A Semantic Map Approach to Adverbial Clauses

A Semantic Map Approach to Adverbial Clauses

Danksagung

An dieser Stelle möchte ich mich zuerst bei meiner Familie bedanken. Ihre Unterstützung in jeder Hinsicht hat mir ein stets sorgenfreies Studieren ermöglicht. Besonders ihr Interesse an meinen Studieninhalten und auch die Fragen über mein Forschungsthema waren für mich immer wieder Motivation und Quelle neuer Ideen. Durch meine Familie habe ich meine Wissbegierde und die Hingabe entwickelt, immer mein Bestes zu geben.

Außerdem möchte ich meinen Freunden danken. Nur durch Diskussionen mit ihnen, durch gemeinsame Gedankenspiele, aber auch durch Unternehmungen und gemeinsames Lachen hat diese Arbeit ihre Form erhalten. Besonders dankbar bin ich Frank Heyne: seine unermüdliche Hilfe beim Korrigieren der Arbeit und vor allem seine motivierende Unterstützung bleiben unersetzlich.

Mein ganz besonderer Dank aber gilt meinen Betreuern Professor Holger Diessel und Doktor Florian Haas. Ich konnte mich jederzeit mit Fragen und Überlegungen an sie wenden und sie haben mich immer in meinen Ideen unterstützt. Sie haben meine Begeisterung für die Linguistik geweckt.

Table of contents 1

Introduction............................................................................................................................. 3

2

Theoretical preliminaries ........................................................................................................ 5

2.1

Conceptual spaces in psychology ........................................................................................ 5

2.2

What is a semantic map? .................................................................................................... 8

2.3

The geometry of semantic maps ......................................................................................... 9

2.3.1

Constructing semantic maps ......................................................................................... 10

2.3.2

The role of contiguity .................................................................................................... 11

2.4

Fields of application........................................................................................................... 13

2.4.1

Synchronic uses ............................................................................................................. 13

2.4.2

Diachronic uses.............................................................................................................. 17

2.5

Limitations of the semantic map approach....................................................................... 19

2.6

Statistical Maps – the solution? ........................................................................................ 22

3

A semantic map approach to concession.............................................................................. 23

3.1

What is concession? .......................................................................................................... 26

3.1.1

Grammar definitions of concession .............................................................................. 26

3.2

Real concession ................................................................................................................. 28

3.3

Hypothetical concession.................................................................................................... 31

3.3.1

Subtypes of hypothetical concession ............................................................................ 32

3.3.2

Source constructions of hypothetical concessive clauses ............................................. 33

3.4 4

Concession and cause........................................................................................................ 35 Functional distinctions of concession ................................................................................... 41

4.1

Real concession ................................................................................................................. 42

4.1.1

Ineffective anticause ..................................................................................................... 42

4.1.2

Restriction ..................................................................................................................... 44

4.1.2.1

German obwohl – on its way to a discourse marker? ................................................... 46

4.1.2.2

Corrective function of concession ................................................................................. 47

4.1.2.3

Attenuating a rejection – a function of disagreement? ................................................ 49

4.1.3

The relation between restriction and correction .......................................................... 50

4.1.4

Contrast and antithetic concession ............................................................................... 50

4.1.5

Einräumung – a function of disagreement .................................................................... 52

4.1.6

Observations.................................................................................................................. 55

4.3 4.3.1

Hypothetical concession.................................................................................................... 56 Scalar concessive conditionals ...................................................................................... 57

4.3.1.1

Formal description..................................................................................................... 58

4.3.1.2

Functional description ................................................................................................... 59

4.3.2

Alternative concessive conditionals .............................................................................. 60

4.3.2.1

Formal description........................................................................................................ 60

4.3.2.2

Functional description ............................................................................................... 61

4.3.3

Universal concessive conditionals ................................................................................. 62

4.3.3.1

Formal description......................................................................................................... 62

4.3.3.2

Functional description ................................................................................................... 65

4.4 5

Observations...................................................................................................................... 65 The geography of the conceptual space of concession ........................................................ 65

5.1

Description and analysis of the language-specific semantic maps ................................... 71

5.1.1

Hypothetical concession................................................................................................ 71

5.1.2

Real concession ............................................................................................................. 74

5.2

Observations and deductions ............................................................................................ 76

5.3

Discontiguity on the German map .................................................................................... 81

6

Diachrony of adverbial concessive conjunctions .................................................................. 83

6.1

The role of grammaticalisation ......................................................................................... 84

6.1.1

What is grammaticalisation? ......................................................................................... 84

6.1.2

Principles of grammaticalisation ................................................................................... 87

6.2

Sources and origins of concessive conjunctions ............................................................... 90

6.2.1

Diachrony of concessive conditionals ........................................................................... 90

6.2.2

Diachrony of concessive conjunctions .......................................................................... 92

6.3

Developmental paths to concessivity................................................................................ 94

6.3.1

Semantic loss vs. semantic enrichment ........................................................................ 96

6.3.2

Three tendencies of grammaticalisation ....................................................................... 97

6.4 7

Diachronic concession on the semantic maps .................................................................. 98 Conclusion ........................................................................................................................... 102

The data of the language sample ................................................................................................ 105 Bibliography................................................................................................................................. 122

1

Introduction In language after language, independently of genetic affiliation and areal location, we find similar grammatical meanings (or functions, or uses), similar patterns of multifunctionality, and similar expression types. Of course, there are also significant differences between languages of different types, but this does not detract from the overall universal pattern. (Haspelmath 1997: 60) This paper deals with a comparatively recent approach to various phenomena of linguis-

tics: the semantic map approach. The technique derives from cognitive psychology and is widely applicable not only in the field of linguistics but also in psychology or computer science. However, a closer look at semantic maps, conceptual spaces and their characteristics reveals that the approach it not so recent after all. In fact, already Aristotle developed similar mechanisms in the Aristotelian Square of Opposition. So what is it about this technique that makes it so persuasive that essentially, it proved useful not only over decades but centuries? What are semantic maps good for? The major use of conceptual networks and semantic maps is the visual representation of abstract concepts, entities or processes. In the field of linguistics, semantic maps are thus an extremely helpful tool. Especially the semantics of linguistic entities are often difficult to capture because one entity might show different meanings in different surroundings. Linguistic meanings may change over time. Hence, semantic maps turned out to be especially useful for the description and illustration of the semantics of grammatical entities of any kind, such as the modality of verbs or the indefiniteness of pronouns. As grammatical markers are often multifunctional, semantic maps are helpful in order to investigate the different functions in more detail and to illustrate the interrelations between the functions. Thus, they provide a basis for further linguistic and psychological research. As the introductory quote implies, semantic maps additionally developed in an effective means to solve one of the greatest mysteries about language: are there universal structures which can be found in all languages of the world? There are over 6000 languages spoken around the world and for decades, linguists have tried to find similarities and differences between them. They have been most interested, of course, in the similarities because those indicate that there must be something universal, some similar structures that are inherent to all languages. In this field, semantic maps turned out to be an excellent tool to illustrate and 3

compare the various markers and functions of a grammatical concept throughout different languages of different language families. The language-specific distributions of grammatical markers on the semantic maps reveal interesting differences and also similar tendencies which can be assumed to be universal. Linguistically, semantic maps may give insights into cross-linguistic variation and universal tendencies throughout languages. Thus, they may function as a basis for psychological considerations concerning the possible interrelations of functions. They may even serve a diachronic function when we assume the interrelations to be developmental paths within the overall meaning of a linguistic entity. The present paper applies the method of building and analyzing a semantic map to the grammatical category of adverbial clauses, more specifically to adverbial clauses of concession. In a first step, I will explain the technique of creating and reading semantics maps and their underlying conceptual foundation. Then, I will turn to adverbial concession. For the development of a semantic map, it is necessary to define the grammatical category of concession and its markers in detail. From there, I will develop the different semantic-pragmatic functions of concession that are inherent to the category. These functions will then be arranged in a network, the conceptual space of adverbial concession. In a last step, the concessive conjunctions of four sample languages will be distributed on the map. The sample of languages is restricted to English, German, French and Latin. Although they are all European languages, they belong to different branches of this language family. The emphasis of the analysis is on English and German. French and Latin function as complementary sources and lead to a balanced sample. Due to the genetic affiliation between German and English, I expect close similarities in the distribution of the markers on the conceptual space. As Latin is the ancestor of Romance languages, I expect a diachronic development from Latin concessive conjunctions to French markers. By comparison, the distributions of the concessive conjunctions on the conceptual space will exhibit similarities and differences. Thus, it will reveal cross-linguistic tendencies both in the synchronic prototypicality and in the diachronic development of the functions. In summary, on the basis of the semantic map approach, I will develop a functional network of adverbial concession. Therefore, I will derive the necessary functions of concession in order to be able to build the semantic map. Then I will analyse the semantic map in order 4

to demonstrate that there are universal aspects and tendencies within the category of concession across languages. Thus, I will investigate if there is a prototypical, universal function of concession and compare the morphosyntactics of concessive markers as well as their semantic ranges. Eventually, I will derive diachronic, possibly universal mechanisms of how concessive semantics develop in the adverbial concessive category.

PART I 2

Theoretical preliminaries In the last years of linguistic research, semantic maps proved to be a valuable tool in

many fields of linguistics. Especially in the area of typology, they offer the means for fruitful cross-linguistic comparison with the prospect of arriving at language universals. Besides their application in the field of typology and in the search for language universals, semantic maps are considered useful in finding answers for the questions of how our conceptual system is structured and how linguistic knowledge is organised and ordered; thus, they also function as an important technique in the fields of cognitive and psycho- linguistics. However, semantic maps do not only lend themselves to synchronic descriptions of linguistic data, they may also be used for the representation of the diachronic development of meaning. But what exactly are semantic maps and what makes them so widely applicable? Before I turn to the investigation and possible answers to these questions, I will have a closer look at the psychological foundations of semantic maps which Peter Gärdenfors laid in his work on conceptual spaces.

2.1 Conceptual spaces in psychology In his book Conceptual spaces: the geometry of thought, Peter Gärdenfors investigates the question how representations in cognitive science should be modelled. He opts for a topological and geometrical mode which leads to the development of conceptual spaces for very different fields of cognitive science, including the field of linguistics and semantics. Other areas which may find the theory of conceptual spaces useful are psychology, computer science and philosophy. Gärdenfors adopts a cognitive point of view when he develops conceptual spaces, trying to create a viable method for different fields of application. Thus, 5

he defines a conceptual space in the broadest terms as a “framework for representing information on the conceptual level (Gärdenfors 2000: 2).” in conceptual spaces, there exists a close connection between relative distance and similarity; a small distance illustrates great similarity between two entities. Consequently, similarity and relatedness of entities can be defined via distance and proximity in the representation on conceptual spaces. Gärdenfors repeatedly emphasises that conceptual spaces are a mode of representation, a “framework for cognitive representations (Gärdenfors 2000: 30)”, which in a next step “can then be turned into empirically testable theories or constructive models (ibid).” So according to him, conceptual spaces can help to construct artificial representations with which researchers can approach cognitive tasks, problems and phenomena. Thus, conceptual spaces are predominantly a visualisation tool (cf. Gärdenfors 2000: 30). They describe how representations can be constructed; they do not point out cognitive processes, developmental processes or types of interrelations. These kinds of information need to be investigated on the basis of deductions from conceptual spaces. According to Gärdenfors, conceptual spaces are theoretical in nature (cf. Gärdenfors 2000: 31); they are not naturalistic, real representations but instruments for further analysis, explanations, and predictions on psychological, philosophical and linguistic questions (cf. ibid.). As will be pointed out later, semantic maps underlie the same functional limitations. Due to the topic of this paper and its restrictions, I will now have a closer look at Gärdenfors’ approach to the topic of semantics and how his psychological theory of conceptual spaces can be useful in the area of linguistics. Gärdenfors defines semantics as “relationship between words/expressions of a language and their meanings (Gärdenfors 2000: 151).” However, this relation between form and meaning needs yet to be explained. Firstly, it is necessary to identify what meaning actually is. According to Gärdenfors, meanings can be realist or cognitive; in a realist approach, they are referential to entities in the real world. In a cognitive approach, meanings are mental entities; this approach leads to the theory of cognitive semantics and is closely tied to the theory of conceptual spaces (Gärdenfors 2000: 152ff.). As semantic maps and conceptual spaces are predominantly used to visualise abstract grammatical meanings and their interrelations in linguistics, the cognitive approach is somewhat more convincing because abstract grammatical meanings are difficult to relate to real entities in the world. As third approach to meaning, Gärdenfors offers the so-called functional approach. Here, linguistic meaning describes the communicative function of an 6

expression or the contribution to the communicative function of the utterance in which the expression occurs. In other words, meaning equates function. This seems to be a useful approach to grammatical entities and constructions as well. In his work, Gärdenfors focuses on the theory of cognitive semantics in order to indicate the usefulness of conceptual spaces in the field of semantics and linguistics. According to the conceptualist (i.e. cognitive) approach, semantics describe the mapping between an expression and the underlying conceptual structure or mental entities. This mapping occurs along the lines of associations which are established and extended when we learn a language. Additionally, the mapping associations undergo diachronic variation both in the individual and in the general language use (cf. Gärdenfors 2000: 154). Gärdenfors then identifies the conceptual structure to which linguistic expressions are mapped as the conceptual space (cf. ibid: 159). Consequently, Gärdenfors can characterise meanings as embodied mental constructs and fundamentally perceptual structures which are based on our every-day experiences. In a next step, semantics are considered as primary to syntax, which is the opposite stance to generative grammarians such as Noam Chomsky who consider semantics as secondary at best and in no way related to the grammatical system. However, Gärdenfors explains that if one considers meaning to be conceptual, these conceptual representations exist before any syntactic expressions are formulated (cf. Gärdenfors 2000: 165). Thus, he implicitly addresses the question of language universals. According to William Croft (2001, 2006), formal structures are only rarely universal across languages. However, the underlying meanings and conceptual structures are far more likely to be universal; in fact, they are the only element that can be universal cross-linguistically, at all. Finally, Gärdenfors allows semantics to include aspects of use or pragmatics (cf. Gärdenfors 2000: 165). Especially for grammatical entities, this is very important as their meaning is usually inseparable from the meaning of the whole utterance. In this regard, Gärdenfors prefers ‘function’ over ‘meaning’ as it does not imply a purely semantic approach to meaning but rather a wide-ranging notion which includes both meaning and use. In his work Conceptual spaces: the geometry of thought, Gärdenfors shows how deeply the semantic map approach to linguistic phenomena is rooted in cognitive semantics. This cognitive perception of semantics has implications for the notion of syntax, language acquisition, the mental linguistic reality of human beings and the notion of meaning itself. Gärdenfors establishes the basic conceptual nature of conceptual spaces when he defines them as 7

mechanism of representing cognitive processes and entities. In the chapter on semantics, he not only proves the usefulness of conceptual spaces for illustrating the meanings of words, he also argues for a generally cognitive approach to semantics and grammar. This allows their representation in semantic maps and, more importantly, sees meaning as an integral part of grammar which influences both the grammatical form and syntax of languages.

2.2 What is a semantic map? According to Martin Haspelmath, a semantic map is a “method for describing and illuminating the pattern of multifunctionality of grammatical morphemes” (Haspelmath 2003: 213). Multifunctionality (also polyfunctionality or macrofunctionality) describes the phenomenon in which one grammatical morpheme, i.e. a representative of the grammatical category visualised with the help of a semantic map, has multiple functions. Consequently, semantic maps only work on the precondition that one linguistic entity has numerous meanings or better: functions. Multifunctionality occurs frequently with function words and grammatical affixes. During their development from lexical items to grammatical items, i.e. during the process of grammaticalisation, they usually get semantically bleached, which leads to more abstract and more general meanings; as a result, they may have multiple functions (cf. Haspelmath 2003: 211). It is necessary for the formation and analysis of semantic maps that these different functions of a grammatical category must be semantically related and that they constitute a network of interrelatedness (cf. Narrog/van der Auwera: 2; Croft 2001: 96). In line with de Haan (2010), a semantic map is thus a network of all functions and all related meanings of one linguistic category; it links the formal aspects of this category to its various semantic aspects, i.e. its uses and functions. As has been indicated above, the basis for the semantic map approach to linguistic data is multifunctionality. There are different options to approach this phenomenon: the monosemist, the polysemist and the homonymist approach. According to the first option, linguists assume a single abstract general meaning of a grammatical marker whose finer functions arise from the context in which this marker is used. In contrast, the polysemist approach differentiates between several meanings and functions of one grammatical entity but sees these meanings as related. Consequently, the formal similarity of multifunctional words is considered as motivated by the relationship between the functions. The homo8

nymist approach does not so much assume multifunctionality of a single linguistic item; it rather sees all the different functions as instances of individual linguistic entities. According to this assumption, these entities share their similar form only by chance (cf. Haspelmath 2003: 212f.). For semantic maps, the distinction between the first two approaches to multifunctionality is irrelevant as both assume a basic, systematic interrelatedness between the different functions of a grammatical marker. Homonymous relationships cannot be represented on semantic maps as these maps rely on similarity in meaning and use of the particular construction or marker in question. This shows how the semantic map approach is bound to the notion of similarity, which in this context refers to the relatedness of functions of a grammatical category in question; similarity also plays a crucial role in the formation and the geometry of semantic maps (cf. Haspelmath 2003: 215f.). In sum, semantic maps are geometrical representations in which the functions of a grammatical category are located on a conceptual space1 (cf. Haspelmath 2003: 213). Thus, semantic maps are a visualisation technique; they constitute a model of linguistic reality and lead to an abstract representation of the meanings (or better, the functions) of a grammatical category.

2.3 The geometry of semantic maps As “[a] semantic map consists of a number of grammatical functions [of one grammatical category] plus a means to link these functions together (de Haan, in progress: 2)”, the geometry of semantic maps and conceptual spaces defines how the similarity, i.e. the relatedness, of functions is expressed and represented on these spaces (cf. Zwarts 2010: 377). Generally, there are two ways of depicting interrelations between functions: either by presenting them as connected with lines or arrows which “denote the proximity of grammatical functions” (de Haan, in progress: 3) or by expressing the relatedness with the help of proximity and relative distance. This differentiation leads to two different types of semantic maps: to classical semantic maps which resemble a network of related functions (also called implicational maps by Martin Haspelmath, in 2003: 213) and to statistics-based maps which 1

In the distinction between ‘semantic map’ and ‘conceptual space’, I follow William Croft (2001). In his work on Radical Construction Grammar, he differentiates between conceptual space as “a structured representation of functional structures and their relationships to each other” (Croft 2001:93) and semantic maps as languagespecific distributions of grammatical markers on this space.

9

employ multi-dimensional scaling (MDS) in order to create a pattern of functional distribution of grammatical markers in a diagram (cf. for example Wälchli 2006/2007, Zwarts 2010). However, not only MDS maps use proximity and distance in order to express varying degrees of similarity between meanings – classical semantic maps indicate the relationship between functions in their arrangement of conceptual space, too. The closer two functions occur together, the more similar these two functions are; in other words: two immediately adjacent functions are more similar and more directly related than two functions that display intermediate functions (cf. Haspelmath 2003:215f.).

2.3.1

Constructing semantic maps

On a classical semantic map and its underlying conceptual space, the individual functions of a grammatical category are depicted as the “nodes” or “values” (Haspelmath 2003: 214) of the semantic network. These nodes are connected with lines which are supposed to show the relatedness and contiguousness of the functions. The resulting network represents the conceptual space of one particular grammatical category (cf. Haspelmath 2003:213). predicatiive possessor

external possessor

direction

recipient

beneficiary

purpose

experience

judicantis

Fig.1 The conceptual space of the dative (Haspelmath 2003: 213).

Before, however, the functions can be brought onto the map and be connected, it is necessary to choose the relevant functions. In order to be represented on the conceptual space, a function must be primitive, i.e. it cannot be further subdivided into other functions. If, for example, two assumedly different functions are never expressed by two different grammatical markers, neither language-specifically nor (more importantly) crosslinguistically, then one should rather subsume these two functions under just one (cf. de Haan, manuscript: 5). The other way around, a function is put on the map if there is at least one pair of languages which differs with regard to the existence of this function and/or its formal marking. In order to construct a comprehensive network, a larger number of languages need to be analysed for the respective functions of the grammatical category in 10

question. According to Martin Haspelmath, 12 languages suffice in order to create a stable conceptual space (cf. Haspelmath 2003: 217). However, one of the shortcomings of the semantic map approach is its vulnerability: if only one language does not support the choice and arrangement of the functions on the conceptual space, the whole map is basically worthless and devaluated (cf. ibid.). The arrangement of the functions chosen is another important aspect of constructing a semantic map because the functions must be arranged in such a way that all multifunctional grammatical markers may cover a contiguous area on the conceptual space (cf. Haspelmath 2003:217; cf. Zwarts 2010: 379). This preliminary is formulated as the principle of contiguity or the Semantic Map Connectivity Hypothesis (cf. Croft 2001): it states that language-specific grammatical markers should always map onto a connected, contiguous area of conceptual space (cf. Croft 2001: 96). Thus, the underlying construction of a conceptual space has to permit this contiguous mapping. In line with William Croft, this further illuminates the distinction between conceptual space and semantic map: while the conceptual space is the underlying network of arranged functions, a semantic map is the representation of a language-specific marker on this network. The marker’s distribution across the functional network is indicated with closed curved lines around the functions expressed by it (cf. Haspelmath 2003: 213). If this procedure is repeated with several markers of one specific language which all encode the grammatical category described by the conceptual space, the different resulting semantic maps may already give interesting insights into semantic development and grammaticalisation processes; they may also be used as starting point for further analyses on other aspects such as pragmatics or synchronic language variation. Most importantly, however, semantic maps prove to be useful tools for cross-linguistic comparisons: the analysis of several different languages reveals both semantic and formal particularities.

2.3.2

The role of contiguity

For the construction of semantic maps and their underlying conceptual spaces, the principle of contiguity is of major importance (compare 2.2.1). The mappings of language- specific grammatical markers have to cover a contiguous area on the conceptual space; thus, the underlying conceptual space has to permit the contiguity with its structure. However, this principle of contiguity may be disrupted. This is the case when linguistic data of one or 11

several languages do not support the architecture of a conceptual space. Hence, the mapping of the language- specific semantic maps is not contiguous. This phenomenon may have different reasons: contiguity may be disrupted because a new marker enters the language and the map right in the middle (that is, in just one function) and does not extend its use from one original function to other related ones. Thus, the intervening marker may disrupt the contiguous mapping of other grammatical markers of that language and take over the function formally expressed by the replaced marker (cf. Haspelmath 2003: 235ff). Usually, this phenomenon renders the conceptual space and the semantic maps worthless (cf. Haspelmath 2003: 235f) because gaps or discontiguities are normally not allowed in the semantic map approach. They rather trigger a different arrangement of the functions on the conceptual space (cf. Zwarts 2010: 383). In a less deontic way, one could say that exceptions are not realised on classical semantic maps but always entail a change in the arrangement of the functions. However, discontiguity is a phenomenon which can be repeatedly encountered. It may disrupt an established conceptual space at any time during the analysis of further linguistic data for either a more fine-grained conceptual space or for the mapping of language- specific grammatical markers. In fact, the phenomenon of discontiguity lies at the bottom of the vulnerability of the semantic map approach as its existence usually invalidates both the maps and all findings and claims based on them. Consequently, it might be worthwhile to rethink the inviolability of the principle of contiguity. The modality map of van der Auwera and Plungian offers such a new approach to the question of how to deal with discontiguity which could alleviate the vulnerability of semantic maps. This particular map includes the language- specific distribution of Dutch mogen which shows gaps in its function mapping (cf. Zwarts 2010: 386; van der Auwera/Plungian 1998). According to van der Auwera und Plungian, the gaps in the semantic map of Dutch modality are instances of diachronic change: once, Dutch ‘mogen’ had had these functions but lost them which led to the discontiguous distribution (cf. van der Auwera/Plungian 1998: 34ff.). This example shows how diachronic explanations can account for discontiguity and may even render it acceptable within the semantic map approach. Indeed, the diachronic loss of functions even from the middle of the map might not be that rare. However, if one takes diachronic information into consideration in order to account for gaps in semantic maps, linguists can no longer rely on synchronic linguistic data alone (which demands the 12

rearrangement of functions in case of discontiguity). Rather, the semantic map approach needs to complement the typological data with historical data and with independent semantic analyses (cf. Zwarts 2010: 386). According to Haspelmath, repeated discontiguity may not only lead to new approaches and constraints concerning the construction and analysis of conceptual spaces and semantic maps if discontiguity turns out to be a phenomenon that cannot be avoided. Repeated discontiguity may also challenge the idea of grammaticalisation as coherent development of a grammatical marker by extending its meaning and function (cf. Haspelmath 2003: 237).

2.4 Fields of application The necessity of using cross-linguistic data to construct stable conceptual spaces and the promising outlook for cross-linguistic comparison with the help of semantic maps show that both conceptual spaces and semantic maps are tightly bound to the field of typology. However, besides this field of linguistics, the semantic map approach also proves useful for cognitive linguistics, historical linguistics or cognitive psychology. As a result, one can distinguish between synchronic and diachronic uses of semantic maps.

2.4.1

Synchronic uses

Semantic maps are foremost a method of typological linguistics and predominantly used for cross- linguistic comparison. Already the choice and arrangement of the functions in question is based on linguistic data from several different languages: this reveals the fundamentally typological character of the semantic map approach. However, not only the preliminaries of their construction are of typological nature: semantic maps offer an interesting answer to the question if and to what extent language universals exist. When examining a particular grammatical category, it becomes obvious that different languages employ different formal markings and constructions in order to encode the meaning of the category. Different languages do not only use different formal markings, they also use a varying number and different strategies of marking. Thus, it is hard to find universals in grammatical categories and their different realisation forms (cf. van der Auwera/Temürcü 2006:131). Still, all languages are thought to be different instantiations of ‘the human language’. Thus, it is not 13

surprising to find that cross-linguistic variation is not random but shows recurring variation patters- a fact which led typologists to believe that there must be an underlying universal grammar to all languages (cf. Haspelmath 1997: 7). However, this theory was challenged by William Croft with his Radical Construction Grammar2 (Croft 2001). One of his central claims is that language universals indeed cannot be found on the level of grammar because they simply do not exist on this level. According to Croft, grammatical categories and their formal realisations will always be language- specific (cf. Croft 2001: 105). What he considers universal is the semantics of linguistic patterns, the arrangement of functions underlying each language-specific semantic map, that is: the conceptual space (cf. Croft 2001: 103f., see also Haspelmath 2003: 213). The languages of the world draw on the same semantic distinctions within grammatical categories although their formal realisations may be highly different and their distribution on the conceptual space considerably diverse (cf. Croft/Cruse 2004: 321). predicatiive possessor

external possessor

direction

recipient

beneficiary

purpose

experience

To (English) judicantis

Dative (German)

Fig. 2 A semantic map of dative functions/ the boundaries of English to and German Dative (cf. Haspelmath 2003: 213 and 219).

So if the grammatical markings of several languages are mapped onto the conceptual space of one grammatical category, the language-specific distributions permits typological insights on the basis of cross-linguistic comparison, such as the recognition of similarities (indicated in overlapping graphs on the map) and differences in marking strategies, genealogical relations between languages, the functional limitations of grammatical markers and 2

Radical Construction Grammar (RCG) with its typological orientation is a branch of linguistics for which semantic maps have proven especially fruitful. It “was developed in order to account for the diversity of the syntactic facts of a single language as well as the syntactic diversity of the world’s languages.” (Croft 2001: 3) RCG radicalizes the propositions of Construction Grammar but is widely compatible with Langacker’s Cognitive Grammar. It is special that RCG considers constructions of any kind as the primitive elements of language while grammatical categories such as syntactic elements derive from their function in a construction. Thus, RCG is strictly non-reductionist. Additionally, RCG denies the existence of syntactic relations and substitutes them by symbolic relations. Consequently, linguistic structures cannot be universal, but the underlying conceptual (symbolic) space is (cf. Croft 2001).

14

the detection of implicational hierarchies and universals (cf. Croft 2001: 103; cf. Haspelmath 2003: 230ff.). Especially implicational universals are of major importance in the typological use of semantic maps. Implicational universals take grammatical variation in account as they do not define closed grammatical categories with particular theoretical characteristics which necessarily have to turn up in all languages of the world. Rather, implicational universals describe “a property that holds in all languages that have a given property (Haspelmath 1997: 8).” In other words, if a language shows a particular property B which implies property A, one can assume that this particular language also shows property A and that there is a relation between the two properties. For instance, all languages that employ the case of nominative also employ the accusative while all languages with ergative case markings also show absolutive case endings. This leads to another implicational universal, namely that if a language marks S (the subject of intransitive sentences) and A (the subject of transitive sentences) similar, it will mark P (the object of a transitive sentence) differently. If, however, a language marks S and P similar, it will mark A differently.

intransitive event transitive event

intransitive event transitive event

A

S

conceptual space

A

S P

intransitive event transitive event A

S P

nominative-accusative pattern

P ergative-absolutive pattern

Fig. 3 Conceptual space and semantic maps of grammatical relations and their marking

Eventually, implicational universals are a mechanism of generalisation as they rely on implicit assumptions and deductions from observable linguistic data. Here, semantic maps prove to be a useful tool. Implicational universals are closely tied to the principle of contiguity as it plays an important role in the deduction of implicational hierarchies. When the mappings of various languages show a similar contiguous mapping on particular functions, this pattern implies that all languages showing one function on this contiguous area (or two functions on this contiguous area separated by intermediate ones) will also employ the other, intermediate functions (cf. Haspelmath 2003: 20). It is, however, necessary to assume contiguity to be a rule without exceptions; only then have semantic maps a predictive power concerning the number and sequence of functions which grammatical markers may 15

cover cross-linguistically, even without sufficient linguistic data (cf. van der Auwera/Temürcü 2006: 132). Semantic maps do not only provide an answer to the question what may or can be universal in linguistics or function as a useful tool for cross- linguistic comparison in typology. Semantic maps also allow sidestepping the distinction and vague continuum of polysemy and homonymy. The identification of one central prototypical meaning is no longer necessary as neither conceptual spaces nor semantic maps distinguish between default and peripheral functions but depict all, regardless of their frequency. However, semantic maps are not only a visualisation tool for typological distinctions; linguists and psychologists also hope to find answers in semantic maps and conceptual spaces about how linguistic knowledge is organised in our conceptual system (cf. Haspelmath 2003: 232). So next to offering insights into typological questions, semantic maps are also hoped to be useful means to representing cognitive structures and answering questions from cognitive linguistics, psycholinguistics and cognitive psychology. As semantic maps and their underlying conceptual spaces base on the concept of multifunctionality, they represent several functions of one grammatical category which all have a similar meaning; the functions and their shades of meaning are perceived as related by the speakers. Consequently, a conceptual space with all its language-specific semantic maps is a visualisation of different functions and their connections and interrelations in the human mind (cf. Haspelmath 2003: 233). Both linguists and psychologists hope that the arrangement and configuration of these functions on the conceptual space correlates with the cognitive structuring of meaning (cf. Haspelmath 2003: 219). Thus, William Croft describes conceptual space as “a universal structure of conceptual knowledge for communication in human beings” (Croft 2001: 105) and even more clearly as “a common human cognitive heritage, indeed the geography of the human mind (Croft 2006: 138ff).” If one assumes that a conceptual space represents a linguistic universal, its structure and composition may very well correlate with that of the human mind and its linguistic capacities. Thus, semantic maps are also an important tool for psychological analyses trying to find answers to the question of what the mental reality of human beings is (cf. Haspelmath 2003: 239). As a result, semantic maps need psychological analysis additionally to linguistic examinations. Their relevance for various disciplines shows how widely applicable semantic maps are – they function as basis for investigation, as visualisation tool and as working hypothesis in typological and cognitive linguistics as well as in psychology. 16

2.4.2

Diachronic uses

Semantic maps do not only give a graphical overview about multiple functions of linguistic constructions across languages, they are also a tool for visualising diachronic change and grammaticalisation paths. According to Haspelmath, the principle of contiguity is again of central importance for a diachronic application of conceptual spaces and semantic maps. A contiguous representation of a single marker across several functions indicates a continuous development from one function to the next. Consequently, it is impossible that markers jump from one function to another, leaving out the intermediate uses (cf. Haspelmath 2003: 233). William Croft sees the potential of semantic maps in giving information about the different (possible) stages of semantic change in the grammaticalisation of particular grammatical markers (cf. Croft 2001: 101/102) – the connecting lines between the functions on a conceptual space can easily be recognised as developmental paths. Thus, semantic maps may illuminate grammaticalisation processes (compare chapter 6.2). If one restricts the relevance of grammaticalisation to the development from lexical items to more grammatical ones, semantic maps are generally assumed to be of little help to represent the grammaticalisation paths from lexical to function word. This restricted approach, however, is questioned by Narrog and van der Auwera (forthcoming) and most importantly by van der Auwera and Plungian who integrate the lexical origins of various modal expressions in their semantic map on modality in a convincing way (cf. van der Auwera/Plungian 1998). As a semantic map is supposed to represent the different semantic aspects of a grammatical category, why not integrate the lexical sources of grammatical markers, even if they need to be located somewhat outside the central map but can still be connected to it? However, grammaticalisation with a more restricted notion can also denote the process of already grammaticalised items changing into other grammatical functions. According to this view, semantic maps come to represent language- specific diachronic information (cf. Narrog/van der Auwera (forthcoming): 5). Furthermore, the underlying conceptual space may indicate universal diachronic meaning relations and allow the deduction of universally valid paths of meaning and function development. The representation of grammaticalisation processes influences the geometry of conceptual spaces and the respective language-specific semantic maps. The connecting lines be17

tween the single functions of a grammatical category change into arrows and come to visualise a dynamic process (cf. Narrog/van der Auwera (forthcoming): 5). Thus, semantic maps indicate the direction of language-specific diachronic processes, i.e. the direction of semantic changes. Eventually, they show how individual markers developed to cover a particular semantic area. Necessarily, this analysis bases in the assumption that the underlying conceptual space shows how a grammatical category came to cover the entire underlying semantic area (cf. van der Auwera/Temürcü 2006: 133). Grammaticalisation processes are generally claimed to be exclusively unidirectional. This claim is fundamental for a diachronic application of semantic maps as only unidirectionality allows assumptions about the semantic development of grammatical markers and the developmental paths. Generally, semantic change always proceeds from the concrete to the (more) abstract (compare chapter 6), which gives an important hint as to how languagespecific semantic maps (and also universal conceptual spaces, for that matter) can be useful for diachronic analyses. However, also the typological conditions for constructing the underlying conceptual space give important clues concerning diachronic information. As Haspelmath notes, the choice and the arrangement of the functions for the conceptual space ultimately derive from cross-linguistic data and analyses of different languages. Additionally, the arrangement of the functions on the space must allow a contiguous mapping of languagespecific grammatical markers. Thus, the conceptual space implicitly depicts the development of meaning of one grammatical category and its respective grammatical expressions both universally and language-specifically (cf. Haspelmath 2003: 233ff., see also van der Auwera/Temürcü 2006:133). Croft specifies this when he explains that semantic maps illustrate how the functions of grammatical markers and constructions extend along the paths of the interrelations between functions on the conceptual space (cf. Croft 2001: 101/102, see also Croft/Cruse 2004: 321). The network of functions which constitute the conceptual space represent the entire semantic area of a particular grammatical category and thus, all the different (possible) stages of its semantic development. The language-specific semantic maps visualise the diachronic semantic development of one or several particular grammatical markers in the different languages. These markers do not necessarily cover the entire semantic area of the conceptual space (indeed, this is rather uncommon) and still offer valuable insights into the different steps of semantic change in various languages. Additionally,

18

they allow deductions and predictions about universal and possible future paths of grammaticalisation. Similar to their synchronic applications, semantic maps are restricted to depicting only the semantic aspects of grammaticalisation which usually encompass processes such as widening, generalisation and semantic bleaching. Other grammaticalisation processes on the level of phonology or morphosyntax remain unconsidered. Thus, semantic maps and conceptual spaces eventually function as visualisation tool and foundation for further research in historical linguistics as well. For diachronic purposes, they of course need to be supported and validated not only by cross-linguistic data but most importantly by historical data. However, if the historical data contradict the assumptions based on the map or its geography, it proves to be invalid and in need of revision (cf. Narrog/van der Auwera (forthcoming): 9). This points to a general weakness in the semantic map approach. The language- specific semantic maps and the typological, psychological, and historical findings based on them are valid only as long as the underlying conceptual space and its arrangement of the functions are valid. One language may render the space, its maps and the resulting insights worthless.

2.5 Limitations of the semantic map approach Although semantic maps are assumed to be a useful tool for a variety of scientific fields which are by no means restricted to the branch of linguistics, they show a number of limitations. Linguistic studies in general and claims based on semantic maps and conceptual spaces in specific may be affected by over- and undergeneralisation by the researcher. Croft (1998) and Sandra (1998) call these dangerous phenomena the generality fallacy and the polysemy fallacy. Both fallacies lead to erroneous conclusions in which linguists assume their own inferences to be generally and universally valid. In the case of overgeneralisation, researchers subsume different meanings of a category under just one wider meaning – a conclusion which they believe to be general, i.e. made by all speakers of one or even all languages of the sample. Thus, they assume monosemy in cases which may be polysemous. In the case of undergeneralisation, researchers differentiate between multiple meanings or functions and assume this distinction to be made by all speakers. Consequently, they may assume more different submeanings or -functions than are actually existent (cf. Haspelmath 2003: 239). Both over- and undergeneralisation in the semantic map approach lead to wrong 19

claims and analyses concerning for example the number, diachronic development or interconnectedness of mental representations in human beings. Eventually, this poses the question of how we can ever arrive at a universally valid depiction of the mental representations/conceptual reality of human beings. The semantic map approach tries to alleviate its own limitations in this respect, however. Usually, conceptual spaces and semantic maps are empirically rooted. The danger of overgeneralisation and consequently of mistakes in choice and arrangement of the functions of a grammatical category is high only when one single language or a very small number of languages is taken into account. However, when the conceptual space roots in a systematic typological comparison of a great variety of languages, overgeneralisation can be forestalled and the construction process should lead to a generally valid, universal conceptual space. This becomes obvious in the fact that functions are chosen to be represented on the conceptual space when as few as two languages differentiate between them. Thus, comparison across languages reveals the full number of functions of one grammatical phenomenon since different markings may indicate different functions. Still, if languages never show different markings for seemingly distinct functions, these functions may indeed be just one function (cf. Haspelmath 2003: 239). This basically typological approach to semantic maps both in their development and their application can help to avoid erroneous conclusions concerning the choice, distinction and arrangement of the functions of one grammatical category. A careful development of the conceptual space increases the potential for the language- specific maps to bring forth valid individual and cross- linguistic information, implicational hierarchies, psychological findings concerning the mental reality of human beings and insights into diachronic grammatical processes. In other words, it brings forth a stable semantic map that may function as a stable starting point for numerous analyses in different scientific areas. Besides the weaknesses depicted above, the crucial limitation of the semantic map approach is the relative instability of both conceptual spaces and semantic maps. The problem here arises from the fact that the validity of a conceptual space, that is the choice and arrangement of the functions, and consequently the liability of the language-specific semantic maps are quite vulnerable. The linguistic data of only a single language can contradict and thus devalidate the arrangement of functions established. This is particularly troublesome if the established conceptual space has already been used to illustrate language-specific distributions of grammatical markers or has led to further analyses of various kinds. In his article 20

on semantic maps, Zwarts describes possible problems concerning the stability of semantic maps and conceptual spaces. He sees danger already in the process of developing the conceptual network when an insufficient amount of linguistic data is considered. The map developed is not supported by enough empirical data; language material which may be applied to the map later contradicts the arrangement of functions and the principle of contiguity cannot be maintained (Zwarts 2010: 383). Thus, only one language will not do in order to arrive at a stable and universally valid conceptual space; only with a larger sample, linguists will discover universal conceptual structures (cf. Zwarts 2010: 383). Otherwise, counterexamples of even just one language will invalidate the conceptual space and its design, request a new arrangement and overthrow all claims made on the basis of this map (cf. Narrog/van der Auwera, forthcoming: 2). Another potential problem according to Haspelmath is the phenomenon of so-called “vacuous maps” (Haspelmath 2003: 218). Here, linguistic data again leads to changes in the prior spatial arrangement of grammatical functions. As a result, all functions of a grammatical category are attested in the sample languages and they are all connected to each other.

Direction

purpose

recipient

Fig. 4 A vacuous semantic map of the Dative (Haspelmath 2003: 218)

The resulting map does not give insights into typological differences or language-specific distributions and it does not allow cross-linguistic comparison (cf. ibid.). These two problematic aspects illustrate how vulnerable the semantic map approach is to contradicting linguistic data. The reliability of both the maps and claims made on their ground are comparatively easy to challenge and prone to changes. It needs carefully selected and comprehensive data to forestall instability and vulnerability already in the development of conceptual spaces. Only with a sufficiently large language sample one may discern the relevant functions for a valid map and arrange them in a stable network. As has been remarked above, this needs the analysis of 12 languages at least according to Haspelmath (cf. Haspelmath 2003: 217), according to de Haan and other linguistics, it needs rather more.

21

In his paper on two approaches to semantic maps, Zwarts tries to give solutions to the limitations of the semantic map approach. Like Haspelmath, he points out that the quality and thus the stability and reliability of a semantic map depends on the amount of linguistic data used. Especially with grammatical categories, their meanings and functions are highly abstract and need concrete linguistic contexts to be sufficiently discernable (cf. Zwarts 2010: 383). If maps are established without or with insufficient data, researchers can develop biases and when confronted with actual data, try to bend the input to the already preestablished space (cf. Zwarts 2010: 383). Thus, Zwarts strongly recommends and supports an empirical approach to the development of conceptual spaces and semantic maps. Spaces and maps should be changed and fitted according to the linguistic findings, not the other way around. However, it is important to keep in mind that semantic maps or their respective conceptual space are “just one particular spatialisation of the data” (Wälchi 2007: 47). Eventually, this reveals the basic limitation and at the same time the fundamental nature of semantic maps and conceptual spaces: they are ‘only’ a means of visualisation of a specific semantic space, they alone do not represent any implications as to the nature of the relations between the functions and meanings, the exact cognitive mechanisms of language change or of psychological mechanisms of accessing and processing linguistic and semantic information. Even the question if there is just the one semantic space of a grammatical (or any other) category is, according to Wälchli, questionable.

2.6 Statistical Maps – the solution? Although the semantic map approach is an extremely useful tool for representing grammatical meaning both language-specifically and cross-linguistically, it does have its limitations and problematic aspects, the crucial one being its restriction to a visualisation technique. Semantic maps do not give any information on the frequency of grammatical functions or on their prototypicality, they do not depict a hierarchy between single functions, and they do not (and cannot) represent cognitive or psychological explanations concerning the interconnections of functions (cf. de Haan (manuscript): 3, Zwarts 2010: 383). These restrictions were sometimes considered a flaw in the semantic map approach so that linguists tried to find complementing visualisation methods that would compensate the apparent lack of information. Especially the negligence of frequency in the representation of semantic 22

maps was seen as a shortcoming of the whole approach and led to the development of statistics-based maps. These maps indicate the relatedness of functions of a grammatical category by spatial adjacency and relative distance alone. As a basis for these maps, large amounts of cross-linguistic data can be examined as they are statistically analysed by computer programmes. A semantic analysis of the data is not necessary, the resulting map is rather a statistical distribution of graphs. This distribution is achieved with multi-dimensional scaling programmes, short MDS (cf. Narrog/van der Auwera (forthcoming): 2). Most importantly, statistical maps take frequency into account by marking frequent functions with capital letters and less frequent ones with small letters (cf. Narrog/van der Auwera (forthcoming): 4). The construction of statistical maps differs considerably from the development of classical maps. A statistical program calculates the spatial arrangement of the data which leads to a particular distributional pattern with functional clusters indicating close similarities of functions (cf. Wälchli 2006/2007: 14). These statistical maps may indicate both frequency and relatedness, and they can cover extremely large data set – an advantage over classical semantic maps. However, statisticsbased maps do not represent cognitive or psychological information concerning the relatedness of functions, either. From them, implicational universals cannot be deduced and grammaticalisation is hardly analysable. Thus, their diachronic application is questionable. Additionally, the statistical computer programs only form the distributional clusters but do not give any parameters according to which the functions are grouped together; the researchers themselves have to interpret the clusters and find metrics which they may apply to them (cf. Narrog/van der Auwera, forthcoming: 4f.). Thus, statistical maps can function as valuable complements to classical maps, but they do not overcome all limitations of the semantic map approach or replace additional analyses.

PART II 3

A semantic map approach to concession According to Ekkehard König, concession can be expressed by various structures such as

prepositions, conjunctions, conjunctional adverbs, affixes or asyndetic linking (cf. König 1988: 146). As this paper specifically deals with the semantic map approach to adverbial clauses, only subordinating, adverbial concessive conjunctions are taken into account when 23

constructing the underlying conceptual space and the language-specific semantic maps. However, if one considers all other expressions and connectors of concession, one may be led to a more elaborate functional distinction and thus, to a more elaborate, complex semantic map. The semantic map of the adverbial relation of concession is necessarily a part of this larger network. For the scope of the analysis in this paper, the number of concessive connectors considered has to be limited. I partly follow the restrictions and parameters of Bernd Kortmann in his book on adverbial subordination. He defines adverbial subordinators as “[...] free forms or bound adverbial morphemes which specify some semantic interclausal (or: circumstantial, adverbial) relation between the subordinate clause over which they operate and the modified matrix clause (Kortmann 1991: 4).” In this paper, I will restrict the analysis to free forms as four sample languages English, German, French and Latin introduce finite adverbial clauses with the help of adverbial conjunctions. The definition of Matthew Dryer gives a more general picture of adverbial subordinators: “By an adverbial subordinator, I mean a word that marks an adverbial subordinate clause for its semantic relation the main clause (Dryer 1992: 53).” Indeed, a definition of adverbial subordination is extremely difficult as subordination itself is already hard to define. According to Kortmann, there is no stable list or set of characteristic criteria which marks all subordinate clauses in general. Rather, subordination and superordination often form a continuum. Thus, it is necessary to assume a prototypical adverbial subordinator in order to obtain any forms which can be analysed. Accordingly, Kortmann attributes a set of characteristics to this prototypical adverbial subordinator. Prototypical subordinators are uninflected, operate over a finite subordinate clause and do not have a core syntactic function, i.e. they are not a grammatical relation. They do not have a flexible position in the sentence but usually occur at the beginning or the end of a sentence. They are of everyday use and unmarked. If the adverbial subordinator is complex, the parts must show signs of fusion, i.e. they must have lost properties of the original phrase. In other words, the meaning of the adverbial connective must be grammaticalised to a point at which its meaning is no longer fully predictable from the meaning of the parts. Thus, the prototypical adverbial subordinator is noncompositional. As a consequence, Kortmann excludes coordinators, conjunctional adverbs, pronominal adverbs, archaic forms and adverbial subordinators modified by a focus particle from the group of adverbial subordinators (cf. Kortmann 1997: 64 ff.). Although this set of criteria is very convincing, bearing in mind that 24

78,2% of Kortmann’s data falls into his category of an ideal adverbial subordinator, it is problematic to exclude connectors modified by a focus particle from the set of adverbial subordinators. In the field of hypothetical concession, subordinators modified by a focus particle are of central importance in all four sample languages. In order to analyse semantics and functions of these types of concessive clauses, it is essential and inevitable to take these connectives into consideration. That hypothetical concession is an important and central area of concessivity makes it only more necessary to include adverbial subordinators with focus particles into the group of subordinators analysed. Additionally, these connectives are frequently named as concessive conjunctions in standard grammars of all four sample languages. Thus and most importantly, I was guided in my choice of concessive subordinators by the primary sources for my analyses: descriptive grammars. The standard grammars of the four sample languages name the most frequent and prototypical subordinators and connectors for concessive adverbial clauses. Thus, they function as guideline and orientation in the choice of the structures and connectors analysed. In most cases, the list of connectors given by the grammars is supported by linguistic essays and analyses on concessive clauses. These sources motivated my choice of subordinators which are semantically as well as formally described and illustrated in the semantic maps on the conceptual space of concession. The restriction to four sample languages arises from the limits of the paper due to time and room restrictions. The four languages English, German, Latin and French stem from two different language families. Although they are all European languages, Latin and French are from the Romance family and English and German have a Germanic origin. I expect that the Romance languages reveal possible etymological developments and relations in their concessive markers. English was heavily influenced by French and Latin. Due to this influence, the concessive structures may show interesting and explainable correlations. However, due to the different genetic origins, there are hopefully also fruitful differences in the marking and use of concessive structures. English and German may probably show many similarities due to their close genetic relationship. Still, German was not as heavily influenced by Latin and French and belongs to a different branch of the Germanic languages, which may lead to interesting differences.

25

3.1 What is concession? Semantically, concession is an extremely complex and difficult relation. Psychologically, it requires both speaker and hearer to relate background assumptions to foreground structures and to deal with contrastive propositions and violated expectations. Thus, the concessive relation is hard to define, which results in a number of many different definitions and explanations in grammars. This is not only the case between grammars of different languages but also within a single one. Consequently, I will have a closer look at the definitions of concession in the major grammars of the language sample before I will present a working definition for this paper. The grammar books are not only the basis for a working definition of concession, they also function as major source of the functional analysis of concessive clauses and thus, of the semantic maps of concession.

3.1.1

Grammar definitions of concession

In A comprehensive grammar of English by Quirk et al., concessive clauses of English are characterised according to the “unexpected, surprising nature of what is being said in view of what was said before that (Quirk et al. 2008: 1098).” Thus, in adverbial relations “[c]oncessive clauses indicate that the situation in the matrix clause is contrary to expectation in light of what is said in the concessive clause (ibid.).” Additionally, Quirk et al. point out that concessive clauses often also “imply contrast between the situations described by the two clauses (ibid.).” In another grammar of the English language, Huddleston and Pullum define the meaning of a concessive adverbial clause (consisting of a subordinate and a superordinate clause) as follows: (i) (ii) (iii)

The subordinate clause is entailed.3 The truth of the subordinate clause might lead one to expect that the superordinate clause would be false. In fact, the truth of the subordinate clause does not detract from the truth of the superordinate clause (Huddleston and Pullum 2002: 734).

The definition of concession in the German standard grammar Duden differs considerably among the various editions of the grammar. In the 1986 edition, the most basic notion of 3

The notion of entailment describes the relationship between the main and subordinate clause. In this relationship, the main clause content refers to the factual truth and thus, requires the subordinate clause content to be factual and trueas well (cf. König/Eisenberg 1984: 317). Consequently, the subordinate proposition is necessarily true in concession, although it is incompatible with the main clause.

26

concession is defined as “ein Verhältnis des unzureichenden Gegengrundes4 (Duden 1986: 794).“ More detailed, in the concessive adverbial clause „wird ein Sachverhalt formuliert, der zwar im Gegensatz zu dem im anderen Teilsatz formulierten Sachverhalt steht, aber nicht ausreicht, um dessen Geltung außer Kraft zu setzen (Duden 1986: 794).“ In a later edition from 2005, concession is treated as a representative of the causal relation in the broadest sense and as semantically related to both the adversative and conditional relation (cf. Duden 2005: 1106); thus, concession is presented as multiply interwoven with other adverbial circumstances. This becomes obvious in the definition of concessive meaning according to the 2005 edition of the Duden Grammatik: Auch die konzessive (einräumende) Verknüpfung korrigiert eine Erwartungshaltung, die durch ein konditionales Verhältnis vorgebeben ist. Diese Korrektur setzt jedoch am Bedingungssatz an und wird als „Verhältnis des unwirksamen Gegengrundes“ bezeichnet (Duden 2005: 1106). In this definition, concession is presented as closely tied to cause and to condition and contrast, i.e. the adversative relation. The connection to cause is established by analysing concession as a “Verhältnis des unwirksamen Gegengrundes”. The close link to contrast and condition is made by a reference to the characteristics of adversative clauses, whose content is contrary to expectation and consequently contrary to an assumed logical chain. Concessive and conditional clauses are semantically linked as conditionals usually establish the logical chain which is contradicted by adversation. Similarly to Germanic languages like English and German, Romance languages like Latin and French employ concessive adverbial clauses as well. The definitions provided by grammars of Latin and French resemble those proposed by English and German grammars. The Latin grammar of Rubenbauer, Hofmann und Heine states that “Konzessivsätze enthalten eine Annahme in Form einer Einräumung und stehen entsprechend der Herkunft der einleitenden Konjunktionen teils im Indikativ, teils im Konjunktiv (Rubenbauer, Hofman, Heine 1989: 317).“ This already indicates one of the major differences in the formal encoding of concessive relations between the four sample languages – concessive conjunctions in Latin and French may obligatorily require the subjunctive or indicative. In German, this is rather a matter of style or speaker intention, while English only rarely distinguishes verbal mood in

4

In own translation: ineffective anticause. Elisabeth Rudolph offers the functional description of an “insufficient objection (Rudolph 1996: 121).”

27

concessive adverbial clauses. Another standard Latin grammar by Hermann Menge adds the semantics of contrast to the notion of conceding: “Konzessivsätze bezeichnen eine Einräumung, ein Zugeständnis, einen starken, meist unauflösbaren Gegensatz zum übergeordneten Hauptsatz (Menge 2000: 859).“ The standard grammar of French, the Grévisse, brings all the different aspects of meaning of the concessive relation together when it states that [l]a proposition de concession indique qu’il n’y a pas eu la relation logique attendue entre le fait qu’elle exprime et celui qu’exprime le verbe principal. Elle énonce notamment une cause non efficace, contrariée, qui n’a pas eu l’effet que l’on pouvait prévoir (Grévisse 2008: 1499). Concerning the definition and analysis of the concessive adverbial relation, all grammars consulted state similar things. At its very basis, concession grounds in the notion of contrast between the two propositions of main clause and subordinate clause – an evaluation which can be found in most of the grammars of English, German, French and Latin. Thus, most grammars see concession as semantically related to adversative relations. The definitions of concessive clauses also note the insufficiency of the proposition in the subordinate clause to invalidate the proposition of the main clause. Additionally, grammars frequently refer to the fact that the proposition in the main clause works against expectations triggered by the subordinate proposition.

3.2 Real concession For the scope of the paper, it is necessary to define concession in more detail; thus, it is necessary to look at concession from a broader perspective in order to be able to find out about various functions of the concessive relation. In her 1996 book Contrast. Adversative and Concessive Expressions on Sentence and Text Level, Elisabeth Rudolph approaches concession as a representative of the relationship of contrast, locating it in a rather broad semantic range. The definitions of the various grammars (cf. Quirk 2008: 1098 or Duden 1986: 794) indicate this connectionas well, but Rudolph elaborates it much further. So what role does contrast play in the semantics of concession? According to Rudolph, the relation of contrast is one of four basic types of clause combinations which include the phenomenon of adverbial subordination. These four combinations are the connection of addition, the connection of contrast, the connection of time and the connection of causality (cf. Rudolph

28

1996: 20).5 In clause combinations that express contrast, “two propositions A and B are conjoined and in contrast to each other [...]. They are simultaneously valid [...] so that they involve a contrast between A and B [...] (Rudolph 1996: 22).” However, the semantics of concession do not rely only on the contrastive character of two propositions, they also ground in a “causal constant (Rudolph 1996: 27)”, that is in causality. The semantic definitions from the various grammars referred to a causal notion in concession, usually by combining it with contrast and naming it “anticause”, “incausal” or “inoperant cause” (cf. König/Siemund 2000: 341). So what is the connection to causality in concession? In perception and reasoning, human beings try to establish a causal chain, a link between two independent events, facts or utterances. This chain does not have to be overtly expressed, “it exists independently from the verbal manifestations (Rudolph 1996: 27).” Eventually, this causal chain is rather a mental process, “a form of background causality which does not have to be mentioned [...] as it will be easily inferred on the basis of the common world knowledge (ibid.).” Thus, the causal chain underlying any expression of concession can be paraphrased as tendencies, regularities and assumptions about the interconnection of two situations based in logic, cause-effect and reason-consequence relationships (cf. König/Siemund 2000: 353, Rudolph 1996: 27). So in a concessive sentence like (1) Although she went to university, she did not find a job.

causality is working in the background. Usually, the expectation is that going to university increases the likelihood to find a job. This expectation grounds in the common world knowledge of a university diploma being the highest regular degree and in the logical chain that the highest educational degree should lead easily to finding a job. This expectation, however, is not fulfilled as both propositions stand in contrast to each other; the non-fulfillment of the expectation is indicated by the concessive conjunction although. The concessive conjunction is the trigger which conveys that the situation expressed in the sentence above does not go hand in hand with the causal constant based in world knowledge, but rather conveys a contrast between the two clauses (and their content).

5

The connection of addition is prototypically expressed by „and“. Both propositions of the combined clauses are simultaneously valid. In clause combinations expressing the combination of time, the two propositions expressed are not valid simultaneously and express a sequence in time (this, for some reason, excludes simultaneity. Causality expresses two propositions linked by a cause, a logical or an empirical necessity; both propositions are valid simultaneously or unsimultaneously (cf. Rudolph 1996: 20).

29

Bernd Kortmann sees concession similarly connected to cause and establishes a further typical connection between concession and another adverbial relation: condition. Cause and condition influence the meaning of concession. Still, all three are clearly separate relations which only partially influence each other (compare chapter 3.3). Kortmann comes up with a more schematic definition of concessive meaning when he describes the underlying structure as “although p, q (Kortmann 1997: 86)”, in which “p and q are asserted6 against the background of an assumption concerning the general incompatibility of such situations, i.e. the standard implication is: ‘normally (if p, then not-q)’ (ibid.).” The “background of an assumption” here is just another wording for a causal chain based in common world knowledge or cause-effect relationships which organise the way human beings process information and establish interconnections between them. Only against this background of an assumption, expectations concerning the informational content of the concessive sentence can be disappointed when the underlying causal chain or assumption about the incompatibility of the situations is violated. However, in contrast to Rudolph, Kortmann sees a clear difference in meaning between concession and contrast: in contrast, there is no standard assumption that p and q are generally incompatible; for him, they rather function as arguments for different conclusions (cf. Kortmann 1997: 86). So from a broader perspective, concession can be defined as follows: “[A] concessive connective implies that the antecedent [i.e. p] and the consequent [i.e. q] are by normal standards incompatible (cf. Harris 1998: 72).” This definition does not totally exclude a notion of contrast, indeed contrast can be anticipated from the incompatibility of the two propositions (although not in Kortmann’s definition of contrast). Clauses in a real concessive relationship have a factual character, that is, they “entail both of their component clauses (König 1988: 146).” Thus, both component clauses refer to the truth and are connected by the presupposition that the content of the component clauses is usually incompatible. This incompatibility is closely connected to Rudolph’s causal chain: against the assumption of logical relationships and common world knowledge, the two incompatible propositions are conjoined (cf. König 1988: 147, cf. Rudolph 1996: 27ff).

6

In linguistics, a proposition is asserted when it is presented as a factual, objective truth. An assertion refers to the meaning of a sentence, i.e. its surface content.

30

A somewhat different but rather insightful approach to the semantics of concession is given by Pötters when he describes concession as a “negated implication (cf. Rudolph 1996: 209)”. The implication refers to a causal relation that “according to our expectation, it is the case that p leads to q (ibid.).” In the case of concession, i.e. in the case of a negated implication, “contrary to the expectation, it is not the case that p implies q, but NEG-q appears exceptionally despite p (Rudolph 1996: 209).” However, concessive relations and concessive clauses do not only include so-called realconcession. Concessive relations also comprise so-called conditional-concessive or hypothetical concessive clauses. Their semantics as well as their formal and functional characteristics are different from real concession; some linguists even argue that hypothetical concessives do not truly belong to the field of concession but rather form a particular sub-type of conditional relations (cf. Duden 1986, König 1988). For this paper, however, they are considered an intermediate type with characteristics of both condition and concession – thus, they may even provide important insights into diachronic processes of grammaticalisation in the field of concession (cf. Harris 1988, Rudolph 1996, König 1985).

3.3 Hypothetical concession Hypothetical concessive clauses, also frequently called concessive conditionals, share features of both conditional and concessive relations. This becomes obvious in the semantics and morphosyntactics of their connectives and clause structure (cf. Harris 1988: 72). Semantically, “a concessive conditional sets up a hypothesis, the realization of which, if it occurs, would bring into being a situation in which a concessive interpretation [...] then comes into play (Harris 1988: 72).” (2) Even if it rains, the Open Air concert will take place.

This sentence exemplifies the hypothetical nature of the adverbial clause; it might rain or it might not rain, but if it does (that is, if the hypothetical situation takes place), the concert will nevertheless take place, in spite of the rain. The realisation of the (hypothetical) rainy weather brings about a concessive situation because usually, one would expect rain to lead to an Open Air concert being concelled. According to this sentence, however, rain will not have this effect. 31

Example (2) illustrates how hypothetical concessive clauses share the notion of probability with conditional clauses and the notion of incompatibility with concessive clauses as one circumstance is presented as being incompatible with the proposition of the main clause (cf. Harris 1988: 73). Probability is expressed by the assumed question if it will rain or not, or in other words, by the hypothetical nature of the content in the subordinate clause. The notion of incompatibility becomes obvious only if one of the possible situations occurs and contradicts common world knowledge or logical relationships, thus violating expectations based on the causal constant (cf. Rudolph 1996: 27f).

3.3.1

Subtypes of hypothetical concession

Hypothetical concessive clauses can be subdivided into three semantic functions of hypothetical concession: scalar concessive conditionals (SCC), alternative concessive conditionals (ACC) and universal concessive conditionals (UCC) (cf. König/Haspelmath 1998: 563). (3) Even if it rains, the Open Air concert will take place (SCC). (4) Du wirst dich der Öffentlichkeit stellen müssen, ob du willst oder nicht (ACC). (5) Whatever he does, everybody loves him (UCC).

These different functions and their respective coding mechanisms can be found in many grammars which refer to similar subtypes of concession as for example Quirk et al. in their grammar of English (cf. Quirk et al. 2008: 1089ff.). However, grammars usually give various sources and origins for the three subtypes of concession; often, they are treated as special instances of alleged source constructions such as particular types of condition for scalar concessive conditionals, embedded interrogatives for alternative concessive conditionals and free,

generalising

relative

clauses

for

universal

concessive

conditionals

(cf.

König/Haspelmath 1998: 563). Although this indicates a variety of semantic aspects, sources and formal encoding mechanisms, all three types of hypothetical concession are unquestionably bound to a basic concessive semantics as example (2) illustrates above. Although many grammars and linguists identify various sources for the three types of hypothetical concession, all three base in conditional semantics. Thus, they show the same tense and mood combinations as conditionals. Alternative concessive conditionals exemplify the conjunction of two conditional subordinate clauses with one being the negation of the 32

other. Universal concessive conditionals in turn may be transformed to alternative ones, thus increasing the similarity to conditionals (cf. König/Haspelmath 1998: 565). However, most importantly, all three types consist of an apodosis (the main clause) and a protasis (the subordinate clause), the typical parts of conditional relations. Still, this is exactly where the major difference between pure conditionals and concessive conditionals can be found: in hypothetical concessive clauses, the protasis refers to a whole set of circumstances. This set is expressed either by a quantification marker such as in UCCs (wh- ever, w- auch immer, cumque), by the disjunction of protasis and its negative part as in ACCs (ob… oder, whether… or, sive… sive) or by a scale expressing an extreme value (even, selbst, auch, même, tout)(cf. König/Haspelmath 1998: 565). Besides the obvious relatedness to condition, hypothetical concessives are also invariably related to concession, i.e. to “[the assertion of] two propositions against the background assumption that the relevant situations do not normally go together [...] (König/Haspelmath 1998: 566).” Concessive conditionals include circumstances in their set of protases which would usually not allow for the apodosis to hold true and which do not usually go together with the apodosis. As an additional feature of real concession, the main clause of hypothetical concessive clauses is factual. In instances of real concession, both clauses are asserted, i.e. express factual information, and both clauses are entailed, i.e. committed to the truth of their content. Hypothetical concessives are semifactual, i.e. their main clause is factual and committed to the truth, the subordinate clause however is usually hypothetical (cf. König/Haspelmath 1998: 567). Additional proof for concessive conditionals being related to real concession is the fact that “concessive conditionals tend to develop into genuine concessives (König/Haspelmath 1998: 568).” Hypothetical concessive markers are a frequent source in the diachronic development of real concessive markers. Many languages even allow the synchronic use of concessive conditional markers in proper concessive contexts; thus, even if and even though may be used interchangeably in English (cf. ibid).

3.3.2

Source constructions of hypothetical concessive clauses

According to Harris, hypothetical concessives are closely related to questions – semantically, the subordinate clause implies a question; if we look again at the rain-concert example, (6) Even if it rains, the Open Air concert will take place. 33

it is unclear if it will rain or if it will not rain. However, the answer to the question is irrelevant as it does not affect the truth of the main clause (cf. Harris 1988: 74). Thus, the sentence has a concessive meaning. König and Haspelmath see the relation between concessive conditionals and embedded interrogatives as well. They differentiate various source constructions when they explain that alternative concessive conditionals are very closely related to polar questions while universal concessive conditionals are rather related to whquestions. Scalar concessive conditionals can also be related to polar questions (cf. König/Haspelmath 1998: 577f.). However, hypothetical concessive clauses exhibit connections to interrogatives not only semantically. Hypothetical concessive clauses, more specifically universal concessive conditionals and alternative concessive conditionals, may also be introduced by interrogative items such as whether in English or wh- items in English and German, thus also formally exhibiting the relationship to embedded interrogatives as source constructions (compare appendix). Besides the semantic and formal links to embedded interrogatives, hypothetical concessive clauses in general and scalar concessive conditionals in particular express a close connection to the conditional relation both semantically and morphosyntactically. The morphology of the subordinators of scalar concessive conditionals shows features of both condition and concession in at least three languages of the language sample: German wenn resurfaces in the concessive conditional auch wenn; Latin si is part of the concessive subordinator etiamsi and English if functions as part of the hypothetical concessive connective in even if (cf. Rudolph 1996: 213). The concessive notion of these subordinators is triggered only by the scalar particles auch, etiam and even. Scalar concessive conditionals may even be formally identical to usual conditional clauses when they employ markers without scalar particles. In these cases, concession derives solely from pragmatic clues such as intonational patterns and the context (cf. König/Haspelmath 1998: 575). Equally, universal concessive conditionals can be linked not only to source constructions such as condition or embedded interrogation. Most frequently, UCCs are related to headless relative clauses; indeed they are “difficult to keep apart from headless (‘free’) relative clauses where the relativized constituent has a nonspecific meaning [...] (König/Haspelmath 1998: 577).” There is, however, a clear indication as to whether a clause is a headless relative construction or a concessive adverbial clause: while adverbial clauses are an additional piece of information, an adjunct in the complex sentence, relative clauses – including headless ones – 34

are a constituent of the main clause. In German, this leads to two different word orders in the main clause; the difference between the two constructions is overtly marked in the different syntactic structures (cf. König/Haspelmath 1998: 577f.). (7) a. Wo auch immer du hingehst, bist du steuerpflichtig. b. Wo auch immer du hingehst, du bist (überall) steuerpflichtig. (König/Haspelmath 577/578)

This example illustrates that in the concessive conditional structure, the subordinate clause does not occur in the immediate Vorfeld but as an adjunct.

3.4 Concession and cause As has been repeatedly pointed out in the various definitions of real concession and hypothetical concession, the concessive relation is inextricably bound not only to condition but also to causality. Condition, cause, concessive conditionals and concession seem to be a continuum of interrelated functions. What exactly is the nature of these interrelations? How is concession tied to causal semantics? Although some of the relations between causality and concession have already been explored in the definition of real concessive semantics (compare chapter 3.2), their interrelations will be examined again and more closely in this paragraph as the relation between cause and concession has led to many different approaches and explanations among linguists. Often, just as definitions from grammars with labels such as ‘anticause’, ‘Gegengrund’, ‘incausal’ or ‘inoperant cause’ suggest, concession is considered the “negative or contradictory counterpart of causal constructions (König/Siemund 2000: 341; compare also Verhagen 2000: 361).” What both causal and concessive clauses have in common is their factual nature, i.e. both clause types “assert the truth of the two clauses [...] combined (König/Siemund 2000: 342).” However, both clause types differ exceedingly when it comes to the nature of the relationship between the two clauses. While causal clauses express a situation, chain of events or two propositions that go along established paths of logic, causeeffect relationships, common world knowledge, general tendencies and sequences of events in the world, concessive clauses do not express anything like that: on the contrary, in concessive clauses, this chain between the propositions is deliberately broken (cf. 35

König/Siemund 2000: 342; Rudolph 1996: 39f.). But does this contrast suffice to describe the interconnections between cause and concession? In their essay on causal and concessive clauses, König and Siemund directly address this question and come up with three arguments which support the thesis that concession and cause are counterparts and opposite relations. One of the fundamental views on the relationship between concession and cause is grounded in “the different interaction between negation and the adverbial adjunct [, i.e.] a difference in the scope of negation (König/Siemund 2000: 344).” One frequently cited example illustrates the difference: (8) This house is no less comfortable because it dispenses with AIR- conditioning. (9) The house is no LESS comfortable, although it dispenses with AIR- conditioning. (König 2000: 343)

Sentence (8) shows how the negator has wide scope over both clauses in the complex sentence, sentence (9) exemplifies narrow scope of the negation only over the main clause. König and Siemund thus show how a causal construction can be paraphrased by a concessive one: when the whole causal construction is negated, it may semantically correspond to a concessive construction in which the main clause is negated. The rejection of the causal connection between the two clauses by the wide scope negator eventually leads to a semantic similarity between negated cause and concession: in both instances, the underlying causal chain is violated. According to this example, both relations can be seen as counterparts and concession is justifiably labelled ‘anticause’. A second convincing argument for a close connection between causality and concession is the diachronic relationship between their markers: “causal connectives may develop into concessive ones, even though the reverse direction has not been attested (König/Siemund 2000: 346).” An example of the English language are complex clause structures introduced by for all which lead to a concessive reading of the sentence and which derive from the causal linker for. Unfortunately, König and Siemund only present causality-based markers for non-adverbial clauses; nevertheless they illustrate a close connection between cause and concession (cf. König/Siemund 2000: 346). The most convincing proof for an opposite relationship between cause and concession are languages which form concessive expressions by negating causal ones, i.e. by “combining 36

a causal connective with a negator (König/Siemund 2000: 347).” Although no language of the sample chosen for this paper uses predominantly this strategy to express adverbial concession, there are examples of negated concessive markers with fundamentally causal meaning such as ungeachtet in German or regardless and nevertheless in English (cf. ibid). The three arguments by König and Siemund show that cause and concession are closely interrelated by causal opposition. While causality adheres to logical cause-effect relationships and regularities of events in the every-day world, concession basically relies on these causal relationshipsas well, but it deliberately rejects them (cf. König/Siemund 2000: 347). Yet, this indicates that concession depends on causal notions and their negation. Besides the arguments by König and Siemund, other linguists found different approaches concerning the question of the relationship between causality and concession. Di Meola, for example, assumes two underlying causal relations as necessary in order to describe the semantics of concession: one causal relation is hypothetical and refers to the logical consequent of the subordinate proposition, the other is factual and refers to the logical cause for the main proposition. If we look again at the university example, this leads to the following assumptions: (10) Although she went to university, she did not find a job.

The hypothetical cause could lead to a complex structure such as ‘Since she went to university, she should easily find a job.’, thus stating the logical consequent of a high educational degree: an economically secure future. The factual cause, on the other hand, provides the logical cause of the main clause proposition; in this instance, something like ‘Since there is an economic crisis, she does not find a job.’ could be imaginable. Both consequents in the two complex structures, i.e. both main clauses, exclude each other. Of the two causes, one is not coded or overtly expressed, usually the factual cause for the main clause proposition of the concessive structure (cf. König/Siemund 2000: 347f.). This theory points out how closely concessive meaning is tied to cause and causal relations which hold between the two clauses of the complex construction. However, this theory also has its limits according to Di Meola himself. He notes that some concessive clauses do not rely on hidden causality. Furthermore, if there exists something like a hidden cause, it refers to information that never surfaces: consequently, its role for the informational content of the sentence and its conversational intentions must be un37

clear or even irrelevant. Additionally, the existence of a hidden cause would violate Gricean Maxims as speakers using concessive constructions would behave uncooperatively in hiding crucial parts of a message. Di Meola deems it unconvincing that an uncooperative conversation pattern would be as firmly established as concessive constructions are in most languages of the world (cf. König/Siemund 2000: 350). König himself offers another approach to the relationship between cause and concession. He considers it “a specific instance of duality (ibid.)” with duality being a relationship that “holds between two scope-bearing expressions if the sentences containing them can be translated into one another, by means of internal vs. external negation (ibid.).” The example offered in König and Siemund is supposed to illustrate the theory of duality in concession. (11)a. The road is not getting wet although it is raining. b. It is not the case that the road is getting wet because it is raining. (König/Siemund 2000: 350)

Theoretically, both sentence pairs can be described as: (a) = although p, q (b) = ¬ ((because p) ¬q) In both structures, p is ineffective, which should render both structures semantically similar if they are in a duality relationship (cf. König/Siemund 2000: 350ff.). However, this approach reveals deficiencies, too. A duality relationship works only from cause to concession; concessive clauses cannot be reformulated to causal ones with the help of external or internal negation. Furthermore, the instances in which a dual situation holds true are comparatively limited. In fact, a relationship of duality only applies to causal clauses with wide scope negation that correspond to concessive clauses whose main clause expresses the negative counterpart of the main clause in the causal construction (cf. König/Siemund 2000: 351). In sum, the theory of duality is another approach in order to explain the relationship between cause and concession. It assumes a very close connection between cause and concession and even succeeds in showing how some instances of causality can be transformed into concessive relations. Still, duality cannot account for all semantic varieties of concession which nevertheless include an undeniably causal notion.

38

An interesting approach to the relationship between cause and concession is offered by Arie Verhagen. She treats the topic quite differently when she sees all parallels and interconnections between cause and concession primarily based in mental spaces and their constructions. According to her analyses, the mental constructions of these two relations display similarities and differences. In concessive relations, Verhagen, in concordance with other linguist, sees “two propositions asserted against the background of an assumption (Verhagen 2000: 365).” The combination of the two propositions leads to an internal contradiction; however, these contradictory semantics need the background of some causal constant or logical chain, of regularities and tendencies of the world to become fully understandable. Verhagen now argues that the semantics of concession are best described with the help of mental spaces and that cause and concession can be compared on the basis of these mental spaces. According to Verhagen, concession involves two mental spaces, one representing the general ‘topos’, a general rule based on common world knowledge, a causal chain. This mental space can be paraphrased by a structure like ‘p, therefore q’ and is identical to the causal mental space. The second space is very similar to the first one but it is distinct nevertheless: here, the causal chain between p and q, i.e. between the two propositions, does not hold true. With the help of mental spaces, cause and concession can be related. In fact, the mental space of cause is mapped onto another, similar but yet distinct space: the surprising, contrasting nature of concession derives from the fact that the assumed causal chain is violated against all expectations (cf. Verhagen 2000: 366f.). On the basis of her analyses with various aspects of concession and cause, Verhagen eventually formulates the major difference and at the same time point of contiguity of the two relations: causality grounds in only one mental space while concession grounds in two, one of them being the mental space of cause and the other being a very similar, yet invalid causal one (cf. Verhagen 2000: 378). What all the theories presented have in common is that they assume basic similarities between cause and concession. Both relations include a presupposition and in both relations, the propositions involved are asserted. While causal constructions rely on the presupposition of proposition p leading to consequent q, according to general tendencies of logic, common world knowledge and regularities in the world, concessive constructions base on the presupposition of proposition p leading to not-q. Thus, the semantics of concession depend on the assumption of usual causal presuppositions or in other words, the semantics of 39

concession derive from the underlying notion that p would normally lead to the very opposite of q. Consequently, concession works against logic, common world knowledge or regularities in the world while still presupposing them as a general underlying common ground (cf. König/Siemund 2000: 352f.). Although both cause and concession obviously express opposite tendencies when it comes to the presupposition between the propositions involved, exactly these propositions are asserted in causal and concessive constructions: they are factual. This suggests that cause is more closely linked to concession proper, i.e. to concession which does not involve hypotheticality and probability. Although hypothetical concessive clauses are generally assumed to be closely connected to conditional clauses, cause has to play a role in concessive conditionals as well. In order to achieve concessive meaning, all three types of hypothetical concessive clauses at least include possible propositions which violate any logical or general presupposition between the subordinate clause and the main clause content. Concerning this topic, it becomes obvious that both cause and concession rely on what Rudolph calls ‘a causal constant’ or ‘a causal chain’ (Rudolph 1996: 31ff.). However, cause and concession behave differently with regard to their adherence to the causal chain. Causal relations, somewhat naturally, stick to expected tendencies, background assumptions and regularities of the world. Concessive relations do not necessarily refer to the exact opposite of what one would expect according to logic and common world knowledge. However, it at least deems these tendencies and regularities irrelevant by neglecting expected consequences based on the causal chain. Thus, causality as causal chain plays a fundamental role in understanding concessive relations, although it is precisely the rejection of causality which establishes concessive semantics. However, the variety of approaches to the relationship between cause and concession shows that the exact nature of the interconnections between the two relations is as yet not fully analysed. Referring back to the topic of conceptual spaces, it becomes obvious that the space of concession must be part of a larger network of (adverbial) relations. The analyses of cause and concession indicate that the conceptual space of concession and that of cause are probably interrelated as the semantics of the two relations are so closely connected. Similarly, conditional and interrogative as well as relative structures may somehow be connected to the conceptual network of adverbial concession because they function as source constructions for various functional aspects of concession, which will become apparent in a more 40

detailed analysis of the functions (chapter 4 and 5) and etymology (chapter 6) of concessive markers.

4

Functional distinctions of concession Interestingly, the very semantics of the term concession do not refer to the relation be-

tween two clauses but they rather “[...] bezeichn[en] eine mögliche Verwendung solcher Sätze (König/Eisenberg 1984: 313).” Thus, the use and function of adverbial clauses and their connectives is of elementary and essential importance for their existence. Due to these fundamentally functional semantics, concession immediately lends itself for an illustration with the help of a semantic map. Generally, concessive conjunctions are complex both formally and semantically. However, their components are easy to identify so that their recent and their original meaning can be derived without problems. Due to the complexity and the complicated structure of concessive constructions, concessives emerge comparatively late both phylogenetically and ontogenetically. Another important aspect concerning their derived semantics and their late emergence in language history and language acquisition is the close relationship of concession to other semantic domains such as condition and cause. König notes that [...] concessive relations are based on a prior experience of compatibility, correlation, connection or causality between two eventualities, as expressed by other, conditional or causal connectives. The assertion of two facts against the background assumption of their general incompatibility must be based on prior knowledge of compatibilities and connections (König 2000: 157, emphasis added). This consideration does not only explain the late acquisition of concessives in first language acquisition because it requires the understanding and use of other, more basic relations, it also indicates how closely concession is tied to other semantic relations. It necessarily has a connection to condition and cause as it depends on semantic notions of these circumstantial relations in order to fully extend its own meaning and in order to be processable and understandable by users of a language. Generally, concession can be subdivided into two types of concessive adverbial clauses: factual real concession and hypothetical concession. These two types employ different subordinators, which in turn show different patterns of use and function in spoken and written utterances along the lines of these two major groups of concessive relations. Consequently, 41

concession qualifies for a semantic map approach in order to differentiate, define and characterise the different functions of concessivity and to illustrate their interrelations and interconnections, their similarities and differences and their semantic development.

4.1 Real concession Real concession is closely tied to the prototypical semantics of concessivity. Two factual but contrasting propositions are asserted; the realisation of the main clause propositions is surprising as it is normally impossible under the circumstances of the subordinate proposition. Nevertheless and in spite of this contrasting circumstance, it is realised because the opposing subordinate proposition does not suffice to invalidate the main clause content. Thus, real concession is closely tied to notions of contrast and opposition. It is noteworthy that both main and subordinate clause refer to circumstances, events and utterances of reality, i.e. they are factual and asserted as truth. In this paper, four different functions of real concession will established.

4.1.1

Ineffective anticause

The semantics of an ineffective anticause are the prototypical core meaning of concession. When concession is defined, the definitions usually assume the relation of an ineffective anticause or “insufficient objection (Rudolph 96: 121).” The meaning of an insufficient objection becomes most obvious in the definition of concession in the Duden: In einem der beiden Teilsätze wird ein Sachverhalt formuliert, der zwar im Gegensatz zu dem im anderen Teilsatz formulierten Sachverhalt steht, aber nicht ausreicht, um dessen Geltung außer Kraft zu setzen (Duden 1986: 794). Thus, the two propositions of a complex sentence are in contrast but the subordinate proposition does not invalidate the main clause proposition although it would normally lead to an opposite consequence; normally, the subordinate proposition and the main proposition are incompatible. This incompatibility, the opposition between the two propositions and their content, is captured by the labels ‘anticause’ or ‘objection’. Its insufficiency refers to the lack of strength to invalidate the main clause proposition. Since the expression of an insufficient objection is the basic, prototypical function of concessive adverbial clauses, it is present in all sample languages and presumably in all lan42

guages showing concessive relations. In German, ineffective anticause is coded by obwohl, more seldom by obgleich and obschon. In English, the prototypical function of concession is expressed by although and (even) though, while French employs the connectives quoique, bien que, malgré que and seldom encore que. The function of an insufficient objection also includes the notion of surprise and unexpectedness. The surprising notion again refers back to the incompatibility of the two propositions: this incompatibility is obvious to speakers due to common world knowledge, generalities and tendencies of events in the world. Nevertheless, the incompatibility is irrelevant as the validity of the subordinate clause does not suffice to invalidate the main clause, thus leading to a surprising or unexpected outcome of the sentence content (cf. Quirk et al. 2008: 1088). According to Rudolph, it is indeed one of the major uses of real concessive clauses to emphasise “striking events which contradict the normal expectations people conclude from their experiences and general world knowledge (Rudolph 1996: 397).” This unexpectedness in the meaning of insufficient objection, i.e. this crucial part of the concessive meaning, is often added by emphatic forms such as focus particles. Forms such as even though or the rare even when reflect this. According to Rudolph, they are used to emphasise the surprise experienced by the speaker about the validity of the main clause content in spite of the unfavourable circumstances (cf. Rudolph 1996: 400). It is also noteworthy that the incompatibility of the propositions does not necessarily ground in a sharp contrast or an expression of the direct contrary. Rather, it is “the affirmation [of the main clause] which is declared to be really the case instead of what might have been presumed to be the implication [due to the content of the subordinate clause] (Rudolph 1996: 391).” Example (15) illustrates a concessive clause which basically expresses the same content as the main clause; the notion of contrast is almost muted. (12)Although Sam had told the children a bedtime story, June told them one too. (Quirk et al. 2008: 1099)

Concessive clauses expressing real concession in general and an ineffective anticause in particular may occur in ante- and in postposition in all sample languages, except for the French encore que which is generally postposed (cf. Grévisse 2008: 1500). Rudolph sees slight semantic differences between the two structures but generally still groups some of the postposed concessive adverbial clauses to prototypical concession. In sentence-initial posi43

tion, “[...] it is clear that an ineffective obstacle is affirmed to be the case which under other circumstances and according to our world knowledge would prevent the main state of affairs from being realized (Rudolph 1996: 419).” Here, a strong connection to the concessive function of the ineffective anticause becomes obvious. When the concessive clause occurs in sentence-final position, the propositions display a different order. At first, we are confronted with the main clause assertion which may have various imaginable causes or reasons. Surprisingly, the subordinate adverbial clause does not mention the causes but exactly the opposite: an objection that would normally render the main cause invalid but in these circumstances remains ineffective (cf. Rudolph 1996: 419). However, these sentence-final concessive clauses may not only express an insufficient objection to the main clause content, they may also display quite different but related semantics: the semantic function of a restriction.

4.1.2

Restriction

The phenomenon of restriction as a function of concessive adverbial subordinators describes that “[t]he state of affairs affirmed in the main clause is restricted in its validity by the state of affairs affirmed in the concessive clause (Rudolph 1996: 411).” Thus, restriction “serves to make clearer what was meant and offers the opportunity of avoiding misunderstandings (ibid.).” (13)The Visigoths became Christians, although it was the wrong sort of Christians, Arians […]. (Rudolph 1996: 411) (14) Er war ein sehr großer Mensch, Anfang dreißig, obwohl er älter wirkte im ersten Augenblick. (Rudolph 1996: 411)

Restriction displays a number of semantic particularities in comparison to the prototypical concessive function of expressing an ineffective anticause. The content of the main clause is not seen as related to a particular cause or logical regularity which then is rejected by the concessive clause. Rather, its content is “checked for its validity (Rudolph 1996: 412).” Eventually, the content of the main clause is made more concrete by the restricting concessive clause. So for example the content of the concessive clause in example (16) offers information about the type of Christianity of the Visigoths, thus stressing its peculiarity. Example (17) does not only describe the appearance of a man but gives additional information on

44

what he really looks like. The concessive clause characterises him more closely and thus restricts the informational content of the main clause. What becomes obvious is that there is no causal chain or presupposition on the basis of world knowledge working in the background of these adverbial clauses. Rather, the concessive clause offers additional information which “limit[s] the meaning of the affirmation [of the main clause] as a whole or of one single aspect of it (Rudolph 1996: 413).” The rejection of a causal chain can be neglected in concessive clauses of restriction because the main clause usually expresses content which does not rely on causal presuppositions. As there is no causal chain working in the background, it naturally cannot be violated by a concession. In addition, the concessive clause itself does not encode an objection or anticause to the main proposition. Instead, it supports the truth of the main clause content by restricting it and thus specifying it (cf. Rudolph 1996: 414). So what is the relation between restrictive concessive clauses and prototypical concession? There is still a contrast which exists between the two propositions of main and subordinate clause. This contrast may be more or less intensive but it is always there (cf. ibid.). In example (17) above, the man described is thirty years old, but his appearance suggests otherwise as his looks are in opposition to his real age. Consequently, the contrastive notion of restrictive concessive clauses relates restriction to other concessive relations such as that of an insufficient objection and an antithetic contrast. Restrictive concessive clauses may only occur in sentence-final position. Semantically, only this position is possible: there must be some linguistic entity with content that can be restricted by the concessive clause, i.e. restrictive concessive clauses need an antecedent to specify its content - a single word, the whole preceding clause or even a whole stretch of discourse (cf. Rudolph 1996: 415). It is an ongoing linguistic discussion whether restrictive concessive clauses are still subordinate adverbial clauses. Rudolph refers to this discussion and decides to consider concessive clauses of restriction as subordinate. According to her sentence material, “the restriction type is of exclusively semantic characteristics (ibid.).” Additionally, she sees her conclusion supported by the choice of connectives in the restrictive function because restrictive concessive clauses are introduced by the same connectors that indicate prototypical concession (cf. Rudolph 1996: 415). According to the rules of the geometry of semantic maps, the 45

polyfunctionality of these connectives calls for a semantic connection between prototypical concession and restrictive concession on the semantic map. Another interesting view on this question is given by Susanne Günthner in her paper on the German concessive conjunction obwohl. She considers exclusively colloquial, spoken German discourse and besides a prototypical concessive notion of incompatibility between two propositions, she also reaches the concessive functions of restriction and even correction. However, her analysis leads her to the conclusion that restrictive and especially corrective obwohl clauses are no longer subordinate adverbial clauses and that obwohl takes over the function of a discourse marker rather than functioning as adverbial subordinator of concession (cf. Günthner 2000: 454ff.). This development is not confined to German concessive sentences as Günthner is able to point out on the basis of similar changes in the use of although in the English language.

4.1.2.1 German obwohl – on its way to a discourse marker? Günthner remarks that German obwohl can be used in subordinate and coordinate word order in colloquial German. She attributes the restrictive and corrective function of concessive markers solely to the coordinate structures: “[w]hereas obwohl-constructions with syntactic integration are used to indicate concession, those with syntactic non-integration are used to limit or correct the validity of the assertion made in the preceding utterance (Günthner 2000: 439).” German overtly distinguishes between syntactic integration and nonintegration due to the different word order structures. Subordinate clauses display verb-final word order while coordinate structures show verb-second word order. According to Günthner, the restrictive use of obwohl is limited to verb-second structure. What’s more, obwohl does no longer function as adverbial concessive subordinator in these surroundings but as discourse marker of correction and disagreement. Günthner observes that while truly adverbial concessive clauses show free word order and may occur in ante- or postposition and within the main clause, with verb-final position as well as in scope of a question (i.e. while adverbial concessive clauses show general features of subordination) restrictive and corrective uses of obwohl do not behave accordingly. However, Günthner identifies restriction as link between real (subordinate) concession and a corrective function of obwohl. The restrictive function of concession still shares features with prototypical concession: in both 46

cases, obwohl links linguistic elements and in both cases, there exists an underlying notion of incompatibility or contrast between those linked linguistic elements (cf. Günthner 2000: 453). Similar to Rudolph, Günthner states that the restrictive function of concessive clauses does not presuppose a causal chain or invalidate the main clause content. Rather, “it restricts the preceding statement or a conclusion which can be drawn from the preceding utterance (ibid.: 454).” As a link between concession proper and the corrective function of concession, restriction shows both subordinate and coordinate features. The subordinate clause restricts the validity of the main clause but does not rely on the presupposition “if p, then normally ¬q”. Still, both clauses in a restrictive concessive structure are entailed and valid, that is they express factual, truthful content (cf. Günthner 2000: 454f.). Restrictive concessive clauses show verb-final order indicating syntactic integration; however, their prosody in spoken language indicates two distinct utterances (cf. Ibid.). As they are necessarily postposed, they exhibit the character of an afterthought. Only after uttering the main clause, the speaker realises the necessity of restricting his/her utterance in order to specify its meaning (cf. Günthner 2000: 456). Although restrictive obwohl clauses can be encoded by verb-final word order, Günthner explicitly states that they “can no longer be classified as a subordinate conjunction (ibid.).” Thus, she takes the opposite stance to Rudolph, who distinguishes restriction from real concession on the basis of solely semantic properties.

4.1.2.2 Corrective function of concession The corrective function of the concessive connective obwohl differs even more from the semantics of prototypical concession. According to Günthner, “corrective obwohl has lost its syntactic, semantic, pragmatic and discourse-functional properties which are characteristic for concessive subordinators [...] (Günthner 2000: 456).” One of the uses of corrective obwohl is that of a self-correction. Here, the main clause content is asserted, but shortly after, a clause introduced by obwohl is added not only to restrict the validity of the main proposition but to correct the utterance (cf. Günthner 2000: 445). The propositions are not entailed and the semantics are no longer related to that of 47

real concession or of an ineffective anticause. It is not the case that the speaker ignores the incompatibility of the two propositions but conjoins them anyway, indicating that (s)he considers the two propositions to be compatible in this particular situation. In fact, the speaker knows about the incompatibility and uses the contrasting semantics in order to rectify the main clause content (cf. Günthner 2000: 445). Günthner very strongly emphasises that the corrective use of concessive connectors is no longer an instance of adverbial subordination. These sentences neither display verb-final word order, i.e. syntactic integration, nor are they prosodically integrated. In spoken language, they are usually separated by a pause and they show their own intonation contours (cf. ibid.). Günthner even interprets them as “major act (ibid.: 446)” both functionally and syntactically as they rectify the preceding clause and thus offer the correct main information; they are no longer syntactically integrated and consequently no longer subordinate. This reveals a very important aspect about the corrective function of concessive clauses. The corrective use of obwohl seems to be an instance of oral, colloquial language because in formal, written contexts, the verb-second position after adverbial concessive connectives is not allowed. The restrictive function with verb-final word order is a concessive function found throughout all registers of the German language and in other languages as well. Just like restriction, the corrective function shows strict postposition because an utterance needs to be made in order to be rectified (cf. Günthner 2000: 446). Corrective obwohl may refer back to “an explicitly stated utterance” as well as to the “correction of a possible implication (Günthner 2000: 448)”; thus, the correction may refer to factual content as well as to epistemic inferences. In addition, corrective concession does not only link clauses but even larger stretches of discourse. This is a strong argument for the thesis that corrective obwohl is no longer an adverbial subordinator but rather in the development of becoming a discourse marker (cf. Günthner 2000: 457ff.). Similarly to German, French employs the corrective concessive function as well. In the Grammaire méthodique du français, the concessive function of an auto-correction is mentioned: “il existe un usage ‘énonciatif’ de la concessive, qui consiste en une auto-correction [...] (Grammaire méthodique 2009: 862). In analogy to German obwohl and English although constructions of this type, these clauses always occur in postposition, i.e. sentence-final. According to the French grammar, their syntactic structure is close to coordinate structures, which is again a similarity to comparable structures in other languages. Consequently, it is 48

questionable whether these structures can still be regarded as a function of adverbial concessive clauses in French when their subordination is in question. Just as in English and German, French shows only one concessive connective with a corrective function, the connective encore que.

4.1.2.3 Attenuating a rejection – a function of disagreement? The very act of conceding becomes most obvious in its semantic and social implications in the function of attenuating a rejection. The Duden explicitly mentions this concessive function. Here, a concessive clause is used to politely reject the content of the main clause, especially in conversation (cf. ibid. 1986: 794). Günthner refers to this use of obwohlas well, when she observes that "[s]econd speakers frequently use obwohl to display their disagreement with their co-participant’s preceding assessments (Günthner 2000: 450).” In example (15), the second speaker Hans at first agrees with Klaus’ proposition that the unnamed beer is very good. However, after expressing an initial agreement, he words his disagreement with the help of an obwohl-sentence, thus politely rejecting Klaus’ claim. (15) 1 Klaus: das is echt s’BESTE BIER. (-) 2

ich mein von den alkoholfreien.

3

(-)

4 Hans: hhm. Obwohl es gibst schon BESSERE. 5

zum Beispiel BECKS is bei weitem TRINKBARER. (Günthner 2000: 451)

This poses the question if this function of expressing disagreement is related to the corrective function of concession – thus, the corrective function would not only cover expressions of self-correction but also corrections of other persons’ utterances. However, the function of attenuating a correction seems to be much more complex than just an expression of disagreement; consequently, this matter will be investigated separately in more detail in chapter 4.1.5. The corrective function both in the case of self-correction and in the expression of disagreement are linked to prototypical concession due to the contrasting, incompatible content of both clauses. However, according to Günthner, the concessive clause does no longer display the necessary characteristics of subordination such as syntactic and prosodic integration.

49

4.1.3

The relation between restriction and correction

In her analyses, Günthner strongly emphasises the German colloquial use of obwohl. However, the German connective is not the only one showing restrictive and corrective functions. Günthner finds similar structures in the use of the English concessive subordinator although which is often used with the notion of however, i.e. as a contrastive discourse marker in oral contexts (cf. Günthner 2000: 459). Japanese and Spanish display comparable structures as well. Still, Günthner notes that the corrective use of concessive connectives is a phenomenon of spoken language and thus not all-pervasive. The restrictive function on the other hand can be found in grammars of various languages and throughout spoken and written language. It still shows traits of subordination such as syntactic integration. According to Rudolph, it even is only a semantic variety of real concession. Like prototypical concession, it involves the incompatibility of two propositions and relies on their contrasting character. One could argue that it is rather a synchronous variation of the use of concessive connectives which may be in the development of grammaticalisation from adverbial subordinator to discourse marker. Thus, restriction can be considered an intermediate state between prototypical adverbial concession and coordinated corrective concession. As for the visualisation of concessive functions with the help of a semantic map, it is necessary to consider the semantic map of concessive adverbial clauses as a part of a larger network of general concession which does not only include adverbial functions but also coordinate structures. Consequently, the concessive function of correction might be a link to a broader network including other linguistic structures of concession, too.

4.1.4

Contrast and antithetic concession

The function of expressing a contrast or an antithesis is a central use of concessive connectives. Both propositions of a complex sentence expressing antithetic concession are entailed, factual and asserted. Thus, the function of concessive contrast is closely linked to the prototypical function of an ineffective anticause. Grammars such as the Duden and Quirk’s grammar on the English language directly refer to the function of concession to express a direct confrontation of two contrasting propositions conjoined by concessive connectives. In fact, concession and contrast are often presented as overlapping or even synonymous relations. Quirk et al. interrelate condition, concession and contrast, which according to them 50

becomes obvious in the multiple, overlapping use of the same subordinators in all three areas. Thus, if is an important part of conditional, concessive and contrastive subordinators and while and whereas introduce both concessive and purely contrastive clauses (cf. Quirk et al. 2008: 1088). In her book on the relation of contrast, Rudolph explicitly attributes concessive and adversative relations to the relation of contrast. According to her analysis, concession and adversation are the two representatives of contrast, with adversative relations being coordinate and concession being mainly subordinate (cf. Rudolph 1996: 14 and 18ff.). That contrast is an inherent notion of concession becomes obvious when we consider the basic semantics of prototypical concession: The two clauses or propositions express normally incompatible situations which are usually in opposition. Quirk even directly mentions in his definition that “[o]ften they [i.e. concessive connectives] also imply contrast between the situations described by the two clauses (Quirk et al. 2008: 1098).” Similarly, the Duden describes a function of concessive clauses that is used in order to express two propositions as being clearly in contrast and opposition (cf. Duden 1986: 794f.). Consequently, contrast is an important and fundamental aspect of concession, but it is not necessarily very obvious. Often, concession rather relies on a notion of incompatibility according to regularities and tendencies of experience in our every-day world (cf. Quirk et al. 2008: 1099). However, the contrastive or antithetic function of concession needs a sharp contrast between the two propositions of a complex clause. According to the Duden, at least in German, the second proposition is the favoured one. This favoured proposition correlates to the main proposition that cannot be invalidated by the contrasting subordinate clause. German expresses antithetic concession with the help of wenn auch. English encodes antithetic concession with while and whereas. (16)Wenn das Buch auch sehr gut ist, (so) ist es (doch) sehr wenig hilfreich. (Duden 1986: 696) (17)While he has many friends, Peter is often lonely. (Quirkt et al. 2008: 1099) (18)Whereas the amendment is enthusiastically supported by a large majority in the Senate, its fate is doubtful in the House (Quirkt et al. 2008: 1097).

whereas is most restricted in its use as it is only possible in antithetic concessive structures and “requir[es] antithesis between two situations [...] (Quirk et al. 2008: 1099).” while is more restricted in its use than for example the most versatile concessive connector al51

though: it is mainly used for contrastive concession but may also conjoin clauses which are not very obviously in contrast, i.e. “in which the contrast is muted (ibid.).” Another less frequent antithetic concessive connector is when, which also requires clear antithesis in order to express concession. It adds the semantics of “in that same situation” to concessivity (cf. Quirk et al. 2008: 1085). The distinction to other functions of concession, such as the prototypical semantics of an ineffective anticause, may be difficult to discern for the concessive connector while. This semantic similarity resurfaces in the closeness of these functions on the semantic map of concession. While English and German grammars directly refer to the antithetic function of concessive connectives, French and Latin do not openly refer to a similar use. French grammars do not indicate any type of concession which is semantically governed by the notion of contrast. According to Mikkola, who analysed concessive relations in Livy, there are ten different types of concession in Latin.7 None of those directly refers to contrastive concession; however, the function of the “komparative Konzessivität (Mikkola 1957: ?)” may be comparable to the antithetic function of concession. In this type of concession, the comparison between two entities or clauses may get a concessive notion. Therefore, both entities have to be dissimilar or in other words in contrast. Additionally, the comparative concession includes the typically concessive notion of surprise and unexpectedness when the comparison makes the dissimilarity between the entities obvious. Still, this comparison does not lead to the same semantics which underlie antithetic concession. Dissimilarity does not necessarily imply a sharp contrast or antithesis. It may simply refer to incompatibility between the two entities.

4.1.5

Einräumung – a function of disagreement

Especially German grammars repeatedly and frequently refer to a particular function of concession called Einräumung. Here, the subordinate clause concedes a situation or proposition that does not – in spite of its conceding force – suffice to invalidate the proposition of the main clause. Thus, it is referred to as Einräumung which somewhat unfortunately translates as “concession” (cf. Duden 1986: 794). As this definition sounds very much alike to the 7

In his book on Livy, Mikkola establishes adjunktive Konzessivität, soziative Konzessivität, komparative Konzessivität, temporale Konzessivität, konditionale Konzessivität, permissive Konzessivität, reservative Konzessivität, limitative Konzessivität, reversale Konzessivität and responsale Konzessivität. These different types of concessivity do not all refer to adverbial constructions but also cover coordinating structures (cf. Mikkola 1957).

52

main function of concession, i.e. expressing an insufficient objection or anticause, the Duden gives further information as to how to differentiate between the prototypical function of concession. While prototypical concession states an ineffective anticause, the function of Einräumung attenuates a rejection. In the case of an ineffective anticause, the concessive clause states an objective, factual and true fact that opposes the validity of the superordinate clause. In the case of an Einräumung, the subordinate clause expresses an irrelevant, conceding fact whose concessive and rejecting force is subjectively known from the beginning (cf. ibid.). Even with the help of these definitions, the difference between the two functions is hard to grasp. The Grammatik der deutschen Sprache proves helpful in search for an unambiguous differentiation between the function of an ineffective anticause and an Einräumung. According to the authors, concessives referring to an insufficient objection or an ineffective anticause offer a contrastive relation on the level of propositions, i.e. on the level of objective facts. Einräumungen rather refer to a subjective judgment as a comment, a so-called “Moduskommentierung (Grammatik der deutschen Sprache 1997: 2313)” originating from the epistemic level. Thus, a concessive clause which functions as Einräumung is based not so much on an objective, factual, insufficient anticause but rather on a contrasting subjective inference that underlies the concessive clause (cf. Grammatik der deutschen Sprache 1997: 2313ff.). (19) Und dieser weite Flug regte ihn auf, wenn er auch noch so gelassen wirkt. (Grammatik der deutschen Sprache 1997: 2313)

In this example, the concession derives from the opposition between the proposition in the main clause and the contrasting, subjective conclusion of the speaker expressed in the subordinate clause. (20) Rufen Sie bitte den Rettungswagen, [...] obwohl hier nichts mehr zu retten ist. (Grammatik der deutschen Sprache 1997: 2313)

Again, the concessive clause expresses a subjective inference and interpretation and not so much an objective anticause. All in all, the authors of Grammatik der deutschen Sprache state that obwohl rather expresses an insufficient objection and thus a so-called “Propositionsmodifikation (Grammatik der deutschen Sprache 1997: 2314). However, it may also be used for a “Moduskommentar (ibid.)”, i.e. for an Einräumung, just like in example (20). wenn auch on the other hand is ex53

clusively used for concessive sentences with the function of an Einräumung. The Grammatik der deutschen Sprache even distinguishes between different types of Einräumungen. obwohl expresses a self-correction, a partial alleviation of the superordinate clause. wenn auch rather encodes a possible objection which subjectively seems irrelevant: “Der vorangestellte wenn-auch-Satz artikuliert dabei einen geäußerten oder auch nur angenommenen Einwand gegen die Gültigkeit der geplanten Aussage, der zwar grundsätzlich anerkannt, jedoch eher distanziert betrachtet und als irrelevant abgetan wird (Grammatik der deutschen Sprache 1997: 2316).“ So in contrast to an objective, factual anticause which turns out to be nevertheless insufficient or ineffective, the function of an Einräumung is characterised by a subjective inference and the judgement to consider the contrasting content of the subordinate clause as irrelevant (cf. ibid.). Similarly to the concessive function of restriction and correction, conceding concessive clauses always occur postposed. Since restriction, correction and Einräumung show syntactic and semantic similarities, the question arises how these three functions of concession are interrelated? Syntactically, it is interesting that all three functions are always postposed. Günthner, as has been mentioned above, sees one of the corrective uses of concessive obwohl in the expression of disagreement, which is used in conversation by “a second speaker to frame an upcoming disagreement (Günthner 2000: 452).” Thus, with the help of a concessive clause, it is possible to reject another person’s statement in a polite and face-saving way as the concessive clause does not express an explicit, direct rejection but rather functions as the attenuation of a rejection (cf. ibid.). A similar use of concessive clauses is mentioned in the Duden. Here, the function of an Einräumung is closely related to the corrective function of concessives as it covers the use as self-correction based on an individual subjective inference concerning the incompatibility of two propositions of the clause. However, Einräumung also covers the correction of others as an expression of disagreement and rejection of the other person’s utterance or idea by presenting contrastive, incompatible content. This does not necessarily lead to a total invalidation of the preceding utterance but it may restrict it in its validity; thus, concessive clauses may function as an attenuated rejection which is more face-saving than a direct rejection of another person’s ideas. Accordingly, “Konzessivsätze werden verwendet, wenn man die Prämisse eines Arguments akzeptiert, aber die Schlussfolgerung ablehnt (König/Eisenberg 1984: 321f.).“ Socially, concessive clauses may be used as a polite way of expressing one’s disagreement with the ideas of the conversation partner. Due to the very se54

mantics of conceding, concessive clauses emphasise that there is a partial consent and acceptance of the premise of the sentence on behalf of speaker. Thus, the rejection is attenuated because a concessive clause at the same time emphasises the shared common ground of the accepted premise and still (if only partially) rejects the conclusion of the conversation partner. So according to Rudolph and Günthner, concessive clauses with the function of attenuating a rejection can also be characterised as a face-saving way of correction. Differences between the three functions of Einräumung, restriction and correction can be found on the level of their syntactic integration. While restriction and particularly correction are no longer considered to be adverbial subordinate clauses but coordinate structures of oral conversation, Einräumung seems to be a much more central function of adverbial concession, at least in German. Although it does have a restrictive and even corrective notion and usually occurs postposed, it is regarded as a function of adverbial concession.

4.1.6

Observations

A closer look at the semantics of the various functions of real concession reveals that the concessivity seems to decrease with the different uses. For the restrictive function, the underlying presupposition of prototypical concession, ‘if p, then normally ¬q’ vanishes in some instances. Similarly, in restriction, the surprising nature of the ineffectiveness of the opposite proposition is secondary. Syntactically, it is questionable whether restriction can still be considered an adverbial subordinate structure as it occurs both syntactically integrated and non-integrated and does not show all features of subordination such as being in scope of a question. The concessive function of correction can no longer be analysed as a subordinate structure as it is neither syntactically nor prosodically integrated. Both functions cannot be freely positioned in the sentence but they always and necessarily occur in postposition. Additionally, both functions do not rely on an underlying causal chain, which is closely connected to the lack of a presupposition in these two functions of concession. Truly adverbial, real concessive clauses on the other hand occur in sentence-initial, inserted and sentence-final position. The concessive function of contrast is closest to the prototypical concessive meaning of expressing an ineffective anticause or an insufficient objection although it depends much more on a sharp contrast or antithesis between the two propositions.

55

4.3 Hypothetical concession Besides real concession, all sample languages also show concessive clauses which do not express a factual but a hypothetical conceding proposition in the subordinate clause. The semantics of hypothetical concession originate in the conjunction of concession and hypotheticality, i.e. when “[...] a concessive conditional sets up a hypothesis, the realization of which, if it occurs, would bring into being a situation in which a concessive interpretation [...] then comes into play (Harris 1988: 72).” Grammars and linguistic analyses (cf. König 1998, 2000; Quirk et al. 2008; Harris 1988) differentiate between three types of hypothetical concessive functions. These three types correspond to the subtypes of concessive conditional clauses and their overall semantics: scalar concessive conditionals (SCC), alternative concessive conditionals (ACC) and universal concessive conditionals (UCC). Before we have a closer look at the definitions and characterisations of the respective functions, it is necessary to distinguish between concessive structures which seem to be very similar at first sight but in fact, can theoretically by differentiated (although the differentiation is complicated in everyday use). English employs the two subordinators even though and even if. Superficially, both seem to be rather similar due to the focal particle even and an undifferentiated use in colloquial language. However, there is a distinct semantic difference between the two concessive connectives. even though is an emphatic form of a real concessive, the anticausal though. The focal particle even emphasises the unexpectedness of the combination of the propositions, thus adding to the concessive notion of the subordinator. even if on the other hand “combines the concessive force of even with the conditional force of if (Quirk et al. 2008: 1099).” Example (21) illustrates the semantic difference that is triggered by the two different subordinators. (21)Even though you disLÌKE ancient monuments, Warwick Castle is worth a visit. (Quirk et al. 2008: 1099)

In this sentence, it is presupposed as truth that the addressee dislikes ancient monuments and consequently, should dislike the visit to Warwick Castle. Thus, it is a contrasting fact and an unexpected surprise that Warwick Castle will be an exception. Both propositions of the clause have a factual character. (22)Even if you dislike ancient monuments, Warwick Castle is worth a visit. (ibid) 56

Here, the adverbial clause leaves open if the addressee dislikes ancient monuments or not; hence, the adverbial clause has a hypothetical character. Similarly, German employs two very similar concessive connectives, auch wenn and wenn auch. auch wenn is used in conditional clauses and irrelevance conditionals, i.e. in hypothetical concessive clauses. It implies only the factual value of the main clause (the consequent) and not that of the subordinate concessive antecedent. Thus, it may be used in hypothetical concessives (cf. Grammatik der deutschen Sprache 1997: 2312). wenn auch asserts the factual value and truth of both sub- and superordinate clause. It is often used in the phenomenon of Einräumung in the function of a Moduskommentierung (cf. Grammatik der deutschen Sprache 1997: 2313ff.). As a last preliminary point, it is necessary to analyse how hypothetical concessives and real concession are related. “Hypothetical concession is predominantly characterised by the fact that only the states of affairs mentioned in the concessive clauses are hypothetical, whereas the states of affairs in the main clauses are clearly not hypothetical but are facts of reality (Rudolph 1996: 438).” So while in the prototypical concessive relation the content of the concessive subordinate clause is asserted to be true and factual but insufficient to be an obstacle to the true and factual main clause, in hypothetical concession, the proposition of the subordinate clause is not asserted, i.e. not factual but hypothetical (cf. Rudolph 1996: 422). This hypothetical content is presented as an irrelevant obstacle to the realisation of the main clause content; however, in hypothetical concessive clauses, it may be two or even any imaginable number of hypothetical events which oppose the content of the main clause (cf. ibid.).

4.3.1

Scalar concessive conditionals

Scalar concessive-conditionals (SCC) express an imaginable state of affairs which may oppose the main clause proposition. There are two subtypes of scalar concessiveconditionals; those “which consist of a conditional clause plus a scalar additive particle (even)” and those “marked by a subordinator that also marks concessive clauses (König/Haspelmath 1998: 584).” Interestingly, scalar concessive-conditionals seldom have special subordinators used exclusively for their structures.

57

4.3.1.1 Formal description The form and semantics of the first type of SCCs are comparatively transparent as they consist of a focus or scalar particle and a conditional subordinator in all languages of the sample. It is the focus particle that adds the concessive semantics to the subordinator. In languages expressing subordination with finite constructions, the scalar particle precedes the conjunction and the conjunction precedes the clause. Typical subordinators of scalar concessive-conditionals are even if in English, auch wenn in German, même si in French and etiamsi, etsi, tametsi in Latin (cf. König/Haspelmath 1998: 585). Semantically, the focus particle has scope over the whole subordinate clause: “[t]he situation expressed by the antecedent [i.e. the subordinate clause] is thus characterized as an extreme case for an open sentence of the form ‘if x then q’ (König/Haspelmath 1998: 586).” Some languages employ the same markers for both concession proper and hypothetical concession. According to König and Haspelmath, in these cases, the only difference in the marking and in the semantics is expressed with the help of mood distinctions. Accordingly, concession proper is usually expressed in the indicative while hypothetical concessive clauses display subjunctive mood. In Early Middle English, even though was used with the subjunctive, thus introducing a hypothetical concessive clause. Nowadays, even though introduces real concession and takes the indicative which indicates a semantic shift from concessive-condition to real concession (cf. König/Haspelmath 1998: 589f.). Many languages, including English and German, allow for their markers of scalar concessive-conditionals, such as even if and auch wenn, to be used in real concessive contexts. This may be considered as an indication that both subordinators develop from primary concessive-conditional meaning to a purely concessive one. However, so far, this purely concessive meaning of scalar concessive-conditionals is secondary as scalarity is a feature of concessiveconditionals, not of pure concession. In fact, pure concessives do not include scalar focus particles (cf. König/Haspelmath 1998: 592). Still, this semantic development hints at a very important point concerning the semantic relationship between real concession and hypothetical concessives. According to Rudolph, scalar concessive-conditionals show a close similarity to real concession, the only difference being the hypothetical nature of the proposition in the subordinate clause. This hypotheticality becomes obvious as “the mentioned state of 58

affairs as well as its opposite [...] are considered as a mere possibility, and it is indicated that it is irrelevant whether they become reality or not (Rudolph 1996: 423).” Languages usually develop special connectives for scalar concessive-conditionals; as a consequence, all sample languages show particular concessive-conditional connectives. For scalar concessiveconditionals, English usually employs even if, German auch wenn, French même si and Latin etiamsi, etsi and tametsi. All languages indicate the close connection between hypothetical concession and condition, especially in scalar concessive-conditionals, as their connectors all use various forms of the conditional subordinators if, wenn, si.

4.3.1.2 Functional description In order to understand the semantics of scalar concessive-conditionals, it is necessary to realise that the proposition of the subordinate clause is considered “an imaginable obstacle which under normal circumstances of our world knowledge would prevent [the main] proposition [...] from being realized (Rudolph 1996: 436).” This definition shows the semantic closeness to prototypical concession. Interestingly, it is irrelevant if the events of the subordinate proposition are existent or not, the main clause proposition is realised in both cases anyway. In other words, the imaginable event and its unexpressed negation are irrelevant for the realisation of the main clause content (cf. ibid.). Similarly to prototypical concession, the causal chain is working in the background and violated by the realisation of main proposition. This similarity indicates relative spatial contiguity on the semantic map of concession. This proximity both in abstract semantics and its visualisation on the semantic map is supported by the emphasis of the ineffectiveness and insufficiency of the subordinate clause content. The main clause proposition can neither be invalidated by the imaginable though hypothetical subordinate proposition nor by its negative counterpart; both possibilities remain utterly ineffective (cf. Rudolph 1996: 436). In some cases, it is difficult to distinguish between real concession and scalar concessive-conditionals, at all. German, for instance, allows auch wenn and wenn auch in many contexts which ask for real concessive connectors such as obwohl. Often, only the larger context gives clues as to which concessive relation is meant. This use is possible as in both cases one event is irrelevant for the realisation of another contrasting event. As the conceding event is without effect, it is not important if it is real or not (cf. ibid.). 59

4.3.2

Alternative concessive conditionals

While scalar concessive-conditionals express only one imaginable but hypothetical objection to the main clause content, alternative concessive-conditionals (ACC) comprise the “explicit expression of two alternatives (Rudolph 1996: 426).” In more detail, Rudolph states that “[f]rom the set of imaginable conditions that may be connected with the main state of affair two alternative states of affairs are selected and declared not to have any relevance for the main state of affairs to be realized (Rudolph 1996: 427).”

4.3.2.1 Formal description According to König and Haspelmath, alternative concessive-conditionals show a more diverse structure as there are five different formal subtypes. The first class of subordinators is based on conditional expressions meaning if... or if. Latin is the only European language that employs an instance of this type of marking with the connector sive…sive. (23)Si-ve pluit si-ve sol lucet, (tamen) discedimus. ‘Whether it rains or the sun shines, we will go (outside).’

At least related to this structure is the English alternative marking of even if... or (even if). The second type of markers is based on interrogatives. In Germanic languages, alternative concessive-conditionals have a very similar form compared to interrogative clauses. In fact, they are more closely related to interrogatives than to conditionals. This becomes obvious in the fact that these languages rather use interrogative markers than conditional ones in alternative concessives. Thus, German does not employ the conditional wenn but the interrogative ob or ob…oder. Similarly, English usually introduces alternative concessive conditionals with whether…(or), i.e. with an interrogative subordinator and a disjunction (cf. König/Haspelmath 1998: 596f.). In a third type of alternative concessive conditionals, the verbs of both the subordinate and the main clause occur in the optative or subjunctive mood. These clauses are characterised by the lack of an overt subordinator. Interestingly, not only French shows examples of this particular sentence type. English displays similar structures as well. However, these structures show a distinctly different syntax with the verb preceding the subject. 60

(24)Come wind or rain, we will climb the mountain. (König/Haspelmath 1998: 599)

In French, this type of alternative concessive-conditionals may be introduced by the general subordinator que or show verb-subject inversion when there is no general subordinator (cf. König/Haspelmath 1998: 598f.). A fourth type of subordinators introducing alternative concessive-conditionals originates from the clause you want. König and Haspelmath state that this type of concessive subordinators is only a marginal phenomenon in Europe. They see it as semantically related to (marginal) expressions of scalar concessive-conditionals originating from let it be or it may be (cf. König/Haspelmath 1998: 599f.). The only sample language with a clear etymological reference to a form of you want is the Latin concessive conjunction quamvis. However, it rather fulfils the semantic and formal characteristics of universal concessive conditionals or even proper concession. Two minor strategies of coding alternative concessive conditionals are “an embedded interrogative clause dependent on an expression of irrelevance (it does no matter, it is all the same) (König/Haspelmath 1998: 601)” and syntactic reduction. According to the analyses of König and Haspelmath, none of the sample languages shows alternative concessive conditionals originating in expressions of irrelevance. While English does employ no matter phrases in hypothetical concessives, these are restricted to the use in universal concessive conditionals. In structures of reduction, “[...] neither the subordination nor the disjunction is overtly expressed (König/Haspelmath 1998: 602).” Thus, these structures do not qualify as subordinate adverbial clauses and are not considered in this analysis.

4.3.2.2 Functional description The semantics of alternative concessive-conditionals are quite different to that of prototypical concession, i.e. of an insufficient objection. Thus, they are also different from scalar concessive-conditionals. In ACCs, the semantics are not so much about contrary anticauses but rather about two “alternatively imaginable causes for the main action (Rudolph 1996: 428).” Since both causes are mentioned, neither of them really is the cause; thus, they are ineffective and irrelevant. Nevertheless, both could have the effect or could be the cause of the main clause content but the speaker does not give any further detail on this matter. Consequently, both causes are ineffective and obviously irrelevant in their informational 61

force (cf. ibid). Compared to prototypical concession, alternative concessive-conditionals express ineffective causes and not an ineffective anticause. Their irrelevance derives from the fact that it is irrelevant which of the two alternative propositions of the subordinate clause is the cause for the main clause proposition. Following this, ACCs do not express an anticause or objection but rather about the exact contrary. A notion of opposition or contrast between the two clauses cannot be observed; alternative concessive-conditionals characterised by a strong notion of irrelevance.

4.3.3

Universal concessive conditionals

“[I]n this third class, a set of states of affairs or a set of properties of one event is imagined, indicating a free choice from any number of alternatives (Rudolph 1996: 428).” Analyses by Haspelmath and König show a wide variety of forms which may encode universal concessive-conditionals throughout different languages. Of the four sample languages, English has a clear and distinct set of expressions introducing universal concessive-conditionals and German shows quite similar structures. French and Latinas well as Romance languages in general, have developed quite different syntactic structures.

4.3.3.1 Formal description Universal concessive conditionals [...] exhibit great formal diversity in the languages of Europe. However, there is one formal element that is shared by virtually all types of UCCs [sic. universal concessive conditionals] in Europe [...]: the parameter that is presented as irrelevant for the validity of the consequent is expressed as an interrogative pronoun or at least as a pronoun based on an interrogative pronoun [...] (König/Haspelmath 1998: 604).

König and Haspelmath distinguish between eight subtypes of universal concessive conditionals. The crucial role that interrogative pronouns play in UCCs supports the theory that hypothetical concessives are closely related to questions. Questions as source construction for hypothetical concessive clauses become especially convincing when one considers that alternative concessive conditionals are related to embedded alternative questionsas well (cf. König/Haspelmath 1998: 606). However, another possible source construction for universal concessive conditionals are headless free relative clauses. Indeed, many grammars even 62

subsume free choice concessive conditionals under the label of free relative clauses, for example the Duden or French grammars. This is possible because free relative clauses in European languages make use of interrogative pronouns which tightly interweaves these three constructions. According to König and Haspelmath, this relation becomes especially obvious in languages which use a special paradigm of pronouns for free relative clauses as this paradigm is usually used for universal concessive conditionalsas well (cf. König/Haspelmath 1998: 606). As there are eight ways of coding universal concessive conditionals, only the formal characteristics of UCCs in the sample languages will be examined. The first and most common type of universal concessive conditional connectives consists of an element like ever, a form of want, also or that following a WH-pronoun. In connectives with ever, this part fulfils the function of emphasising the irrelevance of the subordinate clause content. German and English show whole paradigms of hypothetical concessive subordinators following various WHpronouns. Latin shows similar structures including expressions with elements that etymologically originate in a form meaning ever: in the form of qui-cumque, in which -cumque originally meant when, ever and and (cf. König/Haspelmath 1998: 609ff.). Different expressions of want prove similarly fruitful for universal concessive conditionals and do not only occur in alternative concessive conditionals. Again, the element follows the WH-phrase or –pronoun. Semantically, “[...] the element (you) want expresses the irrelevance of the antecedent by leaving the choice to the hearer (König/Haspelmath 1998: 611).” Structures of you want or the parallel but more objective it may be underline the free choice aspect of universal concessive conditionals as these formal structures encode the unlimited reference and at the same time the irrelevance of the reference of the WH-pronoun or -phrase. The French phrase qui que ce soit is a representatives of a universal concessive structure of that type and is an equivalent for the more common qui que. (25)Qui que ce soit que je rencontre, je lui parle. (König/Haspelmath 1998: 611) ‘Whoever (it is that) I meet, I talk to him or her.’

Especially in French, a common strategy of introducing universal concessive conditionals is the use of that after a WH-element. An example is the subordinator quoi…que (cf. König/Haspelmath 1998: 612f.).

63

A very interesting way of introducing universal concessive conditionals is the strategy of reduplication. The only case known in standard European languages and in the sample languages is Latin; otherwise, reduplication is comparatively rare in these structures in European languages. Latin employs subordinators like quisquis and quidquid which etymologically function as source for the proper concessive marker quamquam. This illustrates the general path of grammaticalisation of concessive connectors which usually develop from a marker of hypothetical concession to a marker of real concession. A very important and central type of universal concessive conditional subordinators are expressions of the no matter-type. As König and Haspelmath observe, “[p]robably all languages allow the possibility of expressing the equivalent of a UCC by means of an embedded interrogative clause dependent on an expression of irrelevance (it does not matter, it is all the same) [...] (König/Haspelmath 1998: 618).” These subordinators most impressively express the total irrelevance of any content of the subordinate clause as the main clause can never be invalidated. Example (26) and (27) illustrate the use of these conjunctions. (26) No matter what she says, he keeps quiet. (König/Haspelmath 1998: 618) (27) Egal was sie auch tut, er liebt sie.

Besides these strategies of expressing universal concessive conditionals which all centre on a WH-element, UCCs may also be introduced by subordinators which do not include such an element. These subordinators are possible in German and in Romance languages, for example in French. Usually, they involve an expression of degree of a property encoded by an adjective (cf. König/Haspelmath 1998: 618). According to the Grévisse, French also employs structures which refer to expressions of degree encoded by an adverb or a noun. Typical subordinators in French are aussi…que and pour…que, which are only found in Romance languages. Structurally parallel to French aussi…que, German employs the connective so…auch (cf. König/Haspelmath 1998: 619). (28)So verschwenderisch du auch bist, das ganze Geld kannst du gar nicht ausgeben. (29)Aussi méticuleux que soit le règlement, il ne parvient pas à tout prèvoir. (30)Ce texte, pour intèressant qu’il soit, n’est pas probant non plus. (König/Haspelmath 1998: 619)

64

4.3.3.2 Functional description In universal concessive-conditionals, the notion of irrelevance again plays a very important role. The main clause content is realised irrespective of any imaginable fact or event expressed in the subordinate clause. Thus, the subordinate propositions are totally irrelevant because “[e]ach imaginable, but not expressed, property may be regarded as an obstacle usually preventing the main state of affairs from becoming reality, but it is ineffective (Rudolph 1996: 437).” In universal concessive-conditionals, the notions of anticause, objection and contrast are more obvious than in alternative concessive-conditionals. However, the exact properties of the opposing set of events are not mentioned, which leads to the hypotheticality of universal concessive-conditionals. This lack of details induces a notion of free choice that originates either in the irrelevance of these details or in the lack of knowledge about these details on the side of the speaker. In both cases, the proposition in the main clause is in focus as it is realised in any case. It is of general validity and is able to fulfill the communicative aim of the speaker even without any details (cf. Rudolph 1996: 429ff.).

4.4 Observations In hypothetical concessive clauses, the notion of contrast is somewhat minor. However, in the descriptions of their semantics, it becomes obvious that irrelevance plays a major role in the meaning and function of concessive conditionals. In fact, irrelevance is the predominant notion of hypothetical concessives and an inherent part of their concessive semantics; it is what makes them concessive in the first place. Thus, the concessivity of hypothetical concessive clauses does not so much consist in the denial of a causal link between the two propositions of the complex clause, it rather originates in the irrelevance of the subordinate clause content for the realisation of the main clause content. Accordingly, they are also often called irrelevance concessives.

5

The geography of the conceptual space of concession In the following paragraph, I will explain the distribution of the functions of concession

in conceptual space. This geography is the basis for the development of the languagespecific semantic maps, their analysis and all deductions drawn from them. 65

The geography of the conceptual space shows that the space is divided into hypothetical concession and real concession.

Einräumung UCC

ACC

SCC

restriction

correction

ineffective anticause contrast

Fig. 5 Conceptual space of concession

The conceptual space of concession is a continuum from hypotheticality to factuality. While universal concessive conditionals at one pole of the space express any imaginable number of irrelevant and insufficient objections and are thus necessarily hypothetical, scalar concessive conditionals are tightly linked to real concession; its markers are often used in real concessive surroundings as the differentiation between hypotheticality and factuality gets weakened and vanishes. Consequently, the use of SCC markers in both hypothetical and factual functions requests a contiguous, adjacent position of scalar concessive condition and real concession (compare Fig. 5). Additionally to the basic division of the conceptual space in hypothetical and real concession, the distance between the various functions and the central and prototypical use of concession mirrors their relation to this fundamental concessive function. The closer the functions are to the prototypical meaning, the closer related they are both syntactically and semantically. Scalar concessive condition is thus semantically very similar to prototypical concession as both express an insufficient objection or ineffective anticause. The major difference is the hypothetical nature of the objection in SCCs and their distinctly extreme degree or value which is nevertheless irrelevant for realization of the main clause content. The extreme nature of the subordinate clause content is not necessarily part of the prototypical meaning of concession but it is necessary in SCCs to underline the irrelevance of the subordinate proposition. In scalar concessive condition, this is what establishes the concessive meaning in the first place. Still, SCCs and prototypical concession are undeniably similar which motivates their adjacent position. The concessive function of Einräumung is overtly and distinctly described only in the Duden and other German grammars. However, the function is not a German particularity. 66

Einräumung is closely related to prototypical concession, which can be seen in the geography of the conceptual space. Einräumung

Einräumung

SCC

UCC

ACC

SCC

restriction

correction

ineffective anticause contrast

Fig. 6 The conceptual space of concession/ SCCs and Einräumung

It is directly adjacent to restriction and the function of an ineffective anticause, but only indirectly related to the function of concessive correction. It clearly belongs to the semantic field of concession as the speaker concedes a fact that does not suffice to invalidate the main clause: thus, subordinate and main clause are in a truly concessive relationship. The semantic particularity of Einräumung is that the subordinate clause content relies on a subjective, individual conclusion or inference of the speaker. In other words, the two propositions are no longer objectively incompatible or in contrast; rather, the content expresses the speaker’s individual and subjective evaluation of the main clause which nevertheless does not suffice to render the main clause untrue. Thus, the concessive clause comes close to a comment on the main clause, which puts it semantically close to restriction. Because of this commenting function, concessive clauses in the function of Einräumung are often used to attenuate a rejection as the notion of a comment alleviates the negative effects of a direct rejection. The third concessive function that is directly linked to prototypical concession is the function of antithetical concession.

Einräumung UCC

ACC

SCC

restriction

correction

ineffective anticause contrast

Fig. 7 The conceptual space of concession/ contrast

Here, the semantic link is clearly the fact that antithetical concession and prototypical concession combine two clauses whose content is clearly not compatible but would be mu67

tually exclusive in the real world. However, the antithesis between the subordinate and the main proposition usually is a very sharp one in the concessive function of contrast; the semantics emphasise the contrastive aspect of the concessive relation. Again, the direct adjacency and link between prototypical concession and the function of concessive contrast is additionally motivated by concessive connectives that both functions have in common. In order to further analyse the relationships between the various functions of concession, one has to think about a fundamental and important question: why is the function of expressing an insufficient objection or ineffective anticause considered the prototypical one? As prototypical function of concession, it is located in the middle of the conceptual space and serves as interception of hypothetical and real concession. Furthermore, it is the concessive function with the most links to other functions, both hypothetical and factual. Most importantly, the central position of this function is motivated by the fact that the expression of an ineffective anticause is the basic definition of concession in grammars and linguistic treatises. Thus, its prototypicality bases in the numerous definitions of concession in linguistic works (compare chapter 3.2). One may safely assume that it is this function which all languages express that employ an adverbial concessive relation. Its central position is additionally supported by its numerous semantic and formal links to other functions like scalar concessive condition, restriction and antithetical concession. Thus, the principle of contiguity illustrates its prototypicality and fundamentality even further. The most frequent, most neutral markers that cover the highest number of functions also encode prototypical concession and in all probability elaborate on this basic, central function. Interestingly, a comparatively high number of connectives throughout languages cover the function of prototypical concession, once more underlining its fundamentality in the relation of concession. Connectives that usually express hypothetical concession may even spread to the prototypical concessive use. This fact additionally motivates the central role of the function of an ineffective anticause in the middle of the network. The distribution of Einräumung, restriction and correction in the domain of real concession on the map is a little more difficult to explain as they are semantically closely related. Their interrelatedness is represented in the visualisation of real concession:

68

Einräumung

restriction

correction

ineffective anticause contrast Fig. 8 The conceptual space of real concession/ Einräumung – restriction – correction

Restriction shows semantic similarities and overlap with the function of an ineffective anticause. Both functions use the same subordinators and both show overt signs of syntactic integration such as verb-final word order in German. However, concessive clauses in the function of restriction are strictly sentence-final and are used in order to specify the validity of the main clause content. The function derives from a syntactic variation of fundamental, prototypical concession (Rudolph 1996: 410ff.). In restriction, the subordinate concessive clause always follows the main clause while prototypical adverbial concessive clauses may occur both in ante- and postposition. When they occur in postposition, the hearer is not prepared to be confronted with an ineffective obstacle to the main clause. Based on experience and our argumentative world view, he or she would expect a cause or logical explanation for the main clause. Surprisingly, the postponed adverbial clause does not only offer the opposite but an ineffective opposite (cf. Rudolph 1996: 392ff.). In the function of restriction, the concessive clause is always necessarily postponed. However, restriction does not offer anticauses or objections to the main clause which violate some underlying logical chain, a relation of cause and effect or tendencies of events co-occurring in the world. Rather, restriction in the sense of the word restricts the validity of the main clause by offering information that is partially opposing or contrasting to the main clause content. Thus, there is a partial incompatibility between the two clauses and their content. This incompatibility of nevertheless co-occurring facts restricts the validity and informational breadth of the main clause content because it specifies the semantic information and consequently the truth of the “remaining” part of the content. So, in restriction, the semantics do not refer to an event or fact that would usually prevent the main clause content from being realised. However, the contrast and incompatibility of the subordinate and main clause content remains – and that is what links restriction to prototypical concession. Additionally, all four languages employ at least one conjunction which covers these two functions, thus demanding direct adjacency to secure a contiguous mapping (compare chapter 2.3.2).

69

At the other pole of the conceptual space stands correction. In the geography of the network, it occurs farthest away from prototypical adverbial concession and thus, constitutes an end point of the semantic continuum of adverbial concession. One reason for this position in the conceptual space is that it is questionable if the concessive function of correction is syntactically still subordinated and still an adverbial relation. This controversy was illustrated with opposing viewpoints by Rudolph and Günthner who express different opinions whether correction is subordinated or coordinated. Proof for a coordinated structure is the verb-second word order in German corrective concessive clauses. Additionally, the concessive connectives in this function are more or less discourse markers and no longer adverbial conjunctions (cf. Günthner 2000: 457ff.). Another reason for the position of correction at the edge of adverbial concession are the semantics of concessive correction. It is more closely related to restriction and Einräumung than to prototypical concession. It still grounds in a basic notion of incompatibility and different degrees of contrast, which keeps the function of correction within the conceptual space of concession. Its semantic relationship to restriction becomes obvious when one considers that restriction narrows down or constraints the informational content and reference of the main clause, thus making it more specific. This function has a slight corrective notion but rather in increasing the correctness of an utterance than in a rectification of it. The semantics of Einräumung include a corrective part as well. As a form of a subjective comment on the main clause content, concessive clauses in the function of an Einräumung may easily develop into a corrective comment either on one’s own utterance or even more on another person’s utterance in a conversation. Still, typical for concession, the two propositions of the main clause and the concessive clause, although it is no longer subordinate, are incompatible when occurring at the same time. These concessive notions make the corrective notion possible but still posit it quite far away from prototypical concession. In summary, restriction, Einräumung and correction are hard to distinguish and semantically close in the domain of real concession. They show syntactic similarities such as strict postposition and an increasing development towards coordinate structures. However, in all sample languages they are encoded by the same subordinators or concessive connectives that also encode prototypical concession. The principle of contiguity demands that all functions are thus distributed in a way that allows contiguous mapping of the markers on the conceptual space. 70

5.1 Description and analysis of the language-specific semantic maps The language-specific semantic maps show similarities and differences which can be explained by their common origins as European languages on one hand and their different language families and branches on the other. Thus, all four languages show distinct, particular markers for hypothetical concession which show very similar structures and origins.

5.1.1

Hypothetical concession

In all sample languages, universal concessive conditionals (UCC) are expressed by connectives that base on interrogative or relative pronouns and an adverbial expression such as ever, auch immer or -cumque. Consequently, free choice concessives are closely connected both structurally and semantically to indefinite pronouns and expressions of indefiniteness. Haspelmath notes something quite similar in his typological approach to indefinite pronouns. He states that concessive structures, in particular parametric, i.e. free choice concessives, may be the source construction for indefinite pronouns and their function (Haspelmath 1997: 136ff.). Thus, in all four sample languages, universal concessive conditionals could then be the source for the development of indefinite pronouns. Additionally, all four languages show connectives for UCCs that express only UCCs. qu…cumque/quisquis quoi…que, (aus)si…que, pour…que UCC

ACC

wh…ever

SCC

Einräumung

restriction

correction

ineffective anticause contrast

w…auch immer Fig. 9 The semantic map of universal concessive condition/English, German, French, Latin

In other words, they do not code any other concessive function but are restricted to free choice concession in hypothetical concessives. This formal restriction to universal concessive condition is probably motivated by the semantics of the function. The geography of the semantic map and therefore ultimately of the underlying conceptual space shows universal concessive condition at the far left, farthest away from prototypical concessive meaning, i.e. the expression of an ineffective anticause. UCCs are characterised not so much by an ineffec71

tive anticause but by a strong notion of irrelevance as well as indefiniteness and indetermination. The subordinate clause does not express a factual objection to the main clause but rather refers to the irrelevance of any possible, imaginable cause or anticause of the main clause content. The nature or any exact references of the subordinate clause contents are not indicated as they are unimportant in the light of the utter irrelevance of the possible objections. They are not factually given. These particular semantics of the free choice concessive function do not include or even approach any other function of concession. Consequently, the formal markers are restricted to this concessive function. Similarly to universal concessive condition, alternative concessive condition (ACC) is encoded by particular conjunctions that do not refer to any other function of concession. In three of the four sample languages, these structures show fundamental similarities. A clear exception is French. French does not employ any overt subordinator in ACCs but uses the subjunctive mood of the subordinate verb forms to indicate alternative concessive condition. Interestingly, the Romance ancestor language Latin does employ a particular connector for ACCs, sive…sive. -ve is a disjunctive particle meaning or which is attached to its antecedent; thus, the Latin marker is quite similar to English and German, which also employ disjunctive connectors in ACCs. si- goes back to the conditional conjunction meaning if. Consequently, sive…sive refers to an alternative state of affairs, the meaning of two alternatives which interestingly may function as cause and not as anticause of the main proposition. Differently from Latin, English and German do not employ recent conditional connectors but historical markers of condition. Nowadays, these markers are used for embedded alternative questions and complements (cf. König/Haspelmath 1997: 597). Alternative concessive condition is semantically quite different from prototypical and real concession. However, the subordinate clause does no longer express any possible or imaginable number of objections as it does in universal concessive condition. Rather, it is restricted to two entities which are not anticauses but more or less possible causes of the main clause content. Often, these causes are in an opposite relationship, which already makes the subordinate clause content implausible and thus irrelevant. Consequently, the concession of ACCs derives again from irrelevance, i.e. the irrelevance of the consideration which of the alternative possibilities is the real cause for the main clause content or if the subordinate clause expresses cause and anticause at the same time. The main clause content is realised no matter what possible objection or possible reason there may be. So simi72

larly to universal concessive conditionals, alternative concessive conditionals are also characterised by a strong notion or irrelevance, but in contrast to UCCs, there is no unrestricted number of objecting alternatives encoded by the subordinate clause. Semantically, alternative concessive conditionals are also close to scalar concessive conditions which express only one possible objection. Thus, they are presented as related on the map; still, there is not a single instance of a marker that occurs in ACCs and SCCs. sive…sive

Einräumung UCC

ACC

SCC

restriction

correction

ineffective anticause

whether…or contrast ob…oder Fig. 10 The semantic map of alternative concessive conditionals /English, German, Latin

Scalar concessive conditionals behave a little differently from other hypothetical concessives. Both semantically and structurally, they are the intersection between hypothetical concession and real concession. They are not only closely related to real concession but also to its fundamental and prototypical function of expressing an ineffective anticause. Thus, the concessive function of expressing scalar concessive condition occurs as link between hypothetical and real concession in the geography of the conceptual space and the languagespecific semantic maps. même si

Einräumung

even if

UCC

ACC

SCC

restriction

correction

ineffective anticause contrast

auch wenn+ etsi, quamvis etiamsi, tametsi Fig. 11: the semantic map of scalar concessive condition/English, German, French, Latin

In German and Latin, the scalar concessive connectives etsi, quamvis and auch wenn can be used in real concessive circumstances in the function of expressing an insufficient objection or an ineffective anticause. Rudolph even goes so far as to call scalar concessive conditionals and concessive clauses of an ineffective anticause essentially synonymous (cf. Rudolph 1996: 436). While prototypical concession refers to a factual objection that does not 73

have enough force to invalidate the main clause, scalar concession refers to a hypothetical objection that – independently of the degree, value or extreme nature of the opposition – does not suffice to invalidate the main clause. Hence, in both cases the main proposition is realised in spite of an ineffective objection; it does not make much of a difference if this objection is factual or hypothetical. Rudolph’s thesis explains why Julius Caesar used the originally hypothetical connective etsi as a real concessive conjunction, why auch wenn can be found quite frequently in prototypical concessive function and why even if and even though are used mutually interchangeably in English. Additionally, this thesis offers an explanation why markers of concessive condition often develop into proper concessive conjunctions, a phenomenon that can be observed quite frequently and cross-linguistically (compare the development of though and especially of etsi, quamvis or auch wenn, Fig. 15). The markers of scalar concessive condition have comparatively similar forms in all four sample languages. As has been indicated in chapter 4.3.1.2, SCCs are semantically and formally close to conditional clauses. This becomes obvious in all four languages. They all employ complex forms with one part of the form originally deriving from the conditional markers si, if and wenn. The second part of the connectives is a scalar or focal particle, either etiam in Latin, even in English, même in French or auch in German. The comparable structure in all four languages shows the close similarities in the coding of this function. This phenomenon might be motivated by the close genetic relationship of the languages, as they are all members of the European language family. However, the structure of these connectives might also be universal in this function, which can only be found out by analyses of more languages of more diverse genetic origins.

5.1.2

Real concession

The prototypical function of concession, the expression of an ineffective anticause, is naturally a part of the repertoire of all languages. If a language expresses a concessive relation, it necessarily includes the prototypical function of an ineffective anticause or an insufficient objection. In all four sample languages, the most frequent and stylistically neutral conjunctions are used in this function of expressing an ineffective anticause: quoique in French, quamquam in Latin, although in English and obwohl in German. Except for French, these connectors are not restricted to the function of an insufficient objection. In English, although 74

covers for different functions, German obwohl expresses even five functions and Latin quamquam is also polyfunctional in the coding of four concessive functions. Thus, the prototypical subordinators of adverbial concession are not only the most frequent ones, but also the most versatile connectors. quamquam

Einräumung UCC

ACC

SCC

restriction

correction

ineffective anticause contrast

quoique obwohl although

Fig. 12 The semantic map of real concession/English, German, French, Latin

One reason for this phenomenon may be that the other functions of concession elaborate on the prototypical, fundamental function. Furthermore, it is a normal process of grammaticalisation that forms become more grammatical and take over more and more elaborate grammatical functions. In the case of adverbial concession, the concessive conjunctions may take over new functions as they are semantically widened. Additionally, the frequency increases with the new functions and in turn, a high frequency leads to forms which are prone to be used in slightly different semantic circumstances because they are ready at hand and easily understood even with a slight extension of meaning. In the concessive function of restriction, the slightly different semantics derive from a different syntactic behaviour, i.e. the strict postposition of the subordinate clause. In the function of Einräumung, the semantic differences derive from an epistemic, subjective inference of the speaker. In the corrective use of concessive conjunctions, the prototypical concessive subordinator develops into a discourse marker and thus becomes even more grammaticalised. In all four sample languages, the prototypical and at the same time most versatile connectors are the link from adverbial, subordinate concession to coordinate structures. This interception indicates a much larger network of concessive functions than only the conceptual space of adverbial concession. Interestingly, French behaves exceptionally. It’s most frequent and neutral adverbial concessive conjunction quoique is not used in any other functions except in the prototypical 75

expression of an insufficient anticause. However, encore que, which is not the fundamental conjunction but restricted to formal and written literary style, covers all the function that the most frequent and most versatile connectors cover in other languages.

5.2 Observations and deductions Considering the semantic maps of all four sample languages, it becomes obvious that restriction and correction are always encoded by the same subordinator like prototypical concession. However, they may be expressed by additional, independent markers as is the case in Latin or German. Here, licet and wenn auch code the function of Einräumung independently of prototypical concession. Besides the function of an Einräumung, wenn auch also refers to the concessive function of contrast in German; however, in this function it requests a correlative so in the main clause. The semantic correlation between Einräumung and contrast could not yet be fully established; indeed, it is unclear what kind of connection there is between the two functions. Again, a corpus analysis might offer helpful insights whether wenn auch is truly used in these functions only or whether new functions need to be added to the map in order to gain a complete one and to establish a semantic link between the two functions. Another explanation for the similar coding of these two functions may be the historical loss of intermediate functions and uses of the marker wenn auch. This startling distribution of the German conjunction will be examined in a later paragraph of this chapter. In Latin, the function of an Einräumung semantically relates to the original meaning of licet, which referred to the act of allowing or permitting. In fact, the act of allowing is not far away from the act of conceding: both acts involve a subjective commitment of the speakers to their utterance based on their individual inferences about the situation and their conversational goals and emphases. Based on the principle of contiguity, the language-specific semantic maps allow predictions about the functions of concessive conjunctions. If concessive connectives cover two functions of the conceptual space which are not directly adjacent, they also have to cover all functions in between. This principle becomes effective for the markers although, quamquam and encore que. Although it is not overtly mentioned in the descriptive grammars, it is acceptable to assume that these three makers are used in order to express the function of Einräumung as they express the two adjacent functions restriction and expressing an insuffi76

cient anticause as well as the function of correction further away on the map. Again, this thesis should be tested with corpus data of concessive adverbial clauses and with descriptive material on other languages in order to see if these markers can be used in this function and if there indeed is a semantic relationship between restriction, Einräumung and correction beyond the four sample languages. As semantic maps and their underlying conceptual spaces can be used to deduce universal linguistic tendencies, one may ask what synchronic and diachronic universal deductions can be made on the basis of the semantic map of concession. Of course, the sample of four languages is far too small to observe any valid, universally accurate implications or phenomena. Still, they might show some tendencies that are worthwhile to analyse and examine with a more elaborate language sample and/or corpus data. Firstly, all languages show a clear differentiation between real concession and hypothetical concession. These two major domains of adverbial concession are expressed by distinct subordinators and may language-specifically even occur in different moods. The subordinators and the different concessive functions of the domains show clear semantic differences: in hypothetical concession, the objections to the main clause are not factual but imagined and hypothetical. Subordinate clauses of hypothetical concession express content ranging from one objection which might even be extreme in its value to any impossible number of irrelevant objections; in any case, they are always hypothetical. Thus, all hypothetical concessive functions are characterised rather by irrelevance than by opposition and ineffectiveness, i.e. the number, degree or value, even the opposing character of two subordinate propositions is irrelevant for the realisation of the main clause. Secondly, in all languages analysed, scalar concessive conditionals function as the link between hypothetical concession and real concession. Interestingly, SCCs seem to be a direct link to the prototypical concessive function of expressing an ineffective anticause or insufficient objection. Except for French, all sample languages allow for scalar concessive conditional markers to be used in prototypical concession. Even in French, it would need an elaborate analysis of the actual use of SCC connectives to prove its exceptionality or its adherence to the tendency. This tendency goes along with the diachronic observation of many linguists that concessive adverbial subordinators often derive from concessive conditional markers (cf. König/Haspelmath 1997: 568). This development can be explained with the fact 77

that both concessive functions are semantically close to each other – the only major difference between scalar concessive condition and prototypical concession lies in the factuality of the content of the subordinate clause. A third tendency which can be observed in the sample languages is that the prototypical markers of adverbial concession are also the most frequent, most neutral and most versatile ones. They do not only code the fundamental function of an ineffective anticause, they also code a variety of other functions. Thus, they are truly polyfunctional within the category of concession. In German, the major concessive connector obwohl covers all functions of real concession; English although, Latin quamquam and French encore que cover at least four of the five functions. Obviously, the other functions of concession elaborate on the most fundamental, prototypical one. Thus, they may be instances of semantic widening. As a typical process of grammaticalisation, the use of these conjunctions spreads to more linguistic environments which results even in the use as discourse markers in coordinate structures such as in the function of correction. A fourth tendency becomes obvious when we look at example sentences of concession and most importantly at their translations in the book on contrast by Rudolph.8 These examples show that (even) though covers more functions than its functional descriptions in descriptive grammars would suggest (cf. Rudolph 1996: 418). However, it is necessary to note that the concessive sentence in this example is abbreviated, i.e. does not have a main verb. though is often used in abbreviated sentences in English. (31) a. Er war blind […] und konnte auch etwas weissagen, obwohl nur stockend und ungenau. b. He was blind […] and he also could foretell a little, though only hesitantly and inexactly.

According to these sentences and their translation, (even) though may be used for the encoding of restriction, Einräumung and contrast as well. Thus, it is almost as versatile as the prototypical concessive subordinator although.

8

As these translations seem to be made by Elizabeth Rudolph herself, they have to be considered carefully as the use of the conjunction may not be necessarily grammatical or without alternatives.

78

Einräumung UCC

ACC

SCC

ineffective anticause

restriction

correction

although

contrast (even) though Fig. 13 even though and although on the semantic map of English

The only function which even though does not cover is correction. Consequently, one can deduce from the distribution on the semantic map that even though is not yet as grammaticalised as although and may not be used as a discourse marker. In descriptive grammars of English, even though is described as an emphasised and rather colloquial expression of concession. Formally, the emphasis in its meaning becomes obvious in the focal particle even. The translations in Rudolph’s book show that even though - just as although in its etymological development - may indeed be on its way to the next fundamental, prototypical and versatile concessive subordinator next to although because it undergoes exactly the same paths of semantic widening (compare appendix). Thus, it is sensible to assume that even though may develop into a discourse markeras well, when it spreads to the function of correction. These example sentences show how the actual use of connectives and conjunctions is an important source for the analysis of their development and distributional behaviour. Thus, they support my claim that a corpus analysis is necessary to back up the geography of the conceptual space and the distribution of the language-specific semantic maps; if required, the data of the corpora may lead to a reorganisation of the networks. When one looks at the descriptions of the various concessive functions in the area of hypothetical concession, it becomes obvious that the notion of concessivity seems to increase. On the conceptual space, this is mirrored in the arrangement of the functions: from universal concessive condition onwards, which occurs at the left pole of the map, the basic concessive notion increases. For this observation, we define basic concession as a structure in which one of two clauses, the subordinate one expresses an objective, factual proposition that normally would prevent the main clause proposition to take place. However, in this particular situation, it does not have the effect of invalidating the main proposition. If we accept this definition of concession as the default definition, it leads to the prototypical concessive function of expressing an ineffective anticause. Consequently, UCCs express a rather aproto79

typical concessivity as they rely on an infinite number of hypothetical propositions which may or may not invalidate the main clause. The exact nature of the anticauses is left open, which increases their semantic irrelevance and leads to a rather subdued notion of concession. ACCS express two proposition of which one may be the cause and the other one the anticause for the main proposition (sometimes, alternative concessive conditionals express even two causes). The irrelevance derives from the impossibility of two opposite facts being causes for the single main clause proposition. So again, the semantics rely more on irrelevance than on real concessivity. In contrast, SCCs are tightly linked to prototypical concession. They express a single hypothetical anticause to the main clause proposition which does not suffice to invalidate this main clause. Thus, scalar concessive condition very closely resembles the functional definition of prototypical concession because it irrelevant whether an ineffective anticause is hypothetical or not: it is ineffective, nevertheless. The notion of concession is dominant and much more obvious than in the functions further left on the map. Interestingly, the hypothetical concessive structures that derive from (and resemble) other types of subordinate clauses, i.e. free relative clauses and embedded interrogatives, are the ones farthest away from prototypical concession and thus from the central point of the network as well as the centre of concessivity. Hence, one may argue that they form the interception to the networks of other subordinate structures. While the notion of prototypical concession increases from the left pole of the map to its centre, it decreases again from the centre towards the right pole. Consequently, the function of correction at the right pole of the conceptual space displays much less prototypical concessive semantics than the central function of an ineffective anticause. One reason is the decrease of characteristics belonging to adverbial subordination. Towards correction, the nature of subordination and thus a fundamental element of adverbial concession get lost. Secondly, subordinate clauses grounding in concessive functions farther to the right on the conceptual space do not express an objective, factual anticause which does not suffice to invalidate the main clause. Rather, they express increasing subjectivity of the subordinate proposition. The concessive clause functions as a subjective comment on the main clause, as restricting element or even as a polite way of expressing a correction of the main proposition. All functions to the right of the conceptual space have basically concessive semantics as the two events or facts of the complex sentence are incompatible but still combined. How-

80

ever, these functions take on additional semantic-pragmatic functions while the prototypical concessive notion diminishes. So although prototypicality of semantics is not fundamental for developing semantic maps and conceptual spaces, the geography of the conceptual space and the description of the various functions on the space can indicate the centrality of functions. It is quite logical that the most prototypical, basic and universal function of concession occurs at the very centre of the network and that less prototypical concessive functions spread from there. Consequently, the function of an ineffective anticause, which displays the strongest notion of concessivity, occurs at the centre of the map. Hypothetical concessive functions increase in their prototypical concessive semantics from left to right as they converge to the centre of the map. Functions of real concession to the right of prototypical concession decrease in their basic concessive notion as they diverge from the centre of the map.

5.3 Discontiguity on the German map A closer look at the distribution of the German concessive markers on the conceptual space reveals that the allocation of wenn auch violates the essential principle of contiguity.

Einräumung UCC

ACC

SCC

ineffective anticause

restriction

correction

wenn auch

contrast Fig. 14 Discontiguity on the German map

As has been explained in the first chapters on the geography and parameters of semantic maps, the marking of functions needs to be contiguous and cover directly adjacent, connected functions. The marking of wenn auch in German covers the function of Einräumung and antithetical concession, two functions that are not directly related and do not stand directly adjacent to each other. According to the principles and rules of drawing a semantic map, the principle of contiguity should not be neglected. If the phenomenon of unconnected marking occurs, the semantic map and the underlying conceptual space are usually devalidated and in need of an organisational change. This leads to the question if the whole con81

ceptual space is invalid due to this phenomenon of lacking contiguity in the German semantic map. Normally, the answer would be yes. However, in their essay on the semantic map of modality, van der Auwera and Plungian (van der Auwera/Plungian 1998) deal with this phenomenon of discontinuity with the help of the example of Dutch mogen. They suggest that discontinuity may be acceptable on semantic maps due to historical changes in a language. Accordingly, markers may lose functions over time, be replaced by other connectives or enter a language at different stages, thus referring to various, unconnected functions within one grammatical domain (cf. van der Auwera/Plungian 1998: 34ff.). As a consequence, I suggest the following different explanations which may account for the discontiguity in the semantic map without discarding the whole conceptual space. In a first explanation, wenn auch developed independently for both functions; thus, it could express them independently from each other. This possibility is not very likely as wenn auch shows the typical form of a concessive marker consisting of a conditional connective and a focal particle and is thus a common representative of concession. Another possibility that might be more likely grounds in exactly this typical concessive (conditional) form of wenn auch. As it formally very closely resembles the scalar concessive conditional marker auch wenn and as its form probably derives from a concessive conditional source, one may assume that wenn auch and auch wenn are more or less cognates. In fact, one finds the suggestion that wenn auch and auch wenn are simply two different realisations of one and the same marker (cf. Rudolph 1996: 423). On the one hand, the examples using wenn auch in an antithetical function could employ a variation of auch wenn, which then would spread from scalar concessive condition via ineffective anticause to antithetical concession. That this explanation should be considered becomes obvious when we take a look at the following examples: b. Wenn das Buch auch gut geschrieben ist, so ist es doch nicht sehr wissenschaftlich. c. Auch wenn das Buch gut geschrieben ist, so ist es doch nicht sehr wissenschaftlich.

To my understanding, these two sentences are semantically identical. Thus, it would be sensible to spread the distribution of auch wenn to contrast. With the help of more elaborate linguistic data, it might even be possible to discern if auch wenn may also cover the function of Einräumung; then, auch wenn and wenn auch would clearly be variations of one and the same concessive connective. 82

On the other hand, wenn auch could be considered an independent marker which, based on its formal make-up, presumably originated from the use for expressing scalar concessive condition. It then spread to prototypical concession and from there to Einräumung and antithetical concession. However, as time went by, it lost the function of at least prototypical concession, which would explain the discontinuous distribution of wenn auch on the semantic map of German. These considerations show that there are ways and solutions to explain the phenomenon of discontinuity at the German map without invalidating the whole conceptual space. Especially historical developments may account for the synchronic lack of continuity. Again, a corpus analyses and the investigation of more languages would allow deeper and statistically verified insights that could facilitate the choice of one of the given options or even offer a completely new explanation that allows maintaining the conceptual space.

6

Diachrony of adverbial concessive conjunctions In a study of conceptual space and semantic maps of adverbial clauses, the question

comes up why is it not only interesting but even necessary to look at diachronic aspects of linguistic structures. According to Haspelmath, there are two main reasons for the importance of diachronic analyses in the semantic map approach to linguistic phenomena. Firstly, language change is a linguistic universal: all languages change constantly. If linguistic analyses have a closer look at these changes, one may hope to find out general claims about language and its workings in general. Secondly, a diachronic look at various languages, i.e. the field of diachronic typology, may help to understand the synchronic state of languages. Diachronic developments may account for synchronic variation in language, a phenomenon that always exists universally across the languages of the world. Thus, historical analyses may help to explain “recent innovations and [...] remnants of earlier regularities that are no longer synchronically motivated [...] (Haspelmath 1997: 129)” and thus, lead to irregularities and variation in the languages of today. But how may typologically oriented semantic maps be useful for diachronic examinations? Haspelmath notes that “implicational maps that account for the cross-linguistic distribution of different functions of grammatical categories also constrain the possible diachronic changes: a category can acquire a new function only if that function is adjacent on the se83

mantic map to some function that the category already covers (Haspelmath 199/: 129, emphasis added).” Consequently, in order to read semantic maps diachronically, one has to bear in mind that “grammatical categories gradually extend their uses along the paths allowed by the map (ibid.).” With regard to the semantic map approach to adverbial concessive conjunctions, grammaticalisation processes explain the sources and the language-specific development of grammatical concessive markers. Thus, semantic maps may give parochial information about languages. Additionally, however, they may also indicate cross-linguistic universal properties of the grammatical category in question, i.e. in this case concessive markers. Semantic maps may reveal answers on questions such as ‘Are there common sources or origins for concessive connectives?’ or ‘Are these origins universal?’ Fundamentally, these questions indicate a more basic one, namely the question if semantic maps have the predictive force to visualise the direction and results of grammaticalisation processes?

6.1 The role of grammaticalisation Before I will have a closer look at the particular diachrony and grammaticalisation of concessive adverbial conjunctions, I will deal with grammaticalisation in general. More specifically, this preliminary part will deal with the following questions: what is grammaticalisation? Which processes does it involve? Are there particular parameters according to which one may examine grammaticalisation processes of linguistic phenomena? According to these pre-established parameters, I will analyse the process of grammaticalisation of concessive conjunctions and its extent. During this analysis, it will show if and in what way the semantic maps will be a helpful tool.

6.1.1

What is grammaticalisation?

According to Hopper and Traugott’s work on grammaticalisation, the term may refer to linguistic phenomena and at the same time to the theories and frameworks to explain these linguistic phenomena (cf. Hopper/Traugott 2003: 1). The basis for the theory of grammaticalisation and considerations of grammaticalisation processes is the distinction between a synchronic and a diachronic perspective on language. While the synchronic dimension of lan84

guages refers to the status quo, to language as it is now, the diachronic dimension refers to various stages, changes and processes through which language developed into its current state. In other words, diachrony is “[...] understood as the set of changes linking a synchronic state of a language to successive stats of the same language (Hopper/Traugott 2003: 2).” Thus, grammaticalisation is an approach to “investigat[e] the sources of grammatical forms and the typical steps of change they undergo (ibid).” These considerations lead to a fundamental definition of grammaticalisation as “[...] that subset of linguistic changes whereby a lexical item or construction in certain uses takes on grammatical characteristics, or through which a grammatical item becomes more grammatical (Hopper/Traugott 2003: 2).” These changes do not occur abruptly and in leaps but according to clines, which means they “go through a series of small transitions (Hopper/Traugott 2003:6).” Historically, clines are pathways along which lexical or already grammatical forms undergo further changes. Thus, clines can also be used to predict the outcome of grammaticalisation processes. Synchronically, they present a continuum between two poles of lexicality and grammaticality (cf. ibid). Interestingly, clines and their directions seem to be universal, i.e. changes develop along the same or very similar and predictable steps throughout languages. One universal process is described by the cline of grammaticality, which develops along the path of content item > grammatical word > clitic > inflectional affix (cf. Hopper/Traugott 2003: 7). As grammaticalisation processes have universal aspects and are predictable across languages because of the existence of clines, semantic maps with their universally valid, underlying conceptual space can be a useful tool in order to visualise these paths or in order to deduce them from the geography of the networks. Clines are generally and usually unidirectional. This behaviour suggests that language change is directional and may be predicted. In the theory of grammaticalisation, this phenomenon is called the principle of unidirectionality. According to this principle grammaticalisation processes usually go from less grammatical to more grammatical (cf. Hopper/Traugott 2003: 16). It is far more likely and a natural development for lexical items to be changed from lexical to grammatical; only seldom, one can find instances of degrammaticalisation, i.e. of developments working the other way around. Unidirectionality should make grammaticalisation processes even more a thing to be predictive and illustratable with the tool of semantic maps. As the unidirectional changes do not suddenly jump from one developmental stage to the next but rather shift from one meaning to another related one, this shift of 85

semantic widening can be represented on a conceptual space or language-specific semantic map. In fact, the unidirectional quality of grammaticalisation is inextricably bound to the contiguity principle of semantic maps which allows not only the representation of the developments and their direction but also represents the various meanings as synchronically and diachronically related. So the assumption that changes progress unidirectionally makes semantic maps a useful tool for diachronic linguistic analyses. So far, the definition of grammaticalisation and its processes has covered mainly the diachronic mechanisms and implications of language change. However, grammaticalisation is not only a description of diachronic phenomena along the basic lines of lexical items developing into grammatical items or grammatical items becoming more grammatical. Grammaticalisation can be seen in very general terms as “[...] the linguistic process, both through time and synchronically, of organization of categories and of coding (Traugott/Heine 1991: 1).” Thus, this linguistic phenomenon asks not only how grammatical coding and grammatical forms came to be as they are, but also “what grammatical coding is possible typologically in a language or in languages and how it is organized either at one period in time [...] or panchronically [...] (Traugott/Heine 1991: 3).” It is important to bear in mind that grammaticalisation is just one process or mechanism of language change. Equally pervasive mechanisms are analogy, metaphor, metonymy and reanalysis. Thus, not all change is grammaticalisation but grammaticalisation is always language change. The principle of unidirectionality is equally complex and does not only refer to irreversible changes from lexicality to grammaticality. The changes of grammaticalisation affect various levels of language as they change it partly from more lexical to more grammatical and from more semantic to less semantic. Thus, grammaticalisation is usually seen as tied to semantic bleaching, loss of semantic and pragmatic meaning and desemanticization (cf. Traugott/Heine 1991: 4). However, this observation cannot be accepted uncritically. Indeed, the very phenomenon of concession challenges this thesis. Traugott and König note that early stages of grammaticalisation rather include semantic widening and enrichment than bleaching. Due to semantic widening, pragmatic enrichment is almost a necessary consequence. In the field of concession, Traugott and König refer to concessive while as an example. Originally, while was a temporal conjunction. When it was used in extended circumstances, its pragmatic meaning increased. Only then, the semantic bleaching started and it became versatile in very different semantic surroundings (cf. Traugott/Heine 1991: 5). This example fur86

ther indicates how important the aspect of frequency is: it functions as catalyst of grammaticalisation.

6.1.2

Principles of grammaticalisation

According to Traugott and Heine, grammaticalisation involves five different processes which may describe the consequent language change: layering, divergence, specialisation, persistence and de-categorialisation. Layering refers to the phenomenon that very often more than one technique is available in a language to serve similar or even identical functions. This formal diversity comes about because when a form or set of forms emerges in a functional domain, it does not immediately (and may never) replace an already existing set of functionally equivalent forms, but rather the two sets of forms co-exist (Traugott/Heine 1991: 23). In concession, this principle of grammaticalisation can be observed when a marker of scalar concessive conditionals spreads into the functional area of prototypical concession, as it is the case with even if in English, auch wenn in German and etsi in Latin. These three forms coexist with the already established forms but are not as frequent as those in this prototypical concessive function. Similarly, layering can be observed when concessive conjunctions spread into coordinate structures in the function of correction and change into discourse markers. although and obwohl then co-exist next to discourse markers with the meaning of however (cf. Günthner 2000). The second principle of grammaticalisation, divergence, is superficially very similar to layering. When a form is grammaticalised, its original form and function may remain. Basically, divergence thus describes a split as the resulting forms are etymologically related but have different functions (cf. Traugott/Heine 1991: 24). Consequently, “one and the same autonomous lexical item becomes grammaticized in one context and does not become grammaticized in another (ibid.).” Again, concessive while functions as an illustrative example as it exists both in a temporal and a concessive meaning, i.e. in two adverbial functions. From the grammatical marker of time, it was further grammaticalised to the function of yet another grammatical marker, the relation of concession. Besides the function of adverbial conjunctions, while also exists as a noun, which in turn served as the origin for the temporal adverbial marker and is the lexical source for the grammaticalised items. Thus, the development of while illustrates the various steps and levels of grammaticalisation because it devel87

oped from a lexical item, a noun, to the semantically related grammatical item, an adverbial marker of time, to an even more grammatical item, the marker of concession. Another example of divergence in the area of concession are the markers of universal concessive conditions, relative and interrogative pronouns. In specialisation, a form emerges whose use in a particular construction becomes obligatory. In other words, the choice of other forms and fundamentally the choice of using this particular form or not becomes narrowed – the construction in question specialises and consolidates in its form (cf. Traugott/Heine 1991: 25f.). In concessive clauses, this becomes obvious in the choice between if and even if. Although simple if can be used in scalar concessive conditionals and even in purely concessive clauses, the more complex form of conditional subordinator and focal particle is much preferred. Similarly, specialisation is and was at work in the choice between though, although and even though. In earlier stages of English, though was the default concessive conjunction and although was a marked form. However, the choice between the two conjunctions narrowed down and although was declared the typical concessive conjunction. Similarly, specialisation may decide in favour of even though and against the use of though. While even though is a comparatively versatile connective whose functional distinctions spread across almost the entire area of real concession on the semantic map (compare Fig. 13), the use of though besomes more and more restricted to particular syntactic structures in sentence-final position and to colloquial language as concessive conjunction. “The [p]rinciple of [p]ersistence relates the meaning and function of a grammatical form to its history as a lexical morpheme (Traugott/Heine 1991: 28).” Here, once more while is an excellent example as the temporal meaning of while is still somewhat apparent in the concessive function and the noun referring to a period of time is clearly apparent in the function of the temporal adverbial conjunction. licet, a Latin conjunction, may be an even better example. The original meaning of allowance still forms an important part of the concessive meaning of the conjunction (compare appendix, the data of Latin). The last of the five principles of grammaticalisation is de-categorialisation. In this process, grammaticalised forms lose characteristics which are singular to their original category. This loss is possible because the category and the affiliation of a form to the category derive from and depend on the use of the form in discourse (cf. Traugott/Heine 1991: 30). So 88

in the grammaticalisation process, forms assume discourse functions and change from propositional to textual use, i.e. to a “meaning or function that is relative to the text or some local construction (ibid.: 31).” For this process, we can again have a closer look at the forms of licet and while. licet once was a verb with the meaning of allowance or permission. In a reanalysis, it became considered a conjunction and grammaticalised from the lexical category of a verb to the grammatical category of an adverbial conjunction. So, licet can no longer be conjugated as verbs can be; thus, it lost the characteristics of its original category. Similarly, while does no longer have all the characteristics of a noun in its function as adverbial subordinator; for instance, it cannot take adjectival attributes. These examples show how concessive adverbial conjunctions were and still are affected by processes of grammaticalisation. At least some conjunctions can be analysed and described according to and with the help of the five principles of grammaticalisation. Thus, one can say that grammaticalisation brought them into being in their function as concessive conjunctions and reveals additional characteristics of the markers such as the polyfunctionality of while and the pragmatic enrichment of even if. Grammaticalisation must have been a historical and developmental process in the diachrony of concessive conjunctions as they are grammatical elements. According to the fundamental definition of grammaticalisation, grammatical markers derive from lexical items, already grammatical items (and thus, eventually from lexical items again) or from demonstratives. As a consequence, concessive conjunctions must have lexical origins or at least less grammatical sources. Interestingly, concession is a derived notion both formally and semantically (cf. Kortmann 1997). According to Kortmann, concessive conjunctions never employ concessive meaning as a primary function and other meanings in secondary position. Usually, the functional attribution is the other way around (cf. Kortmann 1997: 203). Concessive connectives all derive from already existing adverbial notions or pre-existing grammatical or even lexical sources. Also semantically, the notion of concession grounds in other relationships such as cause and condition (compare chapters 3.3.2 and 3.4). This derived and complex nature of the concessive relation is also apparent in the form of its markers. Accordingly, Traugott and Hopper note that “[c]oncessive meanings develop late in the history of specific clause linkage markers partly because the concessive is more abstract, partly because it is more complex logically (Traugott/Hopper 2003: 187).” Thus, the very nature of concessive markers points out the importance to have a closer look at the sources and origins of concessive conjunctions and to in89

vestigate the processes, mechanisms and paths of their development. Eventually, these considerations will show if some paths and directions can be visualised with the help of semantic maps.

6.2 Sources and origins of concessive conjunctions Maybe the most frequent grammaticalisation cline for concessivity is the path of condition > concessive condition > concession (cf. Kortmann 1997: 203). This development is supported by König who notes that scalar concessive conditionals are one of the most pervasive sources for proper concessive markers. However, this developmental path is by far not the only mechanism that leads to concessive conjunctions, just as concessive conditionals are not the only source for markers of real concession. The following chapters will thus have a closer look at mechanisms and sources of the development of both concessive conditional and real concessive conjunctions.

6.2.1

Diachrony of concessive conditionals

In their analysis of concessive conditionals, Haspelmath and König deal with the diachrony of these linguistic phenomena throughout European languages. They note that concessive conditionals are closely related to conditional sentences; for example scalar concessive conditionals are formally derived from conditionals when a focal particle is added. However, pure conditionals cannot be derived by changing the form of concessive conditionals. Similarly, concessive conditional markers may develop into pure concessive markers (compare for example even if), but concessive markers never develop into concessive conditionals. Thus, condition and concession are related along a unidirectional line of development. Besides condition, concessive conditional clauses may also be traced back to original constructions such as embedded interrogative clauses, relative clauses and exclamative clauses (cf. Haspelmath/König 1998: 620). As has already been pointed out above and in chapter 3.3.2, scalar concessive conditionals have their origin in conditionals. One might argue that they also show similarities to polar questions. Alternative concessive conditionals on the other hand clearly go back to alternative questions, which becomes most obvious in their marking that generally is iden90

tical to that of alternative interrogatives. These interrogative source constructions for two types of concessive conditionals suggest, according to Haspelmath and König, that concessive conditionals derive from questions and thus, form the developmental cline interrogative > concessive conditional. However, embedded interrogatives and concessive conditionals are not completely identical. Additionally, concessive conditionals show similarities to other linguistic structures as well. Only alternative concessives marked by structures similar to whether…or can be directly related to the cline above (cf. Haspelmath/König 1998: 621f.). Especially scalar concessive conditionals are formally more closely related to conditionals than to interrogatives. Haspelmath and König note that many languages employ interrogative markers in conditionals and from there, they spread to a concessive use. This observation leads to the second developmental path of concessive condition, i.e. interrogative > conditional > concessive conditional (cf. ibid.: 622). However, this cline does not explain the emergence of conditional and related concessive conditional markers which are clearly distinct from interrogatives as for example German ob (vs. wenn) or English if (vs. whether). Here, different mechanisms and paths must be accountable for the development of these conjunctions, which will be examined at a later point. Free relative pronouns and free relative structures seem to be a fruitful source for universal concessive clauses throughout (European) languages. Consequently, Haspelmath and König assume the clines interrogative > relative > concessive conditional and interestingly, concessive conditional > relative (cf. Haspelmath/König 1998: 623). This second path can easily be imagined as it does not require wide-ranging changes in the syntax and structure of the complex sentences affected. In German, for instance, only the subject or the respective head noun of the main clause needs to be removed. d. Wer auch immer kommt, er wird gut aufgenommen werden. (UCC) e. Wer auch immer kommt, wird gut aufgenommen werden. (free relative) (Haspelmath/König 1998: 623)

It is easy to turn the concessive clause to an argument of the main clause which is the only difference between a universal concessive conditional and a free relative clause in German. The investigation of Haspelmath and König shows that in a larger linguistic network, concessive clauses in general may be seen as connected with interrogative and relative clauses on a larger scale of sentences types and types of subordinate clauses. Not only are 91

relative clauses, complement clauses via embedded interrogatives and adverbial clauses formally related in the interception of conditional and concessive conditional clauses, even their semantic relationship is still traceable due to this diachronic linkage.

6.2.2

Diachrony of concessive conjunctions

The morphology of concessive connectives shows their derived nature as the various parts of the often complex conjunctions quite clearly suggest the etymological sources of the concessive markers. In many of his analyses of the concessive relation, König determines a fixed number of sources of concessive connectives: conditional markers, expressions of universal quantification, expressions of remarkable co-existence and co-occurrence and expressions referring to the human being. In his investigation of 1988, he adds expressions meaning true, fact, well, indeed. Interestingly, concessive connectives of any kind developed comparatively late in the history of languages and also occur quite late in language acquisition processes, probably due to their derived nature and their complex form and complicated semantics (cf. König 1985: 2). The first source investigated for real concessive conjunctions are concessive conditional markers. According to many linguists, foremost Ekkehard König, this grammatical development is one of the most frequent sources for the emergence of concessive conjunctions. An example is the Old English þeah, the etymological source for the modern concessive though and although. Originally, þeah was a correlative occurring in the main and dependent clause. In this use, it was an expression of concessive conditional semantics (cf. König 1985: 7). Initially, it had a meaning similar to that of even if, which means that it was used as a connective of scalar concessive condition. Interestingly, scalar concessive condition is the concessive function closest to prototypical concession on the semantic map. Thus, þeah is diachronic evidence for concessive conditional markers of scalar expressions to develop into the markers of prototypical concession, i.e. ineffective anticause. The semantic maps of present day English, German and even classical Latin show similar tendencies. even if, auch wenn, etsi and quamvis can be used in proper concessive surroundings although they are expressions of scalar concessive condition.

92

Einräumung UCC

ACC

SCC

ineffective anticause

even if

restriction

correction

auch wenn

contrast

quamvis, etsi Fig. 15 The semantic map of SCCs/English, German, Latin

These four conjunctions all derive from pure conditionals and thus, suggest the path conditional > (scalar) concessive conditional > concessive. However, as has been indicated above, conditional conjunctions are not the only source for the development of concessive markers. A common source are so-called universal quantifiers, which can still be recognized in many concessive connectives. Examples are expressions referring to all as in although or tout as in tout… que, but also elements such as auch immer, ever or -cumque which together with relative or interrogative pronouns form free relatives and thus, the connectives for UCCs (cf. König 1985: 10). Other expressions belonging to this group of sources are expressions of volition as for example in quamvis (cf. König 2000: 153). In his etymological analysis of 1985, König attributes expressions such as bien que to these origins as well. In 2000, he established another source for concessive connectives, namely expressions with the meaning true, well and indeed (cf. König 2000: 154-155). The examples offered by him do not suggest any adverbial conjunctions deriving from this source; however, bien que and probably the Latin ut… ita should be attributed to these origins. Another frequent source for concessive expressions are conditional, originally conditional or temporal connectives which are combined with a focal particle comparable to the meaning of even. The frequency of this source becomes obvious in the number of connectives which can be traced back to these origins. Generally, connectives containing elements such as wenn, ob, though and si go back to conditional conjunctions. Temporal examples are quand, cum and while. Accordingly, English even if and even though, German wenn auch and auch wenn, Latin etsi, etiamsi, tametsi and French même si etymologically go back to conditional connectives with a focal particle. Similarly, the German default concessive conjunction obwohl and its less frequent variants obgleich, obschon, obzwar go back to these origins (cf. König 1985: 10-11; König 2000: 153-154).

93

The third source for real concessive connectives grounds in expressions implying remarkable co-occurrence or co-existence of two events or facts. Eventually, these connectives are closely related to expressions of simultaneity. In his papers from 1985 and 2000, König does not name any adverbial conjunctions that derive from this source. However, based on the analysis of Traugott and König (1991), while is a representative of this diachrony. One may even argue that whereas with its locative semantic origins refers to the co-existence or co-occurrence of two facts or events and can thus be considered as deriving from this source as well. Other possible examples are cum in Latin and encore que in French. In his analysis of 2000, König emphasises that these expression also cover cases in which one event does not prevent or hinder another event. Consequently, the connectives centre on negative expressions. However, these instances have not brought forth adverbial concessive conjunctions (so far) (cf. König 2000: 155). The last group of concessive connectives derives from expressions of spite, obstinacy and contempt, i.e. of expressions referring to human beings and human agents. Quite a number of sentential adverbs and coordinating structures stem from these expressions, but adverbial connectives only rarely go back to these sources. French malgré que and German trotzdem are two adverbial subordinating conjunctions deriving from this rather lexical source (cf. König 1985: 12; König 2000: 152).

6.3 Developmental paths to concessivity “Concessive connectives and constructions seem to be quite generally based on expressions or constructions having another, more basic meaning and use (König 2000: 156). Consequently, in his analysis König notes if not universal, then at least very widespread tendencies concerning the semantic changes that lead to concessive expressions. He claims that “some very general processes of semantic change underlie the development of concessive connectives in a wide variety of languages (König 1985: 12).” The development along the cline of conditionals > concessive conditionals > concession takes place according to two mechanisms: strengthening and reinterpretation. Emphatic particles such as all give a factual character to previously conditional, i.e. un-entailed clauses. Thus, although expresses a factual objection while though was originally used in conditionals and later in concessive conditional surroundings (cf. König 1985: 13). even though 94

shows a somewhat different development than although which can be compared to that of even if and other recent markers of scalar concessive condition. The addition of even or other focal particles suggest that there are other conditions beside the given one for which the main clause is true. However, this given condition is most unlikely to occur in combination with the main proposition; thus, main clause and subordinate clause are incompatible (cf. König 1985:13ff.). Here, the path of semantic change can be described rather as reinterpretation than as strengthening. Formerly concessive conditional markers come to be used in factual surroundings, which would ask for proper concessive conjunctions. The factuality then gets reinterpreted as characteristic of the connectives. obwohl for example is a prototypical concessive marker nowadays, but goes back to a conditional connective with a focal particle. Similarly, etsi and quamvis are used in factual surroundings by Julius Caesar. Today, even if and auch wenn undergo the same semantic widening when they stretch into the real concessive function of expressing an ineffective anticause and thereby develop from scalar concessive condition to factual concession. One may even argue that the loss of the subjunctive in French indicates similar semantic changes. Concessive connectors based on lexical expressions of spite, contempt and obstinacy undergo a semantic change from being concrete to more abstract, a process called semantic bleaching. Semantic bleaching is a fundamental and quite common process of grammaticalisation as the development of grammatical items usually starts with lexical expressions which “need to lose” their concrete, referential meaning and assume a more abstract, textual meaning. In order to produce concessive connectives based on remarkable co-occurrence and coexistence, languages need to undergo rather complex and interesting processes. König and Traugott investigated these origins of concessive conjunctions and found a third strategy of semantic change. Due to the complexity of our world, co-occurrence, co-existence and overlap of facts and events are all-pervasive in the world and consequently nothing relevant or specifically exiting. Thus, this information is usually rather irrelevant. However, there are conditions when these situations are interesting or relevant, most importantly when the two co-occurring facts or events are incompatible (cf. Traugott/König 1991: 200). As a consequence, the concessive notion of these expressions developed due to “some interpretative augmentation of their literal meaning (König 1985: 15).” Since an overlap, co-occurrence or co-existence of two events is rather uninformative, irrelevant and ubiquitous, the hearer 95

looks for additional information and further possibilities of interpretation as utterances, according to Grice’s maxims, usually are not irrelevant and uninformative. In other words, the hearer assumes there must be more to the sentence. These considerations lead to the assumption about an incompatibility between the two co-occurring facts because then, the utterance would be highly relevant and informative; indeed, it would even be surprising (cf. König 1985: 16). In their 1991 analysis, Traugott and König call this process “the strengthening of informativeness as a conversational implicature becomes conventionalised (Traugott/König 1991: 190).”

6.3.1

Semantic loss vs. semantic enrichment

With their thesis that the development of grammatical connectives such as conjunctions of cause and concession involves predominantly “the strengthening of informativeness” (1991: 190), Traugott and König express an opposite opinion to standard analyses on the development of semantics in grammaticalisation. Usually, linguists assume that grammaticalisation involves mostly semantic loss and semantic bleaching. Traugott and König, on the other hand, find bleaching only in late stages of grammaticalisation. The development of concessive while is a prime example to investigate the mechanisms of grammaticalisation. Traugott and König observe that the emergence of conjunctions of concession grounds in a “strengthening of the expression of speaker involvement (Traugott/König 1991: 191).” So when temporal while (in turn originating from the lexical noun hwile denoting a time span) developed into concessive while, it suddenly no longer referred to reality and the situation described in the complex sentence but – due to the semantics of concession – to a world construed on the basis of the speaker’s sense of coherence, logic, and correlation – also with respect to the underlying presupposition of the sentence. Thus, the speaker’s beliefs are strengthened on the pragmatic level (cf. ibid.). Traugott and König differentiate between ‘what is said’ and ‘what is meant’. The shift in grammaticalisation always occurs from what is said to what is meant; this tendency is more or less unidirectional. For while this meant that the change took place from temporal to concessive meaning and not the other way around. Similarly, concessive conjunctions are always derived from connectives which have primarily other meanings and functions (cf. Traugott/König 1991: 193; Kortmann 1997). In its further development, the temporal meaning of 96

simple simultaneity, overlap or co-occurrence of while turned into an unmarked, conventionalised meaning and made it possible for other inferences to emerge: in English, this led to concessive inferences, which by no means is a universal development as the German counterpart developed into the causal weil. The exact semantic change probably originated in the notion of surprise based on the overlap and relation of the two events. This reaction led to the inference of a concessive meaning because surprise presupposes a certain contrast or concessivity relation between two situations and indicates the unusualness of their simultaneity. In this process, the subjectivity of the utterance increases as the speakers’ evaluations of the situation and their subjective constructions of the world get expressed because they offer their views about the coherence and correlations of events in the world (cf. Traugott/König 1991: 201-203). A similar development was observable in the semantic change of cum narrativum in Latin. Concessive and adversative cum go back to cum narrativum which was used in order to state the circumstances of the main action and thus, could implicate co-occurrence. Similarly to while, the surprising incompatibility of the events could have easily led to a semantic widening from cum narrativum to cum concessivum. In these instances, the mechanism of grammaticalisation is the conventionalisation of conversational inferences, which are made about the relation between facts and events. Other important mechanisms of language change are metaphor and metonymy. They are at work not only in lexical change but also in grammaticalisation. Spatial expressions, for example, are often used as temporal ones, illustrating the metaphor of TIME IS SPACE. As a result of her analyses on grammaticalisation and lexical change, Traugott formulated three tendencies of grammaticalisation and the changes on the semantic-pragmatic level which take into account various mechanisms and strategies of language change.

6.3.2

Three tendencies of grammaticalisation

In her considerations about language change, Traugott developed three tendencies of grammaticalisation. Of these, only one is applicable and of relevance for the diachronic development of concessive conjunctions. Nevertheless, for a thorough depiction of Traugott’s theory, I will name all three tendencies and shortly explain them.

97

According to the first tendency, the grammaticalisation cline goes from “[m]eanings based in the

external

described

situation

>

meanings

based

in

the

internal

(evalua-

tive/perceptual/cognitive) situation (Traugott/König 1991: 208; Traugott 1989: 34). The second tendency formulates the development from “[m]eanings based in the described external or internal situation > meanings based in the textual situation (ibid.: 208; Traugott 1989: 35).” Here, already more abstract or grammaticalised items may become further grammaticalised and develop textual meaning, i.e. acquire a cohesive meaning and function. The last and most important tendency for this paper states that “[m]eanings tend to become increasingly situated in the speaker’s subjective belief-state/attitude towards the situation (Ibid.: 209; ibid.).” According to Traugott, this is where concessives are affected by grammaticalisation as they are “essentially expressions of speaker attitude to the relationship of elements within the proposition or of propositions to each other, as well as of the compatibility of those relations (Traugott/König 1991: 209).” For the example while, this becomes obvious as it expresses the surprise of the speaker that two propositions are related although they are incompatible. While it seems that metaphor is only of minor importance for these processes, Traugott and König argue that metonymy is much more relevant. For concession, this can again be made obvious with the help of the connective while. while expresses an extended notion of simultaneity, namely of surprising simultaneity and thus of concession. These semantics are still connected to the original meaning of the connective and the original intention of the speaker. This retained connection is the point of contiguity, the basic characteristic of a metonymy. Traugott formulates that “an originally conversational implicature that a marker of simultaneity would not be used unless there was something remarkable about that simultaneity came to index the surprise factor and thus the concessive (ibid.: 211).”

6.4 Diachronic concession on the semantic maps After the theoretical approach to the etymology and diachrony of concessive conditional and real concessive conjunctions, I will now turn to the question if any of these grammaticalisation processes can be illustrated and verified with the help of the language-specific semantic maps. Indeed, the semantic maps of all four languages illustrate the developmental cline of concessive conditional > prototypical concession. English allows even if in real con98

cessive surroundings especially in its spoken form. German acknowledges auch wenn in concessive sentences as correct, which shows how deeply entrenched the functional spread is. In Latin, etsi and quamvis are both considered correct in real concessive surroundings (compare Fig. 15). Only French seemingly does not permit this development; however, one may argue that the loss of the subjunctive in concessive clauses is comparable to the spread of concessive conditional markers into factual concessive functions. Here, linguistic data as for example a corpus analysis may show if and how this development takes place in French.

Einräumung UCC

ACC

SCC

restriction

correction

ineffective anticause contrast

Fig. 16 The development of hypothetical concessive markers to real concessive markers

The geography of the conceptual space of concession indicates that scalar concessive conditionals are not only semantically closest to real concession. This distribution is also supported by developments in form and use of scalar concessive conjunctions. Both in the maps analysed and cross-linguistically on a larger scale, the spread from concessive condition to prototypical concession can be verified. Historically, examples such as though strengthen this observation as its etymological source þeah was originally used in concessive conditional surroundings. As the semantic maps of the sample languages indicate, there is no linguistic evidence that universal concessive conditionals or alternative concessive conditionals develop into real concessives. Their semantics is probably too different from prototypical concession; scalar concessive conditionals lend themselves for the functional spread to real concession because they are semantically very similar to prototypical concession. In addition, there is no formal or semantic indication that the concessive conditional markers derive from each other. They undeniably show semantic similarities in their hypotheticality, in the semantic predominance of irrelevance and a varying number of irrelevant but opposing subordinate antecedents, which in turn leads to an increasing irrelevance. However, universal concessive conditionals and alternative concessive conditionals are not as conditional as for example scalar concessive conditionals. UCCs and ACCs derive either from free relatives or embedded 99

alternative interrogatives. Thus, they may function as linking elements to other forms of subordinate clauses. As UCCs and ACCs are representatives of adverbial clauses but are etymologically related to either relative clauses or a special type of complement clauses, they connect the network of adverbial clauses to larger interwoven structures of linguistics. In the area of real concession, the semantic maps can function as more useful tools of illustrating grammaticalisation and developmental processes. They prove especially useful for visualising the functional development from prototypical concession to the function of correction, i.e. from the default adverbial function of concession to a rather coordinate concessive structure. All languages show a functional widening from a previously solely prototypical concessivity marker to other functions. though, originally a concessive conditional, developed into the default concessive conjunction but was then substituted by although, which emerged via the process of strengthening. This connective spread at first into closely related functions such as Einräumung and restriction, probably due to a conventionalised conversational implicature and the widening of informativeness. Because of these processes, the use of although eventually extended to other, less closely related concessive functions. Consequently, the semantics and pragmatics of the conjunction developed to an increased informativeness as the connective took over other functions and could be used in more, related contexts. A similar development occurred in the other languages as the cases of obwohl, quamquam and encore que show. Interestingly, all the default conjunctions of the various sample languages have different origins. Additionally, encore que is no longer the French default connective but rather specialised in the coordinate functions of concession. Latin grammars directly refer to a function of quamquam in which it is used more like a discourse marker as quamquam correctivum.

Einräumung UCC

ACC

SCC

restriction

correction

ineffective anticause

quamquam, encore que (though) although/ contrast (ob) obwohl Fig. 17 The diachronic development from ineffective anticause to correction

In descriptive grammars and linguistic essays, the function of antithetical concession is not as elaborately investigated as the other functions of concessive conjunctions. Some 100

grammars do not even recognise the function as a form of concession. However, as Rudolph and many others acknowledge, contrast and concession are closely related. Especially Rudolph repeatedly remarks that the adversative and the concessive relation are the two types of contrast, the former coordinated, the latter subordinated. As a consequence, one may assume that primarily contrastive conjunctions such as while, whereas or ut…ita spread towards the concessive relation. This would be a comparatively small developmental step as concession relies on a contrastive incompatibility and opposition between the two propositions of the complex sentence. However, there are also concessive conjunctions which develop towards the function of antithetical concession. A possible example is even though which Rudolph (1996) uses to translate German concessive clauses of antithetical concession. Here, more linguistic data would be needed in order to investigate if even though really spreads to expressing contrastive concession. Similarly, wenn auch, which shows the typical structure of a conditional marker and a focal particle, is restricted to the function of Einräumung according to Grammatik der deutschen Sprache; however, it also occurs in contrast. The semantic and historical connection between the two is not yet obvious as both functions are not contiguous on the map (compare Fig. 12). The German default concessive connective obwohl spread from prototypical concession to the function of antithetical concession, too. Consequently, one can assume that the prototypical function of concession, the expression of an ineffective anticause, allows access to many other related functions; thus, the understanding of these newly acquired functions is secured based on the understanding of the semantics of the deeply entrenched original function which are extended and elaborated in the spreading. Latin concessive cum developed from the narrative function where it referred to the temporal circumstances of the main action. When those circumstances expressed incompatibility or contrast, narrative cum developed to adversative cum. The close relationship between adversation and concession, i.e. between the two derived uses, naturally leads to the use of cum also in a concessive meaning. According to these investigations, what can be generalised by looking at the historical information that semantic maps illustrate? A frequent and almost universal tendency in the diachrony of concession is the one-way-cline from prototypical concession to correction. Here, concession defies the general tendency of a development from coordination to subordination because the general unidirectionality of this cline is not maintained in the adverbial concessive relation. However, the specific development of concession from subordination to 101

coordination is unidirectional as there is no attested spread from the function of correction to the function of insufficient objection. Possible paths are (1) prototypical concession > restrictive concession > corrective concession (2) contrast concession For concession and contrast, the developmental path probably goes in both directions. The two relations are extremely close so that connectives of contrast may easily assume a concessive notion when the contrast surprisingly shows no effect on the main clause. The other way around, concessive connectives may rely on contrast so heavily that they themselves assume a secondary, purely contrastive function. The development of scalar concessive condition to real concession is another developmental generalisation which can be illustrated with the help of semantic maps. This functional spread is rather frequent cross-linguistically. Generally, it seems that connectives covering the prototypical function of concession are the ones that are most likely to add to their pragmatic meaning and also their semantics. This development is not so much a process of bleaching but rather a process of enhancing their range of meanings; the versatility of the conjunctions increases. The functional enhancing works along the lines of metonymy. Inside the domain of concessivity, there is a point of contiguity, namely the incompatibility of the two propositions or the irrelevance of one of them in relation to the other. These basic semantics are changed, enhanced or transferred to other semantic notions but they eventually always go back to these two fundamental aspects. Especially the function of Einräumung and all the following steps to correction express a subjective assessment of a situation by the speaker. Thus, these functions and their emergence are prime instances of Traugott’s third process of grammaticalisation according to which the meaning of grammatical markers become more and more based in the subjective evaluation and view of the world of a person.

7

Conclusion This paper dealt with the technique of semantic maps and applied it to the grammatical

category of concession. It became obvious that adverbial concession immediately lends itself to the visualization with semantic maps due to its semantic complexity. Adverbial con102

cession displays eight different functions divided into two groups of hypothetical concession and real concession. The majority of the functions are expressed by markers of all languages. This is only natural as the conceptual space, i.e. the underlying network of the functions, is assumed to be universal. Yet, one function seems to be the most basic, prototypical concessive function: that of ineffective anticause. If a language expresses concession, it has to be the basic semantics of an ineffective anticause. Therefore, this function occurs in the centre of the map. Furthermore, it turned out that the notion of concessivity is not stable within the category of adverbial concession. From left side to the right side of the map, the notion of concessivity increases and finds it climax at the centre of the map in the function of an ineffective anticause, i.e. in prototypical concession. From this point onwards, the notion of concessivity decreases again to the right. Interestingly, not only the concessivity but also the characteristics of adverbial subordination decrease, which culminates at the right pole of the map. Here, the conjunctions introduce a rather coordinated structure and function more as discourse markers. The notion of prototypical concession is weak and other pragmatic implications such as correction prevail. Concerning the revelation of universal tendencies, several points have been found out. Firstly, the central function of expressing an ineffective anticause can be considered universal as it is the most basic function of adverbial concession. The formal make-up of concessive conditionals turned out to show cross-linguistic similarities as well. Especially in hypothetical concession, scalar concessive conditionals consist of a focal particle and a conditional marker in almost all sample languages. Additionally, universal concessive conditionals all derive from free relative clauses and alternative concessive conditionals regularly resemble alternative interrogatives. However, these similarities may be due to the genetic affiliations of the four sample languages. Thus, it would need a broader sample to check these findings. The diachronic semantic developments of the various concessive markers reveal possible universal tendencies as well. The change from prototypical adverbial concession to more coordinate, corrective functions seems to occur throughout languages. Similarly, the development from hypothetical concession, more specifically from scalar concessive conditionals, to real concession, i.e. ineffective anticause, seems to happen regularly in all language samples. Another example for universal tendencies in the language sample is the close link between adverbial concession and other subordinate structures such as interrogative complement clauses and free relatives. 103

The comparison of the language samples shows that English and German display many similarities both in the distribution of semantic maps on the conceptual space and in the formal make-up of the conjunctions as well as their diachronic development. Surprisingly, Latin and French show no closer connection in their conjunctions at all. Obviously, Latin markers of concession got lost and French (as well as other Romance languages) developed totally new conjunctions. However, the diachrony of French was only shortly investigated in the appendix as a deeper analysis would require more space and time. It became obvious that the claims made on the basis of the interpretation of the semantic maps need a thorough back-up by linguistic data. Especially in Latin or French, some functions are not as well attested as in German or English. Here, a corpus analysis of adverbial concessive clauses would allow deeper insights and would provide more information about the frequency, formal marking or syntactic particularities of various concessive functions in different languages. In parallel, an analysis of a larger sample of languages would be necessary in order to check and support the claims of this paper. Here, the choice of languages should lead to a statistically balanced sample of less related languages from all over the world. Only then can truly universal tendencies in the grammatical category of adverbial concession become obvious. However, in this paper it was proven that adverbial concession immediately lends itself for the semantic map approach and leads to an elaborate functional network even with only four languages. Any further analysis will surely lead to many more interesting findings in this field.

104

APPENDIX The data of the language sample As has been explained above, the limits of the paper lead to a restriction of the sample to four languages. These languages belong to two different language families, the Germanic and the Romance languages. Thus, it is expected that they show different structures and connectors in the area of adverbial concession. With Latin being the ancestor language of Romance languages, one may assume that Latin employs concessive structures and conjunctions that function as origin and source of French adverbial concessive structures. Additionally, French heavily influenced English after the Norman Conquest in 1066. Thus, it is interesting to see whether there are similar constructions and subordinators used in the concessive relation, although languages only seldom borrow grammatical words and structures. German and English on the other hand belong to the same language family and probably show many similarities in concessive clauses. Still, they are representatives of different branches of Germanic languages and consequently may also show grammatical differences. Furthermore, German was not as heavily influenced by Latin and French as English. The sources for the data of adverbial concessive subordinators in the four languages are primarily descriptive standard grammars. Additionally, I used linguistic essays, analyses and monographs in order to support and complete the information on adverbial concessive clauses in all four languages. Still, it is highly probable that the data are not exhaustive and conclusive as they are not counterchecked by native speakers; moreover, some functional distinctions within the adverbial concessive relation cannot be attested on the basis of the sources used. However, this does not necessarily mean that they do not exist in the languages analysed. A corpus analysis on concessive clauses in the four sample languages, the analysis of concessive clauses in more languages and the sighting of more sources besides descriptive grammars may lead to new insights into the functional distinctions and the distribution of concessive connectives in both adverbial and coordinate concessive clauses. Consequently, the conceptual space and the language-specific semantic maps are only preliminary visualisations of adverbial concession and perhaps need to be re-ordered and elaborated.

105

English a. Inventory: English clearly differentiates between real concession and hypothetical concessive clauses. Thus, it employs a rather elaborate set of concessive subordinators. There is considerable overlap in the use and functions of the concessive conjunctions employed both within the relation of concession and between different relations such as concession, condition and contrast, place or time (cf. Quirk et al. 2008: 1088). The concessive subordinators may trigger a correlate in the main clause which is usually optional (ibid.: 1001). Prototypically, real concession is expressed by although, though and even though. Closely connected to real concession is the antithetic concessive use of while and whereas. Subordinators of minor importance are when, whilst, as and that. In hypothetical concessive clauses, English employs even if, whether…or, no matter whether and wh... –ever as well as no matter wh-.... b. Origins: although originally consisted of the two words all and though. The use of all made though more emphatic and emphasised its concessive notion. Due to semantic bleaching, the emphatic nature of the phrase was lost and all though and though became synonymous. In another step, the phrase was reanalysed: all lost its own semantic force and the phrase became one word. though derives from the OE form þéah, ðéah or þéh. Originally, though was the default concessive conjunction which could be strengthened with all. Nowadays, the concessive meaning of though can again be emphasised by even, leading to the concessive conjunction even though. The real concessive connective while derives from the OE phrase þá hwíle þe which translates into during the time that. This originally temporal conjunction was transferred into other contexts such as concession. It still grounds in the semantics of at the same time but additionally has an adversative or concessive meaning. The OED directly states that in concession, while is sometimes almost synonymous with although. It is important to note that the adversative meaning is more basic than the concessive one and consequently leads to the antithetic and contrastive function of concessive while. Similarly to while, whereas derives from a semantically different area. In the early stages of use, the concessive conjunction consisted of two single words, where and as. The source where indicates a locative origin and functions as 106

an adverbial of place. From the expression of a physical location, where was grammaticalised to a more abstract, figural use of confronting two propositions. Today, it commences statements of contrast or adversation and opposition, which is the reason of the notion of sharp antithesis in its concessive use. In scalar concessive conditionals, English employs even if. Morphologically and semantically, even if is very transparent. It consists of the conditional conjunction if and scalar focus particle even. The scalar semantics (and thus the concessive notion) of this type of hypothetical concession derives from the focal particle even, “[i]ntimating that the sentence expresses an extreme case of a more general proposition implied (=French même).” It is usually “[p]refixed [...] to the particular word, phrase, or clause, on which the extreme character of the statement or supposition depends [...]”9, i.e. the subordinate clause of the concessive structure. In universal concessive conditionals, English employs a whole wh-element series combined with the adverb ever. These wh-pronouns convey a generalising, indefinite sense indicating free choice. The exact identification, quality, place, and time of the content, any details of it do not matter. The main clause holds true in any case, condition and circumstances. The last type of connectives are those of alternative concessive conditionals. Here, English employs whether…or. whether derives from OG hwæþer or hweþer. As a conjunction, it is used to introduce a disjunctive clause together with or. According to the OED, these clauses often have a qualifying or conditional force, that is they are adverbial clauses and show a close semantic connection to condition. Similarly, they are closely related to alternative questions which are also introduced by whether. Semantically, whether…or expresses that the main clause holds true in either of the cases mentioned. c. Distribution:

(even) though Einräumung

whether…or

UCC

ACC

no matter wh… wh…ever

9

SCC even if

ineffective Anticause

restriction

correction although

while/whereas

contrast

http://oed.com/view/Entry/65255, last access 06.09.2011.

107

d. Functional distinction: The English language shows a wide variety of functions of adverbial concessive clauses. English concessive subordinators may express an insufficient objection, contrast, restriction, a correction, scalar concessive conditionals, alternative concessive conditionals and universal concessive conditionals. As the principle of contiguity allows deductions about the inventor of grammatical functions, it is assumed that English also employs the concessive function of Einräumung since although and even though cover all adjacent functions on the map. Thus, the semantic map of adverbial concession in English covers (almost) the whole underlying conceptual space. i.

Ineffective anticause (a) Sonia doesn’t speak French although she grew up in Paris. (Huddleston/Pullum 2002: 734) (b) He used my mower even though I told him not to. (Quirk et al. 2008: 1097)

ii.

Contrast (c) While I admit I did it, I didn’t intend to. (d) Whereas the amendment is enthusiastically supported by a large majority in the senate, its fate is doubtful in the House. (Quirk et al. 2008:1097).

iii.

Restriction (e) His skin was suitably pale, though less so than that of many London gentleman […]. (f) The Visigoths even became Christian, although it was the wrong sort of Christians […]. (Rudolph 1996: 411)

iv.

(Einräumung) (g) Although the book is very good, it is of no use to me.

v.

Correction (h) A: I would love to come along, really. Although, I wouldn’t like a ten hour flight. (cf. Günthner 2000: 441)

vi.

SCC (i) I’m going out even if it rains. (Huddleston/Pullum 2002: 737) 108

(j) Even if it rains, the Open Air concert will take place.

vii.

ACC (k) Whether Martin pays for the vase or replaces it, I’m not inviting him again. (Quirk et al: 2008: 1100)

viii.

UCC (l) Whatever I say to them, I can’t keep them quiet. (Quirk et al. 2008: 1101) (m) Wherever that man goes, there is trouble. (Rudolph 1996: 429)

German a. Inventory: Similarly to English, German knows the difference between real concession and concessive conditionals, i.e. hypothetical concession. Consequently, German employs a similarly large number of concessive conjunctions. More often than in other languages, German allows for correlates in the main clause which emphasise the concessive notion of the sentence. Frequent examples are trotzdem and dennoch. Proper concession is prototypically expressed by obwohl, obgleich and obschon with obwohl being the most versatile, stylistically neutral and most frequently used concessive connective. Additionally, German employs wenn auch, wenngleich and wennschon for expressing functions of real concession such as contrast and restriction as well as Einräumung and a correction. Of those three, wenn auch is the most usual concessive conjunction. Besides its use as a correlative in the main clause, trotzdem can also be used as a concessive connector in the subordinate clause. This use is generally restricted to colloquial language and leads to a different stress pattern in the pronunciation of the connective (cf. Grammatik der deutschen Sprache 1997: 2294). In hypothetical concessive clauses, German uses auch wenn in scalar concessive conditionals, ob…oder in alternative concessive conditionals and w... –auch immer in universal concessive conditionals.

109

b. Origins: The etymology of the German prototypical concessive connective obwohl is difficult to characterise. According to Duden, ob originates in the meaning of über, oberhalb von (= above, over). It can also refer to an originally alternative interrogative conjunction meaning something like both. Thus, it would be semantically comparable to English whether in semantics and use; this becomes obvious in its occurrence as conjunction in alternative concessive conditionals, which are closely related to alternative questions (cf. Kluge 2002: 660). The Grimms in their lexicon on the German language attribute the conjunctive use of ob to an old Germanic word meaning doubt. This underlying notion of doubt seems to be a probable origin for the concessive and conditional use of ob. In obwohl and similarly in obgleich, obzwar and obschon, the concessive ob is supported and emphasised by conceding adverbials such as gleich, wohl, schon and zwar. wohl originally derives from wollen meaning want or wish. Verbs of volition are a frequent source for concessive markers. Originally, ob and its intensifiers were phrases consisting of two words; however, due to grammaticalisation processes of bleaching and reanalysis, these phrases developed into default one-word-conjunctions. auch wenn and wenn auch both consist of the conditional subordinator wenn meaning if and a scalar focus particle auch (or gleich and schon). Structurally and semantically, they are similar to English even if (= auch wenn) and even though (=wenn auch). wenn originally derives from wann, a temporal adverb. wenn is usually used in conditional surroundings introducing factual or hypothetical conditions. The concessive use derived from this conditional use immediately. wenn can be either used absolutely in concessive sentences, i.e. alone, or followed or preceded by a focal particle. In universal concessive conditionals, German employs conjunctions in the form of a w-pronoun + auch immer or a structure of egal + a w-pronoun. Similarly to English, these concessive conditional subordinators derive from interrogative or generalising relative pronouns indicating free choice and the irrelevance of an exact identification of the referent.

110

c. Distribution:

obwohl

Einräumung

ob…oder

UCC

ACC

w…auch immer

SCC

restriction

correction

ineffective anticause

auch…wenn

contrast wenn…auch

d. Functional distinctions: As the description of the inventory shows, German displays a very wide variety of functions of adverbial concessive clauses. In fact, German concessive connectors cover all semantic distinctions of the conceptual space of concession. Grammatical analyses of concessive clauses and descriptions in grammars such as the Duden and the Grammatik der deutschen Sprache directly refer to the manifold and diverse functions of concession in the German language when they particularly emphasise Einräumung and correction as well as restrictive uses. i.

Ineffective anticause (a) Obwohl er sich sehr beeilte, kam er zu spät. (Duden 1986: 811) ‘Although he hurried, he was late.’

(b) Auch wenn er sich sehr beeilte, kam er zu spät. ‘Even though (even if?) he hurried, he was late.’

ii.

Contrast (c) Wenn das Zimmer auch sehr klein ist, (so) ist es doch sehr hell. ‘While the room is very small, it is still very light.’

iii.

Restriction (d) Er war ein sehr großer Mensch, Anfang dreißig, obwohl er älter wirkte im ersten Augenblick. (Rudolph 1996: 411) ‘He was a very tall man, around thirty, although he looked older at first sight.’

iv.

Einräumung (e) Wenn das Buch auch sehr gut ist, ist es doch für mich wenig nützlich. (Duden 1986: 794) ‘Even though the book is very good, it is still quite useless for me.’

111

v.

Correction (f) Ich mag kein Bier, für mich also nicht – obwohl, gib mir doch mal eins. ‘I don’t like beer, so none for me – however, give me one, too.’ (g) A: Ich fand den Film schlecht. B: Ja, obwohl – die Idee war ziemlich kreativ. ‘A: I thought the film was horrible. B: Yes, but – the idea was pretty creative.’

vi.

SCC (h) Sie will unbedingt zum Vortanzen, auch wenn sie sehr krank ist. ‘She desperately wants to the audition, even if she is very ill.’

vii.

ACC (i) Ob es regnet oder schneit, ich wandere jedes Silvester auf den Brocken. ‘Whether it rains or snows, I climb on the Brocken every New Year’s Eve.’

viii.

UCC (j) Wer (auch) immer du bist, ich bin an deiner Seite. ‘Whoever you are, I’m at your side.’ (k) Was immer sie tut, er liebt sie trotzdem. ‘Whatever she does, he loves her anyway.’ (l) Wo (auch) immer er hingeht, sie folgt ihm bedingungslos. ‘Wherever he goes, she follows him unconditionally.’

French a. Inventory: Just as other languages of Europe, French differentiates between real concession and hypothetical concession. In both areas of concessivity, it uses distinctly different subordinators and structures. In real concessive settings, French employs quoique, bien que, malgré que and more seldom encore que. bien que is rather restricted to written French, while quoique and malgré que seem to be the prototypical concessive conjunctions. In hypothetical concessives, French shows a rather peculiar pattern of structures and subordinators which clearly differs from the subordinating markers and mechanisms of the other sample languages. Similar to English, German and Latin, 112

French employs concessive connectors based on a conditional one, namely même si which is semantically similar to quand bien même. Both are used in scalar concessive conditionals. Parallel to other European languages, French also employs wh-elements in universal concessive conditionals, leading to the subordinator quoi que. In a parallel way, quelque…que and quel(le)(s)…que are used in universal concessive conditionals as well. Distinctly different are the French connectives tout…que, si…que, aussi…que, and pour…que. They express a variable fact usually referring to an expression of degree or value which, although it may be of a high scale, does not suffice to invalidate the main action or hinder it to hold true. Thus, they may also be used in scalar concessive conditionals as they basically fulfil the same function. Interestingly, French completely lacks subordinators indicating alternative concessive conditionals. According to König and Haspelmath, it expresses alternative concession by a pattern of two main clauses conjoined by or with their verbs in the subjunctive optative mood (cf. König/Haspelmath 1998: 598). It is noteworthy that French verbs in concessive clauses show different moods which are usually motivated by the conjunction of the subordinate clause. The subjunctive is the default mood of concessive clauses although they refer to factual, real events. However, via the influence of spoken French, the indicative spreads into originally subjunctive constructions. b. Origins: Morphologically, the conjunction quoique is a combination of quoi and que. quoi originally is a relative or interrogative pronoun of indefinite character. Thus, the default concessive conjunction of French derives from a frequent source of usually concessive conditional markers, a development found quite frequently throughout languages (cf. Dictionnaire historique de la languge française 2007: 3050). A variant of quoique was que que which poses the question if French employs the strategy of reduplication in order to derive at concessive conjunctions (compare quamquam in Latin). Usually, quoique takes the subjunctive but nowadays is also found with the indicative. The Dictionnaire historique points out that these indicative structures are semantically closer to the coordinated structure of adversation (cf. ibid).

113

The real concessive conjunction bien que originates in the Latin adverb bene meaning good. Thus, the conjunction derives from the antonym of mal, another source for a real concessive conjunction in French, malgré que. Originally, bien could be used on its own in order to express concession; however, in modern French this use is substituted by bien que (cf. Dictionnaire historique 2007: 392). malgré que is a compounded conjunction consisting of mál, a preposition after the French adjective meaning bad, and gré, a noun deriving from the Latin word gratum which translates as agreeable or favourable. Consequently, the noun has the meaning of agreement or consent. The semantics of the conjunction malgré que translate as against the consent of and thus, indicate the surprising nature and the violation of underlying expectations and presuppositions in concession. Another real concessive conjunction is encore que. encore is an adverb which derives from the Latin phrase hinc ad horam or hinc hac hora (= until now). Thus, encore que is fundamentally temporal in its meaning as it originates from an adverb of time. However, it is quite common for temporal adverbs to gain concessive meaning, as can be seen in English while or Latin cum. Nowadays, encore que is only used in written, formal or literary contexts and in the concessive function of restriction and autocorrection (cf. Dictionnaire historique 2007: 1235). In hypothetical concessive conditionals, French employs a wide variety of connectives which are characterised by very different properties. Similarly to Latin, English and German, French also uses a conjunction consisting of the conditional conjunctive si and an emphatic focus particle, même. même derives from the Latin word metipsimus, a form of the emphatic -met and the demonstrative ipse. Thus, même functions as emphatic particle similar to even, auch and etiam indicating either focus, grade or similarity. In universal concessive conditionals, French employs either quoi que, quelque…que or quel(le)(s)…que. All three conjunctions consist of an indefinite relative pronoun (quoi), interrogative pronoun (quel, quelle, quels, quelles) or an indefinite adjective (quelque) and the connective que. quelque…que, according to the Larousse, marks a concession which is characterised by indetermination in its semantics. This indefiniteness is typical for universal concessive conditionals. quel(le)(s)…que, an interrogative adjective or pronoun, derives from the Latin word qualis meaning of what kind or nature. In its function as interrogative adjective, it is 114

used as conjunction in concessive circumstances. However, in recent times, it is usually substituted by quelque…que. Again, typically for universal concessive conditionals, the nature or character of the referent, i.e. the content of the subordinate clause, is irrelevant and totally non-descript (cf. Dictionnaire historique 2007: 3034f.). The other conjunctions used in the function of hypothetical concessive conditionals in French differ considerably from the concessive markers used in other European languages. tout…que has its source in the adverb tout and marks the intensity or degree of an adjective or adverb framed by the two parts of the conjunction, thereby expressing an extreme endpoint of a scale10. According to Haspelmath and König, tout…que as well as the constructions (aus)si…que and pour…que belong to universal concessive conditionals. si...que does not derive from the conditional conjunction si as one would surmise due to condition being a frequent source for concessive markers. It rather grounds in the adverb si which in turn derives from the Latin adverb sic meaning in that way. In the structure si + adjective + que, si may introduce a hypothetical concessive clause (cf. Dictionnaire historique 2007: 3496f.). aussi…que consists of two Latin words, aliud (= other) and the adverb sic. Originally used for expressing comparison, aussi…que can function as concessive marker when it expresses a degree or intensity that in its extreme value does not suffice to invalidate the main clause content (cf. Dictionnaire historique 2006: 260). The French concessive connective pour…que originates from the Latin preposition pro referring either to substitution, in favour of or before (cf. Dictionnaire historique 2007: 2884f.). The concessive structure pour…que is typical for Romance languages and can be found in Spanish and Portuguese, too (cf. Harris 1988). c. Distribution: Einräumung UCC

ACC

SCC

quoique même si quelque…que quel(le)…que (aus)si…que, pour…que, tout…que 10

ineffective anticause

restriction

correction

quoique, malgé que, bien que

contrast

encore que

http://www.larousse.com/de/worterbucher/franzosisch/tout/78775, last access 09.09.2011.

115

d. Functional distinction: The descriptive grammars of French directly name the concessive functions of expressing an ineffective anticause, restriction, an autocorrection, scalar concessive condition and universal concessive condition. There is no descriptive evidence for the concessive function of Einräumung and of contrastive concession in French. At least Einräumung can be deduced as existent due to the principle of contiguity on semantic maps. Alternative concessive condition cannot surface on the semantic map as it is not encoded by an adverbial concessive conjunction. i.

Ineffective anticause (a) Bien que je n’aie pas de temps, je fais une promenade. ‘Although I don’t have time, I go for a walk.’ (b) Malgré qu'il ne pleuve pas, j'ai un parapluie avec moi. ‘Although it doesn’t rain, I carry an umbrella with me.’ (c) Quoiqu'il ne travaille jamais pour l'école, il continue à avoir des bonnes notes. ‘Although he does not work for school, he continues to have good marks.’

ii.

Contrast

iii.

Restriction (by principle of contiguity)

iv.

(Einräumung) (by principle of contiguity)

v.

Correction (d) Il viendra demain. Encore que je n’en sois pas tout à fait sûr. (Grammair méthodique du français 2009: 862) ‘He will come tomorrow. However, I am not totally sure of it.’

vi.

SCC (e) Même s’ il pleut on va se baigner.

vii.

ACC

116

viii.

UCC (f) (Aus)si méticuleux que soit le règlement, il ne parvient pas à tout prévoir. (König/Haspelmath 1997: 619) ‚However meticulous the regulation might be, it cannot succeed in forseeing everything.’ (g) Ce Texte, pour intéressant qu’il soit, n’est pas probant non plus. (ibid.) ‘This text, however interesting it is, is no proof either.’ (h) Quoi qu’il advienne, observe cette règle. (ibid.) ‘Whatever happens, observe this rule.’

Latin a. Inventory: Latin shows a rather diverse pattern of concessive conjunctions. Similarly to the three other sample languages, it differentiates between real and hypothetical concession. The superordinate clause may show correlatives such as tamen, at, certe and combinations of the three. In real concessive clauses, Latin usually employs quamquam and cum. cum here seems to conjoin to antithetical facts, i.e. it is used in antithetical concession. Although quamvis translates as a concessive conditional it can be found in real concessive circumstances. In hypothetical concessive clauses, Latin shows a wide variety of connectives such as etsi, etiamsi, tametsi, licet, ut and quamvis. However, not only quamvis can be used in both hypothetical and real concessive surroundings. etsi is the default concessive conjunction in Caesar and thus, is by no means restricted to hypothetical concessive clauses. In its translation, ut suggests a hypothetical concessive function; however, in the correlative use of ut…ita, it rather has an adversative function and expresses a contrast between the two propositions. Similarly to all the other European languages, Latin also employs indefinite relative clauses in concessive function (more specifically in universal concessive conditionals). They are introduced either by forms of quisquis and quidquid or relative forms conjoined with cumque. The content of the subordinate clause is unassertive, or rather the reference of the subordinator is unasserted and thus, irrelevant for the truth of the main proposition.

117

The conjunctions in Latin demand either the subjunctive, which occurs after the major part of conjunctions, or the indicative, which occurs after quamquam. b. Origins: The Latin concessive conjunction quamquam has the structure of reduplicated quam, meaning how, like, as. In the doubled form, the semantics change to whatever, however, anyhow. Thus, the concessive notion is derived by reduplication, a process that indicates the origin of quamquam in the concessive conditional semantics. Similarly, quam is etymologically part of quamvis. The second part of this connective has its source in velle, i.e. a form of want. Accordingly, quamvis translates directly into however much you want. Originally, it stood next to adjectives and adverbs as can be seen in a context such as ‘Tam quam vis subito venias.’ which translates as ‘You may come as fast as you want.’ By reanalysis, quamvis was attributed to venias and changed into a conjunction (cf. Rubenbauer 1989: 317f.; cf. Walde-Hofmann 1938 (1982): 397f.). Another concessive conjunction of Latin is licet which originally was a verb form in a paratactic sentence in the subjunctive. The speakers of Latin started to analyse the subjunctive as dependent on licet and it developed into a concessive marker. The verb originally had the meaning of being allowed or may – a notion that is still easily traceable in the concessive semantics of licet as even if (cf. Walde-Hofmann 1938 (1982): 797). cum, a very productive and functionally diverse conjunction, may also have an adversative and concessive meaning in Latin. It originally derived from quom, an adverbial accusative singular masculine which in turn grounds in the relative stem of relative pronouns. In its concessive function, it takes the subjunctive; however, the reason is unclear. The adversative and concessive use of cum developed from the narrative use in which it prototypically indicates the temporal circumstances of the main action. Thus, in order to use cum in an adversative or concessive meaning, the content of the clauses must be in contrast or opposition. In Latin, hypothetical concessive clauses are expressed by etsi, etiamsi, tametsi, sive…sive and connectives deriving from relative pronouns. Conjunctions like etsi and etiamsi are both morphologically and semantically transparent. They consist of the conditional conjunction si meaning if and adverbial expressions of also or even. However, they are not only used in conditional concessives as their structural make up would suggest. Caesar, for example, used exclusively etsi in traditionally conces118

sive surroundings. However, Rubenbauer groups these concessive clauses close to hypothetical, i.e. conditional clauses (cf. Rubenbauer 1989: 318). sive…sive encodes a disjunctive structure of two conditions and is thus semantically similar to whether…or and ob…oder. Interestingly, English and German do not employs their conditional conjunctions but rather historically conditional markers which are no longer used in this context (whether and ob, comparatively). Latin, on the other hand, does employ the conditional connective si in alternative concessive conditionals. Universal concessive conditionals can be encoded by two strategies in Latin. Firstly, relative pronouns may be reduplicated in order to form generalising, free relative clauses which may have a concessive meaning. This leads to connectives of the form of quisquis and quidquid. Secondly, universal concessive conditional connectives may have the form of an interrogative pronoun (structurally similar to relative pronouns) which is conjoined to the generalising particle -cumque (cf. Walde-Hofmann 1938 (1982): 309f.). c. Distribution: quamquam

Einräumung

sive…sive etsi, quamvis

UCC

ACC

qu…cumque quisquis quidquid

SCC

ineffective anticause

restriction

correction

licet

contrast

etiamsi tametsi

ut…ita(sic) cum

d. Functional distinction: The functional distinction of concession turns out to be more difficult in Latin than in the three other sample languages. The prototypical concessive function of expressing an ineffective anticause is covered by quamquam and, in the case of Caesar at least, by etsi. Scalar concessive conditional clauses are expressed by etsi, etiamsi and tametsi. quamvis expresses the irrelevance of any degree or intensity of an objection or anticause – it cannot invalidate the main clause. Thus, quamvis expresses fundamentally scalar concessive condition, too. However, it can also be used in prototypical concessive circumstances and then, refers to “eine[r] Einräumung, deren Annahme man in hohem Grade dem Zuhörer überlässt (Menge 2000: 860). sive…sive expresses solely alternative concessive conditionals. 119

Universal concessive conditionals are introduced by forms of quisquis and quidquid as well as by indefinite relative pronouns combined with -cumque. cum expresses two functions of real concession, i.e. contrastive or antithetic concession and prototypical concession. This polyfunctionality originates in the semantic widening from narrative cum to adversative and concessive cum which leaves the functions of this conjunction semantically closely related and often interchangeable. Functionally similar, ut…ita may be used in order to express antithetic concession. licet is somewhat harder to attribute to a particular function. In its semantics, it is closest to the function of conceding, of an Einräumung. As it has its semantic origin in the verb licere, meaning to allow and to permit, it rather expresses a fact that from the subjective point of view of the speaker has a conceding force which nevertheless does not suffice to invalidate the main clause. In its frequent German translations of mag es auch sein, dass and zugegeben, dass, this function of Einräumung becomes clear. Thus, the semantic map of Latin covers almost the entire conceptual space of concession. i.

Ineffective anticause (a) Quamquam gratiarum actionem a te non desiderabam, tamen fuit ea mihi periucunda. (Rubenbauer, Hofman, Heine 1989: 318) ‚Although I did not expect an expression of thanks from you, it nevertheless pleased me a lot.’ (b) Socrates cum facile posset educi e custodia, noluit. (ibid.: 305) ‘Although Sokrates could have been easily abducted from the prison, he did not want to.’

ii.

Contrast (c) Ut reliquorum imperatorum res adversae auctoritatem minuunt, sic Vercingetorigis ey contrario dignitas incommode accepto in dies augebatur. (ibid.: 295) ‘While failings diminished the authority of the other generals, Vercingetorix’ authority on the other hand increased day by day inspite of the defeat.’ (d) A Caesare nemo ad Pompeium transiit, cum paene cottidie a Pompeio milites ad Caesarem perfugerent. (ibid.) ‘Nobody went over from Caesar to Pompeius, while almost daily soldiers went over from Pomeius to Caesar.’

iii.

Restriction 120

(by principle of contiguity)

iv.

Einräumung (e) Licet pericula impendeant, omnia subibo. (Rubenbauer, Hofman, Heine 1989: 318) ‘Although dangers threaten me, I will undergo everything.’

v.

Correction (f) O poetam egregium! Quamquam ab his contemnitur. (ibid.: 318) ‘Oh, what an outstanding poet! However, those despise him.’

vi.

SCC (g) Corpus etiamsi mediocriter aegrum est, sanum non est. (ibid.) ‘Even if the body is only slightly ill, it is not healthy.’ (h) Romani tametsi a fortuna deserebantur, tamen omnem spem salutis in virtute ponebant (ibid). ‘Even if fortune had left the Romans completey, they still put all their hoped on virtue.’ (i) Paupertas si malum est, mendicus beatus esse nemo potest, quamvis sit sapiens. (Rubenbauer, Hofman, Heine 1989: 318) ‘If poverty is bad, no beggar can be happy, even if he is a wise man.’

vii.

ACC (j) Hoc loco libentissime soleo uti, sive quid mecum ipse cogito, sive quid scribe aut lego. (ibid.: 312) ‘I like to be at this place, whether I want to think by myself or to write and read.’

viii.

UCC (k) Enecas me odio, quisquis es. (Rubenbauer, Hofman, Heine 1989: 286) ‘I almost die of hate towards you, whoever you are.’

121

Bibliography

CRISTOFARO, Sonia (2010), “Semantic maps and mental representation.” Linguistic Discovery 8.1, 35-52. CROFT, William (2001), Radical construction grammar: syntactic theory in typological perspective. Oxford: OUP. CROFT, William & Alan CRUSE (2004), Cognitive Linguistics. Cambridge: CUP. CROFT, William (2006), Typology and universals. 2nd ed. Cambridge: CUP. CROFT, William (2010), “What do semantic maps tell us? Comment on ‘Semantic maps and mental representation’ by Sonia Cristofaro.” Linguistic Discovery 8.1, 53-60. DE HAAN, Ferdinand. On representing semantic maps. Manuscript in progress. http://www.u.arizona.edu/%7Efdehaan/papers/semmap.pdf, last access: 13.10.2011. DE HAAN, Ferdinand (2010), “Building a semantic map: top-down versus bottom-up approaches.” Linguistic Discovery 8.1, 101-117. DRYER, Matthew S. (1992), “Adverbial subordinators and word order asymmetries.” in John A. HAWKINS, Anna SIEWIERSKA (eds). Peformance principles of word order. EUROTYP Working Papers, European Science Foundation, 50-67. Duden (1985), Grammatik der deutschen Gegenwartssprache. hrsg. von der Dudenredaktion. Peter EISENBERG et.al. Mannheim [u.a.]: Dudenverlag. Duden (2005), Die Grammatik. hrsg. von der Dudenredaktion. Kathrin KUNKEL-RAZUM et al. Mannheim [u.a.]: Dudenverlag. GÄRDENFORS, Peter (2000), Conceptual spaces: the geometry of thought. Cambridge: MIT Press. GRÉVISSE, Maurice & André GOOSSE (2008), Le bon usage: grammaire française. Bruxelles: De Boeck-Duculot. GÜNTHNER, Susanne (2000), “From concessive connector to discourse marker: The use of obwohl in everyday German interaction.” in COUPER-KUHLEN, E. & KORTMANN, B. (eds.) Cause, Condition, Concession, Contrast: Cognitive and Discourse Perspectives. Berlin: Mouton de Guyter, 439-468. HARRIS, Martin (1988), “Concessive clauses in English and Romance.” in J. HAIMAN & S. THOMPSON (eds.), Clause combining in grammar and discourse. Philadelphia: J. Benjamins Publishing Company. HASPELMATH, Martin (1997), Indefinite pronouns. Oxford: Clarendon Press. 122

HASPELMATH, Martin (2003), “The geometry of grammatical meaning: Semantic maps and cross- linguistic comparison.” in M. TOMASELLO (ed.) The new psychology of language vol. II. Mahwah, NJ: Erlbaum, 211-242. HOPPER, Paul & Elizabeth C. TRAUGOTT (2003), Grammaticalization. 2nd ed. Cambridge: CUP. HUDDLESTON, Rodney D. & Geoffrey K. PULLUM, Laurie BAUER (2002), The Cambridge grammar of the English language. Cambridge: CUP. KÖNIG, E. & P. EISENBERG (1984), "Zur Pragmatik von Konzessivsätzen." in STICKEL, G. (ed.), Pragmatik in der Grammatik. Jahrbuch 1983 des Instituts für deutsche Sprache. Düsseldorf: Schwann. KÖNIG, Ekkehard (1985), "On the history of concessive connectives in English. Diachronic and synchronic evidence." Lingua 66, 1-19. KÖNIG, Ekkehard (1988), "Concessive connectives and concessive sentences. Crosslinguistic regularities and pragmatic principles." in J. HAWKINS (ed.), Explaining Language Universals. Oxford: Basil Blackwell. KÖNIG, E. & E.C. TRAUGOTT (1991), "The pragmatics-semantics of grammaticalization revisited.” in B. HEINE & E.C. TRAUGOTT (eds.), Approaches to Grammaticalization. Amsterdam: Benjamins. KÖNIG, E. & B. KORTMANN (1992), "Categorial reanalysis: The case of deverbal preposition." Linguistics 30, 671-698. KÖNIG, E. & M. HASPELMATH (1998), "Concessive conditionals in the languages of Europe." in J. VAN DER AUWERA (ed.), Adverbial Constructions in the Languages of Europe. Berlin: Mouton, 563-640. KÖNIG, Ekkehard & P. SIEMUND (2000), „Causal and concessive clauses: Formal and semantic relations.“ in COUPER-KUHLEN, E. & KORTMANN, B. (eds.) Cause, Condition, Concession, Contrast: Cognitive and Discourse Perspectives. Berlin: Mouton, 341-360. KORTMANN, Bernd (1996), Adverbial subordination: a typology and history of adverbial subordinators based on European languages. Berlin: Mouton the Gruyter. MALCHUKOV, Andrej J (2004), “Towards a semantic typology of adversative and contrast marking.” Journal of Semantics 21.2, 177-198. MENGE, Herman (2000), Lehrbuch der lateinischen Syntax und Semantik. Darmstadt: Wissenschaftliche Buchgesellschaft. MIKKOLA, Eino (1957), Die Konzessivität bei Livius: mit besonderer Berücksichtigung der ersten und fünften Dekade. Eine stilistische Untersuchung. Helsinki: Akateeminen kirjakauppa. 123

NARROG, Heiko & Johan VAN DER AUWERA. Grammaticalisation and semantic maps. forthcoming. http://webh01.ua.ac.be/vdauwera/25Narrogvan%20der%20Auwera_Nov7.pdf, last access: 13.10.2011. QUIRK, Randolph et al. (2008), A comprehensive grammar of the English language. Harlow, Essex: Longman, Pearson Education. RIEGEL, Martin et al. (2009), Grammaire méthodique du francais. 7th ed. Paris: Presses Université de France. RUBENBAUER, Hans & J.B. HOFMAN, R. HEINE (1989), Lateinische Grammatik. Bamberg: Buchner. RUDOLPH, Elisabeth (1996), Contrast: adversative and concessive relations and their expressions in English, German, Spanish, Portuguese on sentence and text level. Berlin: Mouton de Gruyter. SANDRA, Dominike (1998), "What linguists can and can't tell you about the human mind: A reply to Croft." Cognitive Linguistics 9, 361-378. SANSO, Andrea (2010), “How conceptual are semantic maps?” Linguistic Discovery 8.1, 299309. TRAUGOTT, Elizabeth C. & Bernd HEINE (1991), Approaches to grammaticalization. Vol. I. Amsterdam: Benjamin. VAN DER AUWERA, Johan & Vladimir PLUNGIAN (1998), “Modality’s semantic map.” Linguistic Typology 2, 79-124. VERHAGEN, Arie (2000), “Concession implies causality, though in some other space.” in COUPER-KUHLEN, E. & KORTMANN, B (eds.), Cause, Condition, Concession, Contrast: Cognitive and discourse perspectives. Berlin: Mouton de Gruyter, 361-380. WÄLCHLI, Bernhard (2006/2007), Construction semantic maps from parallel text data. Manuscript in progress. http://ling.uni-konstanz.de/pages/home/a20_11/waelchli/waelchlisemmaps.pdf, last access: 13.10.2011. ZIFONUN, Gisela et al. (1997), Grammatik der deutschen Sprache. Bd. 3. Berlin: Mouton de Guyter. ZWARTS, Joost (2010), “Semantic map geometry: two approaches.” Linguistic Discovery 8.1, 377-395.

ENCYCLOPEDIAS/DICTIONARIES REY, Alain (ed.) (2007), Dictionnaire historique de la langue francaise. Paris: Le Robert. 124

VAN DER AUWERA, Johan & X TEMÜRCÜ (2006), “Semantic maps.” in Encyclopedia of Language and Linguistics. 2nd ed. Keith Brown et al. (eds). Amsterdam [u.a.]: Elsevier. WALDE, Alois & J.B. HOFMANN (1982), Lateinisches etymologisches Wörterbuch. 3. Auflage. Heidelberg: Winter.

INTERNET SOURCES http://oed.com/view/Entry/65255, last access 06.09.2011. http://www.larousse.com/de/worterbucher/franzosisch/tout/78775, last access 09.09.2011. http://ling.uni-konstanz.de/pages/home/a20_11/waelchli/waelchli-semmaps.pdf, last access: 13.10.2011. http://webh01.ua.ac.be/vdauwera/25Narrogvan%20der%20Auwera_Nov7.pdf, last access: 13.10.2011. http://www.u.arizona.edu/%7Efdehaan/papers/semmap.pdf, last access: 13.10.2011.

125