment of suitable tenses indicating posteriority, it is a natural consequence that the specific interpretations of before described in Section 3.8 above, namely, nonveridicality and epistemic and deontic readings, are not apparent using when. However, when seems to be close to after with momentary events, including the associations of causality as indicated above. 3.9.3. Conveying immediate succession: Then and related terms As already indicated, Miller and Johnson-Laird propose a specifically close temporal relationship for the usage of when. They furthermore suggest that the associated notion of immediate succession is conveyed especially clearly, for example, by the expressions as soon as and as soon after as possible. This general idea is pursued in some more detail by Habel (1994, 1995). In his account, using expressions like immediately afterwards or and next, it is possible to determine the next event in time without specifying the exact temporal relationship between the events. Habel points out that

62 Temporal dimensions: Linguistic repertory and issues of application this notion presupposes the conceptualisation of events, and thereby the underlying time structure, as discrete rather than dense or continuous, which would also be conceivable. Two conceptions are crucial to the notion of an immediate successor in time. First, as Avrahami and Kareev (1994) point out, contextually embedded events are cognitively packaged as independent entities which may be separated from each other (Cut Hypothesis; p. 245): A sub-sequence of stimuli is cut out of a sequence to become a cognitive entity if it has been experienced many times in different contexts.

Without this effect, the world would be experienced as a continuous stream of events and states (see also Zacks and Tversky 2001). Second, events can be conceptualised, and linguistically described, on different levels of granularity (Section 3.4). Thus, one event is conceived of as the immediate successor of another event only at a specific level of granularity, involving a certain degree of detail. Speakers may in the ongoing discourse choose to integrate more detail, which means switching to a finer level of granularity. In that case, further events can be introduced that occur between exactly those events which were previously described as directly succeeding each other (Habel 1995). Further expressions that may be used in a similar way could be German anschließend (see e.g., Ehrich 1992, Schilder 2002) and its possible English translation subsequently. However, in the literature dealing with the notion of immediate succession, after and before are usually not mentioned. But Habel’s observations concerning succession on a specific level of granularity are fairly close to the insights on proximity discussed in Section 3.4 above. In example (3.23), repeated here as (3.106) for convenience, it is not far-fetched to assume that the next thing he did was get up again. Thus, it is reasonable to assume that the temporal dimensional terms under analysis are also capable of expressing a notion of immediate succession. Chapter 4 will shed more light on this issue. (3.106) After he fell, he got up again.

Another obvious candidate for this concept is then and its German counterpart dann in sentence-initial positions (compare Glasbey 1993 for an account of sentence-final then, and Rossdeutscher and Stutterheim 2004 on dann in relation to sentence position). Glasbey (1993: 286) notes that sentence-initial then can be paraphrased by next. This idea is considered further by Spejewski (1994), who finds that then is sometimes capable of indi-

Related terms 63

cating a “cognitive connection” (1994: 142) between two events, while next is not. She gives the following examples: (3.107) John took a closer look at the photograph. Then he realized who was in it. (3.108) John took a closer look at the photograph. Next he realized who was in it.

In the latter example, the two events seem to appear as independent entities. Thus, then seems to be similar to before and after in that causal or nearcausal relationships can be inferred by its usage. Non-veridical readings and epistemic–deontic interpretations are not available for then. However, Spejewski (1994: 117) points to an interpretation of then that has already been discussed in Section 3.5 above in terms of TERMINATION. With respect to (3.109) (her example), she notes that: [t]here is an implication (…) that the state ceases after the event, but the implication is defeasible, meaning it can be overridden. (3.109) John was depressed. Then he took a Valium.

This possibility is exemplified by: (3.110) John was fairly happy. Then his girlfriend dumped him. He was still happy. Then his car got totalled. He was still happy. Then he was kicked out of school and became severely depressed. (Spejewski 1994: 117)

Thus, then can express partial overlap between the described events; it only requires that the second event starts after the first started, but not necessarily after it ended. Spejewski (1994) additionally points out that this is only true for (unbounded) states, since bounded events are described as a whole and therefore do not allow for such an overlap reading with then. Similarly, Ehrich (1992) notes that the temporal relationship conveyed by dann can be either fully or partially posterior. This is exemplified by the following (1992: 130): (3.111) Hans kam früher nach Hause als sonst. Er lächelte. Dann (?anschließend, ?danach) zog er einen Blumenstrauß hervor. ‘Hans came home earlier than usual. He smiled. Then (?afterwards, ?after that) he produced a bunch of flowers.’

According to Ehrich, dann (though not the alternative temporal terms) indicates that Hans still smiles when producing the bunch of flowers.

64 Temporal dimensions: Linguistic repertory and issues of application 3.10. Conclusion In the present chapter, I have reviewed previous findings on the syntactic range and application of temporal dimensional terms, including the ways in which they reflect underlying concepts of the domain of time. The main result of this overview is the (nowadays uncontroversial) finding that these terms induce discourse relations and additional inferences that go far beyond a mere description of temporal precedence. Different contexts and applicability conditions lead to different interpretations. Furthermore, different syntactic constructions exhibit different kinds of limitations with regard to the conveyance of underlying non-temporal relations. Although sentences containing the temporal markers were often found to lead to very similar inferences regardless of whether a full clause was used or not, a number of aspects could be identified that only occur with conjunctions or (in some cases) conjunctive prepositions. Thus, Degand’s (2000) findings can be partly confirmed for temporal markers in that full clauses are more flexible and more expressive. But on the other hand, relations similar to discourse relations between full clauses could also be identified for shorter constructions, sometimes even if the temporal relatum is omitted completely, as is the case with the usage as an adverbial. Therefore, Talmy’s (2000) approach can also be partly confirmed in that the conceptualisations of Figure and Ground, and the relationships between them, are fairly consistently reflected by a variety of syntactic constructions. A number of issues have been identified that call for a closer look at empirical data. For example, before as a sentence adverbial seems to allow for a number of diverging interpretations. Furthermore, a range of possible non-temporal associations has been identified in the literature with regard to the application of the temporal terms. A natural language data analysis may allow for a further differentiation of categories of interpretation, and possibly the identification of further kinds of interpretation. For example, the notion of immediate succession has been associated with a range of temporal markers, but not with the dimensional terms under analysis. An analysis of natural language data can show whether this notion can also be conveyed by these expressions. Furthermore, the ability of after to convey TERMINATION has not been empirically investigated so far, and the range of interpretation in the German terms needs to be addressed specifically. The analyses of natural language data presented in Sections 4.1 (for English) and 4.3 (for German) will deal specifically with these questions.

Conclusion 65 non-temporal rel-entities

NONT EMPORAL-T YPE

epistemic deontic

{: “and” relation [ : “or” relation

non-factual

RELAT ION VERIDICALIT Y

non-committal presupposed

temp

unrestricted-precedence T EMPORALRELAT ION

immediate-succession restr-prec

REST RBY

causality event-rel

EVENT RELAT ION

conditionality termination

Figure 3.1. Interpretations of relationships between entities in a discourse that are juxtaposed by a temporal marker

Since it has been claimed that temporal clauses contain presuppositions, it is interesting to find out how speakers actually make use of this kind of construction. Do they use presuppositional clauses only if the contents are already accessible in the discourse context, or do they sometimes refer to elements of information that are in fact completely new to the hearer? Section 4.2 below will deal with the strategies speakers employ in discourse with respect to the conveyance of information in temporal components. In Figure 3.1., the findings of the present chapter are illustrated graphically. The schematic representation summarises the ways in which sentences containing a temporal dimensional term can be interpreted. As discussed in Section 3.3, the entities related by a discourse marker may be full clauses or other kinds of syntactic forms. These are not necessarily interpreted as temporally related to each other at all, since non-temporal readings are also available. If they are, they may or may not be veridical, and in some cases (“non-committal”) this remains underdetermined. At the same time, the temporal precedence relation may be restricted or unrestricted. If the temporal relationship is restricted, this may be the case because the events are conceptualised as following directly after one another (immediate succession). Another possibility is that they are causally or quasicausally connected in some way, for example, by direct CAUSALITY, CONDITIONALITY, or by the reversed causality relation TERMINATION. It is assumed here that such associated relationships between events serve to restrict the temporal relationship in a way compatible with Herweg’s findings on proximity discussed in Section 3.4 above.

Chapter 4 Temporal dimensions: Empirical analysis

While Chapter 3 has provided an overview of previous findings with respect to the semantics, application, and interpretation of temporal dimensional terms along with some tests of hypotheses on constructed sentences, the present chapter turns to the investigation of speakers’ spontaneous usages in natural language. One of the problems often encountered in the literature, and prominent especially with the use of constructed sentences, is the restriction to the sentence level. While such an approach may lead to a number of valuable insights, the pragmatic perspective aims for a broader scope. It addresses phenomena that occur in the wider discourse context, including the conveyance of underlying semantic relationships between events on the basis of various kinds of linguistic contributions. With the identification of such relationships, the application contexts in which temporal dimensional terms are naturally produced become clearer. Previous research on temporal dimensional terms, as reported in Chapter 3, has pointed to two main insights associated with the application of temporal dimensional terms. On the one hand, it has been recognised widely that the juxtaposition of events by temporal markers often involves a conceptualisation of an intricate relationship between the events that is not purely temporal. Most uncontroversially, a causal relationship between the events can often be identified. On the other hand, it is often pointed out that the linguistic components connected by a temporal marker are not equal in terms of news value or foregroundedness; in some accounts temporal clauses are generally treated as presuppositional, in contrast to the contents expressed in the main clause. The present chapter addresses these issues with respect to the spontaneous usage of temporal dimensional terms in English and German. Section 4.1 deals with the interplay between the events involved and the application of temporal dimensional terms in English in light of the wellresearched notion of discourse relations between adjacent clauses (Section 2.4.4). Specifically, it is investigated in how far the (conceptualisation of the) relationship between the events referred to influences the application of the terms. The result of the analysis will be a systematic account of categories of application of temporal dimensional terms. Section 4.2 addresses the question as to how far speakers present the contents of the temporal

Relations between events: Analysis of English data 67

clause as Given, presupposing knowledge on the part of the hearer that may or may not be part of the previous discourse. For this purpose, a different corpus was specifically chosen in which the interlocutors’ knowledge about each other was restricted to the information contained in the discourse itself. In Section 4.3, both research areas are addressed for German by way of comparison to the English results.1 Throughout the present chapter, analyses will be entirely and exclusively qualitative. In other words, there will be no relative frequencies, no statistical measures, and therefore no indications with respect to speakers’ preferences and probabilities of occurrence. This method is adopted, on the one hand, because the data stem from unrestricted, naturally produced corpora collected in contexts unrelated to the analysis of temporal terms; there is no experimental design that could delimit the ways in which the terms under analysis would be applied and interpreted. Therefore, there are indefinitely many factors that might contribute to the preference of a certain kind of choice, rendering frequencies of choices unreliable and insignificant. On the other hand, a thorough qualitative analysis of speakers’ usage is considered a necessary preliminary step that needs to be done prior to developing reliable coding criteria, suitable for enabling inter-annotator agreement (Carletta et al., 1997). Previous research on encoding semantic properties has shown that this is not easy to achieve (Spenader 2002). This achievement is not an aim of the present work, although its preparation and enablement certainly is. For this purpose, tests for recognition of underlying concepts will be introduced that enable the analyst to identify plausible interpretations. These tests are intended to be a guideline, keeping in mind that such judgements are necessarily based on subjective assessment. However, this mirrors the situation in actual communication: it is not always clear what kind of underlying relation is intended by the speaker, and often different interpretations are equally acceptable, plausible, and possibly compatible. Generally, the method employed in the present chapter is to search corpora of natural language data for occurrences of temporal dimensional terms and investigate these utterances in the context in which they occur. Thus, in most cases, a first intuitive understanding of the semantic relations between the two connected components could be gained. In some cases, the discourse context turned out to be too complex to be able to determine the 1.

This chapter summarizes research reported in more detail in Tenbrink and Schilder (2003), Tenbrink (2004), and Tenbrink (2007).

68 Temporal dimensions: Empirical analysis underlying relationships; such cases were not analysed further. This has no direct influence on the results because no relative frequencies were aimed at. Gradually, patterns of usage emerged which are presented and discussed along with suitable examples in terms of categories of discourse relations. A fairly large amount of data is necessary for this procedure since relationships between situations represented in the discourse often involve complex cross-references and associations with previous trains of thought. For a suitable demonstration of the concepts it is necessary to identify relatively clear-cut examples which do not involve further complications, and which are readily and obviously interpreted in the suggested way rather than a different one. This is particularly tricky because sometimes several interpretations are plausible, which is part of the pragmatic flexibility of the terms under analysis. Unlike Chapter 3, in the present chapter no systematic tests of syntactic forms are presented. This is due to the fact that the present chapter relies entirely on natural language data; therefore no syntactic transformations are employed. Since not all syntactic forms occur in the data with equal frequency, a systematic syntactic analysis of the empirical data would have required a different procedure involving relative frequencies, which is left to future investigation. 4.1.

Relations between events: Analysis of English data

In the present section, data from two different online available corpora are analysed. The first corpus is taken from the CHILDES collection (MacWhinney 2000) containing 26 20-minute sessions of interaction between the 2- to 3-year-old Trevor and his father (Demetras 1989). These data were videotaped in 1985–1987, transcribed and made available for computer analysis. Trevor and his father are native speakers of American English. Most relevant utterances stem from the father. No attempt is made here to analyse the effect of his language input on Trevor’s language development. Examples from this corpus are indicated by mention of father or child as speaker. The second corpus is a sample of the Corpus of Spoken Professional American–English (CSPA),2 containing press conference transcripts from the White House, and a record of faculty meetings at UNC and Committee Meetings held at various locations around the country to discuss 2.

http://www.athel.com/sample.html; last visited: September 20th, 2006.

Relations between events: Analysis of English data 69

the creation of different kinds of national tests. Examples from this corpus are indicated by mention of the speaker’s name, which also enables identifying the utterance in its wider context in the original corpus. The sample contains four stretches of discourse, altogether approximately 42,000 words. These two corpora were chosen because they represent two broadly different kinds of context, which are taken into account in order to identify various temporal and non-temporal concepts. The corpora were searched for occurrences of three English expressions that are used to express temporal order, namely, before, after, and then. The latter is included in order to address its relationship to the two temporal dimensional terms before and after, since earlier treatments point to a similar range of applicability (cf. Section 3.9.3). Thus, a specific aim of the present section is to identify similarities and differences in applicability between before, after, and the related term then. In these corpora, the following conceptual categories underlying the usage and interpretation of before, after, and then were identified: x Unrestricted temporal precedence: Repetition and Reference to past event x Proximity involving discourse relations such as EXPLANATION and RESULT (which can be subsumed by CAUSALITY) and NARRATION (which can be further specified through the concepts INSERTION, REGULATION, DEPENDENCY) x IMMEDIACY: Immediate succession on the agent, patient, or narrative level x EXPLICIT LENGTH OF TIME. These different categories will be considered in detail in the following subsections. 4.1.1. Unrestricted temporal precedence Of the three expressions analysed, only clause-final before (i.e., before used as a clause adverbial rather than a conjunction or a preposition) regularly expresses a temporal relationship that is unspecified for proximity. More specifically, clause-final before may be used to indicate that one specific event either happened at some unspecified time or never happened at any time before a reference moment (such as speaking time). In these cases, no non-temporal semantic interconnections are implied, and there are no indi-

70 Temporal dimensions: Empirical analysis cations that the event (if it happened at all) happened in temporal proximity to the reference time. Two different kinds of usage were identified in the corpora. Clause-final before can be used to express the concept that an event that is a current topic in the actual discourse has already occurred once at some unspecified earlier time, such that a repetition is induced. (4.1)

PIKULSKI: The other thing I’m very concerned about is the issue of aggregation of any of this data. This issue has come up before.

This concept may also be used in questions (in which the issue to be negotiated is whether or not there is a repetition) and negated sentences (in which it is denied that there is a repetition): (4.2) (4.3)

BROWN: How many, anybody ever participated in that before? STRICKLAND: I would like to welcome two people who haven’t been with us before.

Throughout these three usages of repetitive before, the perfect tense is (or is intended to be) used. Thus, repetition is induced where before occurs at the end of a perfect clause. With the perfect tense, some effects which have long been recognised in the literature (e.g., Comrie 1976), come into play: the event mentioned in the perfect clause triggers a so-called posterior state (the continuing relevance of a previous situation) after the event that holds up to the reference time, which in the case of present perfect is the present moment. No restriction is required on when this earlier event happened; the only requirement is that it needs to be relevant to the time of speaking. This relevance is present in sentences involving clause-adverbial before via the fact that the event is now repeated. It is the reader’s or listener’s task to search for an event during the time of reference (equivalent to present time in ongoing discourse) that matches the expectation that the event mentioned in the perfect clause is repeated. With negation, the interpretation is as follows. The negation of an event is interpreted as a state (temporally prior to the present moment) in which this event does not occur. Thus, the inference is that one event which has not happened at any time before is referred to at the present moment, such that it is denied that there is a repetition of similar events. The second concept that may be conveyed by clause-final before is reference to a specific (but not necessarily linguistically specified) time (span) in the past, which is not required to be proximate to the present moment. This concept is evoked where a clause-adverbial before occurs with the past tense, as in:

Relations between events: Analysis of English data 71 (4.4) (4.5)

FATHER: you changed his clothes and put them in the corner? Oh, we did, yeah a couple times. You mean when we made a tape before? CHILD: did my microphone come off? FATHER: before, it did. CHILD: did it come off now?

Note that, in the latter example, the father’s utterance is syntactically derived from the clause-adverbial structure it did before by topicalisation. This linguistic choice emphasises the contrast between the present time (now) and the earlier event that the father refers to. Where clause-adverbial before occurs with the past tense, before may be paraphrased as in earlier times while still conveying a similar meaning. The context of the discourse may, however, evoke proximity if it is understood that the earlier moment mentioned in the clause involving before occurred within a certain time frame. In the following example, the time of the meeting is obviously the intended time frame, provided that the principle of relevance (a generally accepted basic principle of communication saying that utterances must be relevant in a given context, e.g., Grice 1975, Sperber and Wilson 1986) is not violated: (4.6)

STRICKLAND: Eunice had her hand up before.

Here, Strickland points to a specific hand-signalling event during the time of the meeting that happened before the present moment, and which can be understood from the discourse context as the reason for Strickland’s assigning speaker rights to Eunice now. In a different interpretation, however, this example may be understood as an elliptical version of a sentence such as: (4.7)

(4.6)’ Eunice had her hand up before Peter had his hand up.

in which a temporally ordered list of speaker rights based on the times of their signals during the present meeting is induced. This should then be interpreted as a case of REGULATION (see below). It is not clear whether such closely related possibilities of interpretation are worked out in detail by the hearer, as long as a decision is not necessary. Both interpretations work well in this context and support the intended consequence, namely, to provide the reason why Eunice should be allowed to speak next. To test whether clause-final before may be interpreted in an elliptical sense, it must be possible to repeat part of the previous lexical material in order to complete the sentence as in example (4.7). In sum. The corpora data show that only clause-final before regularly (i.e., even in the absence of further linguistic markers of temporality) ex-

72 Temporal dimensions: Empirical analysis presses a temporal relationship unspecified for proximity, i.e., one in which no specific temporal restrictions are required for the earlier event. With the perfect tense, clause-final before induces repetition, whereas with the simple past, a specific past event is referred to. Clause-final before may also be used in an elliptical sense, in which case other interpretations become possible as described in the following sections. All other usages of before, after and then generally involve temporal constraints (which may be made explicit or ruled out by contextual clues) that fall roughly into three categories. These are described in the following three subsections: Proximity, Immediacy, and Explicit length of time. 4.1.2. Proximity Before, after, and then can all be used in situations where it is inferable from the context – and possibly further specified by modifiers such as shortly – that the mentioned events happen in close temporal proximity. It stands to reason that the temporal expressions (excepting before in clausefinal position) themselves carry a proximal interpretation as part of their connotational meaning, which may be ruled out by the context, e.g., by adding explicit information about the length of time between the two described events (see Section 4.1.4 below). In some cases, there is reason to assume that the events succeed each other immediately (see Section 4.1.3). In all other cases of temporal proximity, there is some other reason for mentioning the events together, such as a non-temporal semantic interconnection like CAUSALITY, that is either explicitly mentioned in the linguistic context, or that is derivable from the situation or from world knowledge. Semantic interconnections may be identified as the speaker’s reason for choosing to linguistically express the temporal relationship between two events (provided, again, that the speaker intends to contribute a relevant piece of information to the discourse). Such interconnections are typically treated in terms of discourse relations in various approaches. Here, the starting point is Lascarides’ and Asher’s (1993) proposal, chosen because it specifically addresses temporal phenomena. In the present corpora analysis, natural language examples are provided for three of the discourse relations identified in SDRT (Segmented Discourse Representation Theory, Lascarides and Asher 1993): NARRATION,

Relations between events: Analysis of English data 73

EXPLANATION, and RESULT.3 Additionally, the cases are further specified in which the discourse relation NARRATION is found. One distinction concerns informational status, i.e., whether the clause’s content is new to the hearer or given in the dialogue context. Moreover, there may be some kind of DEPENDENCY between the events. Thus, a more fine-grained set of discourse relations is necessary with respect to the usages of the temporal connectors in dialogue. Altogether, three relations are distinguished that can be subsumed by NARRATION: INSERTION, REGULATION, and DEPENDENCY. Since the definition of NARRATION includes that the two events have to share a common topic, the informal description of INSERTION, REGULATION and DEPENDENCY can be seen as a further specification as to how a common topic could be derived. In the following, each of the identified discourse relations is described in detail, starting with causal relations. 4.1.2.1 Explanation and Result: Causality The two causal discourse relations EXPLANATION and RESULT are distinguished only by the order of the two clauses involved, i.e., whether the temporally prior event is mentioned first. Since order of mention is not in the current focus, these two relations are here subsumed by the more general term CAUSALITY. As discussed in Section 3.5, causal relations between the clauses are not easily inferred with before, but they are with after. The present corpora findings support this claim, as illustrated by examples such as: (4.8)

MANDEL: But to understand that this is a next natural progression after many years of work at the local and the state level.

Here, exchanging the temporal term by a causal marker does not seem to have a major impact on the interpretation: (4.9)

3.

(4.8)’ (…) this is a next natural progression because of many years of work at the local and the state level.

Although it is also possible to contrast events with each other in a discourse involving the three connectives under analysis, temporal connectives alone do not seem to be capable of expressing this relationship, but need further expressions to attain this specific meaning, as in but then. Therefore, such cases are left out of the analysis.

74 Temporal dimensions: Empirical analysis Moreover, then may also support CAUSALITY. In (4.10), the later event can naturally be understood as being caused by the prior one: (4.10)

MANDEL: And we can make sure that it gets distributed in a broad range of communities. Then, we get a complete diversity of responses in the review process.

This can be shown by the reformulation given in (4.11): (4.11)

(4.10)’ And we can make sure that it gets distributed in a broad range of communities. Because of that, we get a complete diversity of responses in the review process.

The causal inference gets lost when the clauses are connected by before instead of after or then: (4.12)

(4.10)” We can make sure that it gets distributed in a broad range of communities before we get a complete diversity of responses in the review process.

Thus, causal relationships may easily be conceptualised with events that are juxtaposed using after and sometimes then as well, but with before such relationships are less readily inferred. 4.1.2.2 Narration I: Insertion Where future actions are expressed, speakers may insert one event, or a time span, between the present moment and an expected event, using before.4 In this case, the event described by the temporal clause is given or expected based on the discourse context, and the inserted one is new to the hearer. This kind of NARRATION relation is distinct in that the consequence is expected or can be inferred from the context, while the event which it is the consequence of is new to the hearer. The default order, in which the Given part precedes the New (Halliday 1994), may be changed, indicating special emphasis on the New clause if the clause complex starts with it: (4.13)

4.

STRICKLAND: Just one thing going back to the issue of time before you speak, Eunice.

In the two corpora described in this section, INSERTION was only identified with future actions, although the investigation of other text types reveals that this concept may equally be used with other tenses if the ‘present moment’ is textually projected into the past.

Relations between events: Analysis of English data 75

Here, it is expected that Eunice will speak; however, something happens before this expected event. Note that this meaning may not be expressed using after: (4.14)

(4.13)’ After just one thing going back to the issue of time, you speak, Eunice.

This sentence conveys a different meaning, conceivably one in which Eunice is told that it will soon be her turn to speak. Here, there is no clue as to whether the speaking event is already expected by Eunice or not. Then works similarly, while the effect is less strong here: (4.15)

(4.13)” Just one thing going back to the issue of time; then you speak, Eunice.

Thus, in cases of INSERTION, information (which is new to the hearer) about an event that happens earlier than another expected (and thus presupposed) event is provided. With before, this information appears in the main clause, which is syntactically also the superordinated clause, and which is not presupposed. With after, in contrast, the relevant information would have to be conveyed via a presupposed clause, while the part of the sentence that is expected appears in the main clause in a position which is usually the place for new information. With then, none of the clauses are presupposed. It is then the hearer’s task to infer which one of the clauses contains the relevant information. Cases of INSERTION can be identified by reformulating the utterance in a way that makes the insertion of an unexpected event before an expected one explicit. Reformulating (4.13) above in this way yields: (4.16)

(4.13)’’’ Yes you will speak, Eunice, but before that happens there is just one thing going back to the issue of time.

Another example is given in (4.17) and reformulated in a similar way: (4.17)

(4.18)

MYERS: And then I just thought a little readout from this morning’s health care meeting– VOICE: Before you do that, can I ask who Markey was meeting with? (4.17)’ Yes you will do that, but before that happens can I ask who Markey was meeting with?

In (4.17), the fact that the information in the temporal clause is Given (and thereby expected) is linguistically reflected by the use of the anaphor that instead of providing fully specified content.

76 Temporal dimensions: Empirical analysis A special case of INSERTION are rules: statements which concern not only the event currently under focus, but express a general regularity. This concept is evoked by the presence of a modal verb such as supposed to or must, as in: (4.19)

CHILD: now can we knock it down? FATHER: you’re supposed to ask before you knock it down.

4.1.2.3 Narration II: Regulation In the corpora analysed here, speakers often use before or after to regulate the order of future actions. There seems, however, to be no fundamental reason why speakers should not employ then as well. In cases of REGULATION, both events are known or derivable from the discourse context, and the utterance’s new information concerns the order itself, as in: (4.20)

FATHER: can I drink it now? CHILD: no. after dinner. FATHER: after dinner. okay.

This effect can be reinforced by placing prosodic focus on the temporal term, although this need not be done in all cases by the speakers. Similarly, the order of two known events is at stake in situations of contest, as the following example shows: (4.21)

FATHER: can you get to the light before I do?

In the same vein, REGULATION can occur with other tenses whenever the order of two events is focused on in the discourse context, as in the continuation of the above situation, in which the father talks about the contest after it is over: (4.22)

FATHER: I almost thought I was gonna get it before you got to it.

As a suitable test for this concept, the clause complex may be paraphrased to produce an it-cleft such as: (4.23)

It was before you got to it that I got to it.

It-clefts produce the effect of special focus on the following clause. Such an effect fits well into a context in which the temporal ordering is negotiated, while it seems artificial in others:

Relations between events: Analysis of English data 77 (4.24)

(4.13)’’’’ It is before you speak, Eunice, that I want to mention just one thing going back to the issue of time.

Usually, clause complexes conveying REGULATION will serve as an appropriate answer to the question “When does/did one of the events happen?”. Thus, it is clear that both events need to be known in the discourse context: for one event, the time at which it occurs needs to be specified, while the other event serves as a reference time for this specification. 4.1.2.4 Narration III: Dependency The later event (E2) may be based on the earlier one (E1). This relationship (called CONDITIONALITY, ENABLEMENT, or – here – DEPENDENCY)5 is closely related to CAUSALITY, but it does not involve a direct causal relation (cf. Section 3.5). According to the present analysis it can occur with all tenses, and all three expressions under investigation can be used for it. In these cases, E1 can be viewed as a precondition, or necessary preparation, of E2, in the sense that the second event would be pointless or could not even occur without the earlier one (without being directly caused by it). This kind of discourse relation does not seem to be related to specific kinds of information status. (4.25)

(4.26)

VOICE: So are you saying that there was a meeting; it was for the White House to inform them of the announcement [E2] after it was made [E1] and to brief them in advance, not to tell them please don’t go out and [...] JACKSON: [...] we even send out some surveys for courses that were taught the previous semester [E1] then attempt to get the information back [E2].

The notion of a dependency or preparation relation can be tested for by making this relation between the two connected events explicit, as in: (4.27) (4.28)

(4.25)’ The making of the announcement [E1] is a precondition for the White House to inform them of the announcement [E2]. (4.26)’ Sending out some surveys for courses that were taught the previous semester [E1] is a precondition for attempting to get the information back [E2].

Another example is the following: 5.

The term is chosen because it seems to allow for a greater flexibility concerning the ways in which one event may be based on another.

78 Temporal dimensions: Empirical analysis (4.29)

MYERS: Again, we’ll wait – he’s asked James Lee Witt to call him after assessing the damage and talking with state and local officials and to tell him whether or not that would be useful.

In this case, the later event (calling) is only sensible based on the results of the first event (assessing the damage), as supported by a reformulation test: (4.30)

(4.29)’ Assessing the damage is a sensible preparation for calling him.

Thus, the concept of DEPENDENCY includes various degrees of interdependency; in some cases a direct conditional relationship can be identified, while in other cases, the relationship is less strong but nevertheless implied by the juxtaposition of the events. In contrast to INSERTION, the situations described in clause complexes involving either REGULATION or DEPENDENCY are independent of a (third) specific time span such as the moment of speaking, except for the obvious fact that the temporal relation between the moment of speaking and the described events determines the tense to be used. Recall that INSERTION, in contrast, is characterised by the specific concept that an event is inserted in between the present moment (in cases involving the present tense) and a further (expected) event. In sum. The corpora analysis of discourse segments where two events are represented as occurring in close temporal proximity, but not necessarily directly after one another, reveals that a discourse semantic interconnection between the events is supported in these cases. With after and then, this interconnection may be a causal relationship. A further possibility, available for all three temporal expressions, is non-causal DEPENDENCY; in other cases, the interconnection depends on the information status of the hearer. In order to gain a meaningful interpretation of the discourse segment, the hearer is required to derive the intended interconnection between the events. However, the corpora analysed in the present study did not contain any examples for two other relations previously identified by Schilder (2001), namely TERMINATION and non-factual relations (but see Section 4.2 below). This may be caused by differences in discourse type, i.e., it is conceivable that different kinds of relations are typical for either dialogic or narrative discourse.

Relations between events: Analysis of English data 79

4.1.3. Immediacy In Section 3.9.3, the concept of IMMEDIATE SUCCESSION was introduced; it occurs when speakers conceptualise events as one following immediately after another at one specific level of granularity that is identifiable (i.e., shared) by the discourse participants. In previous work, it had been assumed that this concept is conveyed predominantly by complex expressions such as as soon after as possible or immediately afterwards.6 However, the present analysis shows that at least then is perfectly capable of conveying this concept, and indeed is frequently used in this sense. One example is as follows: (4.31)

STRICKLAND: Audrey, do you want to respond? And then Marilyn.

Here, no further semantic interconnection between the two responding events (involving first Audrey, then Marilyn) is intended beyond the fact that they follow immediately one after the other, in terms of speaker turns. In contrast, the following example illustrates a case of temporal proximity without implying immediate succession: (4.32)

FATHER: so after dinner I get pie? CHILD: yeah. FATHER: what do I get now?

In this example – which can be interpreted as a case of REGULATION – the father does not intend to find out what he will do immediately after dinner; instead, he inquires whether he gets pie (E2) after having finished the regular meal (E1), contrasting the time of E1 with now in the following conversational turn. However, another example indicates that after may in certain contexts also be used to express immediate succession: (4.33)

CHILD: lookit dis air plane. FATHER: where’s he going? CHILD: um right dere. FATHER: oh. where’s he goin after that? CHILD: um it’s not goin to duh zoo.

Where immediate succession is involved, no further events on the same level of granularity (i.e., no comparable events) may happen between the 6.

The concept also corresponds fairly well to that of Sequence in RST (http: //www.sfu.ca/rst/; last visited: June 9th, 2005), where relations are defined without reference to specific lexical items.

80 Temporal dimensions: Empirical analysis two mentioned events. This can be tested by asking whether the second event is intended to be the next event: (4.34)

Is he going to the zoo next?

The answer to this question should either be “Yes” or contain the insertion of a further similar event in order to deny immediate succession of the aforementioned events: (4.35)

No, he’s going to the station first.

However, it should not contain an event on a different granularity level: (4.36)

No, he’s traveling through the air before he’s going to the zoo.

Here, it is not possible to infer that the speaker simply wants to be more specific. Instead, the answer requires an interpretation in which the air travel takes much longer than expected if the next goal is the zoo. With that interpretation, a case of INSERTION is inferred in which an event on the same level of granularity is inserted in between the present moment and the expected event of going to the zoo. Then is used preferentially to express the concept that one event immediately succeeds another; in fact, it can be concluded that immediate succession is the standard (or ‘default’) interpretation for clause complexes involving temporal then. In contrast, the corpora contained only very few examples involving before and after that could be interpreted in terms of immediate succession. Thus, although before and after are also capable of expressing this concept, they are only seldom used for it, and if so, the linguistic context often indicates that the usual interpretation involving a semantic interconnection is overruled, as in: (4.37)

FATHER: then I’ll get you a new piece later. CHILD: yeah. FATHER: (o)kay. CHILD: no get new piece after dis one.

Here, after is contrasted with later, thus it is clear that the child wants to get the piece directly after the first one in contrast to some later time. To further specify the concept of IMMEDIACY, this kind of event conceptualisation may involve different kinds of attentional focus on textual elements, which are here categorised as the agent, the patient, and the narrative level. With this categorisation, two central roles of functionally recognisable discourse participants are captured, together with a higher-

Relations between events: Analysis of English data 81

order level of discourse organisation which does not focus on specific participants but rather on the unfolding of the event structure as a whole. Further investigation may lead to the identification of additional possibilities; in the data used here, these three levels could be recognised. On the agent level, the activities of a person who is represented as the responsible participant of the action are described in temporal order, as in: (4.38)

MYERS: It’s more of a periodic view, and then he will be briefed privately out of the view of your eyes and ears. Then he will have lunch with the Vice President.

In this example, the steps the agent (the President) will take are described one after the other. Obviously, this indicates a listing of events ordered by their temporal occurrence, implying no semantic interconnections, concentrating on the agent. Linguistically, this effect is achieved by a consistent lexical string involving a repetition of he; while the other lexical items change throughout the discourse segment. On the patient level, in contrast, events are described that happen to an entity which is represented as the affected participant of the action. Thus, in the following example, the events that happened to an object found by the father are described in temporal order: (4.39)

FATHER: I found it on the windshield of a car. CHILD: yeah. FATHER: then I uh pulled it out and I showed it to Mommy. and then I brought it home.

This example illustrates that it is possible to represent a succession of events concentrating on the patient even if the patient is not in syntactic subject position, as the example is narrated by the agent in active voice. Note that, in this specific example, both agent and patient level interpretations are possible because of the consistent lexical cohesive strings for both agent (I) and patient (it). However, the wider discourse context reveals unambiguously that the discourse participants are currently talking about a flyer which the father found on the windshield of a car. In a case of immediate succession on the agent level, in contrast, the participants would talk about the father’s successive actions, rather than about the flyer. Finally, narrations may involve a description of neither agents’ actions nor patients’ undergoings, but rather of “what happens next”, i.e. the narrative level, as in:

82 Temporal dimensions: Empirical analysis (4.40)

CHILD: they have to go upstairs. FATHER: there they come. CHILD: weal high up stair. FATHER: way up. CHILD: and then they. it’s way up. and then this guy jumps in. FATHER: he’s in the racquet club pool.

In this example, there is no lexical string involving the same agent or patient throughout the discourse, as in the previous examples. However, the sequence is still coherent because it is understood that the described events happen in close temporal and spatial proximity, as reflected lexically by then and jumps in. Thus, in the case of the (higher-order) narrative level, temporal succession is maintained independently of specific participants. In sum. The corpora analysis revealed that speakers’ conceptualisation of events as happening immediately after one another is usually expressed by sequences involving then. However, after and before are also capable of expressing immediate succession in specific cases. Where immediate succession of events is conceptualised on one of three available discourse levels, i.e., the agent, the patient, or the narrative level, no further semantic interconnections are implied or required. Discourse contexts that involve a temporal ordering of temporally close events, without a linguistic specification of the length of time between the events, therefore allow for either immediate succession or semantic interconnection. 4.1.4. Explicit length of time Neither immediate succession nor semantic interconnections need to be implied if reference times are explicitly given, as in the following two examples: (4.41) (4.42)

VOICE: It’s now a month after the IEA suggested it had to know within weeks. MANDEL: (…) We just had met nine days before this new document was coming out.

In such cases, the time span itself or the placement of an event in time is of interest rather than a semantic relation between two events. The default interpretation of immediate succession for sentence-initial or mid-sentence then can be overruled by the existence of an explicit reference time (ERT), as in:

Relations between events: Analysis of English data 83 (4.43)

MYERS: (...) Saturday, he will give his radio address live at 10: 06 a.m. and then leave that night at roughly 11: 00 p.m., maybe a little bit before, for Brussels.

Where then occurs in clause-final position, ERTs evoke simultaneity (cf. Glasbey 1993), as in: (4.44)

BAYNE: (...) or do we have to wait until 2056. (...) BROWN: Great. And I’ll be dead by then. BAYNE: And you wouldn’t be counted then.

Glasbey distinguishes between clause-final occurrences of then together with an ERT, in which cases then is used anaphorically, and other clausefinal occurrences which support a BACKGROUND or ELABORATION relation between the events. In the present corpora, clause-final then occurs only with ERTs or in questions, which Glasbey does not deal with. In questions, an immediate successor relation can be identified, as in: (4.45)

FATHER: an you jumped on the bed. (...) what’d you do then? CHILD: go go sweep. FATHER: no! you didn’t go to sleep!

Note that, in this example, the father’s question allows for two possible readings, which involve different intonational contours: he might, as well as asking what happened next, intend to find out what happened simultaneously to the jumping, in which case the intonation would involve a stress on do. However, the child’s answer (i.e., the discourse context) rules out this possibility, as it is hard to imagine how one would go to sleep while jumping on the bed. Further investigation is necessary to specify the contribution of intonational contours to the interpretation of clauses involving clause-final then. In sum. When temporal terms occur together with an explicit mention of a time span, then it is not necessarily the case that the events are semantically related, since the length of the time span may be of interest by itself. Clause-final then with explicit mention of a time span invokes a simultaneity reading, i.e., it is used as a temporal anaphor. 4.1.5. Conclusion The following general observations emerge from the analysis so far. The occurrence of clause-final (i.e., adverbial) before calls for further investiga-

84 Temporal dimensions: Empirical analysis tion. According to the present findings, clause-adverbial before induces two different concepts, depending on the tense with which it occurs: With the perfective, clause-final before indicates the occurrence of a repetition; with the past tense, it points to a specific past event. Furthermore, clause-final before may be part of an elliptical sentence, in which case relationships may be induced such as those expressed by the juxtaposition of clauses. The occurrence of before and after (other than as a clause-adverbial) usually induces a discourse relation between the components which integrates both temporal and non-temporal information. This relation needs to be identified and interpreted by taking the discourse context into account. Examples for such relations are EXPLANATION, RESULT, INSERTION, REGULATION, and DEPENDENCY. However, this categorization does not claim to be exhaustive. The identified categories of concepts underlying where before, after, and then are applied can be summarised as follows. When an EXPLICIT LENGTH OF TIME is provided, it is likely that the time span itself is the most important information conveyed by the utterance. In such cases, semantic interconnections are not necessarily inferred. IMMEDIACY: Supplementing previous findings on the concept of immediate succession, it turned out that this was the most natural interpretation of then,7 used frequently throughout the corpora, whereas immediately after and the like did not occur at all. Where events are described as occurring directly after one another, this can be traced linguistically on either the agent, the patient, or the narrative level. Mostly, this option is realised by then rather than after or before, but the latter two can also be employed for it. This concept can be tested by inserting another event on the same level of granularity, which should then change the meaning. A suitable reformulation test would be to express the concept of succession directly by a term like next or a phrase like The next thing that happened / they did was…. CAUSALITY: One of the two events connected by a temporal term causes the other. This concept is often expressed using after; where before is used, causal connections do not seem to be as easily inferable. This concept can be tested for by reformulating the utterances using a term that directly expresses causality, such as because. DEPENDENCY: Before and after can both be used to express the fact that the later event is based (non-causally) on the former. Then, the former event is (in the strong case) a precondition or (in the weaker case) a sensible preparation for the latter event to occur. This concept can be tested for 7.

This finding mirrors studies on French puis carried out by Bras et al. (2001).

Relations between events: Analysis of English data 85

by reformulating the sentences including based on and it is sensible/ possible to. INSERTION: New information about an event that happens earlier than another expected event is provided. It may be presupposed that the expected event occurs after the inserted one, unless a non-veridical relation is inferred: the inserted event may have the effect of precluding the expected event, or it may be indeterminate whether the expected event happened after all. The concept of INSERTION can be tested for by reformulating the utterance making explicit that something else happens before or instead of something that is expected to happen. After cannot express this concept. REGULATION: Before and after can be used to point to the order of known events. This is often the case when scheduling future actions, but also in other cases where the order of two events is considered newsworthy. This concept can be tested for by placing prosodic prominence on the temporal term in order to see if the order itself is in focus. Furthermore, cases of REGULATION can be identified by reformulating the sentences using itclefts, which also serve to reinforce the temporal order, and by testing whether the utterance would be a suitable answer to a When–question. Since the temporal order itself is in focus in cases of REGULATION, causal or conditional inferences are not supported. Therefore, it stands to reason that the usual requirement of temporal proximity between the events may be relaxed with this concept, because it is conceivable that the order of temporally remote events is negotiated. Often, however, temporal order should be relevant (as is the case in occurrences of REGULATION) particularly in cases where events are conceptually related just because they are not too far remote in time. These considerations, of course, support the assumption that proximity is a subjective, gradual, and qualitative concept (cf. Section 3.4) which is hard to formalise or specify further, in spite of its obvious cognitive relevance. Naturally, the identification of the underlying categories proposed in the present section is not always straightforward or unambiguous. Although methods such as reformulation tests are frequently employed in the literature (and the current work follows this tradition) they are nevertheless subjective to a high degree, as they reflect the analyst’s assessment of whether the interpretation of the utterance is changed or sufficiently similar to justify the identification of a concept. Furthermore, the concepts are not necessarily mutually exclusive. There is no reason why one utterance should not support the interpretation of several underlying concepts at once, especially in light of the fact that they are conceptually fairly close. In fact, if the pro-

86 Temporal dimensions: Empirical analysis ducer of the utterance had intended to restrict the interpretation to just one relationship, they would probably have formulated the utterance in a more specified way. In the previous sections, the natural language examples have been related to different concepts, neglecting the possibility of other interpretations. Examples were chosen that seem to suggest the proposed underlying concept. The following example is rather more ambiguous; it could be interpreted in terms of either INSERTION or DEPENDENCY: (4.46)

MYERS: Before we begin the background briefing on today’s event and the empowerment zones, I thought I’d give you an update on the earthquake.

The reformulation test for INSERTION yields the following, which seems consistent with the original: (4.47)

(4.46)’ As you know we will begin the background briefing (…), but before that happens, I’ll give you an update on the earthquake.

However, there could also be a relationship of sensible preparation between the two events, as the reformulation test for DEPENDENCY shows: (4.48)

(4.46)’ Giving you an update on the earthquake is a sensible preparation for beginning the background briefing on today’s event.

In order to see whether (4.48) is indeed consistent with the original utterance, it would be important to know whether “today’s event” is related to the earthquake (which is not entirely clear from the available data). This could be supported, for example, by prosodic focus on update in the original utterance, whereas if earthquake is focused, it is presented as New and therefore likely to be unrelated to the events mentioned in the temporal clause. Thus, various features of the discourse context contribute jointly to conveying the intended meaning (cf. Bateman 2001: 57). 4.2.

Presupposition processes

Several linguistic devices can be used to signal whether a part of the discourse is new or already known to the participants. As a rule of thumb, for example, all that is introduced into the discourse with an indefinite article can be viewed as New, while definite articles signal a Given status; these are also regarded as presuppositional. However, in natural discourse, the interlocutors’ actual information status does not always correspond to this simple distinction. For instance, discourse-new elements are frequently not

Presupposition processes 87

presented as new, and much of the discourse needs to be inferred rather than being presented explicitly. Such phenomena – with different emphasis on various related aspects – are variously treated under the headings of Common Ground (e.g., Clark 1996), information status (van Deemter and Kibble 2002), presuppositions, Given/New dichotomies (plus further distinctions to deal with inferable entities), etc. (see also Tenbrink 2007). In the functional approach (e.g., Halliday and Matthiessen 1999), the Given/New distinction is applied to (shared) meaning rather than to knowledge. In this view, language reflects the speakers’ linguistic strategies to communicate elements as New or Given, irrespective of the hearer’s assumed state of knowledge. Then, backgrounding is achieved by packaging pieces of information in a linguistic form that signals to the listener that the information is not to be focused upon. This happens locally by using specific linguistic constructions, and globally by a Given/New pattern. In Halliday’s (1994) terms, the Given/New structure is reflected prosodically in spoken language, with each clause possessing its own Given/New structure in terms of prominent and non-prominent elements. The unmarked structure is that the Given part is followed by the New. In these cases the Given part corresponds to the clause’s Theme, representing what the speaker is talking about (through longer stretches of the discourse, this signifies thematic organisation). But it is also possible that clauses start with the New part; this is realised by prosodic prominence, and in written language, by marked syntactic structure. Since linguistic Given/New patterns reflect meaning rather than knowledge, elements that are new to the discourse may be introduced as established facts, reflecting the speaker’s stance. This happens, for example, by using presuppositional constructions such as the fact that in situations in which the information to be presented had not been introduced in the discourse before. A key notion for the treatment of discourse-new elements that are introduced in the form of presuppositions is that of accommodation (Lewis 1979). If new information is introduced by the speaker as a presupposition, the common ground of speaker and hearer is tacitly extended in a way consistent with the previous discourse so that it includes the new information. As a very simple example, if a speaker says “my sister” in any kind of linguistic context, hearers will probably from this point start to believe that the speaker has a sister, regardless of whether this information has previously been known or not. However, accommodation may not always be as straightforward as in this example. Alternatively, listeners can explicitly reject the presupposition, resulting in a cancellation of the pre-

88 Temporal dimensions: Empirical analysis supposition (this possibility is captured in formal approaches by distinguishing between local and global accommodation), or they can ignore it either consciously or without being aware that a presupposition has been implied. Then, the speaker will believe something to belong to the common ground that, in fact, is not represented in the listener’s current representation of the discourse. Such misrepresentations may cause communication problems, they may be detected later in the discourse, or they may have no effect at all, being irrelevant. Furthermore, the accommodation process itself can happen either consciously or without awareness. In spite of a large amount of literature dealing with these issues, there has not been much empirical research applying these notions to natural (unrestricted) discourse. One reason for this is that it is notoriously hard to specify for any utterance what exactly the common ground of the interactants may be; and speakers’ reasons for presenting specific linguistic elements as either Given or New are equally hard to determine. In the present section, these issues will be addressed empirically by investigating in how far speakers using temporal dimensional terms take the listener’s prior knowledge into account. This issue is particularly interesting for the investigation of these terms since they have frequently been claimed to trigger presuppositions in various ways (cf. Section 3.6). Time and again, it has even been claimed that they can be (or are even regularly) used “unfairly”, to impose knowledge on the listener by mentioning new information in a backgrounded clause rather than introducing it as news (see also Tenbrink 2007 for further elaboration of this idea).8 As the overview presented in Section 4.1 shows, some of the identified categories of usage of temporal dimensional terms necessitate a specific kind of information status, while others do not: for instance, if REGULATION is inferred then both connected elements should be Given, as the most important information concerns not the connected events but the order itself; and for an interpretation in terms of INSERTION the event described in the temporal component must be Given. But for IMMEDIACY, CAUSALITY, and DEPENDENCY, no such requirements have been identified. A further distinction needs to be made between the two kinds of ordering that can occur with the temporal terms. As just outlined, in the standard case a sentence departs from what is Given in the discourse context. Thus, the ordering of the elements of a sentence has implications with regard to 8.

Dan Scorpio’s online essay on ‘Presupposition’ (http://www.angelfire.com/ nd/danscorpio/presup.html; last visited: June 9th, 2005) deals in considerable detail with the possibility of unfair usage of presuppositions.

Presupposition processes 89

the information flow. In Schilder and Tenbrink (2001) we show that sentences beginning with a temporal connective differ from those in which the connective appears only later in that the former sentences comprise two kinds of discourse relations instead of only one: preposed elements need to be connected in specific ways to the previous context, while postposed constructions have a weaker, and more unrestricted, relationship to the context. Thus, it seems that not all components introduced by before or after can be analysed as common ground for both speaker and hearer. Instead, intricate relationships between the temporal component, the main clause, the previous context, and other sources of knowledge account for the employment of a temporal term. Since it can be assumed that temporal components serve as an anchor for specifying the time of a further event, the question needs to be addressed as to which discoursal requirements can be identified to account for the employment of such an anchor even if the event or state described in the temporal component is discourse-new. This section presents a corpus investigation that addresses this issue. 4.2.1. Method In the corpora analysed in the previous section, the amount and nature of knowledge about the interaction partner could not be controlled. Therefore, a different corpus was chosen for the present analysis of presuppositional effects. This decision also opens up the opportunity to put the suitability of the categories identified earlier to the test, and to identify further categories that may not have been detected earlier (possibly due to peculiarities of the text types involved). The data investigated here were taken from the Switchboard corpus distributed by the Linguistic Data Consortium (LDC),9 a large corpus of conversational speech over long distance telephone lines which involves numerous speakers representing a great variety of sociological and dialectal backgrounds. To collect the linguistic data, speakers (paid volunteers recruited from different companies and agencies) chose a topic about which they wished to talk, out of a repertory of 70 topics. Because of the high diversity of topics, the examples presented in this section contain a brief indication of the topic currently talked about. The participants are considered by the collectors to be broadly representative of adult 9.

http://www.ldc.upenn.edu; last visited: September 20th, 2006.

90 Temporal dimensions: Empirical analysis speakers of American English between 20 and 60 years of age, although a bias toward higher socioeconomic and educational levels was inevitable due to the task requirements. The LDC provides online access to transcriptions (without prosodic information) as well as to the original sound files of all utterances. Thus, the present analysis could be carried out by first searching for occurrences of before and after in the transcriptions and then listening to the original utterances in order to specify the prosodic organisation. This corpus is advantageous for present purposes since the speakers communicating with each other are not acquainted, and it can therefore be safely assumed that they do not draw upon information from previous encounters. Thus, personal knowledge (Clark 1996) which would otherwise be hard to control, is ruled out (in contrast to the corpora investigated in the previous section). Furthermore, since the speakers are talking on the phone they cannot refer exophorically to entities that are perceptually accessible to both partners. The only drawback is that the speakers were allowed to ‘warm up’ before they started the recording by pressing a button, resulting in (according to informants) a minimal amount of shared knowledge that is not available to the analyst. Additionally, of course, they share much knowledge about the world and the community they live in; it will be seen that such communal knowledge (Clark 1996) is indeed used continuously as shared background. To identify regularities in the circumstances of presenting Given or discourse-old, New or discourse-new information in temporal components, naturally occurring instances of before and after in spoken dialogue were analysed with respect to the following issues: x Common ground (based on textual context, communal knowledge, etc.) x Information conveyed in the temporal component: discourse-new, known, or inferable based on the previous discourse context x Linguistic indicators of information structure including prosodic organisation (Given/New distinction based on speakers’ meaning rather than knowledge) x Relationship between the two connected elements and the previous discourse (if relevant) x Listener’s reaction with regard to the information provided in the temporal component, be it Given, inferable, discourse-old or new, or (presented as) New.

Presupposition processes 91

These notions were analysed for each of the categories identified in Section 4.1, working out regular patterns in speakers’ and listeners’ linguistic behavior. 4.2.2. Results The qualitative analysis of spontaneous occurrences of the connectives supports the hypothesis that the categories of usage of the temporal connectives involve different degrees of Givenness in a discourse. In other words, the degree to which the information conveyed in the temporal clause needs to be accessible differs systematically between the diverse conceptual categories. In the following, a few exemplary analyses of short dialogues are presented in some detail in order to illustrate how speakers employ the connectives in context. Further analyses can be accessed in Appendix 1 in a more sketchy form, for further reference and illustration. However, the main body of the analysed dialogues that serve as the basis for the conclusions drawn below cannot be represented for reasons of space. In the following dialogue excerpt (interpreted as a case of IMMEDIACY: one TV program is presented as following another) the information given in the temporal component is presented as New by clear prosodic prominence10 on Who, followed by a pause and elaborated further by a prepositional phrase: (4.49)

(Context: sitcoms) B: (…)11 did you stay up late to watch this Red DWARF. It came on after Doctor WHO (pause 1 sec) A: Uh–huh. B: on Saturday NIGHTS, HERE at least.

The pause (along with a rising intonation) seems to indicate the speaker’s wish for confirmation of the new information, which is immediately given. By backchannelling, the listener informs the speaker that he has managed 10. Items with obvious intonational prominence in the original utterance are set in CAPS to provide an impression of the utterances’ prosodic organisation. A more fine-grained analysis would certainly reveal further differences in relative prominence that go beyond the scope of the present work. 11. Unfortunately for reasons of space some stretches of text have to be omitted. They do not contain information relevant to the present analysis.

92 Temporal dimensions: Empirical analysis to accommodate the information. In spite of the speaker’s obvious aim of having the information safely in place, it is nevertheless not presented as belonging to the main thematic thread of the discourse. It is presented only in a backgrounded prepositional phrase introduced by after, notably not a full temporal clause. Thematically, the speaker is talking about Red Dwarf, and anchors this temporally. To be able to do this, she needs the listener to confirm that he is indeed familiar with this background information, i.e., the interlocutors share common ground based on communal knowledge. The next dialogue excerpt (here interpreted in terms of CAUSALITY) illustrates the special impact of preposing a temporal clause. (4.50)

(Context: painting) A: (…) and then you can just PAINT right over the DARK. (…) After about, uh, half dozen COATS it looks REASONABLE but it kept, the, the dark it was, uh, just a real dark, uh, WOOD grain type PANELING B: Uh–huh.

Here, the temporal component provides discourse-new information, introducing new linguistic material (“about half dozen coats”) in a prepositional phrase, where both actor and process are omitted. This informationally sparse construction is sufficient for the introduction of new information for two reasons: First, prosody supports information transfer by the speaker’s hesitation (uh) before half dozen and by prominence on coats. Second, the information is inferable because its relationship to the previous discourse is transparent: As is common for preposed temporal components (Schilder and Tenbrink 2001), there is an ELABORATION relation to a part of the previous discourse. In this case, half dozen coats elaborate painting over the dark mentioned in speaker A’s previous utterance (cf. Schilder’s considerations of sub-part relations in Schilder 2001). Accordingly, the listener obviously has no problems accommodating the information. In the following example, the expected event is precluded by the main clause, yielding a non-veridical (INSERTION) reading of the before-clause (in which temporal precedence is less important than deontic precedence, i.e., desirability): (4.51)

(Context: Extending school years) A: (…) I understand there’s a lot of talk now about EXTENDING the school year which seems INORDINATELY short anyway, (…) well how do you think THAT would work in GRAMMAR schools and

Presupposition processes 93 B.: I think, uh, in, in SOME sense, um, [lipsmack] what I would do BEFORE I would extend a school year, is I would make it BETTER. BEFORE I would continue the AGONY, ((trailing off)) A: Yeah, yeah, that’s right [laughter].

Here, the speaker apparently wishes to emphasise that the event referred to in the temporal clause should be precluded. This is indicated by the double construction, where, in the first version, the speaker simply repeats the wording of previous utterances to point to the expected event. The second version, in contrast, elaborates the first statement by a clearly negative assessment of the event (agony). Accordingly, the first version only carries some weak prominence on before and better, while the second version shows a fairly strong prominence on agony. Then, the speaker does not finish the sentence, since the main proposition (the inserted event) has already been given and does not need to be modified. The listener explicitly acknowledges the speaker’s evaluation. As indicated in Section 4.1.2, Schilder (2001) describes the capability of before to indicate a TERMINATION relation between the main and the temporal clause; for some reason (possibly related to text type), no example for this category was found in the data analysed in that section. In the Switchboard data analysed here, a TERMINATION relation was found to be fairly frequent, especially with preposed temporal components. TERMINATION only occurs with before, and is indicated by the use of duratives (descriptions of states rather than events) in the main clause. The concept supported in this construction can be specified further as follows: either x the event in the temporal component terminates the state of affairs as described in the main clause where the temporal component describes an event, or x the state described in the temporal component was different to the state of affairs as described in the main clause where the temporal component describes a state. The following Switchboard example illustrates the first of these options. (4.52)

B.: (…) Before I became a mother I was going to law school.

As already suggested in Section 3.5, TERMINATION can be tested for by reformulating the utterance using until: (4.53)

(4.52)’ Until I became a mother I was going to law school.

Since this sentence is equally suitable in the given discourse context, clearly a reversed causal relationship is conveyed here: becoming a mother

94 Temporal dimensions: Empirical analysis was the reason for terminating law school. (The later discourse indeed confirms this interpretation.) To convey this concept, it is not necessary that the event or state in the temporal component is known beforehand, as the following example illustrates. (4.54)

(Context: Gardening) B.: Yeah. BEFORE I moved down here, when I was up, uh, north in Wisconsin living with my folks, THEY lived on, uh, three and a half acres of LAND, and, uh, we started mowing that with a PUSH mower. A.: Oh.

The context of this piece of conversation indicates that the speakers know vaguely about each other where they live, although there was no mention of moving house. Thus, the information given in the temporal clause is discourse-new but can be accommodated easily. Prosodically, there is almost no prominence at all in the temporal clause, the main emphasis clearly being on land in the main clause. The speaker is talking very fast, giving the impression that he wishes to be finished with providing the necessary background information in order to be able to get to his main point, where he considerably slows down his speech. He is, however, obviously fully aware that the listener needs the information, as he gives a further evaluation of it in the paralleled when-clause. The listener does not react to this information directly, but backchannels on the later main proposition. The data contain frequent occurrences like the previous one where discourse-new information is presented as Given in a temporal component. Usually, such information is accommodated on the basis of communal knowledge: the events or states mentioned are simply not unusual, although the interlocutor could not know them to be the case prior to mention. The next example (interpreted as a case of REGULATION: the temporal order itself is focused upon, while the described situations are treated as Given) is similar in this respect. (4.55)

(Context: pets) B.: I had one cat for EIGHTEEN years. I GOT her when I lived in Hawaii. (…) BEFORE I even got married I HAD my CAT. A.: She’s old.

From the previous context, the listener cannot know that the speaker is married. However, this information can be easily accommodated based on communal knowledge (many people get married). The information is not

Presupposition processes 95

foregrounded even though it is discourse-new; the strongest prominence is on cat. The importance of the temporal information is indicated by even. The speaker’s intention in this context seems to be to convey the length of time that has passed since she got the cat. This is immediately supported by the listener’s subsequent remark. Thus, the speaker has in effect ‘imposed’ knowledge on the listener that is not being questioned. However, a different reaction is possible, as in the following example: (4.56)

(Context: Studying) A.: So, are, are you going to go for your DOCTORATE? B.: Probably not. See I’m in the NAVY. A.: Yeah. B.: And, uh, you know, unless the Navy FUNDS it I probably won’t do it until I, if I DO do it, it will be after I get out of the Navy. A.: When do you get OUT? B.: Oh, probably be another TWELVE years. A.: Oh, okay [laughter].

Again, the information in the temporal clause is treated as Given – carrying only very little prosodic stress on after and Navy – even though it is discourse-new. On the basis of communal knowledge, the information that the speaker will get out of the Navy is inferable, though when that will be is not. Therefore, the listener asks for this information, which is then provided by the speaker. Interestingly, this information is indeed important for understanding why speaker B assumes that he will not do his doctorate, since the time span in question is very long. 4.2.3. Discussion The analysis of natural language data has shown that the information in the temporal component need not in all cases be known beforehand, in spite of its being classified (in some accounts in the literature) as presuppositional. The examples discussed here did not appear to be in any sense exceptional; on the contrary, they serve to illustrate a pattern that was systematically revealed by the presupposition analysis: Before and after cannot be said to mark common ground in the sense that the speakers linguistically express what they believe to be their shared knowledge basis, e.g., because it has occurred earlier in the discourse. Discourse-new information presented in temporal components is usually easily accommodated on the grounds of

96 Temporal dimensions: Empirical analysis communal knowledge, or because of its relation (e.g., an ELABORATION discourse relation) to the previous discourse context, as is specifically often the case in preposed temporal components. In some cases, however, the speaker uses temporal components to refer to information that is only treated as Given, although it is actually new and unrelated to the previous discourse or other knowledge. The hearer then has the choice to accept the new information explicitly or inexplicitly, to ignore it or to challenge it. Assessing the speaker’s expectations about the information status of the hearer cannot be achieved in an uncontroversial way. Linguistic devices only reflect the speaker’s presentation of elements as Given or New. Since the data source used for the analysis was particularly suitable for determining how much the interactants know about each other, it could be shown that – in contrast to some previous approaches that do not recognise the necessary distinction between New and discourse-new elements – the hearer’s information status (or the speaker’s assessment of it) is not straightforwardly mirrored by the use of linguistic expressions or prosodic prominence. Instead, some pieces of information are completely new in the discourse and still backgrounded linguistically and prosodically. But information that has been mentioned before can equally well receive prosodic emphasis and linguistic foregrounding. Thus, these devices may or may not be used to indicate the speaker’s expectations with regard to knowledge on the part of the hearer; rather, they seem to reflect the speaker’s intention to place specific emphasis on parts of the linguistic material (in accord with the Hallidayan analysis of Given/New structure). Emphasis can be placed on elements of the temporal component, but the thematic thread of the discourse is mostly carried by the main clause. The various kinds of concepts supported by before and after were found to exhibit the following features, showing different degrees of accessibility of the information conveyed in the temporal component. Further examples for each of these concepts can be found in Appendix 1. Explicit length of time. Explicit statements about length of time do not require the information in the temporal component to be Given, but it should be inferable as it is not presented as important to the main proposition. The main point of the clause complex is either the time span itself or the proposition in the main clause which is supplemented by the information about the length of time. Listeners can challenge or evaluate the information in the temporal component but usually do not do so, following the main thread of the ongoing discourse.

Presupposition processes 97

Immediacy. Immediate succession reflects the speaker’s conceptualisation of events following one another at one level of granularity. The event in the temporal component may be known beforehand, inferable or discoursenew. If it is known, then the utterance anchors a ‘previous’ or ‘next’ event in time in relation to the known event. If it is inferable or discourse-new, then the speaker anchors an event s/he is talking about in time, establishing the reference time as predecessor or successor ‘on the spot’. Causality. Where after is used to express a causal relation between two events (or states), the main emphasis is on the result rather than the cause, which is reflected via the fact that the cause is packaged syntactically into a backgrounded component. Thus, the cause introduced by after does not seem to receive the interactants’ attention. It is possible that discourse-new information is introduced as a cause, but the result is more important. In the data examined in the present analysis, discourse-new information given in the temporal component could be accommodated without problems and was not questioned by the listener. Notice that the causal relation itself is even less challengeable for the listener, since it is not expressed explicitly but needs to be inferred by the juxtaposition of two otherwise unrelated events. Accordingly, no cases were found in which the listener questioned a causal relationship implied by the employment of after. Dependency. DEPENDENCY relationships between the elements connected by temporal terms occur whenever a conditional relationship between events, or a weaker sense of preparation, can be identified. Since the precondition can be either of the events, either the precondition or the event that is based on it can be foregrounded. In either case, the backgrounded event may be known, discourse-new or inferable; but since the speaker focuses on the foregrounded event, the listener naturally follows their train of thought without challenging inferred information. Insertion. In cases of INSERTION, the event in the main clause is inserted between the present moment and the event referred to in the temporal component, which is either expected based on the previous context, or else treated as expected, i.e., presented as Given. Usually, if the information itself is not explicitly stated in the discourse history, it can be accommodated easily. The event in the main clause may prevent the event in the temporal component: in non-veridical instances, this is obviously the case,

98 Temporal dimensions: Empirical analysis while in indeterminate cases, the expected event may or may not take place, this becomes unimportant because of the inserted event. Regulation. In REGULATION, the temporal order of two expected or known events is focussed upon. Usually, the information of both connected elements is discourse-old or easily inferable. In any case, it is linguistically treated as Given. If it is in fact new to the listener, s/he can either accommodate the presupposition or question it. Termination. Instances of TERMINATION are indicated by duratives in the main clause; they support the notion that some kind of change has occurred at some point in time. The dependent component provides the temporal information relevant to the change. It can consist of discourse-new, old, or inferable information; since it is clearly backgrounded linguistically, a congruent listener’s reaction is not to react to it at all. In the investigated data, no listener gave any comment on a before-component indicating TERMINATION. In accord with previous accounts regarding subordinate clauses, it is common to all concepts supported by before and after that the situation described by a temporal clause (or reduced construction) is backgrounded, i.e., seldom focused upon in the discourse. This is mirrored by a nearly consistent lack of reaction on the part of the listener with regard to the information packaged in a temporal component, as well as a general tendency for the information to be inferable on the basis of the given context or communal knowledge. However, the degree to which New information can be conveyed in a backgrounded temporal component differs systematically, reflecting a continuum of inferability (or accessibility) in the discourse context. On one end of the scale, the two connected elements are nearly equivalent with regard to the ‘newness’ of information: this is the case in IMMEDIACY, where in some cases events are listed as happening one after the other without there necessarily being any previous knowledge with regard to either event. Similarly, in cases of EXPLICIT LENGTH OF TIME, TERMINATION, CAUSALITY and DEPENDENCY the main emphasis lies on the semantic relation between the events or the expressed time span, but there is no need for the backgrounded event to be previously known to the hearer. Here, however, the focus is more clearly on the main clause. On the other end of the scale, in INSERTION and REGULATION the information in the temporal component is regularly presented as Given. The two catego-

Presupposition processes 99

ries differ in that in REGULATION, also the information in the main clause is treated as Given. It seems, then, that the information in a temporal component should not in all cases be classified as presupposed, since at least in cases of IMMEDIACY information that does not previously belong to the common ground can be conveyed straightforwardly in the temporal component, as in one example taken from a novel:12 (4.57)

She kissed them both before she brushed their hair.

This example is extremely balanced with respect to newness of information in the two clauses. However, such balanced cases turned out to be rare in the conversational examples analysed here. Overwhelmingly, people took care to present only easily inferable information even in cases of IMMEDIACY. The following example (interpreted in terms of IMMEDIACY as indicated lexically by right) illustrates this point: (4.58)

(Context: credit cards) B.: Yeah, WE’VE tried to do that, WE’VE paid ours off, you know, all the way down to where we had EVERYTHING down to zero and ESPECIALLY right before I, I quit work two YEARS ago, to stay home with the KIDS and right, and BEFORE I left we had everything paid off, we were in GREAT shape, we were putting money, you know, because we were both WORKING, A.: Uh–huh.

The speaker starts introducing information in a temporal clause but stops in mid-utterance to start anew. It is clear from the prosodic curve of this reformulation that the speaker now starts a new clause in order to introduce discourse-new information (I receives prosodic stress after the restart), rather than continuing the temporal clause. The temporal clause is then formulated only later after having made sure that the listener knows what the speaker is talking about. 4.2.4. Conclusion The present analysis of naturally occurring spoken discourse addressed the question as to how far temporal markers trigger presuppositions in the 12. John le Carré, The Russia House.

100 Temporal dimensions: Empirical analysis pragmatic sense, given that the formal semantic approach to presuppositions claims that temporal clauses are always presuppositional. By investigating the discourse context speakers’ actual strategies could be addressed with respect to whether the information introduced by a temporal dimensional term is generally already known, as could be expected by the pragmatic interpretation of presupposition. The results show that this is indeed often the case, although discourse-new information can unproblematically be presented in a backgrounded temporal component if it is easily inferable. However, the degree to which the information needs to be accessible in order to support a particular concept differs systematically between the different concepts supported by before and after. In particular, INSERTION and REGULATION require high accessibility of the information in the temporal component, while IMMEDIACY is much less restricted. 4.3.

Relations between events: Analysis of German data

4.3.1. Material and Procedure In this section, German natural language data are analysed to address the research issues discussed so far for English. For this purpose, Prof. Erhard Mergenthaler of the University of Ulm kindly provided me with an extensive amount of data of the Ulmer Textbank.13 Other suitable German corpora of dialogic spoken language were not accessible to me at the time of research (see Schmidt 2005 for a recent essay on the availability of German dialogue corpora). The corpus contains a collection of transcribed initial dialogic interactions between a psychotherapist and a new patient. This background is especially suited for addressing the presupposition issue since the interactants are not familiar with each other, so that most information that is new to the discourse is also new to the hearer. The only exceptions are general world knowledge, which is to a high degree also available to the analyst, and possibly some preliminary information about the patient that the psychotherapist had access to (but such information is, in the doc13. I was asked to provide the following information: “The clinical material was made available by the Ulmer Textbank. In order to protect personal data, a complete representation of the sources is not possible. For requirements of scientific interests, access to the data is possible at the Department for Psychosomatic Medicine and Psychotherapy at the University of Ulm, Am Hochsträß 8, 89081 Ulm.”

Relations between events: Analysis of German data 101

tor–patient dialogues, usually not taken for granted but re-checked if considered important). However, since the corpus made available to me consisted entirely of transcribed data, the intonation contour could not be used for analysis, which reduces the depth of the analysis of the presupposition issue (in comparison to the corresponding analysis of the English data presented in Section 4.2). Furthermore, topics focused solely on therapeutic issues, and the relationship between the interactants is naturally asymmetric (doctor and patient). In the present section, only one syntactic form will be analysed, namely, temporal dimensional terms occurring as conjunctions. This has practical as well as theoretical reasons. On the one hand, there is a current major interest in Germany in the semantic and pragmatic features of discourse connectives (e.g., Pasch et al. 2003). On the other hand, the German language is full of occurrences of both vor and nach that are interpreted in ways not analysed here, due to their inherent polysemy and frequent occurrence in phrasal constructions. For example, nach can also have a spatial interpretation when used together with a spatial dimensional term, as in nach rechts (to the right) or nach hinten (to the back). The range of applications of both terms as prepositions is rather wide; while this would be an interesting topic of research in itself, it is well beyond the scope of the present work. 4.3.2. Results Concerning the non-temporal concepts conveyed by the temporal terms, basically the same relationships as identified for English were found to be supported by bevor and nachdem as well; additionally a new conceptual category (No Change) was identified. The categories are addressed in more detail in the following subsections. 4.3.2.1. Causality The present discourse type turned out to be specifically interesting with respect to causal aspects, since it is one of the aims of psychotherapeutic dialogues to identify reasons and causes for psychological problems in the life story of the patient, so that temporal and causal aspects are specifically viewed as interrelated and often explicitly reflected in the dialogue.

102 Temporal dimensions: Empirical analysis The suggestion of a causal relationship can clearly be supported by nachdem. The causal relation is not in all cases conclusive or unquestionable, but juxtaposing two events using nachdem often strongly suggests it. This effect is transparent in the following example, where the inference of a causal connection is made explicit by the ensuing clarification question: (4.59)

ja, das eben, daß es mir schlechter ging und dann, äh, wußt ich nicht mehr, was ich tun sollte. und dann wollte ich Schluß machen irgendwie. nachdem es Ihnen schlechter gegangen ist? ja. und nur wegen der Krankheit? ja. ‘ yes, that, well, that I was getting worse and then, uh, I did not know any more what I should do. and then I wanted to end it all somehow. after you were getting worse? yes. and only because of the illness? yes.’

Note that the proximal time frame induced here has indefinite value with respect to the future, since the illness could theoretically provoke a suicidal tendency that lasts indefinitely (or could stop at any time, maybe influenced by some further event). In the next example the proximal time frame is reduced to an interval vaguely defined by gestern Mittag. A causal relationship is explicitly inferred by P2 on the grounds of the temporal coincidence, even though it might not have been intended by the original speaker, P1: (4.60)

gestern mittag waren es neununddreißig sechs. nachdem Sie bei mir waren? ja, (...) na ja, aber das bedeutet, daß diese Auseinandersetzung mit mir gestern, nicht, ziemlich eine –? hmhm. Spannung in Ihnen hervorgerufen hat ‘ yesterday at lunchtime it was 39.6. after you were with me? yes, (…)

Relations between events: Analysis of German data 103 well, but that means that this conflict with me yesterday has, you know, produced kind of a –? hmhm. strain in you’

In other situations, a causal inference is obviously implied by the speaker by juxtaposing facts: (4.61)

(4.62)

wissen Sie, so nachdem ich ja also zehn Jahre jetzt nicht, oder mehr sogar, Patienten mit Kolitis, nicht, und ähnlichen Sachen, Sie, hab ich etwas Erfahrung, wie sehr man vorsichtig is, wenn man über seine Sachen redet, ne. ‘you know, after I now have ten years, or even more, patients with colitis, you know, and similar things, you, I have some experience how very cautious people are in talking about their affairs, you know.’ erstmal ist alles so gespannt und nachdem ich mir dann den Schmerz zugefügt habe, weil, es zieht dann unheimlich im Kopf, habe ich einen Moment Erleichterung ‘at first everything is so tense and then after I have inflicted pain upon myself, because it tugs heavily inside the head then, I have a moment of relief.’

Note that, in all of these examples, the temporal relationship between the mentioned events is still valid. Often, the causal aspect is open to interpretation, as in: (4.63)

also jetzt kenne ich mich nicht mehr aus damit, also daß ich, hm, im Jahre neunzehnhundertneunundsechzig im Frühjahr, nachdem meine Verlobung geplatzt war, daß ich da auch eine Augenentzündung hatte mit starkem Fieber ‘well, I now don’t know any more, well, that I, hm, in the year 1969 in the spring, after my engagement had broken up, that I then, as well, had an eye inflammation with high fever.’

where the speaker declares that s/he does not know how to interpret the sequence of events. No similar (clearly causal) concepts could be detected for bevor. Using bevor instead of nachdem in the above examples deletes the potential for a causal connection, for example: (4.64)

(4.59)’ Bevor ich Schluss machen wollte, ging es mir schlechter. ‘Before I wanted to end it all I was getting worse.’

104 Temporal dimensions: Empirical analysis Here, a concept of TERMINATION may be inferred: The patient felt bad until s/he wished to end his/her life. Thus, the causal relation is reversed – the reason for terminating the state in the main clause is given in the temporal clause. Transforming example (4.61) into a bevor-sentence is interesting: (4.65)

(4.61)’ Bevor ich etwas Erfahrung habe, wie sehr man vorsichtig ist, wenn man über seine Sachen redet, habe ich zehn Jahre Patienten mit Kolitis [behandelt]. ‘Before I have some experience how very cautious people are in talking about their affairs, I have [treated] colitis patients for ten years.’

This construction seems to convey that it took the speaker as long as ten years to learn, which is surely not what s/he wanted to communicate. The reason is that a direct causal connection is not inferable, and that, furthermore, the sense of temporal overlap between the therapeutic work and the ensuing growing wisdom also gets lost. 4.3.2.2. Dependency The concept of DEPENDENCY can be supported by both bevor and nachdem. As the following examples show, linguistic markers like müssen (must) and fähig (capable) are frequently employed to reinforce the underlying necessity for some event to enable a different one. (4.66)

(4.67)

(4.68)

dann k– muß man erst mal aufräumen eine halbe Stunde bevor man hin kann zum Essen ‘then you have to tidy up first for half an hour before you can go to eat’ brauch ich hinterher wieder ne Zeit zum zur Ruhe zu kommen (…) bevor ich überhaupt fähig bin mich auf irgendwas anderes zu konzentrieren. ‘afterwards I need some time to calm down (…) before I am at all capable of concentrating on anything else.’ man muß das erst mal erleben und innerlich abklären, bevor man etwa überhaupt aufschreibt. ‘you first have to experience that and clarify it inwardly before you can write it down at all.’

With negation, it is pointed out that without the event in the temporal clause the event in the main clause will not be enabled:

Relations between events: Analysis of German data 105 (4.69)

Sie kriegen Ihre innere Ruhe nicht, bevor Sie nicht dieses Problem gelöst haben. ‘You cannot calm down before you have not solved this problem.’

Blühdorn (2002) suggests that the (very common) double negation in such sentences is due to a neutralisation of the temporal factor. However, this interpretation is problematic in that there is a clear temporal relationship involved: after the problem is solved, peace of mind can be re-gained. I would suggest that the double negation indicates a relation of negated TERMINATION: there can be no TERMINATION of the problem in the main clause as long as the problem described in the temporal clause is not solved. Thus, the two relations (TERMINATION and DEPENDENCY) are interconnected in this specific example. Certainly, the impact of negation on the interpretation of temporal clauses would be an interesting but complex research area in its own right, which is not pursued further here. With nachdem, often no additional markers are used for conveying DEPENDENCY. In the following example, telling a person about a specific topic enables discussing the topic: (4.70)

wir haben ja, nachdem ich’s ihr gesagt hab, oft genug dadrüber geredet. ‘you know, we have, after I have told her that, talked about that often enough.’

Here is another example containing temporal overlap between the events: (4.71)

ich bin dann aus dem Krieg wieder zurückgekommen (…) und habe dann eben, nachdem man wieder so leben konnte und also über das Essen und Trinken hinaus wieder was anderes denken konnte nun wieder versucht, diese Dinge in Gang zu bekommen ‘I came back from the war then (…) and then I have, you know, after it was possible to live again, and to think about something else besides food and drink, again tried to put these things into operation’

A weaker form of DEPENDENCY is identified with relationships of sensible preparation: (4.72) (4.73)

ich werde Sie einbestellen, nachdem ich mit ihm gesprochen habe ‘I will appoint you after I have talked to him.’ bevor er zur Welt kam, machten Sie sich über die Schwangerschaft der Mutter Gedanken vielleicht? ‘before he was born, were you concerned about the mother’s pregnancy perhaps?’

106 Temporal dimensions: Empirical analysis In both cases, the earlier event seems to be a sensible preparation for the later event. 4.3.2.3. Termination As indicated above, TERMINATION (which is only possible with bevor but not nachdem) can be described as conveying a reversed causal relationship. In this case the event in the temporal clause is responsible for terminating the situation in the main clause, as in: (4.74)

ich war vorher in einer Clique drin gewesen, bevor ich krank geworden bin, aber das hat sich jetzt zerschlagen. ich kann nicht mehr mitmachen so ‘I had previously been in a clique, before I became ill, but that has ended now. I cannot join in anymore like that’

It is also possible to ask how the situation was before it was terminated by the event in the temporal clause: (4.75)

wie haben Sie das denn so gemacht bevor Sie den Unfall hatten? ‘how did you do that before you had the accident?’

4.3.2.4. Insertion The use of German bevor mirrors the English usage fairly directly as regards the concept of INSERTION. Here is an example: (4.76)

wir können, bevor wir unterbrechen, äh, Sie noch etwas fragen, ob Sie träumen manchmal? ‘we could, before we break, uh, ask you something, whether you dream sometimes?’

INSERTION requires that the information in the temporal clause is known to the hearer (cf. Section 4.1.2). In the above example, this is the case because the speaker had previously announced that there would be a break. But in the next example, s/he had not done that: (4.77)

ja, nun bevor wir dieses Gespräch jetzt, nicht, vorübergehend unterbrechen, um miteinander hier zu beraten, wollte ich Sie gerne noch eine einzige Sache fragen

Relations between events: Analysis of German data 107 ‘yes, well, before we now interrupt, you know, this conversation temporarily, in order to debate among ourselves here, I would like to ask you one single thing’

Accordingly, there is a linguistic marker of hesitation and request for confirmation by the interaction partner (nicht) in order to compensate for the fact that information is treated as Given that actually is new to the discourse, and cannot be inferred. The next example is very similar in content but shows a different strategy: the speaker first introduces the future event as new information, than treats it as Given in the temporal clause in order to insert a further event in between the present and the future moment. (4.78)

ich werde sicher jetzt, äh, das Gespräch einen Moment unterbrechen, aber bevor ich uns unterbreche will ich Sie noch etwas fragen ‘I will now surely, uh, interrupt the conversation for a moment, but before I interrupt us I want to ask you something’

INSERTION cannot be supported by nachdem, as exemplified as follows: (4.79)

(4.78)’ (…) ? Nachdem ich Sie noch etwas gefragt habe, unterbreche ich uns. ‘? After I have asked you something I interrupt us.’

which sounds odd in the discourse context. As in English, as hypothesised above, bevor can occur in non-veridical contexts, as the following example (which contains temporal overlap) shows: (4.80)

ja, vielleicht sollten wir, jetzt, bevor wir in ner allgemeinen Diskussion (P lacht) jetzt sind und sind von Drogen, eh, so weit gekommen, doch einmal gucken was, was mit diesem / – ob wir da, ein Stück weiter sehen können. ‘yes, maybe we should, now, before we now are in a general discussion (P laughs) and have now deviated from drugs, uh, so much, now look what, what with this / – whether we can look a bit further ahead.’

Seeing that the patient drifts into a general discussion, this speaker tries to constrain the topic to more relevant issues, in order to prevent the general discussion from being established. The temporal overlap is marked linguistically by the present tense, indicating that the general discussion is about to begin, which again proves that the event in the temporal clause is expected to happen. Thus, the example can be reformulated in the way suggested in Section 4.1.5 without fundamental meaning change:

108 Temporal dimensions: Empirical analysis (4.81)

(4.80)’ ja, vielleicht sollten wir, jetzt, anstatt dass wir jetzt in eine allgemeine Diskussion geraten, doch einmal gucken was, was mit diesem / – ob wir da, ein Stück weiter sehen können. ‘yes, maybe we should, now, instead of getting into a general discussion, now look what, what with this / – whether we can look a bit further ahead.’

4.3.2.5. Regulation REGULATION is used to focus on the temporal order itself where the described situations in both main and temporal clause are known, as in: (4.82)

und im Grunde, also Sie haben die Blasenbeschwerden seit dem dreizehnten Lebensjahr, während diese ganzen Beschwerden die traten alle auf – . ja, nachdem ich verheiratet war, ja. ‘ and in a sense, well, you have had these bladder troubles since the thirteenth year of life, while all these troubles they all occurred – . yes, after I was married, yes.’

The discourse coherence would remain unchanged14 if nachdem was substituted by bevor: (4.83)

(4.82)’ (…) diese Beschwerden traten alle auf, bevor ich verheiratet war. ‘(…) these troubles all occurred before I was married.’

With REGULATION, therefore, nachdem and bevor can be regarded as semantic converses, since no specific semantic relationships are inferred. 4.3.2.6. Explicit length of time Similar to REGULATION, providing an EXPLICIT LENGTH OF TIME – even a qualitative measure such as kurz (shortly) is sufficient – typically leads to a focus on the temporal relationship itself, rather than the events or a nontemporal relation between the events. In the following example, the temporal relationship indicated by the first speaker is denied by the interaction partner, which is rarely the case if a nontemporal relationship is in focus: 14. Unless a causal connection is inferred here as well, which is possible.

Relations between events: Analysis of German data 109 (4.84)

hmhm – äh das war äh also kurz nachdem Sie von zu Hause weggegangen sind damals, +nicht? nein,+ das könnte man nicht sagen. ‘ hmhm – uh that was uh well shortly after you had left home at that time, right? no, you couldn’t say that.’

4.3.2.7. Immediacy In the following example, IMMEDIATE SUCCESSION is indicated lexically by sofort (instantly): (4.85)

und ich darf auch sagen, daß mir Schwester *1442 sofort gesagt hat, nachdem sie raufkam wieder zu mir, nachdem sie Sie runterbegleitet hat, –. ja. daß Sie Empfindungen an den Füßen, oder wie war das, bekommen haben. ‘ and I may also say that nurse *1442 instantly told me, after she had returned upstairs to me, after she had accompanied you downstairs, –. yes. that you had got sensations in the feet, or how was that.’

Another marker for IMMEDIATE SUCCESSION is gleich (immediately): (4.86)

weil ich praktisch gleich nachdem meine Lehre zuende war ähm hatte ich dann eine Zweigstelle übernommen. ‘because I almost immediately after my training was finished uh I had taken over a branch office.’

4.3.2.8. No change (Denial of causality) One of the results of the present analysis is that a new concept has been detected, supported by bevor, which is not described in the literature on either German or English terms, namely that of NO CHANGE. This finding is viewed as an effect of investigating further corpora, i.e., dealing with an increasing amount of natural language data, rather than as a peculiarity of a

110 Temporal dimensions: Empirical analysis particular text type or language. This assumption, of course, requires further investigation. The category of NO CHANGE captures situations in which a state was present even before an event that could have caused it. In the following examples, this is supported linguistically by the additional linguistic markers schon zuvor auch (already before that) and überhaupt (even): (4.87)

(4.88)

aber die Beschwerden schon zuvor auch. mehr oder weniger, da. sagen wir mal so. bevor nun die Mutter bei uns wohnt ‘but the troubles already before that as well. more or less, then. should we say. before the mother now lives with us’ die hat sich auch Sorgen gemacht, bevor sie überhaupt erfahren hat, daß ich ausziehe. ‘she was worrying before she even learned that I was moving out.’

This concept is akin to REGULATION since the temporal order is important, and both situations need to be known to the discourse. However, it differs in it by the element of contrast to expectation. In this way, a causal relationship between the two events is denied. Since it is possible to translate these examples straightforwardly to English without loss of coherence and associations, I assume that the same concept can also be supported by before. Possibly, this concept (which cannot be subsumed by any of the other categories) depends entirely on the presence of additional linguistic markers such as those present in the above. In that case, the linguistic items jointly contribute to conveying the concept of NO CHANGE. 4.3.2.9. Presuppositional effects With regard to presupposition, no fundamental differences to the English results could be found. In many cases in the data, the proposition in the temporal clause is already known from the discourse context, satisfying the presupposition. In other cases, the information was discourse-new but could be inferred, as in the following example: (4.89)

da hab ich mal vor Jahren gearbeitet, noch bevor ich studiert hab ‘I worked there years ago, even before I studied’

Here, accommodating the information that the speaker has studied is no problem since this is a normal part of many people’s lives. The case is substantially different in the next example:

Relations between events: Analysis of German data 111 (4.90)

nachdem man mir jetzt schon zum wiederholten Male gesagt hat, daß es vom Neurologischen fast in Ordnung sei, die Laufbeschwerden aber noch ziemlich stark sind, hat man mich darauf hingewiesen, daß es eben psychische Ursachen haben könnte, die da mit rein spielen. ‘after I’ve now been told many times that from a neurologic point of view almost everything is okay, although the problems with walking are still fairly heavy, they have pointed out to me that psychological reasons may also come into play.’

This case, which can be interpreted in terms of DEPENDENCY, exemplifies what Blühdorn (2002) recognises as factual, but completely without discourse anchor. The information is discourse-new and not inferable from the context. However, it is backgrounded in the discourse and tacitly accepted by the interlocutor, i.e., there is no reaction. In general, it can be seen from the data that, corresponding to the findings for English in Section 4.2, this effect is not possible in all discourse situations. In a situation where the temporal order itself is in focus (either REGULATION or NO CHANGE), it is required that both events described in the main and temporal clause are known or easily inferable. Accordingly, speakers can be found to repair their own usage if they introduce new information, as in the following case of NO CHANGE: (4.91)

bevor ich das erste Penicillin genommen hab, ich hab da so große Amp– äh, Tabletten, Penicillin bekommen, äh, war das an sich schon abgeklungen ‘before I had taken the first penicillin, I have got these big amp– pills, penicillin, uh, it had already died down in a way.’

Here, the striking fact that the health problems were gone before taking medicine is focused upon. During production of this utterance the speaker notices that s/he has not mentioned having taken medicine before, and therefore inserts a new clause in which this information is presented linguistically as new. Similarly, the speaker can repair the utterance by providing further information that is known to the hearer, which in the case of example (4.85) above is confirmed by the hearer. 4.3.3. Conclusion The application of bevor and nachdem in natural discourse, as investigated in the present section, mirrors to a large degree the results concerning the

112 Temporal dimensions: Empirical analysis English terms presented above. However, further insights could be gained on the basis of the additional data, which may well also be valid for English (this, of course, again needs investigation). Many of the possible relations between the two events or states involved are causal in some respect: x CAUSALITY: nachdem can be used to support a direct causal relationship between the two juxtaposed situations. x DEPENDENCY (strong form, conditionality): nachdem and bevor can be used in contexts where the occurrence of the later situation depends on the occurrence of the prior situation. x DEPENDENCY (weak form, preparation): nachdem and bevor can be used in contexts where the prior situation is a sensible preparation for the later one. Which of these relationships should be assumed is often very much subject to interpretation. Speakers often declare explicitly that they are unsure about the underlying relationships between events. Thus, the categories should be viewed as different values on a scale, with the exception that with bevor, direct causal relations are less readily inferred. Instead, bevor can express some further concepts which may also be causal: x TERMINATION: bevor can be used to convey that a situation ends because of – or at least coinciding with the beginning of – another situation. x NO CHANGE (denial of causality): bevor can be used to indicate that one situation is not the cause of another situation (because it happened earlier). x INSERTION: bevor can be used to support the notion that an event is inserted between the present moment (or a textually projected moment in the past) and an expected event, which sometimes prevents the expected event from occurring. In such non-veridical cases the inserted event is the cause for the expected event not to happen. Moreover, nachdem and bevor can be used to focus on the temporal order between two events (REGULATION) that are already known to the discourse participants; and they can be used to express a temporal sequence (IMMEDIACY) which is dependent on a clear conception of an underlying granularity level where one event happens directly after another. Lastly, they can be used together with EXPLICIT TIME measures to provide a direct specification of the temporal relationship between two situations. In all three cases, no causal or near-causal relationship between the events is required, but can still be inferred.

General Discussion 113

4.4.

General Discussion

What are the applicability conditions of temporal dimensional terms in English and German? In this and the previous chapter, this question was approached from several directions, taking a pragmatic perspective. The results revealed a fairly consistent pattern, both across the two closely related languages and across text types. The latter result is especially remarkable because of the limitations and peculiarities associated with some of the sources, for instance, the asymmetric interaction data in the psychiatric interviews (Section 4.3), or the comparison of child-directed speech versus highly intellectual political discourse (Section 4.1). From the data, it has become clear that speakers typically do not simply use the terms to denote the temporal relationship of unrelated events, but that the terms support the inference of an underlying semantic relationship between the events. Additionally, the temporal relationship of known events (whatever their relation may be), or the length of the time span, might be focused upon. These results accord with the structural resources of language: the English and German languages offer many options for expressing temporal relations, such as tense and aspect, temporal adverbials, or implicitly conveying the information via clause order. Thus, if temporal dimensional terms are employed, speakers may intend to convey more than just temporal information. Here, one simple reason for employing an explicit relational term is that the temporal order is not only conveyed but is in focus, as is the case in the categories of REGULATION and NO CHANGE. However, also a number of possible conceptual relationships between the events have been identified and discussed in the present chapter, many of which are causal in some sense. This fact reflects previous findings in cognitive science on the relationship between time and causality (Section 2.3.3): since we, as humans, experience time as a network of causally interrelated events, it is only natural that our linguistic manifestations of temporal relations should be closely related to notions of causality. However, the employment of temporal terms while implying causal relations seems to entail that the speaker wishes to leave some amount of freedom to the listener’s interpretation: otherwise, they could have employed explicit causal relational terms such as because. This observation is in accord with the fact that speakers (in the present corpora) sometimes explicitly mention their uncertainty about the nature of the interrelationship between two events. In addition, speakers can always intend to suggest even more than what is received by the recipient (or the analyst). But in

114 Temporal dimensions: Empirical analysis many cases the underlying relation is supported and further specified by the larger linguistic context – either in the connected segments themselves, or in the surrounding discourse. Thus, the present analysis has attempted to clarify the contexts in which temporal dimensional terms appear, indicating the principles underlying their application. Some semantic relationships are supported by the application of the temporal terms in some kinds of contexts, and others are discouraged. Typically, more content is intended by – and interpreted into – an utterance than merely a temporal relationship, which is the most obvious and uncontroversial aspect of the semantics of temporal dimensional terms (cf. Section 3.4). Since speakers typically do not relate two previously unknown events linguistically, the present chapter also addressed the question of how speakers deal with different degrees of Givenness in discourse. The two research directions are interrelated in that specific kinds of conceptual categories pose different degrees of restrictions on Givenness. This was spelled out in some detail for English. In general, it can be concluded that the information conveyed in the temporal component is mostly known or inferable from the context or discourse external knowledge; but this is not a strict requirement for the application of temporal dimensional terms. Most of the categories identified in the present chapter are not entirely new; they have turned up in a number of ways in the earlier literature. This applies most obviously for causal relationships between events related by temporal markers (e.g., Heinämäki 1974, Lascarides and Oberlander 1993, Schilder 2001). But interestingly, Heinämäki (1974: 63) also proposes a categorisation of five cases of before-clauses, which is in later chapters not employed in any consistent way, and the analysis does not seem to have been put to the test on natural language data – but the concepts mentioned in this categorisation are very similar to some of the concepts described in the present chapter. The first of these categories is described as noncommittal and not involving causality, as in:15 (4.92)

I left the country before anything happened.

In the present work, this occurrence would probably (depending on the discourse context) be classified as INSERTION. The second category is factual, causal, and contains no modifiers (such as temporal measure phrases, e.g., two months before): (4.93)

We had to install a new battery before the car started.

15. Examples (4.92) through (4.96) are Heinämäki’s.

General Discussion 115

This can (in the present approach) be interpreted as DEPENDENCY, indicated by had to. A third category is described as factual without causality: (4.94)

Sachi bought a Toyota before the price went up.

– which may be a case of REGULATION. The fourth category is non-factual and causal with modifiers: (4.95)

Max died two months before he would have retired.

which can be described as either EXPLICIT LENGTH OF TIME or INSERTION, depending on whether the focus is on the length of time or on the fact that something happened before Max could retire (which may or may not be clarified by the wider discourse context). The fifth category is described as non-factual and causal without modifiers: (4.96)

Max died before he saw his grandchildren.

which is another case of INSERTION. This alternative categorisation proposed by Heinämäki exemplifies a phenomenon often encountered in the area of discourse relations, namely, that different categorisations and descriptions are available for similar phenomena, structuring them in various ways (e.g., Knott and Dale 1994). These need not be incompatible, but the relationship between them is often difficult to work out. In the present work, I have relied on, and further expanded, previous approaches as much as possible, which is only complicated by the fact that approaches that consider a wider discourse context of naturally occurring language samples are still rare. The present analysis has led to a deeper understanding of the details of various kinds of relationships between events, several of which are causal in some respect. Furthermore, new conceptual categories such as NO CHANGE were identified, the occurrence of before as a clause-final adverbial was investigated in more detail than in earlier approaches; and along the way, a fair number of occurrences of temporal overlap were identified, confirming previous (mostly introspective) hypotheses in the literature that temporal overlap is indeed possible with temporal dimensional terms (cf. Section 3.4). Naturally, there might be many more options for conceptual relations available to speakers, which are not detected by the present analysis because they generally occur only rarely, not in the analysed text types, or by chance not in the analysed samples. In other words, there is no claim of exhaustiveness with regard to the identified categories, and no guarantee that some options are mentioned in spite of the fact that they represent rare exceptions to speakers’ typical choices. However, the claim that is made

116 Temporal dimensions: Empirical analysis here, which goes beyond the overview presented in the previous chapter and thereby of many earlier findings in the literature, is that all of the examples examined here were produced spontaneously in an unbiased way, instead of by speakers influenced by the analysts’ expectations or the experimental design, or indeed produced by the analysts themselves (via introspection). Furthermore, especially with respect to presuppositional processes the wider discourse context was taken into account to a higher degree than typically encountered in the literature. Thus, while the (quantitative) significance of the results with respect to speakers’ preferences could not be assessed by the present approach, the (qualitative) features of their choices could certainly be addressed. A recent informal investigation of data from the BNC corpus showed no fundamental differences to the insights presented here. Of course, however, further analyses would lead to further detail, possibly to the identification of additional categories (as happened several times throughout the present work), and certainly to a better understanding of distributions with respect to speakers’ preferences of conceptual categories and syntactic forms under consideration of different text types. Figure 3.1 in Chapter 3 summarised previous results with respect to the interpretations of relationships between entities in a discourse that are juxtaposed by a temporal marker. This schematic representation can now be expanded based on the results of the present work, as shown in Figure 4.1 below. This diagram includes the newly identified ways in which the events may be semantically connected. For example, it incorporates the idea (suggested, for instance, in Section 4.1.5) that non-factual or non-committal occurrences of before-clauses represent a subcategory of the INSERTION relation, which is consistent with the data analysed in the present chapter. Further research, however, is necessary in order to clarify whether nonveridical interpretations could also occur with other concepts. Furthermore, the placement of REGULATION is not entirely clear; in the present diagram it has been categorised as not being restricted temporally, since it is conceivable that the order of temporally remote events is negotiated. However, since the cases in the present data all involved temporally close events it could also be the case that REGULATION in fact involves a temporal restriction, in which case it should be represented as a subcategory of ‘event-relation’, where the temporal relationship is restricted by the semantic interconnection between the events. This placement would suggest that the events are in fact always semantically connected even in cases of REGULATION, where the temporal order (not the semantic relationship) is

General Discussion 117

in focus. Further research is needed to clarify this issue, if it can be ultimately resolved at all, considering the fact that there is a high degree of flexibility in both production and interpretation of temporal dimensional terms. Note, however, that as an outcome of this decision, all of the conceptual categories that are here categorised as ‘event relation’ are categories that are causal in some sense, as explained in Section 4.3.3. non-temporal

NONT EMPORAL-T YPE

epistemic deontic repetition ! before as clause-final adverbial ! clause in perfect tense

unrestr_precedence rel

UNREST RICT EDPRECEDENCE-T YPE

REL

reference-to-past-event ! before as clause-final adverbial ! clause in past tense regulation focus on temp. relation of known events agent-level

immediate-succession

IMMEDIAT ESUCCESSION-T YPE

patient-level narrative-level

temp

T EMPREL

causal

CAUSALIT YT YPE

explanation ! after / nachdem ! temporal component in postposed position result ! after / nachdem ! temporal component in preposed position

restr-prec

REST RBY

dependency event-rel

EVENT REL

ST RENGT H

precondition preparation

termination ! before / bevor non-factual

! : inferences from usage of this feature [ : “or” relation

insertion ! before / bevor temporal clause may be non-veridical

VERIDICALIT Y

non-committ presupp

no-change ! before / bevor explicit-length-of-time focus on length of time between events

Figure 4.1. Conceptual categories underlying the application of temporal dimensional terms

Chapter 5 Spatial dimensions: Linguistic repertory and issues of application

This and the following chapter deal with the investigation of spatial language, starting with a thorough overview of findings in the literature, which is (in Chapter 6) followed by original empirical investigations. The first section of the present chapter provides an account of the semantic core and the syntactic variability available in English and German for spatial dimensional terms. The following sections then examine findings concerning their application, including considerations of underlying reference systems (starting from the systematics provided by Levinson 2003), functional factors influencing applicability, and the impact of the discourse situation and task. Insights concerning other kinds of spatial terms are included if considered relevant, but this overview is only aimed to be (near–) complete with respect to dimensional terms. 5.1.

Linguistic repertory: Syntax and semantics of spatial terms

5.1.1. Categories of spatial expressions To begin with, a clarification of terminology and categorisation is in order. In the present work, following Wunderlich and Herweg (1991), I use the category name dimensional for the terms under analysis, namely, expressions like right, left, in front of, behind, above, below, and their German counterparts. This category name highlights the fact that the terms in this category are used to express relative location on the spatial dimensions or axes. In the present work, the name dimensional is used for all syntactic forms that the terms can take in natural language, i.e., not only prepositions. The same category name applies likewise and unproblematically to related expressions in the temporal domain, unlike the near-synonymous term projective which is also used widely in the literature (e.g., Herskovits 1986) for the same class of spatial expressions. Another term that is sometimes used for spatial terms in a similar way, for example, in van der Zee and Slack (2003) is directional; however, this

Linguistic repertory: Syntax and semantics of spatial terms

119

term is highly polysemous, comprising orientation (of an entity, i.e., object or person), location in a specific area with regard to another entity, and direction of movement. Indeed, elsewhere in the literature (e.g., Wunderlich and Herweg 1991, Eschenbach 2005) the term is associated with a predominantly dynamic interpretation (as opposed to a static interpretation for locative terms) that is not in focus in the present work, since motion involves both spatial and temporal dimensions which are kept apart as far as possible here. The analysis focuses on static locative expressions, since the aim is to investigate how objects and events are localized, which is essentially a static conception. A detailed discussion of directionals can be found in Tenbrink et al. (forthc.). The categories labeled ‘locative’ and ‘directional’ comprise diverse kinds of static and dynamic spatial terms (which may sometimes be realised in different ways), i.e., also non-dimensional expressions such as on/onto, in, across, etc. Thus, dimensional expressions are contrasted from other classes of spatial expressions, e.g., topological (expressing neighborhood or contiguity, such as at, on, in), path-related (such as across, through, along), distance-related (such as near, far, close), and other categories (such as between, opposite, etc.; see e.g. Pribbenow 1992 for a more detailed account). Furthermore, they should not be confused with extensional terms: these are sometimes (e.g., Lang 1989) also called ‘dimensional’ because they depend on the one–, two– or three-dimensional extension of objects (cf. Clark 1973), e.g., lang/kurz and long/short, dick/dünn and thick/thin, etc. If all three dimensions are involved, one usually talks about size not extension, which concerns the application of terms like big/groß and small/klein. Such expressions relate to properties rather than spatial relations or locations. As Wunderlich and Herweg (1991: 760) note, locatives are usually relational since they are used to localise objects relative to other objects. This observation also applies to temporal dimensional terms, but not necessarily to directional ones: expressions such as nach rechts require only one participant from which the direction is derived (using information about the participant’s position and orientation). Furthermore, some spatial terms such as between require more than two participants. The present work focuses on relational dimensional expressions.

120

Spatial dimensions: Linguistic repertory and issues of application

5.1.2. Syntactic range of locative spatial-dimensional terms In the area of dimensional terms, Herskovits (1986: 156) lists a range of linguistic options, noting that most are composite expressions: – on the horizontal plane: (at/on/to/by) the (left/right) of (at/on/to/by) the (left/right) {hand} side of (at/on/to/by) the (front/back/side) of in (front/back) of before/behind (right/back/left) of – expressing vertical relations: above/below over/under (at/on/in/by) the (top/bottom) of on top of Herskovits’ list only aims at capturing preposition groups in English. It will be taken as a starting point, with the aim of expanding it to include German terms and other syntactic forms. Note, furthermore, that Herskovits here includes prepositions that are neutral with respect to the half-axis chosen, e.g., at the side of, though others such as beside and next to are missing. The dimensional term denoting the axis (left, back etc.) is often combined with further morphemes (at, on, in, by…) that systematically constrain application (see Section 5.2.6 below). Concerning further syntactic variability, Wunderlich and Herweg (1991) list nouns, adpositions (prepositions and postpositions, the latter of which rarely occur in English or German), adverbs and verbal particles. Locative nouns in English and German correspond to object parts or are used metaphorically, as in (Wunderlich and Herweg 1991: 760): (5.1)

Die Truppen sammelten sich im Rücken des Gegners. lit.: ‘The forces concentrated at the back of the enemy.’

Locative adverbs can serve the same function as prepositional phrases, i.e., they can replace them, as in (Wunderlich and Herweg 1991: 761): (5.2) (5.3)

Anna steht vor dem Haus. ‘Anna is standing in front of the house.’ Anna steht davor. lit.: ‘Anna is standing there–in front of’; ‘Anna is standing in front of it.’

Linguistic repertory: Syntax and semantics of spatial terms

121

Since locative adverbs are semantically relational but do not take a syntactic object, they are used for the implicit localisation of objects relative to contextually given relata. Verbs and verbal particles do not seem to exist for spatial dimensional terms, although there are a few instances of other kinds of spatial meaning carried by such forms, such as enter or surround. In German, case differentiates between directional prepositional phrases (such as (5.4)) and locative ones (such as (5.5)). (5.4) (5.5)

Die Brücke führt über den Fluss. ‘The bridge leads over the river.’ Die Brücke ist über dem Fluss. ‘The bridge is over the river.’

In English, this difference is expressed by the choice of prepositions in some non-dimensional terms (onto vs. on); otherwise the dynamics can only be provided by the verb (leads vs. is). Eschenbach (2005: 73) presents a table listing the syntactic range of German dimensional terms (both directional and locative) which highlights several observations: First, the vertical dimension offers a much wider range of terms in German than the two horizontal ones do. Second, there is a much greater range of variation in directional terms than in locational ones. Third, with respect to dimensional terms the just mentioned case distinction applies for the vertical (über, unter) and frontal (vor, hinter) axes but not for the lateral one. In the lateral axis, prepositions appear with a further case, namely, the possessive (rechts/links des Hauses), which is also possible with the vertical axis (oberhalb/unterhalb des Hauses). Adjectives and adverbs can appear with all axes. These distinctions are represented in Table 5.1 below (omitting Eschenbach’s list of directional terms). Note that Eschenbach’s list does not include adverbial forms with additional deictic morphemes such as davor, nor nouns such as (die) Front as mentioned by Wunderlich and Herweg (cf. above), nor verbal particles (such as vorgehen). Eschenbach’s list may well reflect the intuition that the expressions listed should be the most common ones in German, as opposed to English where the noun front frequently appears in a dimensional preposition group, as analysed in detail by Herskovits (cf. above). Note, for instance, that the frequently used English preposition group to the left can be translated literally into German, yielding the rather infrequently used expression zur Linken which – interestingly – is listed by neither Wunderlich and Herweg nor by Eschenbach. Klabunde (1999: 152) additionally mentions rechterhand and linkerhand (similar to right/left hand side in English)

122

Spatial dimensions: Linguistic repertory and issues of application

which are semantically similar to the adverbs rechts/links, but “their use is confined to a humanlike origo”. Table 5.1. Syntactic variability in German locatives (acc. to Eschenbach) Syntax Prep [+dat] Prep [+poss] Adjective Adverb

UP über oberhalb ober– oben

DOWN unter unterhalb unter– unten

FRONT vor vorder– vorne

BACK hinter hinter– hinten, zurück

RIGHT

LEFT

rechts

links

recht– rechts

link– links

Furthermore, the list does not include superlatives, which are specifically interesting because neither English nor German cover all possible options in any consistent way (i.e., employing the basic dimensional terms as discussed above), as Table 5.2 illustrates. It seems that these terms are often ignored in accounts of spatial terms, especially if there is a focus on “prepositions” as prototypical instances. Table 5.2. Superlative dimensional terms in English and German German English

UP oberst– topmost / uppermost

DOWN unterst– lowest

FRONT vorderst– ? foremost

BACK hinterst– ? rearmost

RIGHT ? rightmost

LEFT ? leftmost

5.1.3. Core semantics for locative spatial-dimensional expressions The aim of the present section is to work out to what extent a core meaning for locative spatial-dimensional expressions can be identified that abstracts from individual discourse situations as far as possible. Herskovits (1986) proposes to assume an “ideal meaning” which is then subject to “bending and stretching” in discourse. This parallels the present approach in that I assume a general distinction between semantic and pragmatic levels. However, unlike Herskovits I do not consider the semantic core meanings as ‘ideal’, nor do I view their variability in actual discourse as ‘distortions’. Rather, I assume that there is some kind of semantic core that pertains to every use of the terms regardless of context. This implies that the semantic core needs to be sufficiently abstract to hold in spite of the many ways in which discoursal factors can influence applicability. In other words, there is

Linguistic repertory: Syntax and semantics of spatial terms

123

no ideal – or logically accurate – meaning that could be (in terms of Bennett 2002) “approximated” in real usage, but the term is sufficiently flexible to be used in a great variety of contexts without loss of its semantic core. Discoursal conditions, pertaining either to the discourse itself or to external factors of the situation, influence the application of spatial terms in various ways, sometimes constraining the applicability, sometimes highlighting specific aspects of the core meanings while neglecting others. Like Herskovits, I assume that such processes are regular and should be accounted for as systematically as possible. Crucially, it is not the core meaning of the dimensional terms that is affected by diverse discourse factors, but rather, application of the term can be more or less appropriate depending on context, or more or less compatible with the spatial situation (Freksa 1999). How this works will be spelled out in some detail in the applicationrelated sections below. In approaching the nature of the ideal meaning of spatial terms, Herskovits (1986: 39) starts by proposing the following: The ideal meaning of a preposition is a geometrical idea, from which all uses of that preposition derive by means of various adaptations and shifts. An ideal meaning is generally a relation between two or three ideal geometric objects (e.g., points, lines, surfaces, volumes, vectors).

Note that this view corresponds largely to that taken in the cognitive linguistics tradition where the nonspatial usages of originally spatial expressions are viewed as metaphorical derivations of basic (abstract, geometric) meanings (called proto-scenes, for instance, in Tyler and Evans 2003). Herskovits (1986: 181) lists the following expressions as having “the most general conditions of application”: to the right of, to the left of, in front of, behind. She describes the ideal meaning of these terms as “graded concepts with the denoted axis as their focal region” (p 184). The other terms used for external relationships (given above) are also based on this ideal meaning but are more constrained in application by pragmatic factors (Herskovits 1986: 188ff.). Herskovits motivates the gradedness of the concepts by pragmatic arguments. In other words, the application of the terms in discourse indicates in several ways that the underlying semantics of the dimensional terms must be based on graded concepts, with the axis itself as ideal interpretation, although pragmatic factors (especially contrast, see Section 5.3.1 below) may affect the influence of the graded structure. This view of including the gradedness of interpretation in the semantics of dimensional terms mirrors a wealth of literature dealing with this phenomenon in the (nearly) two decades since Herskovits’ work. The concept

124

Spatial dimensions: Linguistic repertory and issues of application

of spatial templates, sometimes expressed in terms of prototypicality effects (cf. Rosch 1978, Lakoff 1987), has been developed in detail in psycholinguistic research (see Section 5.2.3). The experiments carried out to determine the structure of the templates are designed to trigger utterances that describe the spatial relation between a reference and a target object for a specific purpose, e.g., to find a hidden object (e.g., Zimmer et al. 1998). Thus, they inevitably interact with discourse-related factors involved in application, and they inevitably involve concrete objects whose properties, as is well known, influence application as well. For example, as Herskovits (1986: 184ff.) shows, the interpretation of the ideal meaning depends on the spatial extension of the reference and target objects, and other discourse-external factors as described in Section 5.2. These facts illustrate the difficulty of separating abstract meaning components from actual usage in discourse – there seems to be no usage-independent evidence for a basic semantics of dimensional terms. Accordingly, psycholinguistic experiments are usually not designed for isolating abstract meaning features of spatial terms but rather aim at identifying cognitive or pragmatic criteria for their usage. Often, however, the term ‘semantics’ is used in a number of different ways for characterising findings on the interpretation of spatial terms in different application situations. In the present approach, it is assumed that the nature of real-world objects and pragmatic factors cannot affect the (abstract) semantics of a linguistic expression that does not directly denote these real world objects, and that is not restricted in usage to a limited range of objects. Thus, the semantic core should be precisely that part of the description of a term that remains largely unaffected by pragmatic factors as well as by the semantics contributed by the discourse context (relating, for example, to the objects involved). It is then the applicability of a term that is affected by such factors, not the term’s semantics, which should be defined in such a way as to allow for flexible usages in various discourse contexts. In other words, the semantics of a term (at any given time, for any given language user or community, as far as this can be determined) should be restricted to aspects that can be specified without knowledge about the situation context. For the present concerns, this means that, as far as possible, factors pertaining to the relation between two objects (as denoted by the dimensional term) should be distinguished from factors pertaining to the objects that are related by the dimensional term. The former concern the semantics of dimensional terms directly, while the latter pertain to the semantics and conceptualisations of the objects concerned. Both factors jointly contribute to the

Linguistic repertory: Syntax and semantics of spatial terms

125

interpretation of any given spatial description, which is further influenced by underlying conceptualisations of configurations as expressed in terms of reference systems, by the level of granularity chosen (perhaps based on the current discourse task), and by other factors of the discourse situation. Clear borderlines are certainly difficult to draw (if feasible at all) since spatial relationships always depend on concrete real-world objects, even though attempts are made to abstract from them, as is the case with Herskovits’ ‘ideal meaning’ that abstracts as far as possible from specific usages. Grabowski (1998: 11f.) points out that, in the case of dimensional terms, the definition of an abstract semantics is especially difficult because of a considerably high variability in application situations and a high degree of context dependency, which stand in contrast to a relatively low degree of specification that is possible for the lexical meaning itself. Thus, the idea of a core semantics may be useful to some degree for explanation, but it clearly has limitations and controversial associations. For present purposes, gradedness effects are viewed as important for the application of dimensional terms (see Section 5.2.3 below), but not classified as part of their core semantics. This view is supported by the observation (Coventry and Garrod 2004: 29ff.) that prototype effects occur with all kinds of linguistic expressions as well as with categories that are created on the spot (i.e., which do not possess a lexical entry at all), such as things to take from one’s home during fire. Coventry and Garrod (2004: 32) conclude that the existence of prototype effects does not mean that the structure of representation in the lexicon is in terms of one (or more) prototypes.

Furthermore, Herskovits (1986: 190f.) ascribes two distinct basic meanings to the dimensional terms depending on the underlying reference system. In the present approach, in contrast, it is assumed that the choice of reference system is basically independent of the dimensional term itself. This assumption is motivated, first, by the fact that most dimensional terms can be used flexibly with different reference systems (Section 5.2.1), and second, by the observation that the application of a dimensional term presupposes a conceptualization in terms of a reference frame that can also be accessed non-linguistically (e.g., Levinson 2003). Thus, dimensional terms and underlying reference systems interact but should not be conflated on the semantic level. This view is reflected in the lexical entries for dimensional prepositions given by Eschenbach (2005: 75), defining the semantic core of each term relative to a reference system that is viewed as a contextual component. For instance, the semantic form of German vor (in front of) is described as follows (G refers to the ground object, i.e., the referent):

126

Spatial dimensions: Linguistic repertory and issues of application

[T]he figure is spatially included in a region that is derived from G and the principal direction FRONT as given by a reference system RSYS.1

The semantic forms of hinter, über, unter, rechts, and links can then be described similarly by replacing the principal directions (half-axes) accordingly. Generally, the proposals presented so far mainly treat the basic dimensional terms representing the six spatial directions (left, right, front, back, above, below) similarly, focusing on the geometric aspects of the abstract semantics of the dimensional terms. In a different approach developed in detail, for example, by Aurnague and Vieu (1993) with regard to French terms, the geometric level represents only part of the semantic description. Another part is the contribution of a functional element, such as SUPPORT in the case of the topological term on, or CONTAINMENT in the case of in (Garrod et al. 1999). According to Aurnague and Vieu, the functional element also belongs to the abstract semantics because it is valid across discourse contexts. Coventry and Garrod (2004) propose a systematic account of functional effects. For example, the lexical entry of in contains a functional component of location control and a geometric component of enclosure, which come into play to different degrees in different discourse contexts (2004: 53): When objects exhibit only weak enclosure (…), then firm evidence of location control is required to support an in relation. (…) However, when the evidence for location control is weak, there needs to be a strong geometric enclosure to support the in relation. The same is true for the on relation.

In the present work, a distinction is made between associated functional features of the spatial expression itself (or the spatial relation that is represented by it), and functional aspects that pertain to the objects involved. Thus, an expression like in is associated with the functional element of containment because it is restricted to spatial relationships of containment. Therefore, the expression can only be used with objects that allow for such a relationship. This inherent feature of in explains the difference in acceptability between sentences using in together with different objects: (5.6) (5.7) (5.8) 1.

The banana is ?in / on the table. The hole is in the table. The banana is in the bowl. For a formal version of this and other lexical entries see Eschenbach (2005).

Linguistic repertory: Syntax and semantics of spatial terms

127

Since objects can be associated with different kinds of functional features, naturally different objects can match with the spatial term to different degrees, also depending on the situational context. These aspects constitute the major part of research concerning functional geometry (a term proposed by Garrod et al. 1999) since it is usually the function of the objects that contributes the evidence for location control and the like. This field of research is addressed separately in Section 5.2.7 below. It will be seen there that the functions of objects can influence the applicability of spatial terms in a number of ways, going beyond the (usually fairly clearcut) functional element of the spatial term itself. Tyler and Evans (2003) propose a similar view as that taken in Aurnague and Vieu (1993), but deal exclusively with English expressions. Their analysis draws on, reviews, and integrates previous accounts, including the ground-breaking work on over by Brugman (1988). In their terms, the semantics of each spatial term is described by a spatial “proto-scene” together with a functional element that serve as basis for all additional senses derived from this basic meaning. Such a proto-scene abstracts away from the details of specific spatial scenes but consists of a “schematic trajector” (TR) and a “schematic background element”, a landmark (LM). Thus, their description of the vertical dimensional term over is that “the TR is higher than but within potential contact of the LM” (p. 65). The functional aspect involved here is that either one of the participants “is conceptualized as being within the sphere of influence” (p. 66) of the other. This description is suited to differentiating between over and the very similar term above that nevertheless has different areas of application. Tyler and Evans (2003: 66) claim that this is because “the proto-scene for above designates a spatial configuration in which the TR is higher than but not in potential reach of the LM”. Table 5.3 below shows an overview2 of the basic senses ascribed to dimensional terms by Tyler and Evans (2003). They do not consider the lateral axis (left, right) at all, presumably because non-spatial senses (the main focus of their work) can scarcely be derived from them. The authors point out that, with regard to the basic senses, pairs of spatial terms often constitute a contrast pair, as is the case with above/below, over/under, up/down, etc. Thus, the spatial description involves a contrast with regard to the direction on the same axis, while the functional element

2.

Unfortunately, Tyler and Evans do not provide any comprehensive overview of this kind; rather, the relevant information is distributed throughout the book. I hope that this overview does not misinterpret their proposal.

128

Spatial dimensions: Linguistic repertory and issues of application

in each case is either equivalent (as with above/below) or contrastive (as with the directional terms up/down). Table 5.3. Basic senses of dimensional terms (acc. to Tyler and Evans 2003) Expression over above under below up

down

in front of

behind

Spatial Proto-Scene the TR is higher than but proximal to the LM the TR is higher than but distal to the LM the TR is lower than but proximal to the LM the TR is lower than but distal to the LM an oriented TR is directed towards the top of an oriented LM (vertical axis) an oriented TR is directed towards the bottom of an oriented LM (vertical axis) an oriented LM is directed towards an unoriented TR (frontal axis) an oriented LM is directed away from an unoriented TR (frontal axis)

Functional Element potential reach (“sphere of influence”) between TR and LM no potential reach between TR and LM potential reach (“sphere of influence”) between TR and LM no potential reach between TR and LM positive value: physically elevated means more visible, more accessible, in a normative position, etc. negative value

(perceptual) accessibility

lack of perceptual access

For many of the terms, it is proposed that the semantics involve the orientation of either the landmark or the trajector. However, especially in the case of in front of/behind this interpretation runs into problems. The main problem lies in the fact that the authors generally rely on examples involving intrinsic reference systems, where either the trajector’s or the landmark’s orientation is used for reference. Only in a footnote on page 162 do they mention the possibility of the vantage point of a third entity. The present approach, in contrast, starts from Levinson’s classification of reference systems (see Section 5.2.2.1 below) which neutrally allows for different vantage points along with a variety of possible reference systems, neither of which are inherent parts of the semantics of dimensional terms. This classification will show that it is only in the case of intrinsic reference systems

Linguistic repertory: Syntax and semantics of spatial terms

129

that the orientation of the entities involved matters. In relative reference systems, the observer’s vantage point takes over this function. The association of accessibility with the frontal axis is mirrored in various other accounts; Eschenbach (2005: 78), for instance, assumes that the front region of a person (in an intrinsic reference system) is the region of his or her best access by vision, manipulation, and motion. The region behind the person is the opposite region, which can be accessed less well. (…) Relative uses of vor and hinter, in contrast, deal with distance and occlusion.

Thus, in her account the semantics of the expressions of the frontal axes is associated with the effects of the reference systems on accessibility. Grabowski (1998) proposes that all horizontal expressions are related to perceptual accessibility from the origin’s position. In his view the spatial relation between the two objects is less relevant than perceptual factors. On the front half-axis, the object is located directly in the visual field of the observer; on the lateral axis, the observer only needs to turn their head to perceive the located object. On the back half-axis the located object is occluded either because it is behind the observer (in an intrinsic reference system) or because it is behind the relatum (in a relative reference system). It can be added (to Grabowski’s observations) that a reversed relation of occlusion also comes into play on the front half-axis in relative (but not intrinsic) reference systems, where a referent in front of another object occludes this relatum from the observer’s point of view. Specifically with regard to hinter, where a lack of perceptual accessibility is proposed as part of the semantic core, this description is compatible with Tyler and Evans’ account. On the grounds of empirical evidence, however, Landau and Jackendoff (1993: 231) ascribe only a secondary role to this factor: Visibility and occlusion may well play a role in children’s learning of spatial prepositions. Johnston (1984) found that 3-year-olds understood behind to mean ‘occluded from sight’. For example, they agreed that X was behind Y only when Y was an object large enough to occlude X. In contrast, adults agree that X is behind Y whatever their relative sizes, as long as the figure object is aligned properly with the reference object’s front-back axis.

While these results do suggest that children overestimate the significance of the occlusion effect, they do not disprove the hypothesis that the semantics of behind at least for some speakers includes an element of at least partial occlusion. But note that the effect interacts with the underlying reference

130

Spatial dimensions: Linguistic repertory and issues of application

system, at least in German: Occlusion effects do not come into play, or at least play a lesser role, for instance, if the observer’s visual field is partitioned into regions (see Section 5.2.2.2). In that case, hinten describes a region that is further away from the observer but usually still visually accessible, though maybe with increased effort. Therefore, presumably the perceived importance of perceptual accessibility should be viewed as conceptually derived from a typical situation in which the spatial arrangement in fact supports or impedes perceptual access, rather than as an inherent part of the terms’ semantics. In the following, I will address relevant factors influencing the application of locative dimensional terms separately, keeping in mind that it is often a matter of theoretical decision to include some of these aspects in the lexical semantics of the terms, unless it is clearly the objects, not the relation between them, that influence the terms’ application. 5.2.

Discourse-external factors involved in the application

5.2.1. Axial structure in language and cognition Unlike time which involves only one dimension and therefore only one basic axis, spatial concepts are applicable in three different dimensions, complicating representation. Typical semantic descriptions of dimensional terms, such as the proposal by Eschenbach (2005) cited in the previous section, involve the notion of axis (or direction in a different terminology). More concretely, interpretation of the terms builds on the conceptualisation of regions surrounding three basic axes, corresponding to each of the three dimensions. Various ways of representing spatial information in language and other modes (by way of axes, half-axes, half-lines, vectors, regions, etc.) capture basically this same idea (e.g., van der Zee and Slack 2003). In principle, it is also possible to conceptualise the spatial directions as half-planes rather than axes, dividing three-dimensional space into halfspaces, rather than two-dimensional space into half-regions. In language use and in most literature on the topic, however, this conceptual option does not seem to play a major role. Even if the concept may be underlying some linguistic representations, this is usually not reflected in the language, as is exemplified with respect to various interpretations of one utterance in Section 5.2.5 below. In other words, most linguistic representations that employ the dimensional terms can be interpreted as though referring to a

Discourse-external factors involved in the application

131

two-dimensional rather than a three-dimensional world, in which case it is sufficient to talk about axes separating half-planes, rather than planes separating half-spaces. This intuition is supported by the observation that usually a maximum of two, not three, dimensional terms are combined in one single utterance. This implies that, usually, the third dimension does not play a role in the spatial relationship represented by the utterance. According to Herskovits (1986: 76) spatial relationships are often described ignoring the vertical plane “because we are normally bound to the ground”. Most empirical work on dimensional terms, furthermore, is restricted to two dimensions by the spatial setting: usually, the vertical dimension is missing, except if pictures are involved, in which case the frontal axis is neglected. The studies presented in Chapter 6 of the present work are no exception in this respect. With regard to the cognitive status of the three axes, some differences can be noted. For instance, the vertical axis is viewed as special because it is defined on the basis of gravity, and thus independently of the orientation of the other axes (e.g., Nikanne 2003). Psychological and psycholinguistic research (e.g., Bryant et al. 2000, Landau 2003) also shows that the vertical axis is cognitively privileged, as proved by results in memory and attention tasks, language acquisition, etc. Furthermore, the frontal axis has precedence over the lateral one because it is directed and derivable from objects’ fronts, while the lateral axis is symmetric and needs to be derived from the orientation of the frontal axis. Retz-Schmidt summarises the results and arguments of earlier work as follows (1988: 97): [T]he vertical dimension has the status of a privileged direction because it is fixed by the gravitation of the earth, whereas in both horizontal dimensions, man [sic] can move freely. Thus, the vertical dimension can be conceived of as the primary dimension. Moreover, the human body is asymmetric in the front-back dimension, enabling man to distinguish more easily between front and back than between left and right. This asymmetry permits the less salient distinction between the front-back dimension as the secondary dimension and the left-right dimension as the tertiary dimension.

Ehrich (1985: 132) notes that this symmetry of the lateral axis is reflected in language by the existence of the direction-neutral lateral term neben (beside), which has no direct correspondence on the other two axes. This observation provides a good reason why Herskovits (1986) and several other authors intuitively (and often without further explanation) include neben, beside and sometimes further terms in the description of dimen-

132

Spatial dimensions: Linguistic repertory and issues of application

sional (projective) terms, which is not self-evident since the terms depart in several respects (e.g., morphologically) from the other expressions. Franklin et al. (1995) specifically examine the relationship between the frontal and the lateral axes in surrounding space (i.e., the horizontal plane in relation to the human body). Although Retz-Schmidt’s summary above shows how the salience of the human body has long been recognised in the literature, previous psychological work addressing spatial regions often dealt with abstract pictures rather than surrounding space. The results of such work then showed no difference between the two horizontal axes. In contrast, the work of Franklin et al. (1995) proves the cognitive precedence of the frontal axis in surrounding space in several respects. For example, frontal regions were treated as larger than others (see Section 5.2.3 below), resolution (i.e., discriminability of directions) inside the regions was higher, and linguistic reference centered around the frontal axis. An alternative to referring to three axes (lateral, frontal, vertical) is to refer to six half-axes or directions (left, right, front, back, up, down). Herskovits (1986: 199) opts for using “half-line axes” for the following reason: A half-line axis is a straight line starting at some point and extending to infinity in one direction only. A description of the frame of reference in terms of six half-line axes is more convenient than one in terms of three oriented axes extending to infinity in both directions, since the composite prepositions make direct reference to six directions, although opposite axes are not always symmetric.

While pointing to the importance of axial structure in both language and cognition, Landau (2003) finds that a specification of half-axes, i.e., direction information, has more effects in linguistic tasks than in memory. Thus, people distinguish positive and negative ends of vertical and horizontal axes in language but not necessarily in non-linguistic tasks. She concludes (Landau 2003: 36): The facts are consistent in suggesting that direction is represented separately from the axes themselves, and the evidence from spatial impairment suggests that direction may be a more fragile component of spatial representation than axial structure.

Since the relationship between axes and half-axes is hyponymical, it is convenient (for present purposes) to refer to the basic axes whenever the directionality is irrelevant, and to the half-axes (or directions) and their

Discourse-external factors involved in the application

133

linguistic labeling whenever a difference becomes apparent, such as in the actual usage of a single term. The acceptable space (the region of applicability) of dimensional terms is based on the relationship of the referent to the relatum with respect to one of the basic axes. The gradedness of acceptability then concerns the degree of deviation from these axes (Section 5.2.3). The frame of reference (Section 5.2.2) determines whether the relatum’s inherent axes are used in the application of the dimensional term (in the case of an intrinsic reference frame) or an axial structure is imposed externally (in the case of a relative reference frame). In both cases, the notion of axis is usually understood (implicitly) as an abstract linear structure without specifications of extensions in other dimensions. Some observations concerning the treatment of extended objects are discussed specifically in Section 5.2.4 below. The possibility of using the relatum’s inherent axes for spatial reference raises the question of how an object’s internal axes are derived. In natural interaction, these may as such be less relevant than their (top, bottom, front, back, left, right) parts or regions, which systematically interrelate with the underlying linear axes. As Landau and Jackendoff (1993: 220) point out, object parts are not subsidiary in the way handles or wings are; instead, “they denote regions of the object based on its inherent orientation”. In order to determine an object’s internal regions, various aspects come into play, most of which are not physically determined but rather based on human cognition and behavioural conventions (Herrmann 1990: 120). Levelt (1996: 87) points out that the attribution of intrinsic sides is not always uncontroversial, and can lead to problems in communication: The felicity of speaker/hearer coordination in the intrinsic system is (…) crucially dependent on the shared image of the relatum. First, coordination in the intrinsic system is only possible if the relatum is oriented. (…) Second, frontness is an interpretative category, not a strictly visual one.

In animate beings the part containing perceptual organs defines the intrinsic front. Herskovits specifies the case of a human being’s axial structure in a prototypical situation (1986: 157): A human being in canonical position defines six half-line axes with origin at him/herself. The down axis follows the direction of the gravitational field and the up axis is opposed to it. The four other axes are in an horizontal plane: the direction in which the observer’s eyes are facing defines the front axis, and the back axis is opposed to the front one; right and left axes are orthogonal to the front-back line.

134

Spatial dimensions: Linguistic repertory and issues of application

Note that this definition is rather vague: it is not clear whether the origin should be understood as an abstract point (in which case the exact position of this point would need to be clarified), or a structure resembling the human’s body; the view direction is also not very clearly defined. However, for analysing natural language such a vague definition is sufficient because language is generally not used in a mathematically precise way. The situation context usually provides further information that enables successful communication in spite of such underdeterminacy. In other cases, typical direction of motion and orientation to the observer play a role (Miller and Johnson-Laird 1976: 400f.). Thus, arrows have an intrinsic front because of their associated direction of motion, and a desk’s front is defined by its usual orientation to a person. Intrinsic tops are dependent on an entity’s characteristic orientation on the vertical axis. Often, functional features come into play that are derived from our characteristic uses of the objects. These are also decisive for the determination of reference regions (see Section 5.2.7 below). With natural entities, it has often been pointed out that axes are not equal (e.g., Marr 1982). Landau and Jackendoff (1993: 221) distinguish between one generating axis (the object’s principal axis, e.g., the vertical axis in humans) and two orienting axes orthogonal to each other and to the generating axis (the lateral and the frontal ones). These axes can be marked as directed or symmetric; the vertical and frontal axes in humans are directed, while the lateral one is symmetric. In contrast to this, consider an arrow which “has a directed generating axis but no significant orienting axes” (Landau and Jackendoff 1993: 221). These distinctions are responsible for the ascription of object parts, since only directed axes can have one preferred direction of typical encounter or motion, which define object fronts. Due to the inherent orientation of the frontal axis, it is possible to conceptualise a linear order with respect to entities positioned on that axis. Thus, people situated closer to the front end of a queue, for example, are viewed as prior in that order. However, ordering situations apply primarily when time is involved. In the queue example, people waiting closer to the front end in a queue will be served earlier than people behind them. Thus, sequentiality here corresponds to an order in time as well as in space. Further situations in which spatial scenes can be conceptualised as linear are motion events, and linguistic representations of complex spatial situations, e.g., in localisation sequences and route descriptions (cf. Section 5.3.1).

Discourse-external factors involved in the application

135

Once the intrinsic front has been determined, the back side is straightforwardly inferred (by its position at the opposite side), but not necessarily the lateral directions. Here, choice of perspective is decisive (Retz-Schmidt 1988). Objects can be viewed from the outside, such as a picture, where the left side is defined by the observer’s position. But some objects can also be viewed from the inside, as is usually the case when talking about the internal parts of a person, but it can also be true for objects that can be occupied by a person, such as a car. This difference is elsewhere often referred to as handedness in the literature (e.g., Pribbenow 1992: 25, Levinson 2003), a term which relates to features of the objects themselves rather than to the perspective taken by an observer. In Retz-Schmidt’s (1988) account, the two object categories are distinguished by the way they are typically viewed, which seems to be uncontroversial. This way of ascribing sides may interact with other factors such as the effects of intrinsic directionality (as with vehicles). Pederson (2003: 291) notes that the interpretation of clockwise also depends on the perspective of an unspecified observer looking at a clock from the outside, while an inside perspective (“the perspective of someone lurking in the back of a grandfather clock”) is easily conceivable, yielding a contradictory interpretation. Determining an object’s axes is thus often a matter of convention and conceptualisation with regard to the object’s features and functionalities. But even in the case of abstract, non-functional objects speakers systematically view specific properties of objects as decisive for determining axes. For example, van der Zee and Eshuis (2003), who explore the influence of a relatum’s three-dimensional properties on intrinsic usages of horizontal prepositions, find that long axes are preferredly taken as frontal axes, while the shortest axes were primarily used as lateral axes in (external) intrinsic reference relations. Relative curvature of the main plane of symmetry and contour expansion along a relatum’s main axis also play a role in assigning reference axes. 5.2.2. Reference systems, topological distinctions, and perspective As has often been noted, the same spatial situation can often be expressed in a variety of different ways. An impressive demonstration of this can be found in Talmy (2000: 226), where the expression in front of can be interpreted in no less than four conflicting ways, all of which are consistent with

136

Spatial dimensions: Linguistic repertory and issues of application

a certain spatial scene – and which, by far, do not exhaust the possible ways in which the spatial relationships present in that scene can be described. The reason for this is that these interpretations rely on the conceptualisation of different underlying reference systems. The aim of this section is to show in which ways reference systems can be employed, and point to the impact they have on the interpretation of dimensional terms. Reference systems have been dealt with extensively in the literature, sometimes proposing that they belong to the semantic definition of dimensional expressions, sometimes dealing with them as a phenomenon entirely independent of language. Various different classifications and models have been proposed (e.g., Retz-Schmidt 1988, Herrmann 1990, Levelt 1996), and the literature offers a variety of overviews and discussions of advantages and shortcomings of specific views (e.g., Levinson 2003, Pederson 2003). Levinson elaborately describes the confusions arising with different and conflicting terminology, including various interpretations of terms like deictic, extrinsic, and intrinsic. Importantly, deixis is often confused with perspective, since both are based on the actual situation. For this reason, in Levinson’s terminology (Levinson 2003) the term ‘deictic’ is avoided in the area of reference systems. Levinson’s account, which is equivalent to Herrmann’s (1990) “6-H-Modell”, is the starting point adopted in the present work, which is extended by further insights stemming from other sources in order to propose a systematic comprehensive overview of possible reference systems in combination with the diverse perspectives available. To my knowledge, such an overview is not available in the literature so far, though many authors have pointed to many different aspects that need to be accounted for. All of the variants presented here are relevant for the interpretation of dimensional terms; other spatial expressions need other kinds of frameworks that are not discussed here. A crucial notion involved in the interpretation of reference systems is that of perspective, representing the spatial point of view of an observer. The term is sometimes used in a broader sense, reflecting the fact that all utterances in some way reflect the speaker’s position with regard to many different (also non-spatial) dimensions (see e.g., Schober 1998a), few of which have direct consequences on the interpretation of dimensional terms. Another spatial sense emerges in work on different kinds of knowledge about spatial relations, where the well-known difference between route and survey representations is often expressed in terms of perspective (e.g., Taylor and Naylor 2002). These distinctions are especially relevant with respect to speakers’ strategies in describing a spatial situation (see Section

Discourse-external factors involved in the application

137

5.3.2.1 below). Tversky (1999) offers further reflections on the notion of (spatial) perspective. In the present work, the term perspective is used solely for the origin of a reference system as used in spatial descriptions. 5.2.2.1. External reference systems Objects can be related either externally or internally; in the latter case one object is located inside another object. Levinson’s account focuses on external relationships. It can be characterised schematically as recognising three different reference systems with three variations each, dependent on whether the speaker, the hearer, or a third entity serves as the origin of the perspective employed. The three different options for reference systems are labeled by Levinson as intrinsic, relative, and absolute. They are examined in this subsection. The three options for origins reflect the distinction in grammar between first person (speaker), second person (addressee), or third person (cf. Herrmann 1990: 131). 5.2.2.1.1. Intrinsic reference systems In intrinsic reference systems, the relative position of one object (the located object or referent) to another (the reference object or relatum) is described by referring to the relatum’s intrinsic properties such as front or back. Thus, in a scenario where a stone (the referent) is situated in front of a house (the relatum), the stone can be unambiguously identified by referring to the house’s front as the origin of the reference system: (5.9)

There is a stone in front of the house.

In such a situation, the speaker’s and hearer’s positions are irrelevant for the identification of the object. However, the speaker’s or hearer’s front or back, or, for that matter, left or right, may also serve as origins in intrinsic reference systems, as in: (5.10)

There is a stone in front of me / you.

In this kind of intrinsic reference, the speaker’s or hearer’s actual position and orientation is decisive, which may be one reason why such sentences have sometimes been classified as deictic, causing the above-mentioned confusion in the intrinsic/deictic distinction.

138

Spatial dimensions: Linguistic repertory and issues of application

In intrinsic reference systems, no further entity is needed, which is why Herrmann (1990) refers to this option as two-point localisation. Put another way, the point of view or perspective, or in Levinson’s terms: the origin, is conflated with the relatum, which is only possible if the relatum is at all capable of providing an intrinsic perspective, by virtue of its possessing internal parts. Eschenbach (1999: 334) gives the following definition for intrinsic reference frames: An intrinsic relation R(F,G) relates the positions of F and G on the basis of a frame of reference that is established by the functional-spatial structure of G.

Here the “functional-spatial structure of G” defines the internal parts of G. Levelt (1996) has claimed that the use of intrinsic reference systems is dependent on the orientation of the relatum: the relatum must be oriented in its canonical position with respect to the dimension used for reference. This claim has recently been empirically addressed by Walker and van der Zee (2005), who found that in some cases there is indeed a tendency for an influence of the suggested principle, but it is by no means a strict rule. Intrinsic reference systems can also be induced by features of the figure rather than the ground, as in: (5.11)

Sarah stood in front of the bush. (Tyler and Evans 2003: 161)

However, apparently only in front of is capable of expressing this kind of intrinsic relationship without implying an external observer, as the following example shows: (5.12)

Sarah stood to the right of the bush.

where an external perspective is needed for interpretation. Possibly, the relation expressed by (5.11) should be viewed as primarily functional rather than spatial, highlighting proximity and accessibility rather than a specific position on a spatial axis. But without an appropriate discourse context this remains speculative. 5.2.2.1.2 Relative reference systems If relative reference systems are used, or, in Herrmann’s terminology, three-point localisation, the position of a third entity serves as origin instead of inbuilt features of the relatum. Thus, the stone (the referent) may

Discourse-external factors involved in the application

139

be situated to the left of the house (the relatum) from the speaker’s, the hearer’s, or a further entity’s point of view (origin): (5.13) (5.14)

There is a stone to the left of the house from your / my point of view. There is a stone to the left of the house, viewed from the hut.

Here, the house’s front and back are irrelevant, which is why this reference system can be employed whenever the position of an object needs to be specified relative to an entity (a relatum) with no intrinsic directions, such as a box. Eschenbach’s (1999: 334) definition is as follows: A relative relation R(F,G,V) relates the positions of F and G on the basis of a frame of reference that is established by the position and functionalspatial structure of an additional entity V. V may be implicitly given, it is different from both F and G, and it may also be a spatio-temporal process, such as a process of motion.3

In this definition, the fact that the view direction is derived from a certain position is expressed by assuming a further entity (such as speaker / hearer) with a “functional-spatial structure”, such as a front, which permits assigning a viewpoint. This entity need not be expressed linguistically: in (5.13) the phrase “from your / my point of view” would typically be omitted in natural discourse (Herrmann and Grabowski 1994). Pederson (2003: 290) proposes that, in English, it is possible to use a shortened version: (5.15)

The cat is to your left of the trash can.

which induces him to introduce a new term for this phenomenon, namely, altercentric or hearer transposed relative (as opposed to altercentric intrinsic as used, for instance, in example (5.10) above). Likewise, a third party transposed relative version is available, as exemplified by simply exchanging his for your in example (5.15). It could be added that it should be equally possible to say to my left of the trash can, using the speaker’s perspective. According to the intuition of some native speakers of English, however, these options do not really seem natural and are usually dispreferred. This intuition is supported by the empirical work presented in Chapter 6 (see Section 6.4.3.1). Notice that this abbreviated way of marking perspective is not available in German at all, which could be interpreted in the way that making the origin explicit in a relative reference system is not grammaticised in Ger-

3.

Eschenbach’s observations on motion are taken up later in this section.

140

Spatial dimensions: Linguistic repertory and issues of application

man. Furthermore, the options in English seem to be restricted (if they are used regularly at all) to the lateral axis, as exemplified by the following: (5.16) (5.17)

? The cat is to your front / back of the trash can. The cat is (*to your) behind / above / below the trash can.

In example (5.14), contrary to Eschenbach’s definition, the point of view from a location is taken that itself does not provide a clear view direction, at least not if the hut’s intrinsic properties are not known. Nevertheless, the spatial relationship can be interpreted uncontroversially (however vaguely), assuming that somebody (an entity possessing perceptual capabilities, which the hut itself does not possess) looks from somewhere close to (or inside of) the hut into the direction of the other two entities mentioned. Thus, although the description in (5.14) is underdetermined with respect to the exact position and view direction of the observer, and also with respect to the observer’s identity and perceptual features, the spatial relationship may be sufficiently described for the purpose at hand – provided that it is given in a suitable situation. This is only the case if the hut (the origin) is located at some distance from the other objects, and not, for instance, between the two, yielding a situation in which there are two possible view directions toward the objects. Furthermore, the situation becomes more difficult if more objects are involved, so that it is not clear where the view direction should focus. If there are several objects of the same class present, it is furthermore possible to use the whole group of similar objects as a relatum. This option is called group-based reference by Moratz (e.g., Moratz and Fischer 2000). In contrast to situations where the relatum is an object of a different kind (here called landmark-based relative reference system), in a group-based relative reference system the relatum consists of one or several objects of the same kind as the target object. An example is the following: (5.18)

The ball is located to the left of the other balls (from my / your / a third point of view).

The term ‘group-based’ refers to the linguistic rather than a cognitive level. It does not aim at determining what kinds of factors (e.g., of distance or similarity) may lead people to conceive of an assembly of objects as a group, but rather, it simply points to the fact that people sometimes treat objects linguistically as a group, possibly based on perception, but maybe based on other factors as well. Such linguistic treatment is achieved by introducing all objects except for the target object as one (possibly plural) participant in a linguistic construction, as in the other balls. This participant

Discourse-external factors involved in the application

141

can also be omitted on the linguistic surface while still being intended as relatum of the spatial description. In that case, the underlying reference system may be underdetermined, as the relatum could be the group of other objects as well as any other kind of relatum present in the scenario. However, this is not necessarily true for dimensional adjectives. Eschenbach (2005: 82f.) claims that German dimensional adjectives like link– (left) can only be employed when there is at least one other object of the same class present: They can be used only as restrictive modifiers to select an object (the figure) in contrast to another object (called here ‘ground’) of the same category. Thus, das obere Buch (‘the upper book’) refers to a book that is in a higher position than another book present. Nevertheless, in the context of a projective adjective the ground object cannot be specified by linguistic means but has to be provided by the situation of utterance.

Eschenbach thus formulates the intuition that projective (i.e., dimensional) adjectives require the presence of another object of the same class, as in (5.19) where the spatial relationship of one ball is determined (implicitly) by the position of at least one other ball present in the situation: (5.19)

Kannst du den linken Ball holen? ‘Can you fetch the left ball?’’

This utterance is equivalent to (5.18) with respect to the localisation of the ball in relation to (one or more) other balls present. Thus, it seems that in Eschenbach’s definition the other object of the same class is used as a relatum, mirroring the present definition of a group-based reference system. The empirical analyses in the present work (specifically Section 6.3.5.3) highlight whether the employment of German (and maybe English, as well) dimensional adjectives is indeed, as could thus be assumed, restricted to cases of group-based reference, or whether other factors come into play as well. Also, English and German may differ in this respect, especially since English (but not German) has the additional option of a dimensional superlative such as leftmost in this situation. Superlatives naturally carry the restriction proposed by Eschenbach (the presence of other objects) in their semantics, since they can only be used in comparison to other candidates. With regard to the frontal axes, Hill (1982) points out that two conceptual alternatives are possible, one of which is used regularly in languages such as English and German, and the other (at least) in Hausa, although there are indications that both options are available in all cultures. The two options are called facing versus aligned. In facing interpretations (as in

142

Spatial dimensions: Linguistic repertory and issues of application

English and German usage), an object that is located in front of another object is closer to the origin than the relatum is. In aligned interpretations (as in Hausa), the opposite is the case.

X Figure 5.1. Two conceptual options for frontal axes in relative reference systems (plan view)

In Figure 5.1, a facing interpretation would be that the circle is in front of the square from the point of view of X, while an aligned interpretation yields that the square is in front of the circle, as Hausa speakers would assume. Since the present work only deals with English and German usage, the aligned interpretation is simply viewed as an interesting conceptual alternative that is not regularly4 put to use in these languages.5 5.2.2.1.3 Absolute reference systems In absolute reference systems, neither a third entity nor intrinsic features are used for reference. Instead, fixed-bearing systems such as north and south or, in some languages, uphill or downhill (Levinson 2003) serve as anchor directions. Thus, the stone may be located to the north of the speaker, the hearer, or the house. Eschenbach’s definition for absolute reference frames is as follows (1999: 334): 4.

5.

According to Vorwerg (personal communication, 2006), the natural language data collected in various experiments consistently show that a low proportion of speakers do use the aligned interpretation in their descriptions in English and German. This finding corresponds to some native speakers’ intuitions. If the entities involved are part of a queue, then interpretation would be based on the functional front end of the queue rather than X’s point of view, which would then lead to similar results as in the Hausa example (i.e., the square is in front of the circle with respect to the queue). Similar effects arise if the entities are in motion (see below), where the direction of motion induces a perspective on the situation.

Discourse-external factors involved in the application

143

An absolute relation R(F,G) relates the positions of F and G on the basis of a frame of reference that is established by the functional-spatial structure of the common surrounding of F and G.

On the vertical axis, gravity determines the direction of up and down and other terms using absolute (rather than relative or intrinsic) reference. Pederson (2003: 290) notes that there are at least two distinct subtypes of absolute reference systems: 1) an absolute RF which relies on co-ordinates which do not derive from the immediate environment and 2) an absolute RF which creates ad hoc coordinates by appeal to perceptual features.

An example for the first option is the usage of the earth’s cardinal directions (north, south, east, west), and for the second, the usage of local landmarks for projective descriptions (Pederson 2003: 290): (5.20)

The cat is towards the wall from the trashcan.

Absolute reference systems are a special case in that they are based on directions rather than entities, which renders the identification of an origin superfluous.6 Pederson attempts to capture the relationship between his second kind of absolute reference systems and intrinsic reference systems as follows (2003: 290): In both cases there is a direction determined from the ground on the basis of perceptually available features. The difference being that in one case the features belong to the ground and in the other case they do not. It is unclear whether to treat towards the wall as a case of transposing co-ordinates onto the trashcan. Subjectively, the example seems to represent a route from trashcan to wall, along which a cat will be found. As such, the wall is essentially the same indicator of direction as a more abstract term like North.

It seems that spontaneously created absolute reference systems (Pederson’s second kind) are not very well researched so far. Absolute reference systems of the first kind are, in Western societies, only used in specific cases, preferably in large-scale space, as opposed to indoor scenarios (which are mostly in focus in the empirical work presented in Chapter 6); and preferably in relation to maps in which the absolute directions are explicitly

6.

In the case of “the object to the north from my point of view”, the speaker conceptualises the direction as relating to an origin (source of perspective) instead of a relatum. As far as the spatial relation is concerned, “from my position” would be equivalent to this conception.

144

Spatial dimensions: Linguistic repertory and issues of application

marked (Levinson 2003). However, in the latter case speakers might simply refer to the side marked “North” as North without having to conceptualize the actual absolute compass directions; therefore, map use may be closer to Pederson’s second (ad-hoc) kind of absolute directions. In some cases this may even happen for depictions that are not aligned with the compass directions, such that the top side is referred to as North even if this is not actually the case. In this way, an ad-hoc “absolute” reference system is created on the grounds of conventional map use. Also, Carroll (1997) reports that native speakers of English, in an experimental study involving models of villages and rooms, used north, south, east, west in a nonabsolute way (1997: 140): The use of the ‘earth’ axes in English is apparently warranted by the fact that rooms and villages are places we can walk around in. The axes are not absolute but are established relative to the position of the observer, so that the north side is typically the side furthest away from the viewing point.

This usage corresponds to the usage of dimensional terms in internal regions in the visual field of an observer (see Section 5.2.2.2 below). Most absolute reference systems employ terms other than the spatialdimensional expressions dealt with in this book. These will not be treated further. The only exceptions are terms used on the vertical axis which are based on gravity, such as many usages of above and below. These will be dealt with in more detail below. Table 5.4. Linguistic examples for reference systems and perspectives origin

relative

intrinsic

speaker (1st person)

There is a ball in front of the tree (from my point of view).

There is a ball in front of me.

listener (2nd person)

There is a ball in front of the tree (from your point of view).

There is a ball in front of you.

third entity (3rd person)

Viewed from the stone, there is a ball in front of the tree.

There is a ball in front of the arrow.

Discourse-external factors involved in the application

145

5.2.2.1.4 Choice and coincidence of intrinsic, relative, and absolute reference systems Table 5.4 exemplifies usages on the horizontal plane in the two reference systems in focus, illustrating the impact of variability of the origin on the language used. Carlson (1999: 367) points out that, on the vertical axis, gravity defines an absolute reference frame that is often, but need not be, coincident with intrinsic and relative interpretations: Sometimes the axes of the three reference frames are aligned, such that the orientation and direction parameters are set similarly. For example, gravity, an upright viewer, and an upright reference object all indicate the same direction for the spatial term ‘above’ (…). However, because people and objects in the world can move and rotate in space, sometimes the axes of the three reference frames are misaligned and assign competing directions to the same spatial term.

This observation holds also for the horizontal plane, except that here it is impossible to align relative, intrinsic, and absolute reference frames in any consistent or generalisable way. This is due to the fact that gravity as source for an absolute reference frame coincides with the prototypical vertical axis (Section 5.2.1), while horizontal directions are usually derived from view directions and internal object parts, both of which are completely independent of absolute directions. As Levinson (2003: 35) puts it: [P]erceptual cues for the vertical may not always coincide, but they overwhelmingly converge, giving us a good universal solution to one axis. But the two horizontal coordinates are up for grabs: there simply is no corresponding force like gravity on the horizontal. Consequently there is no simple solution to the description of horizontal spatial patterns.

More concretely, objects can be conceptualised as having intrinsic properties that define all three axes, such as a car whose bottom can be defined by its wheels and the top by its roof, the front by its prototypical motion direction, and the lateral sides by the positions of the seats for people to sit in. These directions remain true even if the car is turned upside down, in which case two possible ways of talking about the car’s top can be derived – the intrinsic top and the current upper side. However, this applies mainly for internal relations (Section 5.2.2.2) but less straightforwardly for external ones, as research by Carlson (1999) has shown. Her study concerns the activation of different reference frames with respect to the vertical axis, i.e.,

146

Spatial dimensions: Linguistic repertory and issues of application

the assignment of an above relation on the basis of either the object’s features or of gravity (interpreted as an absolute reference frame); the latter was clearly preferred. She concludes (1999: 376): The research presented in this paper focused on the process by which a reference frame is selected. Together, the results indicate that reference frame selection involves the simultaneous and automatic activation of multiple reference frames, followed by a selection process that is assisted by inhibition of the non-selected frames. (…) [D]ifferent reference frames may be used online during apprehension because people have different preferences for using these reference frames to define various spatial terms.

She also notes that, on the horizontal plane, preferences may well be different, possibly favoring relative reference frames for the front region. In describing non-intrinsic above relations, Carlson (1999) conflates absolute and relative reference systems, which, in her work, makes sense because viewers were always in canonical position. Somewhat similarly, Levinson (2003) (who generally focuses on the horizontal axes) claims that the vertical dimension “shares the same major divisions into intrinsic, absolute, and relative” (p66). But it is questionable whether, in cases where the observer is not in upright position, relative reference systems can be derived at all for the vertical direction. In that case, the position of a viewer would need to externally determine a vertical direction that is independent of both gravity and the object’s internal properties. However, parallel to the examples given in Table 5.4 above, it does not seem to make sense to say: (5.21)

* There is a ball above the tree from my point of view.

in a situation where the speaker is not in upright position, and the ball not in a higher position than the tree (in relation to the earth). Or to use a different example: Imagine a person in the air above a swimming pool who has just jumped head-first from a height of five meters. From the perspective of this person, the pool’s ground floor is above the water – but there is no straightforward way the person could actually put this conception into words and be understood. He could, of course, verbalise the perception as such, as in: (5.22)

From this position, it really looks like the pool’s bottom was above the water.

which is somewhat different from stating that the pool’s bottom is above the water from someone’s point of view.

Discourse-external factors involved in the application

147

One study that directly addressed this issue is presented by CarlsonRadvansky and Irwin (1993). They attempted to trigger a relative reference frame for above by placing the viewer in a lying position and showing them pictures. The referent (a point-like fly) appeared in a roughly aligned position with the observer’s head next to the relatum (a chair). Results showed no independent contribution of the relative reference frame: In the described situation, above was only used if the chair was also in a lying position, in which case an intrinsic reference frame was activated. However, above was clearly more favoured in situations fitting the absolute (gravitational) reference frame. According to research by Friederici and Levelt (1990), people use their own head position (i.e., their perspective) for reference in the absence of gravitational cues. The subjects were asked to describe the position of a white ball with respect to a black ball. Since balls do not possess intrinsic axes, intrinsic reference systems could not be employed. As gravity was not available for reference, the subjects could only define the vertical direction by using either their head position or the visual background (pictures of oriented trees) as a cue. The nature of this task suggests to me that the reference system employed was not genuinely relative in the sense that the subjects could have said “from my point of view”. Instead, they employed an absolute reference system based on an assessment of the main axis by cues other than the normally available gravity (see Section 5.2.2.5 below). Thus, either their own head position or the visual background served to define where the vertical axis could be, in order to define the absolute directions. Other kinds of tasks would be necessary to determine whether relative reference systems are available with the vertical axis in the same way as with the horizontal axes: for example, providing different possible points of view that conflict with respect to their “vertical” direction. The main aim and outcome of Friederici and Levelt’s study was not the proof of an existence of relative reference systems with the vertical axis, but rather, to find out how people define the (absolute) vertical in the absence of gravity. Thus, it is a remarkable result that (1990: 265): although on earth gravity plays a dominant role for the choice of spatial reference in adults, mental representations or concepts of spatial orientation can be used quite consistently in its absence. (…) When gravitational cues are absent or task-irrelevant, subjects tend to use the head-retinal reference system.

Pointing to the controversy with regard to whether terms denoting the vertical axis should be included in the group of projective terms (i.e., those

148

Spatial dimensions: Linguistic repertory and issues of application

terms that need to be interpreted on the basis of a reference system), RetzSchmidt (1988: 97) summarises the suggestions of several authors, showing that some major approaches only deal with horizontal terms (in front of, behind, to the left of, to the right of). She then points out that “deictic” (in our terms, relative) uses are not possible with the vertical axis, though “extrinsic” (absolute, gravity-based) and intrinsic interpretations are. Thus, with the vertical axis, the preferred interpretation is absolute rather than intrinsic, and a relative interpretation is (at least nearly) impossible. What about the horizontal axes? Clearly, in this case absolute reference systems play a minor role in Western societies, as opposed to some other cultures where they are used throughout, and where relative reference systems sometimes are not even available at all (Levinson 2003). Miller and Johnson-Laird (1976: 399) claim that people initially assume intrinsic interpretations (see also Ehrich 1985): In interpreting spatial indications, people first determine whether the landmark has intrinsic parts. If it does, they try to interpret the spatial relation intrinsically unless they are explicitly informed to the contrary. If the landmark does not have intrinsic parts relevant to the spatial indication, they must rely on context to provide a deictic interpretation. If both strategies fail, they may ask for more explicit information.

Levinson (2003: 108) observes that the expressions used for relative reference systems are usually based on those for intrinsic ones: Normally, then, the relative frame of reference is expressed through the same classes of expressions as the intrinsic ones, from which they are normally derived by diachronic extension.

But with regard to usage, the MPIP7 research yields a rather different picture (Levinson 2003: 179). In contrast to languages that only offer one kind of reference system for a situation, for example in English or Dutch, both relative and intrinsic frames of reference are available and colloquially used, but the relative frame is clearly predominant for most kinds of spatial description. In the case where more than one frame of reference is available, one may find one frame of reference preferred for one situation, and another for another situation.

Note, furthermore, that the question of variation between reference frames entails several different aspects. Some research (e.g. the work by Carlson 7.

The Max Planck Institute for Psycholinguistics (MPIP) in Nijmegen, see Levinson (2003).

Discourse-external factors involved in the application

149

1999 described above, and other contributions by the Carlson-Radvansky research group) deals with the question of how reference frames are processed, whether one “default” interpretation is activated first, if all possibilities are active at all, and if yes, if that is a serial or a simultaneous process. Taylor et al. (1999) give an overview of research in this direction and apply a technique of cognitive neuroscience (event-related potentials, ERPs) to this question. Comparing processing of intrinsic versus relative reference frames on the horizontal plane, they found a priority selection of the intrinsic reference frame. The relative frame was only activated later, i.e., retrieved from memory if needed for interpretation. The etymological and structural features of the terms applicable for the diverse reference frames as indicated above represent a different direction of research. But the question that is most relevant to present purposes is the preference of reference frames in application rather than cognitive processing, which depends on cultural or language-specific as well as situational factors. In this direction, very few results are available so far. According to Carroll (1997), who compares English and German usage, German speakers tend to rely on their own perspective, while English speakers preferably use the intrinsic properties of the objects involved, in which case speaker perspective is irrelevant (see also Weiß et al. 1996). Carlson-Radvansky and Radvansky (1996) show that a functional relationship (cf. Section 5.2.7), such as a mail carrier who is oriented towards a mailbox, enhances acceptability and production of dimensional terms in an intrinsic reference frame. Where such a functional relationship is ruled out, e.g., if the mail carrier is oriented with his back to the mailbox, relative usages become more likely. This finding is intuitively plausible because “intrinsic terms focus attention on the object” (Carlson-Radvansky and Radvansky 1996: 56) considering that intrinsic sides can only be attributed when considering the object’s properties and functionalities (cf. Section 5.2.1). With respect to room descriptions, Ehrich (1985) notes that speakers seem to take into account that some reference options involve possible conflict with other options, while others do not. Where conflicting interpretations are possible – mainly, with regard to the lateral axis – speakers prefer other kinds of descriptions such as temporal order (dann, danach / then, after that) or use topological or direction-neutral expressions such as an, bei (at) or neben (beside). Furthermore, she observes that keeping a constant perspective (such as that of the speaker) is advantageous because the description would be much more complex using intrinsic reference with different relata. This might, of course, lead to increased use of relative ref-

150

Spatial dimensions: Linguistic repertory and issues of application

erence frames. In the experiment described by Ehrich, perspective is held constant in 95% of cases (Ehrich 1985: 159). Note, however, that speakers in many other cases do switch perspectives, as described in Section 5.3.2.1 below. Apart from these few results, unfortunately, there has not been much research to determine what kinds of reference system are preferred in linguistic tasks dependent on what kinds of factors. This is not surprising considering the fact that comprehensive systematic accounts of reference systems were first presented during the 1990’s. Furthermore, decisions between landmark-based and group-based relative reference systems cannot be predicted at all since the latter kind of reference has not been recognised widely at all so far. Since both these kinds of reference systems are relative, the main distinction here concerns the choice of relatum. Chapter 6 will shed more light on these issues. On a wider scale (encompassing several of the topics addressed in later sections of this chapter), research results provide some insights into speakers’ choices of strategy in spatial descriptions. These will be addressed specifically in Section 5.3.2.1 below. This line of research shows that ease of reference as well as discriminability play a role for linguistic choices. Since spatial axes are cognitively not equal, this may mean that a reference system is chosen that avoids the lateral axis in favour of the frontal one (Tversky et al. 1999). Furthermore, there are first indications (Watson et al. 2004) that processes of interactive alignment (Pickering and Garrod 2004) lead to a preference for a kind of reference system that has already been employed by the interlocutor, even if this entails choosing a different spatial preposition (contrary to the effects of lexical priming). 5.2.2.1.5 Choice and allocation of perspective With respect to the perspective employed, more results are available in the literature. This is also not surprising since the impact of the underlying perspective on the interpretation of the spatial description has been acknowledged in the literature for decades. According to Herrmann and Grabowski (1994), speakers employ the listener’s point of view specifically if there are reasons for this. Obviously, processing is easier for the listener if their own perspective has been used; thus, if the speaker wishes to facilitate interpretation for the interlocutor, they use the interlocutor’s viewpoint. This is especially likely if the partner is expected to have less cognitive or

Discourse-external factors involved in the application

151

linguistic capabilities, e.g., because of younger age or because someone is not a native speaker of the language employed. But it can also be the case in situations where the speaker wants to be polite, e.g., if the interlocutor has a higher status. Furthermore, if actions are involved on the part of the listener, usually the listener’s perspective will be used. Tversky (1999) reports similar results concerning the social implications of adopting another person’s spatial perspective, concluding that there is no default perspective; rather, adopting different perspectives on one occasion is not uncommon. In contrast, Herrmann and Grabowski (1994: 123) propose that, in the absence of specific reasons for taking the interlocutor’s perspective, the listener should assume that the speaker uses their own perspective as a standard strategy. This is because specific mental costs are required for the adaptation of another person’s perspective. As a long tradition in psychology (starting with Piaget’s extensive research, e.g., Piaget and Inhelder 1956) has shown, it takes children several years to be able to mentally adopt a different perspective than their actual one. These costs may differ in situations involving different degrees of divergence from the interlocutor’s view direction, and they are not primarily a problem of language, but rather a general one (Herrmann and Grabowski 1994: 142). The impact of different angles in ego-reorientation on the mental costs involved has been studied extensively by von Wolff (2001), who shows that egocentric axes can be decisive as well as the symmetry axes of a room, depending on the circumstances. Angles diverging from these basic axes are harder to adopt. These results show that the cognitive load of adopting a different person’s perspective is indeed higher than using one’s own point of view. Accordingly, it can be considered more polite to take over the cognitive load, and it is plausible that this is done for specific kinds of addressees and dependent on the scenario. Further evidence for this hypothesis stems from research by Schober (1993,1998b) and Mainwaring et al. (2003). Schober’s findings prove that, with real interaction partners, the proportion of speakers using their own perspective is much higher than with imagined interaction partners. Mainwaring et al. show that the addressee’s perspective is preferred across a range of different configurations and scenarios, except in situations where the target object was situated closer to the speaker than to the addressee. In those cases speakers tended to refer to their own rather than their addressee’s position, using distance terms rather than dimensional terms. In the present framework, this observation does not really (contrary to Mainwaring et al.’s interpretation) relate to perspective choice

152

Spatial dimensions: Linguistic repertory and issues of application

but rather to choice of linguistic strategy, since dimensional terms differ in certain fundamental respects from distance expressions. Crucially, they do not depend on a view direction at all, because distance terms only require a relatum, not an origin (of a reference system). With regard to the third entity perspective, Herrmann (1990) notes that it is unlikely to be used in indoor scenarios. His example (1990: 129) indeed sounds strange: (5.23)

? Vom Stuhl aus gesehen steht die Vase vor dir. ‘Viewed from the chair the vase is in front of you.’

This is particularly deviant because, in this case, the addressee is used as relatum (rather than origin), ignoring the addressee’s orientation, and an inanimate object lacking perceptual abilities is used as origin. Nevertheless, Herrmann’s suggestion that third entity perspectives are especially likely to occur in non–face-to-face situations is intuitively appealing; they will mainly be used in spatial situations that are currently to be imagined, as in route descriptions (1990: 128). Furthermore, if the speaker is not sure about the listener’s current view direction it may be useful to adopt a neutral perspective (Herrmann 1990: 128f.). Other results in the literature concerning choice of perspective in linguistic interaction mostly pertain to discourse-related factors, such as consistency within a discourse and co-ordination with the interaction partner (see Section 5.3 below). Specifically in route descriptions, the speaker can induce the interaction partner to orient themselves in a specific direction, as in (Herrmann and Grabowski 1994: 133): (5.24)

Drehen Sie sich doch einmal um und stellen Sie sich so hin, dass Sie da hinten das chinesische Restaurant sehen. ‘Please turn around and position yourself in a way that you can see the Chinese restaurant over there.’

Kray (2003) has explored this option for application in a mobile tourist guide, a handheld device to be used for navigation through unfamiliar surroundings. He defines it as follows (p.60): An induced frame of reference is a frame of reference that requires the listener to first perform one or more mental or physical actions before the frame of reference is established. These actions include rotation and relocation, which may be applied to the origin and/or the orientation of an original frame of reference, as well as defining the handedness.

Discourse-external factors involved in the application

153

For instance, in contrast to Herrmann and Grabowski’s example the device may not know the actual orientation of the listener and can therefore not simply suggest rotation as in example (5.24). Instead, it may suggest a view direction on the basis of nearby buildings: (5.25)

If you stood in front of the church, the fountain would be to your right.

5.2.2.1.6 Perspective induced by motion Objects in motion (or potentially in motion) induce a further kind of perspective that is independent of intrinsic fronts or perceptual organs. The perspective adopted can be described as though the moving object was viewed from the inside, so to speak, looking in the direction of motion. Thus, even completely symmetric objects, such as a ball, can be ascribed front, back, right and left sides when in motion: (5.26)

John is running to the right of the ball that is rolling down the hill.

This example can also be interpreted with a relative reference system using an external viewpoint, so that the utterance is potentially ambiguous. It is not trivial to predict which of these options would be preferred in natural discourse. In recent experiments addressing this issue, Walker and van der Zee (2005) found that speakers sometimes use this additional source for perspective in the case of motion, but mostly do not, relying on other available reference frames. Furthermore, the assignment of sides in the case of motion holds for scenarios in which the moving entity itself serves as the origin of a relative reference system. In this case the perceptual apparatus does not play a role, but rather the direction of motion, as in: (5.27)

If the ball continues rolling down the hill, it will fall into the hole in front of the wall.

This observation leads Eschenbach (1999: 334) to include motion as a possible origin of reference systems in her definition of relative reference frames (see p. 139 above). She notes (Eschenbach 1999: 341) that, in this case, the relation is judged with respect to a third moving object X (die Steigung ist vor dem Ziel / the slope is in front of the finish). This shows that motion based uses of vor and hinter consider motion of X relative to G (…) rather than motion of G or F.

154

Spatial dimensions: Linguistic repertory and issues of application

In the present approach, the effect induced by motion is not directly related to reference systems but rather to perspective. Thus, any oriented object can serve as the origin of both intrinsic and relative reference systems; the orientation can come about by (potential) motion or by intrinsic features such as perception. Furthermore, similar effects arise by functional ordering relations such as those induced by a queue (see Section 5.2.1 above). 5.2.2.2. Internal relations The discussion so far has concerned solely external relationships, i.e., relations between objects that are spatially separate. These need to be contrasted with internal relationships, which depend either on the conceptualisation of regions inside objects, i.e., object parts, or on the partitioning of a region, such as the observer’s visual field. 5.2.2.2.1 Object parts As shown in the beginning of Section 5.2.2.1, for the application of intrinsic (external) reference systems the origin / relatum needs to have internal properties, such as a front. For the description of these object parts, the same dimensional expressions (right, left, front, back side) are employed as for relationships between objects. Furthermore, other objects can be located inside of other objects and then be described on the basis of the relatum’s internal parts: (5.28) (5.29) (5.30) (5.31)

The table is standing in the back of the room. Der Tisch ist hinten im Zimmer. The woman is sitting in the front of the car. Die Frau sitzt vorne im Auto.

Note that a typical English surface form for such a relationship is a noun in a prepositional phrase using in, while in German the adverb is used (see also Section 5.2.2.4 below). How objects’ internal parts are derived was dealt with in Section 5.2.1. Apart from the processes described there which pertain to permanent object properties, internal object parts can be ascribed spontaneously and temporarily, as noted by Herrmann (1990: 121). Thus, it is possible to interpret an utterance like (5.32) even though trees are usually not ascribed internal parts, lacking typical perceptual organs, functionally salient sides, or direction of typical motion.

Discourse-external factors involved in the application (5.32)

155

The front side of this tree really looks scrubby.

Here, the front side of the tree is not ascribed on the grounds of typical features, but ad hoc, viewed from the outside (cf. Section 5.2.1). There may be various reasons for ascribing sides in an ad hoc way. In this case, the observer’s visual access to the tree may be responsible for the ascription. Conceivably, it may also be possible to ascribe an ad hoc inside view interpretation (Herrmann 1990: 139), for instance, if someone is presently occupying the object: (5.33)

The left side of the tree in which you are sitting really looks scrubby.

Here, the left side of the tree might be determined by the addressee’s position, for example, if the speaker’s own view direction is less clear. However, such a situation seems far-fetched; it is probably usually the onlooker who determines the ad hoc ascription of sides. Object parts can also be ascribed via the perspective induced by motion (see above), as in the following example in which the entity (an army) is partitioned into sections: (5.34)

The front section of the army moves much too fast.

5.2.2.2.2 Partitioning a region into internal parts Regions can be partitioned into internal (relative) sections by adopting a “global perspective” (Carroll 1993, 1997) from an observer’s viewpoint (see also Section 5.3.2.1 below). In that case, the observed region is divided into spatial sections and ascribed part regions that are described by the dimensional terms (front, back, left, right), sometimes explicitly so by referring to sides (such as “on the left/right side”, Carroll 1993: 30). The observed region can be a specific assembly of objects that are perceived as belonging together or being relevant for the discourse situation, or any other kind of region that is within the limits of perception. For example, in German it is possible to say: (5.35)

Siehst du den Apfelbaum dort hinten? lit.: ‘Do you see the apple tree there in the back?’

where the visual field is partitioned into regions in relation to the position of the speaker (and hearer). Then, the area close to the observer is referred to as vorne (front), and the area more distant to the speaker within the visual field is referred to as hinten (back). Note that, in the case of front, left,

156

Spatial dimensions: Linguistic repertory and issues of application

right the regions coincide with the intrinsic perspective of the speaker, but not in the case of back, because the intrinsic back region is located behind the speaker, while in the case of partitioning the visual field the back region is a region in front of, but distant from the speaker. Furthermore, no further objects need to be involved that could provide the origin or relatum of relative or intrinsic reference systems. Thus, utterance (5.35) is based on a different kind of reference system, namely, the partitioning of a certain region (e.g., the visual field) into (internal) subregions of that region. Carroll (1993) claims that English speakers sometimes use the terms denoting cardinal directions of the earth (north, south, east, west) to describe the same partitions (see Section 5.2.2.1 above). According to her research, of the dimensional terms only left and right are used in English for internal regions. However, as noted above in this case there is no way to distinguish a region-based reference system from a speaker-based intrinsic one; even the mention of side may implicitly refer to the speaker’s, not the scenario’s sides. Interestingly, Carroll also claims that English speakers generally prefer intrinsic reference systems that build on the internal features of objects present, while German speakers prefer to partition the perceived scene into internal regions from their own point of view, ignoring the intrinsic features of the objects present. This rather far-reaching claim – involving assumptions about language-specific differences in ‘thinking for speaking’ as well as inbuilt constraints in the languages with respect to the representation of spatial situations – has not, to my knowledge, been confirmed or challenged in a broader range of discourse situations. Note, however, that the option of partitioning a certain region within the visual field into subparts may coincide with referring to the internal regions of a relatum such as a room, notably, if the observer’s view originates at the front region of this region or relatum. For example, hinten im Raum and in the back of the room (based on the room’s intrinsic features, such as the position of the entrance) is in effect equivalent to hinten based on distance from the observer’s viewpoint if the observer is positioned at the entrance. A similar case occurs when the internal regions of a picture are described, using the outside view, cf. Section 5.2.1. Here, also, the perceived region is divided into subregions without relying on intrinsic object features. In contrast to the regions of the visual field or a certain delimited region inside the visual field, however, pictures (being two-dimensional and often viewed upright, as on walls or on a computer screen) are usually described using the lateral and vertical axes, instead of all three axes. Furthermore, what is

Discourse-external factors involved in the application

157

described using the vertical axis in pictures often corresponds to the perceived frontal axis in the real world. In such cases, i.e., if the sides of the relatum are only ascribed through the observer’s view direction, then both kinds of reference systems (internal object parts and partitioning of the visual field) correlate such that English usage does not seem to differ from the German case any more, as data presented in Gorniak and Roy (2004) show. They report participant’s spontaneous descriptions of single objects within a scene containing many similar objects presented in 3D on a computer screen. Here, utterances like the green cone in the back are frequently used for referring to objects most distant from the observer. This parallels the usage in German reported by Carroll (1993) in that the scene is partitioned into several areas using dimensional terms; but equally, the internal regions of the computer screen may have been used as a relatum. Thus, it seems that the nature of the difference in usage between English in German is less clear than previously assumed. 5.2.2.3. Differentiating internal and external relations: An illustration In the situation depicted in Figure 5.2 the adult’s right (not left) arm is patting the child’s shoulder, which exemplifies the inside view of the adult. This conception (used for internal relations and descriptions of parts only) coincides with an external intrinsic perspective, as in: (5.36)

The child is standing to the right of the adult.

where the child’s (external) position is described on the basis of the adult’s intrinsic features. However, an outside observer standing opposite the two people could still say: (5.37)

From my point of view, the child is standing to the left of the adult.

using a relative reference system. This description, in turn, coincides with taking the picture as an internal reference frame, since pictures are viewed from the outside: (5.38)

The child is in the left side of the picture.

Finally, the description could also pertain to the observer’s visual field: (5.39)

The child is on the left (hand side).

158

Spatial dimensions: Linguistic repertory and issues of application

(5.40)

Das Kind ist links / auf der linken Seite (meines Blickfeldes).8 lit.: ‘The child is left / on the left side (of my visual field).’

Figure 5.2. Internal vs. external relations

Thus, different and (on the surface) contradictory descriptions can be used for the same situation, depending on whether internal or external relations are described, an on whether the objects involved are viewed from the inside or from the outside. 5.2.2.4. Topological distinctions expressed linguistically Generally, there is no conclusive (empirically founded) evidence in the literature that specific syntactic forms correspond directly to specific conceptions and reference systems. However, some authors have suggested subtle distinctions with respect to topological (or distance) relations between the entities related by dimensional terms. For instance, according to Herskovits (1986: 173), the application of the dimensional preposition groups (see Section 5.1.2) is constrained by their ability to express not only internal versus external relationships, but also a third topological option, namely contact. Landau and Jackendoff (1993: 229) describe a very similar categorisation of regions pertaining to spatial expressions other than dimensional terms:

8.

As the empirical investigations in the next chapter show, the visual field is never made explicit, though this kind of reference system is indeed used in some frequency at least by German speakers (Section 6.3.5.2).

Discourse-external factors involved in the application

159

[D]istance is digitized into several discrete categories. There appear to be four levels described by English prepositions: (1) location in the region interior to the reference object (in, inside); (2) location in the region exterior to the reference object but in contact with it (on, against); (3) location in the region proximate to the reference object (near); and (4) location distant from the reference object (far and perhaps beyond).

Following Herskovits, topological relations have consequences for the application of dimensional terms as well. She provides an overview of the dimensional preposition groups with respect to their applicability for the first three of the categories proposed by Landau and Jackendoff. The distinctions pointed out by Herskovits are well suited to explain, for instance, the difference between (5.41) and (5.42): (5.41) (5.42)

the child in front of the car the child in the front of the car

where in front of can only be used externally, and in the front of only internally. Both terms cannot express contact, in contrast to on the front of, which can only be used in cases where contact is involved. Other expressions, such as on the left hand side of, are more flexible and can be used internally, externally and expressing contact. Note that Herskovits does not specify whether contact should be internal or external; in Landau and Jackendoff’s categorization it is a subcase of external relations. Talmy (2000: 197f.) proposes a similar classification which accounts for Landau and Jackendoff’s categories (2)–(4). In cases of group-based reference, it is not always clear whether the underlying reference system should be regarded as internal or external, or (potentially) both, since the target object could either be viewed as being part of the group, or its position could be specified externally, relative to the rest of the group. With the usage of dimensional adjectives as in example (5.19) above (repeated below as (5.43) for convenience), this distinction does not seem to yield any difference in interpretation. Example (5.18) (repeated as (5.44)) in which the group-based reference is made explicit, in contrast, seems to induce an external interpretation – the target ball is located to the left of the other balls. The distinction is only important when assigning a spatial template in order to determine the extension of the region denoted by the dimensional term (see Section 5.2.3 below). (5.43) (5.44)

Kannst du den linken Ball holen? ‘Can you fetch the left ball?’ The ball is located to the left of the other balls (from my / your / a third point of view).

160

Spatial dimensions: Linguistic repertory and issues of application

For German, Klabunde (1999) claims that syntactic variants of the dimensional terms can be differentiated with respect to the external/internal distinction (1999: 151): While prepositions like vor locate a primary object with respect to a region that is outside the reference object, the corresponding adverb vorn locates the primary object in an internal region provided by a reference object. However, a preferred interpretation with an internal relation does not hold for the adverbs rechts and links referring to horizontal axis-based regions. (…) [T]he adverbs davor, rechts davon etc. have different meanings based on an external region again. Table 5.5. Realisations of horizontal relations in English (acc. to Herskovits 1986) and German (acc. to Klabunde 1999) external

internal contact internal or external internal, external, or contact

lateral to/ by the {left/right} {side} of right/ left of rechts/ links {da}von rechterhand/ linkerhand in the left/ right {hand} side of in the side of at the {left/ right} {side} of on the left/ right of links/ rechts on the {left/ right} {side} of

frontal before/ behind in front/ back of by the front/ back of back of {da}vor/ hinter in the front/ back of vorne/ hinten on the front/ back of at the front/ back of

Chapter 6 will shed more light on these issues. Table 5.5 integrates Herskovits’ (1986: 174) and Klabunde’s (1999: 153) proposals, combining the categorisation of regions with the axial distinctions. From this overview, it becomes apparent that expressions for lateral contact only appear to be missing in both languages, as well as terms that flexibly express either internal, external, or contact relations on the frontal axis. With respect to the vertical axis, Carroll (1997: 142) concludes from experimental work that at least oben in German can be used to denote both internal and external relations, as opposed to above which is only used for external relations. Ehrich (1985: 147) observes that the German adverbs vorne and hinten lose their intrinsic reading if used together with primary deictic expressions (such as hier/dort/da – here/there). According to her interpretation, expres-

Discourse-external factors involved in the application

161

sions like vorne are interpreted intrinsically with respect to a relatum like a cinema, and expressions that combine deictic terms with dimensional adverbs (such as da vorne, lit.: ‘there in front’) are interpreted deictically with respect to the observer. In the present framework, avoiding the intrinsic / deictic distinction, interpretation is as follows. In both cases internal relations are expressed, but with two different kinds of relata. In the former case, the relatum is the cinema, and in the latter case, it is the observer’s visual field. Then, an expression like vorne can be used in any situation, i.e., without necessitating a situational framework such as that offered by the cinema. Eschenbach et al. (2001) show how German dimensional adverbs can be combined with spatial prepositional phrases denoting internal regions, rather than with an external relatum: (5.45) (5.46)

Die Teller stehen oben im Schrank. lit.: ‘The plates are standing above in the cupboard.’ Das Auto steht hinten in der Garage. lit.: ‘The car is standing at the back in the garage.’

Their observations basically support Klabunde’s suggestions, namely, that German dimensional adverbs are used exclusively for internal relations. The argument of the adverbs may be expressed in a prepositional phrase that itself contains a spatial expression. Alternatively, it can be omitted and be derived contextually. This possibility leads to a semantic and structural ambiguity: Example (5.45) could either be interpreted with respect to the cupboard as a bounded internal region, or it could be interpreted with respect to another region that needs to be derived contextually, as shown in (5.47) where the prepositional phrase im Schrank is not the reference region of the adverb, but rather the prepositional phrase in der zweiten Etage which (in an alternative interpretation) could have been implied in (5.45). (5.47)

Die Teller stehen oben in der zweiten Etage im Schrank. ‘The plates are standing above on the second floor in the cupboard.’

Eschenbach et al. (2001) point to similar ambiguities with regard to combinations of dimensional terms, which involves additional interpretational problems that will not be pursued here.

162

Spatial dimensions: Linguistic repertory and issues of application

5.2.2.5. Environment-induced reference systems Up to now, all reference systems could be said to have been used to represent relationships in the real world with its canonical relations: the vertical is defined by gravity, and other directions are derived from it. Imagine, however, a situation in which real world conditions are less relevant than environment-centered conditions: for example, consider a picture lying on a table, depicting a situation involving a vertical direction (such as people standing upright); as the picture is lying on the horizontal plane the depicted vertical orientation does not correspond to the vertical orientation in the real world. Talking about the picture would surely involve reference to the vertical with respect to the environment depicted there, not with respect to the real world. Similarly, during a dive into a sunken ship whose ceiling is oriented at a certain angle with respect to the usual upright direction, people might well adopt the directions from the ship’s tilted environment, creating a new “vertical”. The results by Friederici and Levelt (1990) can be explained in precisely these terms (see above): in the absence of gravitational cues, a new vertical is created and reference systems employed on that basis (in this case, on the basis of either the head retinal or the visual background). All these situations can be viewed as a kind of pretense: the environment is treated as if it corresponded to canonical conditions in the real world. Then, reference systems are employed as usual. This interpretation highlights a major fact about reference systems, namely, that they are based on humans’ conceptualisations rather than any systems or structures in the real world. 5.2.2.6. Summary of reference systems: Systematic overview As indicated initially, the literature does not provide a systematic comprehensive overview of possible reference systems and their relationships to perspectives. Figure 5.3 sums up the findings of the present section. It is restricted to those variations in which spatial dimensional terms are used. The network represents conceptual-semantic relations and therefore does not make any claims about linguistic realisations. It is assumed here that, in general, specific linguistic forms do not map directly on specific variants of reference frames. However, some suggestions for linguistic restrictions are represented in Table 5.5 above.

Discourse-external factors involved in the application vertical

VERT ICALHALFAXIS

above below

AXIS

horizontal

163

lateral

LAT ERALHALFAXIS

frontal

FRONT ALHALFAXIS

right left

HORIZONT ALDIRECT ION

back front

} FRONT SIDE-OF

relatum referent

} VIEWDIR

intrinsic relatum = origin REFSYS

relative + origin ! horizontal

RELAT IVET YPE

outside inside

group-based landmark-based

absolute ! no perspective required ! vertical

proj-terms + referent + relatum

function object

VIEWDIR-DEF

external

perceptual-organs ! inside view orientation

3rd-pers

3RDT YPE

animated-being ! inside view motion-direction ! inside view

PERSP T OPOLREL

+ : participants involved ! : inferences from usage of this feature }: “and” relation ] : “or” relation

location ! orientation towards entities ! relative 1st-person + speaker ! inside view 2nd-person + addressee ! inside view unknown--underdetermined intrinsic-features

entity internal

ENT IT YFEAT URES

R] INT FEAT

ad-hoc-ascription movement ! inside view

RELAT UM

speaker visual-field

DEFINEDBY

addressee unknown

Figure 5.3. The usage scope of spatial dimensional terms: Options of reference systems, axes, and perspectives

outside-view inside-view

164

Spatial dimensions: Linguistic repertory and issues of application

The network should be read as follows. Since dimensional terms are relational, they always involve two entities to be related, which are here called referent and relatum. Reference systems can be conceptualised either with respect to the real world (which is regarded as the usual, canonical case) or with respect to features of the environment. In any case, first, an axis (lateral, frontal, or vertical) and a topological relation (internal or external) need to be determined. Each of these decisions has further consequences. Internal relations concern either the region defined by the visual field of somebody (usually either the speaker or the addressee), or some entity that either possesses intrinsic features, or is ascribed sides in an ad hoc way, or is currently in motion. If the latter is not the case, then the regions can be defined by conceptualising either a view from the inside, so to speak, or from the outside. External relations rely, on the one hand, on a reference system (intrinsic or relative), and on the other hand, on perspective (except in the case of the gravity-based vertical absolute reference system, which is the only option of absolute systems that uses dimensional terms). Perspectives are directions that can be defined by an origin such as a person or other animated being, an object, or a location; and by the direction of movement. In the case of objects serving as origins, they either have features that resemble perceptual organs, are conceptualised as oriented towards the related entities even if they do not possess intrinsic orientation, or by specific object functions that specify directions (as with a car that usually moves in a certain direction, or a pointing arrow). Intrinsic reference systems can be further specified in that, in the case of in front, also the referent can define the front side. In all other cases, the relatum’s sides are decisive. Furthermore, if object functions define an intrinsic relationship then outside or inside view directions can be adopted (this has only consequences with regard to the ascription of left and right, but not with frontal or vertical axes). Finally, relative reference systems can be either group-based (relying on objects of the same kind as relatum) or landmark-based (relying on an object of a different kind as relatum). It should be noted that intrinsic object features come into play in two different ways: on the one hand, internal regions can be defined by object features, and on the other hand, a view direction is sometimes derived from them. In the former case, it is the relatum’s intrinsic features that are relevant, and in the latter case, the origin’s. Furthermore, the view from outside or inside can play a role both in determining internal regions and in defin-

Discourse-external factors involved in the application

165

ing an intrinsic reference system that builds on the intrinsic features of an object that can be viewed either from the outside or from the inside. Movement always induces a view from the inside, just as the adoption of a person’s perspective (be it speaker, addressee, or a third person) does. This is relevant in the partitioning of an entity’s internal parts as well as in the adoption of a perspective induced by direction of motion. This kind of dependency between features is captured via an exclamation mark (!) to specify which kind of feature can be inferred when a certain feature occurs. Thus, relative reference systems can only occur with the horizontal axes. 5.2.3. Spatial templates The generalisations made so far have abstracted from a number of phenomena which will now be examined. The present section addresses the fact that the entities related by dimensional terms do not need to be situated directly on one of the axes with respect to a reference system. Instead, they must be related to the axis in some way: In one view, they must be situated within a region that is defined by the axis denoted by the dimensional term. In a different view, such as that proposed by Herskovits (1986) (Section 5.1.3), the semantics of the dimensional terms contain a gradedness with regard to applicability, similar to the kind of prototypicality proposed by Rosch (1978). In other words: the closer the object is to the axis, the better the fit to the dimensional terms. The terms are not discrete and mutually exclusive, but applicable to different degrees depending on the distance to the focal axis. The degree of applicability may be influenced further by the overall spatial situation and the discourse task (Section 5.3.1), by functional aspects associated with objects (Section 5.2.7), etc. These two views are not in conflict. While it may be true that the terms are most straightforwardly applied when the object is situated on the axis itself or very close to it, it may also be true that there is some limit where applicability becomes inacceptable in any kind of situation, thus delimiting the region defined by the dimensional term. These effects are captured in the notion of spatial template as used, for instance, by Carlson-Radvansky et al. (1997). A comprehensive and thorough overview of previous insights as well as new experimental results with regard to prototypicality effects of dimensional terms is presented by Vorwerg (2001). Here I summarise the most important results.

166

Spatial dimensions: Linguistic repertory and issues of application

Gradedness effects have been dealt with in the literature in several respects. First of all, psychological research points to the fact that gradedness effects are not restricted to language but can also be observed in nonlinguistic tasks (e.g., von Wolff 2001, Vorwerg 2001: 143ff.). Therefore, it is likely that the observed effects are interrelated with degree of cognitive effort. Franklin et al. (1995) carried out non-linguistic and linguistic tasks to determine cognitive differences between the various regions in surrounding space. Results showed that the front region is considered larger than the others, memory for location is more precise in the front than anywhere else, and linguistic descriptions are more precise for the front region. These findings reflect the special status of the frontal direction compared with the other horizontal directions (see Section 5.2.1 above), at least for surrounding space. Interestingly, the term front is treated as a default that can even be omitted; thus, a description like a little bit to the right means that the entity’s location is very close to the front (not right) axis. These effects may very much depend on the discourse context, as recent research by Klippel et al. (2004) on direction concepts shows. In their experimental setting, the region around the prototypical front direction was much smaller than all other regions, reflecting a narrower concept of straight in the context of route directions. Again, however, the front direction stands out as peculiar in comparison with the other directions. With regard to the application of dimensional terms in localisation tasks, Vorwerg and Rickheit (1998: 208) summarize previous research: There is increasing evidence that spatial categories, such as projective relations, are not discrete, mutually exclusive either–or–categories based on critical features with well defined boundaries. Converging experimental and computational results support the idea that projective relation categories are analog, overlapping, internally structured categories based on prototype comparison and with fuzzy boundaries.

For instance, the expressions left and behind can be used straightforwardly for 90° and 180° angles, respectively. However, the more the angles between the target object and the relatum depart from these focal axes, the more linguistic modifications are used for specifying the spatial relation. Simple expressions are acceptable and applicable in a certain range; outside this range compounds or modifiers such as left front or a little bit to the left are more typical (Zimmer et al., 1998). From findings such as these, the following generalisations can be drawn: First, the most straightforward or “best” case of application is position on the axis (e.g., the front axis) itself. The more the referent’s position

Discourse-external factors involved in the application

167

deviates from this axis, the closer it gets to another axis (e.g., right), and so the more likely a term for the second axis (right) will be used, or a combination of both terms. Thus, front can still be used close to the right axis, depending on the situation (see also Freksa 1999). The applicability areas of each term with its neighbor overlap. In the present view, the graded structure is not interpreted as part of the abstract semantics: therefore, what overlaps is not the concepts of the spatial relations, not the semantics of the neighboring terms, but only the (spatial) situations in which they can be applied. In other words, though front and left are distinctive terms for speakers as part of the vocabulary, just as speakers agree that green and blue are different colours, there are still situations in which either term can be suitably applied. The largest area that has been proposed for applicability of a dimensional term is a half-plane (e.g., Herskovits 1986: 181f., RetzSchmidt 1988). Therefore, while the applicability areas of left and front are not mutually exclusive, those of front and back as well as left versus right surely are. But this seems to be the only hard constraint. Here is a summary of the findings on graded structure discussed so far: A dimensional term denotes a spatial relation between two objects. One object serves as relatum, and the other is positioned within a region surrounding a focal axis with respect to the relatum, based on the conceptualisation of a reference system. The size of the region depends on contextual factors but is at all times limited to a half plane. With unmodified dimensional terms the most likely position is on the focal axis itself; with increasing distance from the axis applicability decreases.

In addition to the effects of distance to the focal axis, there is a further effect of the distance to the relatum (Hayward and Tarr 1995). The acceptability of dimensional terms at some distance from the axis increases with distance between the objects. Thus, the angle between the two related objects seems to be important. On the axis itself, however, distance does not have any effect on applicability. With regard to this latter finding, Landau (2003) reports systematic differences between linguistic applicability and cognitive effects: In language, monomorphemic terms are preferably used along the axes but not outside of the axes, independent of distance; while in memory, there are graded effects of distance both with respect to the axis itself and with respect to distance away from the axis. These observations apply for distance on a small scale, i.e., in the experiments all objects were situated in a region within the visual field. Further aspects of proximity versus distance are discussed in Section 5.2.6 below.

168

Spatial dimensions: Linguistic repertory and issues of application

5.2.4. Spatial properties of the objects involved Herskovits’ (1986) description of the semantic core of dimensional preposition groups starts from idealised conceptualisations of geometric points rather than extended objects. This approach, which has also been followed so far in the present work, corresponds to the insight that the application of dimensional terms does not pose any specific constraints on the spatial properties of the involved objects, in the way along does, which requires an elongated relatum (Talmy 2000, Landau and Jackendoff 1993: 226). This fact makes it possible to a certain extent to abstract away from the spatial features of real-world objects. Intrinsic reference systems rely on the relatum’s axial structure and/or its axially determined internal regions. However, the presence of intrinsic features is no requirement imposed by dimensional terms because – as long as an origin, an observer’s view point, can be determined – it is always possible to employ a relative reference system that is independent of the relatum’s properties. However, the fact that dimensional terms do not require specific object properties does not entail that such properties do not influence application. Herskovits (1986: 184ff.) points out that, in the application of the terms, the extension of the real-world objects as well as their (un)boundedness in any direction does play a role. For instance, if the relatum is a square rather than a (conceptualised) point, the focal axes that are relevant for the description of each of the dimensions (on the horizontal plane) are also extended. This could (depending on the underlying reference system) look as depicted in Figure 5.4 below. Depending on other factors, application of the dimensional terms is also possible outside the focal regions, mirroring the gradedness of applicability when assuming focal axes with (idealised) points as relata. Gapp’s (1995) findings apparently corroborate this account; he concludes that the region of applicability increases with the extension of the reference object perpendicular to the focal axis. In contrast to this view, most approaches tend to treat even extended objects as point-like, using the central axis of each region as focal axis and treating the surrounding area as the field of possible application with gradually decreasing applicability. This approach captures the fact that objects’ “center of mass” is a likely focal point serving as a basis for spatial relations. Apart from the center of mass, functionally salient parts of objects can serve the same purpose (Carlson et al. 2003). In both cases, the object may essentially be treated as point-like, with differing criteria used for determining the focal point of the object in each case.

Discourse-external factors involved in the application

169

back focal region left focal region

relatum

right focal region

front focal region

Figure 5.4. Spatial regions of an extended relatum

Herskovits (1986) also deals with objects that are not as symmetric as that depicted in Figure 5.4. In general terms, she describes the attribution of spatial regions as follows (1986: 185): Enclose the object between parallels to the base axes, so that each such parallel is tangent to the object. The right hand (respectively left hand, front, or back) strip extends toward the object up to the part of the boundary limited by the tangency points first encountered. This procedure requires the object to be bounded so it can be enclosed between those parallels, and thus does not apply to roads, rivers, etc.

Similar ideas can be found in the more recent literature, with different terminology. Carlson et al. (2003: 124) refer to the “strip” as “grazing lines” (with respect to the vertical axis): (…) placement of the located object above or below a horizontal line running through the topmost point of the reference object. We refer to this line as the grazing line.

Similarly, a closely fitting “bounding box” around an object with curved axes may serve as an approximation for determining spatial regions (e.g., van der Zee and Eshuis 2003). With unbounded linear objects, such as roads, reference is possible only with respect to either the frontal or the lateral axis at a time, dependent on the observer’s viewpoint (Herskovits 1986: 186). Thus, if the observer is standing on the road, looking ahead, a description like to the right of the road makes perfect sense, since the road is bounded in the described direction with respect to the observer. However, this is not true in a situation in which the observer is standing beside the road and looking toward it at a right angle, since there is no right region in such a situation due to the perceived unboundedness of the road.

170

Spatial dimensions: Linguistic repertory and issues of application

So far, different spatial properties of the relatum have been considered, as opposed to the referent that is still conceptualised as a point. This approach corresponds to a generalisation Talmy (2000: 187) proposes: [T]he seeming majority of spatial elements schematize the figure solely as a point or related simple form, in contrast with the treatment given the Ground.

As a counterexample to this general tendency, he shows that the pathrelated expression across depends on rather specific features of the Figure as well. But usually, due to the specific generalisable relations between Figures (referents) and Grounds (relata) (see Section 5.3.2.2 below), the Ground’s features are more relevant for spatial reference. However, Talmy’s observations concern speakers’ tendencies for choosing Figures and Grounds in discourse, rather than the attribution of regions once the choice is made. Herskovits (1986) considers some abstract cases in which both objects that are described relative to each other are extended (which, of course, cannot be excluded for real discourse situations). In such a case, the respective spatial regions attributed to them may (partially) overlap. Here, conjoined expressions are expected depending on which of the objects is larger and how much overlap there is (Herskovits 1986: 187). It may also matter whether the whole extension of the located object is described, as in a description like (5.48) (given by Herskovits): (5.48)

The garden is to the right and behind the house.

Another possibility is that only salient parts of the located object are selected for the spatial description, as in example (5.49) (Herskovits 1986: 187) where not the whole river is supposed to be in front of the observer, but only that part which the observer is looking at. (5.49)

The Thames is in front of me.

Finally, bounded located objects may extend along a portion of an unbounded linear relatum, as in (Herskovits 1986: 188) (5.50)

the land to the right of the road

depending, again, on the observer’s position and viewpoint. Since Herskovits’ work, the impact of the involved objects’ extensions has been addressed empirically by several authors. Vorwerg (2001), for example, reports several experiments involving various extended (rectangle-shaped) relata together with a small ring as referent. Results show that orientation and shape of the relatum relative to the observer’s view direc-

Discourse-external factors involved in the application

171

tion systematically influence the application of dimensional terms with respect to the expressions’ graded typicality structure. In Vorwerg (2003), she shows specifically how the relation of the relatum’s longitudinal axis to the view direction influences choice of dimensional terms (“tilt effect”), thereby proving the impact of perceptual aspects (Vorwerg 2003: 328). According to Coventry and Garrod (2004: 55), there is a certain danger involved in treating the objects participating in a spatial representation as abstract points. Ignoring most of the object features (apart from a schematised representation that is required by each spatial term), as Talmy (1983), Landau and Jackendoff (1993) and others propose to do, may lead to a failure to appreciate the influence of object functions on the application of spatial terms (cf. Section 5.2.7 below). 5.2.5. Conceptualisations of dimensions The application and interpretation of dimensional terms can be influenced by variability in the conceptualisation of the dimensions involved. For instance, imagine a tree that could be conceptualised as three-dimensional, as two-dimensional (ignoring the vertical dimension, as is often the case in the application of the horizontal expressions, see Section 5.2.1), or as onedimensional (ignoring the extension on the horizontal plane because of the salience of the trunk). Now, an utterance like (5.51)

The bird is in front of the tree.

could receive any of three different interpretations, depending on the conceptualised dimension: In a three-dimensional conception, the bird could fly anywhere between the tree’s extension and the observer. In a twodimensional conception, the bird sits on the ground between the tree’s horizontal extension (including branches and leaves) and the observer. In a onedimensional conception, the bird is situated on the ground between the trunk and the observer. Alternatively, it is possible also to ignore the frontal axis and refer to height and/or the lateral axis only, again reflecting a twodimensional rather than three-dimensional conceptualisation. Notice that these different dimension conceptualisations do not have any direct reflection on the linguistic surface, in contrast to some other terms (e.g., in the bowl vs. on the bowl; see Section 2.2.5).

172

Spatial dimensions: Linguistic repertory and issues of application

5.2.6. Proximity and distance Intuitively, in order for a dimensional term to be applicable, the related objects need to be situated within a reasonable distance. In Section 5.2.3 above, it was noted that distance within the visual field may have an influence on the applicability of monomorphemic terms, i.e., with increasing distance greater deviations from the focal axes are acceptable. However, there are some limits to the distance where dimensional terms are applicable at all. The two objects must be conceived of as spatially related, i.e., proximal to each other to some degree. In the literature, this idea is sometimes associated with a concept introduced by Miller and Johnson-Laird (1976) with regard to spatial relations in general, namely, the notion of a region of interaction. This is described as follows (1976: 59): In order to take account of spatial relations, the perceptual process must not only register place, but relations between places, which entails perception of a spatial region containing the place of the thing. The region of a thing can be thought of as a rather indeterminate penumbra surrounding it. The advantage of region over place as a perceptual predicate is that regions can overlap even though things cannot. Thus, two things whose regions overlap can be seen in spatial relation to each other. We will say that object x is in the region of object y when x is spatially close enough to y to have the sort of interactions with it that normally occur between x’s and y’s. This definition of region is deliberately vague, because the perceptual attributes of a region are correspondingly vague.

For instance, an expression like here can be used to refer to the personal space around oneself (although it may also have other, more extended interpretations, as exemplified by here on Earth). The conceptualisation of a region of interaction around an entity depends on experience, thus (Miller and Johnson-Laird 1976: 59): Once an object has become familiar, its characteristic region of interaction seems to be appreciated with the same immediacy as many simpler perceptual attributes.

This observation is in line with more recent findings on objects’ affordances, e.g., by Tucker and Ellis (1998) who find that “seen objects automatically potentiate components of the actions they afford”. This line of research goes back to Gibson’s theory of perception (Gibson 1979), stating that perception is intertwined with potential actions on the environment. Thus, forms and functions of objects are conceptually closely related.

Discourse-external factors involved in the application

173

With topological expressions, regions of interaction are undoubtedly relevant. For instance, in utterances like (5.52)

The house is located at a river.

the interpretation of at is certainly dependent on the interactants’ concepts of possible interactions between houses and rivers. But is this notion also applicable for dimensional terms? With the terms denoting the earth’s cardinal directions, there seems, in principle, to be no limit with regard to distance, as the following example shows (which may be used in a context where one wishes to point out this striking coincidence): (5.53)

We will spend our holidays at a hotel that is situated exactly south of our home.

However, exchanging south of for a dimensional expression does not work, even ignoring the fact that the spatial relations do not necessarily coincide: (5.54)

? We will spend our holidays at a hotel that is situated exactly behind our home.

Pribbenow (1992: 160) proposes a number of criteria that account for the intuition that two objects need to be sufficiently close to each other in order to be related (pointing out elsewhere that this does not apply for compass directions). Although her criteria are formulated primarily for topological and distance-related spatial terms, such as bei (at), some of the conditions may well be relevant for the application of dimensional terms: x visual access between the two objects, x noninterruption of the spatial region between the objects (i.e., no other objects or obstacles should be in the way), referred to by Herskovits (1986: 84) as nextness, and x primacy of the relation between these two objects compared to any other objects present which could serve as relatum. Formulated in this way, these are rather strong claims that can (and should) be tested empirically. I do not know of any research so far addressing these issues directly, probably due to the fact that empirical studies dealing with dimensional terms usually employ scenarios in which these conditions are met to begin with, for all objects present in the scenario. In situations where the relatum is not a single object but several objects of the same kind as the referent (i.e., group-based reference), all of the involved objects may need to meet the criteria with respect to each other in order to be conceptualised as suitable for spatial reference. Possibly, the primacy condition would conflict with the fact that the referent is related to

174

Spatial dimensions: Linguistic repertory and issues of application

more than one object, yielding a different situation in which the objects need to be conceptualised as a group, and therefore should be sufficiently close to each other. It would be an interesting question for empirical psycholinguistic research to find out whether the objects need to be close to each other (how close?) to allow for group-based reference, or whether qualitative-functional requirements such as Pribbenow’s constraints would indeed be sufficient. Possibly, a formal definition such as that proposed by Thorisson (1994) could be useful here; he defines a proximity score for the distance of each object in the domain to a particular object. The formula is used, for example, by van der Sluis and Krahmer (2000) in an object reference scenario to determine whether objects are situated within the focus space of another object. It is important to note that the proposed constraints only apply for the relationship between the objects, not the observer’s viewpoint (except in the case of intrinsic reference systems where both coincide). Thus, it is entirely possible to say (5.55)

The ball is in front of the table.

in a situation in which the ball is occluded from sight of the observer, other objects are located in between the observer’s position and the two objects, etc. The only requirement is that the two related objects are sufficiently close in terms of the above claims. (Recall that the usage of behind often even implies that the observer does not have visual access to the referent; cf. Section 5.1.3.) Another point to note is that the constraints may sometimes be cancelled by metric distances. Thus, an utterance like (5.56)

The box is two meters behind the barrel.

does not require that no other objects are placed in between box and barrel, etc. Indeed, proximity is not required at all in that case. Thus, the above example (5.54) might become more acceptable (though still unlikely) by introducing metric measures (see also Herskovits 1986: 84f.): (5.57)

We will spend our holidays at a village that is situated exactly two hundred kilometers behind our home.

Likewise, specific discoursal and situational requirements might have the same effect of cancelling the nextness condition, such as, for example, if the position on one of the axes is specifically important (because they are perceptually related), as in:

Discourse-external factors involved in the application (5.58)

175

Could you please place this picture on the wall on a straight line to the right of the upper doorframe at the other end of the room?

Accordingly, it seems to be a good solution to treat the proposed constraints as default conditions which are cancellable by contextual factors, in the same way as many features of dimensional terms can be affected by the situation. One contextual factor that reinforces (rather than cancels) the nextness condition is if a background grid (such as a network of streets or a chessboard) is available (Herskovits 1986: 82): If there is in the context a grid-like pattern and if that grid-like pattern is relevant to the speaker’s point, then that grid-like pattern will define the tolerance in the use of the projective prepositions.

In general, however, proximity as defined in Pribbenow’s terms does not require that the objects are situated particularly close to each other, as long as the criteria are met. Thus, the notorious problem of context dependency in assessing a suitable measure of closeness is reduced considerably, since proximity is not defined in terms of distance but in terms of functional aspects of the situation. Conceivably, terms denoting the vertical direction behave differently, maybe as an outcome of the special status of the vertical axis and the interconnection with gravity as described above. Tyler and Evans (2003) propose a difference between over/under and above/below with respect to proximity as part of their semantics (see Table 5.3 above); furthermore, research by Garrod and Coventry points to further influences of functional aspects (cf. Section 5.2.7 below). Altogether, it seems that proximity is more relevant to over than to above, sometimes involving contact in the case of over but not above. This idea is exemplified by the associations evoked by the following two sentences (Tyler and Evans 2003: 112): (5.59) (5.60)

Nora twirled over the polished floor. Nora twirled above the polished floor.

Here, it seems that Nora must be dancing in the air in the second case. These observations raise the question whether horizontal expressions also show differences concerning proximity. According to some findings in the literature, this is indeed the case; however sometimes in subtle ways. For example, Talmy (2000: 198) states that immediate adjacency is a prerequisite for the usage of some specific preposition groups of dimensional terms, namely, in front of, in back of/behind, on one side of/beside, on the right/left of. He claims:

176

Spatial dimensions: Linguistic repertory and issues of application

The fact that these expressions cannot be used to localize Figures at a greater distance shows that they indicated relative adjacency to the Reference Object. For example, a bike directly lined up with the front of a church but three blocks away cannot be said to be ‘in front of’ the church.

In contrast to this, other expressions can be used to express some amount of distance, in effect removing the “adjacency condition”. According to Talmy, expressions using to rather than on belong in this group, as in: (5.61)

The bike is to the right of the church.

This sentence can, according to Talmy (2000: 199), be interpreted to mean “anywhere from three feet to three blocks.” This difference between to and on is well represented in the literature (see also Herskovits 1986: 188f.). Another suggestion with respect to different requirements in proximity comes from Landau and Jackendoff (1993: 230), who claim that in back of and behind differ in distance: A tree may be right behind (proximal), way behind (distal), or right in back of a house, but ‘The tree is way in back of the house’ sounds odd or colloquial (to us, anyway).The standard use of in back of seems to be restricted to proximal distance (and possibly contact) whereas behind and the colloquial in back of are unrestricted.

Herskovits (1986: 189) suggests that the use of at the left / right imply even closer proximity than on the left / right: These imply the closest proximity consistent with a reasonable world: the bottle is at your left suggests that the bottle is immediately accessible to hand, while it could be further away with the bottle is on your left.

It can be concluded from this section that, although there seems to be some amount of uncertainty in the literature with respect to the details, there is a general agreement that most dimensional terms require some amount of proximity for applicability, mediated by the specific expression used in each case. This proximity may be induced by perception or conceptualisation rather than “objective” environmental factors. As described above, Pribbenow (1992) suggests qualitative-functional criteria for the assessment of proximity, which may be taken as useful default conditions cancellable by contextual factors, rather than strong constraints.

Discourse-external factors involved in the application

177

5.2.7. Functional features of objects As discussed in Section 5.1.3 above, it has been noted by several authors (e.g., Tyler and Evans 2003) that the spatial relationships expressed by dimensional terms are often associated with additional functional features, corresponding to the situations in which they are most often applied. Also, proximity can be defined on the basis of functional aspects, as just shown. The present section deals with a different aspect of functional geometry (Garrod et al. 1999): namely, the impact of functional relationships between the objects involved rather than the spatial relation itself. While functional aspects of the spatial relationships, such as SUPPORT, SEQUENCE, or ACCESSIBILITY, can be (and have been) represented in a fairly abstract way, objects’ functional features are peculiar to the objects involved. Nevertheless, they may have a strong influence on the application of spatial terms. As Coventry and Garrod (2004: 55) put it: “what objects are fundamentally influences how one talks about where they are located”. Or in the words of Coventry and Mather (2002: 182): If one comes across the x is over the y then what this means will be determined by object knowledge. If the figure and ground are a tablecloth and a table, then information about the functions of these objects is available from lexical entries of the nouns accessed, and the spatial relationship between these objects can then be established. If sense selection requires this to be done anyway, then there is no advantage to lexically representing all the senses given the problems that this entails.

Nevertheless, the question can be asked to what degree the influence of objects’ functions can be generalised in order to make predictions about the applicability of dimensional terms. Both kinds of functional relationships often interact to a high degree: a typical situation of employing an expression like in (which is generally agreed to involve an associated functional feature of CONTAINMENT) is one involving a container, such as a bowl. The applicability of in then interacts with the degree to which the functionality of the object as a container is fulfilled (Garrod et al. 1999). That is, in this case applicability depends on a good match between the associated feature of the spatial term, and the functional features of the object(s) involved. As Coventry and Garrod (2004: 61) point out, this may depend on how the object is conceptualised in the first place: Objects are often associated with particular functions, and retrieval of this functional information from memory may promote the application of different routines. (…) For the same configuration of located object and reference

178

Spatial dimensions: Linguistic repertory and issues of application

object, when the reference object is labelled a plate, on is judged appropriate, but when the reference object is labelled a dish, in becomes appropriate. So how the same object is conceptualised can influence the types of computations that are performed on the visual scene.

It is assumed here that spatial terms do not involve a very high number of associated functional features, defined by the requirement that they hold across discourse contexts and are therefore part of the semantics of the terms. In contrast, there may be an infinite number of possible object functions and features, interacting with possible applications of spatial terms. While the object functions themselves may not be generalisable, the interaction with applicability may well be. Several proposals to this point have already been put forward in the literature. Since the amount of research in this field is vast (see Coventry and Garrod 2005 for an overview), I focus here on the most relevant findings. As a starting point I first consider some intriguing proposals put forward by Herskovits (1986). Consider (Herskovits 1986: 74): (5.62)

the cat under the table

Here, the cat is clearly not under the lowest parts of the table – i.e., under the legs – but under the functionally salient part of the table, namely, the table top. Herskovits (1986: 73) refers to this phenomenon as a pragmatic principle of salience: One can use a noun which basically denotes a whole object to refer to the region occupied by a part of it that is typically salient.

This is not a random principle, but speakers must share some kind of experience in which the relevant part is salient. Such salience can, for instance, consist in the fact that only a part of the object is visible, as in a rabbit under the bush. This principle is well suited to account for findings such as the “toothbrush experiment” reported, for instance, in Carlson et al. (2003). Here, subjects were asked to place a tube of toothpaste above a toothbrush. Not surprisingly, many participants chose the functional part of the toothbrush rather than its center of mass for the placement. Thus, Herskovits’ principle of salience does not only apply to object reference (referring to the table instead of the table top) but also to interpretations of spatial relationships expressed linguistically. Spatial relationships are interpreted to pertain to the most salient or functional parts of the involved objects, rather than to their overall shape. The Attentional Vector Sum Model presented in Carlson et al. (2003) captures this idea on the basis of the em-

Discourse-external factors involved in the application

179

pirical finding that speakers’ attention is focused on the most salient object parts, i.e., those that people typically interact with. A corollary of the pragmatic principle of salience is the following (Herskovits 1986: 74): The geometric description applicable may be the base of the object (i.e., its area of contact with the ground plane).

This principle explains cases in which, of a three-dimensional scene, only two dimensions are relevant for reference, as in the block in the circle. This observation corresponds to the empirical finding that references to a square can be resolved unproblematically (though with increased reaction times) in situations involving only a cube (Duwe et al. 2002). However, according to Herskovits two dimensions cannot be ignored, using only one dimension, as would be the case in the block in the line. In a similar vein, Pribbenow (1992: 143) points out that certain regions may be prioritised for spatial descriptions. For example, the following two utterances both refer to the inside of a room, but still evoke different spatial representations: (5.63) (5.64)

der Teppich im Zimmer ‘the carpet in the room’ das Gemälde im Zimmer ‘the painting in the room’

In (5.63), the carpet is typically associated with the floor (thus confined to a two-dimensional representation of the room, rendering the vertical dimension irrelevant), while in (5.64), the picture is likely to hang on one of the walls. Thus, typical object placements evoke associations with different aspects of the spatial situation, which can of course be cancelled by providing a different kind of context (such as repositing a newly delivered carpet or painting in a room prior to arranging them at their future location). This may also mean that speakers associate specific distances with objects in a typical situation, as in: (5.65)

John was sitting in front of the TV.

Here, John is probably not sitting particularly close to the TV set but rather at a distance suitable for watching TV. Thus, the region of interaction (Section 5.2.6) is influenced by typical associations. Furthermore, the position of the referent may deviate completely from the geometric definition of a spatial term, if the objects involved evoke a purely functional interpretation, as in Klein’s (1991: 100) example:

180 (5.66)

Spatial dimensions: Linguistic repertory and issues of application Unter dem Wams trug er ein rotes Leibchen. ‘Under the jerkin he wore a red bodice.’

where the functional aspect of COVERING or OCCLUSION seems to be decisive. Apart from the salience principles, Herskovits proposes that speakers also adhere to relevance principles (pp. 76ff.) which are closely connected to the functions of the objects involved. For instance, the examples (5.67) (5.68)

There is some milk in the bowl. There is some dust on the bowl.

are distinguished not by the spatial relationship between the entities involved, as the choice of prepositions (in versus on) might suggest, but by the fact that in (5.67) the bowl’s function as a container is important, while in (5.68), it is the bowl’s surface that is relevant. Thus, in situations involving objects’ functions a description is chosen that supports the association of that function. This is accomplished straightforwardly by applying a term which contains a relevant functional aspect in its semantics. With respect to empirical research in functional geometry, there is a large number of publications dealing, for example, with issues of containment in the application of in; see Coventry and Garrod (2005) for an informative classification of findings. One main result of this research direction is that CONTAINMENT (or ENCLOSURE as used in Coventry and Garrod 2004) as well as SUPPORT both reflect the effects of a “dynamic-kinematic extra-geometric routine of location control” (Coventry and Garrod 2004: 89) that has been examined in detail by manipulating features and orientations of the objects involved in a spatial relation situation. The basic idea is that in and on are used together with containers and supporting surfaces, whose essential purpose is to constrain the location of other objects. The applicability of the prepositions then depends on the degree to which the objects involved can fulfill this function of location control. Therefore, the functional geometric relation is a matter of degree and can (in many situations) be transitive. With dimensional terms, results are less abundant; in fact, the only dimension that seems to have been researched in any detail is the vertical. Coventry and Garrod (2004) propose that dynamic-kinematic routines similar to those observed for the application of in and on influence the applicability of the vertical prepositions: the effects of gravitation determine whether the relationship between referent and relatum is one of potential influence. These effects interact with the functionality of objects, such as

Discourse-external factors involved in the application

181

those shown by the already mentioned “toothbrush experiment” (Carlson et al. 2003). However, several results indicate that above is not as much influenced by functional factors as is over. For instance, Coventry et al. (2001) found differences in applicability between the different vertical expressions over/under vs. above/below when used in situations involving an object typically associated with a protective function. Over/under were more suitable to support this associated function than the other prepositions. Similarly, Coventry and Mather (2002) present the results of several experiments in which the applicability region of over was highly influenced by the given context (for example, a plane dropping a bomb on a city), while above was not affected in this way. This is in accord with the proposal of Tyler and Evans (2003) who suggest that the semantics of over and under contains a functional element of potential reach, in contrast to above/below which are not associated such a functional element (see Section 5.1.3 above). Thus, the objects’ functionality influences the area of applicability only when terms sensitive to this functionality are used, which seems to be a matter of degree. Carlson-Radvansky and Radvansky (1996) found that intrinsic reference frames are preferred in situations involving functional features of (familiar) objects, while relative reference frames are more likely in non-functional situations. In their study, differences between axes are unfortunately not spelled out. However, as shown in Section 5.2.1, the frontal axis is privileged in some respects. Therefore, in front of may represent a special case with regard to application in functional situations. For example, interaction between two human beings usually occurs in a “canonical encounter” situation (Clark 1973) where the two persons are oriented towards each other. Likewise, interaction between an object and a person is most likely when facing the object, which itself is oriented with its functionally defined front side towards a person. Accordingly, Tyler and Evans (2003) propose a functional aspect of ACCESSIBILITY for in front of. Additionally, in front of is the only relation that allows for the usage of intrinsic reference frames induced by the features of the referent rather than the relatum (cf. ex. (5.11) above). Therefore, it would be interesting to investigate more details, comparing the axes. From this kind of evidence regarding topological terms, dimensional expressions, as well as some further findings on distance-related terms and between, Coventry and Garrod (2004) develop a functional-geometric framework that offers an account of how these spatial terms are interpreted in natural discourse by way of establishing situation models (p. 146):

182

Spatial dimensions: Linguistic repertory and issues of application

[V]arious constraints work together to determine the appropriate meaning of a spatial expression in context. (…) [S]ituation models offer the vehicles by which these multiple constraints come together. Meaning in the functional geometric framework reflects the result of all these constraints coming together in the situation model, which supports the most informative relation between a reference object and a located object. By informative, we mean that the model should support the strongest inferences that could be drawn about the scene.

Such inferences can include situation-specific assumptions about why the speaker chooses one spatial term in favour of another, as in example (5.69) (cf. Coventry and Garrod 2004: 145) where the spatial relationship may be correctly described by near, but at would be a more usual description if the man is indeed playing. The authors suggest that (5.69) can be used in ironic contexts, implying that the man is not a good pianist. (5.69)

The man is near the piano.

Thus, the choice of a spatial term reflects underlying assumptions about the relationship between referent and relatum, not only with respect to the geometric relation, but also with respect to additional factors that may be context-related, due to dynamic-kinematic aspects, and object-function related. On encountering an assignment of a spatial expression to a relation, the hearer can make suitable inferences according to the discourse and situational context. There is some evidence that another term is even more sensitive to functional aspects, namely, the primarily temporal term before which can also be used in spatial situations, although it is not used frequently in everyday discourse. McIntyre (2001) points out that before can be used in a locational sense primarily if a functional aspect of potential interaction is associated. He contrasts the following examples, which – interestingly – are very similar to example (5.69) above presented by Coventry and Garrod (2004: 145) with respect to the difference between near versus at: (5.70) (5.71)

Gertrude was sitting before the piano. Everyone in the room was fixated on the television. Fran and Stan were on the sofa, Gordon was sitting near the fire and Gertrude was sitting {in front of/ *before} the piano, her back resting on the keyboard.

According to McIntyre, (5.70) entails that Gertrude is looking at the piano, playing it, or otherwise interacting with it. If before is replaced by in front of, this kind of inference is rendered optional. However, a context in which interaction is excluded renders the application of before unacceptable,

Discourse-related factors influencing application

183

while in front of (which according to the findings reported above is also associated with a functional component) is still fine. Thus, McIntyre (2001: 1) claims that before in its spatial sense requires a functional interpretation, while in front of does not: The interactional entailment does not come from the context in which before appears; rather the preposition insists that it be embedded in a context where interaction is possible.

This line of argument, while intuitively appealing, still awaits empirical foundation. Research on the functional components of prepositions as well as the contribution of knowledge about objects has largely focused on the more typical spatial expressions, usually those used for the vertical axis. But note that McIntyre’s hypotheses are well compatible with the approach proposed by Coventry and Garrod (2004). 5.3.

Discourse-related factors influencing application

The present section addresses actual discourse situations in which spatial dimensional terms may be used. The discourse factors influencing linguistic choices broadly fall into three categories: discourse task, spatial strategies, and dialogue processes. 5.3.1. Discourse task Much research on spatial language focusses on the ways in which human speakers describe the spatial relation between two entities in a given context, answering a question like “Where is the object?” (e.g., the MPIP research reported in Levinson 2003, as well as many others of the publications mentioned in the previous sections). But there are many other contexts in which spatial language could be used. Generally, the motivation why a spatial relation should be described is decisive; it inevitably influences the choice of linguistic strategies and the linguistic details of the spatial description. In route descriptions, for example, typically a goal location is described via reference to streets and landmarks which can be easily identified in the real world (e.g., Tversky and Lee 1998). There, salience and dimensions of buildings play a role in the choice of landmarks; spatial relations are often sufficiently outlined via simple and vague expressions. Such a scenario

184

Spatial dimensions: Linguistic repertory and issues of application

differs fundamentally from psycholinguistic spatial localisation experiments where participants are asked to specify an entity’s location relative to another. Crucially, besides the reference to spatial relationships between entities in the scenario, other aspects and elements are also essential in this kind of discourse task, for example, the segmentation of the route into (chunks of) smaller units such as paths between decision points (Klippel 2003), the mention of landmarks (Daniel and Denis 1998, Michon and Denis 2001), the inference or explicit mention of start and end points (Habel 1988, Denis 1997), and so on. All of these jointly contribute to clarifying the spatiotemporal relationships involved in wayfinding tasks (Klippel et al. in press). Another aspect that clearly distinguishes this kind of task from other spatial descriptions is the fact that route descriptions reflect a linear structure. Spatial dimensional terms occur mainly in two ways: on the one hand, directional dimensional terms are used to indicate the direction of movement on a route segment, especially in the case of change of direction at a decision point. On the other hand, when landmarks are mentioned this may be done either by simply indicating their proximity using expressions like near or at, or a dimensional term may precisify the spatial relation involved, with respect to the path, the moving person, or a different entity. A different discourse task in which spatial reference is employed concerns the description of spatial relationships in a given scenario, for example, the structure of an apartment (Linde and Labov 1975), or the interior of a room (Ehrich 1985). The strategies identified for such tasks will be briefly addressed in Section 5.3.2.1 below. Another fundamentally different discourse task from that usually addressed is that one of several similar objects needs to be identified on the basis of spatial location, due to the absence of other cues such as distinguishing features or perceptual salience. The relevant question in such a task is “Which” rather than “Where”. I will call this kind of task referential identification,9 thus pointing to the fact that this area of spatial language interacts with other kinds of reference to entities in the real world, on the basis of other kinds of features. The broader area of object reference, from which a number of relevant aspects can be drawn, is itself well researched in the literature. But spatial referential identification is clearly under9.

This notion corresponds closely to that of referential communication in the psycholinguistics literature (e.g., Krauss and Fussell 1991). Here the aspect of identification is perceived to be more crucial than that of communication, since objects can be identified even in the absence of a communication partner.

Discourse-related factors influencing application

185

researched, in spite of its being highly relevant to the field of spatial reference. Therefore this kind of discourse task is addressed in specific detail in this section and in the empirical Chapter 6. In many scenarios, language is not the only medium through which referential identification can be achieved. Much research deals with the question of how gestures are combined with language in order to facilitate communication (e.g., Sowa and Wachsmuth 2002, Kopp et al. 2004). Gestures accompanying language seem to enhance speakers’ abilities at formulating complex spatial relationships (e.g., Emmorey and Casey 2002); also, the interpretation of spatial terms may be influenced by accompanying gestures (e.g., Landragin and Romary 2003). Kühnlein and Stegmann (2003) found that verbal utterances tend to be shorter and less informative in a proximal area where eye gaze and pointing gestures are viewed as sufficient. With increasing distance from the speaker, utterances become longer and more complex. If the distance to the speaker is too great to allow for object identification via gesture, the utterance is usually formulated in a way that is complex enough to be sufficient for singling out the reference object. In the following, I will specifically address some issues pertaining to the discourse task of object identification using language only. Identifying – or referring to – one of several objects in a spatial scenario differs in some basic respects from specifying a spatial relation. For instance, it should involve taking the overall configuration into account to a higher degree than in “Where” scenarios, in which other objects present may often be ignored (or are simply absent in experimental settings). In contrast, “Which” questions presuppose (shared) knowledge of the situation to such a degree that the target object can be distinguished sufficiently from all other objects in the scenario, making the choice of spatial reference highly dependent on the number and arrangement of objects present. Linguistically, there are some systematic differences between utterances relating to the two kinds of discourse tasks. For example, the German dimensional adjectives (such as das linke) and the English dimensional superlatives (such as leftmost) can only be employed as an answer to “Which”, not “Where” questions: There is no way in which “Where is the object?” can be answered by “The leftmost one”; even stating “It is the leftmost one” could at best be interpreted as a fairly indirect way of answering the question. Other syntactic options, in contrast, are available for both kinds of questions: “The object to the left of the barrel” (a typical answer to “Which object do you mean?”) can easily be augmented to “The object is (located) to the left of the barrel” (a typical answer to “Where is the object?”). Note

186

Spatial dimensions: Linguistic repertory and issues of application

that German dimensional adjectives as well as English dimensional superlatives hardly occur in the literature on spatial language, mirroring the fact that discourse tasks of referential identification have been poorly researched so far. As is well known in the literature, in referring to objects speakers do not describe all features of the intended objects they can conceive of. Instead, they use linguistic representations that are specifically suitable in a discourse situation, for example, in order to distinguish the intended object from others present. In an account of how this is done, Herrmann and Deutsch (1976) formulated general principles of greatest distance, dimension preference, redundant verbalisation, and partner-adapted verbalisation. These principles capture that speakers, in choosing a reference strategy in an object identification task with several different objects, usually do at least the following: First, they analyse the target object with respect to properties that can establish a (maximum) contrast to competing objects. Thus, if there are two black boxes of different size, the speaker chooses size for object reference. If the objects differ in several respects, the speaker chooses the property where the distance to the competing object is most obvious. Thus, if there are two boxes, one of which is very small and dark blue, while the other is very big and black, size – rather than colour – will be chosen for reference (greatest distance). Individual preferences also play a role, especially if the distances are viewed as equal (dimension preference). Second, the speaker encodes as many properties as needed for unambiguous object reference, but usually not more, being economic (cf. Grice’s maxim of quantity, Grice 1975). But if the object reference task is complicated by the availability of multiple options, minimal differences in distance, and equal levels of dimension preference, speakers may encode more properties than needed (redundant verbalisation). Third, speakers adapt to their interaction partner’s view of the situation, taking into account cognitive and social distinctions, etc. (partner-adapted verbalisation). Herrmann and Deutsch are exclusively concerned with object reference in non-spatial settings, designing their experiments purposively in a way that spatial reference is ruled out. A situation in which the opposite is the case would be one in which target objects differ from competing objects only in spatial position, and additionally, objects of a different class may serve as possible landmarks (relata) for reference (cf. Chapter 6). This situation has not often been directly addressed in the literature, since most

Discourse-related factors influencing application

187

research on spatial reference focusses on scenarios in which the position of an already agreed upon object is to be defined by relating it to another present object, usually of a different kind. Previous experiments involving referential identification are reported, for example, in Schober (1993, 1998b) and Mainwaring et al. (2003), with the main focus on spatial perspective choice. Weiß et al. (1996) also present a study in which a dimensional term serves to identify a referent (in this case, a place rather than an object). Their results show that (at least German) speakers are influenced by situational factors such as formality and proximity, and that (at least English) speakers prefer non-dimensional expressions for referential identification when no oriented relatum is present, rendering intrinsic reference less accessible. Furthermore, the results of a study by Pobel et al. (1988) suggest that the principles established by Herrmann and Deutsch can be applied also to spatial scenarios. However, due to the matrix-like arrangement of a high number of objects, the only kind of spatial reference found in their study relied on rows and columns. Herskovits (1986) noted that, although the graded structure of dimensional terms applies in most discourse contexts, there are situations in which an expression like to the right, without modification, can denote the full right side (i.e., a half-plane) with respect to the relatum (p. 182): [T]he loosest interpretation of the preposition is adequate, provided that obvious contrasts in the context allow the expression to fulfill its function of identifying the place of the located object.

This applies if there are no competing objects in the same spatial region. In case of the presence of further objects within the half-plane, unmodified dimensional terms can nevertheless be employed for contrastive reference. In that case, Herskovits’ shifting contrast principle applies (p. 81): If two objects, A and B, are placed in a relation to a reference object in such a way that the ideal meaning of a preposition (…) is truer of A than of B, then one can use that preposition to discriminate A from B so that the locative phrase will be assumed true of A but not of B.

For instance, if to the left is truer (i.e., more valid) of A than of B, i.e., A is closer to the left reference axis than B is, A will be recognised as the target object even if to the left could also be applied to B. Here, gradedness effects again come into play, in that the object closest to the prototypical direction will be identified (see also Freksa 1999). In a spatial context, the immediate real-world situation is specifically decisive. In order to distinguish the inferred object from competing ones, it

188

Spatial dimensions: Linguistic repertory and issues of application

must first be decided which objects within an environment are relevant, i.e., whether they belong in the domain of reference or not, which may or may not be a matter of subjective interpretation in a discourse context. Furthermore, object reference (in formulation as well as interpretation) is highly influenced by the focus of attention with regard to a subset of objects present (e.g, Beun and Cremers 2001, Regier et al. 2005). In ambiguous situations, reference is resolved with respect to a subset in focus (Kessler et al. 1996). Rieser (1997) shows that conversational participants often agree on situationally dependent representations in order to achieve smooth and effective communication. Moreover, especially in cases of qualitative or vague spatial instructions, interactive negotiation supports the decision about how the instruction should be interpreted (Wachsmuth and Cao 1995). Thus, spatial referential identification can be described as a speaker’s decision as to how much, and which, information to convey on the basis of the external situation as well as the dialogue history and the interaction partner. One recent study in which spatial referential identification plays a central role is presented in Gorniak and Roy (2004), who address speaker’s strategies and range of variety in describing objects in a visual scene containing similar objects of two colours. Their study was specifically designed to induce participants to produce spatial descriptions, because objects differed only in position (and colour as a non-discriminative feature). Results show a considerable degree of flexibility and creativity in speakers’ spontaneous spatial descriptions in complex scenarios. In their work, utterances like on the left side are interpreted as referring to the object that is situated at a more extreme position to the left than any other objects in the scene. This indicates that, as hypothesised above, not the relation to the prototypical axis is relevant but the relative position on that axis, as compared to the competing objects. Starting from these diverse observations, concrete research questions – and tentative answers – can be formulated as follows for referential identification in situations that only allow for spatial reference: How do speakers choose a reference system and point of view? If a partner is present it can be expected that participants will often choose their partner’s perspective, especially if the partner is expected to act (Herrmann and Grabowski 1994: 123), and that speakers will adapt their utterances to their interlocutor’s in various respects (Clark 1996; Pickering and Garrod 2004), for example with respect to the choice of reference systems (Watson et al. 2004). These findings further specify Herrmann and Deutsch’s princi-

Discourse-related factors influencing application

189

ple of partner-adapted verbalisation (see above). But the literature does not provide much evidence with respect to which reference systems are preferred if several options are available in an object identification scenario. This is so in part because such scenarios have not been in focus very often in spatial cognition research, and specifically, group-based reference has hardly been mentioned in the literature so far at all. Likewise, little can be said for the case of several options for perspective when there is no interaction partner. In the present work it is assumed that the identification task – requiring reference on the basis of spatial contrast – plays a role in the decision, since situations may arise in which one kind of reference system or perspective enables a clearer contrast than the other ones available. On the grounds of Herrmann and Deutsch’s principles, this means that, in addition to adapting to their interlocutor, speakers choose a reference system and perspective that is suitable (just as a unique object property is) for distinguishing the target object from competing ones within the referential domain. If attention is focussed on specific subsets of objects in a scenario, a reference system and perspective may be chosen that only distinguish objects within that subset. How do speakers choose a reference axis? Within a reference system, the frontal (front/back), the lateral (left/right), and the vertical (above/below) axes are available for reference. With spatial localisations, the reference axis is chosen that the target object is closest to. But with referential identification, this may not yield unambiguous reference. It is hypothesised here that competing objects play a role in deciding about a reference axis: A reference axis is chosen that is best suitable for distinguishing the target object from competing ones within the referential domain and/or the subset in focus, considering the principle of greatest distance. How explicit are speakers about underlying reference systems and origins? The situation may offer various reference systems yielding similar results (e.g., Carlson 1999): in some situations, to the left could equally well be used for an intrinsic reference system conflating origin and relatum, for a group-based reference system using the other objects as relatum, and for a landmark-based relative reference system. In such situations, to the left can simply be used without further specification since no conflicts arise. Similar observations apply with respect to the chosen point of view (origin); if several options are available yielding no difference, speakers do not need to provide an expression like from my point of view. In intrinsic reference systems, the origin is often specified because it coincides with the

190

Spatial dimensions: Linguistic repertory and issues of application

relatum, as in in front of me. The interesting case is when different reference systems and points of view yield different results, so that to the left can be interpreted in different ways. Herrmann and Grabowski (1994: 132) state that speakers are usually not explicit with regard to the perspective used. This may be because they tacitly assume that the interlocutor will understand the intended meaning even without the additional effort, since perspective needs to be expressed by additional linguistic material and since the partner’s perspective is (in many kinds of situations) conventionally preferred. But with respect to reference systems, no such conventions are known so far, so that it is hypothesised here that relata – but not necessarily origins – will be made explicit in case of potential conflict. Under what circumstances do speakers modify and combine dimensional terms? In spatial localisation tasks, speakers increasingly use modifiers and compounds as distance to the reference axis increases. According to Herskovits (1986) this is not expected in identification tasks (see above). Likewise, Herrmann and Deutsch’s principles, as well as Clark and WilkesGibbs’ (1986) principle of minimal cooperative effort, predict that speakers will not provide more information than needed, unless several options with equal properties compete. Thus, it is hypothesised here that speakers use a dimensional term without modification or combination with another dimensional term in case there are no competing objects for which the same description applies to the same degree. If the target object is placed where it could equally well, and equally unambiguously, be referred to by two terms, such as to the left and in front, both may be combined. Thus, according to the findings presented in this section, it can be expected that pragmatic principles such as contrastivity can undermine prototypicality effects like those described in Section 5.2.3 above. Crucially, using minimal effort for achieving contrastivity implies that no detailed spatial description will be given if this is not required by the scenario. This observation is a strong argument against including prototypicality structure in the semantic specification (cf. the discussion in Section 5.1.3). Note, however, that the gradedness in applicability of dimensional terms is not completely ruled out in this case, since unmodified dimensional terms can only be used when there are no competing objects for which the same description applies to the same degree. The validity of the above hypotheses will be addressed in the (spatial) empirical part of this book, Chapter 6.

Discourse-related factors influencing application

191

5.3.2. Strategies of describing spatial relations In addition to – and in interplay with – the influence of the task given in a discourse, linguistic choices are also affected by another crucial factor, namely, by speakers’ individual strategies. While a certain amount of idiosyncrasy must be expected, recent research has pointed to several systematic ways in which speakers can choose among available options for describing spatial relations. The most important of these concern, on the one hand, the level and framework of description, and on the other, the assignment of Figure and Ground. 5.3.2.1. Level and framework of description While temporal information can easily be presented linguistically in the same order as the events occurred (though this is not a constraint), with space there does not seem to be any default order. But in fact, there are some conventions with regard to how spatial information is presented. For example, if entities are conceived of as being ordered on one of the spatial axes (usually the frontal one, see Section 5.2.1) the order of mention may be pre-structured. A different situation is one in which a route is to be described. Then, linguistic descriptions correspond to the order in which the spatial entities on the route will be encountered when the route is actually travelled. Similarly, in instructions of tasks involving the construction of spatial objects, object parts will be mentioned in the order in which they are needed. Experiments presented by Buhl et al. (2000) show furthermore that the order of representation is influenced by the order in which the objects in a spatial situation have originally been encountered, i.e., by the order of knowledge acquisition. The authors call this effect the according-toexperience–effect (Geneseeffekt). Moreover, if knowledge has been acquired in two diverging ways, the first of those is more decisive than the second (anchor effect). This effect was shown to be stronger than another well-known psychological effect called the recency effect, according to which the most recent experience is the most decisive one. If the situation (or its original presentation in the process of acquisition) does not itself provide an order on any kind of axis, it is still possible to impose an order. Thus, Herrmann and Grabowski (1994: 114) describe the option of describing spatial surroundings by way of conceptualising a ge-

192

Spatial dimensions: Linguistic repertory and issues of application

neric wanderer, i.e., an imaginery tour through the spatial setting. This option is often realised in German by employing the impersonal pronoun man (equivalent, perhaps, to English they or the impersonal use of you). It is however also possible to address the interaction partner directly, and represent the situation linguistically (by using a second person pronoun) as though he or she did the “wandering” through the spatial situation. A ground-breaking early study in this area was conducted by Linde and Labov (1975). They had participants describe the structure of their own apartments, and identified several systematic regularities in the participants’ strategies. For example, they follow a mental route, beginning with one room, and systematically describing further rooms according to an imagined tour through the apartment. However, if a branch during the tour leads to a dead end, speakers mentally “jump” back to a previously described place. Speakers indicate directions using terms like left and right, and they differentiate between main and minor rooms. Ehrich (1985) reports a study in which the interior of a room should be described, which was done either by employing the generic wanderer strategy, or by focussing on the spatial relations between objects in functional arrangements (such as tables and chairs). She notes that the former strategy represents a connective, linear description, while the latter may be disconnective, involving spatial leaps and discontinuities. In disconnective descriptions, the usage of dimensional terms is much more frequent than in connective descriptions, where the direction-neutral expressions (neben) and temporal terms (danach, describing the experiences of the imagined wanderer) are more likely to occur. If the aim is to instruct the interaction partner to reach a certain position in space, a common strategy is to start by leading the partner into the target region using large spatial units in the description, then become more and more specific until the target is reached, similar to general structures of search behavior: “from the coarse to the detailed” (Herrmann and Grabowski 1994: 152). Another strategy listed by Herrmann and Grabowski (1994) is to refer to the category or framework of what is to be described. For example, after mentioning a village it is possible to refer to typical parts of villages by relying on world knowledge that can usually be safely assumed to be shared by both interlocutors. Specific knowledge about such frameworks may reduce the complexity of describing spatial relations considerably. Such spatial hierarchical frameworks are often described in terms of conceptual models, which serve as basis for examining how humans deal with wayfinding tasks (e.g., Timpf et al. 1992).

Discourse-related factors influencing application

193

Since this is so, speakers often activate frameworks that are not readily pre-defined in a situation. For example, Garrod and Sanford (1988) examine what they call “discourse models” (or situation models) which are constructed for the specific purposes of the present discourse, similar to the notion reviewed by Zwaan and Radvansky (1998). The basic idea is that situation models are required even when there is a real world scene to talk about, i.e., the conceptualisation of the scene plays a role in referring to it. Thus, in the experiments described in Garrod and Anderson (1987), speakers developed consistent description schemes for reference to objects in a maze, a process which Garrod and Sanford (1988) claim to be dependent on underlying mental models of the maze configuration (1988: 149f.): Thus mental models of space have the effect of breaking down any scene into significant spatial entities – points, lines, regions or volumes of space – associated with the various objects in the scene, and then representing significant spatial relations between those entities.

The imposed relations between the entities in the maze scenario then correspond to other kinds of relations perceived between objects in the real world, e.g., the functional relationships described in Section 5.2.7 above. The analysis shows that the underlying models can constrain or enhance referability and discriminability of entities in the maze situation. Carroll (1993: 25) distinguishes three categories of describing the location of objects, one of which corresponds to Herrmann and Grabowski’s (1994) generic wanderer. Further options are to refer locally and successively to entities in a point-by-point fashion, or to adopt a “global frame of reference” concerning a single spatial structure. These three different options of spatial descriptions are reminiscent of Siegel and White’s (1975) famous distinction between landmark vs. route vs. survey knowledge. Landmark knowledge can be viewed as similar to point-by-point or local descriptions, route knowledge mirrors the perspective of a generic wanderer, and survey knowledge may be linguistically expressed in terms of global frames of reference as described by Carroll. There is not much literature on how the three kinds of knowledge interact with the three kinds of linguistic strategies. Taylor and Tversky (1996) and Tappe (2000) investigated speakers’ descriptions of sketch maps in terms of route, survey, and “gaze” perspectives (the latter reflects a static view on the scene from a certain position). These are distinguished, for instance, by lexical and conceptual choices such as using an observer as relatum as in hinter dir / behind you with a route perspective, versus using absolute directions (e.g., das Rathaus ist im Süden / the town hall is located

194

Spatial dimensions: Linguistic repertory and issues of application

south) with the survey perspective (Tappe 2000: 75f.). Descriptions based on a route perspective represent a linear conception, in spite of the fact that the described picture is two-dimensional. In Tappe’s study, results show that speakers seldom consistently conceptualised the depiction as a route, even in a condition in which the sketches were presented dynamically. The descriptions in both static and dynamic conditions contained features of route descriptions as well as features of descriptions compatible with a survey perspective. One observation often pointed out is that speakers readily switch between different strategies within a discourse (e.g., Herrmann and Grabowski 1994, Carroll 1993). Tversky et al. (1999) address the question why this may be the case, in spite of the fact that it is cognitively costly (as they show). They find that adopting and maintaining a specific perspective (or strategy) also involves certain costs, which at times can be higher than switching to a different strategy that has become available (1999: 404): In some cases, the costs of switching perspective may not be greater than the cost of staying with the same perspective. Some spatial situations may be more readily described from one perspective rather than another. Indeed, some spatial arrays seem to encourage route perspectives, and others, survey perspectives. Having a single prominent pathway through an environment and landmarks at the same size scale, for example, encourages route descriptions.

Specifically, dimensional terms are hard to compute, with the effect that “people avoid using them when other options are available” (ibid). This observation relates to findings reported in Mainwaring et al. (2003), who compare speakers’ strategies of spatial reference in a referential identification task using a number of different configurations. They find that speakers avoid using dimensional terms – specifically, if the lateral axis would be needed for reference – especially if the target object can be specified using distance terms. This is only the case, of course, if the competing object is not equidistant to the relatum (i.e., the entity to which the distance term refers, which can be the speaker, the addressee, or a landmark) as the target object. In that case, speakers prefer near to far even if this means that they specify the target object’s relation relative to their own position rather than their addressee’s, which is usually the preferred strategy in terms of perspective choice. Altogether, Tversky et al. conclude (1999: 409): Ceteris paribus, speakers select reference objects that are salient to communication partners and terms of reference that are relatively easy to produce

Discourse-related factors influencing application

195

and comprehend. Switching perspective may enable the more salient reference object and easier reference terms.

These findings may also shed some light on how speakers choose reference frames at all (as addressed in Section 5.2.2.1 above, in contrast to perspective selection). A different approach to identifying speakers’ strategies has been developed by Fischer (2003), who focusses on the specific situation of linguistic interaction between humans and artificial communicators, i.e., computers or robots. In a setting involving a robot moving to a goal object indicated verbally by a human, Fischer identified several levels of description that represent the users’ strategies of achieving their communicative goal. On the highest level, they refer directly to the goal object, using locative dimensional terms. A lower level is to make the robot move towards the goal object by indicating a direction, i.e., by using directionals rather than locative descriptions of the goal object’s position. Finally, especially in cases of continuous failure, speakers turn to an even lower level, indicating minor actions like turning on the motor (Fischer and Moratz 2001). This account highlights how speakers’ strategies are intricately connected with, on the one hand, the overall discourse situation that they are initially unfamiliar with, and on the other hand, the development of the discourse, involving success or failure with the different kinds of strategy. Furthermore, it shows that it is by no means self-evident that speakers directly refer to a goal object even in a situation where a goal position within the field of vision is to be reached. Rather, the discourse situation may encourage the speaker to turn to more fine-grained, incremental instructions for the same aim. Speakers’ reasons for choosing directional rather than goal-based instructions are as yet not entirely clear. In the present chapter, several aspects have already been mentioned that might provide some explanation: As Tversky et al. (1999) point out, dimensional terms (used for indicating the spatial position of goal objects) are often hard to process and difficult to employ because of the complexity of factors involved in their interpretation. This fact is mirrored in the present account of the diversity of perspectives and reference systems (as depicted in Figure 5.3 above) plus the various discoursal and object-related aspects that come into play in the application of locative dimensional terms. These factors become increasingly serious with increasing complexity of the spatial situation, and decreasing referability and discriminability of the objects involved. For example, they may not be situated directly or close to one of the focal axes with respect to a salient relatum, or several competing objects may be situ-

196

Spatial dimensions: Linguistic repertory and issues of application

ated on the same axis. In contrast, directionals can be used with respect to a moving entity’s position in a fairly uncomplicated and straightforward way, with considerably fewer factors to account for. Thus, it seems reasonable to turn to this simpler strategy especially in cases where uncertainty about the interaction partner’s abilities further complicates the situation, as is typically the case in human–robot interaction, at least if the users are not specifically informed about the robot’s features. 5.3.2.2. Figure and Ground How do speakers select roles for referent and relatum in their spatial descriptions? Talmy (2000) proposes a number of answers to this question in his cognitive linguistics framework, using the terms of Figure (corresponding to our referent) and Ground (corresponding to our relatum) for any kind of linguistic description. According to his theory, Figures are typically moving or conceptually movable and of current relevance, while Grounds tend to be larger and more perceivable, and therefore serve as (stationary) reference entities with respect to which the Figure’s position can be described (Talmy 2000: 183). Herrmann and Grabowski (1994) point out that the choice of Figure and Ground is interrelated with ease of comprehension. For example, a typical strategy is to maintain the same Figure throughout a discourse, yielding a simple thematic structure in which the figure (e.g., a person whose spatial surroundings are described) remains the constant topic and usually syntactic subject. Another strategy is to rely on a Given / New structure which repeatedly takes the New of the previous sentence as a Given in the next, as in the following example (Herrmann and Grabowski 1994: 119): (5.72)

Am Ende des Flurs ist ein großes Zimmer. In dem Zimmer steht hinten ein Tisch. Auf dem Tisch befindet sich ein Globus. In dem Globus steckt dort, wo Europa ist, eine Stecknadel. ‘At the end of the hall there is a big room. In the back of the room there is a table. On the table there is a globe. In the globe there is a pin at the place where Europe is.’

Furthermore, speakers adapt their localisation strategies to the requirements of their listener, in certain situations preferring to use the interlocutor’s perspective, and always coordinating their utterances with the listener’s reactions. As a consequence of this fact, speakers consider the interlocutor’s information status in their utterances as well as their presumed con-

Discourse-related factors influencing application

197

ceptualisations of the spatial situation (Herrmann and Grabowski 1994: 117f.). Further evidence, especially with respect to the aspect of contextual saliency, comes from the literature on the choice of landmarks in route descriptions. Landmarks are typically defined as entities (objects or buildings) in the environment which are salient, easily recognisable and identifiable, memorable, clearly distinguished from other entities close by, and so on (e.g., Presson and Montello 1988, Sorrows and Hirtle 1999). They are the entities that are remembered and oriented to in recalling a route, in contrast, for example, to other entities that are not contrastive in a similar way as landmarks are (e.g., Jin et al. 2004). In route navigation, landmarks serve as anchor points for orientation and re-orientation, for example, at decision points, i.e., places where several alternatives for navigation are available, necessitating a navigation decision. They are therefore typically mentioned in route descriptions in order to identify decision points and to enable reference to a future direction of movement, and they are stored in memory and recalled in repeated navigation tasks. 5.3.3. Dialogue processes Over the past decade there have been increasing efforts in investigating the processes active in natural dialogue (much of this may have been inspired initially by Clark 1996). This concerns, for example, ways in which speakers adapt to their interaction partner, partly on the grounds of preconceptions about the interlocutor and partly – dynamically – via negotiation processes (e.g., the interactive alignment effects investigated by Pickering and Garrod 2004). Some relevant aspects were hinted at time and again throughout the present chapter; for instance, in choosing a spatial perspective it may be decisive who the speaker is talking to (cf. Section 5.2.2.1). Furthermore, the discourse history – a history of successful or unsuccessful communication, of feedback and establishment of common ground – may influence all aspects of linguistic choices. These processes also entail the alignment of reference frames to previous usage (Watson et al. 2004, Moratz and Tenbrink 2006, Vorwerg and Tenbrink 2007). Also, for each linguistic choice it is decisive whether the speaker expects their interaction partner to be able to comprehend and interpret the relevant underlying conceptions. Crucially, in a spatial description that is based on a visual perspective, which has been shown to be the case in most usages of

198

Spatial dimensions: Linguistic repertory and issues of application

dimensional expressions, it is necessary to believe that the interaction partner possesses perceptual abilities (similar to those of the speaker in the default case) in the first place. All these aspects fundamentally influence speakers’ actual choices in dialogue and therefore constitute a vast research area in its own right, in which much empirical research has only just started. Thus, although real dialogue will unfortunately not be pursued further in this book,10 it should be kept in mind that the findings and principles outlined here will always be mediated by actual dialogue processes. 5.4.

Conclusion

The present chapter summarised and explored findings on the syntax, semantics, and application (pragmatics) of dimensional terms. As a schematic result, a systematic overview of available reference systems for locative dimensional terms was presented. Furthermore, it was shown how the lexical concepts can be put to use in diverse discourse contexts depending on the spatial setting and in interrelation with the objects to be related linguistically. Throughout the present chapter, a number of issues became apparent that still require further research. Among these, a crucial area concerns the influence of the discourse task, i.e., eliciting spatial reference in tasks that do not, as in most previous empirical work, answer a “Where” question. The empirical research presented in the next chapter will shed light on the following related open questions: x What are the effects of a discourse task involving referential identification rather than localisation? x What kinds of factors are crucial for the employment of locative dimensional terms rather than other kinds of options, such as directionals or topological expressions? x How do speakers choose a reference axis, reference system, and point of view in a situation that allows for different kinds of linguistic representations? What kinds of spatial or discourse-related factors influence such decisions?

10. A range of central insights in the area of spatial dialogue are collected in Coventry et al. (forthc.).

Conclusion

199

x How explicit are speakers with regard to underlying reference systems and perspectives in light of the fact that many different factors come into play in the interpretation of dimensional terms? x How detailed are speakers’ spatial descriptions in a referential identification task? x Under what circumstances does group-based relative reference come into play, and how is it represented linguistically? Further open questions concern, for instance, the kinds of applicability conditions that hold for dynamic spatial dimensional terms (directionals; see, for instance, Tenbrink et al. forthc.). But most crucially, as mentioned in the last subsection, the effects of dialogue processes on speakers’ linguistic choices need to be investigated in detail, which will be a target for diversified research for years to come.

Chapter 6 Spatial dimensions: Empirical analysis

This chapter presents empirical findings that provide answers to a range of research questions identified in Chapter 5. These are subsumed in terms of two major perspectives on the application of locative dimensional terms: First, the “paradigmatic”1 perspective examines when locative dimensional terms are used as opposed to other candidates, for example, directional terms or those denoting distance. Second, the “syntagmatic” perspective addresses how they are used, i.e., what kinds of factors and principles can be identified in the employment of dimensional terms together with underlying reference systems and perspectives. As most research up to now has focussed on localisation tasks, the present research investigates a different major area, namely, referential identification. In this kind of discourse task, the identity of one out of several candidates needs to be determined on the basis of spatial position, as opposed to tasks in which the relative location of one entity is described with respect to another. The approach taken here is predominantly qualitative; its main focus is on investigating speakers’ employments of available options. This results in a complex network of functional and contextual aspects coming into play in identifying objects on the basis of spatial position. A closer look at the specific scenarios reveals whether or not a specific expression is preferred, sometimes used, or entirely disregarded in a given scenario. This approach does not allow for predictions of probabilities of the occurrence of a term; however, the quantitative data presented provide fairly detailed insights with regard to users’ preferences, and they point to systematic variation as well as exceptional and idiosyncratic usages. This chapter is structured as follows. Section 6.1 introduces the data source that serves as the basis for analysis, namely, a web study carried out in English and German. In the second section, the two main perspectives (paradigmatic vs. syntagmatic analyses) are introduced and motivated on the basis of the findings presented in Chapter 5. In Section 6.3, the German

1.

The labels syntagmatic and paradigmatic are used here as cover terms for the two perspectives employed in the present analysis, although some parts of the analysis might not directly match with the usual application of these notions.

The web study

201

data are analysed in detail with respect to both research perspectives, and Section 6.4 deals with the English data, focussing on the purely linguistic aspects. The conceptual aspects and the hypotheses established in the analysis of the German data are, for English, first addressed in Section 6.5, where the results of the two languages are systematically compared. Section 6.6 presents a general discussion of the web study in relation to the relevant findings in the wider research field as presented in Chapter 5. 6.1.

The web study

In order to address a range of specific research questions as formulated at the end of Chapter 5, along with the more basic aim of assessing the linguistic range of variety in settings involving contrastive spatial object reference, linguistic data were collected in a web-based empirical study. A major advantage of this approach is that large amounts of native speaker data can be collected with very little effort. On the other hand, since no clarification questions can be asked, there is a higher potential for misunderstandings. Furthermore, participants can vary external factors such as potential distraction, pauses between tasks, advice by other people, etc., which are controllable only in a setting involving the co-presence of experimenter and participant. They may also answer untruthfully to questions about their person regarding age, gender, language skills, etc., and they may re-start and participate several times. Further evaluation of advantages and limitations of web experiments can be found in Reips (2002). In the present case, the advantages outweigh the disadvantages especially in light of the fact that most of the uncontrollable aspects are not considered to be central influencing factors. Nevertheless, in subsequent studies (directly testing hypotheses ensuing from this predominantly explorative study) a higher degree of control is desirable, which will probably entail the participation of a smaller number of speakers. Another caveat concerns the artificiality of the setting. Participants were asked to type their answers to questions concerning a spatial configuration shown on a computer screen. Obviously, this situation differs from a realworld discourse setting in which spatial descriptions are produced naturally. However, most previous studies dealing with spatial language are based on similar settings involving pictures, either shown on paper or on a screen. This is probably the case because of the need to control the factors involved in the spatial setting very closely, as even small changes in the

202

Spatial dimensions: Empirical analysis

environment or the discourse setting could lead to substantially different results. In the present work, care was taken to gradually lead participants towards the conceptualisation of a natural spatial setting as far as possible, as will be explained in detail below.2 In line with this, the typed answers are here referred to as “utterances” (rather than “sentences”; the linguistic concept of utterance is based on pragmatic rather than grammatical features) and the participants as “speakers”, both of which should be understood in a generic way. For the analysis, conventional methods of corpus linguistics were employed. 6.1.1. Method The web-based empirical study was made accessible at the “Portal for Psychological Experiments on Language”, maintained by Frank Keller, at www.language-experiments.org for two periods of time: the English version was online between September 23rd and December 31st, 2003; after that, the German version was online until October 4, 2004. Speakers of English and German were asked via mailing lists and personal communication to participate; furthermore, spontaneous visitors of the language experiments portal could participate.3 Participation was voluntary and not paid for. In the English version, approximately 200 self-assessed native speakers of English participated. Since also non-native speakers of English were asked to participate, the overall number of participants was much 2.

3.

Within our research group as part of the SFB/TR 8 Spatial Cognition, a series of experiments in spatial real-world interaction between humans and robots has been carried out (see e.g., Moratz and Tenbrink 2006, Fischer 2006), partly using scenarios that correspond to those used schematically in the web study. Also investigations of spatial human-human interaction are being done. Results show that further, dialogue-related processes combine with those worked out for the abstract and monological setting used in the web study. Most responses occurred directly following the mails asking for participation in the mailing lists. The German version was kept accessible for a longer period of time because a number of additional configurations were added after several weeks, so that only spontaneous visitors of the webpage were tested with respect to these new pictures. The amount of data collected is still considerably smaller here than with the pictures that had originally been employed, in spite of the length of time allowed for collection. The exact numbers of participants for each configuration are listed in Appendix 4-7.

The web study

203

higher. In the German version, it was requested that only native speakers of German participate. Approximately 180 speakers participated this time. Age was asked for in a questionnaire but not provided by many participants; those who stated their age were predominantly between 15 and 50 years old, which corresponds to the target of the experimental design. Since there are no hypotheses regarding the impact of age in adult speakers, age effects and distribution are not considered further. Each of the participants answered 14 different randomly assigned questions in randomised order out of a pool of 28 possible tasks which cover a range of different scenarios.4 The decision to limit the number of tasks for each participant was taken in order to minimise the time and effort required for participation.5 Effects of order are not addressed in the present analysis, since they are regarded as a non-trivial additional factor that needs to be treated with specific care. Randomisation guarantees a fairly even distribution of task positions. The participants read an introductory text as follows. Web Study on Spatial Language We are interested in linguistic representations of spatial arrangements. In the following, we present pictures of possible arrangements along with a question that we would like you to answer. There are no wrong or right answers, but please try to avoid using numbers like “first”, “second” and so on. We would like to ask you to work through all the tasks one after the other – without using the ‘back’ button of your browser – and press the ‘submit’ button after answering the last question. Otherwise, your data might be lost. Altogether, there will be 15 easy tasks for you.

The German version of this text was the following: Web-Studie zu räumlicher Sprache Bitte nehmen Sie an dieser Studie nur dann teil, wenn Deutsch Ihre Muttersprache ist. Wir sind an räumlicher Sprache interessiert. Es geht in dieser Studie darum, Fragen zu möglichen Konfigurationen zu beantworten. Es

4.

5.

After finishing the first 14 tasks, there was one final route instruction task which was presented to all participants. It is not analysed here. After the route instruction task, all participants were asked to fill in a questionnaire containing questions concerning their gender, age, home country (in English) or (federal) state (in German), professional interests, etc. This aspect of the study would not be suitable for statistical analysis, which was not targeted here, but see Vorwerg and Tenbrink (2007) for a statistical follow-up study.

204

Spatial dimensions: Empirical analysis

gibt keine falschen oder richtigen Antworten; vermeiden Sie nach Möglichkeit jedoch bitte Nummerierungen wie “das zweite”. Arbeiten Sie bitte eine Aufgabe nach der anderen durch, ohne die “Zurück”-Funktion Ihres Browsers zu verwenden. Insgesamt wird es 15 einfache Aufgaben für Sie geben.

Thus, the participants were asked to avoid counting, in order to avoid collecting a large corpus of linguistic utterances that rely solely on counting rather than locative dimensional terms, which seemed likely at least with respect to some of the pictures. As will be seen in the analysis below, some contributions nevertheless included counting as modifications of spatial terms. Although this option is analysed in the same way as the other kinds of modifications that appeared in the data, it should be kept in mind that this option is probably employed far less frequently than it would have been without this introductory text. The 28 tasks belong to three conditions explained below that differ with respect to the possible perspectives on the configurations. Each participant received 4 questions in each of conditions 1 and 2, and 6 questions in condition 3. Randomisation of task assignment and order applies only inside conditions. In each condition, the same set of basic configurations of elements is shown, as represented in Appendix 2. In the simplest scenario, S (situation) 1, three identical squares are located in a row. This configuration was used in order to enable unproblematic group-based reference, along with other possible linguistic choices. S2 shows the same scene except that the middle element is not a square but a circular element,6 providing a further option for the choice of a relatum, i.e., an element of a different kind. This configuration was used in order to allow for a comparison to S1 in which no relatum of a different kind was present. It also corresponds more to scenarios in other kinds of studies in the literature, in which an object’s position in relation to a different kind of object needs to be examined, which is not possible in S1. S3 presents only two identical elements in a spatial relationship that does not correspond to any of the focal axes in any kind of reference system. Thus, linguistic modifications such as combinations of dimensional terms can be expected here. Furthermore, especially in the conditions that include a viewer, it is hypothesised that the option of a group-based reference system is not available. S4 shows two squares in a 6.

This term is here used by way of contrast to the marking of the goal element by a circular line. The participants naturally used circle to refer to the circular element.

The web study

205

diagonal position to each other, similar to S3; in addition, a circle is situated below one of the squares. Similar to S2, this situation offers opportunities for relating to the “landmark” (an object of a different kind), but the spatial relationships are more complex this time as compared to S2. In S5, three elements are situated in quasi-diagonal but asymmetric positions with respect to each other, producing non-prototypical spatial relations. S6 and S7 were only added to the study at a later time. Therefore, there are no contributions by native English speakers for these situations, and only relatively few contributions by German speakers. The situations were added in order to gain information about the vertical and frontal axes, since most of the other pictures produce a bias towards the employment of the lateral axis for reference. S6 is a 90° rotated version of S1, and S7 is a 90° rotated version of S2. Thus, the 7 basic configurations offer a range of spatial relationships that can be conceptualised and referred to in various ways. Apart from the aspects just mentioned that motivate the choice of the configurations, a basic idea applying to all scenarios is to offer a set of simple but varying pictures in order to explore speakers’ choices. The variability offered by the different configurations is further enhanced by the options of perspective, which are varied by the three conditions as follows. All of the scenes in the different conditions are depicted in Appendix 2, Figure App.1. Outside view. In the first condition, participants were presented with pictures that only contain squares and circular elements. In each picture, one of the elements is marked by a circle. The question to be answered by the participants was simply, Which element of the picture is marked with a circle? // Welches Element in diesem Bild ist durch einen Kreis markiert?

In this condition, the only available view direction is provided by the fact that the participants look at a picture on a screen (outside perspective). Concerning reference frames, one option is to use the picture itself as a relatum, applying the dimensional terms for the internal parts (regions) of the picture. Another option is to use a relative reference system employing either (some of) the other objects as a group relatum, or the circular element (if present) as a landmark relatum. Inside view. In the second condition, an X appears in the picture in addition to the elements. The instruction was:

206

Spatial dimensions: Empirical analysis

Now imagine that you are looking at the figures from the position marked X. How do you describe now which element is marked with a circle? // Stellen Sie sich vor, dass Sie auf die Elemente von der Position aus schauen, die mit einem “X” markiert ist. Wie beschreiben Sie nun, welches Element durch einen Kreis markiert ist?

This condition departs from the previous situation in that the participants were asked to imagine themselves inside the picture. Thus, the origin in their reference frames is supposed to be X directed towards the objects, but the outside perspective is still available. The reference frames available are those known from condition 1 plus intrinsic or relative reference systems specifying the target object’s position relative to X, either from the imagined view direction of X or from the outside perspective. Furthermore, the visual field of X can be used for reference. In some (but not all) of the configurations, the marked goal element is a different one to that in condition 1, so as to increase variability. Partner view. The third condition was designed to simulate a real world setting as much as possible. Here, the position of an interaction partner, Y, is added to the pictures. Additionally, both X and Y are assigned a view direction. For each of the scenarios, there were two possibilities for the positions of X and Y, so that the number of configurations is twice as high in condition 3 as compared to conditions 1 and 2. In one half of the pictures, referred to as condition 3A (C3A), Y (representing the person to be instructed) has the same position as X has in the inside view condition, and X is positioned either at a 90° angle or opposite Y with respect to view direction. In the other half, condition 3B, the positions of X and Y are reversed. In each case, the participants read: Finally, please imagine that the figures are real world objects. You are located at X, and now your task is to instruct person Y to go to the object shows the direction each of you is facing marked with a circle. A star in.// Bitte stellen Sie sich vor, dass diese Elemente Objekte in der realen Welt sind. Sie befinden sich bei Position “X”. Teilen Sie nun einer Person auf Position “Y” (schriftlich) mit, zu welchem Objekt die Person gehen soll. Das Zielobjekt ist wieder durch einen Kreis markiert. Der Stern zeigt jeweils die Richtung an, in die jede/r von Ihnen schaut.

Thus, in this condition view directions are given explicitly, and the participants are asked to imagine a dialogue situation. However, since there is no real interaction and no feedback from the interaction partner, grounding

The web study

207

and alignment processes such as those described in Clark (1996) and Pickering and Garrod (2004) are ruled out, similar to imagined–partner experimentation as reported in Herrmann and Grabowski (1994). The instructional task differs slightly from the previous ones. This situation requires a lot of imagination by the participants; therefore, tasks in condition 3 are presented only after the first two conditions. Now, in addition to the origins and reference frames available in the previous conditions, there is the option of using Y as origin or relatum, or both. Generally, the same configurations were used in both versions of the study (English and German), with two exceptions. The English version did not contain the vertical scenarios S6 and S7 (as mentioned above), and in addition, one scenario, S1(C3A), takes a different goal element (see Appendix 2, Figure App.1). The speakers’ contributions show that there were very few problems in interpreting the instructions, although they were handled in different ways, which is an intended design feature in an open setting such as the one used here. The contributions were annotated using the text markup tool Systemic Coder7 version 4.5, according to the criteria motivated and discussed in the next section. For each situation, the preferred linguistic options were identified and analysed with respect to the research questions formulated in Chapter 5. Furthermore, differences between situations were examined by comparing frequencies of linguistic categories, and explained on the basis of features of the configurations. Averages across scenarios (within or across conditions) are weighted, i.e., depend on the number of contributions in each scenario. Hypotheses generated in this way were tested further by comparing groups of situations sharing critical features, and examining the relevant features of the linguistic contributions. This procedure is regarded as a suitable and necessary preparation for ensuing experiments (using, for example, psycholinguistic methods), testing the hypotheses directly by varying configurations in the relevant ways, and supporting them statistically. The present work does not include this step, but broadly explores the field, predominantly on a qualitative basis, but supported by relative frequencies, in line with established methodology in corpus linguistics. Thus, the data collected in the web-based study are treated as a linguistic corpus rather than conforming to the standards usually established in the field of psycholinguistics. The exploratory nature of this study is specifically ad7.

Available freely at http://www.wagsoft.com/Coder/ (last visited: October 11th, 2006) thanks to Mick O'Donnell.

208

Spatial dimensions: Empirical analysis

vantageous for present purposes because a broad range of factors require exploration: even in the relatively simple scenarios used, there are many ways in which configurations can systematically differ. The analysis will highlight just how these differences influence the linguistic choices used and preferred by the speakers, opening up a wide range of issues that can be addressed purposively in future work. In the next section, I will present and motivate the two major steps taken in the analysis, characterised as paradigmatic and syntagmatic perspectives. 6.2.

Paradigmatic vs. syntagmatic perspectives

6.2.1. Paradigmatic analysis: Coding motivations and categories In Chapter 5, several alternatives to using locative dimensional terms were identified. On the one hand, Section 5.1.1 discusses dimensional terms as opposed to other kinds of locative terms, such as topological, distancerelated, extensional, and path-related expressions. Furthermore, besides static locative spatial expressions, dynamic directional expressions are available in language. On the other hand, speakers may turn to other kinds of strategies instead of using spatial terms at all, as described in the second part of Chapter 5. For instance, in some situations reference to the class name of an object may be sufficient, or an interlocutor may be asked simply to move without specifying a direction. Weiß et al. (1996) furthermore report that speakers, in a car-driving (dynamic) scenario, often spontaneously produce temporal, sequential, and related expressions such as before, past, beyond, next, first, etc., rather than spatial dimensional terms. Thus, situations in which referential identification is asked for do not necessarily yield only goal-based utterances containing locative dimensional expressions, but also allow for other kinds of speaker strategies. In the present experimental setting as described in the previous section, the configurations were designed to enable referential identification by using locative dimensional terms. The objects involved in these configurations exhibit a number of properties: x They can be referred to by their class name. x They do not overlap, touch, or contain each other, thus excluding the application of topological expressions such as in and on. x Differing objects only differ in shape (circle vs. square), not size or extension, thus rendering the application of extensional or size-

Paradigmatic vs. syntagmatic perspectives

209

related terms such as thick or big unsuitable for referential identification because these descriptions are not discriminative. x Objects can be counted. x The distance between the object(s) and the people present can vary, allowing for the usage of distance-related terms such as close or far. x Objects may be situated between other objects. x Since pictures are used, objects can be perceived as being higher than other objects. x No objects are hidden. Additionally, there is a discourse history that allows for anaphoric reference to previously mentioned entities, as in the other one (see also Gorniak and Roy 2004). It is possible for speakers to omit parts of the proposition by relying on the discourse context, i.e., to use syntactically incomplete constructions. Also, speakers may divide the task into smaller steps, in order to reach the goal incrementally. Based on these considerations and motivated from other work in similar settings, the analysis starts from the following general utterance categories applied to all utterances under analysis. First of all, speakers can refer directly to the target object. This category is called “only–goal” in the tables presented in App. 4–7. Reference to the goal may happen either by using a locative dimensional term, or by using a different term. The former case represents the main target of the present research and is analysed in detail in the syntagmatic part of the analysis (cf. Section 6.2.2 below). In the latter case, the immediate competing candidates for linguistic reference can be determined. One major aim of this analysis is to investigate the kinds of factors that motivate speakers to employ other terms in favor of locative dimensional ones. These may be spatial or non-spatial. A range of distinct subcategories are represented in the scheme shown in Appendix 3, Figure App.2. Second, at least in those situations that involve movement (i.e., Condition 3), speakers can specify a path to the goal object and refer to the object as the final boundary of this path. Note that locative dimensional terms may also appear in this category of “path plus goal” utterances. It represents a complex kind of utterance that calls for a different kind of analysis than that carried out for the first category, and thus goes beyond the scope of the present work. Crucially, in this kind of instruction the information needed for referential identification may be contained either in the path description, in the goal object description, or both. Thus, the goal object description does not need to fulfill one central aim of referential identification, namely,

210

Spatial dimensions: Empirical analysis

that it should be sufficient for contrasting the goal object from other objects present. A combined path and goal description solves this problem in a different way: it establishes a focus space within the spatial scenario by directing the listener’s attention into a specific area. Thus, after describing the path towards the target object it might be sufficient to simply refer to “the object” regardless of other objects in the scenario. Third, speakers can simply indicate a direction towards the goal object without mentioning the path’s end boundary at all (“only–path” utterances). Here, directional rather than locative dimensional terms are used. All utterances that involve a goal object at all (i.e., the first two categories) were analysed with respect to mention of non-discriminative features, such as shape or colour (see van der Sluis and Krahmer 2000), because the literature is inconclusive as to the role of such features. There are some indications to the effect that object type and other features are mentioned even if non-discriminative or otherwise redundant. In the present case, the goal object could be specified in more than one way, contrasting it from other object types in the scenario: On the one hand, speakers could specify the shape of the object by providing an informative class name, such as Quadrat (square). This reference choice is informative because it specifies the object’s shape, and it is partly discriminative in a situation where there are not only squares present but also a circle. In one situation, S4(C3A), the class name is even sufficient for discrimination. On the other hand, speakers could use an adjective to describe the object’s shape, for instance, by saying das eckige Element (“the angular element”). If information about the object type was not provided, this could, on the one hand, be due to syntactic omission of the noun, as is the case in an utterance like the German zum rechten (“to the right (object)”); on the other hand, speakers could use a non-informative class name such as element or object. This was particularly likely in the present setting because the instruction to the participants included these superordinate expressions, which could be used for any of the objects in the scenario. Additionally, a number of phenomena emerged as relevant that require further exploration. Sometimes, for example, speakers offered more than one description for a spatial scene. An example from the corpus is: (6.1)

das rechte Quadrat bzw. das am weitesten rechts stehende ‘the right square, i.e., the one situated furthest right’

In such cases, only the first description was analysed, and the fact that an alternative is offered was noted in the annotation. Only if the first descrip-

Paradigmatic vs. syntagmatic perspectives

211

tion did not contain a dimensional term, but the second one did, was the second description analysed instead of the first, because dimensional terms are in the current focus of interest. Descriptions were regarded and annotated as alternatives if they are distinctly marked linguistically, such as by the German terms bzw. (roughly: or), or also (thus), or a slash, a semicolon, or a period. But in some cases, distinguishing between alternatives and complex descriptions (in which several kinds of modification are combined in one single spatial description) may be non-trivial. Consider the following: (6.2) (6.3)

das quadrat rechts von mir, das am weitesten von mir entfernt ist8 ‘the square to my right, which is most distant from me’ Gehe zu dem Objekt, welches sich von dir aus gesehen auf der rechten Seite befindet und von den drei vorhandenen dir am nächsten ist. ‘Go to the object which from your point of view is on the right hand side, and of all three objects present closest to you’

I have decided to code examples such as these as “complex descriptions” rather than alternatives since this is the more conservative assumption. This analysis implies that there is only one underlying spatial conceptualisation which is specified in detail, sometimes yielding a highly complex description which is still coherent with one underlying spatial relationship. If this is not the case, for example, if two distinct relata are mentioned, then the utterance is coded as “alternative”, as in: (6.4)

gehe zu dem objekt, das rechts vor dir liegt, mir am nächsten ‘go to the object which is situated to your right, closest to me’

Here, two different spatial relationships are evoked: one relationship is that between the target element and X, and the other between the target element and Y. The alternative description offers further information that is helpful for the listener and thus part of the speaker’s discourse strategy.

8.

The examples given here are taken from the web study. Many participants in our experiments, when using written language, did not use capitals in their contributions, as required for nouns in German orthography. In the present work, the utterances are directly copied without correcting this aspect. However, obvious misspellings (e.g., hiterster instead of hinterster) are corrected in order to avoid complications. Speakers’ spelling behaviour does not fall within the scope of the present analysis.

212

Spatial dimensions: Empirical analysis

Finally, some participants point out that there is more than one possible view direction, in which case they may offer two alternatives without further syntactic marking, as in: (6.5)

abhaengig von Blickrichtung: Quadrat links Quadrat rechts ‘depending on view direction: square left square right’

These are then also marked as “alternative”, and only the first description is coded (in order to avoid overvaluation of single participants’ contributions with respect to relative frequencies). Some participants offer scene descriptions such as the following, where the speaker makes the conceptualisation of a specific figure explicit. This kind of additional information is annotated as “scene description”, as in: (6.6)

ein quadrat, das mit einem weiteren quadrat und einem kreis als dreieck liegt; dabei ist das markierte quadrat rechts von beiden mittig angeordnet ‘a square which together with a further square and a circle forms a triangle; the marked square is situated to the right and in the middle of the two’

In other cases, the participant provides information about further objects in the scene by explicitly denying that they are the target objects: (6.7)

Gehe zu dem Objekt, das links von Dir steht, nicht hinter Dir. ‘Go to the object which is situated to your left, not behind you.’

Such cases are also coded as containing a scene description since they offer additional information about the scene that is not directly relevant for the spatial relation that is established in order to identify the target object. Instead, they ensure – by way of contrast – that the listener does not interpret the spatial description as referring to the wrong object. It is noted, however, that all utterances provide information about the scene: even a single word like links presumes (lexicogrammatically) that something must be there on the left side (of something). Thus, the decision made here to include a binary distinction between scene descriptions (offering additional information not used to establish contrast) versus no scene descriptions should be regarded as pragmatic; it serves as a way of recognising and acknowledging the fact that some speakers are extremely thorough in their descriptions. In general, the annotation criteria with respect to alternatives and scene descriptions are interpretative coding decisions: they should not be regarded as reflecting clearcut differences in the speakers’ utterances, but rather, as highlighting (generalised) ways in which they may vary, and providing some orientation with respect to the frequency of such

Paradigmatic vs. syntagmatic perspectives

213

speaker strategies. No far-reaching conclusions can be drawn from this kind of annotations, but they nevertheless reflect systematic conceptual processes on the part of the speaker. 9 A further distinction made in the analysis is whether the description fits the goal element in the scenario at all. The standard assumption here is that speakers are cooperative and try to solve the object identification task. Thus, if utterances are underdetermined, a conceptualisation is looked for that matches with the present scenario in a way that the goal element is targeted by the description. However, the data contain some obvious speaker errors. For example, if the perspective is given explicitly, then the goal element is clearly either on the left or the right half plane. This is because the maximum applicability area of dimensional terms is a half plane. Some speakers confuse left and right, or they confuse X and Y with respect to the assignment of speaker and addressee. Therefore, utterances are coded with respect to correctness; only obvious mistakes are coded as “incorrect”. The distinctions made in the paradigmatic analysis are presented schematically in Appendix 3, Figure App.2. 6.2.2. Syntagmatic analysis: Coding motivations and categories While there are a number of studies pointing to the factors influencing speakers’ choices of perspective, the principles behind the selection of reference systems are largely unknown (cf. Chapter 5). Especially with respect to group-based reference, the literature offers only sparse information, because this kind of reference system has been largely neglected so far. However, group-based reference is specifically likely to occur in the area addressed in the present work, namely, referential identification in scenarios containing more than one object of an object class. This assumption is based on the fact that group-based reference depends on the presence of several objects that can serve as a group relatum, which is not often the case in the literature on spatial object localisation (i.e., Where tasks). 9.

Mention of more than one linguistic option is a frequent feature of language produced in open scenarios; this rather inconvenient aspect is treated in different ways in different approaches. I have explained the present procedure in some detail here as I believe that such reactions are neither trivial nor random (and should therefore definitely not be ignored); rather, they reflect underlying systematic features of the situation and its conceptualisation by the speaker.

214

Spatial dimensions: Empirical analysis

However, identifying underlying reference systems is not a straightforward process, in spite of some accounts in the literature that seem to suggest the contrary. I will therefore motivate the present treatment in some detail. Crucially, in contrast to perspectives which can be given explicitly by phrases like from my point of view, there are no direct linguistic indicators (in everyday language) for reference systems. Although there is a great range of variety in syntactic forms, they do not directly reflect the underlying reference systems. In some cases, hypotheses have been identified in Chapter 5 with respect to correspondences of forms and systems: specifically, for example, visual field-based reference may in German be expressed via adverbs, and group-based reference may be expressed by adjectives. However, such assumptions usually rely on speakers’ intuitions rather than thorough investigation on the basis of natural language data. Thus, in the present work the syntactic analysis is carried out independently of the analysis of reference systems. Specific care will be taken to test the assumptions with regard to correspondences between reference systems and linguistic forms, in those cases where reference systems can be identified unambiguously. The underlying reference system depends on the conceptualisation of a view direction together with a relatum. Thus, only if view direction and relatum are made explicit or are inferable from the context can the reference system be determined unambiguously. Remaining ambiguities with respect to the underlying reference systems do not, however, necessarily lead to problems in reference resolution, since more than one reference system may be used to identify the same intended object. Genuine ambiguities are more frequent with regard to the perspective used than with regard to the relatum, since different perspectives regularly (i.e., in face-to-face situations) yield exactly opposite results, while the applicability areas associated with various relata in a scenario often overlap.

1

2

Observer

Figure 6.1. No conflict

Paradigmatic vs. syntagmatic perspectives

215

In order to find out how speakers react if different reference systems yield different results, first of all a situation must be found where this is indeed the case. Given that there is so much overlap between reference systems, especially if assuming large (overlapping) applicability regions, this is not an easy task to achieve. Not providing enough information to identify a reference system in the description is not problematic as long as there is only one potential referent for one axis. Thus, in Figure 6.1, the observer can identify object 2 on the basis of two different descriptions reflecting at least three underlying kinds of reference system. It could be described as being in front on the basis of either an intrinsic reference system (in which case the competing object 1 would also be in front, but more to the left of the observer than object 2) or an internal visual field-based reference system that divides the scene into regions (in which case object 1 would be on the left). In a group-based relative reference system, object 2 would be situated on the right, and object 1 on the left. However, no matter which reference system is used, interpretation will be straightforward: object 1 cannot be referred to by to the right at all, and if in front were used, object 2 would still be the better candidate, being situated more closely to the frontal axis than the competing object. Therefore, it does not matter whether the interlocutors are aware of the availability of a number of possible reference systems; they either use a specific one of these conceptual options, or they are content with an underspecified conceptual representation which suffices for the task. In this case, it seems that the lateral axis is a more suitable choice, since it establishes greater contrast to the competing object than the frontal axis does. Note that potential conflicts in interpretation here do not arise because of conflicting reference systems, but because the two objects are situated within the frontal half-plane with respect to the observer. In a different case, depicted in Figure 6.2, conflicting reference systems can indeed yield the same description for two different objects. Here, object 3 as well as object 1 can – at least in German – be referred to as being hinten (“in back”). Object 3 is hinten in an internal reference system that partitions the visual scene of the observer into regions. Furthermore, it is situated behind object 2, at least if the view direction is shifted in parallel in the direction of the row that is suggested by the positions of objects 2 and 3 (speakers often conceptualise configurations in terms of rows and figures).10 Object 1, on 10. The real possibility of such an interpretation is supported by our empirical work (Moratz and Tenbrink 2006).

216

Spatial dimensions: Empirical analysis

the other hand, is situated behind the observer and can therefore also be referred to as hinten. Thus, if the observer is instructed, for example, by an utterance like “Geh zum hinteren Objekt”11 (Go to the object in back), there is a conflict in interpretation because of different reference systems.

2

Observer

1

3

Figure 6.2. Conflicting reference systems

This example is designed to demonstrate how conflicts between reference systems can yield different interpretations at all, even if the same perspective is used (a conflict of which is the usual, well-researched source for ambiguities in reference systems). Clearly, this is not the usual case; in most cases different reference systems lead to descriptions employing different axes, so that interpretation is unproblematic, as in Figure 6.1. Nevertheless, a number of factors restrict the flexibility of interpretation: some linguistic forms are not compatible with some reference systems, as discussed in Section 5.2.2.4. The choice of spatial axis (as given in the dimensional term itself) sometimes informs about the possible underlying reference systems, depending on the spatial situation. Furthermore, the relatum as well as the underlying perspective both may or may not be mentioned explicitly in the utterance. In some cases, knowledge about the intended target element is necessary to disambiguate ambiguous underdetermined utterances, though underdetermined utterances need not in all cases lead to ambiguity. In case of explicit mention, there is only one possibility; but if the utterance does not contain explicit mention of relatum and / or perspective, several options may be possible. Therefore, in the present work

11. As yet unpublished data from our corpora collected within the SFB/TR 8 Spatial Cognition prove that this instruction is indeed produced spontaneously by some speakers.

Paradigmatic vs. syntagmatic perspectives

217

the utterances will be analysed, as far as is feasible, in terms of compatibility with the usage of relata and perspectives, not in terms of employed reference systems because this cannot be ascertained in many cases. Some information about possible relata can be gained by expansion tests. An utterance like (6.8)

The one on the right

can be expanded to yield any of the options in (6.9): (6.9)

The one on the right in the picture / of you / of me / of the barrel / of the other boxes

Note that the option of using the internal regions of the picture can be made explicit by a prepositional phrase with in, as described in Section 5.2.2.4. Accordingly, utterance (6.8) is highly underdetermined with respect to the relatum, if the scenario is compatible with each of these options. This depends, first of all, on the usage of the preposition in the prepositional phrase in which the dimensional term appears. According to Herskovits (1986), to can only be used with external reference systems; therefore the option of a picture-internal reference system may not be available if to is used: (6.10)

The one to the right of you / of me/ of the barrel / of the other boxes / ?in the picture

It should be noted, however, that the prepositional phrase in the picture may also be interpreted in a different way, namely, not by using the picture-internal regions as relatum but by determining the referential domain, in which case the phrase to the right relates to a different entity which may or may not be mentioned explicitly. For this reason, utterance (6.10) may be acceptable also with the last variant. An utterance like (6.11) is even less flexible. It can be expanded to no more than two options, as (6.12) shows. In this case, expansion is not possible for syntactic reasons if the relatum is expressed as a singular noun introduced by of. (6.11) (6.12)

The rightmost one The rightmost one in the picture / of the boxes / *of you / *of me / *of the barrel

218

Spatial dimensions: Empirical analysis

Figure 6.3. A situation where group-based and picture-internal reference systems are indistinguishable

This observation suggests that the syntactic form of a dimensional superlative does not necessarily reflect a group-based reference system, but it is not compatible with many other options. Note that it is not possible to use a superlative if there is only one object present, since superlatives presume the identity of related participants (Martin 1992). Therefore, in a scenario in which a picture is viewed from the outside, the group-based reference system and the picture’s internal regions are conflated because the picture itself constitutes the referential domain for determining which objects belong in the relevant area of object reference. Thus, any objects that are situated to the left of other objects will always be furthest to the left with respect to the picture’s internal parts. The scenario depicted in Figure 6.3 would be compatible with such a description.

Observer

Figure 6.4. Picture-internal and group-based reference systems differ

As will be seen in the analysis below, it is possible to assume a different point of view from a position within the picture. Then, it may be the case that the group-based and the picture-internal reference systems differ. In the situation depicted in Figure 6.4, utterance (6.11) can still be employed (using the observer’s point of view with a group-based reference system), and it is also possible to state that all three squares are situated on the right (in the picture, using the outside view of the picture instead of the observer’s

Paradigmatic vs. syntagmatic perspectives

219

perspective). But using the dimensional superlative with a prepositional phrase indicating the picture as a reference region sounds strange in this scenario because the picture is associated as a relatum: (6.13)

The rightmost one in the picture

Following these observations, the data were annotated with respect to the features that could be identified by the linguistic utterance itself. Each occurrence of a locative dimensional term was coded in detail with respect to the syntactic form in which it occurred, and with respect to the modifications that appeared along with it. Since the available linguistic options differ between the two languages under analysis, details will be given in the respective sections. The syntactic analysis and the investigation of modifications allow for a specification of speakers’ preferences with regard to the ways in which locative dimensional terms are applied in relation to generalisable spatial situations. Also, the data were analysed with respect to the choice of axis: whether one or more axes were mentioned, and which. The data showed that, in both English and German, dynamic terms (geradeaus and straight ahead) were used in similar situations as the static locative dimensional terms (e.g., vorne / in front), leading to the hypothesis that these terms can be used interchangeably with the static one. To test this, the available data on direct reference to the goal object using the front direction was coded for dynamic versus static reference. These criteria are depicted schematically in Appendix 3, Figure App.3. Furthermore, the mention of relata was annotated. There are a number of different options for relata: the (imagined) interaction partner (depicted in the picture as Y), the speaker (X), the picture (or its center or corners), a “landmark” (an object of a different kind), and the group of objects present in the scenario. If the group is mentioned explicitly as a relatum, there are still a number of different possibilities with respect to the way in which this is done linguistically: a class name can be mentioned that subsumes all objects that together serve as relatum, the group can be described in detail (especially if it consists of objects of different kinds), and a figure can be conceptualised and used in the linguistic description, such as row or triangle (see also Gorniak and Roy 2004 for further examples of grouping). Perspectives were marked as explicit only if the utterance contains a term referring to perception, e.g., to the line of sight, such as (6.14)

das von mir aus gesehen linke ‘the left one from my point of view’

Thus, an utterance like

220 (6.15)

Spatial dimensions: Empirical analysis Das Quadrat links von mir ‘The square to my left’

explicitly marks the relatum, but not the perspective, although in an intrinsic reference system (which is likely to be the underlying reference system in such cases)12 origin and relatum are conflated. Likewise, (6.14) could be said to express the relatum explicitly because the origin conflates with it. However, linguistically there is a difference concerning whether the entity serving as origin and relatum is represented in terms of the relatum or in terms of perspective. After all, whether or not the two are conflated depends on the interpretation of the underlying reference system, which can only be assessed by comparing the utterance with the marked target object in the picture, not by analysing the linguistic form of the utterance itself. As the following example shows, it is possible to mark the perspective explicitly in addition to the relatum even in the case of conflation: (6.16)

Das markierte Element befindet sich in meinem Sichtfeld links von mir. ‘The marked element is situated in my field of vision to my left.’

Therefore, only examples such as these are annotated as both “explicit perspective” and “explicit relatum”. These distinctions are represented schematically in Appendix 3, Figure App.4. The analysis then specifically addressed the question of how far underlying reference systems can be identified. In some cases, the intended relatum can be identified on the basis of (some of) the criteria mentioned so far. In other cases, several reference systems lead to the identification of the same goal object, which may be one reason why there is no explicit linguistic marking of underlying reference systems. 6.3.

German web data analysis

6.3.1. Data extraction Out of the data pool collected by German participants, only contributions by participants were analysed who had completed the study, and who stated 12. In fact, according to Michael Schober (p.c., Oct. 2005), at least for some English native speakers to my right must always refer to the speaker’s intrinsic right side, and thus cannot be interpreted in a relative way at all. This may be different in to the right of me, and in the German version rechts von mir.

German web data analysis

221

that they were older than 15 years. Furthermore, contributions were eliminated that did not correspond to the instruction (less than 4 %), for example, in that they relied solely on counting. Altogether, a total of 2,332 utterances were analysed, 666 of which occurred in condition 1, 672 in condition 2, and 994 in condition 3. The number of contributions for each task differs (see Appendix 4) because of the rotation and random assignment of tasks to each participant within each condition, and because the vertical scenarios were included only later. An analysis was carried through using the coding criteria described above.13 Section 6.3.2 summarizes the results concerning the paradigmatic analysis, highlighting the choice of locative dimensional terms in comparison to other terms in relation to the different configurations and task conditions. Section 6.3.3 summarizes the results of the syntagmatic analysis, showing the details of how locative dimensional terms are employed. The relative frequencies are given in overview tables in Appendices 4 and 5. 6.3.2. Results of the paradigmatic analysis in German 6.3.2.1. Speakers’ discourse strategies: Reference to the goal On a high level of abstraction, the first decision to be made is whether or not to refer directly to the goal object. The contributions differ with respect to the strategies chosen at this point, dependent on the given task. In conditions 1 and 2, speakers consistently refer directly to the goal object. In condition 3, also, direct reference to the goal object (only–goal descriptions) is the preferred option (82.4%). In addition, 11.8% describe a direction of motion and then refer to the goal object (path–plus–goal descriptions), and 5.8% only give a motion direction (only–path descriptions). Examples (all taken from situation S1(C3B)14) are as follows: (6.17) (6.18)

Gehen Sie zu dem Quadrat ganz hinten rechts (only–goal) ‘Go to the square all the way back and to the right’ gehen sie geradeaus an den quadraten rechts von ihnen vorbei. es ist dann das hinterste quadrat (path–plus–goal) ‘go straight ahead past the squares to the right of you. It is then the backmost square’

13. The details of the analysis can be accessed in Tenbrink (2006). 14. All configurations are depicted in Appendix 2.

222 (6.19)

Spatial dimensions: Empirical analysis gehe geradeaus, leicht nach rechts (only–path) ‘go straight ahead, slightly to the right’

From now on, the number of utterances of the only–goal category serves as the new total in relation to which percentages are given. Utterances reflecting other strategies are not dealt with further, since the results do not provide sufficient numbers to warrant detailed analysis. Across all scenarios, the majority of (only–goal) utterances use locative dimensional terms to refer to the goal object. The average of utterances using other kinds of expressions is 14.1%. However, most scenarios yield considerably lower values of non-dimensional terms. The analysis per scenario yields two main candidates for competing with locative dimensional terms in simple spatial scenarios such as those used in the present study: terms expressing in–between relations (e.g., middle), and distance-based expressions. In those scenarios in which the goal object is situated roughly between two other elements, the average of utterances relying on in– between relations is 28.0%, i.e., almost one-third of the cases, as opposed to (almost) zero utterances using this relation in other scenarios. Thus, if an in–between relation is available for reference, this option will be used frequently by speakers. This is the case even in situations where the goal object is not situated directly between two other objects of the same kind; speakers do not necessarily account for digressions from this prototypical configuration linguistically. Similarly, distance-related expressions are used considerably more often in situations exhibiting a clear distance differentiation, for example because the goal object is located distinctly closer to Y than the other objects are as in S2(C3B). However, whether or not they are indeed used for reference apparently interacts with other factors, such as the ease of application of locative dimensional terms. Apart from terms expressing in–between and distance relations, there are only few alternatives to locative dimensional terms in goal-based reference in the present scenario, all of which are only used exceptionally (exact frequencies are given in Appendix 4). The options used by speakers are: simple use of class name, which usually yields underdetermined descriptions; reference to sequential order (if distinct from counting, such as “last”); discourse history; comparative height; compass directions; clock directions; and a number of further individual choices such as “opposite” and “vis–à–vis from me”. Some of these options could have been expected to be used more regularly (such as compass or clock directions) than they did in fact occur in the present data. As it turned out, mostly the linguistic choices by the participants centered around locative dimensional terms,

German web data analysis

223

with a few systematic exceptions in situations in which distance and in– between terms were available for referential identification. 6.3.2.2. Additional and non-discriminative information As described above, speakers sometimes provided further information by offering an alternative description or by describing the scene. The overall numbers of the two kinds of additional information are rather low (as can be seen in the frequencies table in the Appendix). However, in addition to the fact that there seems to be a slight overall increase with increasing scenario complexity, there are some exceptional values. In S1(C3B), there are as many as 17.7% alternative descriptions and 12.9% scene descriptions, and in S3(C3B), there are 18.9% alternative descriptions, but only 1.1% scene descriptions. A typical utterance containing two alternative spatial relationships in S1(C3B) is: (6.20)

Gehe zu dem hintersten Quadrat, von mir aus gesehen ganz links ‘Go to the backmost square, from my point of view furthest left’

It seems that speakers in this scenario could not quite make up their mind as to which of the two possible descriptions would be easiest, since both are fairly straightforward. However, the fact that the goal object may not be perceived directly by the interlocutor may play a role here. Generally, alternative representations may be provided more often in cases when the spatial relationship to the speaker has advantages over that to the listener, whose perspective is usually preferred. These issues provide an interesting area for further research. Finally, the analysis of object features in all utterances containing reference to the goal object reveals that about half of the utterances contain information about object type (for example, by using a suitable distinguishing class name such as “square”), while colour is almost never referred to at all. Both object type and colour are never discriminative, except for S4(C3A) in which the (only) circle was the marked goal. Here, 18.2% of speakers referred only to the class name in their description. 6.3.3. Results of the syntagmatic analysis in German In this section, only the utterances that contain direct reference to the goal object using a dimensional term are analysed, resulting in a new total sum

224

Spatial dimensions: Empirical analysis

of utterances (N) for each task as the basis for percentages. This is the preferred linguistic strategy in all tasks of the web study except for S5(C2)15 (percentage of locative dimensional terms in goal-directed utterances: 50%) where an in–between relation is expressed frequently, and in S4(C3B) where only 30.3% of the goal-based utterances use dimensional terms, while the other utterances contain either in–between or distance expressions. Furthermore, in S1(C3B) only 61.3% of the goal-directed utterances rely on dimensional terms, because distance terms are frequently used. In all other cases, the proportion of goal-based utterances containing locative dimensional terms is above two-thirds. With respect to linguistic variability, the data were analysed in two independent ways: concerning the syntactic forms of the dimensional term, and concerning the ways in which these basic forms are modified. The following syntactic categories were distinguished (see Figure App.3, Appendix 3). The dimensional term may appear as: x an adjective, as in das rechte (Element): adjective x an adverb or post-nominal modifier, which can occur in various syntactic constructions, e.g, rechts, das Quadrat rechts, das Quadrat rechts im bild, and das Quadrat, das rechts von den beiden anderen liegt: adverb16 x a preposition, as in das Element vor mir: prep x a noun in a prepositional phrase, such as das Element zu meiner Rechten: noun–in–pp x a handedness term, such as das rechterhand stehende Quadrat: handedness The following lexical environments may affect the interpretation of a dimensional term: x another dimensional term, as in das hintere linke and oben links: second–dim–term x a term denoting an in–between relation, such as das rechte mittlere and Mitte rechts: middle

15. All configurations are depicted in Appendix 2. 16. The classification of all occurrences of this form as adverb may be theoretically problematic, but seems in practice to be widespread usage.

German web data analysis

225

x counting,17 as in das dritte von links: counting x appearance as a superlative, such as das oberste Element: superlative x a precisifying adverbial, such as das ganz rechts, das Element rechts außen: precisification x a complex description combining several kinds of modification, as in das erste ganz rechts: complex descr. x a term denoting distance, as in das Element, das am weitesten rechts liegt: modification–distance x a deictic term, such as das Viereck da vorne: modification–deictic x a sequential term, such as gehe zu dem letzten rechts: modification– sequential In general, it is assumed as the standard case that all kinds of syntactic forms can be modified in all kinds of ways. 6.3.3.1. Distribution of syntactic forms in German The linguistic category of adjectives contains three different variants. First, the adjective can relate explictly to the goal object, which is then syntactically expressed by a noun, as in das rechte Element. Second, the adjective can relate explicitly to a different entity (referred to as “adj–other” in the Appendix), such as das Quadrat auf der rechten Seite. Third, the adjective can relate implicitly to the goal object, as in das rechte; as mentioned above, the class name does not always appear in the linguistic descriptions. The first and third categories cover almost all occurrences of adjectives in the data. The remaining few utterances usually refer to either the picture’s or a person’s sides. In condition 1, the only influencing factor on the choice of linguistic form is the spatial arrangement of the objects; perspective is not varied because the only available point of view is given by the fact that the participant looks at the picture from the outside. As can be seen from the distribution of frequencies given in the Appendix, the situations vary with respect to the preference of adjectives versus adverbs, but other forms hardly emerge in this condition.

17. This is exceptional because the short general introduction given to the participants before starting with the pictures explicitly stated that counting should be avoided.

226

Spatial dimensions: Empirical analysis

In condition 2, an X is present in the scenario together with the elements. Consequently, a greater degree of variability emerges. For example, this condition is the only one in which handedness terms occur, although only very infrequently. In general, mostly adjectives and adverbs are used as in condition 1. The most obvious digression from this general picture occurs in S5(C2). In this situation, as noted above, participants preferred the expression of an in–between relation (das mittlere Element) to the usage of a dimensional term. It seems that those speakers who did not employ this option resorted to otherwise less preferred syntactic constructions. In condition 3A, a Y appears in the picture, which is situated at the same place where X is situated in each of the pictures in condition 2. X appears at a different position. Thus, the range of possible linguistic and conceptual representations of the situations is again increased. In fact, the distribution of linguistic terms changes considerably as compared to the previous conditions. First of all, the use of adverbs is highly increased, and the employment of adjectives is decreased. This fact seems to be related to the possibility of including explicit information about the relatum, as in the following typical utterance (in S3(C3A)): (6.21)

gehe zum objekt links von dir ‘go to the object left of you’

This information is not easily included in the usage of an adjective, such as: (6.22)

Gehe zum linken Quadrat. ‘Go to the left square.’

Altogether, the relatum is almost never mentioned together with an adjective. Where it does appear explicitly in the utterance, the relatum is usually not Y but the group of objects, as in the following example taken from S5(C3A): (6.23)

gehe zum linken der drei Quadrate ‘go to the left of the three squares’

While the explicit mentioning of relata will be discussed in more detail in Section 6.3.5 below, it is obvious at this point that the restrictions of language with respect to the syntactic inclusion of a relatum considerably influences the choice of linguistic form. In the vertical arrangements S6(C3A) and S7(C3A), the preference of syntactic form shifts from the usage (in conditions 1 and 2) of adjectives to the usage of prepositions. A typical utterance now is the following:

German web data analysis (6.24)

227

gehe zu dem objekt vor dir ‘go to the object in front of you’

In condition 3B, the positions of X and Y are reversed. Since most participants consistently use the interlocutor’s perspective, this seemingly small difference has a major impact on the distribution of linguistic forms. Furthermore, S3(C3B) also differs from S3(C3A) with respect to the goal object: here, the goal object is situated directly behind X with respect to X’s marked view direction. Therefore, typical utterances in S3(C3B) include the preposition hinter, as in: (6.25)

gehe zum objekt hinter mir ‘go to the object behind me’

In sum. The range of variety and the specific choice of syntactic forms are influenced by at least the following aspects of the scenario: x More complex scenarios yield greater variety. x The preference to include a relatum reduces the frequency of adjectives, to which relata cannot be directly added linguistically. x The choice of syntactic form is influenced by the choice of axis, which in turn depends on the discriminativity within the scenario. 6.3.3.2. Modifications and combinations of dimensional terms in German Across all situations, slightly more than half of all locative dimensional terms (53.8%) remain unmodified. This overall result is astonishing considering the great variety of spatial relationships represented in the pictures, many of which do not correspond to a single prototypical axis. Some syntactic variations are easier to modify than others. Specifically, there is no grammaticalised superlative form of the adjective recht– in present day German, although it does occur in some of the utterances as rechtest–. Naturally, such contributions are exceptions. However, with respect to the vertical and frontal axes no similar problems arise since oberst– as well as vorderst– exist as superlatives of ober– and vorder– respectively. This difference is reflected in the distribution of superlatives according to the situations: where the lateral axis is clearly preferred for reference, the average of dimensional superlatives is 1.4%; but where the vertical or frontal axes are predominantly used, the average usage of dimensional superlatives is 19.8%. Also, the overall average of modifications is much higher with the vertical or frontal axes (63.6%, vs. 36.9% with the lateral axes).

228

Spatial dimensions: Empirical analysis

Clearly, the features of the German language here influences usage to a considerable degree. Precisifying modifications appear in two different ways. On the one hand, they are used frequently in order to establish a contrast to other elements that are situated not as far to the right or left as the goal element, such as rechts außen or ganz rechts. On the other hand, they are used to express a prototypical relationship to the focal axis, for example, the fact that the goal element is situated directly in front of Y. In S6(C3A) and S7(C3A), for example, a typical utterance is (6.26)

zum Quadrat direkt vor dir ‘to the square directly in front of you’

In condition 3B, the proportion of modified utterances is generally fairly high. Many of the scenarios are quite heterogeneous with respect to the kinds of modifications employed. A good candidate for explaining these observations is the factor of contrastivity, i.e., whether or not the goal element is the only one on a half plane with respect to Y (as relatum). Of course, other relata can also be used. But as shown in the literature, and in Section 6.3.4 below for the present data, speakers tend to prefer their interlocutor’s perspective. Then, using the interlocutor also as a relatum, resulting in an intrinsic reference system, is an obvious and straightforward strategy for reference. The analysis according to scenarios (across conditions 2 and 3) revealed that, in those situations in which the goal element is the only one on a half plane with respect to the preferred origin,18 the average of modifications is 37.9%. In the remaining situations, the average of modifications of dimensional terms is 70.3%. This clear difference supports the hypothesis that modifications are used to contrast the goal element from other elements that may be situated on the same half plane with respect to the preferred origin. This explains the high number of modifications in condition 3B, where such situations occur specifically often. However, it could also be the case that modifications are used in order to account for the fact that the goal element may not be situated directly in a prototypical direction from the preferred origin if used as relatum, following from the findings on the graded applicability structure of dimensional

18. As the discussion in Section 6.3.5 shows, determining the intended relatum is a highly complex and often inconclusive process. Therefore, the scenarios are examined here with respect to the relative positions of the entities, independent of whether or not the preferred origin was indeed used as a relatum.

German web data analysis

229

terms (Chapter 5). In order to test this hypothesis, situations in which the goal object is situated in a prototypical direction with respect to the preferred origin, were compared to the situations in which this is not the case. In the former case, the average of modifications is 46.2%, whereas in the latter case, it is 45.2%, both of which are about equal to the overall percentage of modifications throughout the data (46.2%). This result contradicts the hypothesis that the relationship between the goal element and the preferred origin comes into play in the usage of modifications. In sum. The following results emerged in the area of modifications of dimensional terms in contrastive reference. x Dimensional terms are typically not modified if the goal element is the only one on a half plane with respect to an obvious relatum (often: the preferred origin), regardless of whether the spatial relationship between the referent and the relatum is prototypical or not. x Modifications of dimensional terms can occur in a number of different ways, depending on the spatial situation. If two axes are equal candidates for a spatial description, two dimensional terms will be combined. If the goal object is at an extreme position on a spatial axis, a superlative will be used. Since this is not possible in German with the lateral axis, precisifying adverbs or unmodified locative terms are then used instead. If the goal object is specifically close to the relatum, a distance or precisifying term may be used in addition to the dimensional term. If the goal object is situated roughly between two other objects, an in–between relation may be expressed in addition to the dimensional term. x If the spatial situation is conceptualised as specifically complex, several kinds of modifications are combined in one utterance, and linguistic representations become fairly heterogeneous. 6.3.4. Choice and explicitness of point of view In condition 1, no utterance contains an explicit reference to the perspective taken. This is not surprising since there are no alternatives to the outside view of the picture. In condition 2, almost all utterances (94.6% on average) remain implicit with respect to perspective. In those few cases where perspective is explicitly mentioned, the perspective taken is that of X, as requested in the in-

230

Spatial dimensions: Empirical analysis

struction. A typical utterance containing an explicit mention of perspective is the following (S1(C2)): (6.27)

das von mir aus gesehen linke ‘the left one from my point of view’

In condition 3, perspective is mentioned explicitly considerably more often, altogether in 23.0% of cases. In almost all cases, the perspective mentioned is that of Y. The distribution among scenarios is not even: The highest proportion of explicit perspective is 34.1% in S5(C3A), while the lowest is zero in S7(C3A). Conceivably, either perspective or relatum are mentioned whenever there is an obvious alternative (e.g., the perspective of X instead of Y). In S7(C3A), for example, the choice between the perspectives of X versus Y does not lead to different interpretations, so that neither relatum nor perspective needs to be expressed in order to avoid ambiguities. But when do speakers express the perspective rather than the relatum? In intrinsic reference systems, both are conceptually conflated, but they may still be expressed in different linguistic terms. A possibility is that the spatial relationship to the relatum plays a role here, since relata can be expressed linguistically more easily than perspectives can: to express a relatum, it suffices to add an expression like von dir (of you) to the spatial description, as in rechts von dir (right of you). To express perspectives, a more complex expression is needed, such as von dir aus gesehen (from your point of view). Therefore, it is possible that expressing perspectives is preferred specifically in situations where the spatial relationship to the relatum cannot straightforwardly be described by a single term. In S5(C3A) (where the highest proportion of explicit perspective occurs) the goal element is not situated on the left axis of Y, so that a description like links von dir may not be regarded as a suitable description. But von dir aus gesehen links (left from your point of view) is less problematic because the goal object is situated to the left of the other objects from Y’s perspective. Note that the underlying reference system is then not intrinsic but relative groupbased, using Y’s perspective. In the situations in which the goal object is situated in a prototypical direction with respect to Y, 10.0% of utterances contain explicit mention of perspective. This is considerably less than the average of all scenarios in the partner condition (23.0%). The difference supports the hypothesis that the relationship of the relatum to the goal element influences the speakers’ choices of whether to refer to the perspective taken. However, since the perspective used sometimes does not affect interpretation at all, clearly

German web data analysis

231

further factors come into play here that cannot easily be differentiated in detail because of the diversity of aspects involved in the configurations and descriptions. If the goal object had not been given in the present tasks, in many cases the intended object could not have been determined by the analyst because of the high number of underdetermined utterances with respect to perspective. Some ambiguities may arise because of indeterminacy concerning the exact view direction of the origin, which is a common problem also in real life situations in which it may not be clear in which precise direction a person is currently looking. Furthermore, speakers do not seem to be aware that an utterance like rechts von dir (right of you) determines the relatum but not the perspective. The following utterance that occurred in S5(C3B) illustrates the alternative possibility of using a relative reference system with the speaker as relatum: (6.28)

Gehe zu dem Objekt, das, wenn Du mich anguckst, links von mir steht. ‘Go to the object which, if you look at me, is located left of me.’

Here, both perspective (that of Y) and relatum (X) are given explicitly. But this is a very rare exception. Many speakers simply say something like (6.29)

Gehen Sie zu dem Quadrat rechts von mir. ‘Go to the square to my right.’

and presuppose that the reference system used is intrinsic, which means that the perspective used is that of X in this case. Note that both utterances refer (correctly) to the same scenario, using different axes. Now, the question arises as to which perspective is actually used, independently of whether or not it is given explicitly. This can be determined (at least in some cases) to some degree by the use of spatial axes. In some cases, several perspectives are equally compatible with the utterance together with the discourse task. For example, in S1(C2) the lateral axis can be used for reference from the outside view of the picture as well as with X’s perspective, because all three elements are lined up in a lateral row in front of X as well as with respect to the picture. But the frontal axis is only compatible with X’s perspective. Thus, all utterances that only rely on the lateral axis (89.5%) are compatible with both perspectives in S1(C2), and those that (also) refer to the frontal axis are only compatible with X’s perspective. However, the choice of axis is not generally a reliable indicator of the underlying perspective. Therefore, in the present analysis, conclusions are only drawn that follow directly from the linguistic contributions, such as explicit mention of perspective. Furthermore, if the two interaction part-

232

Spatial dimensions: Empirical analysis

ners are situated at opposite directions, the choice of the lateral half-axis unambiguously reveals the underlying perspective. This is true because locative dimensional terms can never be used to denote a larger area than a half plane, so that the two opposite terms of the lateral axis (left vs. right) are mutually exclusive and can only relate to one of the two interaction partners. In other words, if X and Y are in a face-to-face situation, everything that is on the left side (i.e., covering the total half plane) of X (and can therefore be referred to by a term denoting the left half-axis) is reliably on the right side of Y. This is not the case if the two partners are situated at other positions, e.g., in a 90° relation with respect to the view direction (as is the case, for example, in S2(C3B)). It is also not the case if the axis used is the frontal one: an element that is situated in front of X may also be situated in front of Y (if the element is situated between the two partners). In sum. Though limited by these considerations, the analysis of perspectives yielded the following general results. x Perspectives are usually not given explicitly. This often leads to ambiguities that may not be detected by either speaker or hearer, which implies a source for misunderstandings. x In case of obvious ambiguities, speakers tend to make relata rather than perspectives explicit (which are conflated in the case of intrinsic reference systems). If the interaction partner cannot be referred to as the relatum in an intrinsic reference system using a simple dimensional term, then the perspective of a relative reference system seems to be given explicitly more often than in other cases. x Usually, the interaction partner’s perspective is used. The readiness to use the speaker’s perspective instead increases slightly if the goal element is situated clearly more closely to the speaker than to the interlocutor. These results lead to the identification of a preferred origin in each condition of the present study. In condition 1, only the outside view of the picture is available. In condition 2, the preferred origin (indicated by the instruction to the participants) is X, and in condition 3 it is Y (even though the instruction gives the view directions of both X and Y). These findings can be related to the notion of ease of reference: In the usual case, perspective is not mentioned because this would complicate the utterance. Relata are easier to include in the utterance, so they are preferred in case of ambiguities, except if making relata explicit would be even more complicated. With regard to choice of perspective, the principle of partner adaptation seems to be more dominant; however, ease of reference at least

German web data analysis

233

has some influence on the perspective taken, as shown by the effects of distance.19 6.3.5. Choice and explicitness of relata This section deals with the ways in which speakers choose relata and refer to them linguistically. The first subsection compares the explicit mentioning of relata between the various conditions. In the second subsection, hypotheses are formulated and tested that account for possible reasons for mentioning the relatum explicitly, in interaction with the possibility of identifying the intended relatum and reference system. In the third subsection, the question is pursued in how far linguistic surface forms serve as indicators for underlying reference systems, based on the kinds of relata that are compatible with the descriptions. 6.3.5.1. Mention of relata across scenarios between conditions In condition 1, the relatum is explicitly mentioned in only 6.7% of the utterances containing locative dimensional terms. In condition 2, the average is 25.1%, and in condition 3 it is 56.1%. Thus, speakers increasingly mention the relatum with increasing complexity of the spatial situation. In condition 1, pictures contain either two or three possible relata. The picture frame and the other square(s) present are always available, and some pictures (S2, S4, and S7 in all conditions) additionally contain a landmark relatum of a different object class (a circle). In condition 2, X is added as a further possible relatum, which also leads to the new possibility of using X’s visual field as relatum, and condition 3 additionally contains Y, with corresponding consequences. The picture as a relatum is hardly mentioned explicitly at all. Also, explicit mention of the landmark relatum (even considering only the cases in which the circle is present) is rare, similar to the mention of group relata across all scenarios. Thus, independent of the presence of the (imagined) persons X and Y, objects of the same kind as well as objects of a different kind are seldom explicitly mentioned as a relatum. Of course, the circle in 19. In real-world scenarios, using the addressee’s perspective involves mental rotation, which is not mirrored here, since the speaker is not actually positioned at X. Therefore, this factor relating to additional effort is here ruled out.

234

Spatial dimensions: Empirical analysis

the present scenario is not a particularly salient landmark, since it has the same colour and approximately the same size as the goal element. Considering the large amount of literature on the employment of salient landmarks as relata in relative reference systems, it can be expected that greater saliency would lead to a considerable increase in landmark mention. In the present scenario, the circle was, so to speak, treated similarly as the group of same-class objects that also served as relatum, but was hardly mentioned explicitly. X serves as an explicit relatum in condition 2 in 20.0% of cases, and in condition 3 in 9.7% of cases. In condition 3, Y is mentioned explicitly in 42.9% of all dimensional utterances. Since Y’s perspective is clearly preferred as shown in Section 6.3.4 above, a frequent underlying concept in the partner condition is thus the intrinsic reference system, using Y as both origin and relatum. From these results it can be concluded that, in the present setting, the imagined persons within the scenario are substantially more often explicitly mentioned as a relatum than other objects present or the picture itself. 6.3.5.2. Choice of relatum and reasons for explicit mention In condition 1, there is no person present which could serve as an explicit relatum. Accordingly, almost none of the utterances make explicit reference to the relatum, which could either be the picture itself or the other objects present. It is not possible to determine which of these two options is used by the speakers (if the conceptual distinction is drawn at all) because in all cases both yield the same linguistic results. In many situations in condition 2, the situation is similar in that most of the available reference systems are often conflated. If the relatum is mentioned explicitly (25.1% on average), it is usually X (20.0% on average). The situation gets considerably more complex in the partner condition where X and Y compete as possible relata. In condition 3, the overall proportion of explicit mention of relata is considerably higher than in the other conditions. However, the distribution of explicit mention of relatum is not even, as Figure 6.5 shows. The lowest proportion of explicit relata occurs in S1(C3A) (22.8%), and the highest in S3(C3A) (80.9%). A reason for this considerable difference could be that, in S1(C3A), two possible relata (group and Y) do not yield any difference in interpretation, i.e., the goal element is on the left side of Y as well as of the other objects. This is not the case in S3(C3A), where it is hard to see how the group could serve as a

German web data analysis

235

relatum at all, since the other element of the same class is situated behind Y (see Figure 6.6), i.e., the two elements are positioned at different sides of the preferred origin. Indeed, in the two situations where this is the case, namely, S3(C2) and S3(C3A), the group is never mentioned as an explicit relatum. However, this is not sufficient evidence because of the low overall number of explicit group relata. 100,0 80,0 60,0 40,0 20,0 0,0 S1(C3A)

S3(C3A)

S5(C3A)

S7(C3A)

S2(C3B)

S4(C3B)

S6(C3B)

Figure 6.5. Explicit mention of relata in the partner condition

Figure 6.6. S3(C3A)

To test whether the hypothesis can be supported that relata are mentioned less often if group relata and the preferred origin as relatum are conflated, situations such as S3(C3A) were analyzed in which Y (or X, in condition 2) and the group as a relatum lead to different results (always assuming the perspective of the preferred origin). Where this is the case, the average of explicit relata is 52.9%. The overall average of explicit relata in the conditions 2 and 3 is 41.7%, which is not strikingly different. However, it is still possible that conflation of group relata and the preferred origin as relatum has a certain amount of influence. A further possibility is that group relata are not often mentioned explicitly if they are underlying the description, even if a confusion with the preferred origin as relatum could lead to misinterpretation. The results of the

236

Spatial dimensions: Empirical analysis

analysis by configuration20 support this hypothesis. It is also supported by the overall low mention of relatum in condition 1, where X and Y are not present. In light of the results, it seems that the group is a likely candidate for a relatum, even though it is seldom mentioned explicitly. Furthermore, the visual field may also be employed implicitly. When several relata within a scenario yield the same result, reference is perceived as specifically unproblematic, as indicated by the low proportion of explicit mention of relata. In other cases, explicit mention of perspective serves to counteract ambiguities; then, the choice of relatum is often irrelevant once the perspective has been determined. In sum. The following general conclusions emerged from the analysis. x Groups of objects and the visual field are often used as relata, although they are seldom mentioned explicitly. x Different relata often either yield similar results, or they lead to the employment of spatial half-axes that can be applied to only one of the objects present, and are therefore discriminative. Therefore, explicit mention is often not needed for referential identification. x Relata are predominantly mentioned explicitly if they are conflated with the origin, i.e., in intrinsic reference systems. Since there is a preferred origin in each condition (see Section 6.3.4), the most frequently explicitly mentioned relata naturally coincide with the identified preferred origin. x On the basis of the present data, because of the high proportion of conflations and implicit references it is not possible to determine conclusively whether there is a configuration-independent preferred relatum in each of the conditions involved. 6.3.5.3. Correlations of linguistic forms and reference systems As observed in Section 5.2.2.1.2, it seems likely that adjectives are preferably used for group-based reference systems because they can only be used if there are other objects of the same class present in a scenario. Furthermore, adverbs have been proposed for usage with visual field-based reference in German, at least with the frontal axis. Therefore the usage of adjectives and adverbs in a number of clear cases of group-based or visual fieldbased versus intrinsic reference systems was investigated. Results show that adjectives are used preferably with group-based reference systems, and 20. The details of the analysis can be accessed in Tenbrink (2006).

German web data analysis

237

adverbs (denoting the frontal axis) are used preferably with visual field based reference. But in both cases, there are exceptions, i.e., a few usages that point to the employment of intrinsic reference systems together with either adjectives or (frontal) adverbs. For example, a handful of utterances used linkes Quadrat (left square) in reference to S3(C3) (Figure 6.6 above). Thus, the correlation of syntactic forms with reference systems should be viewed as a tendency, not a rule. The data collected in the present study are not suitable for more farreaching conclusions in this respect because of the fact that reference systems are overwhelmingly conflated. This, of course, reflects the situation in most real world situations, where it has the positive effect that fewer misunderstandings may arise. It is, in fact, hard to see how reference systems could be disentangled further. 6.3.6. Choice of spatial axis Altogether, speakers seem to prefer relying on one axis only: the overall average of utterances that use only one axis is 83.3%. There is no clear difference between conditions; however, some configurations seem to call for the usage of a second dimensional term more than others do. Therefore, a detailed analysis per configuration was carried out. The findings indicate clearly that speakers prefer the mention of only one axis in most cases, except if one axis is either not discriminative, or if two axes are equal candidates for reference because the goal element is situated between them. Otherwise, digressions from the prototypical axial direction usually do not induce mention of a second axis. These findings are very consistent throughout the data.21

Figure 6.7. S5(C2)

21. The details of the analysis can be accessed in Tenbrink (2006).

238

Spatial dimensions: Empirical analysis

A further interesting aspect is highlighted by a closer look at situation S5(C2) (see Figure 6.7). Here, the frontal axis (employed as the sole axis for reference by 73.2% of utterances) could be employed in two different ways, using either the front or the back direction, reflecting differing reference systems. Interestingly, these could even be combined, as in: (6.30)

Das Element vor mir, hinten ‘The element in front of me, at the back’

Furthermore, as many as 18.3% of speakers referred to the goal object using the dynamic expression geradeaus, as in: (6.31)

das Viereck in der Mitte, geradeaus von mir ‘the square in the middle, straight ahead of me’

This description is not in conflict with a static one of either kind, as the following examples show: (6.32) (6.33)

Das geradeaus vor mir liegende Quadrat. ‘The square that is situated straight ahead in front of me.’ das geradeaus ganz hinten liegende ‘the one that is situated straight ahead all the way back’

The frequent usage of dynamic terms in this specific situation could be an indicator of differences in applicability between the static locative dimensional terms for the front direction, and the dynamic non-dimensional expression geradeaus. This would be in accord with the different results on the size of the front region discussed in Section 5.2.3. In the present case, all three elements are situated in the front half area of the relatum; therefore, the static term might refer to all three elements (though applicable to different degrees). A dynamic term, on the other hand, indicates movement in a front direction that might not encompass the whole half plane, but may imply moving directly forward, i.e., on the prototypical axis. This hypothesis is supported by the fact that the only other situations in which geradeaus is used with some frequency are S6(C3B) (13.3%) and S4(C3B) (7.4%), where, again, all elements are situated on the front half plane, but a movement on the frontal axis would lead directly to the goal element only. Furthermore, in these situations some of the utterances containing a direction instruction together with mention of the goal element (“path–plus– goal”) are expressed in a very similar way, for instance: (6.34)

geh geradeaus zu dem quadrat ‘go straight ahead to the square’

English web data analysis

239

Thus, at this point strategy decision and lexical choice get very close, distinguished only by the syntactic form: a reformulation of the latter example as geh zum Quadrat geradeaus would lead to a goal-based description employing a dynamic term. In sum. The analysis of axes revealed that speakers prefer reference to only one axis if feasible. The region of applicability seems to be as large as a half plane for static dimensional terms; for dynamic terms, usage may be restricted to a much smaller region at least in the case of the frontal axis. 6.3.7. Intermediate summary In this section, results on the German data of the web study were presented, and hypotheses formulated on that basis, summed up in most cases in a concise way at the end of each subsection. A range of principles and preferences were identified with respect to under what circumstances and in what ways German locative dimensional terms are used in a setting allowing for a number of different conceptualisations and linguistic representations. Such preferences concern choices and explicit mention of relata and perspectives, and choices of spatial axes, which together indicate the underlying reference systems, choices of syntactic forms and modifications on the basis of contrast rather than position relative to a focal axis, correlations between syntactic forms and underlying reference systems, and other aspects, which together establish a general picture of how spatial reference using locative dimensional terms is accomplished in German. The specific hypotheses formulated will, in the next section, be addressed with respect to English. Since many of the specific results presented here conform with earlier, often more general results in the literature which are often based on English data, it is likely that there are underlying principles at work that hold for speakers of both languages. Such principles will be formulated following the analysis of the English data, and potential language-specific differences (which should be pursued further) will be pointed out. 6.4.

English web data analysis

This section briefly discusses the procedure of the English web data extraction and the syntactic appearances and modifications of the dimensional terms. The other parts of the analysis will not be presented here in detail;

240

Spatial dimensions: Empirical analysis

instead, Section 6.5 gives a detailed comparison of English and German data, including choice of strategy, relata, perspectives, and the like. 6.4.1. Data extraction As with the German data, the contributions of those participants were considered who completed all tasks and were not younger than 15 years, and data sets containing only the goal object’s class name in each subtask were eliminated as well as contributions relying on counting alone (because participants were explicitly asked not to use this strategy). The overall number of utterances was lower than for the German version. Altogether, a total of 1,480 utterances was analysed, 417 of which occurred in condition 1, 425 in condition 2, and 638 in condition 3. 6.4.2. Results of the paradigmatic analysis in English In conditions 1 and 2, almost all utterances refer directly to the goal object. In condition 3, in contrast, the average of only–goal utterances is comparably low (49.5%). 28.7% use path–plus–goal descriptions, and 21.6% only describe the direction of movement (only–path). Of the only–goal descriptions, in condition 1 the majority (88.9%) contain locative dimensional terms. In condition 2, most situations are similar in this respect, except for S3(C2)22 in which 20.5% of utterances do not rely on dimensional terms, and S5(C2) where 41.5% use other kinds of spatial descriptions. The average of locative dimensional terms in condition 2 is 84.6%. Also in condition 3 there are some situations in which other kinds of descriptions are used more frequently, resulting in some cases in a fairly low number of utterances containing locative dimensional terms. The average here is 74.4%. The exact number of contributions of each category for each task can be examined in Appendix 6.

22. All configurations are depicted in Appendix 2.

English web data analysis

241

6.4.3. Results of the syntagmatic analysis in English In this section, only the utterances that contain direct reference to the goal object using a dimensional term are analysed. This is the preferred linguistic strategy with all goal-based utterances across scenarios, except for the two situations S1(C3B) (where the percentage of goal-based utterances containing a locative dimensional term is 38.5%) and S2(C3B) (41.9%); furthermore, in S5(C2) the percentage is 58.5%, and in S4(C3B) it is 53.6%, indicating that other strategies are equally likely. In all other cases, the proportion of goal-based utterances containing locative dimensional terms is above two-thirds. Paralleling the constructions identified in German, the following syntactic categories are distinguished. In some cases the categorisation may appear arbitrary – as in categorising in front as an adverb(ial group) but in front of as a preposition (group) – however, this approach corresponds to many accounts in the literature where in front of is typically treated along with other spatial prepositions, rather than differentiating further. The dimensional term may appear as: x an adjective, as in the right square or upper square: adjective23 x a post-nominal modifier, an adverb or an adverbial group, as in the rectangle left or the square in front: adverb24 x a preposition (group), as in in front of / above the circle: prep x a noun in a prepositional phrase, such as on the right: noun–in–pp

23. The vertical term lower, in English, is polysemic: as a potential antonym to higher, it may denote comparative height, in which case it would not be regarded as a locative dimensional term at all. However, it may also be interpreted as an antonym to upper, in which case it is a locative dimensional adjective. Since this ambiguity cannot be resolved here (except in cases of explicit contrast, which are rare in the data), an arbitrary decision is taken, and the term is coded here in the latter interpretation, equivalent to the usage of unter- in German. The term occurs 13 times altogether in the data; as this number is fairly low, this decision does not seem to be of much consequence with respect to the general picture. 24. The classification of this category as adverb may be controversial, since in many cases no verb occurs along with the dimensional term. However, paralleling the classification in German this classification is chosen here as a label for reference. Since the present work has its focus on conceptual and semantic issues no theoretic stance is associated with this decision.

242

Spatial dimensions: Empirical analysis

x a handedness term, such as the right hand one or the square on the right hand side: handedness x a stand-alone term, in which case the syntactic form cannot be determined conclusively, as in right: indeterminate–form The basic forms may be modified by: x another dimensional term, as in top left: combination x a term denoting an in–between relation, such as right middle and the center, right square: middle x counting, as in the third square on the right: counting x the appearance as a superlative, such as rightmost: superlative x a precisifying adverbial, such as just to my right or diagonally to the left: precisification x a combination of several kinds of modification, as in left uppermost: complex description x a term denoting distance, as in far right and closest to the right: modification–distance x a sequential term, such as move to the last object on your right: modification–sequential x other: modification–other In the following, the distributions of the basic syntactic forms in each of the three conditions will be briefly considered. The following subsection then examines the modifications. 6.4.3.1. Distribution of syntactic forms in English Especially in conditions 1 and 2, the distribution of syntactic forms shows considerable variability in the English data. With the lateral axis, the most frequent basic syntactic form is a noun in a prepositional phrase. However, with a specific kind of modification adjectives become equally likely, namely, with the dimensional superlative. This construction serves to point out the extreme position of the goal element on the denoted axis in contrast to other elements. With the vertical axis, adjectives are most frequent, and with the frontal axis together with a relatum, preposition groups are the preferred syntactic form. Without an explicit relatum, adverbs become more likely with the frontal axis, as in the square in front. Thus, the different axes, in English, seem to call for different syntactic forms of the dimensional terms, which is an interesting result considering the fact (discussed in Section 2.4) that many studies focus on prepositional

English web data analysis

243

forms only in the analysis of spatial terms. This is not to say, however, that the listed preferences are the only options available. With the lateral axis, besides nouns in prepositional phrases and adjectives, a further option is what is here categorised as handedness terms (which could also be categorised as adjectives, which would obscure their specific nature). But adverbial usage of right / left is also possible, and it does occur exceptionally in the data, for instance, in the form of the square directly right. With the vertical axis, nouns in prepositional phrases are also possible, but much less frequent than with the lateral axis. The frontal axis offers a large variety of syntactic options: ahead, in front, front square, in front of you, etc. Here, preferences seem to depend mostly on the mention of relatum. One question, related to the theoretical observations addressed in Section 5.2.2.1.2, is whether English speakers make use of a short form of explicit mention of perspective, namely, by an expression like to your left of the square. In the data, the strings your left of and your right of do not occur at all. However, the expression go to your leftmost square appears six times in the data, indicating that, at least with superlatives, abbreviated forms do exist. Note, however, that since this construction does not contain an additional explicit mention of relatum (as would be the case in to your left of the square), it could equally well be interpreted in the way that your makes the relatum, not the perspective, explicit. Since both coincide in the case of intrinsic reference systems, this question cannot be resolved conclusively at this point. 6.4.3.2. Modifications and combinations of dimensional terms in English Nearly half (47.9%) of all locative dimensional terms in goal-based utterances are modified. While the distribution is fairly stable between conditions, the distribution within conditions is not homogeneous but depends on the relationship between the goal object and the other candidates. If the goal element is situated more to the left or right than two competing objects, superlatives are used with some frequency. Another frequent method of establishing spatial contrast is the usage of linguistic modifiers such as furthest (to the right) (mod.–distance). The presence of a distance modifier linguistically enforces a contrast to other objects present, allowing, for example, for other elements to be “not–far to the right” (e.g., to the right but not as far as another element). If the goal element is conceived of as situated between two axes, for example using the perspective of X or Y, speakers frequently use combinations of dimensional terms. If the goal element is

244

Spatial dimensions: Empirical analysis

situated on the same half plane as other objects, for example using Y as a relatum in condition 3B, a variety of modifications is used, and the proportion of complex descriptions is increased. 6.5.

Comparison of English and German

One of the effects of carrying out a number of different kinds of studies in a row, and partly in parallel to each other, is to grow increasingly aware of potential influencing factors – and, inevitably, faults in the experimental design with regard to factors that had not been considered as crucial before. In the present case, such an insight following experience concerns the exact wording of the instruction to participants. As recent work by Vorwerg and Weiß (2004) shows, seemingly marginal differences in lexical choices can trigger surprisingly far-reaching differences in participants’ reactions. In their work, the difference was that between “let me out of the car” and “could you park the car”, reflecting different social situations and tasks. That the difference was truly based on the task formulation in the instructions had not been recognised in earlier analyses of the car parking experiment (e.g., Weiß et al. 1996). In the present work, the German version of the web study does not correspond exactly to the English version. This is mainly due to the fact that, during the design process, each version intuitively seemed more natural in the language it appeared in. Since the notion of an exact translation is regarded as unrealistic in any case, the differences originally did not seem to reflect more than simply natural linguistic preferences. Specifically, in the partner condition, the instructions to the participants were not direct translations of each other: in the English version, the participants’ task is “to instruct person Y to go to the object”, while in the German version, the participants read, “Teilen Sie nun einer Person auf Position “Y” (schriftlich) mit, zu welchem Objekt die Person gehen soll.”, i.e., they are asked to tell the person at Y which object to go to. Some of the differing results may be due to this difference in the wording of the instruction. Conceivably, the lexical item instruct (which only occurred in the English version) suggests that the person to be instructed does not know how to get to the indicated goal. Furthermore, the German version explicitly contains the discourse question which, but the English version does not. These observations may be an explanation for the finding that, in the partner condition, the English participants used strategies that resembled route directions towards the goal

Comparison of English and German

245

(i.e., path–plus–goal and only–path descriptions) to a higher degree than the German participants did, who in the majority of cases referred directly to the goal object. This result is interesting in itself as it points to important effects of the exact formulation of instructions to participants; often, in experimental work, this aspect is not accounted for sufficiently. In the present study, speakers of both languages produced utterances of all three kinds of strategies (goal-based, path-based, and both combined). They only differed in frequency, probably as a result of the instruction they read. Further experimentation would be necessary to gain further insights with respect to the reasons why speakers choose one strategy rather than the other. The main result, in the present work, is that all of the identified strategies are available to speakers in instruction situations. In the following sections, I will address each of the areas analysed previously in detail for German, and partly for English, again, this time by way of comparison between the two sets of data. Crucially, I will re-consider the hypotheses formulated on the basis of the German data, in order to examine whether similar observations hold for the English data as well. Since the hypotheses are formulated as tendencies rather than rules, the differences just described in absolute group sizes between the two data sets are not regarded as causing major problems. Tendencies may be detectable to a higher or lesser degree, depending on relative group size. 6.5.1. Comparison of English and German results with respect to the paradigmatic analysis 6.5.1.1. Speakers’ discourse strategies: Reference to the goal The English and German speakers do not differ in their tendency to use solely goal descriptions in conditions 1 and 2. Similarly, in the partner condition, only–goal descriptions are the most frequent option in both languages, followed by path–plus–goal descriptions and then only–path descriptions. Thus, the overall tendency is preserved although the absolute proportions differ considerably. In all three conditions, the distribution of usage of locative dimensional terms in contrast to other kinds of expressions (such as terms expressing in–between and distance relations) is similar in English, as compared to the German findings. Also, the analysis of situation types resulted in the same pattern: in German as well as English, terms expressing in–between relations are used preferably in situations allowing discrimination on this basis,

246

Spatial dimensions: Empirical analysis

and distance-related expressions are used considerably more often in situations exhibiting a clear distance differentiation than in other situations, although other factors (i.e., other available options for reference) also come into play. This latter fact is confirmed by the observation that in some of the situations listed in which reference via distance would be possible, this option was not employed at all. Apart from in–between and distance relations several further options identified in the German data were used sometimes by speakers. The English data do not differ much in this respect. The class name was used as sole basis for reference in S4(C3A), the only situation in which it was discriminative, by 18.1% of speakers in German and 25% in English. This proportion is considerably higher than in all other situations, but still lower than could have been expected, given that spatial information was not needed for referential identification in this special case. 6.5.1.2. Additional and non-discriminative information In condition 1, only very few instances of redundant scene information or alternative descriptions occurred in the English data, but in the subsequent conditions averages increased slightly, mirroring the German data. Also, the analysis by configuration yielded similar results to those discussed in Section 6.3.2.2. In comparison with the German data, the proportion of utterances containing information about object type by using a suitable distinguishing class name such as square is slightly higher, though also constant throughout the conditions (condition 1: 69.5%, condition 2: 62.5%, condition 3: 64.3%; in the German data the overall average is 43.9%). Colour is again only mentioned infrequently. 6.5.2. Comparison of English and German results with respect to the syntagmatic analysis 6.5.2.1. Linguistic variability In conditions 1 and 2, it is quite remarkable how the variety of linguistic forms is much broader in English than it is in German. In the German data, almost exclusively adjectives and adverbs were used, with only few occurrences of prepositions. In English, adverbs were only rarely used in most scenarios; usages varied between adjectives, nouns in prepositional phrases,

Comparison of English and German

247

preposition groups, handedness terms, and indeterminate forms. Part of the increased variability in English is due to the coding decision that, in German, no forms were coded as indeterminate since German adverbs (rechts, links) differ in form from adjectives (recht–, link–), which is not the case in English (right, left in both cases).25 Furthermore, handedness terms and nouns in prepositional phrases are fairly exceptional in German, though they are possible, as shown by occasional usages such as rechterhand or zu meiner Rechten. The main difference identified by the comparison of the linguistic variability is a much stronger preference for nouns in prepositional phrases in English throughout the data, together with a comparably stronger preference for handedness terms. The frequency of nouns in prepositional phrases through all conditions in English is 42.7%, while it is only 1.3% in German. With handedness terms, the difference is less clear; overall, the frequency is 3.8%, while in German, it is 0.2%. However, in some situations this option is used to a considerable degree in English (e.g., 11.8% in S2(C1)), which is not mirrored in German. In condition 3 the variability of English choices is diminished somewhat. Here, the German and English versions are less differentiated with respect to range of variety. Now, in some situations speakers settle almost unambiguously for nouns in prepositional phrases in English (and in some situations preposition groups, depending on axis), neglecting other options. German speakers then often use adverbs and in some cases adjectives or prepositions. 6.5.2.2. Modifications and combinations of dimensional terms In German, lateral adjectives like recht– are not used as a superlative; therefore, dimensional superlatives in German are restricted to the frontal and vertical axes. In English, in contrast, dimensional superlatives are straightforwardly used with the lateral axis (as in leftmost square). Since most configurations enhance contrast on the lateral axis, the overall propor25. Alternatively, stand-alone terms could have been coded as such in German as well, parallel to the English indeterminate forms. However, this would have triggered further problems because almost all utterances are syntactically incomplete, so that the annotator would then need to decide which exact requirements were necessary to define a term as stand-alone. In English, no such problem occurred because all forms that could either be adjectives or adverbs (if the utterance was expanded to a complete sentence) were coded as indeterminate.

248

Spatial dimensions: Empirical analysis

tion of superlatives is therefore higher in English (11.1% on average) than it is in German (5.7%; where most occurrences belong to the ‘vertical’ scenarios which were not used in the English part of the study). But not all scenarios allow for the usage of superlatives at all, since the application seems to require that the goal element is situated at an extreme position on an axis relative to at least two other elements, which is not the case in all scenarios. This scenario-related fact accounts for the relatively low overall proportion of superlatives in English even though the language itself does not restrict application on the lateral axis, the way it does in German. Also, modifiers of distance are seldom used in German except in two situations (S2(C3B): 11.1%; S4(C3A): 8.3%) in which the goal element is very close to the interaction partner. In all other situations, the proportion of distance modifiers is never higher than 7.1%. In English, in contrast, distance modifiers are often used to denote an extreme position on an axis, similarly to superlatives, in some scenarios in as many as 35.3% of utterances (S5(C3A)). But since distance modifiers are not in all relevant cases used to a similar degree as superlatives, there must be subtle differences in applicability between the two choices, which cannot be resolved here conclusively on the basis of the data. In the German data, the overall frequency of precisifying modifiers throughout all scenarios is 15.1%, while in English, it is 6.4%. While this overall difference is not large, the distributions of precisifiers are very different. In the English data, in condition 1 no precisifiers occur at all, while in German, specifically in S1(C1) and S2(C1) they are fairly frequent. In condition 2, in English they predominantly occur in S5(C2), while in German, they are used fairly frequently in all situations except S3(C2). Thus, clearly English precisifiers are used in a different way than are German ones. A closer look at the data reveals how. English precisifiers are typically expressions such as directly which emphasise a typical relationship on a focal axis, and in some cases expressions that point to a digression from that axis, such as a little (to the right) and diagonally (to the left). In German, similar expressions for the two variants are direkt and schräg links. But in addition, the German data show frequent usage of precisifiers that emphasise an extreme position on a focal axis, such as ganz (rechts) or rechts (außen). This category seems to be missing entirely in the English data. One candidate for a corresponding expression in English is all the way on the left, which occurs once in a path–plus–goal description. Otherwise, such relationships are in English expressed not by precisifying ad-

Comparison of English and German

249

verbs, but by dimensional superlatives, distance modifiers, and unmodified expressions that (depending on the scenario) may serve the same purpose. Thus, the general picture that emerges from these observations is that there are a number of language-specific phenomena that account for different distributions of modifiers throughout the scenarios. Most importantly, while English speakers use distance modifiers and dimensional superlatives for the specification of an extreme position on an axis (specifically with lateral directions, which are in focus in the present scenarios), German employs precisifying adverbs for this purpose. Other observations based on the German data with respect to the influence of the configuration on the choice of modifiers were largely reflected in the English data: Combinations of dimensional terms are used preferably in scenarios where two axes are equally good candidates for description. In–between relations are indicated specifically in situations where the goal object is situated roughly (but not directly) between two other objects. In the German data, the hypothesis was formulated that modifications occur to a higher degree if the goal element is not the only one on a half plane with respect to the preferred origin (X in condition 2, and Y in condition 3). In such situations, the average of modifications in German is 70.3%. In the corresponding English scenarios, the average of modifications is 78.6%, as opposed to 40.2% in the other scenarios (37.9% in the German data). Thus, the same kind of difference is reflected in the English data, supporting the hypothesis that modifications are very often used to contrast the goal element from other elements situated on the same half plane with respect to the preferred origin. A contrasting hypothesis was also tested and rejected for the German data, namely, that modifications are predominantly used in order to express digressions from the prototypical axis with respect to a relatum. This result is also mirrored by the English data; as in German, the difference between the respective kinds of scenarios is negligible. Thus, the findings identified in the literature for other discourse tasks, namely, that modifications increase with increasing distance from the focal axis, are not confirmed for the present discourse task, neither for German nor for English. In sum. The following conclusions can be drawn concerning modifications of dimensional terms in German vs. English in contrastive reference. x In both languages, dimensional terms are preferentially not modified if the goal element is the only one on a half plane with respect to an obvious relatum (often: the preferred origin), regardless of whether

250

Spatial dimensions: Empirical analysis

the spatial relationship between the referent and the relatum is prototypical or not. x Modifications of dimensional terms can occur in a number of different ways, depending on the spatial situation as well as the languagespecific repertory. If two axes are equally good candidates for a spatial description, two dimensional terms will be combined more often than otherwise. If the goal object is at the most extreme position on the lateral axis, in English either superlatives or distance modifiers will be used more frequently than otherwise. In German, precisifying adverbs are used instead in such situations, which are not used in this way in English. If the goal object is specifically close to the relatum, in both languages a distance or precisifying term may be used in addition to the dimensional term. If the goal object is situated roughly between two other objects, an in–between relation may be expressed in addition to the dimensional term. If the goal object is situated directly on a focal axis, a precisifying modifier will frequently be used in both languages to emphasise this relationship. x If the spatial situation is conceptualised as specifically complex, several kinds of modifications are combined in one utterance, and linguistic representations get to be fairly heterogeneous. 6.5.2.3. Choice and explicitness of point of view In condition 1 and 2, the English findings mirror those of German in that perspective is overwhelmingly not mentioned; if it is, the perspective given is that of X in condition 2, and that of the picture in condition 1 (this is only specified by one particular participant). However, the results in condition 3 in English are markedly different from those in German. In German, 23.0% of the utterances were explicit with respect to perspective; in English, in contrast, the average is as low as 1.7%, resembling the usage in the other conditions. This is a surprising result because, in the partner condition, there is more than one perspective available (that of X or Y). In the few cases where perspective is mentioned in the English data, it is always that of Y (the imagined interaction partner). But even in the situation S5(C3A), where as many as 34.1% of utterances in German contain an explicit mention of perspective, all of the English descriptions are implicit. Interestingly, those participants who do mention perspective never do so by an expression that would mirror the German usage. In German, a typical

Comparison of English and German

251

way of mentioning perspective is an expression like von mir/dir aus gesehen. It could be expected (on the basis of literature dealing with perspective, see Chapter 5) that English speakers use from my/your point of view in similar ways, but this is not the case in the present data. This expression does not occur at all throughout the corpus. Instead, some few utterances contain more complex references to face orientation, such as: (6.35) (6.36)

Go to the circle just to your right as you’re facing me. Assuming I can see Mr. Y’s forward facing position, the answer is...”go to the square directly in front of you”.

The analysis of which perspective is actually used, independently of whether or not it is given explicitly, was again carried out on the basis of those situations in which the underlying perspective could clearly be identified.26 Results showed a very clear preference for using the interaction partner as origin wherever possible. Since the relatum also was overwhelmingly the interaction partner in these cases, English speakers clearly prefer intrinsic reference systems, more consistently so than German speakers do. This clear result supports the hypothesis that expressions like to your right predominantly – or perhaps even strictly – refer to the speaker’s intrinsic side, rather than allowing for a relative interpretation. It also indicates that participants succeeded in imagining to be in a real-world situation as depicted in an abstract and schematic way. The observation made with respect to the German data, that speakers often do not seem to be aware of possible misunderstandings based on a lack of specification of perspective, also holds true for English. This is true, on the one hand, by the fact that perspective is seldom mentioned, and on the other hand by the fact that reference systems other than intrinsic ones (where the perspective is inferable if the relatum is given) are always available. Also English speakers in some rare cases make use of a relative reference system using the interlocutor as origin, as the following exceptional example (taken from S5(C3B)) shows: (6.37)

I’m standing just to the right of it.

Remarkably, in this example Figure and Ground are exchanged: here, the speaker serves as referent, and the goal object is the relatum, although it is the goal object’s position that is to be explained. Nevertheless, this example

26. The details of the analysis can be accessed in Tenbrink (2006).

252

Spatial dimensions: Empirical analysis

shows that relative reference systems are also available to English speakers in this scenario, although clearly dispreferred. In sum. The following conclusions follow from the analysis with respect to use and mention of perspective in the present setting. x In both languages, perspectives are usually not given explicitly. This may lead to ambiguities that remain undetected by either speaker or hearer, which implies a source for misunderstandings. x In case of obvious ambiguities, speakers tend to make relata rather than perspectives explicit (which are conflated in the case of intrinsic reference systems). This seems to be the case more consistently in English than in German. Since English speakers seem to associate intrinsic reference systems with expressions like to your right, the utterance is then no longer ambiguous. x In German, if the interaction partner cannot be referred to as the relatum in an intrinsic reference system using a simple dimensional term, then the perspective of a relative reference system is given explicitly more often than in other cases. In English, no such tendency could be detected. x Usually, the interaction partner’s perspective is used. With German speakers, the readiness to use the speaker’s perspective instead increases slightly if the goal element is situated clearly more closely to the speaker than to the interlocutor. With English speakers, no such tendency could be detected. These results confirm the identification of a preferred origin in each condition also for English. In condition 1, only the outside view of the picture is available. In condition 2, the preferred origin is X, and in the partner condition it is Y. 6.5.2.4. Choice and explicitness of relata For the German data, it was found that speakers increasingly mention the relatum with increasing complexity of the spatial situation. This result can be confirmed for the English data: the frequency of explicit relata in condition 1 is 14.1%, 33.6% in condition 2, and 88.7% in condition 3. Furthermore, as in German, the picture, the group, and the circle are all only seldom mentioned explicitly as relatum. X is explicitly mentioned as a relatum in 29.1% of cases (on average) in condition 2. In condition 3, X appears as an explicit relatum in only three situations in English, namely, S1(C3B) (30%), S3(C3B) (70.3%), and

Comparison of English and German

253

S5(C3B) (4.8%). In the German data, tendencies were similar; more than half of the German utterances explicitly refer to X in S3(C3B) although this is not often done otherwise. In this specific situation, the goal object is situated directly behind X, and is the only element on the back half plane, but there is no simple spatial relationship to Y. Thus, the exceptionally high degree of usage of X as relatum is easily explained on the basis of ease of reference. In the other situations where X is used as relatum in the partner condition in German and English, the goal object is usually also closer to X than to Y. Y only occurs in the partner condition, where it is mentioned explicitly as relatum in 74.0% of cases on average. Thus, in the case of the English data there is a clear preference for employing – and explicitly mentioning – the interaction partner, if available, as a relatum. In German, in contrast, less than half of the utterances in the partner condition (42.9%) contained explicit mention of Y as relatum. However, in general, in the English as well as the German data the imagined persons present are mentioned substantially more often as relata than any of the other options are. In German, the distribution of utterances containing explicit mention of a relatum was quite heterogeneous (cf. Figure 6.5, Section 6.3.5.2). There, relata were mentioned more often in situations in which ambiguities could arise otherwise. This result is not mirrored in the English data. As Figure 6.8 shows, the distribution of explicit mention of relata is fairly homogeneous, ranging between 73.1% (in S2(C3A)) and 100% (in S1(C3A)). Thus, the correlations found in the German data between configurations and explicit mention of relata apparently do not hold for English. 100,0 80,0 60,0 40,0 20,0 0,0

S1(C3A)

S3(C3A)

S5(C3A)

S2(C3B)

S4(C3B)

Figure 6.8. Explicit mention of relata in the partner condition in English

However, a different point identified in the German data still needs to be examined, namely, the likelihood of whether a group or visual field relatum underlies an implicit linguistic representation. The analysis of a number of situations in which group or visual field relatum and the preferred origin as

254

Spatial dimensions: Empirical analysis

relatum lead to different results, in German, led to the conclusion that the group is a likely candidate for reference in spite of the fact that it is seldom mentioned explicitly. The analysis of the relevant situations for the English data yields a different result. Although the group or visual field reference systems seem to be available to English speakers just as in German, English speakers use them to a far lesser degree than German speakers do (about one-third in German vs. just two exceptional cases – 4.2% in one situation27 and none in another – in English). This result may account for the constant high proportion of explicit mention of the interaction partner as relatum. In German, only the interaction partner was mentioned explicitly, but not the group or visual field; thus, if the group or visual field was underlying the description, then the description remained implicit, which does not happen in this way in the English data. In sum. The following general conclusions can be drawn. x In German, groups of objects and the visual field are often used as relata, although they are seldom mentioned explicitly. In English, this option is available but seldom employed. x Different relata often either yield similar results, or they lead to the employment of spatial half-axes that can be applied to only one of the objects present, and are therefore discriminative. Therefore, explicit mention is often not needed for contrastive reference. x In the partner condition, relata are predominantly mentioned explicitly if they are conflated with the origin, i.e., in intrinsic reference systems. In English, this is fairly consistently the case, and therefore the relatum is consistently mentioned explicitly in the majority of cases. Where there is only one (imagined) person, or none, in the scenario (i.e., in conditions 1 and 2), relata are mentioned less often. x Because of the high proportion of conflations and implicit references it is not possible to determine conclusively whether there is a configuration-independent preferred relatum in each of the conditions involved. However, the English data point to a tendency towards using a person (either the only one present or the interaction partner) as relatum, since group relata and the visual field are less preferred.

27. The two expressions used were back and centre and the farther back square, both referring to the goal object in S5(C2).

Comparison of English and German

255

6.5.2.5. Choice of spatial axis In German, speakers seem to prefer relying on one axis only in the present setting (83.3% throughout conditions). This result is similar in English, with an overall average of 84.3%. Also considering specific situations, the distribution of choices of axes is fairly similar between the languages, including the employment of both static and dynamic terms for the front direction in S5(C2). Just like the German participants, English speakers also only use dynamic terms in situations where a movement on the frontal axis would lead directly to the goal object. Also, mention of a second axis becomes more likely if the goal element lies on a diagonal with respect to the relatum used. The presence of a second possible relatum does not influence this tendency. Altogether, in both languages, speakers prefer the mention of only one axis except if one axis is either not discriminative, or if two axes are equal candidates for reference. Otherwise, digressions from the prototypical axial direction usually do not induce mention of a second axis. 6.5.3. Summary of the comparison between the English and German data In the present section, the results of the English part of the web study were compared, as far as possible, to the results presented and discussed in detail on the basis of the German data in Section 6.2. Some of the hypotheses formulated there could be confirmed, and some differences could be identified. Many of the differences relate to the structure of the languages involved, as English speakers have a different repertory of forms at their disposal than do German speakers. But they have also different preferences for using and applying this repertory. Notably, for example, nouns in prepositional phrases are a very common way of applying dimensional terms in English, which is not the case in German, although the form does exist. In German, adjectives are clearly preferred; these are also available in English but much less frequently employed. Furthermore, some differences arise with respect to which forms can be used with which kind of axis. On the basis of the present findings, it is now possible to present a more comprehensive table of syntactic forms of English and German locative dimensional terms (Table 6.1 below). Here, choices that were identified in the present study as specifically frequent in this specific discourse task are set in italics (only lateral axis, as the other axes were not in focus). With respect to the prepositions used in the construction here called “nouns in prepositional phrases”, in German, only zu (meiner Rechten) is

256

Spatial dimensions: Empirical analysis

possible, while in English, there is a broad and flexible repertory containing at least to, on, at, in, towards, near, and from, of which the first two are the most frequent in the present scenarios.28 In English, this kind of construction is preferred in many situations; the only constraint is that to is not used for internal relationships, which occur most frequently in condition 1 if the picture is used as relatum. Table 6.1. Overview of syntactic forms in English and German

English

German

Syntactic form adjective

lateral

frontal

vertical lateral

frontal

right left

front back

vorder– ober– hinter– unter–

noun in prep. phr.: [prep] + [det / pronoun] + handedness term

right left

front back

upper / lower bottom / top bottom top

adverb(ial group); postnominal modifier preposition (group) indeterminate form

(to the) right–/ lefthand (side) right in front/ left back (straight) ahead / behind in front/ back of right / front / left back

recht– link–

vertical

Rechten Linken rechter/ linkerhand

above below

above/ below above/ below bottom / top

rechts links

vorne oben hinten unten geradeaus vor / über/ hinter unter

Furthermore, modifications are distributed in different ways in the two languages. For instance, English speakers use distance modifiers and super28. The details of the analysis can be accessed in Tenbrink (2006).

General Discussion

257

latives to indicate an extreme position on a lateral axis, while in German, precisifying adverbs are used in such situations, partly because, in German, the lateral dimensional terms cannot be realised as superlatives. Apart from these differences that can be regarded as predominantly linguistic, there are also some differences with respect to conceptual choices. German speakers seem to use group-based or visual field-based reference to a high degree (without making this choice explicit on the linguistic surface). Here, a preference for using adjectives for group-based reference, and (frontal) adverbs for visual field-based reference could be identified. English speakers, in contrast, prefer reference via a human relatum, on the basis of an intrinsic reference system. This is specifically obvious in condition 3, where the utterances overwhelmingly contain explicit reference to the interaction partner as relatum. This result is in accordance with previous results discussed in Chapter 5, e.g., by Carroll (1993; 1997). In both languages, speakers typically do not mention the underlying perspective explicitly; however, in certain situations German speakers are more inclined to do so, which is not the case for English speakers. With respect to the underlying principles and preferences for applying locative dimensional terms, however, some tendencies could be identified that hold for both languages alike, abstracting from differences that are due to the structure of these generally very similar languages. These are summarised in the next section. 6.6.

General Discussion

In this chapter, I have presented the results of a web study designed to trigger locative dimensional terms, in order to gain insights about the ways in which they are applied in natural discourse. This method was considered suitable because of the immense context dependency of spatial terms: extracting these expressions from naturally occurring texts, as was done for the temporal dimensional terms, is not feasible in this case, because typically the textual context does not contain sufficient information about the spatial setting in which the terms are applied, rendering underlying principles much more difficult or even impossible to identify. In the following, I will summarise the results systematically and discuss them in the light of earlier findings.

258

Spatial dimensions: Empirical analysis

6.6.1. Determining reference systems In the present analysis, the data were examined with respect to choice of strategy, alternatives to locative dimensional terms in goal based instructions, and details about the application of locative dimensional terms, including linguistic variability and conceptual choices such as perspective, relatum, and axes, which together serve as the basis for a reference system. However, the reference system itself cannot be identified directly on the basis of the linguistic form, at least not for those systems that were employed spontaneously by speakers in the present scenario.29 This is due, on the one hand, to the overwhelming lack of explicitness with respect to either relatum or perspective, or both, and on the other hand, to the fact that there is no one-to-one correspondence between linguistic forms and underlying reference systems. The latter idea was motivated in detail in Section 6.2.2, and further discussed on the basis of empirical results, for example, in Section 6.3.5.3. Although the thorough survey of the literature presented in Chapter 5 already points to this conclusion, it has not been recognised widely by researchers, and the consequences of the lack of identifiability of reference systems have not sufficiently been accounted for. Research has mainly focused on the distinctions of perspectives, largely ignoring the diversity caused by the possible underlying relata. If both elements are mentioned explicitly, the reference system can be identified. But as the data show, perspectives are almost never (specifically in English) given explicitly, and relata are mentioned only in a subset of cases, preferably to avoid obvious misunderstandings. Therefore, the present results point to the conclusion that in the majority of situations there is more than one underlying reference system that is compatible with the linguistic representation in relation to the situation at hand. This idea corresponds to earlier findings in the literature. Carlson (1999), for instance, explicitly addresses the phenomenon with respect to intrinsic versus absolute usages of above. Her results indicate that, with regard to apprehension, several reference systems are activated simultaneously, even though (on the vertical axis) an absolute interpretation is favoured. The various applicability regions are added in a merged spatial template. Where reference systems coincide, apprehension is faster.

29. For absolute reference systems, the case would be different, see Section 5.2.2.1.

General Discussion

259

This finding can be interpreted in the way that, in apprehending spatial dimensional terms, humans do not necessarily decide consciously about the underlying reference system, but are content with identifying an actual match between description and situation if possible – regardless of which reference system may be the basis for this match. Such a hypothesis seems plausible especially in light of the fact that, as shown by the systematic account in Chapter 5, the actual range of options for reference systems is much larger than has been assumed in the relevant previous studies. Thus, it seems to be not unlikely that, where reference systems coincide, speakers do not decide consciously which one is used, neither in production nor in comprehension. This would parallel a situation in which two people are positioned beside each other, facing in the same direction: in this case, the spatial regions of the reference systems using either person as origin would overlap to such a high degree that it is unnecessary to specify whose perspective is taken. An utterance based on these reference systems would therefore certainly be underspecified with respect to perspective, not because one specific perspective is implied but because perspective does not need to be determined. Thus, such an utterance could be interpreted in the way that both possible perspectives are implied because they almost coincide. However, speakers do not need to be consciously aware of this fact, since there is no need for discrimination. This observation can be related to recent research on mental models. There has been some discussion on how far mental representations are specified if they are constructed on the basis of indeterminate information. For example, in “A is higher than B” and “A is higher than C” the relationship between B and C is not clear. Vandierendonck et al. (2004) provide evidence for the theory of “annotated mental models”, which is based on the integration of the available information into one single model. Indeterminate information is integrated by ‘annotating’ the model at the relevant places, indicating the necessity of further elaboration when needed. The annotation itself remains implicit; the possibilities are not spelled out. This strategy places less of a load on working memory than a strategy in which all alternatives need to be kept in mind at the same time. Another possibility is to start with one option and change it if it becomes inconsistent with new information. Although this line of research deals with mental models constructed on the basis of text in order to understand relationships (for example, spatial relations) that are not represented pictorially, it seems to parallel the present findings fairly well. The difference is that, in the present case, a spatial scenario needs to be represented verbally via the

260

Spatial dimensions: Empirical analysis

speaker’s conceptualisation of the scene, rather than vice versa. In producing an utterance containing reference to one of the objects in the scene, the speaker needs to enable the listener to identify the intended referent on the basis of the description. Therefore, the description should be just as determinate as necessary. In the data, a consistent tendency in just this direction could be identified, although there are a number of remaining ambiguities which the speakers were probably not aware of. 6.6.2. Principles of application of locative dimensional terms Generally, the web study participants used a broad spectrum of variability on all scales. The analysis shows that linguistic as well as conceptual choices depend heavily on the spatial situation, i.e., the presence of other objects and (imagined) persons, and the available kinds of perspective. Therefore, generalised predictions are difficult to formulate on a linguistic surface level. These results are further complicated by a number of language-specific differences, as summarised in Section 6.5.3. In spite of the high variability, however, regular patterns of usage can be identified that can be analysed in relation to the hypotheses posed in Chapter 5. The systematic variations found are repeatedly explained by much the same idea, reflecting the participants’ motivation to fulfill the task of providing an unambiguous – contrastive – description of the target object in a given situation. Three major principles, at least, familiar from previous findings in the literature (cf. Chapter 5), seem to be at work, influencing speakers’ choices: The principle of contrastivity ensures that the goal object can be identified among the competing objects. The principle of minimal effort leads, on the one hand, to the omission of information that is redundant or easily inferable, and on the other hand, to linguistic and conceptual choices (including dimensional terms versus other kinds of spatial expressions, relata, and spatial axes) that enable referential identification with a minimum of additional information encoded as linguistic modifications. The principle of partner adaptation seems to be chiefly responsible for the choice of perspective, preferring the interaction partner’s point of view to other options. In some cases, the choice of relatum is similarly affected, but this choice is also influenced by the spatial relationship between the goal object and the entities available as potential relata. The interaction of these three principles can be spelled out, in a first approximation, as follows.

General Discussion

261

Speakers using the general strategy of goal-based reference intuitively seem to make their choices along the following lines. 1. If the configuration allows for unambiguous object reference via the class name, the class name (without spatial descriptions) will be used for reference more often than in other cases. 2. If the configuration allows for unambiguous reference via terms denoting an in–between relation or distance, these options will be used for reference more often than in other cases. 3. When using static dimensional terms, a reference system is chosen that allows for unambiguous reference, i.e., that produces at least one spatial region in which an unmodified dimensional term can be used unambiguously (see 4), if possible at all (otherwise see 7). At least for English speakers, this strategy interacts with a general preference to use intrinsic reference systems with a person (preferably the interaction partner, if present) as relatum, if possible. 4. Unmodified static dimensional terms can be used unambiguously in the following cases, and are therefore employed there more often than in other cases: x if the goal object is situated at a more extreme position on or near the prototypical axis than any competing objects also situated on the same half-plane; x if the goal object is the only one on a half-plane with respect to the reference system used (regardless of whether it is located near the prototypical axis corresponding to the dimensional term used or not); or x if competing objects also situated on the same half-plane are clearly farther away from the axis than the goal object is. 5. If 4 applies for more than one spatial region, the axis is chosen that the goal object is closest to (unless 6 applies), and an unmodified dimensional term is used. 6. If 4 applies for more than one region and the goal object is located at equal distance from two axes (e.g., left and front), then either both dimensional terms are combined or one is chosen at random, via individual preferences, or influenced by the previous discourse. 7. If all regions are occupied by more than one object, modifications, combinations of dimensional terms, counting and/or distance expressions are used more often than in other cases, especially if none of the criteria listed in 4 applies. In such a case, distance expressions

262

Spatial dimensions: Empirical analysis

are specifically likely if the goal object is clearly either the closest one or the one that is farthest away from a suitable relatum. 8. Relata are preferably given explicitly in case of interpretational conflict and when conflated with origins in intrinsic reference systems; otherwise they usually remain implicit. Some kinds of relata (group, picture, and visual field relata) are hardly given explicitly at all. 9. Perspectives are seldom made explicit, except sometimes in cases of conflict if the relatum is difficult to express, and in cases where the chosen perspective is the only source of ambiguity. 10. Usually the perspective of the interaction partner (if present) is chosen, independently of ease of reference. A range of dynamic dimensional terms such as straight may also be used in static situations, though they seem to be fairly restricted with respect to the region of application. This finding may be related to the striking difference concerning the size of the cognitive front region found in earlier research (Section 5.2.3): While Franklin et al. (1995), who investigated static situations in surrounding space, found that front regions were conceptualised as larger than the other regions, Klippel et al. (2004) came to the opposite conclusion with respect to direction concepts. The present data support the hypothesis that this difference is due to the involvement of motion with respect to cognition, which may be reflected in the difference between static and dynamic terms with respect to language, independent of whether motion is actually involved in a discourse situation. 6.6.3. Generalisability of the results Largely, the findings summarised in the previous subsection are in accord with previous results of research on object reference (e.g., Herrmann and Deutsch 1976), where it is known that speakers analyse contrast on any dimension available, with respect to the objects present in a given scenario, and use it in order to achieve unambiguous reference. This observation points toward a high degree of compatibility with other kinds of settings and already recognised cognitive principles, in spite of the obvious limits and artificiality of the study. Nevertheless, real world scenarios clearly involve far greater complexity. For example, objects often differ with respect to both spatial position and non-spatial kinds of features, where under certain circumstances reference on the basis of spatial position seems to be preferred (Pobel et al. 1988). However, the factors influencing such choices

General Discussion

263

and the variability with respect to related options seem to be largely unexplored. For example, in the present study there was one situation that allowed for unambiguous object reference via the class name; however, many participants did not make use of this simple option. Considering more complex real world scenarios, then, in what kinds of situations do speakers deem a simple class name as sufficient for reference? Clearly, this strategy does not only presuppose simplistic referability on the basis of a distinctive object name, but also a clear delimitation of the reference area (among further factors, see Freksa 1980). Such aspects may influence speakers’ choices to a considerable degree, especially in an open setting where the referential domain may be unclear. Brown-Schmidt and Tanenhaus (2003) show that both the form of referring expressions and their interpretation are constrained by previous specification of referential domains. Furthermore, attention focus plays a crucial role (e.g., Kessler et al. 1999). In ambiguous situations, reference is resolved with respect to a subset in focus. Attention can be directed by the speaker through focus and foregrounding on a linguistic level, influencing the listener’s interpretation of spatial descriptions. The applicability of spatial terms is further influenced by functional features of the objects involved, and the relationship between them (e.g., Coventry and Garrod 2004; see Section 5.2.7). In natural communication, interactive processes facilitate the achievement of joint reference (Clark 1996, Pickering and Garrod 2004), and previous experience (i.e., the discourse history) may influence later choices, for example in the employment of reference systems (Watson et al. 2004). All of these major influencing factors need to be accounted for when dealing with natural conversation, in contrast to the simple scenario presented here. Due to the open design of the study a considerably broad range of variety, and systematic patterns of choices, could nevertheless be identified, since speakers’ linguistic behaviour was not governed by instructions to a high degree. This ensures the production of intuitive language to a higher degree than more restricted settings would allow (Fischer 2003). Some indication concerning the generalisability of the results can be derived from a recent experiment in human–robot interaction (HRI) carried out within our research group (Moratz and Tenbrink 2006), in which the same set of scenarios was employed as schematically depicted in the web study presented here. Results showed that the major additional factor influencing the speakers’ linguistic choices was the discourse history. Speakers overwhelmingly re-used syntactic formats in ensuing instructions following

264

Spatial dimensions: Empirical analysis

success in instructing the robot. Apart from that, the general principles and preferences in spatial descriptions identified from the HRI data very much correspond to those discussed in the present chapter. 6.6.4. Speaker strategies The present study showed that, apart from the principles involved in establishing spatial referential identification, speakers can decide to use altogether different spatial descriptions. Among the strategies employed in the present discourse task were descriptions of the spatial direction towards the goal object using a directional term, either with or without reference to the goal object itself, and detailed route descriptions towards the goal. The latter were not expected prior to the studies since route descriptions typically relate to large-scale environments rather than indoor scenarios or those that can be perceived at a glance, which are sometimes referred to as vista space (Montello 1993). Nevertheless, they occurred both in the present web study and in the HRI study reported in Moratz and Tenbrink (2006). On second thought, there are good reasons for this result with respect to both kinds of studies: the participants in the web study were not informed about scale, i.e., the size of the objects and the distances between them were unknown. Thus, although they could indeed perceive the whole scene in the study itself, they may have conceptualised the scene as a largescale environment. This is plausible because the study required a lot of imagination from the start. The participants in the HRI study, on the other hand, were not informed about the robot’s perceptual abilities. Thus, they may have assumed that the robot was not capable of comprehending the scene as such; therefore, they described the movements necessary to reach the goal, and sometimes even lower-level activities (see Fischer and Moratz 2001). Once a goal-based strategy is chosen, the next decision to make concerns the level of precision. With respect to other discourse tasks, dimensional terms are associated with spatial templates, where a close relationship to the focal axis represents a typical relationship that can be expressed by unmodified linguistic descriptions; more deviant relationships require more complex descriptions. In the present discourse task, it is the presence of other (competing) objects, not the exact spatial relationship to the focal axis, that influences the employment of modifications in the linguistic descriptions. Thus, the more competing objects are present on the same half

General Discussion

265

plane, the more precise the description will need to be. A recent direct comparison (Vorwerg and Tenbrink 2007) of referential identification and localisation tasks provided statistical evidence that, other things being equal, localisations are indeed formulated in a generally more precise way than descriptions given for referential identification as addressed in the present study. There is an interesting parallel between these findings on referential identification and a different kind of discourse task, namely, route navigation. In route descriptions, it is essential to identify – in the real world – the path one is expected to follow, as well as the landmarks referred to by the producer of the route description. Verbal descriptions are often incomplete, lacking information about start and end points, directions, length of segments, and the like (e.g., Tversky and Lee 1999). Crucially, identification of the intended entities (landmarks and streets in the correct direction) does not depend on the conveyance of complete or precise information, but on a process that essentially resembles that of object reference. Reference to route elements is achieved on the basis of establishing sufficient contrast to competing candidates, similar to spatial identification tasks. Thus, in both cases, the aim is to identify an entity, either an object or a direction; and in both cases, vague descriptions may be sufficient, depending on the presence of competing entities. However, in route directions, locative dimensional terms are generally only modified very infrequently, even in the presence of competing candidates (Klippel et al., in press). Instead, speakers rely on the usage of informative verbs such as veer (right), ordering information, and reference to landmarks and spatial structure in order to identify future directions. Altogether, the speaker perspective on the spatial task of referential identification as dealt with here seems to be to determine which kind of description is most suitable or least ambiguous with respect to the given goal object, also taking into account the interlocutor’s capabilities. In some complex cases, this might also mean identifying a description that might be suitable for the discourse task at all, provided that the hearer is cooperative and accepts imprecise and fragmentary descriptions. The hearer perspective in the discourse is then to identify which object may be the best match for the given description, or, in some complex cases, to identify an object that could – with a generous interpretation – be described by the given description at all. If the assessments made by speaker and hearer do not match, communication fails, and negotiation will be needed to cooperatively solve the discourse task.

266

Spatial dimensions: Empirical analysis

6.7.

Conclusion

The present study can be regarded as an exploration of natural language produced in an open (though artificial) setting by unbiased speakers, which has led, on the one hand, to an assessment of the diversity in linguistic choices in a range of situations allowing for different interpretations and viewpoints, and on the other hand, to the identification of systematic principles underlying speakers’ choices, which need to be confirmed by more controlled and at the same time more restricted experimentation. In addition to that, in order to gain insights with respect to a broader range of settings the validity of the hypotheses also needs to be tested in other configurations, considering further influencing factors, and considering other languages that may differ from the languages investigated here in more fundamental aspects than English and German differ from each other. In general, however, there is a high likelihood that findings carry over to other related settings, and some of the identified preferences may reflect basic cognitive principles that are not language-specific. Because the present kind of discourse task has previously not been addressed in detail, many of the results are new with respect to their significance in spatial descriptions, especially as far as the details spelled out throughout this chapter are concerned. Nevertheless, they are generally in line with previous results on basic processes involved in discourse. The principles concerning contrastivity and partner adaptation are already established in some detail in Herrmann and Deutsch (1976) with respect to other kinds of (non-spatial) object reference. The principles of minimal effort and further processes concerning adaptation to the interaction partner are addressed, for instance, in Clark and Wilkes-Gibbs (1986) and Clark (1996). Related processes of interactive alignment are spelled out in Pickering and Garrod (2004). Thus, the present work mainly spells out how these principles and processes combine in the application of locative dimensional terms.

Chapter 7 Conclusion: Comparing the applicability conditions

In this final chapter, I will first briefly go through the main findings collected in this book so far. This overview provides the basis for drawing the main conclusions targeted in this work: to assess whether the applicability conditions of temporal and spatial dimensional terms can be said to be conceptually dependent on each other. For this purpose, the differences and similarities in the present findings are related to the conceptualisations of the underlying domains as well as to general principles active in discourse. Finally, the chapter concludes with an outlook on open issues and targets for future research that may enhance further understanding of how basic human conceptions influence the ways in which they put language to use in natural discourse. 7.1.

Summarising the results

With respect to syntax and semantics there are many parallels between spatial and temporal expressions, building the basis for the widespread idea that temporal terms are based conceptually as well as historically on spatial terms. The present book has addressed the question whether this parallel could be identified for the applicability of the terms in natural discourse as well. It started out, in Chapter 2, from the point of view of cognitive science, addressing the ways in which the conceptual domains of space and time are structured and dealt with in cognition and language. The widespread hypothesis was presented and discussed that concepts of time are based – perhaps metaphorically – on concepts of space. The main conclusion drawn from this discussion was that the close relationship between the two conceptual domains is not necessarily due to (quasi–)metaphorical processes or dependency relations. Instead, manifestations of concepts of space and time, such as spatial and temporal language, may be closely related to each other just because of the conceptual closeness of the two domains. The similarities between the domains identified in Chapter 2 point to a high degree of overlap in conceptions. However, there are also fundamental differences that in some cases lead to a decidedly different treatment of basically similar conceptual issues within each domain.

268

Conclusion: Comparing the applicability conditions

The next step in Chapter 2 was to investigate the linguistic means for representing the domains of time and space. Here, the diversity of syntactic forms available for each domain was demonstrated and compared in general terms. Most importantly, it was noted that conjunctions – which are generally treated as signalling discourse relations – are primarily used for the temporal domain, while the spatial expressions are more flexible with respect to other syntactic categories, for instance, prepositions and adverbs. Already this difference in linguistic repertory points to a reflection of the underlying diversity of concepts: focusing on relational terms, spatial expressions relate objects but temporal terms relate events; the ontological difference between these two kinds of entities is reflected in the syntactic difference. Chapter 3 systematically addressed the syntactic and semantic range of temporal dimensional terms. Also, their capability of expressing underlying conceptualisations of causal relationships between events was discussed. This chapter fully relied on previous accounts, which were presented and discussed systematically and carried further, using intuitive judgments of sentences produced by syntactic reformulation tests. The chapter closed with a schematic representation of the range of variety in interpretations of relationships between entities in a discourse that are juxtaposed by a temporal marker. Following this systematic account, in Chapter 4 instances of natural language spontaneously produced by native speakers of English and German were analysed in detail. The first section addressed conceptual categories underlying the application of before and after, such as IMMEDIACY, CAUSALITY, DEPENDENCY, REGULATION, and INSERTION. These conceptions typically point to a proximal temporal relationship that is determined by a conceived semantic relation (often a causal or near-causal one) between the two events that are juxtaposed linguistically using a temporal dimensional term, or by the conception that events directly follow one another. This general effect of proximity associated with the usage of temporal terms can be cancelled by explicit information about the length of the time span. The second section empirically addressed the question whether temporal clauses should be viewed as presuppositional in the pragmatic sense. It turned out that speakers typically take their listener’s information status into account, presenting only information as Given in a temporal clause that is either discourse-old or easily inferable using world knowledge. However, this effect is interrelated with the conceptual categories underlying the application of the temporal term; some of these require a higher degree of Given-

Summarising the results

269

ness than others. These results were, in the third section, examined with regard to the German temporal dimensional terms, where basically the same conceptual categories and effects could be identified. The chapter ended with a systematic overview of conceptual categories underlying the application of temporal dimensional terms in both languages. Chapter 5 turned to the domain of space, identifying first the syntactic range and semantic core of spatial dimensional terms. Next, an extensive and detailed overview was presented of conceptual aspects and issues coming into play in the application of the terms, such as axial structure, reference systems and perspectives, spatial templates (regions of applicability), spatial as well as functional properties of the objects involved, granularity, accessibility, and proximity. The impressive diversity of reference systems available to speakers, internal or external, relative or intrinsic, visual fieldbased or entity-based, landmark-based or group-based, etc., turned out to be particularly intricate. A systematic account of this variability had still been lacking in the literature, although most kinds of reference systems had been investigated in some detail by diverse researchers. The present account integrates these approaches starting from, but considerably expanding, the framework presented by Levinson (2003). These findings were discussed in relation to the discourse task; for speakers’ concrete linguistic choices, it makes a difference whether the location of a known object needs to be specified, the identity of an object is to be determined by spatial reference, or the terms occur as part of a route description task. The complexities involved in determining reference systems were discussed in the empirical part of the investigation of spatial dimensional terms, Chapter 6. While perspective can often be identified unambiguously, reference systems depend furthermore on the spatial relationship to a relatum which is often not specified linguistically. Moreover, the same description may be compatible with a number of different underlying reference systems, with respect to various available kinds of relata, such as the visual field, a picture frame, a group of similar objects, or a landmark. A complex annotation scheme was developed that served as the basis for analysing the data collected in a web study for both German and English terms. The analysis yielded a number of specific results, for example, with respect to language-specific phenomena dealing with the preference of certain syntactic appearances and their restrictions concerning certain kinds of reference systems. Crucially, however, the data generally supported and further clarified, for the discourse task of spatial referential identification, a number of previous findings on achieving reference based on principles of contrastiv-

270

Conclusion: Comparing the applicability conditions

ity as well as partner adaptation and minimal effort. Contrastivity, for example, accounts for the fact that modifications of spatial dimensional terms are not necessary if the described object is the only one on a half plane with respect to a reference system, even if the spatial relationship to the relatum is not prototypical. Partner adaptation explains why, in the overwhelming majority of cases, the perspective of the interaction partner is used (if available). 7.2.

Treatment of spatial and temporal terms in natural discourse

In the empirical chapters of the present work (Chapters 4 and 6), the data were obtained and analysed in two fundamentally different ways that cannot be directly compared. While this fact may, at first sight, seen as diminishing the significance of the present work, or a fault in the approach and design, it may also be related to factors inherent to the application contexts of these terms as such. In other words, it may be the case that the applicability of spatial and temporal terms is different in ways so fundamental that a more direct comparison is rendered unlikely from the start. This idea directly addresses the main question posed in the present work, namely, how the conceptualisations of the domains of space and time influence the ways in which the terms are naturally and spontaneously applied. Spatial and temporal dimensional terms differ considerably with respect to their interpretability on the basis of the available information. With temporal language, most of the information needed for interpretation is often present within the discourse itself, due to the fact that temporal entities are not directly perceptually accessible (a feature addressed in Section 2.3.2). The interpretation of spatial language, in contrast, generally depends on and interacts to a high degree with the spatial situation at hand. In both cases, information may also be drawn upon that is neither represented in the discourse nor directly perceptually accessible; in those cases, it must be present in the interlocutors’ internal mental model of the situations talked about. Such cases were not considered in the present work as it seems impossible to investigate these kinds of internal cognitive processes; therefore, overwhelmingly data were investigated that were collected in situations where speakers did not draw upon shared prior knowledge. Since the underlying concepts involved in application were at the heart of the present analysis, it was necessary to use data where these concepts could be identified to a sufficiently high degree to warrant the formulation

Treatment of spatial and temporal terms in natural discourse

271

of speakers’ principles and preferences. For spatial data, this means that the data were collected in a situation where the spatial configurations were fixed by the experimental design. For temporal data, it entails that it was necessary to take the wider discourse context into account in which the terms were produced, to a higher degree than usually done in the literature. The overview on previous results on temporal dimensional terms presented in Chapter 3 clearly showed that the most crucial aspects concerning the interpretation of these terms relate to the notion of discourse relations, together with aspects of the temporal relation between the events. This is an immediate outcome of the nature of temporal dimensional terms: they connect representations of events, which are typically expressed in full clauses or nominalisations. Connected clauses are subject to (explicit or implicit) underlying discourse relations also in other (non-temporal) cases. Thus, it is only natural for temporal dimensional terms to reflect underlying discourse relations. Spatial terms, in contrast, are used to connect representations of objects, which typically happens via noun phrases. Discourse relations between noun phrases, however, have not been claimed to be relevant so far, except for the case of nominalisations (e.g., Knott and Dale 1994, Grote 2003, Grabski and Stede 2004). Accordingly, the analysis of the temporal terms focused on a systematic investigation of the kinds of discourse relations that come into play, which reflect the kinds of relations between the events related by a temporal dimensional term. A study that prescribed the kind of conceptual situation in which the terms should be employed, perhaps paralleling the design of the web study employed in the present work for investigating the spatial dimensional terms, would not have been suitable for detecting the variability available to speakers in this regard. It was therefore necessary to draw upon data collected in situations where the temporal dimensional terms were employed spontaneously, rather than being triggered by an experimental design. With respect to spatial data, in contrast, no corresponding results regarding conceptual variability with respect to the related entities could be expected. As the overview of previous research in Chapter 5 clearly showed, the most central aspects involved in the application of spatial dimensional terms relate to the conceptualisation of the spatial situation in terms of reference systems, and to the influence of discourse-related aspects, such as the current task, on the specific linguistic choices speakers make. As research on some kinds of discourse task (e.g., tasks triggering answers to Where questions) is abundant in the literature, while others – for instance,

272

Conclusion: Comparing the applicability conditions

the task of referential identification (answering a Which question) – have mostly been neglected, the empirical analysis of the present work focused on one discourse task and investigated it in a number of discourse situations in order to work out general structures of application. In this case, extracting spatial data from already existing corpora of language produced in nonspatial settings, paralleling the method used for the temporal data, would not have led to reliable results: the underlying conceptualisations and reference systems are typically impossible to identify if the spatial situation is not available to the analyst. The complexities involved here even in a situation that is as transparent as possible were dealt with extensively in Chapter 6. A brief look at the occurrences of spatial terms within the German Ulm corpus (used for the analysis of German temporal dimensional terms in Chapter 4) exemplifies these observations. In this corpus, there were 229 occurrences of the adverb vorn(e) (front), 103 of which occurred as part of the collocations von vornherein (from the outset) and von vorn(e) (from scratch). The others partly referred to body features as in the following utterance: (7.1)

bis dann hier Januar dann die Kopfschmerzen hier vorne eben einsetzten ‘until here in January, then, the headache started here at the front’

Others referred to the features of a different relatum which is typically not mentioned explicitly, as in: (7.2)

wir können Ihnen die dann vorne geben. ‘then we can give them to you at the front.’

Such occurrences are not particularly informative with respect to the present goals of analysis, mainly because the situational factors cannot be determined conclusively, although parts of them can be inferred: vorne in (7.2) probably refers to a front desk at the doctor’s practice. With other spatial dimensional expressions and syntactic forms, the situation is even worse because they are either highly ambiguous (such as vor which occurs more than 3.000 times in the data with multiple meanings) or hardly occur at all, and then only with reference to body parts. Thus, the adjective linke (left) occurs 67 times for body parts, twice in a political sense, and once in the sense of unfair. Nevertheless, it is conceivable that an analysis of such data – if methods can be found to identify relevant usages automatically – would be informative as a later step in subsequent analysis, based on the results obtained from more restricted settings as in the present book.

Treatment of spatial and temporal terms in natural discourse

273

Thus, the nature of the terms under analysis itself influenced the choice of how to obtain suitable data. The motivation for using the kinds of methodology employed for investigating these data, then, follows directly from these considerations: since the factors influencing the application of the terms differ fundamentally as identified on the basis of the state of the art, the methods for investigating these factors must also differ. Clearly, the assumptions concerning what are the most relevant aspects involved in the application of spatial and temporal dimensional terms cannot be proved or disproved in this way; the present analyses built on these assumptions. They carried further previous results and hypotheses, formulated them in more detail, and in some cases spelled out additional hypotheses which, in turn, need to be addressed directly (see Section 7.4). However, the present analyses are neither intended nor suitable for supporting the hypothesis that, for spatial terms, underlying discourse relations are not relevant, and, for temporal terms, different kinds of reference systems do not come into play as much as for spatial terms. No evidence for these hypotheses (which are well-founded from previous literature) was looked for in the data, and the different ways of data collection and analysis certainly impede a direct comparison in this regard. Nevertheless, it seems to be safe to say that there was nothing in the data that actually disproved or cast doubt on these basic assumptions. But this does not mean that the results cannot be compared at all. On the contrary: it was one of the prominent tasks of the present work to identify ways in which underlying principles of application can be investigated. The more theoretical discussions of previous literature in Chapters 3 and 5 aimed particularly at working out in detail in how far principles of application for spatiotemporal terms have already been identified in the literature, and in what ways these findings can be supplemented by further empirical work, which was then carried out in Chapters 4 and 6. Thus, the realisation that different kinds of data, and different methods of analysing them, were required for the treatment of spatial and temporal terms can be regarded as a partial result of the present work. This realisation certainly points to fundamental differences in application, contrary to the assumptions in the literature that temporal and spatial terms are so close that they even depend directly on each other. While this dependency hypothesis might in fact hold true to some degree with respect to the purely spatial and temporal aspects, i.e., the core semantics identified for both kinds of terms, it does not seem to be valid for the factors influencing application.

274

Conclusion: Comparing the applicability conditions

Miller and Johnson-Laird (1976: 456) address a similar issue by pointing to the differences in psychology between the two domains: [A]lthough the language of time borrows heavily from the language of space, the psychology of time is very different from the psychology of space. Such differences are not explicated by formal semantic analysis.

These authors also treat the domains of space and time differently because of the underlying conceptual discrepancies (1976: 457): We have chosen to be inconsistent for several reasons. From a psychological point of view, the most important reason is that searching for an event is very different from searching for an object – moving around in time is impossible.

Further reasons for a different treatment of spatial and temporal language, according to these authors, are the obligatory marking of tense in temporal language which remains unparalleled in spatial language, and the differences in referring to absolute temporal locations using “an elaborate system of names” including numbers and the conceptualisation of a time line, which is not available for absolute spatial locations in the same way. Thus, in spite of the many parallels between space and time which have been identified in the literature, notably also by Miller and Johnson-Laird themselves, there are general conceptual differences that lead to farreaching consequences with respect to psychology and – as hypothesised in the present work – linguistic pragmatics, i.e., the applicability conditions and circumstances of spatiotemporal expressions. In the following, the nature of these interrelations between application and conceptualisation will be examined in some detail. 7.3.

Comparing the applicability of spatial and temporal dimensional terms

The main interest in the present work concerned the spontaneous usage of spatial and temporal dimensional terms in natural discourse, explored both via a systematic account of earlier findings and by new empirical investigations. The following generalised applicability conditions emerge. For temporal terms, the main conceptual aspects affecting application are concerned with the relationships betweeen events, interrelated with the nature of the events themselves, and the knowledge status of the interlocutor. Thus, temporal dimensional terms are employed whenever two events need

Comparing the applicability of spatial and temporal dimensional terms

275

to be juxtaposed that are conceptually interrelated in some way, which is often causal in some sense. A purely temporal relationship between the events is the reason for application only in a subset of the cases, namely, when the temporal relationship itself is in focus, or its duration via mentioning an explicit length of time between the events. Otherwise, temporal relationships may be expressed by other, more indirect linguistic means such as clause order, tense, and aspect. Apart from the cases in which the temporal relationship itself is at stake, the presence of a temporal dimensional term usually points to the conceptualisation of some sort of interconnection between the events. However, these expressions are explicit only with respect to the temporal relationship between the events, but not with respect to any kinds of associated (causal or other) relationships. Therefore, a further requirement seems to be that the speaker does not wish to directly impose and represent a stronger relationship. Thus, the application of a temporal dimensional term leaves much freedom of interpretation to the interlocutor, which is sometimes reflected by the wider discourse context where implications such as the conceptualisation of a causal relationship may be made explicit. Concerning linguistic variability, it seems that specific syntactic forms of temporal dimensional terms are not directly associated with specific concepts, but most findings apply to most forms in both German and English (with the exception of English before as a sentence adverbial). However, the different conceptual categories underlying the application of temporal terms are associated with different degrees of Givenness in the discourse. For example, if the temporal relation itself is in focus (in REGULATION), both events should be already known or at least easily inferable. For most conceptual categories, there are specific linguistic markers that may further enhance the conveyance of the associated relationship, such as even, directly, or the explicit mention of a specific time length. With respect to spatial terms, the main conceptual aspects affecting application are concerned with the association of an underlying reference system, the discourse task (which influences the degree of specificity and explicitness as well as the choice of axis), and the functional relationship between the objects, which includes the nature of the objects themselves. The wider discourse context may influence the choice of reference system as well as the syntactic form. The application of spatial terms is warranted in a situation where spatial relationships are to be represented for some reason, since language does not offer any other (more indirect) means of doing so. Out of the overall repertory of spatial terms, spatial dimensional

276

Conclusion: Comparing the applicability conditions

terms are applied whenever a spatial axis is relevant, for example, because a spatial direction needs to be specified, or because alternative terms, such as distance expressions or those expressing in–between relations, are not contrastive in a situation requiring contrastivity. Perspectives are seldom mentioned explicitly (and, at least in the scenario used in the present web study, even less in English than in German), and relata are mentioned mostly if this counteracts ambiguity. Linguistic modifications seem to be more prominent in discourse tasks involving the description of a spatial relation rather than the identification of an object out of several possible candidates, where modifications only come into play if reference is otherwise not unambiguous. Areas of applicability are furthermore affected by the functional relationship between the objects involved. Thus, it seems that, with temporal dimensional terms, what (often) remains implicit is the conceptualised relationship between the entities, while with spatial dimensional terms, it is often one or several of the entities themselves (relatum and origin) which remains implicit, resulting in underspecification with respect to the underlying reference system. This directly reflects the ontological difference between objects and events, discussed in Section 2.2.1 in terms of differences in abstraction: objects are directly perceivable and therefore in some cases do not need to be specified linguistically, while events are more abstract and must therefore be retrievable from or delimited by the discourse itself. Since the relationships between events are not always clearly defined or conceived of, they sometimes remain underspecified in linguistic representations. But in both cases, the interlocutors do not necessarily differentiate between the possible interpretations if more than one works well in the context and no ambiguities arise. Thus, there is an interesting parallel between discourse relations and reference systems: in each case, there are a number of associated concepts that are easily inferable, which may or may not be reinforced by additional linguistic material, and which are often not determinate, that is, there may be more than one possible associated concept that works well in the discourse context without raising communication problems. With respect to the discourse tasks in which the dimensional terms may be applied spontaneously, there does not seem to be much overlap. The main parallel occurs with tasks in which the spatial or temporal relationship itself is in focus, i.e., in utterances that may serve as answers to pointed questions such as “Where is the [known] object [in relation to another known object]?” and “When did the [known] event happen [in relation to another known event]?” In other words, in the rare situation where both

Comparing the applicability of spatial and temporal dimensional terms

277

entities (objects or events) are known beforehand and only their relation needs to be negotiated, then the domain-specific dimensional relationship is in focus, and no further aspects need to be associated or inferred. However, as has become clear, this is often not the case. The dimensional terms of both domains may serve many other discourse functions. For spatial terms, the analysis has been concerned most extensively with referential identification, a task which is not straightforwardly transferred to the temporal domain: it is hard to imagine a situation in which an event needs to be singled out from a number of possible candidates using a temporal dimensional term. This, of course, is again due to the ontological difference between the entities involved (cf. Section 2.3.2). Non-simultaneous events cannot be perceived together, therefore they cannot be “singled out” easily on the basis of their temporal relation. This notion simply does not apply in the same way for temporal relationships. Furthermore, even in the case of a discourse task involving “Where”, which seems to parallel the “When” question as just mentioned, it seems to be the case that “Where” questions much more often involve spatial dimensional terms than “When” questions involve temporal dimensional terms. This is, in the first place, again due to differences in accessibility: it is possible to retrieve the spatial position of an object if one learns about its spatial relationship to another object. However, events are not accessible as objects are; reference to another event therefore does not enhance retrieval in general, but is typically only informative if the two events are somehow conceptually related. Second, as mentioned by Miller and Johnson-Laird (see above), for temporal positioning absolute measures are predominantly used; therefore, a much more typical answer to a “When” question would be to simply mention the event’s date and time of occurrence, rather than describing its relationship to another event. Both domains of space and time equally come into play in situations involving sequence or motion; in those cases, spatial, temporal, and specific sequential (such as next, last) or ordering terms (such as numbers) may be used for the expression of relationships. For spatial dimensional terms, the main area of investigation in the present work has been a setting in which contrastivity was specifically crucial. Notice, however, that the notion of contrastivity is not necessarily restricted to discourse tasks – such as referential identification – in which the establishment of a contrast to competing candidates is specifically prominent. In a certain sense, it is inherent to all language: every utterance has the effect of ruling out conceivable alternatives. This aspect of linguistic communication has been widely discussed and investigated from a number of perspec-

278

Conclusion: Comparing the applicability conditions

tives (e.g., Sperber and Wilson 1986, Levinson 2000, Carston 2002, and ‘Alternative Semantics’ which extends the idea to claims of truthconditional effects, cf. Rooth 1992). It will not be pursued in much detail here, although a few points need to be made in order to work out how this idea might enhance our understanding of dimensional terms. For this purpose I start from some claims formulated by Nemo (1999), who talks about the difference at stake: “what is said (or asked) is relevant insofar as it makes a difference” (Nemo 1999: 353). This difference may be viewed as the set of alternatives that the utterance distinguishes. Such alternatives can be made explicit to some degree by exchanging the term or phrase in focus by alternative terms or phrases. Thus, if an utterance like the box on the left has its (prosodic) focus on left, then the contrasting alternative would be right, as made explicit by the box on the left, not the one on the right. If box were in focus instead of left, then the contrasting alternative could be expressed by the box, not the hat on the left. Similarly, if a temporal term (before) is in focus, the contrast is after. But if other participants in the clause are in focus, temporal order is not at stake; the alternatives concern other elements. Therefore, in cases of REGULATION, where the temporal relationship itself is focused upon, it is possible to make the contrast explicit by saying I coughed before, not after I sneezed. However, with other conceptual categories this is not as straightforward. Take the following well-known example for a causal implication conveyed by a temporal dimensional term: (7.3)

He fell after he stumbled over a stone.

Here, it seems to be inappropriate (if a causal relation is indeed inferred) to exchange the temporal term by its opposite in order to make the implicit alternative explicit: (7.4)

(7.3)’ He fell after, not before he stumbled over a stone.

Instead, it seems much more likely that the speaker wants to rule out alternatives like the following: (7.5)

(7.3)” He fell after he stumbled over a stone, not just out of clumsiness.

Of course, such underlying alternatives cannot be determined unequivocally by linguistic tests, due, on the one hand, to the general subjectivity of such tests, but on the other hand, also to the indefinite range of possible alternatives that may arise in some cases, such as example (7.5). In (7.4), pointing out a suitable contrasting alternative is straightforward because the term in focus possesses a clear opposite. Nevertheless, the underlying gen-

Comparing the applicability of spatial and temporal dimensional terms

279

eral principle is clear: the temporal or spatial relationship is only at stake if it is in focus or otherwise particularly relevant. In other cases, it contributes to the general proposition, which may convey somewhat more than just a spatial or temporal relationship. Nemo (1999: 364) goes so far as to claim that “semantic meaning must explain use in general, and not primarily (apparently) conventional uses”, which comes close to saying that the lexical semantics of all expressions can only be understood by way of considering the application in context. He claims furthermore (1999: 376) that the semantic content of a word is not a description of what it conventionally refers to, but a comparison of different things or states to isolate the specificities of something.

Thus, according to his view, the semantics of each term can only be defined on the grounds of differences to other terms, and the interpretation in context is dependent on the grounds of differences to alternative utterances that might have occurred in this context. This regularly leads to the inference of further information, as exemplified by the following example given by Nemo (1999: 400): (7.6)

Bill Clinton is alive.

This utterance “can represent only a moment when something has happened (an accident, a heart attack, an assassination attempt, etc…)”. Otherwise, i.e., if there is no reason why Bill Clinton should not be alive, the utterance is meaningless. Therefore, the utterance “represents both the fact that he is alive and the fact that this might (strongly) not have been the case.” In a similar vein, Freksa (1999) points to the fact that the interpretation of spatial terms may be dependent on the alternatives available in a certain situation. In the light of Nemo’s account, this observation seems to hold for the interpretation of all kinds of words and phrases. The general idea builds on the insights gained within Relevance Theory (Sperber and Wilson 1986) and other theories dealing with the idea that an utterance rules out alternatives that the context makes available (e.g., Halliday’s Given/New structure, cf. Section 4.2). Thus, the insight is neither new nor restricted to accounts dealing directly with discriminativity. However, with respect to the terms under analysis in the present work, it provides a suitable background for accounting for the variability of influences that the discourse context may have on the interpretation of spatial and temporal dimensional terms. For the spatial terms, the idea of discriminativity has been particularly rele-

280

Conclusion: Comparing the applicability conditions

vant due to the nature of the discourse task in the empirical studies. However, a recent study (Vorwerg and Tenbrink 2007) comparing the particular task of referential identification with that of localising an object in relation to another in answer to a Where question showed that, although utterances tend to be significantly more precise with Which questions, competing candidates (contextual alternatives) clearly also play a role in answering Where questions. Furthermore, similar processes are also active in other kinds of discourse, contributing, for example, to the inferability of further, non-temporal relationships between events juxtaposed by a temporal term as in example (7.3) above. Thus, general discourse processes contribute to the applicability of expressions in context, affecting their interpretation and applicability conditions in different ways depending on the nature of the terms themselves as well as of the discourse factors active in each kind of context. As discussed in Sections 2.3.1, some fundamental conceptual differences between the two domains concern the notions of linearity and directionality, where the temporal domain is much more restricted than the spatial domain. These aspects can be related to an interesting general conceptual distinction addressed, for example, by Vorwerg (2004). She describes two kinds of attribute categorisation, which she opposes in terms of quantitatively versus qualitatively variable attribute dimensions. Thus, some kinds of attributes behave in a ‘circular’ way, such as colours, which “may vary continuously, but there is no ranking order of values except within partial segments of the dimension, with relation to prototypical values”. Therefore, “[a] stepwise change of value leads back to starting value (…) [a] value of zero is meaningless.” Other attributes, in contrast, behave in a ‘linear’ way, such as size or loudness: with these, “there is a continuous variation of values, which possess a ranking order”, leading to the effect that “[a] stepwise change of value leads further and further away from starting value”. For both categories, context effects lead to the establishment of reference values: for example, one mountain may be higher than another mountain, so that the latter appears as small in comparison; nevertheless, it is still high as compared to a hill. With qualitative dimensions, prototypical values are associated as “cognitive reference points” (Rosch 1975). Vorwerg argues that distance belongs to the ‘quantitative’ category, while direction is an example of a qualitative variable attribute. On the grounds of this distinction, it is perhaps possible to say that the conceptual domain of time behaves more like a quantitatively variable dimension, whereas the domain of space is more similar to a qualitatively

Comparing the applicability of spatial and temporal dimensional terms

281

variable dimension. Thus, dimensional expressions in the temporal domain are to be understood solely on a linear scale, while (even related) dimensional expressions in the spatial domain are cyclic. There are no digressions from the linear scale in temporal expressions, only differences in distance. But with the spatial domain, locative dimensional terms always have a more or less prototypical relationship to the focal axis (“a stepwise change of value leads back to starting value”). Of course, effects of distance also play a role in the spatial domain, but always with respect to a spatial direction. This distinction explains, for example, why effects of prototypicality (accounted for in terms of spatial templates for spatial dimensional terms) do not seem to play any role for temporal dimensional terms. In this respect, the superficially similar terms for the frontal direction and temporal order (in front / before, vor / bevor) differ fundamentally, in contrast to their assumed conceptual interrelation. For time, it has furthermore been noticed that the understanding of time is intricately connected to the notion of causality. Events that are perceived as temporally related are typically also conceived of as causally related. A set of temporally neighbouring incidents would probably not be considered as an event, as a whole, without there being some perceived – perhaps causal – connection between subevents. As our conception of temporality influences our linguistic representations of it, it comes as no surprise that the conceived interrelationship between events plays a major part in the applicability of temporal dimensional terms. This phenomenon is not mirrored directly for space: at least, objects in a spatial surrounding are not causally related. But they may be related in different ways, which indeed has an influence on the applicability of spatial dimensional terms, as shown by the research on functional geometry. Thus, with respect to the nature of the entities involved, there seems to be another interesting parallel: in both cases, functions and conceptions associated with the entities, and with the relationship between the entities (cf. Section 2.3.3), have a clear influence on application. In the case of spatial dimensional terms, this applies most prominently to the research area of functional geometry discussed in Section 5.2.7 as well as to some functional features associated with the spatial terms themselves (cf. Section 5.1.3). These findings all point to the fact that speakers using spatial language conceptualise the spatial relationship not simply in geometric terms but also with respect to the associated interaction between the objects. In the case of temporal dimensional terms, a perceived interaction between the events juxtaposed by a temporal marker seems to be the normal case rather

282

Conclusion: Comparing the applicability conditions

than the exception, as language offers other (less explicit) possibilities for conveying temporal information. Thus, a range of conceptual categories representing different kinds of conceived relationships between events account for the application of temporal dimensional terms. Crucially, however, functional relationships between objects influence the application of spatial terms in that the applicability areas for specific terms in relation to a focal axis may change in relation to function. With the temporal terms, it is the applicability itself that is at stake: without a conceptual relationship between the events, temporal dimensional terms would typically not be used at all, since they support the inference of an underlying relationship. Here, again, it seems to be the case that, although not all underlying concepts and relationships are made explicit in the discourse, it is still a normal and legitimate process to make appropriate and partly far-reaching inferences on the basis of world and situation knowledge. These processes are suggested in the literature in a strikingly parallel way by authors concerned with only parts of the above mentioned discourse aspects: Recall Coventry and Garrod’s (2004: 146) suggestion with respect to the most informative relation (see Section 5.2.7, repeated here for convenience): [V]arious constraints work together to determine the appropriate meaning of a spatial expression in context. (…) [S]ituation models offer the vehicles by which these multiple constraints come together. Meaning in the functional geometric framework reflects the result of all these constraints coming together in the situation model, which supports the most informative relation between a reference object and a located object. By informative, we mean that the model should support the strongest inferences that could be drawn about the scene.

This proposal parallels Asher and Lascarides’ (2003: 20) hypothesis of maximum coherence (see Section 2.4.4) which straightforwardly applies to the kinds of discourse relations associated with temporal dimensional terms: the current utterance is relevant only if it’s rhetorically connected to something in the context (…) Interpreting discourse so that it’s maximally relevant amounts to interpreting discourse so that it’s maximally coherent. (…) All else being equal, the more rhetorical connections there are between two items in a discourse, the more coherent the interpretation.

Thus, in both cases the authors suggest that the inferences associated with linguistic representations should be as strong as possible, considering the constraints of the discourse. The more intricate the interrelations between the connected entities are, the more coherent and informative the discourse

Comparing the applicability of spatial and temporal dimensional terms

283

becomes. By these processes, the application of a seemingly simple dimensional term – carrying only information about spatial or temporal relationships in its semantics – becomes a highly complex process triggering a vast number of possible associations and inferences that work together to create a coherent model of the situation at hand. Thus, besides fundamental differences there are also a number of striking parallels in the application of temporal and spatial dimensional terms. But is there any indication that these parallels are due to a dependency relation; does the application of temporal dimensional expressions in some way depend conceptually on the application of the morphosyntactically and semantically similar spatial dimensional terms? Such a conceptual dependency relation for applicability conditions would surely entail that the ways and situations in which the “source” terms, i.e., the spatial ones, are applied, are somehow transferred to the ways and situations in which the “target” terms, the temporal ones, are applied. Thus, there should be some striking overlaps with respect to the application areas, and the conditions should be interrelated in a straightforward way. However, in fact the application areas of the two kinds of terms in their core senses (spatial terms with spatial meanings, temporal terms with temporal meanings) do not seem to have very much in common. It will be recalled that the English terms before and after have often been described as being capable of describing spatial relations as well, an observation which, indeed, serves as a motivation for the claim that the temporal interpretation is based on the spatial one. However, throughout the present data, in both kinds of spatial studies, these terms were almost never used in a spatial sense. An exception is the following utterance, which is also sequential: (7.7)

Go to the cube just in front of you to the right, before the sphere

This occurred once in the web data. Otherwise, the rare occurrences of before and after in the web corpus were used with a temporal meaning throughout. Thus, it seems that spatial terms are used in different kinds of situations than temporal terms are, and there is almost no conceptual overlap to be found here at all. The explanation for this is immediately obvious on the basis of the present findings: temporal dimensional terms are associated with a range of conceptual relationships between events, while spatial dimensional terms are used for the specification of a primarily spatial (though sometimes also functional) relationship. It comes as no surprise that such situations do not overlap to a high degree. Furthermore, temporal relationships may often remain implicit, while spatial relationships need to

284

Conclusion: Comparing the applicability conditions

be expressed directly in language: there simply is no default assumption for implicit spatial relationships paralleling, for example, Dowty’s (1986) Temporal Discourse Interpretation Principle which serves as a default for interpreting implicit temporal relationships (see Section 3.1). Moreover, the applicability principles identified for the two kinds of terms in the present work do not seem to be conceptually related even in those cases where interesting parallels were identified. There is no way in which, for example, the conceptualisation of a causal interrelation between two temporally related events may be viewed as a metaphorical (or otherwise dependent) extension of the functional relationships that may be present between two spatially related objects. This parallel exists simply because both kinds of entities, objects and events, may have some kind of relation to each other apart from their spatial or temporal relationship, and this has natural consequences for the applicability of the terms. But since the kinds of relationships that may be present differ fundamentally between the conceptual domains, the kinds of consequences for applicability differ as well: While in the temporal case, the relationship between the events is often the reason for using a temporal term at all, in the spatial case the interpretation of the spatial relation is affected by the relationship between the objects. There are a number of further parallels: proximity, for example, is a standard assumption with the usage of dimensional terms in both domains that can in both cases be overridden by the usage of metric information. In both domains, the notion of immediate adjacency, or the assumption of a direct uninterrupted spatial or temporal relationship is relevant at least for some usages of the dimensional terms. Also, phenomena related to the conceptualisation of a Figure and a Ground are relevant in both domains (cf. Chapter 2). However, in all of these cases there is no reason to assume a dependency relationship either. It is much more likely that general principles such as relevance and choice of topic and focus are at work, which participate in all linguistic communication, in addition to shared conceptualisation principles concerning the two domains. Taken together, the findings identified and discussed in the present work all point to the conclusion that spatial and temporal dimensional terms share a number of basic features with respect to their application in discourse, without there being a dependency relationship. Parts of these parallels between the two kinds of terms can be traced back to shared issues of the underlying conceptual domains of space and time. Other kinds of shared features are due to general principles operating in discourse, such as inference processes and the effect of ruling out possible alternatives. In

Outlook

285

addition, this work has identified features of applicability where the two kinds of terms clearly differ. With respect to these features, the relation to fundamental conceptual differences between the domains is typically straightforward: time is associated with a number of aspects that are irrelevant for space, and vice versa. Therefore, temporal dimensional terms exhibit different kinds of applicability conditions than spatial dimensional terms do. 7.4.

Outlook

This book has opened up a new and innovative perspective on a popular and well-examined topic, namely, the relation between space and time in general, and between spatial and temporal dimensional terms in particular. Although a number of basic as well as specific results have been obtained, there still remains a lot of work to be done. Obviously, there is no claim for the present findings to be exhaustive; in fact there are good reasons for pursuing the methods and approaches used in the present work further. The analysis of other, more complex configurations in addition to those analysed in Chapter 6, may, on the one hand, support the hypotheses presented here or provide arguments against them, and on the other hand, lead to the discovery of further application principles. In a similar vein, the analysis of further data containing temporal dimensional terms will surely lead to further refinement of the conceptual categories suggested here. Furthermore, it will be interesting for later work to “fill the gaps”, so to speak. As mentioned above, the investigation of spatial terms has focused on different kinds of data and research questions to that of the temporal terms. Now, on the basis of the detailed results within each of these empirical approaches, it should be possible to supplement the findings with respect to what has purposively been neglected in the present work. Thus, it should be interesting to investigate further kinds of discourse tasks for spatial dimensional terms, based on their spontaneous usage (as done here with temporal terms) and identify usages that differ from the main applicability areas identified here in terms of “Which” questions and “Where” questions. One further area that calls for intensive treatment on its own – and which, in fact, has already been addressed from a number of perspectives – concerns the investigation of the discourse task of describing routes (e.g., Denis 1997, Ligozat 2000), where spatial dimensional terms are used to identify streets and landmarks, and to specify future directions.

286

Conclusion: Comparing the applicability conditions

First results that establish the relationship between the present findings and other work in the wayfinding research area are presented in Klippel et al. (in press). Similarly, it should be possible to design experimental studies triggering the usage of temporal dimensional terms without actually prescribing their employment. The aim would be to mirror the spatial web experiment used here which enabled (among other things) the investigation of what kinds of alternatives speakers have at their disposal in a situation in which spatial reference is likely to be employed. It would also be interesting to find a way of determining whether there are, after all, subtle differences with respect to the occurrence of the various conceptual categories together with the diverse syntactic forms in which temporal dimensional terms may appear. In the present work, for methodological reasons no assessment of relative frequencies in this respect could be made for the empirical data. One major outcome of both theoretical considerations and empirical investigations in Chapters 5 and 6 is the insight that reference systems do not need to be specified in detail; in speakers’ utterances, information about the origin or the relatum, or both, is often left out. Speakers’ mental conceptualisations may not necessarily contain specifications of the underlying reference systems that go beyond the information conveyed linguistically. This may be the case especially if the situation itself does not require a differentiated specification. This result maps directly on psycholinguistic findings (e.g., Carlson 1999) which lead to the conclusion that several reference systems are activated simultaneously in spatial reference. Also, it has been shown that readers do not necessarily build up specific models of the spatial (or other) information contained in texts (Vandierendonck et al. 2004). Since the lines of research seem to converge at this point, it seems worthwhile to investigate further in how far linguistic representations reflect underlying conceptual underspecification, especially if the situation does not require a differentiation. The methods employed in the present work were entirely qualitative (though partly supported by relative frequencies of choices), which was suitable and advantageous for present purposes. Starting from here, it is a particularly promising endeavour to combine the discourse-analytic approach adopted here with the additional weight that elementary statistical methods can offer. In part, statistically supported follow-up work has already been done (Vorwerg and Tenbrink 2007); for other areas (e.g., the distribution of conceptual categories in temporal term usage) it may not appear as feasible, though other supporting measures may be developed.

Outlook

287

Within the area of psycholinguistics, a number of further relevant issues have been identified that were discussed in Chapter 5 but not addressed empirically in the present work. Thus, the applicability of spatial terms is further influenced by the functional features of the objects involved, as well as a range of aspects of the discourse context such as attentional focus and referential domains (cf. Section 6.6). For example, it is an open question how the contrastive principle identified here as crucial for referential identification tasks would interact with the functional principles at work, for example, in determining whether an umbrella is over a person, protecting them from the rain. Finally, although the background of the present work is the perspective of cognitive science which is (at least in principle) language-independent, the empirical analyses and most findings discussed from earlier research were restricted to only two closely related Indo-European languages, English and German. With respect to this limitation, it would certainly be a rewarding venture to investigate similar terms in other languages, also from different language groups. For the semantics and morphosyntax of the dimensional terms, this has already been done (e.g., Haspelmath 1997), the results of which were the original starting point and motivation for the present work. Here, the applicability conditions for the expressions in only two languages could be investigated: results show a distinctive behavior of spatial versus temporal terms with respect to application, contrasting with the findings on semantics and morphosyntax, and furthermore, there are some systematic differences between even these two closely related languages. These findings provide good reasons for assuming that the deep-rooted conceptual differences between the two domains of space and time would result in different applicability conditions also in other languages.

Appendix

App. 1. Corpus examples from the Switchboard corpus This section provides further examples of dialogue analyses according to the criteria motivated in Section 4.2. Corpus example 1.

(Context: death penalty)

B: (…) I just saw on the news last NIGHT, that they said the average time a sentenced MURDERER, you know, is in jail is two YEARS before he’s PAROLED, and a rapists is like SIX MONTHS, (…) A: That’s pathetic. (…5 turns…) and the way the law reads, uh, if they sentence you to life in prison, then he’s available for parole. If it’s life and a day, then he’s not eligible for parole.

Category: Explicit length of time Common ground: communal knowledge (convicts can be paroled) Information: inferable: close semantic relation to previous context of legal procedures Given/New: strong prosodic prominence is on years, weaker stress on paroled: the information in the temporal clause is news to the hearer, but not as prominent as that in the main clause. Listener’s reaction: at first no comment on the temporal component; several turns later: evaluation (conceivably the backgrounded information took more time to react to) Corpus example 2.

(Context: movies)

A: (…) I must admit I did look at my WATCH after about an hour. B: Did you?

Category: Explicit length of time Common ground: communal knowledge on time Information: discourse-new Given/New: presented as Given; no prosodic prominence on the temporal component Listener’s reaction: no comment

Corpus examples from the Switchboard corpus

Corpus example 3.

289

(Context: owning houses)

A: We just bought our first HOME (…) after being in school and so, we’re just first time homeowners. We’ve only been living here about six MONTHS. B: Okay, uh, well, I’ve lived in the, in the Dallas and Richardson area for about, uh, twenty-five and a half years, twenty-six years.

Category: IMMEDIACY (covering large time spans) Common ground: communal knowledge (people go to school) Information: discourse-new Given/New: presented as Given: prosodically backgrounded, Agent omitted Listener’s reaction: no comment on the temporal component Corpus example 4.

(Context: Pollution)

A: (…) I don’t know if you ever saw that sixty minutes thing where they showed a chicken PROCESSING plant. (…) I couldn’t bear to think about eating a CHICKEN after seeing THAT. And, but the government says it’s SAFE. Uh–huh. B: Uh, yeah.

Category: CAUSALITY Common ground: discourse context Information: discourse-old, anaphorical construction Given/New: prominence on that, indicating the speaker’s intention to point to the importance of the event referred to (enough to serve as a reason for the speaker’s aversion to eating chicken) Listener’s reaction: no comment Corpus example 5.

(Context: politics)

A: The only, the only thing I see about Cuba though, is, uh, after Fidel Castro dies, I don’t think they’ll be a communist POWER anymore. I, I can’t see communism in that country carrying on PAST him. B: Well, when you take a, a situation where I think in particular in Salvador (…)

Category: CAUSALITY Common ground: world knowledge (people die) Information: discourse-new Given/New: presented as Given, no prosodic prominence Listener’s reaction: no comment on the temporal component

290

Appendix

Corpus example 6.

(Context: Puerto Rico as a state of the US or independent)

B: And I think, uh, having listened to you relative to the ECONOMY thing, I think if I were being forced to make a DECISION I would plead ignorance and WAIT to do more research before PICKING one of these. So I’m ul–, I guess I’m ultimately in favor of status quo also at this point, A: Well that’s interesting.

Category: DEPENDENCY (doing more research as a precondition for picking) Common ground: discourse context Information: inferable from the previous discourse (close semantic relation) Given/New: prosodic prominence on picking, thus emphasizing the point that picking is a major process that needs preparation Listener’s reaction: no comment on the temporal component Corpus example 7.

(Context: camping)

A: And we managed to, uh, fortunately the rain didn’t last long but we managed to get them all back up and, uh, got SOME sleep in that night. And after we got, uh, BACK from the weekend, we checked with one of the parents of one of the other COLLEGE kids and they were a little concerned. THEY had heard on the RADIO that the winds on the DAM there anyway, had gotten up to NINETY-SEVEN miles an hour, B: Good grief.

Category: DEPENDENCY Common ground: returning from a holiday: inferable from the discourse context of camping Information: discourse-new; informationally sparse because it is not stated where back is supposed to be, nor was there any mention of a weekend in the previous discourse Given/New: presented as Given by a definite article; prosodic prominence on back, indicating that the most important part of the temporal clause is that we needed to come back in order to be able to check with one of the parents Listener’s reaction: no comment

Corpus examples from the Switchboard corpus

Corpus example 8.

291

(Context: automatic phone calls advertising goods or services)

A.: I’ve even had some of them, the, they’re voice activated and you’ve got to say hello twice before they’ll do anything. B.: Uh–huh, uh–huh.

Category: DEPENDENCY (indicated linguistically by ‘ve got to) Common ground: discourse context, communal knowledge Information: inferable (elliptical/ anaphorical) Given/New: presented as New: prominence on anything Listener’s reaction: no comment Corpus example 9.

(Context: Puerto Rico as a state of the US or independent)

A: So, if they’ve just taken such ACTION, it would seem to indicate to ME either they’re doing it because they’re afraid they MIGHT become a state and want to declare this BEFORE they become a state, or maybe because they don’t WANT to become a state for fear of losing the Spanish, or Hispanic HERITAGE. B: Uh–huh, well that’s interesting.

Category: REGULATION (indicated by prosodic prominence on before) Common ground: discourse context Information: discourse-old Given/New: the information in both the main and temporal clauses is Given: declare this points back to taken such action, while become a state is an exact repetition of the wording used before. The temporal term itself is presented as New, as indicated by prominence. Listener’s reaction: no comment on the temporal component Corpus example 10.

(Context: Money / credit)

B.: (…) and then these kids start CHARGING and, and they get themselves into some deep TROUBLE. (…) BEFORE they even [laughter] start making MONEY, they’re already in debt, you know. A.: No kidding.

Category: REGULATION supported by the linguistic indicators even and already that point to the peculiarity of the temporal order indicated by the speaker Common ground: communal knowledge Information: discourse-new

292

Appendix

Given/New: prosodic prominence on money; weaker prominence on before to further support the importance of the temporal ordering Listener’s reaction: no comment on the temporal component Corpus example 11.

(Context: ice storms)

B.: when you get, I mean, here they start even to forecast the weather, you know, it’s going to be an ice storm tomorrow, they’ll come, before the ten clock news they’ve already got schools calling in saying they’ve canceled. A.: Really?

Category: REGULATION (indicated by already) Common ground: communal knowledge on news Information: discourse-new Given/New: weak prominence on before, prominence on news: important information on something that happens even earlier than the (known) ten o’clock news (definite article) Listener’s reaction: no comment Corpus example 12.

(Context: cross-stitching)

A: Uh–huh [throat_clearing]. There’s also a couple of large projects, BABY afghans, and I got ONE done and I think the other one’s gonna the child is gonna be in HIGH school before I get – B: [Laughter] On that, uh –

Category: Indeterminate INSERTION Common ground: discourse context Information: discourse-old Given/New: the expected event (probably: finishing the project) is so obvious that it is not even expressed by the speaker Listener’s reaction: this idea invokes laughter on the part of the hearer: the project will take much too long to make sense, i.e., the inserted event renders the project meaningless Corpus example 13.

(Context: Reading newspapers)

A.: And I live in kind of a bad area where if I have the paper delivered it’s stolen before I can get out and get it [laughter]. B.: Gosh that’s, that sounds pretty bad to me. I mean who would steal a newspaper?

Category: Counterfactual INSERTION Common ground: communal knowledge (papers get delivered outside)

Corpus examples from the Switchboard corpus

293

Information: inferable from the discourse context; anaphoric (it) and elliptical construction (get out) Given/New: presented as Given, weak prosodic prominence on the first get Listener’s reaction: no comment on the temporal component Corpus example 14.

(Context: Credit cards)

A.: Do you use credit cards a good bit? (…) B.: Before I got married, uh, about the only credit cards I used were gas cards, because I didn’t like carrying, you know – A.: Uh–huh.

Category: TERMINATION Common ground: world knowledge Information: discourse-new Given/New: prosodic prominence on married Listener’s reaction: no comment Corpus example 15.

(Context: movies)

A.: Because that, that just reminds me of, you know, like back in school and, you know, the girlfriends just taking off for a day [laughter] or something. (…) Just getting in trouble and probably not that much trouble but, (…) Before we found the shopping malls that’s [laughter] what we did. B.: Uh–huh.

Category: TERMINATION Common ground: communal knowledge Information: discourse-new Given/New: presented as Given, definite article; weak prominence on before Listener’s reaction: no comment

294

Appendix

App. 2. Web study (Chapter 6): Configurations

S1(C1)

S2(C1)

S3(C1)

S4(C1)

S5(C1)

S6(C1)

S7(C1)

S1(C2)

S2(C2)

S3(C2)

S4(C2)

S5(C2)

S6(C2)

S7(C2)

Web study (Chapter 6): Configurations

S1(C3A) (German)

S1(C3A) (English)

S3(C3A)

S4(C3A)

S6(C3A)

S1(C3B)

Figure App.1.

S2(C3A)

S5(C3A)

S7(C3A)

S2(C3B)

S3(C3B)

S4(C3B)

S5(C3B)

S6(C3B)

S7(C3B)

Scenarios (S 1–7) in the web study: Conditions 1 – 3B

295

296

Appendix

App. 3. Web study (Chapter 6): Coding schemes locative-dimensional

REFERENCE

OTHER-

other-spatial TYPE

not-locative

KIND

in-between-relation distance comparative-height comparative-size compass-directions clock-directions deictic other-spatial-desc

class-name sequential-order discourse-history other-feature

only-goal

no-alternative alternative-offered no-scene-description SCENE scene-description-offered correct CORRECTDESCRIPTION incorrect ALTERNATIVE

utt

STRAT

path-plus-goal only-path denial

Figure App.2.

“Paradigmatic” analysis: Choice of linguistic strategy.

Web study (Chapter 6): Coding schemes

297

adj-goalobject adjective

ADJECT IVET YPE

adj-other adj-goalobj-omitted definite-article

noun-in-pp

DET ERMINER

SYNT ACT ICFORM

possessive-pronoun det-missing

handedness-term adverb preposition-group indeterminate-form unmodified second-dim-term middle

MODIFICAT ION

counting superlative

loc-dim

modified

MODIFICAT IONT YPE

precisification complex-description other-mods mod-distance mod-deictic mod-sequential

lateral one-axis

ONEAXIS-T YPE

front

FRONT T YPE

static dynamic

back

AXIS

vertical two-axes

Figure App.3.

frontal

DIRECT ION

T WOAXES-T YPE

lateral-frontal lateral-vertical

Coding scheme for locative dimensional terms (“syntagmatic” analysis): Lexical options

298

Appendix

implicit-perspective PERSPECT IVE

addressee explicit-perspective

]

PERSPECT IVET YPE

speaker picture

loc-dim

implicit rel-addressee rel-speaker

RELAT UM

rel-picture rel-landmark explicit

EXPLICIT RELAT UM

group-class-name group

GROUPT YPE

detailed-info figure

center corner Figure App.4.

Coding scheme for locative dimensional terms (“syntagmatic” analysis): Conceptual elements

Relative frequencies: Paradigmatic / German App. 4: Cond. C1 N

S1 113

S2 106

S3 128

S4 129

S5 132

S6 29

S7 All - C1 29 666

only goal 100.0 100.0 100.0 100.0 100.0 100.0 100.0 path plus goal only path denial only goal: new N locative dimensional other than dimensional other spatial class name sequential order discourse history other feature

113 97.3 2.7

106 93.4 6.6

128 97.7 2.3

129 83.7 16.3

1.8

6.6

0.8 1.6

13.2 3.1

299

132 29 29 98.5 100.0 100.0 1.5

100.0

666 94.6 5.4 2.7 2.6

1.5

0.9

0.2

in-between relation distance comparative height comparative size compass directions clock directions deictic other spatial description

12.4

2.4

0.8

0.8

0.3

no alternative alternative offered

98.2 1.8

98.1 100.0 1.9 0.0

96.9 3.1

99.2 100.0 100.0 0.8

98.6 1.4

no scene description scene description offered new N: only goal / path plus goal

97.3 2.7

96.2 3.8

95.3 4.7

99.2 100.0 100.0 0.8

97.7 2.3

113

106

128

129

132

29

29

666

informative class name object shape mentioned no mention object type no class name uninformative class name no colour mentioned colour mentioned

42.5

44.5

48.3

48.3

55.5 43.0 12.5 97.7 2.3

53.5 1.6 45.0 32.6 12.4 97.7 2.3

41.7

57.5 46.0 11.5 97.3 2.7

46.2 0.9 52.8 45.3 7.5 99.1 0.9

58.3 51.7 42.4 41.4 15.9 10.3 98.5 100.0 1.5

51.7 41.4 10.3 96.6 3.4

45.9 0.5 53.6 41.6 12.0 98.0 2.0

App. 4: Cond. C1

S1

S2

S3

S4

S5

99.2 0.8

S6

S7 All - C1

300

Appendix App. 4: Cond. C2 N

S1 105

S2 103

only goal 100.0 100.0 path plus goal only path denial

S3 126

S4 143

S5 142

S6 23

S7 All - C2 30 672

99.2 100.0 100.0 100.0 100.0

0.8

99.9

0.1

only goal: new N 105 locative dimensional 100.0 other than dimensional

103 98.1 1.9

125 88.8 11.2

143 98.6 1.4

142 50.0 50.0

23 73.9 26.1

30 80.0 20.0

671 84.9 15.1

other spatial class name sequential order discourse history other feature

1.0 1.0

11.2

1.4

49.3

26.1

16.7

14.6 0.1 0.1 0.1

in-between relation distance comparative height comparative size compass directions clock directions deictic other spatial description

1.0

3.3 0.7

9.6

0.7

0.8 0.8

39.4 5.6

26.1

16.7

8.5 4.8

0.7

0.1 0.3

0.7

3.5

0.9

no alternative alternative offered

97.1 2.9

99.0 1.0

96.8 3.2

94.4 5.6

92.3 7.7

95.7 4.3

96.7 3.3

95.7 4.3

no scene description scene description offered new N: only goal / path plus goal

96.2 3.8

96.1 100.0 3.9

98.6 1.4

97.2 100.0 2.8

96.7 3.3

97.8 2.2

105

103

125

143

142

30

671

informative class name object shape mentioned no mention object type no class name uninformative class name no colour mentioned colour mentioned

42.9

46.6

40.8

49.0 39.4 52.2 26.7 0.7 50.3 60.6 47.8 73.3 34.3 39.4 39.1 53.3 16.1 21.1 8.7 20.0 99.3 100.0 100.0 100.0 0.7

43.2 0.1 56.6 38.0 18.6 99.7 0.3

App. 4: Cond. C2

S1

57.1 53.4 59.2 38.1 35.0 39.2 19.0 18.4 20.0 99.0 100.0 100.0 1.0 S2

S3

S4

S5

23

S6

S7 All - C2

Relative frequencies: Paradigmatic / German

301

App. 4: Cond. C3A N

S1 71

S2 78

S3 99

S4 85

S5 106

S6 18

S7 24

only goal path plus goal only path denial

84.5 8.5 7.0

91 3.8 5.1

72.7 17.2 10.1

90.6 5.9 3.5

84.9 7.5 7.5

72.2 27.8 0.0

87.5 8.3 4.2

only goal: new N locative dimensional other than dimensional

60 95.0 5.0

71 98.6 1.4

72 94.4 5.6

77 77.9 22.1

90 97.8 2.2

13 92.3 7.7

21 85.7 14.3

other spatial class name sequential order discourse history other feature

3.3

5.6

2.6 18.2

2.2

7.7

14.3

1.3

1.1

7.7

14.3

1.3

1.1

in-between relation distance comparative height comparative size compass directions clock directions deictic other spatial description

1.7

1.4 1.3

3.3

4.2

1.4

no alternative alternative offered

93.3 6.7

97.2 2.8

95.8 4.2

96.1 3.9

96.7 3.3

100.0 0.0

100.0 0.0

no scene description scene description offered new N: only goal / path plus goal

95.0 5.0

91.5 8.5

97.2 2.8

98.7 1.3

98.9 1.1

100.0 0.0

100.0 0.0

66

74

89

82

98

18

23

informative class name object shape mentioned no mention object type no class name uninformative class name no colour mentioned colour mentioned

36.4 1.5 62.1 9.1 53.0 100.0 0.0

44.6 2.7 52.7 6.8 45.9 100.0 0.0

40.4 1.1 58.4 9.0 49.4 98.9 1.1

52.4 11.0 36.6 6.1 30.5 97.6 2.4

42.9

38.9

52.2

57.1 10.2 46.9 99.0 1.0

61.1 11.1 50.0 100.0 0.0

47.8 8.7 39.1 100.0 0.0

App. 4: Cond. C3A

S1

S2

S3

S4

S5

S6

S7

302

Appendix App. 4: Cond. 3B N

S1 78

S2 80

S3 S4 108 108

S5 92

S6 24

only goal 80 path plus goal 19.2 only path 1.3 denial

81.3 11.3 7.5

83.3 82.4 7.4 13.0 9.3 4.6

78.3 17.4 4.3

79.2 20.8 0.0

78.3 17.4 4.3

82.4 11.8 5.8

92.5 5.0 2.5

only goal: new N 62 locative dimensional 61.3 other than dimensional 38.7

65 69.2 30.8

90 89 91.1 30.3 8.9 69.7

72 77.8 22.2

19 78.9 21.1

18 88.9 11.1

819 79.6 20.4

2,156 85.9 14.1

other spatial 25.8 class name sequential order 12.9 discourse history other feature

29.2 1.5

8.9 69.7

22.2

15.8

5.6

17.0 1.8 1.3

11.8 1.5 0.6 0.1 0.1

in-between relation distance 25.8 comparative height comparative size compass directions clock directions deictic other spatial descr. no alternative 82.3 alternative offered 17.7

5.6 5.3

29.2

App. 4: Cond. 3B

32.6 7.8 33.7

92.3 7.7

4.7 6.3 0.1

0.2

0.2

0.6

0.5

94.4 5.6

92.3 7.7

95.3 4.7

98.9 96.6 100.0 100.0 100.0 1.1 3.4

96.9 3.1

97.4 2.6

22

936

2,273

45.6 37.5 50.0 45.5 1.0 1.1 53.4 61.4 50.0 54.5 8.7 6.8 12.5 13.6 44.7 54.5 37.5 40.9 99.0 100.0 100.0 100.0 1.0

42.8 1.8 55.3 7.7 47.6 99.4 0.6

43.9 0.9 55.2 26.6 28.6 99.1 0.9

S2

98 103

S3

22.2

15.8

5.6

3.4

81.1 91.0 18.9 9.0

40.3 39.2 42.9 1.3 1.0 58.4 60.8 56.1 6.5 5.4 4.1 51.9 55.4 52.0 98.7 100.0 100.0 1.3 S1

0.2 3.5 12.6

1.1

no scene description 87.1 100.0 scene descr. offered 12.9 new N: only goal / path plus goal 77 74 informative class name object shape ment. no mention object type no class name uninformat. class name no colour mentioned colour mentioned

S7 All - C3 All Ger. 23 994 2,332

S4

94.4 5.6

88

S5

89.5 10.5

24

S6

S7 All - C3 All Ger.

Relative frequencies: Syntagmatic / German App. 5: Cond. C1 N loc. dim.: new N adjective noun-in-pp handedness-term adverb preposition-group adj-other unmodified modified second dimensional term middle counting dimensional superlative precisification complex description modification-distance modification-deictic modification-sequential new N: loc.-dim. and adjective adj-goalobject adj-other adj-goalobj-omitted unmodified adjective modified adjective App. 5: Cond. C1

303

S1 113 110 62.7

S2 106 99 71.7

S3 128 125 57.6

S4 129 108 57.4

S5 132 130 51.5

S6 29 29 82.8

S7 29 29 79.3

All - C1 666 630 61.6

37.3

28.3

42.4

42.6

48.5

17.2

17.2 3.4

1.8 67.3 32.7

3.0 75.8 24.2

47.2 52.8

5.6 60.2 39.8

1.5 20.8 79.2

37.9 62.1

62.1 37.9

38.3 0.2 2.1 52.2 47.8

1.0 1.0 2.0 19.2 1.0

1.9 21.3 1.9 1.9 5.6 1.9 5.6

48.5

1.8 5.5 1.8 20.0

25.4 2.3 2.3 0.8

55.2 6.9

31.0 6.9

62 43.5 9.7 46.8 83.9 16.1 S4

67 46.3 3.0 50.7 35.8 64.2 S5

24 58.3

23 56.5

41.7 33.3 66.7 S6

43.5 60.9 39.1 S7

50.4

3.6

69 40.6 2.9 56.5 91.3 8.7 S1

0.8 0.8

0.8

71 40.8 4.2 54.9 94.4 5.6 S2

72 40.3 59.7 76.4 23.6 S3

20.3 4.1 1.6 10.3 8.6 1.0 1.9

388 44.1 3.4 52.6 72.9 27.1 All - C1

304

Appendix App. 5: Cond. C2 N loc. dim.: new N adjective noun-in-pp handedness-term adverb preposition-group adj-other unmodified modified

second dimensional term middle counting dimensional superlative precisification complex description modification-distance modification-deictic modification-sequential new N: loc.-dim. and adjective adj-goalobject adj-other adj-goalobj-omitted unmodified adjective modified adjective App. 5: Cond. C2

S1 105 105 56.2 1.0 41.9 1.0 1.9 68.6 31.4 8.6

S2 103 101 55.4

4.0 74.3 25.7

67.6 32.4

S4 143 141 54.6 3.5 0.7 40.4 0.7 5.7 68.8 31.2

7.9

26.1

9.2

5.6 22.5

1.8 0.9 3.6

3.5 12.1 2.1 3.5 0.7

2.8 25.4 7.0 5.6

35.3 17.6

77 46.8 10.4 42.9 85.7 14.3 S4

15 46.7 13.3 40.0 53.3 46.7 S5

13 53.8

19 36.8

46.2 53.8 46.2 S6

63.2 42.1 57.9 S7

44.6

S3 126 111 49.5 1.8 0.9 45.0 2.7

S5 142 71 21.1

S6 23 17 76.5

S7 30 24 79.2

59.2 19.7 2.8 31.0 69.0

11.8 11.8

12.5 8.3

41.2 58.8

37.5 62.5

2.9 16.2 1.0 2.9

2.0 13.9 1.0 1.0

59 42.4 3.4 54.2 91.5 8.5 S1

56 42.9 7.1 50.0 91.1 8.9 S2

55 47.3 52.7 92.7 7.3 S3

4.2 45.8 12.5

5.9

All - C2 672 570 51.6 1.2 0.5 42.6 4.0 2.8 62.6 37.4 11.1 2.8 0.7 4.6 13.0 1.9 3.2 0.2

294 44.9 5.4 49.7 83.3 16.7 All - C2

Relative frequencies: Syntagmatic / German App. 5: Cond. C3A N loc. dim.: new N adjective noun-in-pp handedness-term adverb preposition-group adj-other unmodified modified second dimensional term middle counting dimensional superlative precisification complex description modification-distance modification-deictic modification-sequential new N: loc.-dim. and adjective adj-goalobject adj-other adj-goalobj-omitted unmodified adjective modified adjective App. 5: Cond. C3A

S1 71 57 38.6 1.8

S2 78 70 31.4 4.3

S3 99 68 8.8 5.9

S4 85 60 18.3 6.7

S5 106 88 26.1 2.3

S6 18 12 16.7

S7 24 18 16.7

57.9 1.8

77.9 7.4

8.3 75.0

11.1 72.2

66.2 33.8

73.3 1.7 5.0 58.3 41.7

71.6

54.4 45.6

62.9 1.4 2.9 58.6 41.4

45.5 54.5

16.7 83.3

27.8 72.2

5.3

8.6

14.7

18.3

17.0

1.8 3.5 28.1 1.8 3.5

1.4 22.9 4.3 2.9

8.3 16.7 50.0 8.3

11.1 5.6 44.4 11.1

1.8

1.4

22 86.4 13.6 86.4 13.6 S1

14.7 2.9 1.5

5.0 10.0 8.3

25.0 5.7 6.8

22 6 86.4 83.3 9.1 4.5 16.7 90.9 100.0 9.1 0.0 S2 S3

11 63.6 27.3 9.1 90.9 9.1 S4

23 2 78.3 100.0

3 66.7

21.7 78.3 21.7 100.0 S5 S6

33.3 66.7 33.3 S7

305

306

Appendix

App. 5: Cond. C3B N loc. dim.: new N adjective noun-in-pp handedness-term adverb preposition-group adj-other unmodified modified

S1 78 38 44.7 2.6

S2 80 45 15.6 2.2

S3 108 82 13.4 1.2

S4 108 27 22.2

S5 92 56 10.7

S6 24 15 33.3

52.6

26.8 58.5 2.4 63.4 36.6

40.7 37.0

89.3

13.2 15.8 84.2

75.6 6.7 8.9 31.1 68.9

20.0 46.7 6.7 53.3 46.7

second dim. term middle counting dim. superlative precisification compl. description mod.-distance modification-deictic mod.-sequential new N: loc.-dim. and adjective adj-goalobject adj-other adj-goalobj-omitted unmod. adjective modified adjective App. 5: Cond. C3B

5.3

13.3

7.3

5.3 18.4 26.3 10.5 5.3

11.1 13.3 20.0 11.1

26.8 2.4

22.2 77.8

25.0 7.4 3.7 3.7 29.6 29.6 3.7

13.2 17 52.9 29.4 17.6 11.8 88.2 S1

1.8 35.7 64.3

S7 All - C3 All Ger. 23 994 2,332 16 652 1,852 12.5 21.9 44.5 2.6 1.3 0.2 81.3 60.3 47.4 6.3 15.2 6.6 2.8 2.5 31.3 47.5 53.8 68.8 52.5 46.2

14.3 14.3 7.1 3.6

46.7

56.3 6.3 6.3

7 11 28.6 63.6 57.1 18.2 14.3 18.2 42.9 100.0 57.1 S2 S3

6 83.3

11.2 0.3 1.8 2.1 23.2 7.8 4.4 0.3 1.2

14.3 2.4 1.4 5.7 15.1 3.7 3.2 0.2 0.4

6 5 2 143 825 50.0 40.0 50.0 70.6 49.0 16.7 20.0 12.6 5.7 16.7 33.3 40.0 50.0 16.8 45.3 83.3 100.0 100.0 100.0 76.2 77.2 16.7 23.8 22.8 S4 S5 S6 S7 All - C3 All Ger.

Relative frequencies: Syntagmatic / German App. 5: Cond. C1 S1 S2 S3 S4 S5 S6 S7 loc. dim.: new N 110 99 125 108 130 29 29 implicit perspective 100.0 100.0 100.0 100.0 100.0 100.0 100.0 explicit perspective perspective-y perspective-x one axis 100.0 100.0 48.0 97.2 49.2 100.0 100.0 two axes 52.0 2.8 50.8 lateral 100.0 100.0 12.8 91.7 1.5 frontal 1.5 3.4 3.4 vertical 35.2 5.6 46.2 96.6 96.6 lateral-frontal 1.6 0.9 lateral-vertical 50.4 1.9 50.8 front 1.5 3.4 back 3.4 static 1.5 3.4 dynamic implicit relatum 93.6 88.9 96.0 89.8 96.2 96.6 93.1 explicit relatum 6.4 11.1 4.0 10.2 3.8 3.4 6.9 relatum y relatum x relatum picture 0.9 0.8 1.9 1.5 relatum landmark 6.1 2.8 3.4 relatum group 5.5 5.1 3.2 5.6 1.5 3.4 3.4 corner 0.8 roup: class name or omitted 2.7 2.0 3.2 3.7 1.5 3.4 3.4 detailed information 2.0 1.9 figure description 2.7 1.0 App. 5: Cond. C1 S1 S2 S3 S4 S5 S6 S7

307

All - C1 630 100.0

78.7 21.3 51.7 0.6 26.3 0.5 20.8 0.5 0.2 0.5 93.3 6.7

1.0 1.6 4.0 0.2 2.7 0.6 0.6 All - C1

308

Appendix

App. 5: Cond. C2 loc. dim.: new N implicit perspective explicit perspective perspective-y perspective-x one axis two axes lateral frontal vertical lateral-frontal lateral-vertical front back static dynamic implicit relatum explicit relatum relatum y relatum x relatum picture relatum landmark relatum group corner roup: class name or omitted detailed information figure description App. 5: Cond. C2

S1 105 93.3 6.7

S2 101 93.1 6.9

S3 111 94.6 5.4

S4 141 95.0 5.0

S5 S6 S7 71 17 24 94.4 100.0 100.0 5.6

All - C2 570 94.6 5.4

6.7 90.5 9.5 89.5 1.0

6.9 91.1 8.9 91.1

5.4 73.0 27.0 72.1 0.9

5.7 3.8 1.0

5.0 4.0

16.2 10.8 0.9

5.0 89.4 10.6 80.9 0.7 7.8 7.1 3.5 0.7

5.6 94.4 100.0 100.0 5.6 19.7 73.2 76.5 87.5 1.4 23.5 12.5 5.6

0.9

0.7

76.5

4.2 83.3 4.2

5.4 88.1 11.9 69.1 15.6 3.3 7.5 4.4 8.2 7.4 6.0 2.3 74.9 25.1

1.0

42.3 31.0 23.9 18.3 67.6 32.4

76.5

82.4 17.6

83.3 16.7

78.1 21.9

77.2 22.8

69.4 30.6

76.6 23.4

12.4 1.0

10.9

30.6

22.0 0.7

31.0

11.8

4.2

0.7

1.4

5.9

4.2 8.3

5.9

4.2

0.7

1.4

S4

S5

S6

4.2 S7

8.6 2.9 5.7 S1

7.9 4.0 1.0 2.0 1.0 S2

S3

20.0 0.4 1.6 3.2 1.1 0.7 1.4 All - C2

Relative frequencies: Syntagmatic / German App. 5: Cond. C3A loc. dim.: new N implicit perspective explicit perspective perspective-y perspective-x one axis two axes lateral frontal vertical lateral-frontal lateral-vertical front back static dynamic implicit relatum explicit relatum relatum y relatum x relatum picture relatum landmark relatum group corner roup: class name or omitted detailed information figure description App. 5: Cond. C3A

S1 57 75.4 24.6 22.8 1.8 94.7 5.3 93.0 1.8

S2 70 68.6 31.4 30.4 1.4 87.1 12.9 85.7 1.4

5.3

12.9

1.8

1.4

77.2 22.8 19.3

54.3 45.7 34.3

3.5

5.7 5.7

1.8

5.7

S3 68 91.2 8.8 8.8

S4 60 85.0 15.0 15.0

82.4 17.6 77.9 4.4

73.3 26.7 71.7 1.7

14.7 2.9 2.9 1.5 2.9

26.7

19.1 80.9 77.9 2.9

1.7

S5 S6 S7 88 12 18 65.9 91.7 100.0 34.1 8.3 31.8 8.3 2.3 80.7 100.0 94.4 19.3 5.6 79.5 11.1 1.1 100.0 83.3 17.0 2.3 1.1 100.0

1.7 36.7 63.3 60.0 3.3

100.0 1.1 55.7 44.3 40.9

5.6 83.3

25.0 75.0 75.0

77.8 5.6 27.8 72.2 72.2

S6

S7

3.4

3.4 1.8 S1

S2

S3

S4

S5

309

310

Appendix

App. 5: Cond. C3B loc. dim.: new N implicit persp. explicit persp. perspective-y perspective-x one axis two axes lateral frontal vertical lateral-frontal lateral-vertical front back static dynamic implicit relatum explicit relatum relatum y relatum x relatum picture relatum landmark relatum group corner class n. / omitted detailed information figure description App. 5: Cond. C3B

S1 38 55.3 44.7 36.8 7.9 84.2 15.8 55.3 28.9

S2 45 84.4 15.6 13.3 2.2 75.6 24.4 64.4 11.1

15.8

24.4 11.1

28.9 11.1 60.5 39.5 23.7 5.3

26.7 73.3 66.7 4.4

S3 82 86.6 13.4 13.4

S4 27 77.8 22.2 22.2

90.2 9.8 30.5 58.5 1.2 9.8

88.9 11.1 14.8 74.1

3.7 54.9 2.4 1.2 23.2 76.8 20.7 54.9

74.1

11.1

66.7 7.4 37.0 63.0 59.3

2.2 10.5

1.2

3.7

5.3 2.6 2.6 S1

1.2

3.7

S3

S4

S2

S5 S6 S7 All - C3 All Ger. 56 15 16 652 1,852 58.9 93.3 81.3 77.0 90.2 41.1 6.7 18.8 23.0 9.8 39.3 18.8 21.4 7.6 1.8 6.7 1.5 2.2 60.7 100.0 100.0 83.4 83.3 39.3 16.6 16.7 60.7 33.3 87.5 63.3 61.2 66.7 12.5 19.9 12.0 0.2 10.0 39.3 15.8 8.0 0.8 8.7 66.7 10.6 6.4 12.5 9.4 5.6 53.3 9.5 5.3 13.3 1.1 1.1 53.6 46.7 68.8 43.9 70.2 46.4 53.3 31.3 56.1 29.8 30.4 40.0 18.8 42.9 15.1 14.3 13.3 9.7 9.6 0.4 12.5 1.1 1.4 1.8 2.5 3.2 0.1 1.4 1.7 1.8 0.8 0.7 0.3 0.8 S5 S6 S7 All - C3 All Ger.

Relative frequencies: Paradigmatic / English App. 6: Cond. C1 N

S3 90

S4 88

S5 83

All - C1 417

only goal 100.0 100.0 100.0 path plus goal only path denial

98.9

100.0

99.8

only goal: new N locative dimensional other than dimensional other spatial class name sequential order other feature

S1 79

S2 77

311

79 91.1 8.9

77 88.3 11.7

8.9

9.1 2.6

1.1

0.2

90 92.2 7.8

87 77.0 23.0

83 96.4 3.6

416 88.9 11.1

6.7 1.1

13.8 8.0

2.4 1.2

4.8 3.8 2.2 0.2

1.2

2.4

1.2

1.4 0.2 0.5

1.1

in-between relation distance comparative height comparative size compass directions clock directions deictic other spatial description

10.3 5.6 1.1

1.1 1.1

1.1

no alternative 100.0 100.0 alternative offered 0.0 0.0

98.9 1.1

0.2

98.9 1.1

100.0 0.0

99.5 0.5

no scene description 100.0 100.0 100.0 100.0 scene description offered 0.0 0.0 0.0 0.0 new N: only goal and path plus goal 79 77 90 87

98.8 1.2

99.8 0.2

83

4

informative class name object shape mentioned no mention object type no class name uninformative class name no colour mentioned colour mentioned

65.8

71.4

71.1

68.7

34.2 27.8 6.3 97.5 2.5

28.6 23.4 5.2 96.1 3.9

28.9 25.6 3.3 97.8 2.2

70.1 1.1 28.7 23.0 5.7 96.6 3.4

31.3 26.5 4.8 97.6 2.4

69.5 0.2 30.3 25.2 5.0 97.1 2.9

App. 6: Cond. C1

S1

S2

S3

S4

S5

All - C1

312

Appendix App. 6: Cond. C2 N

S1 81

S2 89

S3 89

S4 83

S5 83

All - C2 425

only goal path plus goal only path denial

100.0

98.9 1.1

98.9

100.0

98.8 1.2

99.3 0.5

only goal: new N locative dimensional other than dimensional

81 96.3 3.7

88 95.5 4.5

88 79.5 20.5

83 92.8 7.2

82 58.5 41.5

422 84.6 15.4

other spatial class name sequential order other feature

2.5

3.4

20.5

7.2

41.5

14.9

1.2

1.1

in-between relation distance comparative height comparative size compass directions clock directions deictic other spatial description

1.1

2.5

3.4

0.2

0.5

25.6 12.2

5.0 5.2 0.7

2.4

2.4

1.2

1.2

2.8 0.5 0.2 0.5

12.5 1.1

1.2 2.4

3.4 2.3 1.1

no alternative alternative offered

98.8 1.2

97.7 2.3

95.5 4.5

97.6 2.4

93.9 6.1

96.7 3.3

no scene description scene description offered new N: only goal and path plus goal

100.0

100.0

98.9 1.1

98.8 1.2

98.8 1.2

99.3 0.7

81

89

88

83

83

424

informative class name object shape mentioned no mention object type no class name uninformative class name no colour mentioned colour mentioned

65.4

61.8

60.2

65.1

60.2

62.5

34.6 28.4 6.2 97.5 2.5

38.2 30.3 7.9 97.8 2.2

39.8 31.8 8.0 96.6 3.4

34.9 30.1 4.8 97.6 2.4

39.8 34.9 4.8 97.6 2.4

37.5 31.1 6.4 97.4 2.6

App. 6: Cond. C2

S1

S2

S3

S4

S5

All - C2

Relative frequencies: Paradigmatic / English

313

App. 6: Cond. C3A N

S1 45

S2 57

S3 64

S4 62

S5 76

only goal path plus goal only path denial

53.3 22.2 24.4

49.1 29.8 21.1

57.8 29.7 12.5

51.6 27.4 21.0

46.1 26.3 27.6

only goal: new N locative dimensional other than dimensional

24 70.8 29.2

28 92.9 7.1

37 94.6 5.4

32 71.9 28.1

35 97.1 2.9

other spatial class name sequential order other feature

29.2

3.6

5.4

3.1 25.0

2.9

in-between relation distance comparative height comparative size compass directions clock directions deictic other spatial description

25.0 4.2

no alternative alternative offered

83.3 16.7

92.9 7.1

89.2 10.8

93.8 6.3

97.1 2.9

no scene description scene description offered new N: only goal and path plus goal

100.0 0.0

100.0 0.0

100.0 0.0

96.9 3.1

94.3 5.7

34

35

56

49

55

informative class name object shape mentioned no mention object type no class name uninformative class name no colour mentioned colour mentioned

64.7

68.9

66.1

35.3 17.6 17.6 94.1 5.9

31.1 17.8 13.3 95.6 4.4

33.9 21.4 12.5 96.4 3.6

63.3 10.2 26.5 16.3 10.2 98.0 2.0

67.3 1.8 30.9 12.7 18.2 90.9 9.1

App. 6: Cond. C3A

S1

S2

S3

S4

S5

3.6

3.6

2.7

2.9 3.1

2.7

314

Appendix App. 6: Cond. 3B N

S1 65

S2 61

only goal path plus goal only path denial

40 49.2 10.8

50.8 24.6 24.6

60.8 49 18.9 26.3 18.9 24.6 1.4

35.8 34.3 29.9

49.5 28.7 21.6 0.2

78.0 12.5 9.3 0.2

only goal: new N locative dimensional other than dimensional

26 38.5 61.5

31 41.9 58.1

45 28 82.2 53.6 17.8 46.4

24 87.5 12.5

316 74.4 25.6

1,154 83.4 16.6

other spatial class name sequential order other feature

53.8

58.1

17.8 46.4

12.5

22.2 2.5 0.9

13.3 2.1 1.2 0.1

54.8

7.1 11.1 39.3

12.5

2.5 17.1

0.6 0.6 0.3 0.9

3.4 6.6 0.8 0.1 1.4 0.3 0.2 0.5

91.7 8.3

92.1 7.9

96.4 3.6

no scene description 100.0 100.0 100.0 96.4 100.0 scene description offered 0.0 0.0 0.0 3.6 0.0 new N: only goal and path plus goal 58 46 59 43 47

98.7 1.3

99.3 0.7

499

927

informative class name object shape mentioned no mention object type no class name uninformative class name no colour mentioned colour mentioned

63.8

64.3 1.6 34.1 14.8 19.2 95.6 4.4

65.3 0.7 34.0 23.2 10.8 96.6 3.4

App. 6: Cond. 3B

S5 All - C3

All Eng.

in-between relation distance comparative height comparative size compass directions clock directions deictic other spatial description no alternative alternative offered

S3 74

S4 57

S5 All - C3 67 628

7.7

53.8

3.2 96.2 100.0 3.8 0.0

2.2 2.2 2.2 84.4 92.9 15.6 7.1

57.6 1.7 40.7 10.2 30.5 96.6 3.4

69.8

70.2

36.2 17.2 19.0 96.6 3.4

52.2 2.2 45.7 15.2 30.4 95.7 4.3

30.2 11.6 18.6 95.3 4.7

29.8 8.5 21.3 95.7 4.3

S1

S2

S3

S4

All Eng. 1,470

Relative frequencies: Syntagmatic / English App. 7: Cond. C1 N locative dimensional: new N adjective noun-in-pp handedness-term adverb preposition-group indeterminate-form adj-other unmodified modified second dimensional term middle counting dimensional superlative precisification complex description other modifications modification-distance modification-sequential new N: loc.-dim. and adjective adj-goalobject adj-other adj-goalobj-omitted unmodified adjective modified adjective

S1 79 72 45.8 43.1 5.6

S2 77 68 36.8 38.2 11.8 1.5

S3 90 83 74.7 15.7

5.6 5.6 41.7 58.3

11.8 2.9 61.8 38.2 1.5 1.5 1.5 17.6

9.6 9.6 48.2 51.8 44.6 1.2

26.4 4.2

16.2

33 72.7 12.1 15.2 24.2 75.8

25 76.0 8.0 16.0 40.0 60.0

App. 7: Cond. C1

S1

S2

1.4 1.4 25.0

S4 88 67 31.3 44.8 10.4 3.0 10.4 4.5 50.7 49.3

3.6

20.9 4.5 10.4

2.4

7.5

S5 83 80 63.8 18.8 3.8 1.3 1.3 11.3 6.3 45.0 55.0 35.0 1.3 17.5

315

All - C1 417 370 51.9 31.1 5.9 0.5 0.8 9.7 5.9 49.2 50.8 18.1 4.6 1.6 14.6 1.9

4.5

1.3

9.2 0.8

62 83.9 12.9 3.2 48.4 51.6

21 81.0 14.3 4.8 42.9 57.1

51 88.2 9.8 2.0 47.1 52.9

192 81.8 11.5 6.8 42.2 57.8

S3

S4

S5

All - C1

316

Appendix

App. 7: Cond. C2 N locative dimensional: new N adjective noun-in-pp handedness-term adverb preposition-group indeterminate-form adj-other unmodified modified second dimensional term middle counting dimensional superlative precisification complex description other modifications modification-distance modification-sequential new N: loc.-dim. and adjective adj-goalobject adj-other adj-goalobj-omitted unmodified adjective modified adjective

S1 81 78 29.5 50.0 3.8 5.1 1.3 10.3 5.1 47.4 52.6 11.5 1.3 1.3 15.4 2.6 2.6 1.3 16.7

S2 89 84 27.4 50.0 3.6 2.4 4.8 11.9 2.4 64.3 35.7 10.7 1.2

S3 89 70 18.6 42.9 4.3 15.7 10.0 8.6 5.7 60.0 40.0 30.0

S4 83 77 32.5 45.5 3.9 3.9 6.5 7.8 3.9 58.4 41.6 10.4

13.1 2.4 2.4

1.4 4.3 4.3

15.6 2.6 6.5

23 73.9 17.4 8.7 13.0 87.0

23 78.3 8.7 13.0 26.1 73.9

13 69.2 30.8 53.8 46.2

25 72.0 12.0 16.0 36.0 64.0

App. 7: Cond. C2

S1

S2

S3

S4

6.0

6.5

S5 All - C2 83 425 48 357 18.8 26.1 18.8 43.4 2.1 3.6 37.5 10.6 16.7 7.0 6.3 9.2 4.2 4.2 29.2 53.8 70.8 46.2 2.1 13.4 8.3 1.7 0.3 4.2 10.6 22.9 5.6 14.6 5.3 0.3 18.8 9.0

09 66.7 22.2 11.1 11.1 88.9

93 73.1 16.1 10.8 28.0 72.0

S5 All - C2

Relative frequencies: Syntagmatic / English App. 7: Cond. C3A N locative dimensional: new N adjective noun-in-pp handedness-term adverb preposition-group indeterminate-form adj-other unmodified modified second dimensional term middle counting dimensional superlative precisification complex description other modifications modification-distance modification-sequential new N: loc.-dim. and adjective adj-goalobject adj-other adj-goalobj-omitted unmodified adjective modified adjective App. 7: Cond. C3A

S1 45 17 5.9

S2 57 26 23.1 76.9

11.8 82.4

17.6 82.4

S3 64 35

S4 62 23

88.6

78.3

2.9 8.6

73.1 26.9

82.9 17.1

69.6 30.4 8.7

11.4 5.7

13.0 8.7

11.5

6 100.0

35.3

0

0

16.7 83.3 S1

S2

2.9 2.9 52.9 47.1

2.9 8.8

15.4

0

S5 76 34 14.7 79.4 2.9

17.4 4.3

5.9 76.5

317

5 80.0 20.0 20.0 80.0

S3

S4

S5

318

Appendix App. 7: Cond. C3B S1 S2 S3 S4 S5 All - C3 All Eng. N 65 61 74 57 67 628 1,470 loc. dim.: new N 10 13 37 15 21 231 958 adjective 20.0 7.7 5.4 13.3 14.3 9.1 31.9 noun-in-pp 70.0 69.2 16.2 26.7 76.2 60.2 42.7 handedness-term 2.7 0.4 3.8 adverb 7.7 5.4 6.7 3.5 5.0 preposition-group 10.0 7.7 70.3 53.3 4.8 25.1 9.0 indeterminate-form 7.7 4.8 1.7 7.6 adj-other 0.4 4.0 unmodified 30.0 7.7 59.5 6.7 38.1 51.9 51.6 modified 70.0 92.3 40.5 93.3 61.9 48.1 48.4 second dimensional term 9.5 1.7 12.4 middle 2.4 counting 10.0 30.8 6.7 3.5 1.6 dimensional superlative 10.0 7.7 5.4 6.7 9.5 6.1 11.1 precisification 15.4 29.7 33.3 14.3 17.7 6.4 complex description 10.0 38.5 5.4 40.0 9.5 8.7 4.8 other modifications 0.1 modification-distance 30.0 6.7 19.0 10.0 9.3 modification-sequential 10.0 0.4 0.4 new N: loc.-dim. and adjective 2 1 2 2 3 21 306 adj-goalobject 100.0 100.0 100.0 100.0 100.0 95.2 80.1 adj-other 4.8 12.4 adj-goalobj-omitted 7.5 unmodified adjective 9.5 35.6 modified adjective 100.0 100.0 100.0 100.0 100.0 90.5 64.4 App. 7: Cond. C3B

S1

S2

S3

S4

S5 All - C3 All Eng.

Relative frequencies: Syntagmatic / English App. 7: Cond. C1 locative dimensional: new N implicit perspective explicit perspective perspective-y perspective-x perspective-picture one axis two axes lateral frontal vertical lateral-frontal lateral-vertical front back static dynamic implicit relatum explicit relatum relatum y relatum x relatum picture relatum landmark relatum group center corner group: class name or omitted detailed information App. 7: Cond. C1

S1 72 100.0

319

S2 68 98.5 1.5

S3 83 98.8 1.2

S4 67 98.5 1.5

S5 80 98.8 1.3

All - C1 370 98.9 1.1

1.5 98.5 1.5 98.5

1.2 51.8 48.2 6.0

1.5 94.0 6.0 86.6

1.3 65.0 35.0 2.5

1.1 80.3 19.7 55.1

45.8

7.5

62.5

25.1

1.5

48.2

6.0

35.0

19.7

95.8 4.2

85.3 14.7

84.3 15.7

82.1 17.9

82.5 17.5

85.9 14.1

4.2

4.4 8.8 1.5

2.4 2.4

6.0 4.5 4.5 3.0

12.5

5.9 3.0 1.1 0.8 3.2 0.3 0.8

100.0 100.0

1.2 9.6 1.5 S1

S2

5.0 1.5 3.0

S3

S4

S5

All - C1

320

Appendix

App. 7: Cond. C2 locative dimensional: new N implicit perspective explicit perspective perspective-y perspective-x perspective-picture one axis two axes lateral frontal vertical lateral-frontal lateral-vertical front back static dynamic implicit relatum explicit relatum relatum y relatum x relatum picture relatum landmark relatum group center corner group: class name or omitted detailed information App. 7: Cond. C2

S1 78 96.2 3.8

S2 84 97.6 2.4

S3 70 91.4 8.6

S4 77 97.4 2.6

3.8

2.4

8.6

2.6

8.3

4.8

87.2 12.8 87.2

86.9 13.1 85.7

3.8 9.0

1.2 3.6 9.5

65.7 34.3 52.9 2.9 10.0 7.1 27.1 2.9

83.1 16.9 66.2 1.3 15.6 7.8 9.1 1.3

2.9

1.3

93.8 6.3 39.6 50.0 4.2 2.1 4.2 45.8 4.2 22.9 22.9 58.3 41.7

82.9 17.1 69.2 7.6 6.2 5.0 12.0 7.0 0.6 3.9 3.1 66.4 33.6

37.5 4.2

29.1 2.0 1.7 0.6

75.6 24.4

70.2 29.8

54.3 45.7

68.8 31.2

21.8 1.3

23.8 1.2 3.6 1.2

42.9 1.4 1.4

24.7 2.6 2.6 1.3

1.3 1.2 S1

S2

S3

S5 All - C2 48 357 91.7 95.2 8.3 4.8

1.3

0.3 0.6

S4

S5 All - C2

Relative frequencies: Syntagmatic / English App. 7: Cond. C3A locative dimensional: new N implicit perspective explicit perspective perspective-y perspective-x perspective-picture one axis two axes lateral frontal vertical lateral-frontal lateral-vertical front back static dynamic implicit relatum explicit relatum relatum y relatum x relatum picture relatum landmark relatum group center corner group: class name or omitted detailed information App. 7: Cond. C3A

321

S1 17 94.1 5.9 5.9

S2 26 96.2 3.8 3.8

S3 35 97.1 2.9 2.9

S4 23 95.7 4.3 4.3

S5 34 100.0

100.0

100.0 100.0

82.6 17.4 82.6

100.0

5.9 94.1

94.3 5.7 85.7 5.7 2.9 5.7

100.0

13.0 4.3

94.1 5.7 82.4 11.8 0.0 100.0 100.0

26.9 73.1 65.4

2.9 97.1 94.3

7.7

2.9

4.3 95.7 95.7

17.6 82.4 79.4

2.9

2.9 S1

S2

S3

S4

S5

322

Appendix

App. 7: Cond. C3B S1 S2 S3 S4 S5 All - C3 All Eng. loc. dim.: new N 10 13 37 15 21 231 958 implicit perspective 100.0 100.0 100.0 100.0 100.0 98.3 97.4 explicit perspective 1.7 2.6 perspective-y 1.7 0.4 perspective-x 1.8 perspective-picture 0.4 one axis 90.0 76.9 94.6 86.7 90.5 93.1 84.3 two axes 10.0 23.1 5.4 13.3 9.5 6.9 15.7 lateral 90.0 76.9 18.9 40.0 90.5 69.7 63.9 frontal 73.0 46.7 22.5 8.2 vertical 2.7 0.9 12.2 lateral-frontal 10.0 23.1 5.4 13.3 4.8 6.1 3.3 lateral-vertical 4.8 0.9 12.3 front 5.4 46.7 10.8 5.2 back 67.6 11.7 3.0 static 2.7 40.7 9.1 3.7 dynamic 2.7 6.7 1.7 1.6 implicit relatum 20.0 15.4 5.4 6.7 19.0 11.3 60.6 explicit relatum 80.0 84.6 94.6 93.3 81.0 88.7 39.4 relatum y 50.0 84.6 24.3 93.3 76.2 74.0 17.8 relatum x 30.0 70.3 4.8 13.0 14.0 relatum picture 3.0 relatum landmark 1.3 2.0 relatum group 0.4 0.7 center 0.3 corner 1.4 group: class name or omitted 0.3 detailed information 0.4 0.4 App. 7: Cond. C3B

S1

S2

S3

S4

S5 All - C3 All Eng.

References

Anscombe, Gertrude E.M. 1964 Before and after. Philosophical Review 73, 3–24. Asher, Nicholas 1993 Reference to Abstract Objects in Discourse. Dordrecht: Kluwer Academic Publishers. Asher, Nicholas and Alex Lascarides 2003 Logics of Conversation. Cambridge University Press. Aurnague, Michel and Laure Vieu 1993 A three-level approach to the semantics of space. In C. ZelinskyWibbelt (ed.), The Semantics of Prepositions. From Mental Processing to Natural Language Processing, 393–439. Berlin: Mouton de Gruyter. Avrahami, Judith and Yaakov Kareev 1994 The emergence of events. Cognition 53, 239–261. Barkowsky, Thomas and Christian Freksa 1997 Cognitive requirements on making and interpreting maps. In S. Hirtle and A. Frank (eds.), Spatial Information Theory: A Theoretical Basis for GIS, 347–361. Berlin: Springer. Bateman, John 2001 Between the leaves of rhetorical structure: Static and dynamic aspects of discourse organisation. Verbum 23(1), 31-58. Bateman, John and Klaas Jan Rondhuis 1997 Coherence relations: Towards a general specification. Discourse Processes 24, 3–49. Beaver, David and Cleo Condoravdi 2003 A uniform analysis of before and after. In R. Young and Y. Zhou (eds.), Proceedings of SALT XIII, 37–54. Ithaca: CLC Publications. Bennett, Brandon 2002 Physical objects, identity and vagueness. In D. Fensel, D. McGuinness, and M.A. Williams (eds.), Principles of Knowledge Representation and Reasoning: Proceedings of the Eighth International Conference (KR2002), San Francisco, CA: Morgan Kaufmann. Beun, Robbert-Jan and Anita Cremers 2001 Multimodal reference to objects: An empirical approach. In H. Bunt and R.-J. Beun (eds.), Cooperative Multimodal Communication, 64– 84. Berlin/Heidelberg: Springer.

324

References



Blühdorn, Hardarik 2002 Semantische Variation bei Temporalkonnektoren. Am Beispiel von nachdem und bevor. Kolloquium “Brücken schlagen”. Zur Semantik der Konnektoren. Mannheim. Bras, Myriam, Anne LeDraoulec, and Laure Vieu 2001 French adverbial puis between temporal structure and discourse structure. In M. Bras and L. Vieu (eds.), Semantic and Pragmatic Issues in Discourse and Dialogue: Experimenting with Current Dynamic Theories, 109–146. Amsterdam: Elsevier. Brown-Schmidt, Sarah and Michael K. Tanenhaus 2003 Referential domains and the interpretation of referring expressions in interactive conversation. Proc. DiaBruck, 7th Workshop on the Semantics and Pragmatics of Dialogue, Sept. 4th–6th 2003, Wallerfangen, 15–19. Brugman, Claudia 1988 The Story of OVER: Polysemy, Semantics and the Structure of the Lexicon. New York: Garland Press. Bryant, David J., Barbara Tversky, and Margaret Lanca 2000 Retrieving spatial relations from observation and memory. In E. van der Zee and U. Nikanne (eds.), Cognitive Interfaces: Constraints on Linking Cognitive Information, 94–115. Oxford University Press. Buhl, Heike M., Steffi Katz, Karin Schweizer and Theo Herrmann 2000 Einflüsse des Wissenserwerbs auf die Linearisierung beim Sprechen über räumliche Anordnungen. Zeitschrift für Experimentelle Psychologie 47 (1), 17–33. Carletta, Jean, Amy Isard, Stephen Isard, Jacqueline Kowtko, Gwyneth DohertySneddon, and Anne Anderson 1997 The reliability of a dialogue structure coding scheme. Computational Linguistics 23 (1), 13–31. Carlson, Laura A. 1999 Selecting a reference frame. Spatial Cognition and Computation 1 (4), 365–379. Carlson, Laura, Terry Regier, and Eric Covey 2003 Defining spatial relations: Reconciling axis and vector representations. In E. van der Zee and J. Slack (eds.), Representing Direction in Language and Space, 111–131. Oxford University Press. Carlson, Laura and Emile van der Zee (eds.) 2005 Functional Features in Language and Space: Insights from Perception, Categorization and Development. Oxford University Press. Carlson-Radvansky, Laura A., Eric Covey, and K.M. Lattanzi 1999 “What” effects on “where”: Functional influences on spatial relations. Psychological Science 10, 516–521.



References

325

Carlson-Radvansky, Laura A. and D.E. Irwin 1993 Frames of reference in vision and language: Where is above? Cognition 46, 223–244. Carlson-Radvansky, Laura A. and Gordon D. Logan 1997 The influence of reference frame selection on spatial template construction. Journal of Memory and Language 37, 411–437. Carlson-Radvansky, Laura A. and Gabriel A. Radvansky 1996 The influence of functional relations on spatial term selection. Psychological Science 7, 56–60. Carroll, Mary 1993 Deictic and intrinsic orientation in spatial descriptions: A comparison between English and German. In J. Altarriba (ed.), Cognition and Culture: A Cross-Cultural Approach to Psychology, 23–44. Amsterdam: Elsevier. 1997 Changing place in English and German: Language-specific preferences in the conceptualization of spatial relations. In J. Nuyts and E. Pederson (eds.), Language and Conceptualization, 137–161. Cambridge University Press. Carston, Robin 2002 Thoughts and Utterances: The Pragmatics of Explicit Communication. Oxford: Blackwell Publishers. Clark, Herbert H. 1973 Space, time, semantics, and the child. In Moore, T.E. (ed.), Cognitive Development and the Acquisition of Language, 28–64. New York: Academic Press. 1996 Using Language. Cambridge University Press. Clark, Herbert and Wilkes-Gibbs, Deanna 1986 Referring as a collaborative process. Cognition 22, 1–39. Comrie, Bernard 1976 Aspect. Cambridge University Press. Coventry, Kenny R. and Simon C. Garrod 2004 Saying, Seeing and Acting: The Psychological Semantics of Spatial Prepositions. Hove/New York: Psychology Press. 2005 Towards a classification of extra-geometric influences on the comprehension of spatial prepositions. In L. A. Carlson and E. van der Zee (eds.), Functional Features in Language and Space, 149–162. Oxford University Press. Coventry, Kenny R. and Gayna Mather 2002 The real story of over? In Kenny R. Coventry and Patrick Olivier (eds.), Spatial Language: Cognitive and Computational Perspectives, 165–184. Dordrecht: Kluwer Academic Publishers.

326

References



Coventry, Kenny R., Merce Prat-Sala, and Richards, Lynn V. 2001 The interplay between geometry and function in the comprehension of ‘over’, ‘under’, ‘above’ and ‘below’. Journal of Memory and Language 44, 376–398. Coventry, Kenny R., Thora Tenbrink, and John Bateman (eds.) (forthc.) Spatial Language and Dialogue. Oxford University Press. Daniel, Marie-Paule and Michel Denis 1998 Spatial descriptions as navigational aids: A cognitive analysis of route directions. Kognitionswissenschaft 7, 45–52. Davidson, Donald 1967 Causal relations. Journal of Philosophy 64, 691–703. van Deemter, Kees and Roger Kibble (eds.) 2002 Information Sharing: Reference and Presupposition in Language Generation and Interpretation. Stanford: CSLI Publications. Degand, Liesbeth 2000 Prepositional causatives or causal connectives? Discursive constraints. Journal of Pragmatics 32 (6), 687–707. Demetras, Marty 1989 Working parents’ conversational responses to their two-year-old sons. Technical Report. University of Arizona. Denis, Michel 1997 The description of routes: A cognitive approach to the production of spatial discourse. Cahiers de Psychologie Cognitive 16 (4), 409– 458. Dowty, David R. 1986 The effect of aspectual class on the temporal structure of discourse: Semantics or pragmatics? Linguistics and Philosophy 1 (9), 37–62. Dunlea, Anne 1989 Vision and the Emergence of Meaning: Blind and Sighted Children’s Early Language. Cambridge University Press. Duwe, Ingo, Klaus Kessler, and Hans Strohner 2002 Resolving ambiguous descriptions through visual information. In K.R. Coventry and P. Olivier (eds.), Spatial Language: Cognitive and Computational Perspectives, 43–67. Dordrecht: Kluwer Academic Publishers. Ehrich, Veronika 1985 Zur Linguistik und Psycholinguistik der sekundären Raumdeixis. In Schweizer, H. (Hrsg.). Sprache und Raum: Ein Arbeitsbuch für das Lehren von Forschung, 130–161. Stuttgart: Metzler. 1992 Wann ist jetzt? Anmerkungen zum adverbialen Zeitlexikon des Deutschen. Kognitionswissenschaft 2, 119–135.



References

327

Emmorey, Karen and Shannon Casey 2002 Gesture, thought, and spatial language. In K.R. Coventry and P. Olivier (eds.), Spatial Language: Cognitive and Computational Aspects, 87–101. Dordrecht: Kluwer Academic Publishers. Eschenbach, Carola 2005 Contextual, functional, and geometric features and projective terms. In L. Carlson and E. van der Zee (eds.), Functional Features in Language and Space: Insights from Perception, Categorization, and Development, 71–91. Oxford University Press. 1999 Geometric structures of frames of reference and natural language semantics. Spatial Cognition and Computation 1 (4), 329–348. Eschenbach, Carola, Christopher Habel, and Annette Leßmöllmann 2001 Multiple frames of reference in interpreting complex projective terms. In K.R. Coventry and P. Olivier (eds.), Spatial Language: Cognitive and Computational Perspectives, 209–231. Dordrecht: Kluwer Academic Publishers. Eschenbach, Carola and Lars Kulik 1997 An axiomatic approach to the spatial relations underlying left–right and in front of–behind. In G. Brewka, C. Habel and B. Nebel (eds.), KI–97: Advances in Artificial Intelligence, 207–218. Berlin: Springer. Evans, Vyvyan 2004 The Structure of Time: Language, Meaning and Temporal Cognition. Amsterdam/Philadelphia: John Benjamin Publishing Company. Fischer, Kerstin 2003 Linguistic methods for investigating concepts in use. In Stolz, Th. and Kolbe, K. (eds.), Methodologie in der Linguistik, 39–62. Frankfurt a.M.: Lang. 2006 What Computer Talk Is and Is not: Human–Computer Conversation as Intercultural Communication. (Linguistics – Computational Linguistics 17). Saarbrücken: AQ. Fischer, Kerstin and Reinhard Moratz 2001 From communicative strategies to cognitive modelling. Workshop Epigenetic Robotics, Lund. Franklin, Nancy, Linda A. Henkel, and Thomas Zangas 1995 Parsing surrounding space into regions. Memory and Cognition 23, 397–407. Freksa, Christian 1980 Communication about visual patterns by means of fuzzy characterizations. XXIInd Intern. Congress of Psychology, Leipzig, July 1980. 1997 Spatial and temporal structures in cognitive processes. In C. Freksa, M. Jantzen, R. Valk (eds.), Foundations of Computer Science, 379– 387. Berlin: Springer.

328

References

1999



Links vor – Prototyp oder Gebiet? Probabilistische und possibilistische Raumbeschreibung. In G. Rickheit (Hg), Richtungen im Raum, 231–246. Wiesbaden: Westdeutscher Verlag. Friederici, Angela and Willem J.M. Levelt 1990 Spatial reference in weightlessness: Perceptual factors and mental representations. Perception and Psychophysics 47 (3), 253–66. Friedman, William J. 1990 About Time. Cambridge: MIT Press. Gapp, Klaus-Peter 1995 An empirically validated model for computing spatial relations. In I. Wachsmuth, C.R. Rollinger, and W. Brauer (eds.), KI–95: Advances in Artificial Intelligence. 19th Annual German Conference on Artificial Intelligence, 245–256. Berlin/Heidelberg: Springer. Garnham, Alan 1999 What’s in a mental model? In G. Rickheit and C. Habel (eds.), Mental Models in Discourse Processing and Reasoning, 41–56. New York: Elsevier. Garrod, Simon and Anne Anderson 1987 Saying what you mean in dialogue: A study in conceptual and semantic coordination. Cognition 27, 181–218. Garrod, Simon, Gillian Ferrier, and Siobhan Campbell 1999 In and on: investigating the functional geometry of spatial prepositions. Cognition 72, 167–189. Garrod, Simon C. and Anthony J. Sanford 1988 Discourse models as interfaces between language and the spatial world. Journal of Semantics 6, 147–160. Gibson, James J. 1979 The Ecological Approach to Visual Perception. Boston: Houghton Mifflin. Glasbey, Sheila R. 1993 Distinguishing between events and times: Some evidence from the semantics of then. Natural Language Semantics 1, 285–312. Gorniak, Peter and Deb Roy 2004 Grounded semantic composition for visual scenes. Journal of Artificial Intelligence Research 21, 429–470. Grabowski, Joachim 1998 Ein psychologisch–anthropologisches Modell der einheitlichen semantischen Beschreibung dimensionaler Präpositionen. In P. Ludewig and B. Geurts (Hg.), Lexikalische Semantik aus kognitiver Sicht, 11–40. Tübingen: Narr.

References

329

Grabski, Michael and Manfred Stede 2004 Intra-clausal coherence relations illustrated with a German preposition. AG Konnektorensemantik, 26. Jahrestagung der Deutschen Gesellschaft für Sprachwissenschaft, Mainz, 25. – 27. Februar 2004. Grady, Joseph 1997 Foundations of meaning: primary metaphors and primary scenes. PhD dissertation, Dept. of Linguistics, UC Berkeley. Grice, Herbert P. 1975 Logic and conversation. In Cole, P. and Morgan, J. (eds.), Syntax and semantics. Volume 3, 41–58. New York/San Francisco/London. Grote, Brigitte 2003 Signaling coherence relations in text generation: A case study of german temporal discourse markers. Dissertation, Universität Bremen, Germany. Guyau, Jean-Marie 1988 La genèse de l’idée de temps. Transl. by J.A. Michon, V. Pouthas, and C. Greenbaum, The origin of the idea of time by Jean-Marie Guyau. In J.A. Michon (ed.), Guyau and the Idea of Time, 37–148. Amsterdam: North-Holland Publishing Company. Habel, Christopher 1988 Prozedurale Aspekte der Wegplanung und Wegbeschreibung. In H. Schnelle and G. Rickheit (eds.), Sprache in Mensch und Computer, 107–133. Opladen: Westdeutscher Verlag. 1990 Propositional and depictorial representations of spatial knowledge: The case of path-concepts. In R. Studer (ed.), Natural Language and Logic, 94–117. Berlin: Springer. 1994 Discreteness, finiteness, and the structure of topological spaces. In C. Eschenbach, C. Habel and B. Smith (eds.), Topological Foundations of Cognitive Science, 81–90. Papers from the Workshop at the FISI-CS, Buffalo, NY. July 9–10, 1994. 1995 Representing space and time: Discrete, dense or continuous? – Is that the question? In C. Eschenbach and W. Heydrich (eds.), Parts and Wholes. Integrity and Granularity, 97–107. Graduiertenkolleg Kognitionswissenschaft Hamburg, Report 49, August 1995. Habel, Christopher and Carola Eschenbach 1997 Abstract structures in spatial cognition. In C. Freksa, M. Jantzen and R. Valk (eds). Foundations of Computer Science – Potential – Theory – Cognition, 369–378. Berlin: Springer. Habel, Christopher, Michael Herweg und Simone Pribbenow 1993 Wissen über Raum und Zeit. In G. Görz (Hrsg.): Einführung in die künstliche Intelligenz, 139–204. Bonn: Addison-Wesley.

330

References



Habel, Christopher and Heike Tappe 1999 Processes of segmentation and linearization in describing events. In v. Stutterheim, C. and R. Klabunde (eds.), Processes in Language Production, 117–153. Opladen: Westdeutscher Verlag. Halliday, Michael A.K. 1994 An Introduction to Functional Grammar. 2nd ed. London: Edward Arnold. Original Edition, London/Baltimore: Edward Arnold, 1985. Halliday, Michael A.K. and Ruqaiya Hasan 1976 Cohesion in English. London/New York: Longman. Halliday, Michael A.K. and Christian M.I.M. Matthiessen 1999 Construing Experience: A Language-Based Approach to Cognition. London/New York: Continuum. Haspelmath, Martin 1997 From Space to Time. Temporal Adverbials in the World’s Languages. München: Lincom. Hayward, William G., and Michael J. Tarr 1995 Spatial language and spatial representation. Cognition 55, 39–84. Heinämäki, Orvokki 1974 Semantics of english temporal connectives. Dissertation, Dept. of Linguistics, University of Texas, Austin. Herrmann, Theo 1990 Vor, hinter, rechts und links: das 6H-Modell. Psychologische Studien zum sprachlichen Lokalisieren. Zeitschrift für Literaturwissenschaft und Linguistik 78, 117–140. Herrmann, Theo und Werner Deutsch 1976 Psychologie der Objektbenennung. Bern u.a.: Hans Huber Verlag. Herrmann, Theo und Joachim Grabowski 1994 Sprechen. Psychologie der Sprachproduktion. Heidelberg: Spektrum. Herskovits, Annette 1986 Language and Spatial Cognition. Cambridge University Press. Herweg, Michael 1991 Temporale Konjunktionen und Aspekt. Der sprachliche Ausdruck von Zeitrelationen zwischen Situationen. Kognitionswissenschaft 2: 51–90. Hill, Clifford A. 1982 Up/down, front/back, left/right. A contrastive study of Hausa and English. In J. Weissenborn and W. Klein (eds.), Here and There. Cross-linguistic Studies on Deixis and Demonstration, 13–42. Amsterdam: John Benjamin Publishing Company.



References

331

Hobbs, Jerry R. 1985 Granularity. In A.K. Joshi (ed.), Proceedings of the 9th International Joint Conference on Artificial Intelligence, 432–435. Los Angeles, CA/San Francisco: Morgan Kaufmann. Jackendoff, Ray 1983 Semantics and Cognition. Cambridge: MIT Press. Jekat-Rommel, Susanne 1994 Zeitkonzept und Zeitreferenz. Eine Untersuchung zum bilingualen Erstspracherwerb (französisch/deutsch). Frankfurt/Main: Peter Lang. Jin, Yu, Sabine Gillner and Hanspeter Mallot 2004 Study of eye movements in landmark recognition: An experiment in virtual reality. In T. Barkowsky, C. Freksa, M. Knauff, B. KriegBrückner, and B. Nebel (eds.), Spatial Cognition 2004: Poster Presentations, 11–14. SFB/TR 8 Report No. 002-10/2004, Bremen/Freiburg. Johnston, Judith R. 1984 Acquisition of locative meanings: Behind and in front of. Journal of Child Language 11, 407–22. Kamp, Hans and Uwe Reyle 1993 From Discourse to Logic: Introduction to Modeltheoretic Semantis of Natural Language, Formal Logic and Discourse Representation Theory. Dordrecht: Kluwer Academic Publishers. Kessler, Klaus, Ingo Duwe und Hans Strohner 1996 Sprachliche Objektidentifikation in ambigen Situationen: Empirische Befunde. Report 96/1, SFB 360, Bielefeld. 1999 Grounding mental models: Subconceptual dynamics in the resolution of reference in discourse. In G. Rickheit and C. Habel (eds.), Mental Models in Discourse Processing and Reasoning, 169–193. Amsterdam: Elsevier. Klabunde, Ralf 1999 Logic-based choice of projective terms. In W. Burgard, T. Christaller, A.B. Cremers (eds.), KI–99, LNAI 1701, 149–158. Berlin/Heidelberg: Springer. Klein, Wolfgang 1991 Raumausdrücke. Linguistische Berichte 132, 77–114. 1994 Time in Language. London: Routledge. Klippel, Alexander 2003 Wayfinding choremes – Conceptualizing wayfinding and route direction elements. Monograph Series of the SFB/TR 8 Spatial Cognition, Vol. I. ISBN 3-88722-590-2.

332

References



Klippel, Alexander, Carsten Dewey, Markus Knauff, Kai-Florian Richter, Dan R. Montello, Christian Freksa, and Esther-Anna Loeliger 2004 Direction concepts in wayfinding assistance systems. In J. Baus, C. Kray, R. Porzel (eds.), Workshop on Artificial Intelligence in Mobile Systems (AIMS’04), SFB 378 Memo 84, 1–8. Saarbrücken: Universität des Saarlandes. Klippel, Alexander, Thora Tenbrink, and Dan Montello (in press) The role of structure and function in the conceptualization of directions. In E. van der Zee and M. Vulchanova (eds.), Motion Encoding in Language and Space. Oxford University Press. Knott, Alistair 1996 A data-driven methodology for motivating a set of coherence relations. Doctoral dissertation, University of Edinburgh. Knott, Alistair and Robert Dale 1994 Using linguistic phenomena to motivate a set of rhetorical relations. Discourse Processes 18 (1), 35–62. Kopp, Stefan, Paul Tepper, and Justine Cassell 2004 Towards integrated microplanning of language and iconic gesture for multimodal output. Proceedings of the International Conference on Multimodal Interfaces (ICMI’04), 97–104, ACM Press, 2004. Krause, Olaf 2002 Progressiv im Deutschen. Eine empirische Untersuchung im Kontrast mit Niederländisch und Englisch. Tübingen: Niemeyer. Krauss, Robert M. and Susan R. Fussell 1991 Perspective-taking in communication: representations of others’ knowledge in reference. Social Cognition 9, 2–24. Kray, Christian 2003 Situated interaction on spatial topics. Dissertation, Fachbereich Informatik, Universität Saarbrücken. Kühnlein, Peter and Jens Stegmann 2003 Empirical issues in deictic gesture: referring to objects in simple identification tasks. Report 2003/3, SFB 360, Univ. Bielefeld. Lakoff, George 1987 Women, Fire, and Dangerous Things: What Categories Reveal About the Mind. Chicago University Press. Lakoff, George and Mark Johnson 1980 Metaphors we Live by. Chicago University Press. Landau, Barbara 2003 Axes and direction in spatial language and spatial cognition. In E. van der Zee and J. Slack (eds.), Representing Direction in Language and Space, 18–38. Oxford University Press.



References

333

Landau, Barbara and Ray Jackendoff 1993 “What” and “where” in spatial language and spatial cognition. Behavioral and Brain Sciences 16, 217–265. Landragin, Frédéric and Laurent Romary 2003 Referring to objects through sub-contexts in multimodal humancomputer interaction. Proc. DiaBruck, 7th Workshop on the Semantics and Pragmatics of Dialogue, Sept. 4–6, Wallerfangen, 67–74. Lang, Ewald 1989 The semantics of dimensional designation of spatial objects. In M. Bierwisch and E. Lang (eds.), Dimensional Adjectives: Grammatical Structure and Conceptual Interpretation, 263–417. Berlin: Springer. Langacker, Ronald 1986 An introduction to cognitive grammar. Cognitive Science 10, 1–40. 1987 Foundations of Cognitive Grammar. Vol I: Theoretical Prerequisites. California: Stanford University Press. 1991 Foundations of Cognitive Grammar. Vol II: Descriptive Application. California: Stanford University Press. Lascarides, Alex and Nicholas Asher 1993 Temporal interpretation, discourse relations and common sense entailment. Linguistics and Philosophy 16, 437–493. Lascarides, Alex and Jon Oberlander 1993 Temporal connectives in a discourse context. In Proceedings of the European Chapter of the Association for Computational Linguistics (EACL93), 260–268. April 21–23, Utrecht, The Netherlands. Levelt, Willem J.M. 1989 Speaking. Cambridge, MA: MIT Press. 1996 Perspective taking and ellipsis in spatial descriptions. In P. Bloom, M.A. Peterson, L. Nadel and M.F. Garrett (eds.), Language and Space, 77–107. Cambridge, MA: MIT Press. Levinson, Stephen C. 2000 Presumptive Meanings: The Theory of Generalized Conversational Implicature. Cambridge, MA: MIT Press. 2003 Space in Language and Cognition. Cambridge University Press. Lewis, David 1979 Scorekeeping in a language game. Journal of Philosophical Logic 8, 339–359. Ligozat, Gérard 2000 From language to motion, and back: Generating and using route descriptions. In D.N. Christodoulakis (ed.), NLP 2000, LNCS 1835, 328–345. Berlin/Heidelberg: Springer. Linde, Charlotte and William Labov 1975 Spatial networks as a site for the study of language and thought. Language 51, 924–939.

334

References



Mackie, John L. 1965 Causes and conditions. American Philosophical Quarterly 2, 245– 64. MacWhinney, Brian 2000 The CHILDES Project: Tools for Analyzing Talk. 3rd edition. Hillsdale, NJ: Lawrence Erlbaum. Mainwaring, Scott D., Barbara Tversky, Motoko Ohgishi, and Diane J. Schiano 2003 Descriptions of simple spatial scenes in English and Japanese. Spatial Cognition and Computation 3 (1), 3–42. Marr, David 1982 Vision: A Computational Investigation into the Human Representation and Processing of Visual Information. New York: Freeman. Martin, James R. 1992 English Text: System and Structure. Amsterdam: John Benjamin Publishing Company. McIntyre, Andrew 2001 Functional interpretations: Borderline idiomaticity in prepositional phrases, complex verbs and other expressions. Ms., available at www.uni-leipzig.de/~angling/mcintyre/ (last visited: May 4, 2005). Michon, John A. 1988 Guyau’s idea of time: a cognitive view. In J.A. Michon, V. Pouthas and J. L. Jackson (eds.), Guyau and the Idea of Time, 162–197. Amsterdam: North-Holland. Michon, Pierre-Emmanuel and Michel Denis 2001 When and why are visual landmarks used in giving directions? In D.R. Montello (ed.): Spatial Information Theory: Foundations of Geographic Information Science, 292–305. Berlin/Heidelberg: Springer. Miller, George A. and P.N. Johnson-Laird 1976 Language and Perception. Cambridge University Press. Montello, Dan R. 1993 Scale and multiple phychologies of space. In Frank, A. and Campari, I. (eds.), Spatial Information Theory: A Theoretical Basis for GIS, 312–321. Berlin/Heidelberg/New York: Springer. Moratz, Reinhard and Thora Tenbrink 2006 Spatial reference in linguistic human–robot interaction: Iterative, empirically supported development of a model of projective relations. Spatial Cognition and Computation. 6 / 1, 63–106. Moratz, Reinhard and Kerstin Fischer 2000 Cognitively adequate modelling of spatial reference in human–robot interaction. Proceedings of the 12th IEEE International Conference on Tools with Artificial Intelligence, ICTAI 2000, November 13–15 2000, Vancouver, Canada. 222–228.



References

335

Nemo, François 1999 The pragmatics of signs, the semantics of relevance, and the semantic/pragmatic interface. In Turner, K. (ed.), The Semantics–Pragmatics Interface from Different Points of View, CRiSPI Series, 343– 417. Amsterdam: Elsevier. Nikanne, Urpo 2003 How Finnish postpositions see the axis system. In E. van der Zee and J. Slack (eds.), Representing Direction in Language and Space, 191–208. Oxford University Press. Oxford English Dictionary (OED) 1961 Oxford English Dictionary. First edition. Oxford University Press. Pasch, Renate, Ursula Brauße, Eva Breindl und Ulrich Herrmann Waßner 2003 Handbuch der deutschen Konnektoren. Berlin: de Gruyter. Pederson, Eric 2003 How many reference frames? In C. Freksa, W. Brauer, C. Habel, and K.F. Wender (eds.), Spatial Cognition III: Routes and Navigation, Human Memory and Learning, Spatial Representation and Spatial Learning, 287–304. Berlin: Springer. Petkov, Pavel 1979 Die Temporalleistung der Konjunktionen im Deutschen. In H. Weydt (ed.): Die Partikeln der deutschen Sprache, 215–222. Berlin/New York: de Gruyter. Piaget, Jean 1946 Le développement de la notion de temps chez l’enfant. Paris: PUF. Piaget, Jean and Bärbel Inhelder 1956 The Child’s Conception of Space. London: Routledge and Kegan Paul. Pickering, Martin and Simon Garrod 2004 Toward a mechanistic psychology of dialogue. Behavioral and Brain Sciences 27, 169–226. Pobel, Rupert, Christiane Grosser, Roland Mangold, and Theo Herrmann 1988 Zum Einfluß hörerseitiger Wahrnehmungsbedingungen auf die Überspezifikation von Objektbenennungen. Arbeiten der Forschergruppe “Sprechen und Sprachverstehen im sozialen Kontext”. Heidelberg/Mannheim, Bericht Nr. 17. Presson, Clark C., and Montello, Dan R. 1988 Points of reference in spatial cognition: Stalking the elusive landmark. British Journal of Developmental Psychology 6, 378–381. Pribbenow, Simone 1992 Zur Verarbeitung von Lokalisierungsausdrücken in einem hybriden System. IWBS-Report 211, April 1992, IBM.

336

References



Radden, Günter 1997 Time is space. In Birgit Smieja und Meike Tasch (eds.), Human Contact Through Language and Linguistics, 147–166. Frankfurt/Main: Peter Lang. Regier, Terry, Laura Carlson, and Bryce Corrigan 2005 Attention in spatial language: Bridging geometry and function. In Carlson, L. and E. van der Zee (eds.), Functional Features in Language and Space: Insights from Perception, Categorization and Development, 191–204. Oxford University Press. Reichenbach, Hans 1956 The Direction of Time. Berkeley: University of California Press. Reips, Ulf-Dietrich 2002 Theory and techniques of web experimenting. In B. Batinic, U.-D. Reips, and M. Bosnjak (eds.), Online Social Sciences, 229–250. Seattle: Hogrefe and Huber. Retz-Schmidt, Gudula 1988 Various views on spatial prepositions. AI Magazine 9 (2), 95–105. Reyle, Uwe 1993 Dealing with ambiguities by underspecification: Construction, representation, and deduction. Journal of Semantics 10 (2), 123–179. Rieser, Hannes 1997 Repräsentations-Metonymie, Perspektive und Koordination in aufgabenorientierten Dialogen. In: Carla Umbach, Michael Grabski und Robin Hörnig (Hg), Perspektive in Sprache und Raum, 1–26. Wiesbaden: Deutscher Universitäts-Verlag. Rooth, Mats 1992 A theory of focus interpretation. Natural Language Semantics 1, 75– 116. Rosch, Eleanor 1975 Cognitive reference points. Cognitive Psychology 7, 532–547. 1978 Principles of categorization. In E. Rosch and B.B. Lloyd (eds.), Cognition and Categorization, 27–48. Hillsdale, NJ: Lawrence Erlbaum. Rossdeutscher, Antje und Christiane von Stutterheim 2004 Semantische und pragmatische Prinzipien der Positionierung von dann. AG Konnektorensemantik, 26. DGfS Jahrestagung, Mainz. Schilder, Frank 2001 Presupposition triggered by temporal connectives. In M. Bras and L. Vieu (eds.), Semantic and Pragmatic Issues in Discourse and Dialogue: Experimenting with CurrentDynamic Theories, 85–108. Amsterdam: Elsevier. 2002 Temporale Konnektoren im Diskurs. Kolloquium “Brücken schlagen: Zur Semantik der Konnektoren”, IdS Mannheim, 6.–7.12.2002.



References

337

Schilder, Frank and Thora Tenbrink 2001 Before and after: sentence-internal and -external discourse relations. Workshop From Sentence Processing to Discourse Interpretation: Crossing the Borders. Utrecht (The Netherlands), 2–3 July 2001. Schmidt, Thomas 2005 Datenarchive für die Gesprächsforschung: Perspektiven, Probleme und Lösungsansätze. Gesprächsforschung – Online-Zeitschrift zur verbalen Interaktion (ISSN 1617-1837) 6, 103–126. Schober, Michael F. 1993 Spatial perspective-taking in conversation, Cognition 47, 1–24. 1998a Different kinds of conversational perspective-taking. In S.R. Fussell and R.J. Kreuz (eds.), Social and Cognitive Psychological Approaches to Interpersonal Communication, 145–174. Mahwah, NJ: Lawrence Erlbaum. 1998b How addressees affect spatial perspective choice in dialogue. In P.L. Olivier and K.-P. Gapp (eds.), Representation and Processing of Spatial Expressions, 231–245. Mahwah, NJ: Lawrence Erlbaum. Siegel, Alexander W. and Sheldon H. White 1975 The development of spatial representations of large-scale environments. In H.W. Reese (ed.), Advances in Child Development and Behavior, 9–55. New York: Academic Press. van der Sluis, Ielka and Emil Krahmer 2000 Generating referring expressions in a multimodal context: An empirically motivated approach. In Walter Daelemans, Khalil Sima’an, Jorn Veenstra and Jakub Zavrel (eds.), Computational Linguistics in the Netherlands 2000, Selected Papers from the 11th CLIN Meeting, 158–176. Amsterdam: Rodopi. Sorrows, Molly E. and Stephen C. Hirtle 1999 The nature of landmarks for real and electronic spaces. In C. Freksa and D.M. Mark (eds.), Spatial Information Theory – Cognitive and Computational Foundations of Geographic Information Science, 37–50. Berlin: Springer. Sowa, Timo and Ipke Wachsmuth 2002 Interpretation of shape-related iconic gestures in virtual environments. In Wachsmuth, I. and T. Sowa (eds.), Gesture and Sign Language in Human–Computer Interaction. International Gesture Workshop, London, April 18–20, 2001, 21–33. Berlin: Springer. Spejewski, Beverly 1994 Temporal subordination in discourse. Dissertation. University of Rochester, New York. Spenader, Jennifer 2002 Presupposed propositions in a corpus of dialogue. In Kees van Deemter and Rodger Kibble (eds.), Information Sharing: Reference

338

References



and Presupposition in Language Generation and Interpretation, 35– 60. Stanford: CSLI Publications. 2003 Between binding and accommodation. In P. Kühnlein, H. Rieser and H. Zeevat (eds.), Perspectives on Dialogue in the New Millennium, 79–110. Amsterdam and Philadelphia: John Benjamin Publishing Company. Sperber, Dan, and Wilson, D. 1986 Relevance: Communication and cognition. Cambridge, MA: Harvard University Press. Stalnaker, Robert C. 1978 Assertion. In P. Cole (ed.), Syntax and Semantics 9: Pragmatics, 315–332. New York: Academic Press. Strawson, Peter F. 1950 On referring. Mind 52, 320–44. Svorou, Svetlana 1994 The Grammar of Space. Amsterdam: John Benjamin Publishing Company. Talmy, Leonard 1983 How language structures space. In H. Pick and L. Acredolo (eds.), Spatial Orientation: Theory, Research, and Application, 225–282. New York: Plenum. 2000 Toward a Cognitive Semantics, 2 vols. Cambridge, MA: MIT Press. Tappe, Heike 2000 Perspektivenwahl in Beschreibungen dynamischer und statischer Wegeskizzen. In C. Habel and C. von Stutterheim (eds.), Räumliche Konzepte und sprachliche Strukturen, 69–95. Tübingen: Niemeyer. Taylor, Holly A., Susan J. Naylor, Robert R. Faust, and Phillip J. Holcomb 1999 “Could you hand me those keys on the right?” Disentangling spatial reference frames using different methodologies. Spatial Cognition and Computation 1 (4), 381–397. Taylor, Holly A. and Susan J. Naylor 2002 Goal-directed effects on processing a spatial environment. In K.R. Coventry and P. Olivier (eds.), Spatial Language: Cognitive and Computational Perspectives, 233–253. Dordrecht: Kluwer Academic Publishers. Taylor, Holly A. and Barbara Tversky 1992 Spatial mental models derived from survey and route descriptions. Journal of Memory and Language 31, 261–292. 1996 Perspective in spatial descriptions. Journal of Memory and Language 35, 371–391.



References

339

Tenbrink, Thora 2004 Kausalität in temporalen Konnektoren. AG Konnektorensemantik, 26. Jahrestagung der Deutschen Gesellschaft für Sprachwissenschaft, Mainz, 25. – 27. Februar 2004. 2005 Identifying objects on the basis of spatial contrast: An empirical study. In Freksa, C., M. Knauff, B. Krieg-Brückner, B. Nebel, and T. Barkowsky (eds.), Spatial Cognition IV: Reasoning, Action, Interaction, 124–146. Berlin/Heidelberg: Springer. 2006 Localising objects and events: Discoursal applicability conditions for spatiotemporal expressions in English and German. Dissertation, University of Bremen, 2005. Mikrofiche, Staats- und Universitätsbibliothek Bremen, ej 6922/E03. 2007 Imposing common ground by using temporal connectives: The pragmatics of before and after. In A. Fetzer and K. Fischer (eds.), Lexical Markers of Common Grounds, 113–139. London: Elsevier. Tenbrink, Thora and Frank Schilder 2003 (Non)temporal concepts conveyed by before, after, and then in dialogue. In Peter Kühnlein, Hannes Rieser and Henk Zeevat (eds.), Perspectives on Dialogue in the New Millennium, 353–380. Amsterdam/Philadelphia: John Benjamin Publishing Company. Tenbrink, Thora, Andi Winterboer, and Reinhard Moratz (forthc.) Motion instructions by dynamic projective terms: Linguistic considerations and experiences in human–robot interaction. In E. van der Zee and M. Vulchanova (eds.), Motion Encoding in Language and Space. Oxford University Press. Timpf, Sabine, Gary S. Volta, David W. Pollock, and Max J. Egenhofer 1992 A conceptual model of wayfinding using multiple levels of abstraction. In I. Campari, A. U. Frank, and U. Formentini (eds.), Theories and Methods of Spatio-Temporal Reasoning in Geographic Space, 348–367. Berlin/Heidelberg: Springer. Thorisson, Kristinn 1994 Simulated perceptual grouping: An application to human computer interaction. Proceedings of the 16th Annual Conference of Cognitive Science Society, 876–881. Atlanta GA. Traugott, Elizabeth C. 1978 On the expression of spatio-temporal relations in language. In Greenberg, J.H. (ed.), Universals of Human Language, Vol III: Word Structure, 369–400. Stanford University Press. Tucker, Mike and Rob Ellis 1998 On the relations between seen objects and components of potential actions. Journal of Experimental Psychology: Human Perception and Performance 24 (3), 830–846.

340

References



Tversky, Barbara 1999 Spatial perspective in descriptions. In P. Bloom, M.A. Peterson, L. Nadel and M.F. Garrett (eds.), Language and Space, 109–169. Cambridge, MA: MIT Press. Tversky, Barbara and Paul Lee 1998 How space structures language. In C. Freksa, C. Habel and K.F. Wender (eds.), Spatial Cognition. An Interdisciplinary Approach to Representing and Processing Spatial Knowledge, 157–175. Berlin: Springer. 1999 Pictorial and verbal tools for conveying routes. In C. Freksa and D.M. Mark (eds.), Spatial Information Theory– Cognitive and Computational Foundations of Geographic Information Science (COSIT ‘99), 51–64. Berlin: Springer. Tversky, Barbara, Paul Lee, and Scott Mainwaring 1999 Why do speakers mix perspectives? Spatial Cognition and Computation 1 (4), 399–412. Tyler, Andrea and Vyvyan Evans 2003 The Semantics of English Prepositions: Spatial Sciences, Embodied Meaning, and Cognition. Cambridge University Press. Vandierendonck, André, Vicky Dierckx, and Gino De Vooght 2004 Mental model construction in linear reasoning: Evidence for the construction of initial annotated models. Quarterly Journal of Experimental Psychology 57A, 1369–1391. Vater, Heinz 1994 Einführung in die Zeit-Linguistik. Hürth: Gabel. Vendler, Zeno 1957 Verbs and times. The Philosophical Review 66, 143–160. Vorwerg, Constanze 2001 Raumrelationen in Wahrnehmung und Sprache: Kategorisierungsprozesse bei der Benennung visueller Richtungsrelationen. Wiesbaden: DUV. 2003 Use of reference direcions in spatial encoding. In Freksa, C, Brauer, W, Habel, C. and Wender, K. F. (eds.). Spatial Cognition III: Routes and Navigation, Human Memory and Learning, Spatial Representation and Spatial Learning, 321–347. Berlin: Springer. 2004 Two kinds of attribute categorization. 16th Annual Convention of the American Psychological Society, Chicago, May 27–30, 2004, poster presentation. Vorwerg, Constanze and Gert Rickheit 1998 Typicality effects in the categorization of spatial relations. In C. Freksa, C. Habel and K.F. Wender (eds.), Spatial Cognition. An Interdisciplinary Approach to Representing and Processing Spatial Knowledge, 203–222. Berlin: Springer.



References

341

Vorwerg, Constanze and Thora Tenbrink 2007 Discourse factors influencing spatial descriptions in English and German. In T. Barkowsky, M. Knauff, G. Ligozat, and D. Montello (eds.), Spatial Cognition V: Reasoning, Action, Interaction. Berlin: Springer. Vorwerg, Constanze and Petra Weiß 2004 Einfluss der Verbsemantik auf die Interpretation von Richtungspräpositionen. In D. Kerzel, V. Franz, and K. Gegenfurtner (eds.), Experimentelle Psychologie. Beiträge zur 46. Tagung experimentell arbeitender Psychologen, 271. Lengerich: Pabst Science Publishers. Wachsmuth, Ipke and Yong Cao 1995 Interactive graphics design with situated agents. In W. Strasser and F. Wahl (eds.), Graphics and Robotics, 73–85. Berlin: Springer. Walker, Karen and Emile van der Zee 2005 Motion encoding in language: The use of reference frames. Thematic Session on Motion Encoding, 21st Scandinavian Conference of Linguistics, Trondheim, June 1–4, 2005. Watson, Matt E., Martin J. Pickering, and Holly P. Branigan 2004 Alignment of reference frames in dialogue. Cogsci 2004: 26th Annual Meeting of the Cognitive Science Society, August 5–7, 2004, Chicago. Weiß, Petra, Joachim Grabowski, and George A. Miller 1996 Factors affecting spatial deictic communication: A comparison of German and American English. In Proceedings of the Second Colloquium on Deixis “Time, Space and Identity”, March 28–30 1996, Nancy, 45–73. Nancy: Centre de Recherche en Informatique. Wilson, Deirdre and Dan Sperber 1998 Pragmatics and time. In R. Carston and S. Uchida (eds.), Relevance Theory: Applications and Implications, 1–22. Amsterdam: John Benjamin Publishing Company. von Wolff, Annette 2001 Transformation und Inspektion mentaler Umraumrepräsentationen: Modell und Empirie. Vienna: GeoInfo Series. Wunderlich, Dieter 1985 Raum, Zeit und das Lexikon. In Schweizer, H. (ed.). Sprache und Raum: Ein Arbeitsbuch für das Lehren von Forschung, 66–89. Stuttgart: Metzler. Wunderlich, Dieter and Michael Herweg 1991 Lokale und Direktionale. In A. von Stechow and D. Wunderlich (eds.), Semantik: Ein internationales Handbuch der zeitgenössischen Forschung, 758–785. Berlin/New York: de Gruyter.

342

References



Zacks, Jeffrey and Barbara Tversky 2001 Event structure in perception and conception. Psychological Bulletin 127 (1), 3–21. van der Zee, Emile and Rik Eshuis 2003 Directions from shape: How spatial features determine reference axis categorization. In E. van der Zee and J. Slack (eds.), Representing Direction in Language and Space, 209–225. Oxford University Press. van der Zee, Emile and Jon Slack (eds.) 2003 Representing Direction in Language and Space. Oxford University Press. Zimmer, Hubert D., Harry R. Speiser, Jörg Baus, Anselm Blocher, and Eva Stopp 1998 The use of locative expressions in dependence of the spatial relation between target and reference object in two-dimensional layouts. In C. Freksa, C. Habel and K.F. Wender (eds.), Spatial Cognition. An Interdisciplinary Approach to Representing and Processing Spatial Knowledge, 223–240. Berlin: Springer. Zwaan, Rolf A. and Gabriel A. Radvansky 1998 Situation models in language comprehension and memory. Psychological Bulletin 123 (2), 162–185.

Subject index

abstractness, 3, 13, 18, 19–20, 25, 39, 276 accessibility, 3, 25, 27–29, 98, 128– 130, 173, 177, 269–270, 277 accommodation, 87–88, 92–100, 110–111 adjective, 8, 30–33, 121–122, 141, 159, 185–186, 214, 224–227, 236–237, 241–243, 246–247, 255–256 alternative, 210–212, 223, 277–280 ambiguity, 26, 85–86, 161, 214–216, 231–232, 252, 260–263, 276 applicability regions, 133, 168, 215, 239, 262 attribute dimensions, 280–281 axis, 26, 130–135, 237–239, 255 between, 222, 224, 229, 242, 245, 249–250, 261, 276 causality, 29, 36, 41–42, 47–51, 73– 74, 97, 101–104, 281 cognitive science, 2, 9, 113, 267 conceptual categories, 69, 83–86, 101–110, 112–117, 288–293 concreteness, 3, 7, 13–14, 18–20, 25, 27 conditionality, 50, 65, 77 conjunction, 8, 31, 33–34, 37, 42, 44, 101, 268 context, 5–8, 21, 24, 29, 32, 43, 45, 54, 66, 122, 134, 175, 183, 214, 257, 271, 275, 280, 287

contrast, 28, 33, 79, 110, 141, 186– 190, 201, 210, 228–229, 243, 249–250, 260, 262, 265, 277 corpus, 6, 68–69, 89–90, 100–101, 116, 202, 207 cyclicity, 25 dann, 61–63 deixis, 21–22, 26, 35, 136–137, 160 deontic, 57 dependency, 9, 11, 24, 69, 73, 77– 78, 97, 104–106 diachronic, 14–15, 148 dialogue, 197–198, 206 dimensionality, 22–23, 25, 39, 171 dimensional terms, 2, 118 directionality, 25–26, 118–119 discourse analysis, 1–2, 35 discourse–new, 53, 86–87, 90, 95– 100, 110 discourse relation 35–37, 42, 50, 54, 64, 66, 69, 72–73, 89, 96, 115, 268, 271, 273, 276, 282 discourse task, 183–190 discriminativity, 188, 209, 223, 227, 236–237, 254, 279 earlier, 58–60 elaboration, 39, 83, 92, 96 epistemic, 52–53, 57, 61 explicit length of time, 82–83, 96, 108–109 figure, 40–42, 196–197

344 Subject index functional concepts, 29, 126–130, 177–183, 281

origin, 8, 129, 134, 137–140, 144, 164, 228, 232, 236, 249, 252

ground, 40–42, 196–197 grouping, 140, 219, 226, 236

perspective, 26–27, 135–138, 150– 154, 219–220, 229–233, 250– 252 preposition, 13, 30–33, 37, 42, 44, 55, 118, 120–122, 180, 224, 241, 246–247 presupposition, 40, 51–54, 86–89, 99–100, 110–111 pragmatics, 2, 66, 101, 113, 122– 124, 190, 198, 274 proto-scene, 17, 123, 127–128 prototypicality, 36, 124–125, 165– 167, 190, 228–229, 250, 261, 280–281 proximity, 22, 35, 45–47, 72–78, 102, 172–176, 284

immediate succession, 47, 61–62, 64, 79–82, 97, 109 insertion, 74–76, 97, 106–108 interaction partner, 89, 107, 151– 152, 186, 188–189, 196–198, 206, 219, 232, 251–254, 260– 262 internal part, 133, 135, 138, 154– 155, 165, 205, 218 landmark, 26, 127, 140, 150, 164, 183–184, 193, 197, 205, 234, 265, 285 later, 58–60 linearity, 25–26, 39, 280–281 locative, 119 metaphor, 1, 12–19, 28, 37, 284 motion, 13, 18, 23–25, 119, 134, 153–154, 262, 277 modifiers, 72, 114–115, 166, 187, 190, 204, 227–229, 243–244 no change, 109–110 non-veridicality, 54–57, 85, 97, 107 object, 3, 19–23, 28, 34, 133, 154– 155, 168–174, 177–185, 193, 271, 275–277, 284 orientation, 19, 22, 119, 128, 131, 134, 138, 154, 162, 164, 170, 180, 251

reference system, 135–165, 214– 219, 258–260 absolute, 142–148, 164, 258 environment–induced, 162 group–based, 140–141, 150, 174, 204, 213–215, 218, 236–237, 257 internal, 154–161, 215, 217–218, 256 intrinsic, 137–138, 145–149, 156–157, 181, 215, 220, 228, 231–232, 234, 236, 243, 251– 252, 259 relative, 138–142, 145–150, 231–232, 251–252 visual–field based, 155–158, 161, 214–215, 236–237, 253– 254, 262

Subject index 345 regulation, 76–77, 98, 108 relationships, spatial / temporal terms, 14–18, 26, 30–37, 270–285 English / German, 8, 31–32, 43, 53, 111–112, 121–122, 141, 149, 156–157, 160, 185, 187, 245– 257 space / time, 12–13, 18–19, 24– 25, 30, 277, 280, 284 spatial, 5, 21–22, 26, 29, 36, 125, 136, 177, 184, 260, 264, 270, 275, 279, 283–284 temporal, 5, 21–22, 26, 29, 36, 38, 45–47, 65, 69, 72, 83–86, 103, 108–109, 113–117, 279, 283–284 relatum, 27, 64, 133, 137–141, 154, 168–171, 180–182, 214, 217, 233–236, 252–254 representation, 3, 12–13, 18–21, 37, 38–42, 57, 88, 130, 156, 186, 215, 239, 250, 259, 271, 281 route, 24, 26, 134, 136, 152, 166, 183–184, 191–194, 197, 244, 264–265 scene description, 212 semantics, 2, 31, 45, 122–130, 273, 283, 287 sequence, 13, 24–26, 34, 36, 38, 40– 41, 45–46, 50, 60, 112, 134, 177, 208, 225, 242, 277, 283

simultaneity, 27, 83, 277 spatialist conception, 12 spatial template, 124, 165–167, 281 speaker strategies, 87, 100, 151, 191–197, 221–223, 240, 245– 246, 261–262, 264–265 structure abstract, 17 axial, 130–133, 168 graded, 123, 166–167, 187, 190 temporal, 38–39, 62 succession, 5, 27, 40, 46, 61–63, 69, 79–82, 97 syntax, 30–33, 42–44, 69–72, 120– 122, 159–161, 224–227, 241– 243, 246–247 temporal order, 5, 26–27, 39, 81–82, 85, 98, 110–113, 149, 278, 281 termination, 50–51, 60, 78, 93–94, 98, 106 then, 61–63 topology, 119, 126, 149, 158–159, 173, 181, 208 underspecification, 4, 215, 259, 276, 286 verb, 34–35 when, 60–61