The conventionality of figurative language

Transcription

1 Language in Performance LiP Sandra Handl The conventionality of figurative language A usage-based study

2 The conventionality of figurative language

3 46 Edited by Werner Hüllen and Rainer Schulze Advisory Board: Thomas Herbst (Erlangen), Andreas Jucker (Zürich), Manfred Krug (Bamberg), Christian Mair (Freiburg i. Br.), Ute Römer (Hannover), Andrea Sand (Trier), Hans-Jörg Schmid (München), Josef Schmied (Chemnitz) and Edgar W. Schneider (Regensburg)

4 Sandra Handl The conventionality of figurative language A usage-based study

5 Bibliografische Information der Deutschen Nationalbibliothek Die Deutsche Nationalbibliothek verzeichnet diese Publikation in der Deutschen Nationalbibliografie; detaillierte bibliografische Daten sind im Internet über < abrufbar. Zugl. Dissertation an der Ludwig-Maximilians-Universität München, Narr Francke Attempto Verlag GmbH + Co. KG Dischingerweg 5 D Tübingen Das Werk einschließlich aller seiner Teile ist urheberrechtlich geschützt. Jede Verwertung außerhalb der engen Grenzen des Urheberrechtsgesetzes ist ohne Zustimmung des Verlages unzulässig und strafbar. Das gilt insbesondere für Vervielfältigungen, Übersetzungen, Mikroverfilmungen und die Einspeicherung und Verarbeitung in elektronischen Systemen. Gedruckt auf säurefreiem und alterungsbeständigem Werkdruckpapier. Internet: info@narr.de Printed in Germany ISSN ISBN

6 Table of contents Acknowledgements 1 Introduction 2 Literality, figurativity, and conventionality 2.1 Literal vs. nonliteral: A matter of oppositeness? 2.2 The many meanings of literal 2.3 Literality on different levels 2.4 Summary 3 Metaphor, metonymy, and cognition 3.1 Metaphor and metonymy in pre-cognitive linguistics 3.2 Figurative language and figurative thought The conceptual theory of metaphor Metonymy as a cognitive-linguistic notion Basic assumptions Metonymy as a reference-point construction The ubiquity of metonymic processes Mental spaces and conceptual integration Basic assumptions Metaphor, metonymy, and the theory of conceptual blending Summary 4 Motivation and conventionality in language and thought 4.1 Motivation: From structuralism to cognitive linguistics Non-cognitive approaches to motivation and conventionality Motivation and cognition Approaches to conventionalization and conventionality The conventionalization of language: Institutionalization and lexicalization Conventionality and cognition: A preliminary sketch

7 The notions of salience and entrenchment Factors influencing the entrenchment and conventionalization of figurative language Typologies of conventional and dead metaphors and metonymies Typologies of metaphor Creative, conventional, dead, and literal metaphors The death of metaphors and semantic change Typologies of metonymy Functional criteria and the synchronic-diachronic distinction Types of metonymy and the notion of domains Metonymy as a prototypical notion Criteria of conventionality and the death of metonyms Processing nonliteral language: Linguistic theory and psycholinguistic evidence The standard pragmatic model: Sequential processing The direct access view: The influence of context on comprehension The entrenchment-based view: Conventionality, frequency, and language processing The graded salience hypothesis Processing well-entrenched and less entrenched meanings Summary 5 Investigating the conventionality of figurative language 5.1 Figurative language and corpus linguistics: Methods and problems General considerations The choice of the source domain vocabulary The metonyms The metaphors The procedure: Looking for metonymies and metaphors in the BNC Metonymic senses

8 KWIC concordances and the problem of underspecification Conjoined phrases, multiple mappings, and meanings Further problems: The metaphor-metonymy interface, metonymic compounds, and metonymic chains Metaphorical senses The sub-corpora Extracting and organizing metaphorical senses Problematic cases The presentation of the results Evidence for the conventionality of metonymic meaning and thought 6.1 Frequent and less frequent metonyms: An overview 6.2 The conventionality of metonymies: A matter of linguistic idiosyncrasy or conceptual regularity? Salience and the conventionality of metonymy Wholes and parts WHOLE FOR PART, PART FOR WHOLE, and PART FOR PART metonymies Reversible mappings: Preferred routes and nondefault cases The influence of conceptual structure and salient attributes Conventional and less conventional metonymies: Ontological salience and attribute salience ATTRIBUTE OF ENTITY FOR ENTITY metonymies PRINCIPAL PART FOR WHOLE metonymies CENTRAL FACTOR FOR INSTITUTION/PROFESSION/ ACTIVITY metonymies LOCATION FOR ENTITY metonymies INSTITUTION FOR PEOPLE metonymies WHOLE FOR PRINCIPAL PART metonymies Further metonymic mappings Types of salience and the conventionality of metonymic mappings and meanings: An intermediate summary

9 6.3 Target-in-vehicle salience and the emergent structure of metonymies Salient and non-salient metonymic targets Types of attribute salience and the question of emergent structure Summary Evidence for the conventionality of metaphorical meaning and thought 7.1 The results of the metaphor study: An overview 7.2 Frequent and infrequent metaphorical mappings and meanings 7.3 Preferred metaphorical construals 7.4 Metaphors and attribute salience: Emergent structure and conventional metaphorical meaning The relativity of attribute salience Lean and rich mappings and their attributes Summary 8 Conclusion References Dictionaries Corpora Appendix A: The results of the metonymy study Appendix B: The results of the metaphor study 8

10 Acknowledgements IDEAS ARE CHILDREN, as the Master Metaphor List states, and just like children cannot develop their capacities in isolation, the ideas gathered in this book owe very much to the help and guidance of those surrounding me. First of all and most of all I would like to thank Prof. Dr. Hans-Jörg Schmid, my dissertation advisor at the Ludwig-Maximilians-University Munich, for his constant support, invaluable advice, and patience during all stages of the project s development. Many thanks also go to Prof. Dr. Richard W. Janney, whose critical comments often helped me see things in a different light. I am deeply indebted to Prof. Dr. Leonhard Lipka for showing me the way into the wonderful world of linguistics and for sparking my original interest in questions of conventionality. I am very grateful to Prof. Dr. Rainer Schulze, the series editor of LiP, for his assistance in the publication process. Furthermore I would like to thank Mareike Reichelt and Karin Burger from Narr publishers for their professional expertise and their readiness to answer all my questions. I also owe many thanks to Kathleen Rabl, who proofread the entire manuscript and provided many additional examples. Special thanks go to my friend and colleague from American Studies, PD Dr. Kerstin Schmidt, for her genuine interest in my work and for her priceless help with everything. Most crucially, however, I am indebted to my parents, Brigitte Handl and Manfred Singer, and the rest of the family, Susanne Handl, Torsten Krauß, Laura Handl and Philipp Handl, for encouraging and sustaining me, and for supporting me in countless other ways. 9

11

12 1 Introduction One of the fundamental assumptions of cognitive linguistics is that the most vital mechanisms of thought are those which are entirely subconscious. This is particularly evident in cognitive approaches to figurative language. Metaphors have been demonstrated to be profuse in everyday discourse. They are so common that we no longer notice them, they are conventional. Moreover, and more importantly, conventional metaphorical expressions form coherent systems based on well-established metaphorical patterns of thought. These metaphorical ways of thinking help us to understand and cope with the world around us by helping to structure abstract concepts. In this way, metaphors determine not only what we think, but also what we do. While metaphors have been in the limelight of cognitive-linguistic research ever since the publication of Lakoff and Johnson s (1980) classic book, metonymy has only recently attracted attention. Just like metaphor, metonymy is guided by highly systematic conceptual mappings that lead to regular linguistic expressions. And just as in the case of metaphor, it is assumed that the mappings which are cognitively the most central are those which are so deeply entrenched that the figurative nature of the linguistic expressions they have given rise to is not noticed. For both metaphor and metonymy researchers have compiled substantial lists of examples meant to prove the conventionality of the expressions and, correspondingly, also of the mappings underlying them. But are all of these examples really equally conventional ways of speaking? Is a reference to persons by way of the metonymic expression We need some new faces around here really as common as a reference to the United States by way of the metonym America? And is a metaphorical expression like I m shopping for a better philosophy equally essential from a cognitive point of view as a construal of a verbal ARGUMENT in terms of WAR as in He attacked every weak point in my argument? Even though the concept of conventionality is the foundation of many key assumptions in the conceptual theory of figurative language and thought, the meaning of the term is frequently taken for granted. With respect to metaphors, researchers have proposed some tangible criteria, such as unidirectionality and the invariance principle, which can help to distinguish unconventional from more conventional expressions. However, the principles that are supposed to govern the conventionality of linguistic metonyms and metonymic ways of reasoning, such as the laws of ontological salience are often somewhat more vague, difficult to apply and have hardly ever been put to the test. Therefore, the primary focus will be on metonymy rather than on metaphor throughout this study, in the theoretical as well as empirical parts. 11

13 Given that the conventionality of metonyms and metaphors is intrinsically connected to the question of which figurative ways of thinking are at the heart of human cognition, we should attempt to understand better what it means for metonyms and metaphors to be conventional. In what follows, I will thus search for conceptual rules and regularities that govern the conventionality of figurative language: Which preconditions have to be met by figurative expressions in order to be accepted by the speech community at large? Are all types of metonyms and metaphors subject to the same constraints, or do the distinct types have to fulfil different criteria in order to gain wider currency? Are all examples that belong to one and the same type equally well-established, or are there exceptions to the more general tendencies? To answer these questions, it is necessary to introduce several basic theoretical frameworks. Since the nonliterality of conventional figurative expressions is characteristically no longer consciously apprehended, I will first address the general distinction of literal and nonliteral language and its theoretical and practical value (chapter 2). Chapter 3 discusses the key concepts of cognitively-oriented theories of figurative language. Since some of the assumptions usually considered to be unique to the cognitive framework are actually indebted or at least related to earlier, non-cognitive treatments of metaphor and metonymy, I will begin this chapter with a brief overview of some of these earlier approaches (chapter 3.1), before taking up the most central claims of the conceptual theory of figurative language and thought in chapter 3.2. Here, the first part deals with metaphor (chapter 3.2.1), the second, more extensive one, with metonymy as a cognitive-linguistic notion (chapter 3.2.2). The last section of chapter 3 then introduces the second major cognitive theory which is relevant for an examination of the conventionality of figurative language, the theory of conceptual blending (chapter 3.3). Chapter 4 concentrates on the chief concerns of my study, i.e. theories of the conventionality of figurative language and thought. The chapter begins with a consideration of the relationship between motivation and conventionality (chapter 4.1) and then turns to the process of conventionalization, its preconditions and consequences (chapter 4.2). As explicit examinations of conventionality are rare in cognitive linguistics, structuralist theories are also briefly reviewed in chapters and As far as the cognitive perspective is concerned, two terms, namely salience and entrenchment are of major importance for a description of the conventionalization of figurative language, and these notions are thus introduced in chapter before an initial, rather sketchy picture of the conventionalization of figurative language is drawn in chapter While the latter provides the basis for the discussion and critical evaluation of extant typologies of conventional and dead metaphors and metonymies in chapter 4.3, the remainder of chapter 4 is devoted to the question of the psychological reality of the conventionality of figurative language: It examines the evidence found for 12

14 and against the position that the frequency and familiarity of figurative language bear directly upon its on-line processing. The factors identified as relevant for the conventionalization of linguistic metonyms and metaphors in the theoretical discussion, like frequency of occurrence and salience, are then used as a starting point for the empirical investigation of authentic linguistic data in chapters 5 to 7. This part is mainly based on the results of a large-scale corpus-linguistic study of metonymic and metaphorical expressions. Once again, as in the theoretical section, the focus will be on metonymy. In many theoretical accounts, no matter whether they belong to the realm of structuralism or cognitive linguistics, the factor of frequency is quoted as one of the main indicators of conventionality an assumption that has been supported by many psycholinguistic experiments, and therefore the frequency of metonyms and metaphors in natural language use was analyzed. A description of the general procedure and the problems involved in such a study is given in chapter 5. In the following two chapters, the results of the corpus study are presented and the relative frequencies found for the single instances of different metonymic and metaphorical mappings are related to the criteria of conventionality established in the theoretical section. Chapter 6 focuses on metonymies, and chapter 7 on metaphors. The reasons why some figurative construals are clearly preferred to others are discussed, and it is claimed that the conventionality of figurative language cannot be adequately described if only the level of the conceptual mappings is considered. With regard to metonymies, it is shown in the first sections of chapter 6 (6.1 to ) that the laws of ontological salience often proposed to account for the degree of conventionality of different mappings cannot explain all the relative frequencies of the linguistic metonyms investigated. A comprehensive account of the conventionality of different metonyms presupposes a consideration of the structures of the concepts involved in the mappings, since the degree of conventionality hinges on the salience of single attributes in speakers mental representations of these concepts. The notion of attribute salience is therefore introduced in chapter and applied as an additional explanatory criterion of the corpus-linguistic results in the following sections (chapter 6.2). In contrast to ontological salience, attribute salience can explain regular cases and numerical outliers alike. This becomes evident in chapter 6.3, which is partly based on the results of an attribute-listing task (chapter 6.3.1). In addition, it is argued that metonymies can be characterized by two different types of emergent structure, salience-upholding and salience-changing emergent structure, and that the first results in high relative frequencies of the expressions, whereas the latter leads to low relative frequencies (chapter 6.3.2). Parallel to chapter 6, chapter 7 begins with a preliminary overview of the results regarding the usage of metaphors (chapter 7.1). On the whole, the scope of the metaphor study is more restricted, and the results thus 13

15 allow much less generalizations than for the metonymies. In chapter 7.2, it is demonstrated that many examples commonly employed to illustrate the conventionality of metaphorical speaking are actually not very frequently used in natural discourse. For example, verbs belonging to certain general frames, like COMMERCIAL EVENTS, seem to be particularly resistant to metaphorical meaning extensions. And for some conceptual metaphorical mappings, there are obvious preferences to express them with certain linguistic means. How this is connected to metaphorical highlighting is examined in chapter 7.3. The last section of chapter 7 is devoted to the role of attribute salience for the conventionality of metaphors. The application of rich mappings of the type ARGUMENT IS WAR or A THEORY IS A BUILDING does, as a rule, not disturb the saliency of the attributes of the respective source domains. Lean mappings of the type A PERSON IS A PIG, however, behave differently; they typically single out only one aspect of the source domain and this can result in changes of the salience of attributes in the source, a process which is best described as generating conventionally fixed emergent structure. Attribute salience is thus not only relevant for the conventionality of metonymies, but also for the conventionality of metaphors. The findings of the complete study are then summarized in chapter 8. 14

16 2 Literality, figurativity, and conventionality 2.1 Literal vs. nonliteral: A matter of oppositeness? A study of the conventionality of figurative language seems to presuppose that the notion of figurative language refers to a well-defined and unified concept. At least our folk model of language 1 is based on the assumption that there is a strict dividing line separating what is literal from what is nonliteral. Certainly, the average language user would not hesitate to categorize (1) as figurative and (2) as literal. (1) Juliet is the sun! (2) A dog is a domesticated carnivorous mammal that typically has a long snout, an acute sense of smell, non-retractile claws, and a barking, howling, or whining voice. (NODE; s.v. dog) So far, the issue seems uncontroversial. But deciding upon the literality or nonliterality of sentences or utterances becomes more difficult in cases like the following: (3) He had to digest the bad news. (4) There were many new faces around. On the one hand, (3) and (4) seem to be perfectly normal utterances that could occur in many everyday contexts. They are by no means poetic or otherwise stylistically marked. They are easy to understand, and do not, contrary to example (1), seem to leave much room for creative interpretation. On the other hand, on closer inspection it becomes apparent that (3) and (4) contain lexical items that are not used in their original and most literal senses. The verb digest literally means break down (food) in the stomach and intestines into substances that can be used by the body and the noun face actually refers only to the front part of a person s head from forehead to the chin (NODE; s.v. digest, v., face, n.). These, however, are not the senses principally intended in (3) and (4), where digest describes the way a person tries to cope with some information, and faces refers to people and not to their countenances only. Therefore, (3) and (4) are nonliteral, but much less obviously so than (1). English native speakers are quite 1 That human thinking is determined by oversimplified versions of reality is the rule rather than the exception. What is stored in speakers minds and forms the basis for cognitive processes are, in Lakoff s (e.g. 1987) terminology, idealized cognitive models (ICMs), i.e. idealized versions of the world abstracted from reality in various ways. Good examples are the everyday concept of a MOTHER (cf. Lakoff 1986: , 1987: 74-76, 79-84), or the seven-day calendric cycle. Even though time comes as a continuum with no boundaries and structure, cultures have imposed an organization upon it: We have a work week of five days and a weekend that lasts two (Lakoff 1987: 68-69). 15

17 used to talking about ideas in terms of food, and, consequently, about understanding information as digesting it. Similarly, they are accustomed to conceptualizing persons in terms of their faces, since the face is what we look at first when we meet other people. The fact that it is so common to talk about information in terms of food and about persons in terms of their faces is responsible for the unobtrusiveness of the nonliterality of (3) and (4). If average speakers noticed it at all, they would do so only at second glace, and still might end up confused about whether it is really correct to categorize the examples as figurative, because the senses of digest and face realized in (3) and (4) are so ordinary. Merely by virtue of being so familiar, they somehow seem be or have become literal in a way. The discussion below demonstrates that linguistic description is confronted with similar problems when it comes to distinguishing between literal and nonliteral language. For a long time linguists concentrated on the study of literal language and considered figurative uses as derived and parasitic, 2 thus implying a neat boundary between the two. More recent research in cognitive semantics, psycholinguistics and pragmatics has, however, shown that the criteria proposed for the distinction cannot account for all the cases, so that the dividing line between literality and nonliterality turns out to be much more blurred than commonly assumed. 2.2 The many meanings of literal The uncertainty about what should be regarded as literal and what belongs to the realm of figurative language is by no means restricted to laypersons judgements. Even though much work has been devoted to this question (e.g. Bartsch 1996; Dascal 1987, 1989; Gibbs 1984, 1989, 2002; Katz 1977; Lakoff 1986; Récanati 1989, 1995, 2004; Searle 1978), there is no single, allencompassing answer to it, the main reason being that the concept of literal meaning is not a unified one. As Lakoff (1986: 292) and Gibbs (1993, 1994: 75) point out, the term literality can be understood and has been investigated in up to five different senses: 3 1. Conventional literality (i.e., plain, direct language) 2. Subject-matter literality (i.e., expressions ordinarily used in connection with a particular subject-matter) 2 This position is not restricted to semantics. Many, more traditionally oriented pragmatic approaches also hold that any successful understanding of a nonliteral expression, e.g. an indirect speech act or a linguistic metaphor, relies on the rejection of the corresponding literal interpretation. Grice (1975; 1998) and Searle (1979) are often quoted as prime representatives of this view; cf. also chapters 3.1 and Other linguists also state that literal meaning has been characterized in many different ways: as direct, fixed, explicit, and necessary, as fully compositional, context-free, and truth-conditional (cf. e.g. Ariel 2002a: 362; Dascal 1987: ; Katz 1977: 14; Searle 1978: ). 16

18 3. Non-metaphorical literality (i.e., language that is not even partly understood in terms of something else) 4. Truth-conditional literality (i.e., language that fits the world) 5. Context-free literality (i.e., language as interpreted in a null context) Even if the different senses of literality are considered in isolation, they are difficult to apply. Searle (1978), for example, convincingly argues that there is no such thing as a null context. Even if we encounter language without any meaningful situational (or linguistic) context, we invoke default assumptions which are based on our experiences and determine our interpretation of the linguistic input. For Searle s famous example The cat is on the mat, we automatically assume that the cat and the mat are within the gravitational field of the earth and not floating in outer space. But if context invariably plays a role for any interpretive process, literality cannot be defined as independent of the context and the speaker. 4 An account of literal meaning in terms of the principle of compositionality, which goes back to Frege ([1892] 1952), does not solve the problem either, as this framework presupposes that all morphemes and words have fixed literal meanings. Therefore, understanding sentences or rather their literal meanings should consist of an activation of all the meanings of the words in the sentence, supported by the hearer s knowledge of the syntactic rules according to which the words have been combined. Apart from the fact that background assumptions will always exist, this view has another shortcoming: That word meaning is fixed cannot be maintained for all cases. Particularly good examples to illustrate this are adjectives, as their interpretation often varies with the nouns they modify. 5 Clark (1991), for example, has demonstrated that this even holds for fairly concrete adjectives like red. If red is used as a predication of hair, subjects tend to imagine an orangish colour, if it is used as a predication of wine, the meaning is shifted towards a purply red (cf. also Ariel 2002a: 371, Gibbs 1994: 39). The meaning of red is vague. 6 In such cases it seems almost impossible to speak 4 The idea that semantic description should be restricted to sentence meaning (as opposed to utterance meaning in context) is especially prominent in the generative framework (e.g. Katz and Fodor 1963). Cf. also Katz (1977) who defines the ideal speaker/hearer s semantic competence with reference to the so-called anonymous letter criterion. 5 For an attempt to apply the principle of compositionality to larger chunks of linguistic input, cf. Bartsch s (1996) discussion of cut. She argues that literality should not be searched for on the word-level, but on the level of propositions. 6 The terminology used to refer to the different semantic aspects belonging to one lexeme is unfortunately not uniform. In general, one can distinguish between ambiguity and vagueness. The former describes cases where one word has various distinct senses, while the latter refers to underdetermined or underspecified meanings, as in the case of red. Another example of vagueness is cousin, which is underdetermined with respect to the sex of the person, and ambiguity can be illustrated with port harbour vs. port wine. A useful survey of the different terminological systems in this field can be found in Schmid (1993: ). Cf. also Geeraerts (1993a), Taylor (2003: 17

19 of one (or more) fixed literal meaning(s) of the word. There is no clear oneto-one or one-to-many correspondence between form and meaning(s) as suggested by our folk understanding of language as a conduit (cf. Reddy 1979, Ritchie 2006: 11). Truth-conditionality, too, cannot be regarded as an infallible criterion for literality. Cases in point are negated predications (cf. 5) or ambiguous sentences (cf. 6): (5) Life is not a bed of roses. (6) Anchorage is a cold city. (Leezenberg 2001: 175) Example (5) is literally true, but a literal interpretation misses the metaphorical point of the utterance. Example (6) poses a similar problem. Depending on the context, it could receive either a literal ( low temperature ) or a nonliteral ( unfriendly ) interpretation. If uttered with the second intention in mind, truth-conditions cannot cope with it, as due to the existence of the literally true interpretation the nonliteral reading does not violate them (cf. Leezenberg 2001: ). Another argument against a truth-conditional approach is that not all statements that are literally false have to be figurative. There is no criterion that separates nonsensical sentences such as A dog is a plant from nonliteral ones like My husband is a pig. 7 Neither compositionality nor truth-conditionality are definitive indicators of literality. Moreover, literal meaning is not fixed and contextindependent. While these criteria fail for various reasons if considered in isolation, the weaknesses of classic accounts of literality are even more evident if the different senses of literality are applied to one and the same example, because usually, the various senses or types of literality are tacitly assumed to converge. But this is not necessarily the case. What is of particular significance for my concerns is that linguistic semantics tends to think of the first and third sense of literality, i.e. conventional and non-metaphorical literality, as being identical (Lakoff 1986: ). Metaphorical language is considered unconventional and regarded as an antonym (if not as a complemetary term) to literal language. But as the conceptual theory of metaphor 8 convincingly claims, metaphoricity and conventionality are by no means mutually exclusive. Many quite conventional and ordinary utterances such as Your claims are indefensible or This esp. chapters 6 and 8), Tuggy (1993). In cognitive linguistics, the different semantic facets of vague lexemes have also been described as active zone phenomena (cf. e.g. Langacker 1987, 1993). 7 What is more, to specify a set of truth conditions for a sentence presupposes that some interpretation has already been given to that sentence (Gibbs 1994: 64). Without any knowledge about Anchorage and cold and the relation between the two, a truthconditional analysis is impossible. This caused many researchers, such as Searle (1978) or Récanati (1989, 2004), to modify their concept of truth-conditionality to fit the data at hand, while others flatly reject it as a criterion for literality (e.g. Ariel 2002a; Gibbs 1994, 2002). 8 An overview of this research paradigm can be found in section

20 theory needs more support have a metaphorical motivation and are thus literal in one sense, but nonliteral in another. The same holds true for metonymically motivated language (e.g. The White House isn t saying anything, He drives a BMW), although metonymy is usually not addressed in the literature on literality. This proves that the classic approaches to literality failed mainly because they tried to capture too many, too diverse aspects with the help of one (seemingly) unified notion. A promising way out of this dilemma is Ariel s (2002a) idea of abandoning the notion of literality and replacing it with three concepts of minimal meaning (2002a: 388), which will be introduced in the next section. 2.3 Literality on different levels Ariel (2002a) starts with the observation that the problems connected to the traditional notion of literality have led to so many modifications that there are now many different concepts of literal meaning (2002a: 362) which are only partly justified and, what is even more important, partly clash with each other. She therefore proposes giving up the search for a single, unified notion of literality and argues that many difficulties can be avoided or overcome by focusing on different types of basic, or as she calls them, minimal meanings. Minimal meanings occur on three levels of linguistic description the linguistic, psycholinguistic, and pragmatic (or interactional) levels and each of them reflects the primary interest of the corresponding linguistic discipline. The linguistic level is concerned with the difference between coded and inferred meanings. On this level, a meaning is minimal, if it can become active by mere decoding. Linguistically minimal meanings are not restricted to the lexeme level, but embrace larger grammaticized units such as collocations or idioms. Crucially, not only fully truth-conditional and non-figurative meanings can be linguistically minimal. In contrast to other approaches, Ariel (2002a: 392) includes figurative readings within the realm of the linguistically most central meanings. As regards the psycholinguistic level, 9 meanings are minimal if they are automatically and obligatorily accessed without causing any undue processing effort. The defining criterion on this level is speed of accessing (Ariel 2002a: 393). Psycholinguistically minimal meanings are different from linguistically minimal ones, since not all coded meanings are accessed 9 Originally, Ariel (2002a: 393) uses the expression dynamic construction of meaning in context in real time. In the strict sense, however, highly salient meanings (in Giora s 1997, 2003 terminology) have to be stored, as salience is considered (among other things) a function of familiarity and frequency. The term meaning construction should be reserved for more dynamic processes that involve at least a full sentential context. 19

21 with equal speed. More frequent and more familiar meanings are processed faster than less frequent and familiar ones (cf. e.g. Giora 2003; cf. also chapter ). For example, two of the encoded meanings of the form drop tiny amount of liquid and act of falling differ with respect to their psycholinguistic minimality, as the former is more frequent and therefore accessed faster, than the latter (cf. Ariel 2002a: 393). Only the interactional perspective takes context into account. Here, minimal meanings are less stable and therefore referred to as possible privileged interpretations (Ariel 2002a: 396; original emphasis). From a pragmatic point of view, meaning is always a question of negotiation, and hearers interpretations of utterances do not necessarily converge. Interactionally minimal meaning is characterized by its relevance for the ongoing discourse. That this level is more subjective than the other two can be seen if competing opinions occur as to which interpretational formula is the most contextually appropriate. This is demonstrated in example (7). The librarian feels it is necessary to cancel the contextually appropriate interpretation of one minute as short period of time. This illustrates the significance of interactionally appropriate meanings as those to whose sincerity the speaker is being committed. (7) Woman at the library: D you have a minute? Librarian: Literally, one minute. (Ariel 2002b: 1009) What is gained by giving up one unified notion of literality and focusing instead on the different underlying motivations of the term? One advantage is that Ariel s approach considers, as frequently demanded (cf. Gibbs 1984, 1989, 1994), the psychological reality of literal meaning. But even though it has been claimed that meaning is never entirely contextindependent, as seemingly context-free interpretations are invariably the result of the hearers assuming a default context (cf. Searle 1978), Ariel does not completely do away with the idea of literal meaning on the linguistic level either. By splitting up literality into three different minimal meanings, she takes away much of the pressure a single unified notion has to cope with. Since her model explicitly allows that a meaning is minimal on one of the levels, but non-minimal on another, it is more successful than other approaches in accounting for the contradictory findings that abound in the relevant literature. 10 The idea of minimality on different levels turns literality into a much more flexible notion. It seems reasonable to assume that a linguistically encoded but not very frequent meaning will not necessarily be accessed at all, if the context points in a different direction. According to the linguistic and psycholinguistic status of the meaning in question and the situation at hand, the different scales shift their position, to allow a unified perception of the current meaning (cf. Figure 1). 10 For an overview of these contradictory findings, cf. chapter

22 Linguistic level: encoded meaning inferred meaning Psycholinguistic level: Contextual level: fast access contextually relevant current meaning slow access contextually irrelevant Figure 1: Three scales of minimal meaning (based on Ariel 2002a) Ariel s approach manages more than the prototype view of literality proposed by Dascal (1987). The latter tries to retain many of the traditional ideas about literality (e.g. truth-conditionality and compositionality) no longer considered central by Ariel. Even though Dascal realizes that the conventionality of a meaning plays a significant role for its processing, his explanation is not very helpful. His argument is that the conventionalized meanings of indirect speech acts, idiomatic expressions and metaphors are in fact the literal ones (1987: 267). In this way, he destroys the very basis of literality as a necessary counterpart to nonliteral usage by incorporating large amounts of nonliterally motivated language into his concept of moderate literalism (Dascal 1987: 260; cf. also Dascal 1981, 1983, 1989). 11 The boundary between literal and figurative language is not as easy to draw as our everyday conception of language makes us think. However, giving up the notion of literality altogether, as Gibbs (e.g. 1994, 2002) proposes, does not solve the problem either. How would it be possible to identify a metaphor such as He attacked every weak point in my argument, if not by stating that we conceive of ARGUMENTS in terms of WAR, i.e. by pointing out that our conceptualization is not a literal one? Just because such expressions and the patterns of thinking they rely on are conventional, does not mean they are literal. Conventionality is not the same as literality and does not cause literality. Conventionality is a phenomenon that essentially cuts across the two fuzzy-edged categories of literality and figurativity. 11 For metaphorical language and many idioms, this corresponds to what Gibbs (1994: 10) calls dead-metaphor view, a view he vigorously challenges in many publications (cf. e.g. Gibbs 1993; Gibbs and O Brien 1990). That metaphors still fulfil important conceptual functions, even if their linguistic instantiations have become conventions, is also one of the major claims of conceptual metaphor theory (cf. e.g. Barcelona 2000b: 5; Lakoff and Johnson 1980: 55); cf. also chapter

23 2.4 Summary Neither literality nor figurativity can be measured in absolute terms. The traditional strict distinction of literal and figurative language has to be replaced by a psychologically realistic model. While it cannot be denied that strong contextual implications can favour even highly unfamiliar meanings, it is also safe to assume that speakers work with default assumptions, if such a strong context is lacking. Accordingly, meanings that are foremost in speakers minds, i.e. meanings that are highly conventional, still enjoy a somewhat privileged status. What matters is the meaning we conventionally associate with linguistic expressions. Let me illustrate this with two individual lexemes: to attack and America. The question is whether we are more likely to think first of their respective literal meanings, i.e. take aggressive action against (a place or enemy forces) with weapons or armed force, typically in a battle or war (NODE, s.v. attack, v.) and a landmass of the western hemisphere consisting of the continents of North and South America joined by the Isthmus of Panama (NODE, s.v. America), or whether we are so accustomed to their figurative senses, i.e. criticize or oppose fiercely and publicly (NODE, s.v. attack, v.) and name for the United States (NODE, s.v. America), that these are the first that come to our minds when we encounter the lexemes either in a weak context or without any context at all. Cognitive linguists have convincingly claimed that figurative thinking and speaking can be conventional. However, the extent to which individual conceptual mappings and figurative expressions are conventional is a question that has, so far, not been addressed in detail. Even though conventionality is unquestionably a matter of degrees, it seems to have been regarded as an absolute. Moreover, the question of what factors foster or inhibit the conventionality of metaphors and metonyms has not attracted much attention either. These are important issues, however, since drawing conclusions about the cognitive weight of different figurative reasoning patterns only becomes possible on the basis of an assessment of the degree of conventionality of linguistic metaphors and metonyms. To approach these questions, it is first necessary to provide an overview of the main claims of the relevant theoretical frameworks, i.e. the conceptual theory of metaphor and metonymy and the theory of conceptual blending, which will be done in the following chapter. As some of the major insights commonly ascribed to these theories are, however, not as unique to them as usually assumed, the chapter will start with a brief consideration of what other, earlier, and partly still influential theories of figurative language have to say with regard to the relationship between figurative language, thinking, and conventionality. Section 3.1 thus serves as background for the following discussion of the conceptual theories by pointing out similarities with and differences to later cognitive approaches. 22

24 3 Metaphor, metonymy, and cognition 3.1 Metaphor and metonymy in pre-cognitive linguistics Traditionally, metaphor was regarded as a matter of words, as a mere substitution of one word for another that is based on the criterion of similarity. However, in Aristotle s description it is already evident that figurative language is at least to a certain degree a necessity, and not only an ornamental feature of language. Aristotle admits that there are cases where gaps in literal language are filled by metaphor (Leezenberg 2001: 35-39). Moreover, he does not deny that metaphor is used in everyday language and by the average language user. Still, his focus is clearly on poetic, i.e. unconventional examples, as these achieve the greatest rhetorical effects. This focus is probably responsible for the fact that Richards (1936), who developed the so-called interaction theory of metaphor, begins his account with a severe criticism of the Aristotelian tradition. He points to the evil presence of three [...] assumptions which have ever since prevented the study of this greatest thing by far [i.e. metaphor; SH] from taking the place it deserves (1936: 89): 1. Good metaphors can only be made up by geniuses who have a special eye for resemblances. 2. Individuals cannot learn how to use metaphor. 3. Metaphor forms the counterpart to normal language use, is therefore exceptional. For Richards, metaphor is ubiquitous so that [w]e cannot get through three sentences of ordinary fluid discourse without it (1936: 92). As he (1936: 118, 130) points out, metaphorical language can become part of the lexicon, can become a conventional way of expressing ideas. This, however, does not alter the status of the expression as a figurative one. Moreover, metaphor is said to determine our thinking to a large extent (Richards 1936: 92, 94; cf. also Black 1977: ): [W]hen we use a metaphor we have two thoughts of different things active together and supported by a single word, or phrase, whose meaning is a resultant of their interaction. (Richards 1936: 93) The interaction view 12 holds that much is lost if a metaphor is translated into a literal statement (Black 1962: 46, 1977: 441), as metaphors make us see things in a new light. This is the result of the interaction between the two constituents of a metaphor, called tenor and vehicle by Richards (1925, 12 More recent versions of the interaction view can be found in Indurkhya (1992) or Kittay (1987). 23

25 1936). 13 Metaphor is considered a double unit in which knowledge commonly associated with the vehicle is projected onto the tenor. Which parts are selected depends on the speakers knowledge about the tenor. As a result, the tenor is linked to selected aspects of the vehicle and acquires novel semantic aspects. But the vehicle is also changed as the aspects used in the metaphor have to be aligned with the tenor. Consider example (8): (8) Man is a wolf. (Black 1977: 433) The speakers knowledge of wolves consists of a whole array of different features. They are animals, have four legs, have a fur, live in groups, are often considered malicious and aggressive, etc. From all these different aspects, only one is chosen when wolf is connected to man. This changes the structure of the concept WOLF significantly. But the selected facet cannot be applied to the tenor in a straightforward way. The mental picture of aggressive animals has to be aligned with the knowledge about the tenor: If humans are metaphorically conceptualized as wolves, the behavioural patterns conjured up are certainly different from those of the vehicle. Interaction theory can certainly be criticized because of the unclear idea of the central term interaction and its lack of examples (cf. Jäkel 1997: ). Nevertheless, the theory anticipates some of the most vital ideas for which the conceptual metaphor theory has become famous, although proponents of the conceptual view may be reluctant to accept this (cf. Lakoff and Turner 1989: ): Metaphorical language and thought are not considered dispensable. Instead it is argued that metaphors are cognitive instruments (Black 1977: 454) which make us perceive connections that are otherwise not present. The notion of metaphor is widened to include those processes in which we perceive or think of or feel about one thing in terms of another (Richards 1936: 116). Within pragmatically oriented approaches, exemplified by Searle s (1979) account here, cognitive aspects, evidently, play a minor role. Searle (1979: 83) defines metaphors as follows: a speaker utters a sentence of the form S is P and means metaphorically that S is R. As metaphorical utterances are characterized by the difference between what is said and what is meant, analyses of metaphors consist of two sentences: the metaphorical utterance and a literal paraphrase (Searle 1979: 81-82). (9) (MET) It s getting hot in here. (PAR) The argument that is going on is becoming more vituperative. (Searle 1979: 82) However, Searle is well aware that literal paraphrases are often inadequate (1979: 82) and that there are cases where no paraphrase can be 13 In Juliet is the sun!, quoted as example (1) above, Juliet is the tenor and sun the vehicle. Black, the second major early proponent of the interaction theory, uses different terms. Richards s tenor is called principal subject (1962) or primary subject (1977) and the vehicle is called subsidiary subject (1962) or secondary subject (1977). 24

26 given because the language does not offer any literal way of expressing certain ideas (1979: 83). Figurativity is, thus, a necessary part of language. The relations between P and R are diverse. Searle (1979: ) offers a list of eight principles, which is, as he points out (1979: 107), potentially incomplete. Interestingly, this list also embraces metonymic relations: When one says, S is P, and means that S is R, P and R may be associated by such relations as the part-whole relation, the container-contained relation or even the clothing and wearer relation (Searle 1979: 110). As far as metaphor proper is concerned, R can, for example, be a characteristic property of P things (e.g. Sam is a giant for Sam is big ). Or P things can be stereotypically believed to be R (e.g. Richard is a gorilla for Richard is mean, nasty, or prone to violence ). Some metaphors, he emphasizes, do therefore not rely on any relation between P and R that exists in the extralinguistic world. Sometimes it is just a matter of perceptions, sensibilities, and linguistic practices (1979: 97) that speakers see a connection between P and R. Searle quotes the example of coldness, which is conventionally associated with a lack of emotion, stresses that this metaphor is not culturespecific and that it leads to many expressions, like heated argument, warm welcome, and sexual frigidity (Searle 1979: 98), that are connected by the association of temperature with emotions (cf. also Leezenberg 2001: 122). Searle calls such instances dead metaphors, but for him they are still worth considering: Dead metaphors are especially interesting for our study, because, to speak oxymoronically, dead metaphors have lived on. They have become dead through continual use, but their continual use is a clue that they satisfy some semantic need. (Searle 1979: 83) Even though his approach is fundamentally pragmatic, it cannot be denied that metaphors and metonymies can and do play a role for the lexicon of a language. Figurative meanings can become established semantic interpretations of words, or, to put it differently, conventional. As this short overview has shown, some of the central ideas of the conceptual theory of metaphor are anticipated in earlier theories: First, metaphorical language cannot be entirely avoided, since we sometimes lack literal expressions. Thus, metaphors have, second, a firm place in everyday language, and figurative meanings can and do become conventional. Third, metaphors can go beyond language and play a role in thinking. Fourth, the relationship between the two constituents of a metaphor, whether called tenor and vehicle, or P and R, is sometimes only perceived because of the metaphor, and not otherwise given or extant. And fifth, the relationship between these constituent parts can extend to larger coherent systems of associations. The role of metonymy, however, is grossly underestimated in earlier research. It is often treated as a sub-type of metaphor (cf. Searle s approach), and typically only mentioned in passing, even though Bredin (1984: 45), for 25