The semantic predecessors of "need" : from old to early modern English

UNIVERSIDADE DE SANTIAGO DE COMPOSTELA FACULTADE DE FILOLOXÍA DEPARTAMENTO DE FILOLOXÍA INGLESA THE SEMANTIC PREDECESSORS OF NEED : FROM OLD TO EARLY MODERN ENGLISH Lucía Loureiro Porto 2005 Universidade de Santiago de Compostela Facultade de Filolox í a Departamento de Filolox í a Inglesa THE SEMANTIC PREDECESSORS OF NEED : FROM OLD TO EARLY MODERN ENGLISH Tese de Doutoramento realizada por Luc í a Loureiro Porto e supervisada pola Dra. Elena Seoane Posse Visto e Prace da directora Asdo. Luc í a Loureiro Porto Asdo. Dra. Elena Seoane Posse To my brother, in memoriam ACKNOWLEDGEMENTS This piece of research is the result of several years of investigation, and I would like to express my gratitude to those people and institutions which have contributed to its fulfilment. First of all, I would like to express my deepest gratitude to my supervisor and friend, Dr. Elena Seoa ne, not only f or her constant advice and words of encouragement, but also for her being at my disposal 24 hours a day for the l ast years. Without her this piece of research would have never been possible. I also want to thank a number of people who have contributed to the improvement of this study with their comments on earlier drafts: Professor Teresa Fanego, whose advice has proved pr ovidential for the progression of this piece of research, Dr. Bel é n M é ndez, who has also selflessly helped me with the translation of some examples, Dr. Mar í a Jos é L ó pez Couso, who was a witn ess of the birth of this piece of investigation, Dr. Tom á s Jim é nez Juli á and Dr. Benedict Buono, who were members of the examination board of my M.A t hesis. Thanks go also to Dr. Christopher Williams for his comments on stylistic matters. Needless to say, errors or omissions that remain are my resp onsibility. I am deeply indebted to Dr. Alejandr o Alcaraz, who ha s provided me with vital corpus resources in a sign of extraordina ry generosity. Li kewise, I thank Dr. Cristina Su á rez G ó mez for her big-heartedness in supplying me with different dictionaries, corpora and bibli ographical sources. Thanks go out to Dr. Mercedes Casal, Jos é Luis Sim ó n Cameselle and Dr. Ramom Chantada for having c ontributed to partly neutralize my computer ignorance. Several people have kindly let their lives be disturbed by m y bibliographical needs, namely Laura Barrio, Mar í a Blanco, Roberto Fdez. Sobrado, Bob Foster, Dr. Ana Mart í nez Insua, Dr. Paloma N úñ ez, Dr. Ignacio Palacios, and Andr é s Pi ñ eiro. I truly appreciate their help. My friends have also played an important role in the fulfilment of this study. Although I have already mentioned some and names are too many to say them all, I would like to extend my tha nks to H é ctor Agrafojo, Rosa Cougil, Mar í a Fari ñ a, Anabel Gonz á lez Cruz and Fernando Rodr í guez-Gallego, because they have been indeed good companions in this journey. My most special thanks go to Á lex, whose unconditional support, prodigious patience, infinite generosity and good sense of humour have never surrendered to the unidirectional compulsions which invaded me during the writing of this dissertation. Last but not least, I give the most he artfelt thanks to my parents for their love and serenity throughout my life, and especially to my mom, who has always devoted her life to satisfy other pe ople ’ s needs. I am also indebted to the Galician Ministry of Education ( grant numbers PGIDT01PXI20404PR and PGIDIT02PXIC20402PN), the Spanish Ministry of Education and Science (grant numbers : BFF2001-2914 and HUM2004-00940) and AEDEAN, for their economic contribution to this piece of research. Cover design: Alejandro Herrada Mederer. Santiago de Compostela, February 2005 Table of contents i TABLE OF CONTENTS LIST OF TABLES v LIST OF FIGURES xii 1. INTRODUCTION 1 1.1. Aim of the study 1 1.2. Outline of the study 6 2. GRAMMATICALIZATION, MODALITY AND IMPERSONALITY 7 2.1. Grammaticalization and l anguage change 7 2.1.1. Why do languages change? 8 2.1.2. Mechanisms of language change 10 2.1.2.1. Reanalysis 11 2.1.2.2. Analogy 12 2.1.2.3. Se mantic change 16 2.1.3. Grammaticalization 19 2.1.3.1. Processes and parame ters of grammaticalization 21 2.1.3.2. The role of reanalysis and analogy within grammaticalization 25 2.1.3.3. Grammaticalization: a unidirectional phenomenon? 29 2.1.3.4. English modal s: a paradigmatic example of grammaticalization 31 2.2. Present-Day English need and need to : an insight into modality 35 2.2.1. Morphosyntactic features 35 2.2.1.1. Traditional considerations 35 2.2.1.2. Re cent approaches 42 2.2.2. Semantic features 45 2.2.2.1. The concept of modality 46 List of tables viii Table 4.14: Types of strong external forces expressed by active and passive neden v.1 269 Table 4.15: Origin and intensity of the forces expressed by ME neden v.2 273 Table 4.16: Types of strong external forces expressed by nede n v.2 with specification of clause polarity 273 Table 4.17: Types of strong internal forces expressed by ME neden v.2, with indication of clause polarity 276 Table 4.18: Neutral ge neral nede n v.2 with indication of clause pol arity 280 Table 4.19: Types of forces expressed by M E neden v.2 according to origin, strength and clause polarity 282 Table 4.20: Type of infinitival theme of nede n v.2 without an explicit experiencer 287 Table 4.21: Experienc er verb constructions of neden v.2 with a nomi nal theme 289 Table 4.22: Experienc er verb constructions of neden v.2 with an infinitival theme 294 Table 4.23: Syntactic features of neden v.2 taking into account the presence or absence of the experiencer and the nature of the theme 301 Table 4.24: Experienc er verb constructions with ME neden v.2 by sub- periods 302 Table 4.25: Distribution of ME bihove n by subperiods 304 Table 4.26: Origin and intensity of the forces expressed by ME bihoven 305 Table 4.27: Types of neutral gene ral forces expressed by bihoven , with indication of clause polarity 309 Table 4.28: Types of sentential the mes of bihoven without an explicit experiencer 315 Table 4.29: Experienc er verb constructions of bihoven with a sentential theme 322 Table 4.30: Syntactic features of ME bihoven taking into acco unt the presence or absence of the experiencer and the nature of the theme 331 Table 4.31: Distribution of experiencer and non-ex periencer verb constructions with ME bihoven by subperiods 332 Table 4.32: Freque ncy of the six verbs in Middle English 336 Table 4.33: Origin and intensity of the forces expressed by each ME verb 338 Table 4.34: Semantic implicati ons of the six ME ‘ need ’ - verbs 340 List of tables ix Chapter 5: Table 5.1: Number of words in the eModE section of the Helsinki Corpus per subperiod 374 Table 5.2: Number of words per decade and text-type in the CEECS and the Lampeter Corpus 375 Table 5.3 Number of words per subperiod in my eModE corpus 376 Table 5.4: Frequency of each ver b in the eModE corpus 377 Table 5.5: Distribution of eModE need by subperiods 378 Table 5.6: Origin and intensity of the forces conveyed by eModE need. 379 Table 5.7: Types of strong external forces conveyed by eModE need with indication of clause polarity 380 Table 5.8: Types of strong internal forces conveyed by eModE need with specification of clause polarity 384 Table 5.9: Types of ne utral general forces conveyed by eModE need with specification of clause polarity 390 Table 5.10: Syntactic features of eModE need without an experiencer: chronological distribution 396 Table 5.11: Distribution of animat e and inanimate experiencers with eModE need in a variant of Allen ’ s ( 1995) Type II construction 402 Table 5.12: Type of sentential theme and evolution of experiencer-anim acy with eModE need in Allen ’ s Type ‘ Personal ’ construction 405 Table 5.13: Use of auxiliaries with eModE need in Type ‘ Personal ’ constructions 406 Table 5.14: Presence of morpheme {-eth} or {-es} whe n eModE need does not have an experiencer 417 Table 5.15: Presence of morpheme {-eth} or {-es} whe n eModE need has an experiencer 418 Table 5.16: Distribution of eModE behove by subperiods 419 Table 5.17: Origin and intensity of the forces expressed by eModE behove 420 Table 5.18: Types of neutral general forces expressed by eModE be hove , with indication of clause polarity 420 Table 5.19: Types of theme of eModE behove without an experiencer 423 Table 5.20: Types of theme found with eM odE behove with an experiencer 425 Table 5.21: Freque ncy of the two eModE ver bs 428 List of tables x Table 5.22: Origin and intensity of the forces expressed by each eModE verb 429 Table 5.23: Semantic implications of the two eModE verbs 430 Table 5.24: Types of experiencer verb constructions found with eModE need and behove 432 Chapter 6: Table 6.1: Frequency of each verb from Old English to early Modern English 435 Table 6.2: Frequency of all verbs per chronological subperiod 436 Table 6.3: Origin and intensity of the forces expressed by tharf per subperiod 439 Table 6.4: Types of forces and barriers expressed by tharf from Old to late Middle English, with indication of clause polarity 440 Table 6.5: Themes exhibited by tharf per subperiod 443 Table 6.6: Experiencer verb constructions in which tharf is found 446 Table 6.7: Origin and intensity of the forces expressed by betharf per subperiod 448 Table 6.8: Types of forces and barriers expressed by betharf from Old to Middle English, with specification of clause polarity 448 Table 6.9: Themes exhibited by be tharf per subperiod 449 Table 6.10: Experienc er verb constructions in which betharf is found 450 Table 6.11: Origin and intensity of the forces expressed by need throughout its history 452 Table 6.12: Types of forces and barriers expressed by need from Old to early Modern English, with specification of clause polarity 454 Table 6.13. Chronological distribution of act ive and passive need v.1 and need v.2 throughout history 459 Table 6.14: Types of compleme nt of active need v.1 throughout history 460 Table 6.15: Types of compleme nt of passive need v.1 throughout history 460 Table 6.16: Presence of the experiencer with need v.2 from Old to early Modern English 461 Table 6.17: Themes of need v.2 without an experiencer: chronol ogical distribution 462 Table 6.18: Themes of need v.2 with an experiencer per subperiod 462 List of tables xi Table 6.19: Evoluti on of experiencer verb constructions with need v.2 throughout history 463 Table 6.20: Chronological distribution of sentential themes in Allen ’ s Type ‘ Personal ’ constructions with need v.2 464 Table 6.21: Animacy of the experiencer of need v.2 in Type ‘ Personal ’ constructions from Old to early Modern English 465 Table 6.22: Origin and intensity of the forces expressed by behove 467 Table 6.23: Types of forces expressed by be hove from Old to early Modern English, with indic ation of clause polarity 468 Table 6.24: Presence of the experiencer with behove from Old to early Modern English 470 Table 6.25: Type of theme of behove without an explicit experiencer per subperiod 470 Table 6.26: Type of theme of behove with an experiencer per subperiod 471 Table 6.27: Evoluti on of experiencer verb constructions with behove 471 List of figures xii LIST OF FIGURES Figure 2.1: Development of auxiliary be going to (from Hopper and Traugott 2003: 93) 15 Figure 2.2: Modality: from the extra-linguistic world to grammar (adapted from Palmer 1986) 48 Figure 2.3 : Grammaticalization of modality (based on Palmer 1986: 4) 49 Figure 2.4 : Logical relations between necessity and possibility 52 Figure 2.5: Meanings conveyed by modal verbs in the history of English: from the physical to the mental world 57 Figure 2.6: Basic constituent structure of Type S impersonal constructions (from Elmer 1981: 26) 73 Figure 3.1: Degree of grammaticalization of OE pre-modals cunnan and *sculan (from Goossens 1987: 138) 103 Figure 3.2: Scale of de semanticization of modals (from Goossens 1987: 118) 104 Figure 3.3: Meanings implied by modal verbs in the history of English: from the physical to the mental world (adapted from Traugott 1989 and Sweetser 1990) 172 Figure 3.4: Frequency of the four verbs in early and late Old English 203 Figure 4.1(a): Freque ncy of the six verbs in the four ME subperiods 336 Figure 4.1(b): Frequency of four of the verbs in the four ME subperiods 337 Figure 5.1: Frequency of the two verbs in the three eModE s ubperiods 428 Figure 6.1: Frequenci es of my verbs from Old to early Modern English 437 Figure 6.2: Semantic evolution of German d ü rfen 441 Figure 6.3: Semantic evolution of English tharf 441 Figure 6.4: Semantic confluence of need v.1 and need v.2 in M3 (1350-1420) 456 Figure 7.1: Take over of need v.1 with need v.2 485 Chapter 1. Introduction 1 CHAPTER 1 INTRODUCTION 1.1 Aim of the study This study is part of a larger research project called “ Variation, linguistic change and grammaticalization, with special reference to English, ” which is being carried out by a number of researchers at the De partment of English of the Universidade de Santiago de Compostela. 1 As is well-known, the study of grammaticalization is burgeoning, as is attested by the recent publication of numerous volumes (cf., among others, Traugott and Heine 1991, Giacalone Ramat and Hopper 1998, Fischer et al. 2000, Bybee and Hopper 2001, Wischer and Diewald 2002). Furthermore, there is considerable interest in the grammaticalization of modals (cf. Plank 1984; Goossens 1987; Heine 1993; Warner 1993; Bybee et al . 1994; van der Auwera and Plungian 1998; Krug 2000, 2001, 2002; Sturiale 2002; Traugott and Dasher 2002, Aijmer 2004, Tagliamont e 2004, among many others). More specifically, the marginal position of Present- Day English (henceforth PDE) need has been highlighted by several scholars (to cite just a few, Bolinger 1942, Jacobsson 1974, Duffley 1994, Leech 2003, Smith 2003, Taeymans 2004a). Most of these works concentrate on the twofold character of need , which may be cons idered modal and non-modal (cf. Huddleston 1984, or Quirk et al . 1985, for instance). However, these studies concentrate on the synchronic features of need and tend to neglect its historical 1 These researchers are the members of the Research Group on Variation and Language Change (cf. <http://www.usc.es/ia303/vlc/main.htm l>), led by Professor Teresa Fanego. Chapter 1. Introduction 2 evolution. Further explorations of the development of need are, therefore, necessary to explain the complex nature of this verb. A preliminary overview of the history of need reveals some interesting developments. As a modal verb, need differs morphologically from the central modals, since, unlike the central modals, need does not derive from the Old English (henceforth OE) preterite-present verbs. An insight into Ol d English shows that, at that time, preterite-present verbs already included a ver b semantically and syntactically equi valent to PDE need , namely OE þ ur fan. Þ urfan , however, has not survived into Present-Day English, so we may hypothesize that its disappearance favoured the auxiliarization of need . Further overviews of Old a nd Middle English (henceforth ME) reveal, however, that þ urfan and need are not the only verbs meaning ‘ need ’ in the history of Englis h, but that there are other verbs, such as OE be þ urfan , OE behofi an or ME misteren , which are semantically equival ent and which, in principle, could also have replaced þ urfan . Surprisingly enough, need i s the only verb which survives with its meaning in Present-Day English and, what is more, it has bec ome one of the 100 most frequent verbs in spont aneous speech (cf. Krug 2000: 291). From this brief outline we can easily gather that only a thorough analysis of the history of need and of those verbs which may have competed semantically with it can disclose the reasons why need has ousted its semantic competitors and the factors determining its PDE double nature as auxiliary and non-auxiliary. For this reason, the aim of this work is to elucidate the evolution of PDE need and its semantic predecessors. The historical period selected for my analysis is from Old to early Modern English (henceforth eModE) for the following reasons. Firstly, a good number of studies on PDE need are already available, as well as research in progress which will be published shortly. 2 Secondly, as Rissanen (1999: 189) states, “ [t] here are, in fact, very few major syntactic changes after the end of the 18 th century, although change in language is of course an ongoing and neve r-ending process. ” In other words, we expect the major changes to occur before the late Modern English (lModE) period. Finally, a preliminary study of the relevant literature revealed that after early Modern English the occurrences of nee d and behove , the surviving verbs, were not indicative of further changes which might prove relevant for the general purpos es of this study. In this period, need and behove had bec ome the verbs they are in 2 Cf. the work of Soili Nokkonen, which is being supervised by Professor Terttu Nevalainen at the Research Unit for Variation and Ch ange in English (VARIENG), University of Helsinki. Chapter 1. Introduction 3 Present-Day English from a semantic and syntactic perspective, and the semantic predecessors of need have disappeared or become specialized with a different meaning. More specifically, the aims of this study are to find out how need and its semantic predecessors compete for the expression of the same meaning, how they develop morphologically, syntactically and semantically through time, and the reasons why they survive, change or disappear from the language. In addition, through the analysis of nearly 1000 years of linguistic evolution I intend to provide information about the grammaticalization of need and any of its semantic competitors as modal auxiliaries. Grammaticalization is here understood in the traditional functional approach in which most scholars posit it. Thus, in this piece of wor k I follow works such as Lehmann (1995 [1982]), Hopper (1991), Heine (1993), and Hopper and Traugott (2003), and consider that grammaticalization is the result of subsequent changes in the semantic, morphosyntactic and phonological components of language, through processes such as desemanticization, decategorialization or erosion. In addition, modal verbs are said to be born out of reanalysis (cf. Hopper and Traugott 2003). I will try to ascertain the extent to which some of my verbs undergo some of these processes. Since the semantics of the verbs analysed here falls within the modal notion of necessity, it will be necessary to interpret the findings from a modal perspective. For this reason, modality is crucial in this piece of work. Among the different approaches to modality, which classify it according to various criteria, I will examine two widely recognized classifications, namely that of Palmer (1979, 1986, 2003), and that of Sweetser (1990), based on Talmy (1988). I will try to prove that, even though views of modality such as Palmer ’ s (1979, 1986, 2003) have proved essential for the study of this semantic category in sync hronic works, a more dynamic view of modality is necessary for the purposes of this diachronic piece of research. For this reason, the approach to modality followed here is the result of a combination of Coates ’ (1983) and Sweetser ’ s (1990) classification of modality as divided into root and epistemic, together with Talmy ’ s (1988, 2000) cognitive semantic approach to modal meanings in terms of force dynamics. In other words, modality will be analysed from a functional- cognitive point of view. This descriptive framework will prove indispensable for the explanation of the semantic evol ution of need , since this stems from cognitive forces. Chapter 1. Introduction 4 Together with their liability to undergo grammaticalization and their modal meanings, the verbs analysed in this study converge on another lingui stic area which concerns syntax and semantics, namely their occurrence in impersonal constructions. In fact, the expression of necessity has traditionally been considered to favour impersonal constructions, i.e. constructions in which the experiencer is inflected for the oblique case instead of the nominative. There are numerous classifications and interpretations of impersonal constructions in the literature, and I will review the most relevant, namely Elmer (1981), Fischer and van der Leek (1983, 1987) and Allen (1995), and I will then justify my decision to follow Allen (1995). Her approach to impersonality accounts not only for the OE description of these constructions, but also for their evolution t hrough history, which proves very suita ble for a diachronic study such as this one. We will see that the adherence of the verbs to one or the other type sheds light on their degree of grammaticalization an d in their semantic development. The convergence of m y verbs on these three theoretical aspects, namely grammaticalization, modality and impersonality, accounts for the coherence of my verbs both semantically and syntactically. Since the analysis covers nearly one millenni um, the semantic predecessors of PDE need differ notably from one peri od to another. Beginning with Old English, the predecessors of need attested are þ urfan, be þ urfan, ne odian and behofian . OE þ urfan bel ongs to the group of preterite-present verbs which, as mentioned, have evol ved, in many ca ses, to PDE modal auxiliaries (e.g. OE *sculan > PDE shall ). Although, as is well-known, þ urfan does not survive into Present-Day English (except in some nort hern dialects, cf. OED s.v. tharf v. ), in Old English it exhibits a high freque ncy of occurrence and c onveys a wide range of meanings. From þ urfan a m orphological derived verb is recorded, namely be þ urfan , formed with the addition of the prefix <be->, which also means ‘ need. ’ As for neodian , it is the etymological ancestor of PDE need and is included in the study for obvious reasons. Finally, behofian is the etymological predecessor of PDE behove , which in Old English meant ‘ need, ’ rather than ‘ be fitting, ’ its PDE meaning. The list of semantic predecessors of need in Middle English is larger than that in Old English. Thurven (<OE þ urfan ) c ontinues to be used i n the language and is phonologically confused with a nother preterite-present, durren (>PDE dare ). On some occasions, durren occurs instead of thurven , while on other occasions we find blends of both verbs (e.g. þ ar t , whose initial part seems to Chapter 1. Introduction 5 belong to thurven , and whose fi nal part seems to belong to durren ). Bethurven (<OE be þ urfan ) also remains in the group, but is only a ttested in the very early years of Middle English. Neden and bihove n (<OE neodi an and behofian respectively) become the prevalent ver bs of the group, especially at the end of the period. Finally, misteren , a French loanword , will also be used as a ‘ need ’ - verb in this period, but will have an ephemeral life in the English language. Finally, the set of verbs anal ysed in early Modern English is reduced to two, namely need and behove , the only ver bs which survive from Middle English and the only ones which exist in Present-Day English, although, as is well- known, they no longer compete semantically. The analysis of these OE, ME and eModE verbs consists of two different parts. The first is a revision of specialized literature, which will serve to provide a preliminary description of their morphologi cal, semantic and syntactic features. The second part is of an empirical nature and it will focus on the analysis of real linguistic data extracted from several computerized corpora with the aim of testing and enlarging the information dr awn from the literature (cf. Mair 2004, for instance, for the necessity of c orpus-linguistics in grammaticalization studies). In fact, the focus of this piece of research is the detailed corpus-based analysis of each of the verbs in the three periods. With the aim of examining a representative selection of texts, a series of prestigious corpora have been s ubject of scrutiny, as will be duly explained in sections 3.4.0, 4.4.0 and 5.3.0. Briefly, the corpora selected are, (i) the complete Helsinki Corpus of English Texts , which comprises 1.5 million words distributed into Old, Middle and early Modern English, (ii) an 800,000-word selection of texts from the Dictionary of Old English Corpus , (iii) a 600,000-word selection of texts from the Corpus of Middle English Prose and Verse ; and, finally, for early Modern English, (iv) the Corpus of Early English Correspondence Sampler and the Lampeter Corpus . All in all, my diachronic corpus contains 4.1 million words, which will be used as the source of examples of the semantic predecessors of need from Old to early Mode rn English. All such examples will be introduced in a database and analysed a ccording to semantic, syntactic and morphological criteria. The analysis of the corpus examples is twofold. Firstly, I will offer a synchronic analysis of each verb in each chronological period (Old, Middle and early Modern English), payi ng special attention to their semantic, syntactic and morphological features, especially to those which may be indicative of their Chapter 2. Grammaticalization, modali ty and impersonality 12 complement (cf., among many others, McMahon 1994: 130 ff.; Trask 1996: 139; Schendl 2001: 40), as in (2.1) (2.1) þ am cynge licoden peran the (dat) king (dat) were-pleas ing pears (nom) ‘ Pears were pleasing to the king ’ = ‘ The king liked pears. ’ The dative object, normally expressing a human referent, usually appeared in pre-verbal position as in (2.1). In Middle English, with the weakening of the unstressed syllables and the final loss of infl ectional endings, a sentence such as (2.1) became (2.2): (2.2) The king liceden peares the king were-pleasing pears ‘ Pears were pleasing to the king ’ = ‘ The king liked pears. ’ The only vestige of pears being t he grammatical subject is the plural form of the verb. However, this agreement suffix was also lost in the course of time. Without any number agreement, speakers of the late ME peri od reanalysed the segment the king as the subject of the verb, and the segment pears as the object. This came to be known when speakers began to us e new number markers which agree with the new subject. In fact, we say now The king likes pears , or The kings like pears . 4 Reanalysis, therefore, proves to be a crucial mechanism for language change of both morphological and syntactic type. Much has been written on reanalysis as related to grammaticalization, so I will go over reanalysis again below when discussing grammaticalization as another mechanism of la nguage change. 2.1.2.2. Analogy Analogy is another well-known mechanism of language change. In fact, Meillet already referred to it, and to what we have called reanalysi s, as ways of development of new grammatical forms (as mentioned in Hopper and Tra ugott 2003: 63 ff.). As opposed to reanalysis, analogy operates in the paradigmatic axis. According to Campbell (1998) analogy may be of two types: proportional or non-proportional, which roughly corresponds to McMahon (1994: 70-76) 4 Despite the apparent clarity of this example as an instance of syntactic rean alysis, scholars specialized in the study of impersonals disagree with this interpretation, as will be duly explained in section 2.3. Chapter 2. Grammaticalization, modali ty and impersonality 13 systematic and sporadic analogy. Non-proportional anal ogy refers to unexpected changes in language such as hypercorrection (e.g. * for you and I instead of for you and me ). This kind of language change will not be the subject of this piece of research. Consequently, I will just concentrate on so-called proportional analogy. 5 Proportional analogy involves analogical levelling (or just levelling) on the one hand, an d analogical extension (or four- part analogy), on the other. A look at both of them will make it clear why these two types of analogical processes are called proportional. In the first place, analogical levelling is the result of a paradox, known as Sturtevant ’ s paradox (cf., for example, McMahon 1994: 91; Trask 1996: 108), which states that “ sound change is regular, but produc es irregularity; analogy is irregular, but produces regularity. ” This is better understood with an example. If we consider OE verb c Ɲ osan ‘ t o choose, ’ we observe that its paradigm is very irregular phonologically, since there is alternation between three sounds, namely [z], [s] and [r], due to some sound changes undergone by Proto-Germanic which involved intervocalic /s/. These sound changes produced the irregular paradigm found in Old English. However, in t he course of time, all the differences of pronunciation of these sounds were levelled to [z], producing a re gular paradigm, as shown in (2.3): (2.3) OE PDE Pres. c Ɲ osan [z] > choose [z] Past. Sing. c Ɲ as [s] > chose [z] Past. Pl. curon [r] > chose [z] Past. Pple. gecoren [r] > chosen [z] Thus, all the differences which were produced by sound change in this verbal paradigm have been levelled out by t he mechanism called analogical levelling. Other verbs undergoing the same kind of le velling are OE fr Ɲ osan ‘ to freeze, ’ and ( for)l Ɲ osan ‘ to lose; ’ this one only retains the original sound alternation in some isolated forms such as lovelorn and forlo rn . 5 In fact, among what Campbell (1998) calls non- proportional analogy (cf. also McMahon 1994: 75-76 sporadic analogy) there are other examples, such as “ folk etymology ” (e.g. crayfish, understood as com pound of fish , < French single morpheme crevice ) “ back formation ” (e.g. pea < OE pease ) or “ metanalysis ” (e.g. adder < OE n æ ddre ) Fro m my point of view, however, these processes could be included in the m ore general label reanalysis. It is i mportant to bear in m ind, though, that analogy can be a powerful driving force for reanalysis to take place, since some of these forms have been reanaly sed due to analogy with other existing forms in the language. Chapter 2. Grammaticalization, modali ty and impersonality 14 Secondly, analogical extension refers to the addition of new form s to an already existing paradigm, based on anot her paradigm. For instance, English has several nouns of Latin origin ending i n – us which make their plural forms changing that ending for – i , as in cactus/cacti . As a conse quence of this, when speakers find a word such as octopus , they may extend the – us/-i pa radigm to this word, creating a plural form octopi , instead of octopuses . Other well-known examples include those of PDE strong verbs being inflected for the past as weak verbs, as in shell : shelled :: swell : swelled , instead of swollen . Analogical extension, as well as reanalysis, is not only a mechanism for morphological change, but also for syntactic change. In fact, it is one of the three only processes of syntactic changes recognized by Campbell (1998: 226). Although the view followed in this piece of work is broader and involves other kinds of mechanisms, one of Campbell ’ s (1998) examples will help illustrate what syntactic analogical extension is. Let us the consider Spanish reflexive construction Juan se visti ó ‘ John has dressed himself up. ’ The pronoun se is the marker of the reflexive construction. However, this se may also be found in ambiguous sentences such as El rico se entierra en la iglesia , in which se may be a reflexive marker, implying that ‘ the rich one has himself buried in the churc h, ’ or, on the other hand, may be a passive ma rker, meaning ‘ the rich one gets buried in the church. ’ This ambivalent interpretation does not reveal that the construction has undergone analogical extension. We can only notice it when we find sentences such as Los vinos se venden en esta ciudad . In this sentence the reflexive interpretation is not possible because the subject is inanimate, so the only possible interpr etation is ‘ the wines are sold in this city. ’ Therefore, the reflexive construction with se has undergone analogical extension and it has broadened to be used with any type of transitive verb and any type of subject. A similar example of analogical extension is the already illustrated case of the English verb like . It has been said that its construction was reanalysed: the original dative object becomes the subject and the ori ginal subject becomes the object. This type of reanalysis is supported by the exis tence of similar kinds of constructions in the language. It is very common in English to have a subject designating a human referent, and this ha s made possible the extension of that kind of construction to the verb like . This cooperation of reanalysis and analogy is anything but rare. In fact, in language change norm ally several mechanisms operate together. Chapter 2. Grammaticalization, modali ty and impersonality 15 Another example is provided by Hopper and Traugott (2003), concerning the development of the English auxiliary be going to , as shown in the next figure: Syntagmatic axis Mechanism: reanalysis Stage I be going [to visit Bill] PROG Vdir [Purp. clause] Stage II [be going to] visit Bill TNS Vact (by syntactic reanalysis/metonym y) Stage III [be going to] like Bill TNS V (by analogy/metaphor) P a r a d i g m a t i c a x i s Mechanism: analogy Figure 2.1: Development of auxiliary be going to (from Hopper and Traugott 2003: 93). This figure shows the development of be going to in the syntagmatic and paradigmatic axes (reanalysis and analogy respectively) from an original structure of the directive verb to go in the progressive aspect followed by a purpose clause specifying the reason for the movement (stage I). Due to syntactic reanalysis in the linear constituents, the infinitive marker to is attached to the verb to go , without any “ surface manifestation, ” resulting in an element expressing tense followed by a n activity verb, which is no longer a purpose clause (stage II). This reanalysis is also a metonymic change, since it “ involves specifying one meaning in terms of another that is present, even if onl y covertly, in the context ” (Hopper and Traugott 2003: 93). The last step of the process involves the paradigmatic axis. That is, once be going to be comes a fixed structure, the following verb may be of any type, even a stative verb, such as like . As seen in stage III, the verb like may occur in construction wit h be going to , which in t he previous stage could only be followe d by a verb which could ma tch the directive status of go . This means that the paradigm of verbs which may follow be going to is expanded by analogical extension. It has been seen that in the processes of reanalysis and analogy, elements are prone to undergo semantic change also. E.g. OE lician ‘ to be pleasing ’ turns Chapter 2. Grammaticalization, modali ty and impersonality 16 to PDE like , the verb go in the be going to construction loses its meaning of movement to imply future tense. Semantic change plays, therefore, an important role in language change. 2.1.2.3. Semantic change It is more difficult to systematize semantic change than morphological or syntactic change, because there may be various mechanisms are involved in semantic change. For the purposes of this study, I will follow W. Lehmann (1992: 260-274), McMahon (1994: 174- 184), Campbell (1998: 256-266) and Schendl (2001: 29-34) in order to provide a list of the main processes of this kind of change. The three main reasons for semantic change are: change in linguistic contexts, change in the referent and change due to language c ontact (cf. W. Lehmann 1992: 260). Changes in the linguistic context may provoke a change in the meaning of a word. For example, French pas has its main meaning ‘ step ’ in most of the contexts. However, when it appea rs in negative constructions, its meaning is not the same: pas becomes just a negative particle. Meaning changes depe nding on the context may be classified as follows (cf. Campbell 1998: 261 ff.; McMahon 1994: 179): Degeneration . “ [A] downward move in evaluative attitude ” (McMahon 1994: 179). Due to the linguistic context in which a word appears, its mean ing may undergo pejoration, and become less positive, gaining negative value. For instance, madam may refer to a lady or to t he female head of a house of prostitution depending on the context. Elevation . A word may under go melioration, that is, it may acquire positive connotations in the minds of the speakers due to its use in a certain context. An example is the noun knight , which originally meant ‘ boy, youth, attendant ’ (Schendl 2001: 31), or the adjective sophisticated , which now means ‘ worldly-wise, intellectually appealing, cultured ’ rather than ‘ artificial, ’ its original meaning (McMahon 1994: 179). Taboo replacement . Due to their use in a context of taboo content, the noun ass , once used to designate an animal is being replaced by donkey , since the latter is not associated with taboo topics. The same happens with cock , which has progressively ceased to be used to refer to the animal because it has obscene connotations, and is gradually being replaced by rooster . Chapter 2. Grammaticalization, modali ty and impersonality 17 Hyperbole . Due to exaggeration by overstatement, the English word lame came to mean ‘ stupid, awkward, socially inept, ’ from its original meaning ‘ crippled, having an impaired limb. ’ The second reason for semantic change recognized by W. Lehmann (1992) has to do with a change in the referent of a word. For instance, English pen derives from Latin penna ‘ feather, ’ since a feather u sed to be the instrument for writing. Once modern tools were created to accomplish the same goal, the name was retained. Some mechanisms which imply a change of referent are the following (c f. McMahon 1994: 182-184; Campbell 1998: 256 ff.): Widening . The meaning of a word may widen through time, and for instance the English word dog has today a general meaning, while the corresponding OE docga used to refer to ‘ a specific breed of dog. ’ The same happens to the Spanish word p á jaro , which has evolved from the Latin word passer which referred to a specific kind of bird, namely ‘ sparrow. ’ Narrowing . A change in the opposite dire ction: from more general to more specific. For instance, the referent of English meat w a s f o o d i n O E ( O E mete ), and the noun used to refer to dogs in general was hound (OE hund ). Metaphor . This is the most classic change of referent. It may be defined as the “ transfer of a term because of a real or imagined similarity, ” as, for instance, the use of the term neck to refer to the part of a bottle which is somewhat similar to the part of a body (Schendl 2001: 126) An interesting example of fossilised metaphor is the English word bead , which has evolved from OE bed, beode ‘ prayer, ’ since prayers were usually accompanied by a rosary, which, at the same time, was made of beads. Therefore, the referent of OE bed, beode changed from ‘ prayer ’ to ‘ bead, ’ due to a metaphoric relation (Campbell 1998: 258). Metonymy . Metonymy is said to arise from contiguity of meanings and to involve “ real rather than imagined links between concepts ” (McMahon 1994: 183), and is defined as the “ semantic change in which an attribute of a thing is used to denote the whole ” (Sche ndl 2001: 126). A much cited example is the use of White House instead of the American president . Both metaphor and metonymy are considered the main mechanisms of semantic change (cf. Traugott and Dasher 2002: 27-34) and they imply a high degre e of subjectification. 6 In 6 The term subjectification is used in this study as understood by Traugott (1989, for instance) and Traugott and Dasher (2002), i.e. the “ semasiological process whereby SP/Ws [speakers / Chapter 2. Grammaticalization, modali ty and impersonality 18 addition, metaphor, metonymy and subjectification are highly relevant in grammaticalization, and, for this reason, they will be dealt with below (section 2.1.3.1). 7 Synecdoche . The referent of a word may change to more comprehensive or to less comprehensive. For instance, hand may mean ‘ hired hand, employe r. ’ Poetically, moon may not refer to the satellite, but to a month. Litotes . Sometimes, exaggerations are used in language and a word may acquire a different meaning. For instance, French poison , from which English poison was borrowed, used to mean ‘ potion, draught. ’ Due to litotes, it came to mean ‘ poison. ’ The third reason for semantic change involves contact between langua ges. Usually language contact brings about many calques and loanwords, by means of borrowing (cf. W. Le hmann 1992: 266-274), such as the above mentioned case of the word of French origin army replacing OE here . Together with this type of lexical change, cultural contact can also produce semantic changes, although more rarely than borrowing. In this connection, Campbell (1998: 266) cites an example of Lake Mikow (California), whi ch originally ha d the word k ó :no meaning ‘ bow. ’ The meaning of this word changed to ‘ gun ’ after the colonization, due to the contact of two worlds, and the language de veloped a new way of calling the bow: hint í :l k ó :no ‘ old-time gun ’ (for a thor ough explanation of language contact and its effects, see Lass 1997: 184-207). Summing up, the three main reasons for semantic change are cha nge in the linguistic context, change in the referent and change due to language contact. A number of mechanisms may interact in each of these circumstances, of which only some have been mentioned here. The different types of language change which ha ve been the subject of our attention in section 2.1.2 and its subsections , namely morphological, syntactical and semantic changes, may occur together, so it is necessary to analyse lingui stic writers] come over time to develop meanings for Ls [lexemes] th at encode or externalise their perspectives and attitudes as constrained by the communicative world of the speech event, rather than by the so-called “ real world ” characteristics of the event or situation referred t o ” (Traugott and Dasher 2002: 30). This way I follow scholars such as Mortelmans (2003), as opposed to Langacker (1999, for example), whose notion of subjectification im plies the “ attenuation in the degree of control exerted by an agentive subject ” (1999: 297) . 7 For theoretical discussions on the relation between metaphor and metonymy, see the collection of papers in Barcelona (2000a). Chapter 2. Grammaticalization, modali ty and impersonality 19 changes from these three levels. This is especially the case with the study of grammatical elements since they differ from lexical elements in syntax, morphology and semantics, as is the case, for example, of English auxiliary verbs as against lexical verbs. The main purpose of this piece of research is to analyse the evolution of verbs meaning ‘ need ’ in the history of English, and to find the paths followed by these verbs which have le d to the PDE binomial need / need to , verbs which have undergone or are undergoing grammaticalization to some extent. With the aim of identifying the criteria to reco gnize grammaticalization, the next section provides the basic notions to understand this mechanism of language change. We will analyse the extent to which some of the mechanisms examined so far may be subsumed into the more general label grammaticalization, and to what extent they are indepe ndent mechanisms. 2.1.3. Grammaticalization The first studies on grammaticalization date back to 18 th century French philosophers such as de Condill ac and Rousseau (cf. chapters on the origin of grammaticalization in Lehmann 1995 [1982], Heine et al. 1991, and Harris and Campbell 1995). However, from a linguistic point of view, it is not unt il the beginni ng of the 20 th century that we come across a definition of the term (Meillet 1912: 131): “ l ’ attribution du charact è re grammatical à un mot jadis autonome ” ( ‘ the attribution of grammatical character to an erstwhile autonomous word ’ ). Although Meillet ’ s account of grammaticalization (like all the writing of his time) considers linguistic change a deterioration of the language, his initial paper is regarded as “ the germs of modern ideas on gra mmaticalization ” (Hopper and Traugott 2003: 25). In fact, more recent definitions of the term seem to be paraphrases of Meillet ’ s. See, for instance Kury á owicz ’ s (1965), Lehmann ’ s (1995 [1982]) and Hopper and Traugott ’ s (2003) defi nitions of the term: Grammaticalization consists in the increas e of the range of a morpheme advancing from a lexical to a grammatical or fro m a less grammatical to a more grammatical statu s, e.g. from a deri vative formant to an inflectional one (Kury á owicz 1965: 52) Grammaticalization is a process leading from lexemes to grammatical formatives (Lehmann 1995 [1982]: viii) We define grammaticalization as the process whereby lexical items and constructions com e in certain linguistic contexts to serve grammatical functions, Chapter 2. Grammaticalization, modali ty and impersonality 20 and, once grammaticalized, continue to develop new grammatical functions (Hopper and Traugott 2003: xv) These definitions have much in common with Meillet ’ s: grammaticalization is the process whereby a give n form gains grammatical status. Grammaticalization, therefore, is expected to take place within the grammar of a given language, and not only within syntax, as claimed by Tr ask (1996: 143). Whe n an element is said to be grammaticalized it has undergone several processes which affect its morphology, its syntax, its semantics and sometimes its phonology. A paradigmatic example of grammaticalization is the fut ure tense of Romance. The original Latin future was a synthetic form, as, for example, cantabo ‘ I will sing. ’ At the same time, Latin had periphrastic for ms such as cantare volo ‘ I want to sing, ’ and cantare habeo ‘ I must sing. ’ In the course of time, the periphrastic structure containing the verb habere , ‘ to have, ’ became more and more frequent in contexts conveying future meaning, and it finally gave birth to the future tense of Romance languages, as in Spanish cantar é ‘ I will sing ’ (< cantar he ). It is widely recognized that these forms are grammaticalized, because they have undergone different processes. At first sight, it is obvious that ther e has been a syntactic change, since the original periphrastic structure has become a single form. Syntax, therefore, has led to morphol ogy. This is what a now classic slogan by Giv ó n (1971: 413) states: “ Toda y ’ s morphology is yesterday ’ s syntax. ” The form has also undergone phonological changes as a result of the merge of the two original elements. Finally, there has also been a change of meani ng, since the Latin periphrastic form implied the modal meanings of obligation or intention and the current form conveys mainly future tense. Grammaticalization, therefore, involves syntax, mor phology, semantics and phonology. In addition, grammaticalization also takes pragmatics into consideration, since the relationship between the speech act participants (SAPs) is crucial in the study of language change, as mentioned in section 2. 1.1 (cf. Hopper and Traugott 2003: 72; Harris and Campbell 1995: 54; Haspelmath 1999: 1055). 8 For instance, in the grammaticalization of the English epistemic modal must it is important not to forget about the pragmatic context, since a sequence such as you must have experience may convey obligation (e.g. in a job offer), or deduction (in a context which refers to your capability to manage with something). 8 As general works on pragmatics, see Levinson (1983) and Green (1989). Chapter 2. Grammaticalization, modali ty and impersonality 21 Grammaticalization, consequently, is a complex mechanism which explains the changes undergone by some elements of the language, adopting different perspectives and involvi ng different linguistic disciplines. This is what has made detractors of grammaticalization deny its independent status, because “ what it claims to explain is explained already by other well-understood mechanisms which lie behind it and, as is generally agreed, it cannot “ explain ” without appeal to these other mechanisms and kinds of change ” (Campbell 2001: 151; cf. also Campbell 1998: 242; Campbell & Janda 2001; Janda 2001; or Joseph 2001). 9 Whether grammaticalization has independent s tatus or is a derived mechanism for language change, what is relevant for my purposes is that it is a comprehensive mechanism which describes changes in the grammar of a language. For that reason, the following sections will describe the parameters for the identification of grammaticalization. 2.1.3.1. Processes and parame ters of grammaticalization From the last quarter of the 20 th century scholars have tried to provide an appropriate test with parameters for the identification of grammaticalization. Lehmann (1995 [1982]; 1985) is the first one who proposes a tidy set of criteria, on the basis of the characteristics of a lexical form and of a grammatical one. Other scholars follow and identify some principl es underlying grammaticalization (Hopper 1991) and yet others describe the grammaticalization chain according to four processes (Heine 1990, 1993). Finally, other scholars provide new evide nce for grammaticalization on the basis of general characteristics of the process, such as reanalysis, analogy or unidirectionality (cf. Hopper and Traugott 2003). Contrary to all of them, there are authors who prove t hat some of the earlier-mentioned symptoms of 9 On a different line, Nuyts (2001) not only considers that grammaticalization is not independent, but claims for its explanation in cogniti ve-functional terms: “ grammaticalizatio n is not an independent process and cannot in itself serve as a principle explaining form changes (...) grammaticalization is an observational term covering a cluster of phenomena which are somehow determined by, and should thus be explained in term s of cognitive-functional factors ” (2001: 203). This implies that, in addition to the form al processes which are identified in grammaticalization (see 2.3.2.4.1 and 2.3.2.4.2 be low), Nuyts relies on conceptual fact ors determined by the human mind. A sim ilar idea is found in Kuteva ’ s (2004) work on auxiliation. In her own words: “ the way in which interlocutors arrive at an innovative grammatical use of a complex verb expression in a specific discourse context involves cognitive-semantic ‘ reasoning ’” (2004: 178). However, Kuteva does not enter the debate of the independence of grammaticalization; she just relates it to the cognitive factors of relevance theory (cf., for instance, Klinge 1993; Nicolle 1998). Chapter 2. Grammaticalization, modali ty and impersonality 28 pease > PDE pea, also known as back-formation, cf. Campbell 1998). Theref ore, bidirectionality in reanalysis means that the restructuring of an element may operate in both directions: from the stem outwards and fr om out towards the stem. The fourth difference between grammaticalization and reanalysis concer ns ambiguity (cf. Table 2.1). Although not much attention is paid to this criterion in Haspelmath (1998), we can conclude that grammaticalization does not produce ambiguous structures, while reana lysis does. However, this does not seem to be a definitional criterion, since ambiguity is not a necessary characteristic of reanalysis (cf. for instance, (2.4) above). The last difference between the two type s of changes, according to Haspelmath (1998: 327), concerns their origin . Grammaticalization is said to be due to language us e, and reanalysis due to language acquisition. That grammaticalization is due to language use is clear, it is a gradual process carried out by the speakers of several generations. 11 The idea of reanalysis being originated in language acquisition needs further comment. Let us recall the above-mentioned example of biki ni . It has been said that it became evident that this word was reanalysed the first time th at the noun monokini appeared in language. Relying on language acquisition as the origin of reanalysis woul d imply that the word bikini was reanalysed by a generation of speakers as they learned it and did not relate the word to the Bikini Islands. What is definitely sure is that the first generation of speakers who learned the word monokini considered the original bikini as a derived word. Summing up, grammaticalization and reanalysis have be en proved to be different phenomena. This explains why there exists grammaticalization without reanalysis and reanalysis without grammaticalization. However, their differences do not set them so apart that they do not overlap. Radical views such as Heine et al. ’ s (1991: 217) saying that “ grammaticalization and reanalysis seem to be inseparable twins, ” or Haspelmath ’ s (1998) argumentation that “ pure ” grammaticalization “ should be explaine d within the framework of a theory of grammaticalization, without reference to reanalysis ” will not be followed in this piece of research. My view of grammaticalization and reanalysis, therefore, will 11 In fact, language use has been claimed to be the only origin of la nguage change. Paul (1920 [1880]: 32) stated that “ Die eigentliche Ursache f ü r die Ver ä nderung des Usus ist nichts anderes als die gew ö hnliche Sprecht ä tigkeit ” ( ‘ the real cause of the change of (linguistic) conventions is nothing other than ordinary language us e, ’ translation by Haspelmath 1999: 1066). Chapter 2. Grammaticalization, modali ty and impersonality 29 be that posited by Hopper and Traugott (2003: 58-63), that is, that they are close types of language cha nge that may or may not be related. In the previous paragraphs I have examined the role of analogy and reanalysis as processes which may intervene in grammaticalization. There is still one feature of grammaticalization which deserves specia l attention: the claimed unidirectional nature of this linguistic change. The next section briefly explains this controversial issue. 2.1.3.3. Grammaticalization: a unidirectional phenomenon? The unidirectionality of grammaticalization is one of the most controversial issues of this mechanism of language change, as prove s the fact that scholars do not cease to publish articles on examples and counterexamples (cf., among t he most recent ones, Brinton 2004, Rosenbach 2004, Tsangalidis 2004, Ziegeler 2004). Some authors firmly consider that grammaticalization is uni directional (e.g. Haspelmath 1998, 1999, 2004), others acknowledge that there may be exceptions (e.g. Traugott 2001, Burridge 1998, Hopper and Traugott 2003), and finally some authors consider that grammaticalization is bidirectional (e.g. Ramat 1992, Campbell 2001, Nuyts 2001). Traugott (2001) and Haspelmath (2004) offer a comprehensive review of most of the claimed counterexamples to the unidirectionality criterion. Haspelmath (2004) reaches the conclusion that out of the ca. 100 examples quoted in the literature, only eight are illustrative of a certain antigrammaticalization (his terminology), among which we find the well- known case of the English and Mainland Scandinavian genitive suffix – s , which gave way to the clitic ‘ s (cf., for instance, Newmeyer 1998, and compare with Traugott 2001: 6, which considers it a putative counterexample to unidirectionality). Many of the alleged counterexamples to the unidirectionality criterion are identified as instances of different phe nomena by Haspelmath (2004). Thus, the much cited example of the noun ism deriving from the derivative suffix – ism in words such as cubism or dadaism (cf., for instance, Ramat 1992: 549) is indeed an example of “ delocutive word-formation ” (cf. Haspelmath 2004: 29-30). 12 This 12 According to Haspelmath (2004: 29), “ [a] delocutive lexeme is one that was derived by some regular word-formation process from an other lexeme whose use in speech somehow determines the meaning of the derived lexeme. ” Chapter 2. Grammaticalization, modali ty and impersonality 30 explanation also seems to hold for bus < Latin omnibus (ablative pl ural of omnis / omne ), which Hopper and Traugott (2003: 58) cite. As for the alleged de grammaticalization of adver bs such as up into the homomorphic ver bs (cf. Newmeyer 1998: 273), Haspelmath believes that it is nothing more than a case of conversion, very much like that of the verb bottle from the homomorphic noun. This phenomenon is also referred to as lexicalization (cf. Heine et al . 1991: 50). Another case of degrammaticalization postulated in the literature is the Spanish and Italian derivational suffix – ante / - ente , which derives from the inflectional ending of the Latin pres ent participle. Haspelmath explains this as the loss of the Latin inflectional category which leaves some traces in items which ar e productive not inflectionally, but derivationally. Another well-known case of degrammaticalization is English dare , whose lexical facet far overrides its auxiliary use (cf. Beths 1999). This phenomenon, which has been called “ amphibian nature, ” “ twin role, ” split or divergence (cf. Abraham 1990; Burridge 1998: 28; Hopper and Traugott 2003: 118-122) is better explained as retraction, because it implies that a given form prefers to reinforce its original lexical component rather than under going a full grammaticalization process (cf. Haspelmath 2004: 33-35). Haspelmath bases himself on Traugott ’ s (2001: 9) explanation as for this case: “ the earlier main verb use was marginalized in the early periods and then the grammaticalized one was marginalized in turn and then lost in later periods. ” In other words, lexical and grammatical dare appear to have coexisted since early periods and while in some periods the grammaticalized element predominated over the lexical one, nowadays the grammaticalized dare is marginalized in favour of the lexical dare . We will see below that PDE need seems to be another case of retraction (cf. also Taeymans 2004a). The last refutation to an alleged counterexample to the unidirectionality of grammaticalization concerns examples such as the development of the French particle ti , as in examples (2.5a) and (2.5b): (2.5a) Votre p è re NP part V – il PRO ? (standard French) Æ ‘ Does your father leave? ’ (2.5b) Votre p è re NP par[ Ø ] V ti PTCL ? (colloquial French) ‘ Does your father leave? ’ In colloquial French, the third person singula r pronoun il , a grammatical element, grammaticalizes into an interrogative particle, ti , due to its usual col location after Chapter 2. Grammaticalization, modali ty and impersonality 31 verbal forms ending in – t . Neither the personal pronoun is more grammatical than the interrogative particle or vice versa. This is, then, a case of regrammaticalization, a term coined by Gree nberg (1991). An attempt to account for the development of new grammatical functions of already grammaticalized elements has been done by Lass (1990), who pr oposes the term exaptation, which he borrows from biology (cf. also Lass 1997: 316-324). Exa p tation would refer to those cases in which a grammatical element which had become marginal (what he labels “ junk ” ) acquires a new grammatical value and becomes productive again (cf. Wright 2004 as an example of exaptation of Englis h plural be in AAVE). The difference between Lass ’ s exaptation and Greenberg ’ s regrammaticalization is that the latter does not require the marginalization of the early grammatical element (cf. Traugott 2001). Therefore, most of the alleged counterexamples to the unidirectionality of grammaticalization have been accounted for as examples of different phenomena (cf. Traugott 2001, Haspelmath 2004). The few examples of degrammaticalization (antigrammaticalization, in Haspelmath ’ s terms) are so scarce and so rare that the unidirectionality of grammaticalization seems to be close to universal. 2.1.3.4. English Modals : a paradigmatic case of grammaticalization Auxiliary verbs are a paradigmatic example of grammaticalization and have recently been subject of numerous studies (to cite just a few, Heine 1993, Warner 1993, Bybee et al . 1994 and Kuteva 2004). Accordi ng to Heine ( 1993: 70), an auxiliary “ is a linguistic item covering some range of uses along the Verb-to- TAM chain. ” Such a chain refers to the grammaticalization chain mentioned above, which, according to Heine, consists of four processes: desemanticization, decategorialization, cliticization and erosion. As Kuteva (2004) says, “ depending on [ … ] their location along that chain, they will be more grammatical or less grammatical. ” Among these, Warner (1993) considers the PDE central modals (i.e. can, could, may, might, shall, should, will, would and must , according to Quirk et al . 1985: 137) prototypical auxiliaries, stemming from Rosch ’ s (1977, 1978) prototype theory or theor y of categorization. As is well-known, this theory has to do with the human mental categorization of the world. We tend to have in mind the simplest element of a class, and that element becomes a prototype to which we refer in order to check if, and to what extent, a particular item enters Chapter 2. Grammaticalization, modali ty and impersonality 32 that category. Therefore, Warner implies that PDE modals are the prototype to which we must refer in order to categorize auxiliaries. The consideration of auxiliaries as a class is not free from controversy and, as Kuteva (2004: 5) mentions, up to eight different proposals have bee n postulated in the literature. A widely-acknowledged set of criteria to identify auxiliaries is that proposed by Quirk et al . (1985: 137), who compare auxiliaries to main verbs and, among auxiliaries, modals are further described according to four additional criteria, as seen in Table 2.2: AUXILIARY CRITERIA (Op = operator) AUXILIARY MAIN VERB (a) Op in negation He cannot go *He hopes not to go (cf. note i) (b) Negative contraction can ’ t * hopen ’ t (c) Op in inversion Can we go? * Hope we to go? (d) Emphatic positive *Yes, I D Ò can come. Yes, I D Ò hope to come. (e) Op in reduced clause I can come if you can . *I hope to come if you hope. (f) Position of adverb We can always go early We always hope to go early. (g) Postposition of quantifi er They can all come ?They hope all to come. (h) Independence of subject Ann can do it. ~ It can be done by Ann. He hopes to do it. *It hopes to be done by him. MODAL AUXILIARY CRITERIA MODAL AUXILIARY MAIN VERB (a) Bare infinitive I can go .* I hope go /I do go (cf. note ii) (b) No infinitive forms * to can, *canning, *canned to hope, hoping, hoped, to do doing, did (c) No – s form *She cans come. She hopes to come/ She does come (cf. note ii). (d) Abnormal time reference You could leave this evening. [no past tim e] You hoped to leave this evening/ You did leave this evening. [past time] NOTE i [original] : He hopes not to go is acceptable in the sense ‘ He hopes that he will not go; ’ but this is then a case of the negation of to go , not of hopes . Note ii [mine]: I do go or She does come are correct in emphatic contexts such as that exemplified in Yes, I do hope to come above, that is, it is an alternative construction t o the unmarked one I go (cf. obligatoriness of I can go ). Table 2.2: Formal criteria for auxiliary verbs and modals (adapted from Quirk et al. 1985: 137). As the examples in the second and third column of Table 2.2 show, modals constitute a class of words different from main verbs and from non-modal auxiliary verbs. 13 They are the result of subsequent ch anges of an OE class of 13 From a cognitive-pragmatic point of view, the current grammatical nature of modal auxiliaries is born out of the speakers ’ need to code the “ conceptuall y elementary and systematically recurrent ” categories (i.e. root and epistemic categories) into li nguistic structure, and it is intensified by the fact that bo th root and epistem ic qualifications are closed classes (Nuyts 2001: 270). Chapter 2. Grammaticalization, modali ty and impersonality 33 verbs known as pre-modals 14 in the history of English. The OE pre-modals cunnan, magan, *sculan, *motan, willan, *durran and þ urfan are the ancestors of the PDE modals can, may, shall, must, will and dare (as will be repeatedly mentioned, þ urfan dies out earlier). 15 The gra dual process by which pre-modals develop into PDE m odals implies grammaticalization (see section 2.1.3, for a description of this type of language cha nge), and has been the subject of outstanding pieces of research (cf., for example, Lightfoot 1979; Plank 1984; Heine 1993; Warner 1993; Krug 2000). It is a widespread belief among scholars that the grammaticalization of l inguistic items is gradual (cf., among others, Plank 1984, Giv ó n 1984, Krug 2000, Hopper and Traugott 2003), which contrasts sharply with the improvised linguistic accidents proposed by scholars such as Lightfoot (1979) or “ hopping rules ” (cf. Pullum and Wilson 1977). To cut a long story short, we may say that at the least grammaticalized end of the grammaticalization c h ain of modals stand the OE pre-modals, and at the most grammaticalized end stand PDE modals such as should or may . The characteristics of the OE pre-modals differ from one verb to another syntactically and semantically. Thus, it is worth-mentioni ng that some of these verbs show auxiliary-like characteristics as early as in Old English (for a detailed explanation see section 3.2; cf. also Goossens 1987; Denison 1990a; Warner 1993; Beths 1999). Among the syntactic characteristics which relate these words to auxiliaries is their possibility to occur in elliptical and impersona l constructions. There are also semantic features which relate OE pre-modals to PDE modals. Warner (1993) recognizes the deontic (comprised in which I will call root modality) modal meaning as the main meaning conveye d by these verbs. In the course of the ME period, these verbs gain weight as an auxiliary group due to the development or intensification of formal fe atures, such as the fact that they cease to take nominal direct objects. In addition, it has often been claimed that the disappearance of the inflectional subjunctive left a gap which was soon filled by the modals, which in combination with an infinitive became periphrastic verb phrases expressing modality (cf., for instance, Fischer 2002). In this sense, (pre)modals were reanalysed as analytic mood markers at the same time as the following infinitive was reanalysed as the main verb of the sentence (cf. section 14 The term used to refer to this class of OE verbs is not free from controversy. I use the term pre-modal without inverted commas for the reasons adduced in section 3.2. 15 Since the status of the pr e-modals throughout history will be described in detail in the first half of chapters 3, 4 and 5, this section only gives a cursory look at their evolution and pays special attention to the gramma ticalization mechanisms involved in the overall developm ent. Chapter 2. Grammaticalization, modali ty and impersonality 34 2.1.3.2) for an explanati on of reanalysis in combination with grammaticalization). Further developments took place in the eModE period, when modals began to occur in inversion and in negative contexts without do -s upport. In addition, semantics also plays a role in this period, since epistemic meanings arise between the ME and eModE period, probably due to subjectification (cf., for example, Traugott 1989; Goossens 2000; Pelyv á s 2000; Hopper and Traugott 2003). Interestingly enough, scholars such as Nuyts (2001: 176-178) and Pelyv á s (2003), consider that epistemic modals are closer to the grammatical (i.e. auxiliary) end of the chain than root modals. Although most of the members of the class of OE pre-modals have survived into Present-Day English, the modal class has been subject of some losses and some gains throughout history. Among the losses, we may highlight that in Middle English some preterite-present verbs dropped out of the language probably due to their synonymy with other verbs, such as, for instance, tharf , which was replaced by neden ‘ to need, ’ or cunnan , which was replaced by witan ‘ to know ’ (see Plank 1984: 312). 16 In early Modern English mote drops out of t he language and leaves must unpaired. However, the class may also grow larger and in this period need and ought enter the class, so that the modals in the central part of the eModE period are can / couth, dare / durst, may / might, shall / should, will /would, must, need and ought (cf., for instance, G ö rlach 1991: 114). As mentioned above, the PDE class of central modals is more reduc ed than this one, which implies that it has not ceased to undergo losses. For example, durst is out of the paradigm, whereas need and ought are considered marginal modals. At the same time that the central class of modals appears to get more reduced, other verbs acquire new functions and meanings and get closer to the modals althoug h they do not fulfil all their characteristics, such as be going t o or have to (cf., for instance Heine 1993, Krug 2000). On some occasions, the newly bor n emerging modals, such as have to , push out old, traditional modals such as must (cf., for instance, Smith 2003). These gains in the modal group are in tune with Croft ’ s (1990) idea that grammaticalization is in constant operation. The fact that not all modal auxiliaries developed at the same time and in the same way implies that one must analyse them gradually and considering 16 This avoidance of “ homonymic clash, ” as labelled by Hopper and Traugott (2003: 102), follows the perspective of “ one meaning-one form, ” postulated, am ong others, by Geeraerts (1986). Chapter 2. Grammaticalization, modali ty and impersonality 35 different factors as is suggested by Mortelmans (2003). For this reason, diachronic research seems to be the ideal perspective to study this phenomenon. At the same time, synchronic data, such as synchronic variation, may reveal facts about the origin and reasons f or the grammaticalization. This double perspective of grammaticalization is summarized under the label panchronic or metachronic (see Heine 1993: 76). These terms refer to anything which is neither exclusively diachronic nor synchronic, but which wor ks in both dimensions. In this piece of research I will analyse each single period of the history of English synchronically, and at the end I will provide the diachronic development of each of my verbs. The intention, therefore, is to adopt a pa nchronic or metachronic perspective, which can help me obtain an explanator y account of the evolution of the semantic predecessors of need . 2.2 Present-Day English need and need to : an insight into modality Need and need to constitute a complex phenom enon: apart from the pair dare / dare to , there appears to be no other pair of English verbs which exhibits twofold morphosyntactic features, namely those typical of auxiliary verbs and those of lexical verbs. Due to this double nature, a controversial issue arises: should we treat need and need to separately, as two different ve rbs, or as one verb with two different syntactic realizations? With the aim of answering this question, in the remainder of this section I will examine the morphosyntactic and semantic features of need and need to . 2.2.1. Morphosyntactic features From a morphological perspective, need and need to exhibit important differences which ha ve been variously analysed by scholars. The next two sub- sections review the most widely acknowledged accounts of the classification of these verbs, both traditional and moder n ones. 2.2.1.1. Traditional considerations Traditionally, the PDE verb need is said to be a (marginal) modal verb which has a homomorphic non-modal counterpart 17 (cf. Huddleston 1984 and Quir k et al . 17 This is a simplification of a three-term classification which distinguishes modal auxiliary need ( He need not sign up ), non-modal full verb need ( I need a ticket ) and catenativ e verb ( He needs Chapter 2. Grammaticalization, modali ty and impersonality 36 1985). This idea is supported by the morphological and syntactic differences between both: while non-modal need behaves like any regular transitive verb (which takes both nominal and sentential complements), modal need complies with all the requisites of a modal auxiliary, with the s o-called NICE properties (cf. Huddleston 1980). Let us compare, for instance, the following pairs of examples: (2.6a) Your needn’t do that ( Æ negation) (2.6b) You don’t need to do that (2.7a) Need I sign in again? ( Æ inversion) (2.7b) Do I need to sign in again? (2.8a) John needn’t do the exercises and neither need Susan ( Æ ‘ code ’ ) (2.8b) John doesn’t need to do the exercises and neither does Susan . The (a) examples show instances of modal need , while the (b) examples show the non-modal counterparts. As is obvious, the latter do not exhibit any of the NICE properties (negation, inversion or code, in the case of need ). 18 Another auxiliary- like syntactic characteristic of need is its possibility to occur in tag-questions ( She needn ’ t do it, need she? ), as mentioned by Jacobsson (1974: 56). Apart from these strictly syntactic features, modal need also shares morphological characteristics with the other modal auxiliaries, as Coates (1983: 4, 50) notes: - No – s form for third person singular (* he needs not do it ) o VS. he needs to do it - No occurre nce with another modal (* he will need not do it ) o VS. he will need to do it . - No past forms (* he needed not do it ) o VS. he needed to do it In connection with the last morphological feature, it must be pointed out that Quirk et al. (1985: § 3.40) mention another characteristic of modal auxiliaries: their abnormal time reference. According to t hem, the so-called modal past forms – also called distal forms (cf. Sweetser 1990: 62 ff.) – namely might, would, could and shou ld , can be used to express present time with a nuance of r emoteness (as to write a paper ), as proposed by, for example, Jacobsson (1974: 56). The sim plification to modal and non-m odal leaving aside the construction with a nominal object focuses on the constructions with a following infinitive, which is our main concern here. 18 Not all authors agree on the extent to which need exhibits the NICE properties. Palmer (1979: 4, 127) points out that need does not have the properties of code and emphatic affirmation. However, as can be seen in example (2.8a), code is possible with need . Chapter 2. Grammaticalization, modali ty and impersonality 37 in he might win the race ). In addition, some of the s o -called present forms (i.e. may, will, can, shall or must ) may refer to the past when combined with a perfect infinitive (e.g. must , as in he must have come yesterday ). As for need , it can be inflected for the past tense only when it is a non-modal verb. See, for instance: (2.9) I need to go (2.10) I needed to go (2.11) I didn ’ t need to go (=I didn ’ t have to go). In contrast, modal need does not have a past tense form. It may nevertheless express past time when used i n indirect speech, as in She told him he needn ’ t come , or when in combination with the perfect infinitive, yielding need not have plus past participle (cf. Jacobsson 1974: 56). Needn ’ t have commonly implies that it was not necessary that the action expressed by the proposition took place. The proposition in (2.12) is positive ( ‘ I have gone ’ ), but the modality, needn ’ t (=absence of obligation, e xemption) is negative. Thus, (2.12) means ‘ it was not necessary for me to go, but I went: ’ 19 (2.12) I needn ’ t have gone. Therefore, need seems to qualify as a modal auxiliary, since it complies with nearly all the requisites. There are only two exceptions; the first is the expression of emphatic affirmation, as already mentioned. The second concerns morphology. Modal verbs do not have non-finite forms, i.e. forms such as * to can or * musting are ungrammatical. Nonetheless, the form to need is possible. Although it is not possible to determine whether this form belongs to modal need or non-modal need to , scholars do not address this issue, and omit any reference to this non-finite form. One NICE feature which is absent in modal need is the property of emphatic affirmation (cf. footnote 18 this chapter), as exemplified in (2.13), from Coates (1983: 4): (2.13) Ann COULD solve the problem. 19 As mentioned, the most general difference between sentence (2.11) and sentence (2.12) is that (2.11) implies that the action was not carried out, while (2.12) im plies that the action was actually carried out. The difference between both sentences is neutralised, however, when in (2.11) the stress falls on need . This characteristic of oral speech suggests that the action was accomplished (cf. Westney 1995: 141). Chapter 2. Grammaticalization, modali ty and impersonality 44 and dare (to) oscillate between the central modals (e.g. will, may ) and this new emerging class (Krug 2000: 239). Going to, have to, want to and got to have enough characteristics in common to be grouped together in “ a subcat egory within the higher-level class of modal verbs ” (Krug 2000: 214). The reasons that Krug (2000: 215-217) adduces to justify the introduction of this new sub-cat egory, i.e. emerging modals, are the following: 1. The traditional classification of auxiliaries (quasi-auxiliaries, semi- modals, secondary auxiliaries, etc.) does not allow for the identification of an evolving class of verbs. 2. According to Kr ug (2000: 214), “ It would not seem helpful t o exclude all verbs taking infinitival to complements from auxiliarihood simply because they do not share the syntactic pr operties of the central modals. ” Of course, this also holds for the marginal modals need (to) , ought to and dare (to) . 3. The phonological reductions observed in goi ng to, have to, want t o and got to are “ perfectly regular. ” E.g. want to >wanna; got to > gotta . It must be noted that the sa me reduction is recorded with ought to and need to ( need to > neeta/needa (cf. Krug 2000: 285-286, note 60)). 24 4. The fact that highly grammaticalized paradigms are usually smaller than less grammaticalized ones (cf. Lehmann 1995 [1982]: 132 ff.) gives Krug (2000: 217) enough evidence to consider emerging modals more grammaticalized than other items taking to-infinitive complements, because “ not all can serve as hosts to cliticized to (e.g. * attemma from attempt to ( … )), ” and, therefore, they form a restricted group. 5. The existence of a modal category makes possible the emergence of “ a new modal layer. ” From Krug ’ s (2000) point of view, then, there are enough reasons to consider going to, have to, wan to and got to as the prototypical members of the s ub- category of emerging modals. This wa y he unifies the apparently m essy characterization of these verbs in the literature and pays attention to the dynamic nature of language. As for need ( to), ought to and dare (to) , Krug claims that they oscillate between the central modal group and the prototypical emerging modals. With the help of his gravitation m odel, he proves that need (to) is the 24 The existence of needa/neeta as a phonological r eduction of need to , parallel to widel y- acknowledged contractions such as wanna or gotta , has also been noted by Gramley and P ä tzuld (1992: 161) and Westney (1995: 33). It has, however, been rejected by Pullum (1997: 82), who attributes such pronunciation to “ rapid or very casual speech. ” Chapter 2. Grammaticalization, modali ty and impersonality 45 closest to the new emerging class and, in consequence, the farthest from the central class of modals (2000: 238-239). As can be gathered from the preceding paragraphs, need t o is considered to be a non-modal verb (Huddleston 1984, Quirk et al. 1985), a semi-modal (Leech 2003, Smith 2003) or a marginal modal very close to the group of emerging modals (Krug 2000). I object to the analysis of authors such as Huddleston (1984) and Quirk et al. (1985) on the gr ounds that they ignore the features that all emerging modals have in common and which bring them together, namely, their regular phonological reduction and the place they occupy in the grammaticalization scale, as opposed to verb s such as attempt to , fo r example. On the contrary, they analyse these verbs as radically different items. Huddleston (1984), for example, resorts to three different labels to refer to these verbs: ought to is closer to the modal class than an y of the other verbs, and is said to be in the “ periphery of the class ” (1984: 165); want to and have to are considered “ catenative verbs ” (1984: 142, 166); and need to is said to be a main verb, as already mentioned (1984: 165). In turn, Quirk et al. (1985) coincide with Huddleston ’ s classification of ought to as a marginal modal (1985: § 3.40). As for the other verbs, have to is a semi-auxiliary (1985: § 3.40); have got to is a modal idiom (1985: § 3.40); want to is explicitly called non-catenative (1985: § 3.49) and need to is a lexical verb (1985: § 3.41). None of these authors make reference to got to without have . In other words, the view adopted in these grammars is too general, while Krug ’ s (2000) cla ssification seems more appropriate to account for the syntactic, morphological and semantic characteristics of the verbs are concerned. Therefore, for the purposes of this dissertation I will follow Krug ’ s (2000) view about the existence of a “ new ” category of emerging modals, to which PDE need to is close. 2.2.2 Se mantic features Once I have described the morphological and syntactic characteristics of need and need to (cf. 2.2.1), in this section I will try to explain their semantic features. Since the limits between one category and anot her are not so clear in semantics as they are in morphology and syntax, I will first describe the semantic framework used for the classification of my verbs. For this purpose, this section consists of three sub-sections. 2.2.2.1 is devoted to the definition of the term Chapter 2. Grammaticalization, modali ty and impersonality 46 modality. The need for an insight into the category of modality arises from the fact that, as seen in the previous sections, PDE need ( to) is considered to be close to the group of emerging English modals (cf. Krug 2000). In 2.2.2.2 the different types of modality are examined, with a discussion on the different theoretical approaches to this linguistic category. Finally, 2.2.2.3 analyses the semantic features of PDE need and need t o from the point of view of modality. 2.2.2.1. The concept of modality Though much has been written on modality, linguists do not seem to reach an agreement as to how to define this category. As already mentioned, modality is a semantic category, i.e. it is related to meaning in the same sense that syntax is related to structure, or morphology to word forms. In this way, modality is parallel to concepts such as time or sex, since the three of them belong to the extra-linguistic reality. Time refers to th e temporal axis of the world, and it is grammaticalized in language as tense. Thus, we distinguish between make (present tense) and made (past tense). Likewise, sex is an extra-linguistic category that divides t he world in male and female. The corresponding linguistic category may be gender; hence, we fi nd masculine and feminine nouns and adjectives in some languages, such as Spanish, for e xample, to mark this an other distinctions. Of course, there is no one-to-one correspondence between time and tense on the one hand, and sex and ge nder on the other hand. Tense and gender are grammatical categories which, as such, may have no parallel in the extra- linguistic reality. Thus, for instance, the construction If I went there, I would see her contains a past tense form of the verb go , went , which does not refer to past time. At the same time, the Spanish word mesa ( ‘ table ’ ) is grammatically feminine, although it does not refer to any female being in the extra-linguistic world. Modality also refers to t he extra-linguistic field. It is the term used to name the speaker ’ s judgement of his proposition, according to its truth or falsehood, its probability, etc. This is the most widely acknowledged view (cf. Halliday 1970, 1985; Jim é nez Juli á 1989). In Halliday ’ s (1970: 335) words: a form of participation by the speaker in the speech event. Through modality, the speaker associates with the thesis an indication of its status and validity in his own judgement; he intrudes and takes up a position. Chapter 2. Grammaticalization, modali ty and impersonality 47 He later expresses the same idea as (1985: 75): “ Modality means the speaker ’ s judgement of the probability, or the obligations, involved in what he is saying. A proposition may become arguable by being presented as likely or unlikely, desirable or undesirable – in other wor ds, its relevance specified in modal terms. ” By assuming this idea, these authors base their definition of modality exclusively on epistemic modality, i.e. that related to the mental world, “ the one which most clearly is relevant to normal language ” (Coates 1983: 18). A clear definition of epistemic modality is that proposed by Nuyts (2001: 21): (the linguistic expression of) an evaluation of the chances that a certain hypothetical states of affairs under consideration (or some aspect of it) will occur, is occurring, or has occurred in a possible world which serves as the universe of interpretation for the evaluation process, and which, in the default case, is the real world. However, epistemic modality developed relatively late in the history of all languages (cf. Shepherd 1982, Bybee and Pagliuca 1985, or Bybee et al . 1994 as studies devoted not just to English). The items which came to express epistemic modality existed in the languages before this meaning was grammaticalized. In English, the linguistic items which have finally expressed epistemic modality are the so-called modal verbs, 25 a class of verbs characterized by a preterite-present morphology and by an auxiliary-like syntax. In other words, most of these items have common morphol ogical and syntactic features which go back to Old English times (with the exception of will ). Therefore, the items which came to express epistemic modality in English already formed a class of their own in syntactic and morphological terms. The question to be answered, therefore, would be whether these verbs were also semantically similar before the grammaticalization of epistemic modality or, on the contrary, they became semantically close when epistemic modality was grammaticalized. It is commonly accepted that this class of Eng lish verbs used to have a common kind of meaning from which epistemic modality derived. This common kind of meaning is called ‘ root ’ and it is related to the real world, instead of the mental 25 There are other non-grammatical way s of expressing epistemic modality, such as adverbs (e.g. possibly ), adjectives (e.g. possible ), or mental state predicates (e.g. I think that … ), as mentioned, for instance, by Nuyts (2001: 29). Nevertheless, the p olysemy of modal auxiliaries (which express both deontic and epistemic modality) is attested in ma ny West European languages and also in many t ypologically completely different ones (cf. Steele 1975, as mentioned by Nuyts 2001: 171). Chapter 2. Grammaticalization, modali ty and impersonality 48 world present in epistemic modality (cf. Sweetser 1990; Traugott 1991, 1992; Bybee et al . 1994, among others). As a consequence, the concept of modality in English is generally broadened to include the root meanings of the modal verbs which later derived into epistemic modals. 26 Thus, not only the senses of like lihood or the degree of truth of a proposition are included in the concept of modality, but also meanings such as obligation, permission and the like, all implying the power which is being exerted on the doer of the action expressed by the infinitive following the modals. Modality, therefore, implies the reflection of the speaker ’ s attitudes and opinions through his speech (cf. Palmer 1986: 16). This reflection must take place by means of linguistic items. In othe r words, first the speaker decides to include his opinions in his speech (modality), and secondly, modality needs to be expressed by means of language (i.e. modality is conveyed by language, which is equated with the grammaticalization of modality), as can be seen in the following figure: Speaker ’ s attitudes Æ Modality Æ Grammaticalization and opinions of modality Figure 2.2: Modality: from the extra-linguistic world to gram mar (adapted from Palmer 1986). The expression of modality may be realized by several linguistic items. In English different parts of speech may carry modal meaning (cf. Perkins 1983), namely adverbs (e.g. certainly ), verbs (e.g. can, suppose, think ) or adjectives (e.g. possible ; cf. Huddleston 1984: 166). Adverbs and adjectives are lexical words, i.e., words carrying full meaning. They are grammatically independent, since they do not require the support of any ot her part of speech. The English modal verbs, however, are a kind of intermediate stage between fully lexical items and purely grammatical items (like conj unctions or prepositions). That is, they have a stronger referential meaning th an, for example, the preposition of , 26 I must specify that this broadening of the concept of modality takes place mainly in English linguistics, because in other languages, such as Spanish, the concept of modality is restricted to the so-called epistemic meanings, expressed by inflectional mood (cf. Jim é nez Juli á 1989 for a detailed explanation of the origin of modality – modus clausal in his own term inology- in Spanish). Chapter 2. Grammaticalization, modali ty and impersonality 49 but, at the same time, they share all the features of auxiliaries (e.g. primary auxiliary do , cf. Quirk et al. 1985: § 3.40). There is another way of conveyi ng modality, namely inflectional mood. This device is not found in Englis h but is common in Romance langua ges such as Spanish. The main verb in the sentence is i nflected to express som e of the modal meanings. For instance, if the speaker believes that at the time he is speaking it is five o ’ clock, he may resort to any of the three devices we ha ve just mentioned (the instance of inflectional mood is provi ded in Spanish for obvious reasons): (2.32a) It is possibly the case that it is five o ’ clock . (modal adverb) (2.32b) It must be five o ’ clock . (modal verb) (2.32c) Quiz á sean las cinco. (present subjunctive of the verb ser ‘ to be ’ ) Thus, we could draw another figure to represent the last stage in Figure 2.2, grammaticalization of modality, which may be linguistically expressed in, at least, these three ways, ranging from less to more grammaticalized means: Adjectives and adverbs Æ Modal verbs Æ Inflectional mood - grammaticalized + grammaticalized Figure 2.3 : Grammaticalization of modality (based on Palmer 1986: 4). Once we have described the different formal realizations of modality, we should turn to the analysis of the possible modal meanings. As has been suggested earlier, the classification I will follow here is that of the root/epistemic dichotomy. However, ot her classifications are available. In the next section, I will comment on different theories for the classification of modality, and discuss the framework within which I am going to analyse my examples. 2.2.2.2 Types of modality: root and epistemic The myriad of notions associated with modality allow for a wide variety of classifications in the literature. 27 Some of them divide modality into two sub- groups, while others distinguish three or four sub-groups. As note d by Siemund (1997: 281), Mindt (1995) proposed up to 17 different types of modality. 27 Some of the notions proposed are: subjectivity, non-assertion, non-factivity (Palmer 1986: 4), permission, obligation, volition, prediction (Quirk et al. 1985: § 4.49), and possibility and necessity, which are co mmon in the interpretations of both works. Chapter 2. Grammaticalization, modali ty and impersonality 50 As for the division into two types of modality, I will highlight three major approaches . The three of them are based on the human control of events. In other words, they assume that there may exist some kind of intrinsic control over the event expressed in the proposition or, on the contrary, there may be no control, but just some sort of human judgement. According to these criteria, modality types may be intrinsic vs. extrinsic (cf. Quirk et al. 1985), de ontic vs. epistemic (Huddleston 1984) or root vs. epistemic (Sweetser 1990). The difference between these three approaches is not just one of terminology, but of perspective in general. While Sweetser resorts to the historical evolution of modality from root, i.e. basic, meanings, to epistemic ones, the major grammars by Huddleston (1984) and Quirk et al . (1985) provide a broa d list-like classification of the English modals. Other specific works on modal ity reveal the need for a third type of modality where meanings such as volition or ability would fall in. This is the case of Palmer (1979, 1986, 2003), who, ba sed on Lyons (1977), acknowledges three types of modality: deontic, epistemic and dynamic (the latter term is taken from modal logic, cf. von Wright 1951). This threefold division is also defended by authors such as Goossens (1985, 1987), who uses the term ‘ facultative ’ to refer to the third type of modality, Hengeveld (1988), who prefers the term ‘ inherent, ’ Vihla (1999), and Warner (1993: 14-17). In addition to these types of modality, other authors include s ubordinating moods as a subtype of modality. This idea, supported by Bybee et al . (1994), is based on a typological analysis of a large number of languages of the world. Since modality is the grammaticalization of subjective attitudes and opinions, and is expressed by means of different devices, these scholars draw s ome diagrams which show the development of the different modal meanings which stem from three basic notions, namely, obligation, desire and ability (Bybee et al . 1994: 240, § 6.13). Although thi s study is very interesting from a typological point of view, it does not seem appr opriate, however, to adopt this classification for the study of the English language intended here. Moreover, if we exclude the subordinating moods, their analysis is parallel to the threefold pattern proposed by Lyons (1977) and Palmer (1979, 1986, 2003), because it identifies three types of modality, namely speaker-oriented, epistemic modality and ‘ agent-oriented, ’ Chapter 2. Grammaticalization, modali ty and impersonality 51 which includes those notions which are di fficult to classify, such as ability and volition. 28 At first sight it could seem that the more types of modality, the more exhaustive an analysis would be. This was my view i n an earlier version of this piece or research, when I resorted to Palmer ’ s threefold classification of modality (deontic, epistemic and dynamic) in search for a comprehensive classification which did not leave a ny question unanswered. However, a deeper look into the possible classifications revealed that the scientific quality of a fine-grained theory is not directly proportionate to the number of modality types distinguished, but to the criteria used to delimit each type. Palmer (1979, 1986, 2003) follows the tradition of modal logic presented by von Wright (1951) and Lyons (1977). Stemming from the basic meanings of possibility and necessity, he describes the three types of modality, epistemic, deontic and dynamic, which are to be recognized in many la nguages, the first being “ solely concerned with the speaker ’ s attitude to the status of the proposition, ” the other two being related “ directly to the potentiality of the event signalled by the proposition, ” deontic modality is concerned with external circumstances and dynamic modality with internal ones (Palmer 2003: 7). The logical relation between possibility and necessity implies that if X is not possible, then not-X is necessary and, consequently if X is not necessary, then not-X is possible. That is, if raining is not possible, not raining is necessary. Actually, the relations between possibility and necessity are acknowledged to play a fundamental role in linguistics, as can be observed in the development of German modal d ü rfen ‘ to be allowed. ’ This verb originated as a pol arised necessity modal conveying negative necessity (namely ‘ need not ’ ). Later it acquired a negative possibility meaning (namely ‘ cannot, ’ fr om an original necessity not to, i.e. ‘ must not ’ ), and finally it developed the current positive possibility meaning ‘ may, to be allowed ’ (cf. van der Auwera and Plungian 1998: 99). 28 A more recent typological study based on Bybee et al. ’ s (1994) work is van der Auwera and Plungian (1998). They subdivide modality into non-epistem ic and epistemic. The former has three subdivisions: participant-internal, participant-external and deontic meanings. Van der Auwera and Plungian ’ s study is typologically-oriented, and their classification will not be followed in this piece of research, since our main concern is English modals and other classifications seem more appropriate to capture their nature and evolution. Chapter 2. Grammaticalization, modali ty and impersonality 52 Following the logical relations established between possibility and necessity, we can formulate what is implied in the following figure: ‘ not possible ’ Æ ‘ necessary not ’ ‘ not necessary ’ Æ ‘ possible not ’ Figure 2.4 : Logical relations between necessity and possibility According to Palmer, this formula explains the relation betwee n the two kinds of meanings conveyed by each type of modality, obligation and permission (deontic modality), deduction and probability (epistemic modality), and desire and ability (dynamic modality), as shown in the following table: DEONTIC EPISTEMIC DYNAMIC NECESSITY Obligation- exemption Deduction (inferred certainty, logical necessity) Desire, volition, necessity POSSIBILITY Permission- prohibition Possibility, probability Ability Table 2.5 : Types of modality and modal meanings stemming from the basic notions of necessity and possibility . This table clarifies Palmer ’ s classification of modality. Stemming from the basic meanings of necessity and possibility, and filtering them through modality, we obtain a variety of meanings ranging from obligation to ability. Palmer ’ s classification, therefore, is very tidy and seems to be m ore fine- grained than twofold classifications, sin ce it distinguishes the third, polemical type of modality, namely, dynamic modality. Palmer himself (1979: 36) characterizes dynamic modality as a peripheral kind of modality: dynamic m odality is subject-oriented in the sense that it is concerned with the ability and volition of the subject of the sentence, rather than the opinions (epistemic) or attitud es (deontic) of the speaker (and addressee). It could well be argued that, because of this, dynamic modality is not strictly a kind of modality at all, m odality being essentially subjective ( … ), for CAN and WILL m erely make objective statem ents about the subject of the sentence, as do most other verbs. These are the reasons which made me lean on Palmer ’ s classification in a first approach to this study. However, a later review of these thoughts led me to discard part of Palmer ’ s classification for the reason that follows. However clear and tidy Palmer ’ s classification may be, its clarity and neatness works only in the world of logic, where clear-cut distinctions between one category and a nother are Chapter 2. Grammaticalization, modali ty and impersonality 53 possible. In the world of language, howe ver, such clear-cut distinctions are not always possible, and we often have to resort to the notion of gradience as a link or intermediate stage between prototypes. 29 In fact, Palmer ’ s classification is born out of the analysis of core examples, such as You must finish your homework before I ’ m back (deontic obligation), or You may choose whiche ver you want (deontic permission), or He can play piano (dynamic ability). However, as Coates (1983: 21) notes, these examples are only rarely found in real language. What we find is, rather, peripheral examples of the three classes which Palmer recognizes. In addition to this, Palmer ’ s classification does not establis h any connection between the different types of modality, up to the point that it is difficult to understand what the three of them have in common. Despite its logical status, Palmer's classification is not, therefore, the most appropriate one whe n dealing with langua ge, especially with corpus-data, since the meanings of the modal verbs will vary according to the context and normally cannot fit into any of the categories he proposes. As an alternative to Palmer ’ s classification, I have reviewed other authors ’ conception of m odality and the different types of modal meanings they put forward, paying special attention to those scholars who recognize two types of modality. Thus, Coates (1983) and Sweetser (1990) distinguish between root and epistemic modality. This may seem at first sight too broad a classifica tion, since obviously they are grouping together Palmer ’ s deontic and dynamic modality under the label ‘ root. ’ Their classification, nevertheless, is language- based, instead of logic-based, and it clearly specifies that core meanings are just a reference, while gradient examples are the most representative set in any corpus (cf. Coates 1983: 18-22). An example of the language-based nature of their account is that the label ‘ root ’ (used instead of deontic, which is taken from logic) is historically accounted for, since it is the meaning which gives birth to epistemic modality. 30 For the purposes of this work, I have combined Coates ’ (1983) and Sweetser ’ s (1990) view of root and epistemic modality with Talmy ’ s (1988, 2000) cognitive semantic approach to modal meanings in terms of force 29 On the theory of prototypes and categorization, se e, for example, Rosch (1977, 1978), and Taylor (1995). On the theory of gradience, see, among others Bolinger (1961), Lakoff (1987), and Aarts (1997). 30 This conception of epistemic meanings rooting from socio-physical ones is related to the theory proposed by the German philosopher von Hum boldt (1825), who suggested that in th e earliest stages of language only concrete ideas could be expressed. Grammatical forms are, then, the abstract result of the evolution of those concrete ideas (cf. Hopper and Traugott 2003: 19- 20). Chapter 2. Grammaticalization, modali ty and impersonality 60 (2.41) Clay pots … must have some protection from severe weather ( Æ objectiv e root) These examples, taken from Coates (1983: 21), are extracted from some of the corpora she uses in her study on the semantics of the English modals. As can be easily seen, (2.40) implies the existence of a subject imposing obligation on somebody else (animate human antagonist – the speaker- and animate human agonist – the interlocutor). Example (2.41), however, states a piece of advice as for the maintenance of clay pots by making reference to their fragile nature (the antagonist is an objective rule, as opposed to the passive non-human non-animate agonist, namely clay pots). Root modals may also show gradienc e with respect to the strong-weak continuum. Obviously, there is a difference in strength be tween examples such as the following: (2.42) She must pay taxes every year ( Æ stronger root) (2.43) She must buy a new pair of shoes ( Æ weaker root) In example (2.42) the agonist feels a strong threaten from the superior antagonist (i.e. the state). That is, there are severe consequences derived from the fact that she does not pay taxes. However, in exampl e (2.43) the agonist feels that in the antagonist ’ s belief system, she would fit be tter in society if she bought a new pair of shoes. The last scale of gradience as for root modals is that which concerns the origin of the force or of the barrier which conditions the event (Talmy 2000). The origin of the force or the barrier may be external to the subject or internal: (2.44) I must turn in this paper tomorrow ( Æ external force). (2.45) I need to call her now ( Æ internal force). As for the epistemic gradience, there does not seem to be a strong / weak contrast in epistemic modality, and the external / internal scale is obviously out of question, because epistemicity has to do with mental notions, and is, therefore always internal. The only scale of gradience which applies to epistemic modality is, therefore, the subjective / objective scale. According to Lyons (1977), objective epistemic modality expresses a mathematically computable chance that the state of affairs is true or untrue. Subjective epistemic modality, on the other hand, merely involves a subjective guess as for the truth of the statement. Chapter 2. Grammaticalization, modali ty and impersonality 61 Objective epistemic modality, however, is rare, as mentioned and verified by Coates (1983: 18) with the following corpus examples: (2.46) Paul must be in Liverpool now ( Æ subjective epistemic) (2.47) The simple truth is that if you ’ re going to boil eggs communally, th ey must be hard . ( Æ objective epistemic) This example is also used by Warner (1993: 14) to illustrate objective epistemic modality. However, Traugott and Dasher (2002: 111) consider it a case of deontic modality ( “ it is necessary for the cook to boil eggs hard ” ). Some scholars (cf., for instance, van der Auwera and Plungian 1998: 117, note 1) deny the existence of objective epistemic uses of the modals, since “ for reasons of logic, the speaker ’ s certainty with must happens to be absolute, but it remains the certainty of the speaker. ” Another aspect I will take into consideration when dealing with the semantics of root and epistemic modals is the distinction between necessity and possibility, based on Lyons (1977), Palmer (1979, 1986) and van der Auwera and Plungian (1998). I consider this distinction relevant for my study, since I am concerned with necessity verbs. The difference between necessity and possibility can be observed in the following examples: (2.48) They must answer 50% of the questions ( Æ root necessity = obligation) (2.49) They may use the dictionary ( Æ root possibility = permission) (2.50) They needn ’ t answer 100% of the questions ( Æ root necessity = exemption) (2.51) They cannot use the dictionary ( Æ root possibility = prohibition) Root necessity and root possibility are two radically different concepts, if considered in terms of barriers and forces. While root necessity, as in (2.48) and (2.50), implies some forces compelling or exempting the doer from performing the action (parallel to the lexical verb make ), root possibility, as in (2.49) and (2.51) implies open or closed barr iers for the action to be performed (parallel to the lexical verb let ). My overall classification of the different types of modality will be based on the parameters shown i n the following table: Chapter 2. Grammaticalization, modali ty and impersonality 62 Objective Subjective Strong Weak External Internal Necessity X X X X X X ROOT Possibility X X X X X X Necessity X X EPISTEMIC Possibility X X Table 2.6: Gradience within root and epistemic modality . Together with all these semantic differences between root and e pistemic modality, namely degree of subjectivity, strength of the force, and origin of the force, Coates (1983: 21) also mentions some syntactic features “ linked with Root meaning. ” Those features are the presence of an animate subject, an agentive verb and the possibility to appear with a verb in the passive voice, as illustrated in (2.52) and (2.53). (2.52) You must cook dinner tonight (animate subject, agentive verb). (2.53) Work must be finished by next week (passive voice). However, these features are not exclusive of root modality, as in (2.52) and (2.53). They may also be characteristic of sentences with epistemic meaning. See, for example: (2.52b) He may cook dinner tonight ( ‘ it is possible that …’ ). (2.53b) [ I ’ ve seen them working hard, so I think ] work may be finished by tomorrow ( ‘ it is possib le that …’ ). The semantic ancestors of need found in my corpus will be analysed taking into consideration all these semantic and syntactic features. The possible meanings of PDE need and need to are shown in the next section. 2.2.2.3 Semantic features of Present-Day English need and need to This section complements the morphosyntactic description of PDE need offered in section 2.2.1 and de scribes the semantic features of PDE need and need to . As suggested above, the alleged original semantic differences between both verbs – need to being concerned with internal necessity, and need with external necessity – have been neutralized in Present-Day English and the overwhelmingly frequent need to expresses both internal and external necessity (cf., for instance, Vihla 1999, Taeymans 2004a: 105). These verbs, then, prove to be semantically equivalent at least in the expression of root meanings, as will be seen in the following paragraphs. Chapter 2. Grammaticalization, modali ty and impersonality 63 Root meanings may be originated exte rnally or internally, as seen above. External root need and need to are not at all common in affirmative contexts expressing obligation, that is, the existence of a an external force imposed by the antagonist on the agonist. In Present-Day English this meaning is m ainly expressed by must and, with increasing frequency, by have to (cf. Smith 2003). Need , however, may be used to convey an external piece of advice on the agonist, as in the following example: (2.54) You need to get a hair-cut Æ weak obligation. (example from Leech 1987: 101) In sentence (2.54) need expresses a weakest external force which may make the agonist feel a somewhat urgent necessity to get a hair-cut. According to Smith (2003: 245) and to Taeymans (2004a: 107), PDE British nee d expresses external obligation only rarely and, when it does, as in (2.54), speakers resort to the internal quality of need to obtain some advantage, that is, they combine their wish with the addressee ’ s best interest. In other words, the speaker knows that by using the verb need , the agonist will react more willingly than by using, for instance, must or have to . The expression of rough exte rnal obligation is, then, not among the main semantic features of PDE need and need to. However, when the c ontext is non-affirmative, need and need to are most common expressing exemption or absence of obligation, that is, the antagonist releases the agonist from an alleged obligation, as in (2.55): (2.55) You needn ’ t type the report, I ’ ll do it later Æ exem ption The very common meaning of absence of external obli gation expressed by need in sentence (2.55) can be sketched as follows: A. The agonist does not want to VP. B. The agonist has the belief that the antagonist wants him/her to VP. C. The antagonist has the power to make the agonist VP or not VP. D. The antagonist releases the agonist from the obligation to VP. On rare occasions, non-affirmative need does not expr ess absence of obligation, but a force not to, i.e. prohibition, as in the following example provided by Coates (1983: 51): (2.56) and you needn ’ t glare at me like that! Chapter 2. Grammaticalization, modali ty and impersonality 64 By using the verb needn ’ t in (2.56) the ant agonist is clearly not exempting the agonist from the act of glaring, but implies the speaker ’ s dislike for the addressee doing so. This meaning is clearly to be kept apart from the ones sketched above. Here the antagonist does not want the agonist to VP, i.e. to glare at him, and places a strong barrier which obstructs the agonist from doing so, or else he should assume the consequences. This prohibition meaning of need (to) is expected to be exceptional and only found in particular contexts. Need may express root internal necessity , that is, both the obligation and the exemption may be internally originated. Consider (2.57) a nd (2.58): (2.57) I need /have to stay home tonight to study for the test Æ internal obligation. (example from Sweetser 1990: 53). (2.58) I need not read it again, I know it by heart Æ internal exemption. In both examples the antagonist is located in the self of the agonist. The self is split and opposing forces fight internally. Thus, in (2.57) the agonist ’ s self splits into one half which imposes on the other half the f orce to stay up all night to study. In (2.58), the most severe half of the agonist ’ s self releases the other half from reading the text again. These are examples of strong internal forces, but these may also be weak as in, for example, (2.59): (2.59) I need to buy a new pair of shoes . In this sentence need to expresses weak internal force and the semantic sketch could be the following: A. The agonist wants to VP. B. There is no external antagonist obstructing the agonist. C. The agonist ’ s self seem s to require som e inner permission (from the internal antagonist) to VP (otherwise, the sentence would be I WILL / AM GOING TO VP ). D. The agonist ’ s self (i.e. the antagonist) seems to grant itself perm ission to VP. Another example of need expressing internal force is Sweetser ’ s (1990: 62) he needs to go to the grocery store , in which the internal forces of wanting to eat compel the agonist to perform the action. In addition to external and internal root meanings, ne ed also expresses epistemic necessity , 33 that is, forces related to the world of logic and which are 33 According to Taeymans (2004a) only need (not need to ) can express epistemic necessity and, indeed, the scarce examples I have found seem to corroborate her statement. Chapter 2. Grammaticalization, modali ty and impersonality 65 originated in the mental domain. According to Sweetser (1990: 154, note 17), epistemic need only occurs in non-affirmative contexts, just like epistemic can does. One of such examples is (2.60): (2.60) No, he needn ’ t be a New Yorker – he could just have lived there a long time, or imitate accents well. (example from Sweetser 1990: 62). The meaning of needn ’ t in this sentence is clearly born out of logical factors, i.e. it is originated in the mental domain. Sentence (2.60) implies, on the one hand, that there appears to be enough evidence to think that he is a New Yorker, such as his revealing accent. However, this sentence also implies that that there are logical reasons why he could seem a New Yorker, while actually he is not one. The speaker ’ s knowledge of the world allows him to state that not everybody who sounds like a New Yorker must be from New York, because in his belief system he knows that people may imitate or learn new ac cents. In addition to negative examples of epistemic need such as (2.60), I have also found an affirmative example in Visser (1963-1973: § 1346), which seems to be a counterexample of Sweetser ’ s (1990) claim that epistemic need is restricted to non-affirmative contexts. Such an example is (2.61): 34 (2.61) I need look somewhat changed … for I have undergone some suffering, both of mind and body. (1838-1839 Dickens, Nicholas Nickleby XX) The implications of (2.61) are: s ince I have undergone so much suffering, it is necessarily the case that I look somewhat change d. Need in this example is, then, fairly similar to must in the above-mentioned e xample he has got three houses, four cars and a yatch; he must be rich . Our knowledge of the world forces us to assume that bodily and mental suffering has an effect on the physical aspect of the sufferer. With these examples of need expressing epistemic modality, I close this section devoted to the morphosyntac tic and semantic features of PDE need and need to . The latter may be summarized in the following list: (i) absence of external obligation, (ii) internal obligation and necessity, (iii) lack of internal obligation and necessity, and (iv) epistemic necessity. We have also seen that it has been marginally found expressing weak external obligation a nd prohibition, 34 This sentence is, according to Nykiel (2002), the first instance of epistemic need in his study on need from Shakespeare onwards. Chapter 2. Grammaticalization, modali ty and impersonality 66 although these meanings are not at all central to this PDE verb. In the analysis of the corpus (chapters 3, 4, 5 and 6) we will see whether or not the semantic ancestors of this verb exhibit some or all of its semantic features. 2.3. Impersonal verbs and constructions In addition to their propensity to undergo grammaticalization and their ability to convey modal meanings, the third common characteristic of the verbs analysed in this study is that all of them occur in so-called impersonal constructions at some point of history. In fact, impersonals and modals appear to be tightly related (cf. Pantaleo 2002). For this reason, this section aims at describing the notion of impersonality and clarifying some relate d concepts. Firstly, I will discuss a number of terminological issues (section 2.3.1). Secondly, section 2.3.2 offers some definitions of impersonal construction as found in the literature and describes each of the obligatory and optional constituents that make up impersonal constructions. Finally, in 2.3. 3 some remarks are made about the evolution of impersonal constructions in the history of English. 2.3.1. Terminologic a l issues The term ‘ impersonal ’ is used in the literature to refer to rather different concepts, such as (i) clauses whose verbs have no personal argument, (ii) clauses whose verbs have pers onal arguments with a function other than that of subject, (iii) clauses whose subject is not personal, and finally (iv) any verb occurring in any of the previous contexts (Denison 1993: 62). These four contexts have in common the absence of the subject or of some property of the subject in a kind of construction or verb (cf. Fern á ndez Soriano and T á boas Bayl í n 1999: 1725). The polyvalence of the term ‘ impersonal ’ may sometimes result in ambiguity (cf., for instance, Allen 1997: 1-2), but such an ambiguity may be increased if alternative terms are used for the same notion suc h us, for instance, ‘ subjectless, ’ ‘ quasi-impersonal, ’ ‘ nominative-less, ’ etc. (cf. M é ndez Naya and L ó pez Couso 1997: 185). The va riety of labels propose d by different scholars is an attempt to use a clear terminology and to delimit the boundaries of the concepts they posit. For instance, traditional studies such as va n der Gaaf ’ s (1904) distinguish between (true) impersonals, namely constructions containing weather verbs, and quasi-impersonals, namely constructions featuri ng any of the above-mentioned constituents (ii-iii). Chapter 2. Grammaticalization, modali ty and impersonality 67 However, as mentioned, the rise of so many different terms has only served to cause more confusion. The most frequent alternative label for the term ‘ impersonal ’ is ‘ subjectless, ’ defended, for instance, by Elmer (1981) and von Seefranz-Montag (1984). This label seems to be more precise, because it does not imply the absence of a personal argument in a clause, but the absence of a (syntactic) subject. Nevertheless, the term ‘ subjectless ’ is by no means free from controversy. Some of the probl ems concerning this label, as mentioned in Denison (1993: 61-62) are: (i) how t o account for the presence of a dummy subject it ( hit in Old English), as in the OE contrast rin ð vs. hit rin ð (Lit.: ‘ rains ’ vs. ‘ it rains ’ ); (ii) how to analyse content clauses introduced by that ( ðæ t in Old English) when they are the only argument of a verb, are they a subject or an object?; and (iii) some scholars consider the oblique animate noun phrase as subject, so that the label ‘ subjectless ’ turns out to be, to say the least, paradoxical. Among the scholars who analys e the oblique noun phrase as subject are, for instance, Elmer (1981), who resorts to the term ‘ subjectless, ’ and Allen (1995), who prefers the term ‘ impersonal ’ in order t o avoid the above-mentioned contradiction, t hough she acknowledges that ‘ impersonal ’ is not wholly trouble- free (cf. Allen 1995: 20). Summing up, terminology is inexact in the field of impersonality. Neither the label ‘ impersonal ’ nor the label ‘ subjectless ’ prove unproblematic. For the purposes of this paper, I will use the label ‘ impersonal, ’ following the traditional term defended by most scholars, though I am aware that, as Visser (1963- 1973: 29) says, it is a term used for convenience and for want of a better one. 2.3.2. Impe rsonal constructions: definition and structure The first decision we must make when trying to provide a definition of impersonal constructions is the perspective we will adopt, that is to say, should the perspective be semantic of syntactic? Mitchell (1985) adopts a strictly syntactic point of view: an impersonal construction is “ one which has only the formal subject hit ( … ), or which has no expressed subject and for which no subject other than the formal hit can be supplied ” (1985: § 1025). However, von Seefranz-Montag (1984) considers that impersonal constructions have a clear semantic component in common: [impersonal constructions] are a productive syntactic device to encode expressions of a specific semantic class: verbs denoting physical, emotional and Chapter 2. Grammaticalization, modali ty and impersonality 68 mental experiences ( … ), but also needs and obligation, possession and sometim es perceptions and abilities, existence and happenstance – processes and situations, in which a person is unvolitionally / unselfcontrollably (McCawley 1976: 194) involved. Her definition, therefore, takes syntax as a starting point, but uses semantics to specify the scope of the constructions . According t o M é ndez Naya and L ó pez Couso (1997: 186), the question of adopting a syntactic or a semantic perspective for the definition of impersonal constructions goes back to Wahl é n (1925), who points out that the term ‘ impersonal ’ may be applied to a group of verbs defined on semantic grounds, and to a type of construction defined on syntactic grounds. Denison (1993: 62) claims for the necessity to resort both to semantic and syntactic considerations when analysing i mpersonal constructions. This idea is also supported by M é ndez Naya and L ó pez Couso ( 1997), who propose to define semantically the traditionally-called im personal verbs and, therefore, differentiate between verbs of natural phenomena and ver bs of experie nce, as well as to define them syntactically as constructions characterized by the presence of a verb in the third person singular and the lack of a nominative noun phrase (1997: 190-191) . Therefore, verbs occurring in impersonal constructions refer either to natural phenomena, namely the so-called weather verbs (e.g. rain, snow , etc.), and verbs of experience, which include all the meanings mentioned by von Seefranz-Montag (1984), namely physical, emotional experiences, necessity, obligation, possession, ability, existence and happenstance. 35 The constructions in which these two kinds of verbs occur may vary. Firstly, weather verbs such as rain do not take an argument in Old English, though they may optionally take a dummy subject, namely it (OE hit rin ð = OE rin ð ). 36 Secondly, verbs of experience may occur in two different constructions, that is, with one argument 35 Since this study pays exclusive attention to verbs, in the analysis of impersonal constructions only structures with so-called impersonal verbs will be taken into account. It m ust not be forgotten, however, that OE shows an ample range of impersonal constructions involving the copula BE and an adjective or noun, as for example me is neod ‘ it is necessary for me. ’ See van der Wurf (1992) for a detailed account of this kind of phrasal impersonals. 36 Breivik (in a lecture given at the USC in April 2003; and also 1983: 257) considers that hit in this kind of constructions is an em pty slot-filler. This is a traditional treatment of this element, maintained also by Wahl é n (1925) and Mitchell (1985: § 1031). However, verbs such as rain may also take a subject of a different nature. Three different types may be distinguised. Firstly, these verbs may take a “ cognate object (or may be subject) ” (cf. Denison 1993: 93). Examples of this construction are recorded in the OED (s.v. rain v. 3). Secondly , weather verbs may also take an animate subject whose semantic role is recipient in glosses fro m Latin (cf. Denison 1993: 93). Finally, these verbs may also take an animate subject whose refer ent is a deity (cf. OED s.v. rain v. 2). Chapter 2. Grammaticalization, modali ty and impersonality 69 (e.g. OE hyngran ‘ hunger, ’ as in ac si ðð an him (DAT SG) hingrode (3 SG) ‘ and afterwards he hungered ’ ), 37 or with two argum ents (e.g. OE tw eogan ‘ doubt, ’ as in ðæ t nanne mon (ACC SG) þæ s (GEN SG) ne tweo ð (3 SG) þæ t se sie strong on his m æ gene þ e … ‘ that no man doubts (it) that he is great in strength who … ), 38 as mentioned by Denison (1990b: 140). It is precisely because of the multiplicity of constructions in which verbs of experience occur, as well as their evolution in the history of English, that scholars pay much attention to constructions featuring this type of verbs, rather than constructions with weathe r verbs. Thus, in the most comprehensive contribution so far to the analysis of the Old and Middle English impersonals, Elmer (1981) leaves out of the count “ genuine subjectless constructions like weather expressions, ” because “ they do not occur with a pseudo-s ubject ” (1981: 3, note 1). In a similar line, one of the most influential papers on the study of impersonality in English, namely Fischer and van der Leek (1983), also leaves weather verbs out, because they do not “ express a physical or mental/cognitive experience ” and they have no arguments (1983: 346). This piece of research will also pay attention to impersona l constructions involving ver bs of experience exclusively, not only because they provide a wider range of constructions, but also because the verbs analysed in this study are obviously verbs of experience rather than verbs of natural phenomena. In order to centre the discussion on impersonal c onstructions containing verbs of experience, as, for instance, the above-mentioned OE hyngran ‘ hunger, ’ tweogan ‘ doubt, ’ or, neodian (cf. Bosworth and Toller s.v. neadi an, neodian ), I would like to pr ovide a definition of such constructions. A very conci se definition of this kind of construction is provided by Elmer (1981: 3), as follows: “ OE predicates which generally occur with a pseudo-subject in dat ive/accusative case instead of a nominative subject noun phrase. ” This definition applies to Elmer ’ s work, because he considers that the oblique animate NP which typically occurs before the verb is a pseudo-subject (the reasons for this analysis will be seen below). However, if the syntactic status of pseudo-subject is denied, Elmer ’ s (1981) definition is no longer vali d. 39 Therefore, we have need to find a somewhat general definition of impersonal construction, su ch as that proposed by 37 Example taken from the OED (s.v. hunger v. 1). My translation. 38 Example and translation taken from Denison (1990b: 141). 39 As will be explained below, the function and r ole of the animate NP has been subject to different interpretations. Both Fischer and van der Leek (1983) a nd, on a different framework, Allen (1997) use the term ‘ experiencer ’ to refer to it. Chapter 2. Grammaticalization, modali ty and impersonality 76 us now examine Allen (1995), a final and m ore recent classification of impersonal constructions, before specifying which classification and terminology will be used in this piece of work. 2.3.2.3. Allen (1995) Allen ’ s (1995) comprehensive a nd thorough study provides a detailed analysis of both impersonal constructions and some related ‘ personal ’ verbs such as OE lician , ‘ to like. ’ These two types of construction have some common features which have made scholars bring t hem up together in grammars and specialized books (cf. below, in section 2.3.3, Jespersen ’ s invented example for the evolution of the OE impersonals). Such common characteristics are, for instance, the fact that both contain an animate noun phrase inflected for the oblique case which plays the semantic role of experi encer, 45 on the one hand, and the fact that the PDE counterparts of both structures di splay an exper iencer in the nominative case and governing verb agreement, on the other hand. These two factors are the reasons why Allen (1995) groups both construc tions under the label ‘ experiencer verbs. ’ She classifies these verbs following Elmer ’ s (1981) terminology, although she makes some innovations. Thus, Allen (1995: 69) classifies “ 2NP constructions with the experiencer verbs ” – that is, constructions in which an experiencer verb occurs with two nominal arguments – in the following way: - Type N: Dative/Ac cusative 46 experiencer + Genitive theme. (Elmer ’ s Type N). - Type I: Dative experiencer + Nominative theme. (Elmer ’ s Type I). - Type II: Nominative experiencer + Genitive theme. (Elmer ’ s Type II). Allen (1995), therefore, follows Elmer ‘ s (1981) taxonomy, but she mentions two additional types of construction. In one of them, the experiencer verb has two nominal arguments inflected for the accusative, what Allen calls respectively the experiencer and the theme (cf. Fischer and van der Leek ’ s 1983 ‘ cause ’ ). In the second type, the experiencer is inflected for the dative, and the theme is inflected for the accusative. However, these constructions are very restricted, in fact, Allen (1995: 74) only finds one possible example of the first type, and three possible 45 Though Allen (1995) capitalizes the initial <e> of this term, I do not find any semantic difference between Allen ’ s Experiencer and Fischer and van der Leek ’ s (1983, 1987) experiencer, and will, therefore, adopt this label in low case. 46 Sometimes it is ambiguous whether the experiencer is dative or accusative, but Allen (1995) opts to refer to it as dative in all cases. Chapter 2. Grammaticalization, modali ty and impersonality 77 examples of the second type. For this reason, these types “ must be regarded either as mis-analyses or as constructions which were at best peripheral ” (1995: 95). There are, then, three main constructions for the experiencer verbs when they are complemented by two nominal arguments, and those are the same as those proposed by Elmer (1981) : Types N, I and II (see examples (2.64), (2.65) and (2.66) above in section 2.3.2.1). The classification of the constructions in which experiencer verbs may occur also includes those instances where one of the ar guments of the verb is a sentential theme, either tensed or infinitival. Allen (1995: 86, and passim ) us es the label PROP construction to refer to these instances. PROP constructions may be of three different types: 47 - Type S: non-nominative experiencer + sentential theme. (2.73) Ð onne ð am menn ne lyst on his life nan god don when the-dat. man-dat. not wishes in his life no good do ‘ When the m an does not wish to do any good in his life. ’ (example from Allen 1995: 86) - Type hit : non-nominative experiencer + formal hit or þæ t + sentential theme. (2.74) þ a gelicode hit ð am leodebiscope … þæ t he his lichaman up ð a gelogode then pleased it the-dat. bishop … that he his body up then placed ‘ Then it pleased the bishop to enter his body ’ (example from Allen 1995: 87) - ‘ Personal ’ : nominative experiencer + sentential theme. (2.75) Ne tweoge ic naht, þæ t gode weras w æ ron on þ ysum lande not doubt I nought, that good men were in this land ‘ I do not doubt at all, that there were good men in this land ’ (example from Allen 1995: 97) Type S parallels Elmer ’ s (1981) Type S and Denison ’ s (1993) type (i/ii), since it is ambiguous whether the sententi al argument plays the syntactic role of subject or of object. Allen (1995), like Elmer (1981) and Fi scher and van der Leek (1983, 1987), does not doubt about the object role of on his life nan god do , because the gloss she provides for the verb lystan does not allow for any other interpretation. However, this verb also has the meaning of ‘ to please, cause pleasure or desire, ’ (cf. Bosworth and Toller s.v. lystan v.), and according to this meaning, example (2.73) may also be interpreted as ‘ when doing no good in his 47 As with 2NP types, Allen (1995) only takes into account exam ples where both the experiencer and the theme are expressed. Chapter 2. Grammaticalization, modali ty and impersonality 78 life does not please the man, ’ which would render evidence for the subject status of the sentential argument. The reason why Allen ( 1995) does not hesitate to consider the status of ð am menn is, just like in Elmer ’ s (1981) theory, the subject nature of the experiencer, as will be seen below in this section. Allen ’ s Type hit differs from Type S in having a formal subject, which may be hit or the demonstrative þæ t . The use of hit in this type is rare in Old English, according to Allen (1995: 88), since she only finds nine examples. However, the frequency of use of hit as a formal subject increases when the experiencer is not expressed (as will be seen, this is another of the reasons why Allen (1995) consider s experiencers as the subject of experiencer verbs). The use of þæ t , on the contrary, is much more common (1995: 88). Finally, in the ‘ Personal ’ Type the experiencer is inflected for the nominative and controls verb agreement. It is, therefore, a truly canonical personal construction, in which the experiencer clearly plays the syntactic role of subject. Allen ’ s (1995) work, however, breaks the canon in considering that the experiencer is the subject in all six types of impersonals she regards, no matter the case for which it is inflected. In fact, there are some modern languages which allow non-nominative subjects, as, for example, Japanese, Korean, Georgian or Icelandic, a Germanic language, and Allen (1995) relies on this evidence to justify the possible existence of non-nominative subjects in early periods of English. Allen (1995: 3) claims that earlier analyses of early English experiencer verbs stem from the wrong assumption that subjects are based on morphol ogical grounds, that is to say, from the assumption that subjects are always infl ected for the nominative case in English. She considers that the category subject, however, “ must be determined on the basis of syntax, rather than morphol ogy, ” and, therefore, she gives syntactic evidence in favour of considering the experiencers as subjects. Allen (1995) reinforces the evidence provided by Elmer ’ s (1981) for a subject analysis of experiencers in this kind of cons tructions. Firstly, Elmer (1981) points out that the experiencers (what he calls oblique NPa) are generally in preverbal position, whic h is a location usually occupi ed by subjects. In this respect, Allen (1995) compares, as already mentioned, the occurrence of hit as formal subject in the PROP constructions and its relation to the occurrenc e of oblique ex periencers. She finds out that the presence of hit increases when the experiencer is not expressed. This clearly leads to the subject analysis of those experiencers. Secondly, another piece of evidence pr ovided by Elmer (1981), Chapter 2. Grammaticalization, modali ty and impersonality 79 which has also been mentioned in section 2. 3.2.1, is that in coordina te constructions in which a personal and an impersonal verb are linked, the nominative subject of the personal verb can be elided, because it is easily gathered from the oblique e xperiencer of the impersona l verb (this phenomenon is referred to as “ coordinate subject deletion ” by Alle n 1995). This leads to the conclusion that both elements act as subjects of their respective clauses, since, in general, only subjects controlled this kind of elision in Old English. Allen ’ s (1995) contribution to this explanation is the comparison she draws between preposed dative experiencers controlling coordinate subject deletion and preposed dative objects of ditransitive verbs. Her findi ngs prove that coordinate subject deletion occurs much more often with preposed dative experiencers than with preposed dative objects (Allen 1995: 442). As for the findings related to these experiencers in Middle English, Allen ’ s (1995: 247-248) evidence becomes stronger, and she provi des three new contexts in which the tendency is reflected: 1. The postposed theme in Type I c onstructions (that is, constructions which in Old English have a dative experi encer and a nominative theme, such as OE lician ‘ to like ’ ) fails to agree with the verb. 2. Preposed dative experiencers are fairly frequent in Middle English, despite the fact that fronted pronouns had become quite unusual. 3. When both the pronominal experiencer and the pronominal theme occur in preverbal position, the pronominal experiencer precedes the theme. Since the usual word order of pronominal forms is SO (subject object), the experiencer must have been understood as subject by new language learners. All these syntactic features of the preposed dative experiencers both in Old and Middle English seem enough for Allen (1995) to consider them subjects of the constructions in which they appear, irrespectively of the case marking they exhibit. Summing up, Allen ’ s (1995) classification of the possible constructions in which impersonal verbs may occur stems from the type of arguments which such constructions exhibit. If both ar guments are noun phrases, there may be three types: Type N, Type I and Type II. If one of the arguments is of sentential nature, Chapter 2. Grammaticalization, modali ty and impersonality 80 there may also be three types: Type S, Type hit and ‘ Personal ’ Type. All six types of constructions may have, and most often do have, an animate experiencer which Allen (1995), like Elmer (1981), considers to be a subject, independently of its morphological inflections. The English language will gradually lose its ability to mark subjects obliquely, but this is a feature still found in other modern Germanic languages such as Icelandic (cf. Allen 1995: 3). For the purposes of this work, I will follow Allen ’ s (1995) classification of the impersonal constructions, since it is the most comprehensive, thorough and detailed of all the studies on impersonality, both from a synchronic and a diachronic perspective. More over, her descriptive approach is close to the one adopted for this study, while Fischer and van der Leek ’ s (1983) GB theory approach, for example, is not in keeping with the gene ral purposes of this piece or research. In addition to her 1995 work, I will also follow other works on impersonal verbs and constructions, such as Allen (1986a, 1986b and 1997). Before proceeding any further, a word of clarification is in order here. This section has shown that Allen (1995) uses the label “ experiencer ” to refer to the personal NP argument in these constructions, irrespective of whether it occurs in nominative or oblique case; likewise, the label “ theme ” is applied to the complement, whether sentential or nominal. Following this author, in the analysis of my corpus I will use the term “ experiencer ” exclusively to refer to the (animate) noun phrase irrespective of the case in which it appears in the different periods of the language. Similarly, I will use the term “ theme ” to refer to the constituent which encodes the thing needed, be it nominal or sentential. For example, in a sentence such as he needs to go , he will be analysed as the experiencer, and to go will be analysed as the theme, although it is well-known that the most widely-spread labels to refer to these arguments are “ subject and “ complement ” (cf. among many others, Traugott 1992, Warner 1993). In fact, in the introductory sections of chapters 3, 4 and 5, the labels “ subject ” and “ complement ” may be used for coherence with the bibliographical references. The decision to use “ experiencer ” and “ them e ” in my corpus analysis is informed by the fact that the labels “ subject ” and “ complement ” are suitable for PDE need constructions, but prove problematic for earlier periods of the language. Toge ther with the terms “ theme ” and “ experiencer, ” I will also be using the terms “ agonist ” and “ antagonist, ” as the opposing forces intervening in the expression of necessity (cf. section 2.2.2.2). Even if the agonist is normally encoded as an Chapter 2. Grammaticalization, modali ty and impersonality 81 experiencer in experiencer verb constructions, I will keep these two pairs of labels separately, because they correspond to different levels of analysis. 2.3.3. Ev olution of the Old English impersonals Most of Old English impersonals (e.g. the above-mentioned example me hreowe þ … ) evolve i n the course of time to personal co nstructions (e.g. I rue … ), that is to say, the oblique animate noun phrase which precedes the verb becomes nominative, and fulfils, therefore, the morphological features of subjects. This interesting evolution has led many scholars to try to find an explanation for it. From Jespersen (1909-1949) onwards, 48 the most widely acknowledged theoretical explanation of the changes has been that of reanalysis. The most famous example used by scholars supporting reanalysis is the development of t he verb like , which, according to Allen (1995), in Old English is an experiencer verb occurring in a personal construction. Jespersen ’ s invented example with all the stages undergone by like is illustrated under (2.76): (2.76) (a) þ am cynge licodon peran the king (dat.) pleased (pl.) pears (nom . Pl.) 49 (b) the king likeden peares (c) the king liked pears (d) he liked pears Following the stages outlined in (2.76), the change from (a) to (d) is explained in terms of reanalysis due to the loss of inflections. In stage (a), the experiencer is inflected for the dative, while the second, inanimate, noun phrase (Allen ’ s 1995 theme) is inflected for the nominative and controls verb agreement, as is reflected in the plural verbal f orm licodon . Stage (b), in turn, still shows agreement between the inanimate NP and the verb, but the experiencer is neutral as for case, due to the loss of inflec tions in nouns. In stage (c), we witness the disappearance of verbal inflections, which, t ogether with the loss of nominal endings, yields an ambiguous clause, since both elements seem to be potential subjects (with a slight difference in the meaning of the verb, that is, from ‘ please ’ to ‘ like ’ ). Finally, stage (d) illustrates the ultimate interpretation given to the structure involving reanalysis of the experiencer as subject. This reanalysis is, therefore, 48 Actually, van der Gaaf (1904) already recognizes th e same theory, that is, he suggests that the change from impersonal to pers onal constructions is due to the am biguity resulting from the morphological coalescence caused in the ME period by the loss of inflectional endings. 49 Gloss provided by Denison (1993: 74-75). Chapter 2. Grammaticalization, modali ty and impersonality 82 explained basically as a c onsequence of two factors. The first one concerns the morphosyntactic ambiguity caused by the decay of the inflectional system, which brings about structures such as (c). The second reason has to do with t he rigidification of the SVX word order, which leads to the interpretation of the preverbal element as subject. There have been many responses to Jespersen ’ s analysis ever since it was first published. To mention just a few, von Seefranz-Montag (1984: 529-530) argues that there are two facts which contradict Jespersen ’ s alleged development of impersonals. On the one hand, many experiencers are disambiguated as oblique co mplements by the use of the dummy subject it (e.g.1205 hit me rwes þ at ‘ I rue that, ’ from Elmer 1981: 86. ex. 8). On the other hand, the hypothesis of the SVX word order cannot explain the loss of constructions wit h dummy it as subject (e.g. 1304 hit him of- þ inche ð ‘ it causes grief to him, ’ from the OED s.v. ofthink v.2), or the abolition of non-personal nominative subjects in favour of personal ones (e.g. ano þ er drem dremede me yet , which yields I dreamed anot her dream , from von Seefranz-Montag 1984: 530). On a more specific line, McCawley (1976) points out t hat in oral speech very few experiencers would be third person singular as the king in (2.76) above, but they would most probably be first or second person pronouns, and these were not morphologically ambiguous. Reinforcing this argument, Allen ’ s (1986b) monographic paper on like reflects that the proportion of ambiguous case NPs in sentences with two nominal arguments i s notably low. Furthermore, the frequency of pre-verbal experiencer position with the verb like , and all Type I verbs, in general, is fairly low as well, a fact which is also noticed by Fischer and van der Leek (1983: 351; cf. also Allen 1995: 111 for a re-statement of this idea). Allen (1995) provides further evidence against the traditional view that there was reanalysis triggered by the loss of inflections. As already mentioned (cf. section 2.3.2.3), Allen cons iders that t hose explanations are based on the wrong assumption which states that there is a close relationship between case marking and grammatical relations. According to this assumption, reanalysis would explain the assignment of nominative case to the prepose d NPs which formerly were inflected for the dative, because the preposed position makes new language learners reanalyse those NPs as subjects. Allen (1995) is obvi ously against this interpretation, since, as already mentioned, she consi ders that preposed dative experiencers are subjects in Old English, despite their morphological endings. Anot her piece of evidence against the reanalysis Chapter 2. Grammaticalization, modali ty and impersonality 83 interpretation is, according to Allen (1995), the lack of agreement between dates. If reanalysis were the right interpretation, we “ would predict that we should begin to find examples of clearly nominative (i.e. pronominal) Experiencers with all formerly PDE [=preposed dati ve experiencer] verbs as soon as nouns were no longer regularly marked for dative case ” (1995: 324). In other wor ds, the loss of inflections only affected nouns, since pronouns still show case marking in Present Day English. Therefore, a ME speaker could select an apparent nominative noun as the experiencer of an experiencer verb, but could select a clear dative pronoun. Allen (1995) specifies that a clear piece of evidence for reanalysis in Middle English should exhibit a nominative pronoun, because that would be the ultimate proof that speakers chose a nominative experiencer with some verbs. However, clear nominative experiencers do not appear with the majority of Type I verbs suc h as like until more than one century after the loss of distinction between nominative and dative. In addition to that, impersonal constructions do not start losing ground until the 15 th century, while the us e of preposed dative experiencers decreases already in the 14 th century. Finally, reanalysis presupposes t he sudden death of one system in favour of another. For example, Lightfoot (1979, 1988) adopts the generally assumed date in which word or der changed from OV to VO, that is, the 12 th century, as the date in which reanalysis took place. 50 However, the loss of impersonal constructions cannot have been due to a sudden reanalysis, because, as Allen (1995) mentions, such a loss was not a matter of variation across speakers, but variation across verbs within the language of the same speaker. In other words, if reanalysis were the key concept to interpret the evolution of impersonals, one would expect that some speakers reanalysed all constructions as personal, whil e other speakers had not reanalysed them yet. That is to say, there would be variation across speakers. However, that does not seem to be the case in Middle English. What we actually find is variation within the language of the same speaker. Even after the loss of morphological distinctions the same speaker assigns oblique or nominative case to the NPa dependi ng on the verb. In Allen ’ s words, speakers managed to “ abduce grammars in which particular verbs could assign case lexically to their Experiencers ” (1995: 451). This seems to imply that 50 However, Lightfoot seems to change his mind as far as this sudden reanalysis is concerned, and opts for an explanation of the loss of impersonal constructions as a gradual process in his 1991 work. Chapter 2. Grammaticalization, modali ty and impersonality 84 the change from impersonal to personal construction has not been sudden due to the birth of new speakers. Therefore, if Jespersen ’ s theory is not applicable, how shoul d the development of impersonal constructions be accounted for? Von Seefranz- Montag (1984) examines a number of Indo-European and non-Indo- European languages and comes to the following conclusion (1984: 546): The change of ‘ subjectless ’ constructions is a consequence of historical changes in the functional and coding properties of the grammatical relation ‘ subject ’ : The gradual acquisition of syntactic and morphosyntact ic subject properties by experiencer arguments of impersonal verbs is proportional to the establishm ent of grammatical relations in a language. This account goes hand in hand with Allen ’ s (1986b: 398) assertion that from the 13 th to the 15 th centuries “ a preposed cause was marked nominative and a postposed cause was marked dative. ” Allen (1986b) also underlines the fact that in Old English nominative was the default case, and, therefore, any postposed cause could get nominative case, while in Middle English, case assignment to objects was structural, and any postposed element is inflected for the dative. This seems to be directly related to what von Seefranz-Montag (1984) calls “ establishment of grammatical relations in a language. ” Once inflectional cases are based on structural grounds, pre-verbal elements start to acquire nominative case, while post-verbal elements obtain da tive case. This is in direct connection with Allen ’ s (1995) postulation of an OE dative subject, since at that period, morphological case marking was not connected to syntactic functions. Nevertheless, neither von Seef ranz-Montag ’ s (1984) explanati on, nor Jespersen ’ s picture can account for the rise of new impersonals in th e ME period, such as, for instance, the impersonal uses of ME neden, 51 lacken, happen or thurven (cf. von Seefranz-Montag 1984: 526; Anderson 1986; Pocheptsov 1997: 479-480). In any case, these new members of the set will not stay in the language for a long time, because by the end of the 15 th and the beginni ng of the 16 th centuries, impersonal constructions disappear “ along three avenues ” (von Seefranz-Montag 1984: 526): (i) either the verb disappears in favour of a nearly 51 It has already been mentioned that the OE verb ne odian is characterized as impersonal by Bosworth and Toller (s.v. neadian, neodian ) . Von Seefranz-Montag ’ s (1984) account of impersonal ne den as a ME innovation, however, does not seem to be wrong in the light of the OE data retrieved from the 1.2 million-word corpus where not a single example of impersonal neodian has been found. This comes to prove that OE impersonal neodian cannot have been frequent. Chapter 2. Grammaticalization, modali ty and impersonality 85 synonym personal verb (e.g. OE þ yncan ‘ to seem ’ disappears in favour of think ); (ii) a dummy hit is introduced as an obligatory element (e.g. it pleases me ); and (iii) preverbal oblique experiencers are assigned nominative case (e.g. I think something, I like something ). Another well-accepted explanation for the development of the impersonal constructions in English is Fische r and van der Leek ’ s ( 1983). 52 Their main point of argumentation states that, instead of consi dering that impersonal verbs change their meanings in the history of English (e.g. OE lician ‘ to please ’ > PDE like ), a more accurate explanation is that in Old English (1983: 337-338): both meanings existed side by side, systematically associable with different syntactic constructions. Due to the weakening of the OE case system , the various constructions collapsed into one; this resulted in semantic ambiguity, which in its turn led to the obsolescence of one or the other m eaning of the verbs in question. By way of illustration, Fisc her and van der Leek (1983: 352) resort to an example of OE lician . Example (2.77) is an instance of this verb meaning ‘ to have pleasure, to like, ’ rather than ‘ to please ’ (from Fischer and van der Leek 1983: 352): (2.77) þ u eart sunu min leof, on þ e ic wel licade you are son m y dear, on whom I (nom) well was pleased ‘ you are my dear son in whom I was well pleased ’ ( Mark ; Skeat, 1871-1887: 11) The ver b licade is said to have the meaning ‘ be pleased ’ in this sentence, and this interpretation is probably favoured by the nominative experiencer ic ‘ I. ’ With this example, Fischer and van der Leek (1983) justif y their explanation for the evolution of impersonal constructions without implyi ng, as Jespersen and others do, that the OE impersonal change d to a personal construction. What they maintain, on the contrary, is that the OE impersonal ve rbs could be construed in both personal and impersonal constructions. In most of the cases, the impersonal construction was lost in favour of the personal one, and this disappearance is derived from the fact that from the 16 th century onwards no verb c an occur with 52 Their theory is very well accepted by Denison (1993: 80 ff.) Chapter 3. Old English þ ur fan , be þ urfan, neodian and behofian 92 However, the origins of the weak class “ although obscure, were strictly Germanic ” (Hogg 1992c: 146). The formal difference between these two types of verbs concerns the f ormation of the preterite. Strong verbs form their preterite by means of vowel variation in their stem (cf. PDE sing-sang-sung ), while weak verbs form their preterite by means of suffixation, adding a denta l suffix to the stem (cf. PDE love-loved-loved ). Paradoxically, the Germanic innovation overrode the Indo-European conj ugation and became the more pr oductive paradigm. Both neodi an an d behofian belong to the weak class of verbs. However, not all OE verbs can be assigned to one of these two categories. According to Mitchell (1985: § 600), there are two more types of verbs: “ preterite-present ” and “ anomalous ( willan, don, gan and beon ). ” Hogg (1992c: 146), however, resorts to the general label “ irregular ” to refer to all these verbs. The reason why their classification is different is that Mitchell ’ s classification is historically based, since the origins of the preterite-present and the anomalous verbs are not the same. Hogg ’ s classification, by contrast, is based on OE synchronic data; in other words, by the time when Old English was spoken all those verbs were just different from the norm and this accounts for the label ‘ irregular. ’ Among the preterite-presents we find þ urfan and be þ urfan . As for the frequency of occurrence, despite the fact that the vast majority of OE verbs belong to the weak class, the three types of verbs (weak, strong and irregular) are very similar in rate of occurrence. As Hogg (1992c: 146) points out, although the verbs which he calls irregular verbs constitute a small minority, they are highly freque nt verbs; similarly, among strong verbs t here are also many high-frequency verbs. Concerning the inflection of OE verbs, the y can be inflected for person, number, tense and mood. Old English e xhibits first, second and third per son singular verbal forms, and a single form for the plural, as Old Saxon and Old Frisian (cf. Mitchell: 1985: § 17). 2 Like in all Germanic languages, verbs in Old English have two tense forms: pr esent and past both in the indicative and in the subjunctive moods, while in t h e imperative mood the y only mark the second person singular and plural of the present tense. With regards to mood, together with forms belonging to the well distinguished indicative, subjunctive and imperative moods, Mitchell also recognizes what he calls “ ambiguous forms such as woldest, ” because “ the 2 Plural verbs forms are also used with the nominative of the dual of the first and second person pronouns (cf. Mitchell 1985: § 600). Chapter 3. Old English þ ur fan , be þ urfan, neodian and behofian 93 ending – est in the preterite may be either indicative or s ubjunctive ” (1985: § 601). However, the origin of these unclear forms need not be a linguistic one, but the result of the fact that texts were written by different scribes (cf. Mitchell 1985: § 601a for a list of these “ ambiguous ” forms). Thus, person, number, tense and mood ar e the four basic grammatical categories that Old English could be inflected for. The expression of anot her grammatical category, namely voice, is periphrastic (in combination with the verbs beon/wesan or weor þ an ), with the exception of the verb hatte “ is/was called ” , plural hatton “ are/were called ” (cf. Mitchell 1985: § 600). This brief summary of the morphological features of OE verbs reveals that the verbal system of Old English was highly inflected as compared with that of Present-Day English. This summary also makes it possible to classify þ urfan and its derived verb be þ urfan as preterite-present verbs, and neodian and behofian as weak verbs. In the next sections the syntactic features of these four verbs will be examined. 3.2. Preterite-present verbs an d Old English pre-modals This section is divided into two parts. In the first part (3.2.1), I give a brief account of preterite-present verbs, a class to which two of my verbs belong. In the second part (3.2.2), the OE preterite-presents þ urfan and be þ urfan are described. 3.2.1. Morphologic al, syntactic and semantic characteristics Preterite-present verbs are not very numerous in Old English if we compare them to weak and strong verbs. However, most of them are high-frequency verbs, and, therefore, they occur very often in OE texts. I will start by describing preterite- present verbs from a morphological perspective, and will turn later to their syntactic features as well as to their semantic dimension. As already mentioned, scholars differ as to where to include these verbs in a morphological classification of OE verbs. They may be either included in the ‘ irregular ’ group (within the general classification strong, weak and irregular) or they may constitute a class of their own. This obviously implies that these verbs are ‘ special ’ as far as their morphological nature is concerned. Roger Lass (1994: 169) provides a suitable definition of these verbs: Chapter 3. Old English þ ur fan , be þ urfan, neodian and behofian 94 The origin is a non reduplicating IE [Indo-European] perfect, which developed present tense [ … ]. Since the past sense was lost in these historical perfects, new pasts had to be constructed; and since the weak conjugation even in early tim es was the only productive one, this is the natural source. Some of these verbs are of course ancestors of our modern auxiliaries; the fact that th e present is ‘ really ’ (historically) a strong preterite accounts for one m ajor structural anomaly: the lack of 3 sg inflection ( he can , not *can-s ). Since the strong PRET1 has no ending here [ … ], the descendants of these OE presents don ’ t either. Theref ore he can is really equivalent to he sang , not he sings . These verbs, therefore, derive from original preterites of strong verbs, and after they lose their past time reference, they develop a new preterite following the weak conjugati on, since this is the productive one in Old English. 3 Denison ’ s (1993: 296) definition is very similar to Lass ’ s, but he provides a new piece of information. He states that there is a di fference between the inflectional ending of the second person singular in the preterite of regular strong verbs and in the present of preterite-present verbs: while the ending for the preterite of the first group is – e , the ending f or the present in preterite-present verbs is – st . This seems to be a clear indicator that these verbs had lost their past time reference and had co me to express only present tim e. Consequently, Denison ’ s definit ion of these verbs is as follows: “ a set of verbs with a present tense just like the past (preterite) tense of a strong verb (apart from the – st of the 2 SG) and a past tense formed on an irregular stem with the endings of the weak past ” (Denison 1993: 296). The fact that these verbs have historically been conjugated according to two different classes has led some scholars to call them ‘ strong-weak. ’ However, this term is not very accurate, since, as Campbell (1959: § 726, fn. 1) points out, this label “ implies commitment to the view that the dental element in the preterite of these verbs is identical in origin with that of the Gmc. [Germanic] weak preterite. ” In other words, it is a mistake to imply that the dental inflectional ending of preterite-present verbs is the sa me as the one of weak verbs, since these endings did not a ppear at the same point of time. The te rm strong-weak seems to imply that the verbs under that la bel belong to both classes simultaneously, a nd this is not true. As already mentioned, they used to belong to the strong class and then transferred to the weak one after having lost their past time reference. 3 The same class of verbs may be found in all Germanic languages, as well as in Latin (e.g. coepi , ‘ I begin ’ ) and Greek (e.g. oida , ‘ I know ’ ), which evidenc es the Indoeuropean origin of this class of preterite-presents (cf. Warner 1993: 140). Chapter 3. Old English þ ur fan , be þ urfan, neodian and behofian 95 Preterite-present verbs may be subdivided into two groups: the ‘ non- modal ’ verbs and the so-called ‘ pre-moda ls ’ (Traugott 1992: 193, Denison 1993: 296). Non-modal verbs include: beneah/ geneah ‘ suffice, ’ deah ‘ avail, ’ gemunan ‘ remember, ’ (ge)unnan ‘ love, grant, ’ and witan ‘ know. ’ 4 None of these verbs survives in Present-Day English. The ‘ pre-modal ’ group is the most relevant one for t he purposes of this study. According to Mitchell (1985: § 990) pre-modals include: agan ‘ own, ’ cunnan ‘ can, ’ *durran ‘ dare, ’ magan ‘ may, ’ *motan ‘ must, ’ *sculan ‘ shall, ’ and þ urfan ‘ ne ed. ’ Denison ’ s (1993) classification coincides with Mitchell ’ s, except for the fact that Denison treats OE agan as a marginal modal, since “ its syntactic properties were significantly different from the rest ” (1993: 295). Nearly all these verbs survived in Present-Day English, and they belong to a special class of verbs, the modal verbs, a group which also includes PDE will . We know that all these verbs are somewhat different from the rest in Present-Day English, and scholars agree in that they are also different from the rest in Old English. However, there is no agreement as to what label would be suitable to cover all of them plus willan in Old English. OE willan does not have a preterite- present origin, 5 and therefore the term ‘ preterite-present ’ is not a c omprehensive one for all these verbs. As already pointed out, Traugott (199 2: 193) uses the term ‘ pre-modal, ’ while Mitchell (1985: § 991) calls them ‘“ modal ” auxiliaries. ’ The term auxiliary, however, is not ver y suitable for these verbs, since they could also be used independently as full lexical verbs. In addition to this, the term modal is also misleading, because the verbs included under this label do not necessarily convey modality in Old English. Denison (1993: 292) states that “ For simplicity I shall stick to modal , without inverted commas. ” I will follow Traugott (1992) and use the term pre-modal without inverted commas, since, even though they did not behave exactly like PDE modals, they are obviously their ancestors. 4 Since these verbs are taken from Denison (1993), the forms appear in the third person singula r, as opposed to the infinitive, which is the form used by Mitchell (1985) to quote OE verbs. 5 Even though OE willan exhibits t he morphology of a regular weak verb, Warner (1993: 142) mentions a couple of anomalies, which bring it close to preterit e-presents. The first one is the second person singular of the present indicative, þ u wilt , which resembles preterite-presents rather than the expected form for a weak verb (* willest ). The second anomaly concerns the third person singular of the present indicative, he wile , which clearly contrasts with the expected weak form (* wille þ ). These two features, nevertheless, are not enough to include OE willan into the group of preterite-presents. Chapter 3. Old English þ ur fan , be þ urfan, neodian and behofian 96 I have tried to show that most schol ars agree that preterite-present verbs should not be treated as a single category, but included into the class known as OE pre-modals. I will follow this classification in the remainder of this section, where I deal with the syntactic and semantic characteristics of these verbs. However, since I am overtly dis cussing modals, I should not forget to, at least, mention what Denison (1993: 315-323) calls marginal modals in Old English, as opposed to ‘ non-modals ’ and ‘ pre-modals. ’ These are: agan (which Mitchell 1985 included into the main modal group), have ‘ be obliged, ’ be ‘ of necessity, obligation or future, ’ uton ‘ let ’ and ongi nnan ‘ begin. ’ These verbs fall outside of the scope of this study because they do not pr ovide any relevant information related to the group of pre-modals, to which one of t he verbs under analysis, OE þ urfan , belongs. In ot her words, my concern here has to do with the OE preterite- present verbs that have yielded any PDE modal, alongside OE willan , since its syntactic and semantic features bring it very close to preterite-present verbs . By excluding marginal modals I avoid dea ling with marginal cases, which could constitute a hindrance to my attempt to offer a clear description of the background of OE þ urfan . Table 3.1 provides a graphic summary of all the groups of verbs that have been mentioned so far, with the aim of delimiting the scope of this study. In order to determine the classification of these verbs, I have used the following sources. The list of OE pre-modals follows Traugott (1992) and Mitchell (1985). The inventory of the OE preterite-present verbs has been taken from Campbell (1959). The label “ marginal modals ” has been borrowed from Denison (1993). And finally, I have followed Huddleston ( 1984) and Quirk et al. (1985) for the identification of PDE modals. The Xs in bold type in this table signal the verbs which I attempt to describe in this section. All of them are OE central pre-modals, PDE modals and, with the exception of OE willan , all of them are preterite-present verbs. In the following paragraphs I describe this set of verbs along thre e different dimensions. Firstly, their morphological features. Secondly, t heir syntactic idiosyncrasy. And, finally, a semantic outline of these verbs. Chapter 3. Old English þ ur fan , be þ urfan, neodian and behofian 97 OE Pre- modal Preterite- present OE marginal modal PDE modal witan ‘ know ’ X agan ‘ possess, ought ’ XX 6 (X) 7 *dagan ‘ avail ’ X unnan ‘ grant ’ X gemunan ‘ remember ’ X cunnan ‘ can, know ’ XX X þ urfan ‘ need ’ XX X 8 *durran ‘ dare ’ XX X *sculan ’ owe, shall ’ XX X *motan ‘ must ’ XX X magan ‘ may, be able to ’ XX X willan ‘ wish, desire ’ XX habban ‘ have ’ X( X ) 9 be of necessity, obligation or future X (X) uton ‘ let(us) ’ ? 10 X onginnan ‘ begin ’ X Table 3.1: OE pre-modals, OE preterite-present verbs, OE marginal modals and PDE modals. Campbell (1959: § 767) classifies preterite-present verbs morphologically according to their original strong class. There are seven classes of strong verbs according to the Ablaut series, that is, according to the stem vowel in the following forms: infinitive, preterite singula r, preterite plural and past participle. In Old English, there are preterite-present verbs belonging to class I ( wat, ah, ), to class II ( deag ), to class III ( ann, cann, þ earf, dearr ), to class IV ( sceal, geman, be-, geneah ) and to class VI ( mot ). There is also a verb whose class is uncertain according to Campbell (1959: § 767), namely m æ g . Warner (1993: 142) considers that the group of pre-modals is coherent as early as Old English, since it is linked to st ative semantics, probably due to the perfect origin, which focuses on the state achieved. 6 As already mentioned, I follow Denison (1993: 295) in treating OE agan as a marginal modal, rather than as a central modal, a claim defended by Mitchell (1985: § 990). 7 Huddleston (1984: 165) states that ought “ is on the periphery of the class, ” the class being PDE modal verbs. Quirk et al . (1985: § 3.40) classify this verb as a “ marginal modal. ” This is the reason why brackets have been used in this category. 8 Huddleston (1984: 165) argues that both PDE need and dare may be main verbs, and Quirk et al . (1985: § 3.41, 3.42) agree in that “ for each there is also a ho momorphic verb (DARE, NEED) constructed as a main verb. ” 9 According to Huddleston (1984: 165), both PDE have and be “ in certain uses have some affinities with the modals. ” 10 Denison (1993: 318) poi nts out that for some scholars uton ‘ kn ow ’ i s a f o r m o f O E witan ‘ depart, die, ’ a rare strong verb, rather than a form of the preterite-present witan . Chapter 3. Old English þ ur fan , be þ urfan, neodian and behofian 98 Another special morphological feature of OE pre-modals concerns the possibility of contraction with the negative particle ne . This particle merges with certain common words, which have an initial vowel (e.g. ic nam < ne + am ), initial [w] (e.g. ic nylle < ne + wille ), or initial [h] (e.g. ic n æ bbe < ne + h æ bbe ), as noted by Warner (1993: 150-151). Present-Day English still exhibits a vestige of this phenomenon in willy-nilly , a contraction of ne + will (Denison 1993: 309). Although in a different order of elements (i.e. verb + negation), negative contractions are also exclusive of auxiliaries in Present-Day English (e.g. won ’ t ). A last morphological characteristic of OE pre-modals is their defective paradigm, since they rarely occur as non-finite forms. In fact, the infinitives of some of these verbs are not recorded (those asterisked in Table 3.1), and that is the reason why Campbell (1959), for instance, uses the third person singular to quote them. Therefore, these verbs are already defective in Ol d English, since the original infinitive is no longer valid for the new meaning they have acquired. Visser (1963-1973: §§ 1649-1651, 1684-1687, 1722-1723, 1839, 2042, 2134) provides some examples of this rare use of non-finite f orms of OE pre-modals, and Warner (1993: 145) offers a complete list of the recorded non-finite forms of these verbs in Old English. Therefore, the verbs under analysis in this section are, from a morphological point of view, quite different from the vast majority of OE verbs. Now I will examine their syntactic behaviour, in order to check whether they also differ from the main stream of verbs, in a way that can explain their characteristics in Present-Day English. In section 2.3.2.5, I described PDE modals as auxiliary verbs which have undergone a complex process of grammaticalization in the course of history. In the following paragraphs I will examine their ancestors in the OE period in order to determine their degree of grammaticalization at this early stage. OE pre-modals could be construed accordi ng to four different patterns: intransitively, as in (3.1), with a NP as object, as in (3.2), followed by an infinitive, as in (3.3), and, in the case of OE magan and willan , with a þæ t - complement, as in (3.4): (3.1) Eac neah þ an ealle þ a ð ing þ e ð anon Also nearly then all those things that thence cuma ð , wi ðæ lcum attre magon . come, against every po ison they-prevail Chapter 3. Old English þ ur fan , be þ urfan, neodian and behofian 99 ‘ But nearly all those things that are ex tracted from it can be used as antidote to any poison. ’ ( Bede 1 1.30.3) (3.2) …þæ t he geornor wolde sibbe wi ð hiene þ onne gewinn. ...that they rather wanted peace with him than conflict ‘… that they wanted peace with him rather than conflict. ’ ( Or 3 1 96.17) (3.3) … swa þæ t hi n æ fre ne mihton ne noldon sy ðð an fram his willan gebugan … so that they never not might nor not-wanted since from his will bend ‘ so that they never were able or wanted after th at to revolt from this will. ’ ( Æ CHom I, 1 12.7) (3.4) Deme ge nu, swa swa ge willon þæ t eow sy eft gedemed . judge you now, as as you wish (subj ) that to-you be afterward judged ‘ Judge now as you wish to be judged later. ’ (BlHom 5) 130) (examples and translations from Traugott 1992: 193, 194, 263) Therefore, OE pre-modals seem to behave as lexical verbs, at least in some contexts, namely when they are used absolutely or when they are complemented by an NP or a wh- finite complement clause. However, they are very commonly found with infinitives as complements. 11 This preference for the infinitive seems to point towards an auxiliary-like nature of these verbs in Old English, as Bolinger (1980, as cited in Hei ne 1993: 27) states: “ [t]he moment a verb is give n an infinitive complement, that verb starts down the road of auxiliariness. ” As for Old English, we must differentiate between the plain or bare infinitive (e.g. singan , ‘ to sing ’ ) and the inflected infinitive which is always preceded by to (e.g. to singenne, to singanne, to singan , ‘ to sing ’ ). The origin of the inflected infinitive is a prepositional phrase implying direction or purpose, by means of the preposition to followed by a dative. However, in Old English – enne and – anne are no longer datives (cf. Warner 1983: 200 ff.). The plain and the inflected infinitive are normally distributed as follows. The plain infinitive occurs with a few high frequency verbs, and, therefore, is very frequent in Old English. The inflected infinitive, on the contrary, is recorded with a wide range of OE verbs. However, the vast majority of OE verbs shows variation as for the choice of the infinitive, that is, most verbs 11 According to Visser (1963-1973: § 548), whether the pre-modal was complemented by a noun phrase or an infinitive is not a relevant feature, because, since infinitives were nouns, the relationship between them and the modal was the same as the relationship between the NP and the pre-modal. Chapter 3. Old English þ ur fan , be þ urfan, neodian and behofian 100 may select either the plain or the inflected infinitive. The only verbs which never occur with the inflected infinitive are cann, dearr, m æ g, mot, sceal, þ earf, uton and wile (cf. Warner 1993: 137). More specifically, Warner (1993: 138) observes that the inflected infinitive is preferred in constructions in which the (assumed) subject of the infinitive is the same as the one of the verb, as in PDE I want to go , as opposed to constructions in which the infinitive and the verb have different subjects, e.g. verbs of perception such as OE seon , ‘ to see, ’ as in I saw you go . Pre-modals are an exception, since, in spite of occurring in shared subject constructions, they only select the plain infinitive. 12 The fact that OE pre-modals select the plain infinitive exclusively seems to imply that this group has a distinctive syntactic feature which evokes their incipient auxiliariness as early as in Old English. There are, in addition, further pieces of syntactic evidence in favour of such an interpretation. OE pre-modals occur in a number of constructions which reveal their non- fully lexical character, that is to say, they prove to have undergone decategorialization, one of the four processes of grammaticalization mentioned by Hopper (1991) and Heine (1993: 58 ff.), which implies, among other cha nges, a reinterpretation of syntactic function (cf. section 2.3.2. 5). These frequent constructions impersonal and elliptical constructions. Impersonal constructions (see section 2.3 for a complete descriptio n) have been thoroughly studied in coocurrence with OE pre-modals, since the ability of pre-modals to occur with a non-nominative argument by influence of the accompanying infinitive reveals that their status is subordinate to the infinitive, which is the syntactic head (Denison 1990a; Warner 1993, among others). See, for example, (3.5): (3.5) For þ on ne þ earf nanne man tweogan, þæ t seo forl æ tene cyrice ne hycgge because not need no man (acc) doubt that the forsaken church not take-care ymb þ a þ e on hire neawiste lifg ea þ . about those that in her neighbourhood live ‘ Because no m an need doubt of this, that the forsaken church (will) not take- care for those that live in her neighbourhood. ’ (BlHom 41.36) (example and translation from Warner 1993: 123) In sentence (3.5), the third person singul ar of the pre-modal þ urfan , ‘ need, ’ namely þ earf , occurs together with the impersonal verb in the infinitive form, 12 Mitchell (1985: § 996) mentions the exceptional behaviour of agan (a marginal modal, according to Denison 1993: 295) which demands th e inflected infinitive, unless two or more infinitives are joined; in such cases, only the first one is inflected. Chapter 3. Old English þ ur fan , be þ urfan, neodian and behofian 101 tweogan , ‘ doubt, ’ and the experiencer of the construction, nanne man , ‘ no man, ’ is inflected for the accusative, as tweogan demands, instead of nominative, as was expected from þ earf . The fact that þ earf loses its usual way of marking the experiencer in favour of the s yntactic requirements of tweogan seems to imply that this OE pre-modal has less syntactic weight than the impersonal verb. In other words, þ urfan seems to function as an auxiliary verb. However, as Warner (1993: 132) points out, “ the fact that a verb may ‘ intervene ’ in a impersonal construction is in itself a poor argument f or its auxiliarihood [ … ]. We need a set of interrelated properties. ” That is, we cannot draw a conclusion on the basis of the fact that a group of verbs occurs characteristically in a give n construction, but further characteristics of that group are needed in order to evidence their status as a coherent group. Indeed, according to Warner (1993: 111-116), further syntactic features are common to this group, namely, elliptical and pseudo-gapping constructions. Elliptical constructions are those in which an element of a clause is elided. It is said that an auxiliary verb occurs in an elliptical construction when the omitted element is the infinitive which should occur after it, as illustrated in (3.6): (3.6) deofol us wile ofslean gif he mot. devil us will kill if he is-allowed ‘ (the) devil will kill us if he can. ’ ( Æ CHom i270.10) (example and translation from Warner 1993: 112) In this sentence, the pre-modal motan , ‘ be allowed, ’ is not followe d by the expected infinitive ofslean , ‘ to kill. ’ Instead, such an infinitive is elided and must be retrieved from the preceding context. The clause, therefore, may be said to be an instance of elliptical construction. Pseudo-gapping constructions may be considered a variant of elliptical constructions in which the infinitive is elided and the complement of the infinitive is retained and occurs after the auxiliary, as illustrated in (3.7): (3.7) We magon monnum bemi ð an urne ge ð onc & urne willan, ac we we may men (dat.) hide our thoughts (acc.) & our desires (acc.) but we ne magon Gode. not may God (dat.) ‘ We can hide from men [lit.: from-men hide] our thoughts and our desires, but we cannot [lit.: not c an] from God. ’ (CP 39.12) (example and translation from Warner 1993: 114) Chapter 3. Old English þ ur fan , be þ urfan, neodian and behofian 108 the semantic features of OE pre-modals concern the expr ession of their basic root meanings and, in some cases, temporal meanings. In addition to the variety of meanings which OE pre-modals may convey, it is also important to notice that, on occasions, they prove to be semantically empty, since they are accompanied by a synonymous non- pre-modal (cf. Beths 1999). Consider, for example, (3.15): (3.15) Hwa dear nu gedyrstl æ can, þæ t he derige þ am folce? Who dare now dare that he harm (subj) this people (d at) ‘ Who would now dare to harm these people? ’ ( Æ HomI vii, 306) (example, gloss and translation from Beths 1999: 1081) Sentence (3.15) contains the thir d person singular of the pre-modal * durran , ‘ to dare, ’ namely dear , followed by an infinitive, gedyrstl æ can , which also means ‘ to dare. ’ This is not an isolated exampl e (see Beths 1999 for more examples), and it might indicate that the meaning of the pre-modal is totally bleached, because otherwise it would be redundant to have two verbs expressing the same meaning in the same verbal unit. In ot her words, it is redundant to say who dar es to dare? , so the presence of the second dare must be due to the semantic voi d of the pre-modal * durran . Summarizing the features of the OE pre-modals, there is a cline from fully lexical to partially grammaticalized verbs. In some cases, the OE pre-modals may function as fully lexical verbs, keeping their original meaning, and occurring i n syntactic constructions in which t hey prove to be t he head. In other cases, the OE pre-modals behave as auxiliaries, since they lose their syntactic idiosyncrasy in favour of the infinitive which follows them, and they may be used to express root modality and temporal meanings. We observe, then, that as early as in Old English, some of the pre-modals have undergone two of the formal processes claimed by Heine (1993: 58 ff.) to be part of the grammaticalization chain, as seen above in section 2.1.3.1. Thes e two processes are, on the one hand, decategorialization, or change in morphosyntax, and, on the other hand, desemanticization, or change in semantics. Once I have examined the morphological, syntactic and semantic features of OE pre-modals (which, as repeatedly mentioned, include the preterite-present verbs plus willan ), I will have a look at the pre-modal verb that is the concern of this study, namely OE þ urfan , and at its derived verb be þ urfan . Chapter 3. Old English þ ur fan , be þ urfan, neodian and behofian 109 3.2.2. Preliminary approac h to Old English þ urfan and be þ urfan As already mentioned, OE þ urfan is a preterite-present verb belonging t o class III (Campbell 1959: § 767). Its possible forms are: 1 st sg. þ earf 2 nd sg. þ earft 3 rd sg. þ earf INDICATIVE Pl. þ urfon PRESENT SUBJUNCTIVE þ urfe(n), þ yrfe(n) PRETERITE þ orfte INFINITIVE þ urfan PRESENT PARTICIPLE þ earfende 15 Table 3.2: Forms of OE þ urfan (from Campbell 1959: § 767) Campbell also mentions that in the Rushworth Go spels the present participle form ð orfende and ð urfende are recorded. In addition to that, the Northumbrian dialect has a re-formed present ð orfe ð , used as a third person singular and pl ural. Finally, in the Lindisfarne Gospels the form ð urfu is also attested. The label þ urfan includes, in this piece of research, other semantically and morphologically similar verbs, namely, OE þ earfan, ‘ to need, suffer need, ’ þ orfan , ‘ to need, ’ and þ earfian , ‘ to be in need, ’ according to the respective entries in Bosworth and Toller. These verbs are clearly synonyms of þ urfan in the sense of Bosworth and Toller ( s.v. þ urfan v.), and they are morphologically similar as well. In fact, the main difference between these verbs and þ urfan is the radical vowel (<u>, <ea> or <o>). This is not, however, such a substantial difference, because it is possible for þ urfan to exhibit either of these vowels, as shown in Table 3.2. In addition, the present participle of þ urfan , namely þ earfende , is also the present participle of þ earfan (cf. Bosworth a nd Toller s.v. þ earfan ). Thus, the apparent difference between both ver bs seems to be neutralized in the present participle forms, which, furthermore, are usually adjectives meaning ‘ poor, needy ’ in both cases, according to the entries give n by Bosworth and Toller. The same signification is carried by the present participle of þ orfan , namely þ orfende , though this may also be used substantively ( cf. Bosworth and Toller s.v. þ or fend m.), and that of þ earfian , namely þ earfigend , as can be gathered from the onl y example of this verb offered by Bosworth and 15 According to Bosworth and Toller ( s.v. þ urfan v.), the present participle of þ urfan is þ urfende, þ yrfende . Although these forms will also be search ed for in the corpus, I list þ earfende as the most common form of the present participle of þ urfan in Table 3.2, f ollowing Quirk and Wrenn (1955: 57) and Cam pbell (1959: § 767). Chapter 3. Old English þ ur fan , be þ urfan, neodian and behofian 110 Toller (s.v. þ earfian , v.). The fact that the dictionary does not provide any personal form of the verb reveals that such forms are marginal in Old English. Therefore, although the initial panorama offered by Bosworth and Toller contains four different verbs, þ urfan, þ earfan, þ orfan and þ earfian , their semantic and morphological features do not prove to be so specific; rather, they converge at the same morphological and semantic point. The y may be morphologically referred to as þ *rfan verbs, and they all convey the meaning ‘ need. ’ In addition, the fact that these verbs seem to merge in Middle English under the verb thurve n ( MED , s.v. thurven , v.) seems to indicate that they are close relatives. For all these reasons, in the analysis of the corpus data, the label þ urfan will include not only the verb referred to in the lexical entry of s uch a word in the dictionary, but also its morphological variants þ earfan, þ orfan and þ earfian . Having made these morphological clarifications, it must be noticed that, as far as the syntax of þ urfan is concerned, it behaves much like PDE modal ne ed , that is, it is essentially restricted to non-affirmative co ntexts, interrogative and negative (cf. Denison 1993: 295). The syntactic pattern in which it occurs depends on the meaning it conveys in each example. For this reason, I will start by analysing the meaning of this verb following the semantic indications found in Bosworth and Toller ( s.v. þ urfan v. I-III). In this dictionary we can observe that OE þ urfan may ha ve three different meanings: 16 1.- ‘ to be in need, have need of something. ’ In this sense, the verb coul d be used: - absolutely (3.16) Gif ð u cla þ a þ e ma on h æ fst, þ onne ð u þ urfe . if you (nom) clothes (gen.) part. more on have (2 sg) than you need (sg. subj.) ‘ If you have more clothes than you need. ’ - with genitive of thing needed (3.17) ne ð u mines þ earft . neg. you (nom.) mine (gen.) need (2 sg) ‘ You don ’ t need mine. ’ - with accusative (3.18) Mu þ a gehwylc mete þ earf . mouth each meat (acc.) needs 16 Unless otherwise stated, glosses and translations are mine. Chapter 3. Old English þ ur fan , be þ urfan, neodian and behofian 111 ‘ Each mouth needs m eat. ’ 2.- ‘ to need to do something. ’ Obviously, in this sense, þ urfan will be followed by an infinitive. The nuances that describe the necessity m ay be of different types: a.- “ where a want has to be satisfied, a purpose accomplished or the like. ” (3.19) Hi witan hw æ r hi eafiscas secan þ urfan . they know (3 pl) where they river-fish (acc- pl.) seek need ‘ They know where they must seek the river-fish (if they are to find them). ’ b.- “ where the need is based on the grounds of right, fitness, law, morality, etc., to be bound to do something because it is right, etc. ” (3.20) Gif he gewitnesse h æ bbe, ne þ earf he ðæ t geldan. if he knowledge has neg need (3 sg) he that pay ‘ If he knows that he is not bound by law to pay. ’ c.- “ with the idea of compulsion, or where the inevitability of a consequence is expressed; in some cases the word might be taken almost as an auxiliary, of much the same force as shall : to be obliged, be compelled by destiny. ” (3.21) Ge ne þ urfon her leng wunian. you (pl) neg need (pl) here long dwell ‘ You shall not be obliged to stop here any longer. ’ d.- “ to have a good cause or reason for doi ng something. ” (3.22) Ne þ earf he gefeon. neg need (3 sg.) he rejoice ‘ He has no reason to rejoice. ’ e.- “ where the need arises from an advantage to be gai ned, or purpose to be served, to be use, to be good for a person to do something. ” (3.23) Ne þ earf ic yrfestol bytlian. neg need (1 sg) I hereditary seat build ‘ It is no good to me to build an hereditary seat. ’ 3.- to owe (cf. * sculan ) (3.24) Ne þ ear ic N. sceatt ne scilling. neg need (1 sg) I N. property neg shilling ‘ I owe no property of money. ’ As can be observed, the semantics of OE þ ur fan does not include any epistemic meaning. However, the information in the dictionaries contrasts with Borgenstierna ’ s (1988) findings. In her monographic work devoted to the verbal expression of modality in Old English, s he comes to the conclusion that OE þ urfan is rarely found expressing other meanings than epistemic. This assertion Chapter 3. Old English þ ur fan , be þ urfan, neodian and behofian 112 strikingly contrasts with the infor mation offered in section 3.2.1, where we have seen that there is a general agreement on the absence of gra mmaticalized epistemic meanings in Old English. The reason for the mismatch between Borgenstierna ’ s conclusions and t he other scholars ’ lies on her criterion for the identification of epistemic meanings. She consi ders that the fact that OE þ urfan is often followed by verbs of thought (e.g. think, doubt, consider , etc.) implies that this pre-modal has epistemic values. However, I am of the opinion that the carrier of the epistemic weight is the infinitive and not þ urfan . Therefore, the most frequent meanings of þ ur fan range from ‘ need something ’ and ‘ need to do s omething, ’ up to ‘ to be obliged or compelled. ’ Consequently, the modal meanings expressed by OE þ urfan s e e m to be r e s t ri ct ed to root necessity, either weak, or strong, i.e. ‘ obligation. ’ If we take into account that this verb used to appear in non-affirmative contexts, both meanings will converge into the more general meaning of ‘ lack of obligation. ’ In addition, OE þ urfan may also be a synonym of OE * sculan in its basic ‘ owe ’ meaning. Syntactically, OE pre-modal þ urfan may be used abs olutely. This is not to be confused with apparent cases of post-verbal ellipsis, because, as mentioned above, the possibility of a pre-modal to occur in absolute uses is one of the exceptions for the existence of syntactic ellipsis, as mentioned by Warner (1993: 113-114). It may also be followed by an NP or by an infinitive, that is, it may have a nominal or an infinitival theme. The wide variety of possible semantic nuances of OE þ ur fan when it has an infinitival theme seems to reveal that this is the most frequent type of c onstruction for this OE pre- modal. The preference for an infinitival theme seems to suggest that OE þ urfan has auxiliary status. In fact, this pre-modal verb occurs also in impersonal constructions of the t ype mentioned above, in which the pre-modal loses its syntactic characteristics and adopts those of the impersonal infinitive (cf., for instance, Warner 1993: 125). Consider, for example, (3.25): (3.25) þæ t us (DAT) þ onne ne ð urfe sceamian . that us then not need be ashamed ‘ that we need not be ashamed. ’ (Foerst VercHom 9) (example and translation from Allen 1997: 15) In this example, the non-nominative experiencer us occurs, instead of the nominative we , because þ urfan adopts the synta x of sceamian , ‘ to be ashamed. ’ OE þ urfan , therefore, has lost its syntactic weight in this construction in fa vour Chapter 3. Old English þ ur fan , be þ urfan, neodian and behofian 113 of the impersonal infinitive which follows it, which can be analysed as a case of decategorialization. It may be said, therefore, that its function is that of an auxiliary, as we saw above in example (3.5). Further evidence of this auxiliary status is seen in the fact that þ urfan may occur with a passive construction, and it, therefore, becomes a “ sentence modifier ” (in Warner ’ s 1993: 160 terms). Consider, for example, (3.26): (3.26) ac witodlice þæ t gesegen beon m æ g, ne þ earf þæ t ben gelyfed. but indeed what seen be can not need (3 sg.) that be believed ‘ but indeed what can be seen [lit.: seen be can] does not need [lit.: not needs that] to be believed. ’ (GD 269.15) (example and translation from Warner 1993: 161) Warner (1993) co nsiders that the pre-modal þ earf functions as a sentence modifier in this example. That is, here þ urfan acts as most PDE modals, since it does not mark a relationship between the subject and the object (the experie ncer and the theme). In a sentence such as John drinks water , the verb drink mediates between the subject John and the object water . In addition, drink always selects its subjects, that is to say, neither the noun water nor the pronoun what may occur as subjects of drink . PDE modal verbs, however, do not select their subjects, because they do not mediate between them and the object; they are merely sentence modifiers. This is what Warner (1993) claims for examples such as (3.26), where þ urfan is just modifyi ng the sentence in which the main verb occurs in the passive voice. In spite of these auxiliary-like features of OE þ ur fan , it must also be noted that þ urfan is not expected in contexts involving pseudo-gapping, because, as mentioned by Warner (1993: 133-134), this verb, together with OE cunnan, *durran and motan , is not attested in such a construction. In the analysis of the corpus data, however, we will see the frequency of each of the possible constructions and meanings of this OE pre-modal, and, t herefore, check the accuracy of the information found in the literature. The second preterite-present verb expressing necessity which will be part of my study is OE be þ urfan , a verb derived from þ urfan by means of the OE prefix be- . This is one of the most common OE verbal prefixes, which in stressed contexts may be realized as bi - (cf. Kastovsky 1992: 379). It may ha ve the following effects on the verb: (a) transitivization, i.e., it may make an intransitive Chapter 3. Old English þ ur fan , be þ urfan, neodian and behofian 114 verb transitive (e.g. feohtan , ‘ fight, ’ befeohtan , ‘ take by fighting ’ ); (b) intensification, i.e., it intensifies the meaning of the original verb (e.g. brecan , ‘ break, ’ bebrecan , ‘ break to pieces ’ ); and, finally, (c) it may not change the meaning of the verb at all (e.g. beodan, bebeodan , ‘ offer, announce ’ ), as mentioned in Kastovsky (1992: 379). The fact that be -, together with other OE prefixes (e.g. a-, ge- ), may not have any semantic effect on the verb, and the subsequent existence of two synonymous verbs which may alternate in the same text leads to an inevitable decay of these prefixes in Middle English (c f. Kastovsky 1992: 377). It will be interesting to check in the corpus data if the frequency of be þ urfan is even throughout the Old English period, or if, as could be expected, it decreases along the period. The forms of be þ urfan are those listed in Table 3. 2 for þ ur fan , though with an initial be - or bi - . There are not any morphological variants alternating the radical vowel, as was the case with þ urfan . Semantically, both þ urfan and be þ urfan are synonyms in their basic meaning ‘ to need, to be in need ’ (cf. Bosworth and Toller, s.v. be þ urfan , v., and þ urfan , v.). However, as far as syntax is concerned, the information we obtain fr om Bosworth and Toller ( s.v.be þ urfan , v.) reveals that the use of both verbs is somewhat different. The prefix be- seems to reduce the range of possible syntactic constructions where this verb may appear, especially as compared to the ample variety of constructions available for OE þ urfan . The examples provided in the An Anglo-Sax on Dictionary share a particular characteristic: in no case has be þ urfan a sentential theme. Instead, genitival noun phrases are found, sometimes preposed and sometimes postposed to the verb. Consider, for instance, (3.27) (3.27) Wisdomes be þ earf. wisdom (gen. sg.) requires ‘ he requires wisdom. ’ A different interpretation for sentences like this one is suggested by Krug (2000: 123). In his opinion, OE be þ urfan may be a synonym to OE þ olian when constructed with a genitival theme, meaning ‘ to lose, lack. ’ I have not found the meaning ‘ to lack ’ in any of the entries of be þ urfan in the two Old English dictionaries used for this study (B osworth and Toller, and Clark Hall). Howe ver, it is undeniable that there is a direct relationship between the meanings ‘ to lack ’ and ‘ to need. ’ This information will be checked in the analysis of the examples from the corpus, in section 3.4.1. Chapter 3. Old English þ ur fan , be þ urfan, neodian and behofian 115 3.3. Preliminary approac h to Old English neodian and behofian In this section, I will provide a description of the syntactic features of OE neodian and behofian , as found in the relevant literature. 3.3.1 deals with neodian , and offers and explanation of the complex range of ve rbs which are analysed under this form. In its turn, 3.3.2 concentrates on behofian , with s pecial reference to its claimed impersonal nature. 3.3.1 Ol d English neodian As mentioned, the aim of this section is to analyse the syntactic features of neodian , which basically concern its impersonal c haracter. However, before undertaking this task it is necessary to clarify and also justify the forms that I have decided to include under the form neodian . The editors of the OED state that the etymological predecessor of the PDE modal and non-modal verb nee d is OE neodi an (cf. OED s.v. need v.2), and, therefore, all possible forms of that verb are to fall under the scope of my analysis. In addition, Old English has another ver b neodi an (cf. OED s.v. need v.1), which is said to mean ‘ compel, force, urge. ’ In order to find and analyse all the possible examples of the etymological predecessor of PDE need meaning ‘ to be necessary ’ or ‘ to need, ’ I opted to analyse all other possible variants of OE neodian , based on the morphological inf ormation provided i n the An Anglo- Saxon Dictionary edited by Bosworth and Toller, as is explaine d below. From a morphological perspective, OE neodian (cf. Bosworth and Toller, s.v. neodian, neadian ) is an OE weak verb class 2, since it exhibits the characteristics of this class of verbs, as opposed to weak verbs class 1: the absence of an i -mutated vowel and of a geminated consona nt in the stem, and the presence of /i/ in the infinitive (cf. Hogg 1992c: 157-162). As is well-known, verbs belonging to th e weak class are derived from other lexical items of the language. Therefore, we expect neodian, neadi an to be derived from a noun such as neod or nead . In Old English there actually exists a noun neod , 17 for which Bosworth and Toller ( s.v. neod , n.) provide two entries: 17 In fact, the construction of this noun together with the verb beon/wesan or habban , is very frequent in OE as an expression of necessity (meaning ‘ it is necessary ’ or ‘ have need ’ ). This type of construction will not be considered in this paper, since my purpose is to account for verbs exclusively. However, it is my intention to analy se, in future research, constructions such as neod beon / neod habban and þ earf beon / þ earf habban , since the noun þ earf , from the pre- Chapter 3. Old English þ ur fan , be þ urfan, neodian and behofian 116 Neod, ned, nied, nyd f. ‘ desire, eagerness, diligence, earnest, endeavour. ’ Neod (=nead) ‘ necessity. ’ According to Bosworth and Toller (s.v. neod(=nead) ), there seem to be no formal distinctions between both words. In addition, the dictionary refers us to the entry nid , which exhibits the alternative spellings nead, ned, neod, nied and nyd , and a series of possible meanings, from ‘ necessity, inevitableness ’ to ‘ difficulty ’ or ‘ compulsion. ’ Therefore, there are three OE nouns, namely neod, nead and nid whose meanings refer to ‘ necessity. ’ It is foreseeable, then, that Old English also exhi bits three weak verbs derived from these three nouns, verbs such as neodian, neadi an and nidan . This is actually the situation we find in Old English, and we also find variants from these verbs according to the different spellings of the nouns from which they derive, that is, we may find verbs such as nedan, niedan or nydan . Also, as shown in the Bosworth and Toller dictiona ry, these verbs may ha ve ge- variants, that is, variants beginning with the prefix ge- . For the sake of clarity, Table 3.3 summarizes the different spellings of the se necessity weak verbs, as related to the nouns from which they derive. In Table 3.3 we see that the meanings of the verbs in the right hand column may be easily divided into two groups: one group c ontaining those verbs expressing ‘ be necessary ’ and the like, and a second gr oup containing those verbs expressing ‘ force, compel ’ and related meanings. The meaning of both gr oups falls within the scope of necessity as understood in the force-dynamic conception of modality described in section 2.2, that is, as analysed in terms of forces and barriers. Thus, when something is necessary (in the sense of 2a neodi an , for example), there is some force for it to exist or to be present. In a similar line, when a person compels or forces somebody to perform an action (i n the sense of 3 neadi an or 5a nidan , for instance), such a person (i.e. the antagonist) exerts some kind of force on the other person (i.e. the agonist). In other words, we may say that the verbs in the latter group, meaning ‘ force, compel ’ (listed as 3, 5a, 7 and 8 in Table 3.3) are causative, while the verbs in t he former group, meaning ‘ be necessary ’ (listed as 2a, 4 and 6) lack such a nuance. modal verb þ urfan is also comm on in OE, and also ned þ earf beon / habban , showing a combination of both stem s (cf. Taeymans 2004b). Chapter 3. Old English þ ur fan , be þ urfan, neodian and behofian 117 NOUN + MEANING VERB + MEANING 1) Ne ó d, n é d, ni é d, n ý d ‘ desire, eagerness, diligence, earnest endeav our ’ 2) Ne ó d (= ne á d ) ‘ necessity. ’ See passages under n í d 2a) Ne ó dian ‘ to be necessary ’ , v. ne á dian No entry for ne á d 3) Ne á dian (v. n í d VI) ‘ to force, compel, constrain ’ 4) Ne á dian, ne ó dian (v. n í d IV) ‘ to be necessary ’ 5) N í d, ne á d, n é d, ne ó d, ni é d, n ý d (cf. 2) above) I ‘ necessity, inevitableness ’ II ‘ necessity, need, urgent requirement ’ III ‘ a necessary business ’ , duty ’ IV ‘ need, what one wants ’ V ‘ necessity, need, difficulty, hardship, distress ’ VI ‘ force, compulsion ’ VII ‘ the nam e for the rune ’ 5a) N í dan ‘ to force, compel, urge ’ No entries for nom inal ge -forms 6) Ge-ne ó dian ‘ to be need ’ I. of something for a person II. that something be done 7) Ge-ne á dian ‘ to compel ’ 8) Ge-n é dan, -ni é dan, -n ý dan ‘ to compel, force, urge ’ Table 3.3: Entries for neodian, neadian and related items in Bosworth and Toller (1898). 18 Given the above-mentioned difference be tween the two groups of verbs, it could be argued that each group of verbs should be dealt with separately. However, I have embraced them all unde r the label neodian for a series of reasons. Firstly, these weak verbs can all be said to ultimately derive from the very same noun, namely OE nid (listed as 5 in Table 3.3). Secondl y, in Old English the verbs neodian (2a) and neadian (4), on the one hand, and neadian (3) and nidan (5a), on t he other hand, are synonymous, which makes ne adian a polysemous verb ( ‘ to be necessary ’ and ‘ t o force, compel, urge ’ respectively). Thirdly, both meanings, ‘ need, to be necessary ’ and ‘ c ompel, force, ’ fall under the scope of the force-dynamic interpretation of modality followed in this study (cf. section 2.2.2.2 above). Finall y, in Middle English, all three ve rbs yield the 18 The entries provided in Table 3.3 are taken from An Anglo-Saxon Dictionary , both from the edition by Bosworth & Toller (1898) and from the appendix by Alistair Campbell (1972). Chapter 3. Old English þ ur fan , be þ urfan, neodian and behofian 124 The next task was to randomize the list of texts, so that the sample is not biased by any external influence, such as, for example, the criteria followed by the editors when orde ring the texts in t he 77 files. The electronic tool used to randomize the texts was Microsoft Excel, 22 which produced a randomly ordered list of all the Old English texts which appear in the Dictionary of Old English Corpus . The selection of the texts to be include d in my corpus was then safely extracted from this list, where the order of the texts is not biased by any factor. A preliminary overview of the list of the texts revealed that the chronological distribution of the Old English texts was extremely uneven. As is well-known, the OE period is traditionall y divided into early and late Old English. The definition of the term “ earl y Old English ” is not uncontroversial. The Cambridge History of the English Language (Hogg 1992a) follows the traditional distinction between ea rly and late Old English. Accordi ng to it, early Old English or Alfredian Saxon refers to the language produced before 950, or, in other words, the language used in the texts written in the court of King Alfred, who ruled from 871 to 899 (cf. Hogg 1992b : 6). After that date, and due to the standardization introduced by Æ thelwol d (abbot of Abingdon from 954, and bishop of Winchester from 963), the orthog raphy changed considerably up to the point that “ there is no direct chronolo gical line of descent between Early and Late Old English ” (Hogg 1992c: 83- 84). Therefore, according to The Cambridge History of the English Language , the main representative of early Old Englis h is King Alfred, while Æ lfric, one of Æ thelwold ’ s pupils and abbot of Eynsham, is the most outstanding figure of late Old English (cf. Hogg 1992c: 78). The compilers of the Dictionary of Old English Corpus s e e m t o a g r e e o n the delimitation of late Old English, since they label as early Old English all texts dated before 950. However, as for early Old Englis h, they consider not onl y those texts written at the court of King Alfred, but also all other written texts, such as old runic inscriptions. As a consequence, in this study I will consider that early Old English includes all the English texts written before 950, and that late Old English comprises those texts written between 950 and 1150. For historical reasons, the number of extant texts dating from early Old English is much more restricted than those of late Old English. For this reason, I decided to extract all such texts, with the exception of those which appear in t he Helsinki Corpus and 22 The procedure is the following. The list of texts must be pasted in Column A in a new document of Microsoft Excel. In Column B, a formula m ust be applied so that a random number from 0 to 1 appears in each cell. Then the list is ordered according to the number in Column B. Chapter 3. Old English þ ur fan , be þ urfan, neodian and behofian 125 those which belong to Latin-Old English Glossaries, since such material does not provide a running context for a linguistic study such as the one I intend to carry out. The routine followed in the extraction of the texts is the following. Using the Cameron Number I located each text within the 77 files of the Oxford Text Archive , 23 with the aid of the Index t o the Old English Cor pus (OTA), whic h I downloaded from <http://www.georgetown.edu/faculty/ballc/englisc/oecorpus- index.html>. Then I opened the relevant file, searched the text, copied it, and pasted it on a new Microsoft Word document. After counting the words in each text the document was saved as a TXT file. 24 This pr ocess was repeated as many times as necessary up to completing the number of words needed for my 1. 2 million word corpus. As for the chronological distribution of texts, I extracted 140 texts of early Old English, which come up to 153,802 wo rds. Added to the early OE words in the Helsinki Corpus , we obtain a total of 248,042. Therefore, the totality of the extant early Old English texts (-950) is included in my corpus. That leaves us with a list of texts containing wor ks dating from late Old English (950-1150) and also wor ks whose date is indeterminate. In my selection of the 1.2 million words I needed to represent OE, I decided to include only texts whose date of composition was known, since this is a diachr onic study which aims at describing the chronol ogical evolut ion of some verbs. Therefore, I left out the texts which cannot be classified as belonging to either early or late Old English. My goal, therefore, was to obtain a randomly selected list of late Old English texts which would be representative of that period. A total of 373 texts were included in my corpus, making a total of 638,603 words, which added to the late Old English material present in the Helsinki Cor pus comes up to 957, 613 words. Therefore, my Old English corpus contains 1,205,655 words chronologically distributed as shown in the following t able: 23 I acknowledge my debt of gratitude to Dr. Alejandro Alcaraz for his invaluable help and generosity in providing me with the files extracted from the Oxford Text Archive formatted by himself. 24 DOC files have a nu mber of drawbacks when working with the concordance program Wordsmith Tools. One drawback is that Wordsmith Tools does not provide a complete wordlist from the DOC files. Another disadvantage of the DOC files is that Wordsmith concordance repeats some of the examples, when they occur at the beginning of the file. The TXT files, however, yield accurate data when working with Wordsmith Tools. Chapter 3. Old English þ ur fan , be þ urfan, neodian and behofian 126 EARLY OLD ENGLISH (-950) LATE OLD ENGLISH (950-1150) TOTAL Helsinki Corpus 94,240 319,010 413,250 Dictionary of Old English Corpus 153,802 638,603 792,405 TOTAL 248,042 957,613 1,205,655 Table 3.4: Number of words per corpus and OE subperiod . The examples retrieved from this corpus have been analysed according to morphological, syntactic and semantic va riables. All in all, the following 43 variables have been taken int o consideration:  GENERAL INFORMATION 1) Code. Both the HC and the DOEC codify the texts they contain accordi ng to some parameters, such as author or title. 2) Example. It contains the example in the original language with a large enough linguistic context to interpret the verb under scrutiny. 3) Translation. The example is transl ated into Present-Day English. 4) Name of the text. 5) Author. 6) Subperiod. Though the information given i n the Helsinki Corpus is very specific (namely, O1 (-850), O2 (850-950), O3 (950-1050) and O4 (1050- 1150)), the DOEC subdivides Old English only into early and late OE, and this is the subperiodization used in this study. 7) Dialect: The HC provides the following dialectal information A (Anglian), AM (Anglian Mercian), AN (Anglian Northumbrian), K (Kentish), WS (West Saxon). The DOEC , on the other hand, only classifies examples as Saxon (which roughly corresponds to West Saxon), Anglian (including Mercian and Northumbrian) or unknown. 8) Verse or prose. Verse may favour the particular syntactic constructions, which might concern the verbs under study. 9) Text-type. This variable will allow for generalizations as regards the use of some of the verbs in certain text-types. 10) Contemporaneity between the original text and t he manuscript which is kept nowadays. This information may be relevant for the interpretation of anachronistic constructions, becaus e it may reveal that the copyist changed some aspects of language influenced by the langua ge of the time Chapter 3. Old English þ ur fan , be þ urfan, neodian and behofian 127 when he copied the document. Unfortunately, this piece of information is only provided in the HC . 11) Relationship to foreign original (gloss, translation, etc.). It may be the case, as will be seen below, that the foreign language in which a text is originally written may bias the translat or or glossator towards the use of a given form or construction. 12) Foreign original (e.g. Latin).  LINGUISTIC INFORMATION MORPHOLOGY 13) Archi-verb: 25 þ urfan, neodi an, be þ urfan and behofian . For example, the archi-verb þ urfan stands for the orthogra phical variants þ urfan / ð urfan , the archi-verb neodian stands for neodian / neadian / nydan, and so on. This label is useful when dealing with orthographical variants of the same verb. 14) Verb. The base form of any of the orthographical variants mentioned above (e.g. ð urfan, nedan, nydan ). 15) Verb form. The actual form found in the corpus (e.g. þ earf, nedde, behofa ð ). 16) Person/number. 17) Tense/mood. Tense is a crucial aspect when dealing with modal verbs, because changes in tense may entail changes in the subjectivity conveyed by the verb (cf. Sanders and Spooren (1997: 103 ), as quoted by Mortelmans 2003). 18) Voice. This field includes i nformation as regards the voice of the verb, as well as the voice of the infinitive accompanying the modal verb. The latter proves crucial for the interpretation of the status of the modal verb as being more or less grammaticalized (cf. Warner 1993: 160). 25 The label “ archi-verb ” is mine, and it is used he re to refer to the basic orthographical form of each of the verbs under analysis. I have selected this label based on linguistic labels such as archi-phoneme , which, as stated in the OED ( s.v. archiphoneme n.) refers to “ A phonological unit comprising the totality of distingui shable features common to two or more phonem es. ” In the same sense, it could be said that an “ archi-verb ” is a verbal unit comprising the totality of distinguishable features common to two or more verbal variants. Chapter 3. Old English þ ur fan , be þ urfan, neodian and behofian 128 SYNTAX 19) Complement or theme. This variable only states the presence or absence of a complement or a theme, and, therefore, the absolute uses of the verbs as opposed to other uses. 20) Type of complement or theme. Here I specify the nature of the complement or theme of the verb, namely, transitive uses or nearly auxiliary function when the complement or theme is an infinitive. 21) Complement verb (if any). The information contained in this variable is important when dealing with grammaticalization, more specifically with semantic bleaching, because if a verb is followed by an infinitive with the same meaning, for example, this implies that the former has lost part of its semantic weight. 22) Main / subordinate clause. 23) Matrix verb. I specify the verb which occurs in the matrix clause (if my verb occurs in a subordinate clause), because it may also be revealing of semantic bleaching or of a certain degree of subjectivity. 24) Negation. Negation is recorded for it can be a marker of subjectivity (cf. Langacker 1991: 134; Mortelmans 2003). 25) Type of negation: not, nowhere, raising, etc. 26) Scope of negation. When negation occurs in a const ruction with two or more verbs, it may affect any of them, and it, therefore, may have different semantic implications. 27) Experiencer-verb c onstruction. As already mentioned, my verbs are prone to occur in this kind of construction. Moreover, impersonality may also be a marker of auxiliarization. 28) Allen ’ s (1995) Type. The possibilities of this variable are those mentioned by Allen (1995), that is, Types N, I, II, when the theme is a noun phase, and S, hit and ‘ personal, ’ when the theme is a clause (cf. 2.3.2.3 a bove). 29) Dummy subject. It specifies the presence of a dummy hit subject, or any other dummy subject (e.g. there ). 30) Experiencer case: nominative, oblique, prepositional phrase, etc. 31) Cause / theme. It specifies the case for which the thing needed is inflected. Chapter 3. Old English þ ur fan , be þ urfan, neodian and behofian 129 32) Interrogative. As was the case with negation, interrogative clauses are non-affirmative contexts which are generally considered to be more subjective than declarative clauses. 33) Time reference. One of the PD E features of modal verbs is the abnormal time reference, i.e. a past form need not refer to a past event, but it may just imply some kind of remoteness. SEMANTICS 34) Modality. As repeatedly mentioned, the types of modality distinguis hed in this study are root and e pistemic. 35) Subject force. From a cognitive point of view the subject of a verb expressing necessity may be ag onist or antagonist (cf. section 2.2.2. 2 above). 36) Force. The type of force implied in the i nteraction between agonist and antagonist may be of different types: general, legal, religious, inner (with split of the agonist), and so on. 37) Strength. The force exerted by the antagonist may be strong or weak. 38) Origin. The origin and the strength of the force exerted on the agonist oscillate in a scale from strong external to weak internal (cf. section 2.2.2.2). 39) Force not to. Depending on the scope of the negation, a given ve rb may imply a force not to (or a prohibition), as in you must n ’ t do that , of l ack of force (of absence of obligation), as in you needn ’ t do that (variable 37). 40) Lack of force. 41) Translation of the verb. This is a rudimentary variable which facilitates the search for a given example or set of examples. 42) Animacy of the experiencer / subject. In origin, the meaning of necessity is restricted to human beings who might need something; when non human or non animate experiencers start to occur with these verbs, we might think of a higher degree of grammaticalization (cf. Heine et al . 1991: 156; Krug 2000: 90; Mortelmans 2003). 43) Modality somewhere else. The presence of two or more modality markers such as subjective hedges (e.g. I think, I suppose ) m ay underline the subjective character of some modal verbs (Mortelmans 2003). Also, with the help of this variable we may retrieve examples in which our modal verbs appear close to other semantically similar modal verbs. Chapter 3. Old English þ ur fan , be þ urfan, neodian and behofian 130 For the analysis of this corpus, I have resor ted to the computer program Wordsmith Tools , which has pr oved very suitable for a number of reasons, such as its ability to obtain a longer context for each exampl e. After having scrutinized more than 1,500 potential examples (see Appendix I below for details), the number of occurrences of each verb expressing necessity in the OE corpus is the following: 26 VERB NUMBER N.F. % Þ URFAN 158 (13.11) 46.61% BE Þ URFAN 47 (3.89) 13.86% NEODIAN 104 (8.62) 30.68% BEHOFIAN 30 (2.48) 8.85% Total 339 (28.12) 100% Table 3.5: Frequency of each verb in the OE corpus. Table 3.5 also shows, in brackets, the normalized frequencies for 100,000 wor ds and, finally, the percentage which each of the verbs represents in the total number of occurrences. These examples have been introduced int o a Microsoft Access computer database, and later analysed as regards the feat ures mentioned. The findings will be illustrated in this study with examples which will take the following form: Ne þ urfan we us ondr æ dan þ a deoflican costnunga . not need (pl) we us fear the devilish temptations (acc) ‘ We need not fear devilish tem ptations. ’ (2.524 helsinki\coaelet3) or ( Æ Let 2 (Wulfstan 1), 96) The OE example is followed by the gloss and the translation into Present-Day English; finally, I include in brackets the codification used in each of the cor pora I have analysed. If the example is taken from the Helsinki Corpus , the code contains the following information. The number refers to the location of the verb form (in this case þ urfan ) within the text in whic h it occurs. The code of the t ext is specified at the very end of the bracketed codification; in this case, coalet3 , which is the convention used in the Helsinki Corpus to refer to the text Let Wulfstan 1 , by Æ lfric (cf. Kyt ö 1991). 26 As already mentioned, in Old English there exist other linguistic means to express the same kind of necessity of my verbs, namely constructions consisting of the nouns þ earf, neod or ned þ earf in combination with the verb beon / wesan , or hab ban , as in Him is þ earf / Him is neod / Him is ned þ earf þæ t... ‘ it is necessary for him that... ’ (cf., among others, Taeymans 2004b). Chapter 3. Old English þ ur fan , be þ urfan, neodian and behofian 131 If, on the other hand, the example is retrieved from the Dictionary of Old English Corpus , the codification differs. The information provided contains the short title of the text in which the example occurs ( Æ Let 2, in this example), the author, the editor or the recipient of the text (Wulfstan is the recipient of Æ lfric ’ s letter, in this case), and, finally, a series of numbers which refer to the location of the segment within the text, which may be the line, par agraph, or page of the text (line 96, in this instance). In what follows, I will deal separately with each of my verbs. Section 3.4.1 deals with OE þ ur fan and OE be þ urfan . Section 3. 4.2 is devoted to the analysis of OE neodian . Finally, section 3. 4.3 concentrates on OE behofian . 3.4.1 Ol d English þ urfan and be þ urfan in the corpus These two verbs are examined in the same section, because, as mentioned in section 3.2.2, be þ urfan is derived from þ ur fan , and it will be interesting to see up to what extent their morphology makes them differ semantically and syntactically. OE þ urfan is much more frequent (46.61 %) than be þ urfan (13.86 %) and this is so both in early and late Old English, as seen in Table 3.6 27 which displays the actual number of occurrences of each of these verbs together with the normalized frequencies calculated for hypothetical 100,000-word s u bperiods, i.e. for early (O1 and O2 in the Helsinki Corpus ) and late (O3 and O4 in the Helsinki Corpus ) Old English: O1-O2 N.F. O3-O4 N.F. TOTAL N.F. Þ URFAN 48 19.35 110 11.49 158 13.11 BE Þ URFAN 8 3.22 39 4.07 47 3.89 TOTAL 56 22.57 149 15.56 205 17.00 Table 3.6: Distribution of OE þ urfan and be þ urfan by subperiods. The normalized frequencies reinforce the evidence provided by the sheer number of occurrences of each OE verb. As for the frequency of each of the verbs, Table 3.6 shows that þ urfan undergoes a slight dec rease at the end of the period, while 27 The dating of the OE texts is fairly ambiguous in the Helsinki Corpus . Many of the texts selected by the editors of this corpus are tagged as, for example, O2/3, which i mplies that the original text was written in O2, and that the copy used by the compilers of the editors dates from O3. In cases such as these, I have considered the text as belonging to subperiod O2. On other occasions, the texts are tagged as OX/3, which i mplies that, although the original text is of unknown date, the copy used by the compilers of the corpus dates from O3. In these cases, I have considered the text to be from O3. Chapter 3. Old English þ ur fan , be þ urfan, neodian and behofian 132 be þ urfan exhi bits practically the same frequency. Let us now turn t o the analysis of these verbs paying attention to their semantics first, and to their syntactic features later. 3.4.1.1 Semantic features of Old English þ ur fan and be þ urfan In this section I will analyse the meanings conveyed by OE þ ur fan and be þ urfan , in order to see the extent to which they represent the meanings conveyed by PDE need . We will also observe how the meanings of þ urfan and be þ urfan overlap, which opens the way to the semantic analysis of the other OE verbs. As seen in section 3.2. 2, Bosworth and Toller ( s.v. þ urfan v.) suggest three main meanings for þ urfan : ‘ to be in need/have need of something, ’ ‘ to need to do something ’ and ‘ to owe ’ (the latte r being nearly synonymous with *sculan , ‘ be obliged ’ ). These meanings are related to the notion of necessity, and that is, indeed, the case of 157 examples of þ urfan in my corpus, while in one sentence þ urfan conveys possibility or, rather, absence of possibility. Consider example (3.34): (3.34) Þ urh so þ e bireousunge þ eo soule reste onfo þ . Ac ne þ earf ic nefre resten through true mercy the soul ’ s rest begins but not need I never rest þ urh þ ine bireousunge, ac altog æ dere ic am forloren þ urh þ ine lu þ ere through your mercy but altogether I am destroyed through your wicked deden. deeds ‘ Through true mercy the rest of the soul begins. But I cannot ever rest through your mercy, but altogether I am destroyed by your wicked deeds. ’ (HomU 5.6 (Buch F) 13) As seen in section 2.2.2.2, possibility meanings are described as barriers in force- dynamic terms. Hence the absence of possibility expressed by þ urfan in (3.34 ) can be considered a ba rrier. The meaning is clear from the context: “ I cannot rest, because I do not receive mercy, but wicked deeds. ” It seems obvious that a necessity meaning does not fit in this context at all (i.e. a sequence such as “ I need not rest, because I do not receive mercy, but wicked deeds ” does not seem to make any sense). Despite the marginality of this example, it constitutes another piece of evi dence of the polysemic values of OE þ urfan . This polysemy arises as a consequence of the logical relations between necessity and possibility, as claimed by Lyons (1977) and Palmer (1986) (cf. section 2.2.2.2), which establish that if X is not necessary, not-X is possible. These relations seem to Chapter 3. Old English þ ur fan , be þ urfan, neodian and behofian 133 operate in other languages than English, because, as noted by van der Auwera and Plungian (1998), German d ü rfen , the cognate of þ urfan , develope d its current possibility meaning, ‘ be allowed to, ’ from its original necessity meaning ‘ to need. ’ 28 According to these scholars, the development is highl y conditioned by the non-affirmative contexts in which d ü rfen used t o occur. From an original meaning ‘ need not, ’ expressing absence of necessity, it developed the meaning of prohibition ‘ must not, ’ and, then, due to the logical relations between necessity and possibility (which establish that if you must not do X, you can do not-X), it developed its current possibility meaning ‘ may, may not. ’ In the analysis of the ME corpus, we will observe how this meaning gains ground with ME thurven and is also possible with ME nede n v.2 (cf. section 4. 4.2.1). Going bac k to the meanings of OE þ ur fan as offered by Bosworth and Toller ( s.v. þ urfan v.), we must say that the three of them are root necessity meanings, which are defined in the cognitive terms of forces, as mentioned in section 2.2.2.2. In that section, devoted to the description of root modality, I also point out that one of the adva ntages of defining necessity in terms of forces is the possibility of resorting to three scales of gradience. The first scale I mentioned concerns the degree of subjectivity of the forces implied. It is said that root modality is subjective when it is ba sed on subjective referents (cf. you must get out of the bath now ). On the other hand, objective root modality is that stated in general truths (as clay pots must have some protection from severe weather ). The second scale refers to the strength of the force, and it goes from very weak (as in she must buy a new pair of shoes ) to very strong (as in she must pay taxes every year ). Finally, the thi rd scale concerns the origin of the force, which may basically be internal (exemplified earlier on with I need to call her now ), or external (as in I must turn in this paper tomorrow ), as mentioned above; however, in the analysis of the corpus data I have found examples of a third type of force, namely that of general or igin (in general statements such as the adjective need not always precede the noun ). The combination of the latter two scales, namely strength and origin of the force will yield different types of forces, that is, social (which may have different nuances such as religious or hierarchical), legal, inner, general, and so on. When illustrating the scarce examples of general types of forces it will be observed that among them we will find the few examples which can be considered cases of 28 Interestingly enough, its derived verb bed ü rfen has rem ained as a verb meaning ‘ need ’ in German (cf. Molencki 2002). Chapter 4. Middle English thurven (& durren ), bethurven, neden, bihoven & misteren 236 The noun phrase trees, herves and gras functions as the experiencer of nede þ , and the prepositional phrase of hete of þ e sonne is the theme of the verb. In fact, in Middle English a to -prepositional phrase may also be found instead of an of - phrase, as illustrated in (4.25): (4.25) Fremmde menn.. nedenn to þ in hellpe . ‘ Foreign men need your help. ’ (?c1200 Orm.(Jun 1) 6161) (from MED, s.v. neden v.2 2(b)) Considering an of -prepositional phrase equivalent to a genitive noun phrase seems to be straightforward, but considering a to -prepositional phrase equivalent to a genitive may be more problematic. For that reason, it would not be accurate to classify sentences such as (4.25) as Allen ’ s Type II constructions concerning experiencer verbs. There are, finally, other types of constructions in volving ME neden v.2, that is, those constructions in which it has a nominative experiencer and a sentential theme, which is in most of the cases an infinitival clause. Actually, the MED ( s.v. neden v.2 2(c)) only provides instances concerning infinitives, but Visser (1963-1973: 1426, § 1347) notes that Shakespeare uses need followed by a that -clause: But I, who never knew to entreat, Nor never needed that I should entreat, Am starv ’ d for meat (1596 Shakespeare, Taming Shrew IV, iii, 7). Perhaps it is only a later eModE development, or a sign of Shakespeare ’ s idiolect, but in any case it must be taken into consideration when analysing the examples from the ME corpus. The first recorded instance of personal neden v.2 followed by an infinitive is dated from 1380, according to Visser (1963-1973: 1426, § 1347): (4.26) More than he nedi þ for to have . ‘ More than he needs to have. ’ (c1380 Wyclif, Select. Wks. III, 348) The infinitive is marked by for to , which in Middle English alternated with to . The marked infinitive is the only type recorded in the MED ( s.v. neden v. 2 2(c)). We know that the plain infinitive after need must have reached its climax in the eModE period, since, according to Krug (2000: 202), in Shakespeare “ the ratio of plain to marked infinit ives is approximately eight to one ” . Even if the MED does not register any occurrence of nee d + plain infinitive, Visser (1963-1973) records the first occurrence quite early in the history of English: Chapter 4. Middle English thurven (& durren ), bethurven, neden, bihoven & misteren 237 (4.27) þ ou mai Ñ t not longe endure, And nedes dye ; henne þ ou mote. ‘ You may not suffer for a long time and need die when you must. ’ (c1390 In a Pistel (in Brown, Relig. Lyr. XIVth C.) 75) According to Visser, dye is an infinitive which complements the verb ne des , supposedly inflected for the second pers on singular, which could be possible in a northern dialect (cf. Lass 1992: 137). Although he does not record any other instance of plain infinitive after personal need until the last quarter of the 17 th century, the occurrence of examples of nede n with a plain infinitive is potentially possible in the corpus. In any case, when neden is construed with a nominative experiencer and a sentential theme of any type, it may be s aid to be one of Allen ’ s (1995) ‘ Personal ’ constructions with experiencer verbs with sentential themes. Therefore, ME neden , v.2 seems to be a rich experiencer verb, since it may be construed in Allen ’ s (1995) six possible types, including nominal and sentential themes, that is, Types S, hit , N and I, when it occurs with a non-nominative experiencer, and Types II and ‘ Personal ’ when it has a nominative experiencer. In addition, the MED records another possibility: a reflexive construction. Consider (4.28): (4.28) Thow ned the to fyght..With youre flesche, and with the fende . ‘ You need to fight (yourself) with your flesh and against the fiends. ’ (a1500 God of hewine (Tit A.26) 197; from MED, s.v. neden v.2 2 (c)) In this sentence there is a nominative pronoun, namely thow , and its oblique counterpart, namely the . For this reason, the editors of the MED consider that this is an instance of a reflexive construction with the verb neden v.2. I wonder whether an interpretation of ned as a form of nede n v.1 is possible. In this sense, (4.28) might well be interpreted as ‘ you compel yourself to fight. ’ If such an interpretation holds true, this sentence would be another instance of ambiguity between neden v.1 a nd nede n v.2. There is yet another instance of se mantic overlapping between both verbs. One of the meanings provided by the editors of the MED for need v.2 is ‘ to be required, to be oblige d ’ ( MED s.v . neden v.2 2 (c)). In other words, one of the meanings of need v.2 is the correspondi ng passive voice of nede n v.1 ‘ compel, force. ’ This has already been suggested i n the section devoted to the analysis of the data retrieved from the OE corpus (3.4.2.2). There it was mentioned that OE neodian occurs very frequently in the passive voice, which results in a change in Chapter 4. Middle English thurven (& durren ), bethurven, neden, bihoven & misteren 238 the syntactic structure of the sentence in which it occurs, because the subject is no longer the antagonist, but the agonist, just like in the examples of OE þ urfan . The fact that ME neden v.2 develops in that way constitutes a significant piece of evidence of the semantic connection between ME neden v.1 a nd neden v.2. The last feature of ME neden v. 2 which I want to highlight conc erns the polarity of the sentences in which it occurs. A look at the entry neden v.2 in the MED reveals that in senses 1b (meaning ‘ it is necessary …’ ) and 2 (m eaning ‘ to need, be obliged …’ ) negative instances outnumber positive examples (cf. OE þ urfan ). Polarity in this verb i s so relevant that Visser (1963-1973) tr eats negative and interrogative instances as separate constructions, independent of affirmative ones. As far as negative constructions are concerned, he mentions that the first occurrence of negative neden with a plain infinitive dates from 1470, that is, from the very end of the ME period (1963-1973: 1428, § 1348): (4.29) The woman … Cawkit ilk Ñ ett, that thai neid nocht gang by. ‘ The woman … Cawkit (?) each gate that they need not go by. ’ (c1470 Henry the Minstrel, Wallace VII, 414) The infinitive gang , ‘ go, ’ follows the negated form of neden , namely neid noc ht , in this first example with a plain infinitive, which is also recorded in the OED ( s.v. need v.2, 8b). In fact, this is the first example of any structure of ne den v.2 which the editors of the OED consider to have a bare infinitive, since they do not record the 1390 example quoted above as (4.27). As far as interrogatives are concerned, they also deserve special comment, since, according to Visser, they are construed impersonally until the second half of the 15 th century, just the same as affirmatives (1963-1973: 1429, § 1351). Consider, for instance, (4.30): (4.30) What nedith it thane a new lawe to bigynne? ‘ What need it th ere then to begin a new law? ’ (1377 Langland, P. Pl. B XVII, 30) This sentence consists of an opening what , the main interrogative type as for need until the eModE period (cf. Visser 1963-1973: § 1351), a formal it , and a to- infinitive . Another possible interrogative sentence would have an oblique personal pronoun instead of it , specifying the experiencer of the necessity. In its way towards a personal type, i nterrogative need also takes the bare infinitive, whose first recorded example is (4.31): Chapter 4. Middle English thurven (& durren ), bethurven, neden, bihoven & misteren 239 (4.31) What nede Ñ e be abast ? ‘ What need is there for you to be abashed? / ‘ Why should you be abashed? ’ (c1460 Towneley Pl. p. 143) (from Visser 1963-1973: 1430, § 1351) The passive infinitive be abast follows the verb nede without any particle. As mentioned above, the use of plain infinitives in Middle English is very rare, and almost limited to the descendants of the OE pre-modal verbs. To summarize the characterization of ME neden as found in the literature, we should begi n by saying that nede n v.1 and neden v.2 pr ove to overlap to so large an extent that the best way to capture their evolution is to analyse the m jointly. Secondly, both verbs develop in the expected way, that is, neden v.1 ‘ compel ’ becomes less frequent than in the OE period, and nede n v.2 ‘ need, be necessary ’ not only increases its frequenc y, but also develops new uses. Thirdly, neden v.2 may be both a transitive and an intransitive verb, and as a transitive verb, it may take both nominal and sentential themes / complements. We must not forget, in this connection, that the presence of a sentential infinitival complement is one of the first features which auxiliaries exhibit (cf. Bolinger 1980), and the presence of a plain infinitive seems to be especially related to modal verbs already in the ME period. Fourt hly, the frequent use of neden v.2 in impersonal constructions, added to t he fact that ME thurven also occurs in this type of structure, come to corroborate the idea that impersonality and the expression of necessity are closely associated (cf. Fischer 1992, Warner 1993, Pocheptsov 1997, and Krug, 2001, 2002). Finally, neden v.2 exhibits a close connection with non-affirmativeness, like OE þ urfan and PDE need . 4.3.3. Middle English bihoven Like ME thurven and neden , bihoven also undergoes important c hanges in the ME period. We have seen that in Old English, behofian is primarily a verb construed with a nominative experiencer and either a genitive theme or a sentential theme (cf. sections 3. 3.2 and 3. 4.3). In Allen ’ s (1995) terminology, we could ascribe OE behofian to Type II and Type ‘ Personal ’ constructions. In Old English the main meaning of this verb is ‘ to need, ’ despite the fact that scholars such as Bosworth and Toller, for example, claim that behofian may also mean ‘ be appropriate, be fitting. ’ We expect a series of changes to take place in the history of this verb, on t he grounds that its PDE counterpart, behove , exhibits Chapter 4. Middle English thurven (& durren ), bethurven, neden, bihoven & misteren 240 syntactic and semantic features which differ considerably from those in the OE period. Indeed, some of the changes which affect behove be gin in the ME period. Allen (1997: 8) considers that bihoven started to appear with non-nominative experiencers in the late 11 th century, and in addition, she mentions that such a usage is the norm in texts written in the 12 th century, as in (4.32): (4.32) alswa micel swa heom behofe ð . as much as them (dative) behoves ‘ as much as is suitable for them . ’ (Ch 1110 Harm 62) (example, gloss and translation from Allen 1997: 7) However, bihoven may also occur with nominative experiencers in non- original manuscripts, that is, in ME copies of OE composition, as in (4.33): (4.33) for þ an mancynn behofe ð godcundre lare. for that mankind (nom inative) behoves godly (gen.) learning (gen.) ‘ because mankind needs good learnin g. ’ (LS 28 (Neot) 1) (example, gloss and translation from Allen 1997: 7) The ME copyist of this text opted to be faithful to the OE manuscript and retained the noun mancynn in the nominative, although by this time the use of non-nominative experiencers was extended (c f. Allen 1997: 8). Therefore, the ME period seems to be the turning point in the syntax of behove , since it moves from the Type II construction consisting of a nominative experiencer and a genitive theme which can be found in Old English (as in (4.33)), to a Type I construction, consisting of a non-nominative experiencer and a nominative theme (as in (4.32)). This is not, however, the only type of construction which emerges in the ME pe riod. According to Allen (1997: 10), the combination of bihov en with sentential elements increases considerably in the 13 th century. Consider, for instance, (4.34): (4.34) Bihofde nawt þæ t swuch were leafdi of castel . behoved not that such were lady of castle ‘ It would not be fitting that a lady of a castle were like that. ’ (AW 58.7) (example, gloss and translation from Allen 1997: 10) In sentence (4.34), the that -clause þæ t swuch were leafdi of castel is the only argument of the verb bihofde . The absence of an explicit experiencer is also a ME innovation. If a sentence s uch as (4.34) had an oblique experiencer, it would Chapter 4. Middle English thurven (& durren ), bethurven, neden, bihoven & misteren 241 be considered an instance of Allen ’ s Type S constructions with experiencer verb constructions. The basic difference between Type S construction and the ‘ Personal ’ Type in which OE behofi an can be found concerns the nature of the experiencer. While in Old English the e xperiencer is inflected for the nominative, in Middle English the experiencer is either non-nominative or absent from the sequence. The last type of syntactic construction in which ME bi hoven may occur is Type hit , that is, the construction consisting of the formal subject hit , an oblique experiencer, and a sentential theme, as can be seen in (4.35), from the MED ( s.v. bihoven v. 2b (a)): (4.35) It byhoveth the to ben obeisaunt to the maneris of thi lady it behoves you (obl.) to be obedient to the manners of your lady [i.e. fortune] . [i.e. fortune] ‘ It behoves you to be obedient to (the manner of – OED s.v. manner n.1 2a) your lady [i.e. fortune]. ’ (?a1425(c1380) Chaucer Bo.(Benson-Robinson)) The use of bi hoven in this type of construction seems to emerge in Middle English, and it will remain in the language up to Present-Day English. While in Middle English the sentential theme may be a that -c lause or a to -infinitival clause, the sentential constituent is in Present-Day English mostly a to -infinitive, rather than a that -clause (cf. OED s.v. behove v. 4). Therefore, in Middle English we must expect bihoven to exhibit a wide variety of syntactic constructions, which ranges fr om the OE relic of Type II (nominative experiencer and geni tive theme) and the ‘ Personal ’ Type (involving a nominative experiencer and a sentential theme), to the ME innovations, namely Type I (non-nominative experiencer and nominative theme), Type S (non- nominative experienc er and sentential theme) and Type hit (formal hit , non- nominative experiencer and sentential theme). This scenario, therefore, includes all the types mentioned by Allen (1995) but one, namely Type N constructions, which consist of an oblique e xperiencer an d a genitive theme. According to Allen (1997: 19, note 11), “ unambiguous instances of this verb with two non- nominative arguments are not t o be found. ” It remains to check out whether t he ME corpus contains any sentence with these features. In addition to occurring in all these types of experiencer verb construct ion, ME bi hoven is also likely to be Chapter 4. Middle English thurven (& durren ), bethurven, neden, bihoven & misteren 242 found without a ny experiencer, a construct ion which is considered archaic in Present-Day English (cf. OED s.v. behove v. 4 b) Having explained the syntactic changes undergone by ME bihov en , we must turn now to another linguistic level of analysis, namely semantics. As already mentioned, the PDE connotations of behove are mainly concerned with the notion of appropri ateness, while OE behofian expresses basically volitional necessity. The shift towards the PDE notion takes place in the ME period, as illustrated in example (4.34), where bihofde means ‘ be fitting. ’ Although we may suspect that this semantic change originates in the syntactic changes undergone by this verb in Middle English, Allen (1997: 10- 11) considers that such a nuance of appropriateness is mainly bas ed on the natural development of the semantic notion of necessity. It is not difficult to conceive that from an internally rooted necessity such as that expressed by OE behofian , a new meaning may emerge conveying externally generated necessity, such as ‘ be obligatory or hi ghly advisable, ’ finally, the change from ‘ be obligatory ’ to ‘ be fitting ’ is subtle and straightforward. According to the MED ( s.v. bihoven v.), this verb may express types of necessity rooted in different external entities, such as destiny, circumstances, and morals or doctrine. It is in this light that we must interpret the meaning suggested for byhoveth in (4.35), namely ‘ be proper; ’ in this sentence bihoven expresses appropriateness based on destiny. In addition, ME bihoven may still express the older meanings, which include volitional necessity and obligation. An example of the former is, for instance, (4.36), from the MED ( s.v. bihoven v. 1a (b)): (4.36) Þ e oost dryed vp þ e ryueres, for hem byhoued so moche water. the host dried up the rivers for them behoved so m uch water ‘ The host dried up the rivers, because they needed so m uch water. ’ ((a1387) Trev. Higd.(StJ-C H.1) 3.237) Despite its syntactic features, which might lead to an appropriateness reading of the meaning of bihoven , this sentence expres ses internally rooted necessity (close to strong volition), and hence the translation provided is ‘ need. ’ This is the most common meaning of OE behofi an , but it is still common in the ME period. Another old meaning which can be found in ME bihoven is that of externally imposed necessity (obligation), as illustrated in (4.37), from the MED ( s.v. bihoven v. 2b (a)): Chapter 4. Middle English thurven (& durren ), bethurven, neden, bihoven & misteren 243 (4.37) Sayle hem bihoued holliche al a ni Ñ t. sail them were-obliged wholly all night ‘ They were obliged to sail wholly all night. ’ (a1375 WPal.(KC 13) 2721) In this sentence bihoven expresses an externally ge nerated necessity (based on the conditions of the sea), and the crew is obliged to sail the whole night in order to save their lives. Therefore, the verb expresses an obligation, rather than appropriateness, and for that reason bihoued is better translated as ‘ be obliged. ’ We can gather from the preceding paragraphs that ME bihove n expresses a wide range of possible meanings, from volitional necessity and obligation (cf. examples (4.36) and (4.37)), to general appropriateness (as in (4.32) and (4.34)), and specific types of appropriateness based on destiny (as in (4.35)). The syntactic and semantic complexity of ME bihoven is not surprising, because it takes place in the period of English in which most linguistic changes occur. Since bihoven is in a transitional stage from its OE status to its modern idiosyncrasy, we expect the ol d and the new structures and meanings to co-occur until the prevalent one overrides the others. In the corpus section devoted to the ME period, it will be seen which of the syntactic and semantic features of bihoven are more common and are, therefore, prone to survive in the eModE period. 4.3.4. Middle English misteren In this final section, I provide an outline of the features of a new verb meaning ‘ need, ’ which enters the language in the ME period via French: misteren . The first example of misteren recorded in the MED ( s.v. misteren v. 1(d)) dates back to 1412: (4.38) Richard sall gette all the stuff of the stane that misters more of the makyng of the Kirk of Katrik than that stuffe that is founde within the kirke yerde. ‘ Richard shall get all the stuff of the stone that is m ore necessary for the making of the Church of Katrik than that stuff that is found within the church yard. ’ (1412 in Salzman Building in Engl. 487) However, the OED ( s.v. mister v.1 2) provides an earlier example, from 1375: (4.39) And alkynd othir apparaill That mycht availl, or Ñ eit mysteir Till hald castell . Chapter 4. Middle English thurven (& durren ), bethurven, neden, bihoven & misteren 244 ‘ And all kind of other apparels that might be of value or be necessary for you to the old (?) castle. ’ (1375 Barbour Bruce xvii. 215) Sentences (4.38) and (4.39) ha ve a common characteristic, which is the meaning of misteren in both cases, namely ‘ be necessary. ’ However, the constructions in which the verb occurs are different. In (4.38) we have an intransitive construction, while in (4.39) the verb is construed transitively. As will become clear from the following paragraphs, these are indeed the two main possible constructions in which ME misteren may occur. When misteren is used intransitively, it has a single argument. An instance of such constructions is (4.38) above, which has as only argument the relative pronoun that . Despite the fact that such a pronoun cannot be said to be nominative or oblique, the MED includes this example in the personal sense of misteren . In this respect, misteren resembles the intransitive use of ME neden exemplified above with (4.14), namely In a goode spouse and wif nedi þ þ ese condiciouns ( ‘ These conditions are necessary in a good spouse and wife ’ ). In both cases the verb means ‘ be necessary, ’ and takes only a s ubject argument (presumably) inflected for the nominative. Another intransitive use of misteren may be that in which the verb has a formal subject it , or an oblique experiencer in subject position, as illustrated in (4.40) and (4.41) respectively: (4.40) Yf it mystier , we shal guyde & lede you wel. ‘ If it is necessary, we will guide and lead you well. ’ (c1500 Melusine (Roy 18.B.2) 222/36) (4.41) Þ e enchauntere rade on his Mule.. þ at bar him, whanne him mysteryd , by þ e way. ‘ The enchanter rode on his mule, which bore him by the way, when it was necessary for him. ’ (a1500(?a1425) Lambeth SSecr.(Lamb 501) 104/28) As mentioned, the only differenc e between (4.40) and (4.41) is the nature of the constituent in subject position. While in the first it is a formal it , in the second sentence there is an oblique pronoun, him . Both it and him occupy the subje ct position and, as mentioned in section 2.3, there are scholars who consider that oblique ex periencers may function as syntactic subjects. Anot her aspect which (4.40) and (4.41) have in common is their occurrence in parenthetical constructions introduced by conditional items, namely yf and whanne . The use of Chapter 4. Middle English thurven (& durren ), bethurven, neden, bihoven & misteren 245 expressions of necessity in parenthetical constructions seems to have remained important even in Present-Day English (cf. PDE if need be in OED s.v. need n. I 3a). As far as the transitive uses of misteren are concerned, they may be classified according to the nature of the experiencer, which may be nominative or non-nominative. Whe n the experiencer is nominative, the ver b may be followed by an unmarked theme, as (4.42), for instance, from the MED ( s.v. misteren v. 1 (a)): (4.42) He sayde he was a synner & mysterd forgyfnes of his syn. ‘ He said he was a sinner and needed forgiveness of his sin. ’ (c1450 Alph.Tales (Add 25719) 75/24) The experiencer of mysterd in this sentence is clearly nominative, namely he . The theme, on the contrary, is unmarked ( forgyfnes ), which implies either that it is accusative or that it has no inflections as a result of the erosion of inflectional endings in Middle English. This type of construction m ay be said to be a variant of Allen ’ s (1995) Type II, because in that type the theme takes the genitive case. Another variant of Allen ’ s Type II is the following, in which ME misteren is followed by an of -prepositional phrase, as in (4.43): (4.43) Þ ou mysters not of my lectuarie , for þ ou erte a leche . ‘ You do not need my electuary, because you are physician. ’ (c1450 Alph.Tales (Add 25719) 136/4) Sentence (4.43) also has a nominative experiencer, þ ou , but the theme is not an ambiguously marked noun phrase, but an of - prepositional phrase, of my lectuari e . If we interpret, as has been done above, that an of -prepositional phrase is equivalent to a geniti val noun phrase, sentence (4.43) would be an instance of Allen ’ s (1995) Type II constructions with experienc er verbs. This is another respect in which misteren and neden function alike (cf. examples 21 and 24 above). The preposition of seems to be som ewhat frequent in the introduction of complements of verbs meaning ‘ need. ’ The last transitive use of misteren with a nominative experiencer involves infinitival themes, although Visser (1963-1973: 1424, § 1344) mentions that it is rarely found in such a collocation. In fact, the MED a nd Visser mention only one and the same ME example, given below as (4.44): (4.44) I sall so ordand at þ ou sall nott myster to be a thief no mor . Chapter 4. Middle English thurven (& durren ), bethurven, neden, bihoven & misteren 252 As Table 4.5 makes clear, my ME corpus contains 1,248,765 wor ds. Subperiod M2 remains the less represented age of Middle English, and its proportion wi th respect to the total is the same as that in the Helsinki Corpus , namely 16% of the total of Middle English. Despite this weakness, the corpus I have selected appears to be a representative one for several reasons. Firstly, it doubles up the size of the Helsinki Corpus , which is in itself a representative one. Secondly, the texts cover all the ME decades, when possible. Thirdly, the texts are instances of different text-types, different dialects, and the y represent both original compositions and translations from Latin or French. Therefore, it looks as if the corpus described in this section will be an appropriate one for the analysis of my verbs. After having scrutinized more than 15,000 potential examples (see appendix II below for details), I have found out that the number of occurrences of each verb in the corpus is the following: 16 VERB OE ME OE N.F. ME N.F. OE % ME % THURVEN 159 55 13.19 4.40 46.77% 12.82% BETHURVEN 47 4 3.89 0.32 13.82% 0.93% NEDEN 105 161 8.62 12.89 30.59% 37.53% BIHOVEN 30 206 2.48 16.49 8.82% 48.02% MISTEREN 0 3 0.0 0.24 0% 0.70% Total 341 429 28.20 34.35 100% 100% Table 4.6: Frequency of each verb in the ME corpus as compared to Old English. The differences in the frequency of occ urrence with respect to the OE data are striking. If we recall the figures for Old English, we observe that the frequencies of thurven and bihoven reverse those of þ urfan and behofian respectively, and neden increases its frequency to the detriment of bethurven . There are no OE data about misteren , because, as is well-known, this is a ME loanword. Following the same steps as for the analysis of the OE corpus-data, I have included all these ME examples into a Microsoft Access computer database, and have analysed them according to the series of variables listed in section 3.4.0. The findings will be illustrated with examples retrieved from my corpus. As far as the examples taken from the Corpus of Middle English Prose and Verse are 16 In Middle English, as was the case in Old English, there exist other linguistic means of expressing the same kind of necessity as these verbs, namely expressions consisting of the noun nede or myster in combination with the verbs be and have (e.g. hym is nede, he hath nede, he hath myster ). Taeymans (2004b) shows that the frequency of have need in M iddle English is similar to that of the verb need . Although I am aware of the important role played by these expressions, they have been left out of m y an alysis, which is only concerned with verbs. Chapter 4. Middle English thurven (& durren ), bethurven, neden, bihoven & misteren 253 concerned, I will provide the year of composition, the title of work, the author, and, in the case of verse works, the number of the lines where the example occurs. In what follows, I deal separately with each of my ME verbs with the same structure adopted for the analysis of Old English. Section 4.4.1 deals with ME thurven (and durren ) and bethurven . Se ction 4.4.2 is devoted to the analysis of ME neden . Section 4.4.3 concentrates on ME bihoven , and, finally, section 4.4.4 analyses the scarce number of examples of ME misteren . 4.4.1 Mi ddle English thurven (and durren ) and bethurven in the corpus ME thurven and bethurven are the descendants of OE þ urfan and be þ urfan . Despite the predominant use of þ urfan in Old English, its ME counterpart becomes less frequent in favour of other verbs (cf. sections on ME nede n and bihoven below). Nonetheless, ME thurven occurs in all the four subperiods of Middle English. Table 4.7 shows the actual number of occurrence s of thurven in each of the four subperiods (second column), 17 as well as the normalized frequencies per 100, 000 words (third column). Subperiod Number of Occurrences Normalized Frequencies M1 31 10.74 M2 7 3.38 M3 7 1.91 M4 10 2.58 TOTAL 55 4.40 Table 4.7: Distribution of ME thurven by subperiods. The normalized frequencies reveal that more than half of the examples of thurven (10.74) occur in M1, which impl ies that it became less and less frequent in the following subperiods. In fact, the frequenc y of occurrence of thur ven seems to decrease from one subperiod to another, with the exception of M4, when it undergoes a slight increase. However, the results of this table are tentative and this slight increase might be due to textual factors, sinc e most of the texts of M4 belong to fiction. 17 Six out of the 31 examples of M1 are actually coded as MX/1 in the Helsinki Corpus . The seven instances of M2 also include examples of M2/3 and M2/4, since the date of com position is M2, although the manuscripts used by the compilers of the Helsinki Corpus belong to M3 and M4 respectively. Chapter 4. Middle English thurven (& durren ), bethurven, neden, bihoven & misteren 254 ME bethurven , in turn, undergoes a drastic decay from Old English, since my ME corpus only records four instances (as compared to the 47 OE examples), and all of them belong to non-contemporary manuscripts from M1, which may imply that they are copies from OE originals. If it were so, we could conclude that in M1 copyists take for granted that the recipients of their works would be able to understand this verb, even if, as it seems, bethurven is no longer productive in the ME period. After these preliminary remarks on the frequency of use of thurv en and bethurven in Middle English, I proceed now to offer the analysis of the examples retrieved from my corpus, paying special attention to semantics (section 4. 4.1.1) and syntax (section 4.4.1.2). 4.4.1.1 Semantic features of Middle English thurven (and durren ) and bethurven As mentioned in section 4.3.1 above, ME thurven may express a wide range of necessity meanings, and also, at times, possibility. This is indeed verified in my corpus, where four out of 55 total examples express possibility rather than necessity, as can be seen in (4.47) and (4. 48): (4.47) 4172 Nu is Iulius awei ifloen; (...) now is Julius away fled 4174 ne þ urfe we nu nauer-mare; iseon hine cumen here . not can we now never-more see him come here ‘ Now Julius has fled (...) we cannot / will not have the occasion to see him come here any m ore. ’ (1205 Layamon ’ s Brut, lines 4172-4174) (4.48) he wax so mylde and so meke, he was so mild and meek A mylder man þ urt no man seke . a milder m an could no man see ‘ he grew so mild and so gentle, no one could seek a m ilder man. ’ (6,024 helsinki\cmhansyn) These two examples illustrate clearly the possibility meanings conveyed by ME thurven , which does not express a cognitive force, but a cognitive barrier, conveying the meaning ‘ cannot. ’ In section 4.3.1 I formulated the hypot hesis that thurven expresses possibility only when its spelling combines features of this verb and of ME durren . This seems to be the case in (4.48), where þ urt exhibits the absence of any [f] or [v] sound. However, this is not the case of sentence (4.47), which is also used in the MED to illustrate the possibility meaning of Chapter 4. Middle English thurven (& durren ), bethurven, neden, bihoven & misteren 255 thurven ( s.v. thurven 7a (a)), and, therefore, it may be taken as prototypical. The morphology of the verb form, þ urfe , reveals that such a form belongs to the paradigm of ME thurven , rather than to that of durren . This should not be surprising either, because, as seen in section 3.4.1.1, the meaning of OE þ urfan may also be that of absence of possibility, that is, it may express the presence of a cognitive barrier. The construction illustrated in (4.48), namely a negative verb of necessity expressing possibility when following a comparative adjective, is also possible and rather frequent with ME neden , as will be seen below. Leaving aside the four instances of thurven where it expresses possibility, 51 examples expressing necessity remain to be analyse d. Table 4.8 sketches the semantics of thurven in terms of cognitive forces, taking into account their origin and strength: ORIGIN STRENGTH N. OF EXAMPLES TOTAL STRONG 28 EXTERNAL WEAK 28 STRONG 8 INTERNAL WEAK 6 14 GENERAL NEUTRAL 9 9 STRONG 36 WEAK 6 TOTAL NEUTRAL 9 51 Table 4.8:Origin and intensity of the forces conveyed by ME thurven. As was the case in Old English, ME thurven exhibits a tendency to express strong forces (36 instances) and external forces (28 instances), though it also expresses other types of forces with more frequency than in the pr evious period. Thus, in Middle English thurven may express internal forces with relative frequency (14 instance s), as well as general types (nine instances). Such gener al forces are originated in a nebulous, generalized authorit y, as has been described above, and the degree of strength of such forces is neither strong nor weak. I n the paragraphs which fo llow, however, I expand each of the types of for ces mentioned here in order to analys e the different nuances expressed by t hurven in each case. Let us begin with strong external forces , the most frequent t ype of meaning expressed by thurven , as was the case of OE þ urfan . Ta ble 4.9 shows the different notional types of strong external forces expressed by thurven , namely forces exerted by a religious or a hierarchical authority on the agonist. It also shows the polarity of the sentences in which thurven occurs, because polarity conditions the semantics of the verb. When the verb occurs in an Chapter 4. Middle English thurven (& durren ), bethurven, neden, bihoven & misteren 256 affirmative context it expresses the presence of a given force; when, on the contrary, the verb occurs in a non-affirmative context (cf. section 2.2.1.1 above for examples of non-affirmative contexts), it may express either the absence of a force (absence of obligation), or the presence of a force not to act in a given way (prohibition). The latter meaning, however, is not recorded for ME thurven in my corpus, while OE þ urfan expresses it on 14 oc casions: NON-AFFIRMATIVE CLAUSE POLA RITY TYPE OF FORCE AFFIRMATIVE LACK OF F ORCE TOTAL RELIGIOUS 2 19 21 HIERARCHICAL 7 7 TOTAL 2 26 28 Table 4.9: Types of strong external forces conveyed by ME thurven , with indication of clause polarity. If we read this table horizontally, we observe that, as was the case with OE þ urfan , ME thurven expresses strong external forces predominantly exerted on the basis of a religious authority. If, on the contrary, this table is read vertically, the numbers indicate that thurven shows a pronounced tendency to express lack of obligation, as was also common f or OE þ urfan . Nevertheless, this ME verb may also express strong external forces in affirmative contexts, such as (4.49): (4.49) Þ u schalt (...) wakien i moni care. Nawt ane for þ e-seolf; ase you shall weaken in many mental-sufferings not only for your-self as þ erf godes spouse. must God ’ s spouse ‘ You shall become weak due to m any mental sufferings, not only for yourself, as must (is proper of) a spouse of God (i.e. nun). ’ (5,237 helsinki\cmhali) The meaning expressed by thurven in this sentence is that referred to above as ‘ what is fitting ’ (cf. section 4.3.1). From this context we gather that a nun knows that her religious career implies mental sufferings, that both things are correlated. Therefore, it is an external authority, religion, that inflicts an imposition on nuns, but such an imposition is not an obli gation, it is inherent to God ’ s spouses. It does not mean that nuns must suffer, but that it is proper for nuns to suffer. In this sense ME thurven is similar to some uses of ME bihove n , as will be described below. Moving on to the negative examples of thurven (Table 4.9), where it expresses absence of obligation, the release of the imposition may come from Chapter 4. Middle English thurven (& durren ), bethurven, neden, bihoven & misteren 257 two different types of authorities, namely religious or hi erarchical. A prototypical example of absence of religious force is (4.50): (4.50) Ase techeth holy bok, as teaches holy book Þ arf me no þ ing drede, need me nothing fear Sathan shal nout spede Satan shall not succeed Wy þ wrenches ne wy þ crok . with tricks nor with deceit ‘ As the holy book teach es, I need not fear any thing, Satan will not succeed with tricks or with deceit. ’ (d1333 The Works of William Herebert) As in other examples mentioned in the OE section of this study, thurven is followed by an infinitival clause headed by a verb meaning ‘ fear, ’ which remains the most frequent verbal meaning associated with this preterite-present verb in Middle English. In this example, the doctrine taught in the Bi ble makes the speaker feel with enough strength to feel released from the fear towards Satan. Although in a lesser rate, the absence of obligation expressed by thurven may also be based on a hierarchical superiority, as in sentence (4.51): (4.51) 7226 heo wulle ð bi-witen þ i lond; (...) they will protect your land 7227 Þ enne mihte þ u mid winne; þ i lif al uor-werien. (...) then might you with possessions your life all wear-out 7229 ne þ r æ t þ u nauere habben kare; of uncu ð e leoden. not need you never have care of unknown people ‘ They (Octa and Ebissa) want to /will protect your land (...) then you might wear out your life with your possessions (...) you need not take care / worry about unknown people. ’ (1205 Layamon ’ s Brut , lines 7226-7229) The speaker tells the listener that the presence of two hierarchical superior people, such as Octa and Ebissa, implies the absence of any need to worry about instability in the region. Therefore, sentence (4.51) expresses absence of a force based on the hierarchical principle that superior authorities provide protection f o r the people. Another example of thurven expressing absence of force from a hierarchical perspective is (4.52), where the verb occurs in a interrogative sentence introduced by what : (4.52) 799 an þ e o þ er ne can sweng but anne (...) and the other not can strike but one Chapter 4. Middle English thurven (& durren ), bethurven, neden, bihoven & misteren 258 803 [w]at þ arf he recche of a mo swenge what need he trouble of a more strike ‘ and (MED says "if") the other cannot strike but one (...) why should he trouble / worry to strike one more? ’ (1910-1210 The Owl and the Nightingale , lines 799-803) This is the first example found in my corpus for this kind of construct ion involving the interrogative pronoun what . The translation I suggest is ‘ why should? ’ following the MED ( s.v. thurven 3a (a)), instead of the expected ‘ what does he need? ’ As will be seen in the next chapter, this kind of construction with interrogative pronoun what will be especially common in early Modern English with the verb need . If we go back to Table 4.8, we observe that thurven doe s not express weak external forces in Middle English. In fact, the expression of this kind of f orces was not very frequent in Old English either, sinc e it only occurred on four occasions out of 159 examples of þ urfan . The absence of this t ype of exampl es leads us to the next line of Tabl e 4.8, that is, that containing the examples in which thur ven expresses forces originated in the agonist ’ s self. In eight cases thurven expresses the absence of an internally rooted strong force , as in (4.53): (4.53) 104 I swere you by this light. I swear you by this light 105 ffor whosoeuer may get thise close, for whosoever may get this close 106 he ther neuer rek where he gose, he need never care where he goes 107 ffor he semys nothyng to lose. for he seems nothing to lose ‘ I swear to you by this light, for whosoever may get this close, he need never care where he goes, for he seem s [to have] nothing to lose. ’ (1460 The Towneley Plays ) The verb thurven in sentence (4.53) expresses the internally rooted strong necessity related to the instinct of self-protection. Such a force is cancelled when the agonist feels that he has nothing to lose , because in this case, he need not protect what he does not have. Therefore, (4.53) seems to be a clear example of absence of strong internal force. ME thurven also expresses weak internal force on six occasions, that is, a type of force originated in the agonist ’ s self and which is close to volition rather Chapter 4. Middle English thurven (& durren ), bethurven, neden, bihoven & misteren 259 than to obligation. Only on one occasion is the force positive. In the other five cases thurven expresses the absence of a weak internal force, as is in (4.54), for instance: (4.54) By this proverbe thou shalt understonde, by this proverb you shall understand Have thou ynogh, what thar thee recche or care have you enough what need you worry or care How myrily that othere folkes fare? how happily that other folks travel (The proverb implies that he is m ost contented who is not envious of others) ‘ By this proverb you shall understand, if you have enough, what need have you to concern yourself with how happily other people live? ’ (from Blake 1980: 187-188, n.327, 329-330) / “ why should you care a curse how well-off other people are? ” (from Wright 1985: 227) (4,840 helsinki\cmctvers) In this fragment of The Wife of Bath ’ s Tale , thurven expresses the absence of a weak internal force which makes the agonist envy others. The double translation I offer is taken from two different editions of the Canterbury Tales . The PDE versions of Blake (1980) and Wright (1985) reveal that these authors also consider the force expressed by thar to be weak, since they choose respectively ‘ what need have you? ’ and ‘ why should you?, ’ instead of a strong ‘ why must you? ’ The last line of Table 4.8 shows that ME thurven also expresses general types of forces on nine occasions, that is, forces which are exerted by a nebulous, generalized authority, as defined by Langacker (1999: 308). As already mentioned, the strength of such f orces is neutral, that is, neither strong nor weak. The clause polarity is non-affirmative in all nine cases, and the meaning conveyed is lack of force. Sentence (4.55) is a good illustration of this kind of force: (4.55) Þ o qua þ þ e hule "[W ] u schal us seme, then says the owl who shall us reconcile þ at kunne & wille rit us deme?" that can & will right us judge "Ich wot wel" qua þ þ e nitingale, I know well says the nightingale "Ne þ aref þ arof bo no tale. not need thereof but no conversation Maister Nichole of Guldeforde, Mister Nichole of Guldeforde Chapter 4. Middle English thurven (& durren ), bethurven, neden, bihoven & misteren 260 he is wis an war of worde. he is wise and prepared of words ‘ Then said the owl: “ Who shall reconcile us that can and will judge us right? ” “ I know well, ” said the nightingale, “ There need/shall be no conversation / talk, Mister Nicole of Gulderforde, he is wise and prepared with words ” . ’ (1190-1210 The Owl and the Nightingale , lines 187-192) The verbal form þ aref in this example expresses the absence of a force, and such a force is not concretely defined, but general. The sentence means, then, ‘ there is no general necessity for a conversation. ’ Anot her common context for the expression of this kind of force concerns those cases in which it is generally stated that ‘ no one need ask or tell how anything happened. ’ This context f or the expression of a general type of force, which is exemplified in (4.56), is also commonly found with other ME verbs such as neden or bihoven : (4.56) So þ at he cam to caunturburi : and dude ase ri Ñ t was # þ ere, so that he came to Canterbury and did as right was there And Erchebischop was i-maud : is vn þ onkes þ ei it were. and archbishop was made is unwilling though it were Ne þ arf no man þ ar-of esche : Ñ we þ ur he toke on wel i-nou Ñ not need no man thereof ask whether he took on well enough And wel wissede holie churche : and to eche guodness drou Ñ ! and well guided holy church and to each goodness moves ‘ So that he came to Canterbury and did as was right there, and was m ade archbishop, though it was unwillingly. No man need/m ust ask whether he took on well enough or guided well the holy ch urch, and to each goodness moves! ’ (6,034 helsinki\cmseleg) With this example, which will be referred to below when analysing the features of other ‘ need ’ -verbs, I close this section of the semantic characteristics of ME thurven . As far as ME bethurven is concerned, the onl y type of meaning conveyed by the four examples in my ME corpus is strong internal force . One o f such examples is (4.57): (4.57) & hwitere gose smere anes sceallinges. wyht . & euforbeo swa micel. & & white goose grass one schilling weight & euphorbia so much & wyne æ l togadere. & do in ane boxs. & nime sy þþ an swa oft. swa dry altogether & do in one box & take afterwards so often so he be þ urfe . he needs Chapter 4. Middle English thurven (& durren ), bethurven, neden, bihoven & misteren 261 ‘ and the weight of one shilling of white goose grass & the same amount of euphorbia and dry it all together and put it in one box and afterwards take so often as he needs. ’ (4,898 helsinki\cmperidi) This sentence is an excerpt of the medicine handbook Peri Didaxeon , and it represents a piece of advice or instruction about how to proceed in case of sickness. The meaning of thurve n in this sentence seems to be clearly that of a strong internal force, since the necessity is related to health (hence it is a strong force) and it originates in the agonist ’ s own body (hence it is internal). In addition to (4.57) there are other three examples, two of which also belong to the same text, and in all three cases bethurven refers to the strong necessities of the patient who wants to be healed. Interestingly enough, only one out of the four examples of bethurven occurs in non-affirmative contexts (i.e. 25%), while thurven shows a strong preference for such contexts (94.5%). This was also the case of their OE counterparts, since OE þ urfan occurs in non affir mative contexts 90.5% of its occurrences and OE be þ urfan 25.5%). Therefore , thurven and bethurven seem to have a complementary distribution as far as clause polarit y is concerned. We will see below whether such a distribution is also witnessed from a syntactic point of view. The main conclusions whic h can be drawn from the analysis of thurven and bethur ven are similar to those of Old English, although th ese verbs have undergone a clear reduction in the range of the possible meanings which they may express. Such a reduction is most evident in bethurven , which is highly restricted to the expression of i nner necessity in the very early years of Middle English. As for thurven , we have seen that it rarely expresses the presence of a force. In non-affirmative contexts, it never expresses prohibition, but a bsence of necessity. The following table graphically represents the possible semantic values of this verb, as compared to its OE c ounterpart. Revealingly enough, ME thurven does not convey any meaning which has not been attested for OE þ urfan , that is, it only undergoes semantic losses. Chapter 4. Middle English thurven (& durren ), bethurven, neden, bihoven & misteren 268 4.4.2. Middle English neden in the corpus As mentioned in section 4.3.2, t he label neden comprises two ME verbs which coalesce under the same form. They both express some kind of force, and have evolved from the OE form neodi an . We have seen that in Old English m o st of the instances of neodian convey the meaning of ‘ press, compel, force, ’ while those forms of neodian meaning ‘ need ’ or ‘ be necessary ’ are extremely rare (only one case in my 1.2 million-word corpus). In Middle English this proportion is reversed: out of 161 total examples of neden , only 15 express ‘ press, compel, force; ’ these will be referred to as neden v.1, following the MED . As expected, the disappearance of the use of neden to express ‘ compel, force ’ is also attested as a gradual process within the subperiods of Middle English, since it occurs seven times in M1, twice in M2, six times in M3, and, fi nally, no instance is recorded in M4, the last ME subperiod. On the contrary, my corpus contains 146 examples of neden forms meaning ‘ need, be necessary, ’ which will be referred to as forms of neden v.2, following the MED . Their frequency increases as Middle English advances with two cases in M1, no instance in M2, 72 instances in M3 and other 72 instances in M4. The following table illustrates graphically the decay of neden v.1 a nd the rise of neden v.2, as found in my corpus: Neden v.1 Neden v.2 TOTAL VERB PERIOD NUMBER N.F. NUMBER N.F. NUMBER N.F. M1 7 2.42 2 0.69 9 3.12 M2 2 0.97 2 0.97 M3 6 1.64 72 19.68 78 21.32 M4 72 18.59 72 18.59 TOTAL 15 1.20 146 11.69 161 12.89 Table 4.13: Distribution of ME neden v.1 and v.2 by subperiods. 19 This table shows that neden v.1 is primarily found in early Middle English, while neden v.2 reaches its peak of frequency in late Middle English. Sections 4.4.2.1 and 4.4.2.2 deal with the semantics and syntax of these verbs respectively. 19 The seven examples of ned en v.1 in M1 also contain one instance coded as MX/1 in the Helsinki Corpus , and the 72 instances of neden v.2 in M3 contain six examples of texts from M3/4. Chapter 4. Middle English thurven (& durren ), bethurven, neden, bihoven & misteren 269 4.4.2.1. Semantic features of Middle English neden In the OE section of t his study, I analysed all neden for ms together, because the chances to find instances of ne odian meaning ‘ need ’ were so low that it was pointless to analyse all the semantic aspects of each verb indivi dually. However, the numbers in Middle English allow for an analysis for each of the verbs separately, and, therefore, I will begin my semantic analysis with the forms of neden v. 1, and then analyse those of neden v.2. ME neden v.1 ‘ press, compel, force ’ expresses only strong external forces, the most common type of meaning expressed in Old English. Like OE neodian , ME neden v.1 is une xpectedly frequent in the passive voice, namely in 40% of the occasions, that is, above the OE ratio of 25%. Since, as mentioned, voice determines semantic aspects such as the nature of the subject as agonist or antagonist, the semantic analysis of this verb must necessarily take into account this syntactic parameter. Table 4.14 sketches the types of strong external force expressed by neden v.1: VOICE TYPE OF FORCE ACTIVE PASSIVE TOTAL PHYSICAL 2 2 HIERARCHICAL 7 7 RELIGIOUS 6 6 TOTAL 9 6 15 Table 4.14: Types of strong external forces expressed by active and passive neden v.1. As can be seen in Table 4.14, the number of possible notional types of strong external forces expressed by neden v. 1 is reduced from that in Old English (cf. Tables 3.28 and 3.29, section 3.4.2.1). However, ME ne den v.1 still keeps important values which illustrate the wide range of semantic implications which fall under the scope of forces. I will not break down this table into others devoted to each of the notional types of forces, because the numbers are too low and do not require further specification. I must say, however, that three out of the fifteen examples occur in negative contexts and express absenc e of force. Now I proceed to illustrate and examine the three types of forces expressed by neden v.1. From a force-dynamic perspective, the expr ession of meanings related to the physical world is, as mentioned in section 2.2.2.2, the most distant semantic notion with respect t o the expression of modal meanings. Since I have been Chapter 4. Middle English thurven (& durren ), bethurven, neden, bihoven & misteren 270 describing necessity in terms of forces, it seems appropriate to allude to the fact that the expression of physical force is the natur al predecessor of modal necessity, and this evolution has been observed in OE neodi an (cf. section 3.4.2.1). Middle English also exhibits instances of neden v.1 expre ssing physical force, such as (4.65): (4.65) he crepe ð cripelande for ð , it creeps creeping forth his craft he ð us ki ð e ð , his craft it does shows Seke ð a ston ð at a ð irl is on, looks-for a stone that a hole is on Narwe, buten he nede ð him, narrow but it (nom.) presses it (obl.) Nime ð vnne ð es ð urg, takes with-difficulty through for his fel he ð er lete ð . for its skin it there leaves ‘ (the adder) it creeps forth, (and) thus shows its craft, looks for a stone with a hole; narrow but it presses itself, goes through with difficulty, for it leaves its skin there. ’ (748 helsinki\cmbestia) This is a clear example of physical force, because the adder presses itself to go through a small hole in order to l eave its skin there. As repeatedly said, it is from these physical meanings that metaphorical meanings s uch as ‘ to press someone to do something, to oblige ’ evolve. The following example illustrates this metaphorical force based on the existence of a hierarchical superiority: (4.66) Ð e hali gast hem warne ð , and sei ð : 'Godd ð e Ñ eu # haue ð icleped the holy ghost them warns and says God who you (obl.) has summoned ut of ð are lease woreld in to gode liue, he ne nett # Ñ eu naht ð er to, out of the false world in to good life he not compel you (obl.) not thereto ac sei ð : (\Si uis perfectus esse,\). but says if you-want perfect be ‘ The holy ghost then warns and says: “ God, who has summoned you out of the false world into the good life, he does not compel you thereto, but says: if you want to be perfect... ”’ (5,767 helsinki\cmvices1) Sentence (4.66), which is also quoted in the MED ( s.v. neden v.1 (b)), illustrates a type of obligation or, rather, absence of obligation, based on hierarchy: since God stays over all things, he is a superior entity able to compel or not compel anyone. He is the antagonist who forces the agonist ( you ) to perform his will. Chapter 4. Middle English thurven (& durren ), bethurven, neden, bihoven & misteren 271 Finally, we may have a look at one of the examples of neden expressing force based on religion. As seen in Table 4.14, all such examples occur in the passive voice, which implies that the subject is no longer the antagonist, but the agonist. Consider, for instance, (4.67): (4.67) as holy writt maketh mynde / oure lorde wepte notably thre tymes: o ne tyme as holy writ makes notice our lord wept notably three times one time / in the deth of laare / the wrecchedn esse of mankynde wherby he is nedede in the death of lore the wretchedness of mankind whereby he is com pelled to deye for the firste syn ne. to die for the first sin ‘ as the holy writs notice, our Lord wept notably three tim es: one time in the death of doctrine, the wretchedness of mankind whereby it is com pelled to die for the first sin. ’ (fl1410 Mirrour of the blessed lyf of Jesu Christ ) The agonis t of this sentence is expressed by the nominative pronoun he , which stands for the noun mankind , although this noun is originally neuter (cf. OE mancynn n., as found in Clark Hall). A paraphrase of this sentence could be because of the first sin (committed by Adam and Eve when they ate the apple and disobeyed God), mankind must die . In this paraphrase I have replaced the passive form is compelled for a modal verb, must , because their connotations are fairly close, at least when must expresses root modality. It could be said, therefore, that examples such as (4.67) represent the closest stage of neden v.1 to the field of modality, in which neden v.2 is included. Though the number of examples of neden v. 1 in my ME corpus is very low, it allows for some conclusions. Firstly, the use of this verb decreases gradually in the ME period. Secondly, its meanings keep the main distinctive OE features, that is, it may express both physical and social force, which in turn may be based on hierarchical or re ligious reasons. And, thirdly, the passive constructions of nede n v.1 represent a bridge between the expression of physical force and the expression of metaphorical forces of the type that may be found with neden v.2, because in both cases the subject is the agonist, and the meaning expressed is related to a type of necessity, which m ay be of a weaker or a stronger character. The following para graphs concentrate on the analysis of the examples of neden v.2 as found in my corpus. To begin with, I must point out that there is an important difference between neden v. 1 and neden v. 2, as regards polarity. While Chapter 4. Middle English thurven (& durren ), bethurven, neden, bihoven & misteren 272 neden v. 1 occurs in non-affirmative examples only on 25% of its occurrences, neden v.2 exhi bits quite a high ratio of non-affirmative instances, 82 cases, namely more than 56% of its occurrences. As is well-known, PDE modal need is also mainly concerned with non-affirmative examples. Therefore, the ME data seem to begin to show this tendency. In addition, ME thurven like its predecessor OE þ urfa n, is also especially frequent in non-affirmative contexts (see above 4.4.1, and Table 4.10) . This may imply that ME thurven and neden v.2 compete to some extent in the expression of the sa me kind of meanings. This is what I aim to show in the paragraphs which follow. To begin with, ME nede n v.2, the same as ME thurven and OE þ urfan , does not always express necessity, but it may also express possibility, which may be defined in terms of cognitive barriers. When barriers occur in non-affirmative contexts, the event is blocked. This is the case of three examples out of the total 146 examples of neden v.2. Consider (4.68): (4.68) These dide merveilously wele, and so dide Sir Ewein, that a better knyght these did marvellously well and so did Sir Ewein that a better knight than he neded no man to be-holde. than he needed no man to behold ‘ These did marvellously well, and so did Sir Ewein, that no m an could behold a better knight than him. ’ (1450-1460 Merlin ) This is one of the three examples in which neden v.2 expresses the existence of a barrier. The context is similar to those of thurven , mentioned above, and illustrated with sentence (4.48) in this chapter, because in both cases the verb follows a comparative adjective in order to imply that there cannot be anyt hing better than what is being described. The MED ( s.v. neden v.2) does not include this type of construction, but, as me ntioned, it occurs thrice in my corpus. All of them appear in the text Merlin and, therefore they may be considered to be influenced by text ual factors such as the idiolect or the dialect of the author. In any case, these three examples represent a piece of evidence that neden v.2 and thurven express similar types of meaning in similar types of construction. The other 143 examples of neden v.2 express some kind of necessity which can be described in terms of cognitive forces, as shown in Table 4.15, which follows the s ame model used above to outline the general semantic features of each verb, accounting for the origin and strength of the force: Chapter 4. Middle English thurven (& durren ), bethurven, neden, bihoven & misteren 273 ORIGIN STRENGTH N. OF EXAMPLES TOTAL STRONG 33 EXTERNAL WEAK 33 STRONG 48 INTERNAL WEAK 22 70 GENERAL NEUTRAL 40 40 STRONG 81 WEAK 22 TOTAL NEUTRAL 40 143 Table 4.15: Origin and intensity of the forces expressed by ME neden v.2. We observe that nede n v.2 may express forces originated in external or internal entities, as well as forces of diffuse origin, i.e. general forces. The strength with which they are exerted may be strong, weak or neutral. It is necessary now to break down this table into other tables describing with detail each of t he forces expressed. As shown in Table 4.16, ME neden v.2 e xpresses strong external force on 33 occasions: NON-AFFIRMATIVE CLAUSE POLARITY TYPE OF FORCE AFFIRMATIVE LACK OF FORCE FORCE NOT TO TOTAL RELIGIOUS 15 7 22 HIERARCHICAL 1 6 7 APPROPRIATENESS 3 3 LEGAL 1 1 TOTAL 17 13 3 33 Table 4.16: Types of strong external forces expressed by neden v.2, with specification of clause polarity. This table shows that strong external neden v.2 i s mostly concerned with religious matters, though it may also be due to other types of notional f orces. In the polarity axis, it is worth pointing out that nearly half of the instances of strong external neden v.2 are non-affirmative, and that in most of them the verb expresses absence of force, lack of necessity. As for affirmative religious force a good example is (4.69): (4.69) … and God lovith wol tenderly us while (we) be in synne, and so us nedyth … and God loves well tenderly us while (we) be in sin and so us need to doe our neybor. to do our neighbour ‘… and God loves us tenderly while we are in sin, and so we must do to our neighbours (lit.: so it is necessary to do to our neighbours). ’ (1373 A Revelation of Love ) One of the Ten Commandments says that we must love our neighbour. Sente nce (4.69) expresses this idea, that is, there is a strong religious force which inflicts Chapter 4. Middle English thurven (& durren ), bethurven, neden, bihoven & misteren 274 on us the imposition to love our neighbours. This justifies the choice for must in the translation rather than need . A similar kind of strong external force expressed by neden v.2 is that based on a hierarchical superiority, which only occurs once in affirmative contexts in my corpus. Consider (4.70): (4.70) so seid he to his folkes: "I pray you remembre wele thies matiers, for it nedith to take good aduise; for our abidyng here is fulle noious and dangerous / & oure departyng shold be shameful ý . ” ‘ so he said to his folks: “ I pray you remember well these matters, for you must take good advice; for our abiding here is fully annoying and dangerous and our departing should be shameful. ”’ (1500 The Three Kings ’ Sons ) As is the case of (4.69), neden v.2 expresses the existence of a strong external force which is best translated with the modal of obligation must . In the case of (4.70), the source of the obligation is a kni ght who is giving commands to his people and reminds them that it is their obligation to take good ad vice. The only difference between examples such as (4.69) and examples such as (4.70) lies on the nature of the external authority which inflicts an imposition on the agonist. A third type of imposition is that which I have labelled as legal (cf. Table 4.16, fourth line). The ver b neden v.2 is also found expressing an obli gation inflicted by law on one occasion. I turn now to the analysis of those instances of neden v.2 expressing the absence of a strong external for ce, namely lack of obligation or of necessity. According to Ta ble 4.16, neden v.2 may e xpress absence of for ces based on religion and hierarchi cal superiority. A para digmatic example of absence of religious obligation is (4.71): (4.71) Here may we sen that we arn al bound to God for kinde, snd we arn al bound to God for grace. Here may we sen us nedith not gretly to sek en fer out to knowen sundry kindes, but to holy church. ‘ Here we may see that we are all bound to God for kind and grace. Here we may see we need not (lit.: it is not necessary for us to) search far out to know different kinds, but the holy church. ’ (1373 A Revelation of Love ) In sentence (4.71) nedith expresses the absence of necessity for believers to look for other types of religion, because they find all they need in the Holy Church. This means that the agonists, the believers, are exempted from wandering in search of explanations by an external antagonist, namely Christianity. In this respect, neden v.2 is semantically similar to thu rven , because the latter also Chapter 4. Middle English thurven (& durren ), bethurven, neden, bihoven & misteren 275 shows a tendency to express abse nce of religious necessity or obligation, as seen above (cf. ex. (4.50)). Finally, let us consider those instances in which neden v. 2 occurs in non- affirmative contexts and expresses a force not to act in a given way. An interesting type of force not to is seen in sentence (4.72), where the source of the force is strong and external, and is based on the appropriateness of circumstances: (4.72) … for she wolde speke to moche and clatre there it nedithe not. … for she would speak too much and clatter there it is-necessary not ‘ (one out of three sisters is selected to be the queen, and the second has been rejected) for she would speak too much and there it is not appropriate to clatter. ’ (1484 Book of the Knight and La Tour-Landry ) The previous conte xt of this sentence, which I provide between brackets, is crucial to understand the meaning of nedithe . It seems fairly obvious that if speaking too m uch were not necessary, it would not be a reason to reject the second sister, it would just be an unnecessary quality. However, it appears that her logorrhea is a handicap or disadvantage for her to become a queen. Therefore, the meaning of nedithe in (4.72) must be one of absence of appropriateness, and the meaning is that it is not appropriate for a queen to clatter. It could be claimed that this mean ing is closer to expressing the existence of a barrier than the existence of a force. In other wor ds, (4.72) and the other two examples which express absence of appr opriateness might be understood to be examples of neden v.2 expres sing possibility rather than necessity (sentence (4.72) would then mean ‘ it is impossible to clatter there ’ ). Since the MED ( s.v. neden v. 2) does not mention that such a semantic value is possible, I have decided to analyse these examples as instances of absence of strong external force not to act in a given way. They are not examples of harsh prohibition, but rather the context explains that circumstances make some types of behaviour inopportune, and there is an implied force for the agonists not to act according to such types of behaviour. After examining the possible semantic nuances of neden v. 2 when it expresses strong external forces, we must move on in Table 4.15, and obser ve that in no case does this verb express weak external force. As repeatedly mentioned, the interpretation of a force as weak or strong is fairly subjective. As Chapter 4. Middle English thurven (& durren ), bethurven, neden, bihoven & misteren 276 a consequence, the a bsence of examples expressing this kind of force should not be necessarily interpreted as a gap in t he semantic map of neden v.2, but may be attributed to my subjective analysis of the examples. So far we far examined all possible types of strong forces expressed by neden v.2; now I will concentrate on the internally-rooted forces, paying attention to strong internal forces first, and to weak internal forces afterwards. ME neden v.2 expresses strong internal forces in 48 instances of my corpus, as sketched in Table 4.17: NON-AFFIRMATIVE CLAUSE POLARITY TYPE OF FORCE AFFIRMATIVE LACK OF FORCE FORCE NOT TO TOTAL INNER 25 19 1 45 INNER-PHYSIOLOGICAL 3 3 TOTAL 28 19 1 48 Table 4.17: Types of strong internal forces expressed by ME neden v.2, with indication of clause polarity. Table 4.17 displays the possible types of strong internal forces expressed by neden v.2. Up to now all internal forces expressed by my verbs have bee n characterized simply as inner, because they have their or igin in the agonist ’ s self. However, Table 4.17 introduces a new type of internally rooted force which is not related to the agonist ’ s self, to his will, desires of emotional needs, but to the agonist ’ s physiological necessities. Table 4.17 also shows the polarity of t h e sentences in which the verb occurs. The percentage of non-affirmative sentences decreases as compared to strong external forces, but it still represents a fairly high proportion. Finally, as was the case with external forces, we also observe one negative example of strong internal force which expresses a force not to, rather than the absence of a force. Beginning with the affirmative examples, (4.73) illustrates the use of neden v. 2 whe n expressing the presence of a strong internal force: (4.73) 600 (...) he bad his folk leuen, he bad his folk leave 601 And only seruen him-self · & hijs rewle sechen, and only serve himself and his rule seek 602 And all þ at nedly nede þ þ at schuld hem nout lakken. and all that necessarily is-necessary that should them not lack ‘ He bad his folk leave and only serve himself and search his rule, and all that is necessarily necessary, which they should not lack. ’ (1394 Pierce the Ploughman ’ s Crede ) Chapter 4. Middle English thurven (& durren ), bethurven, neden, bihoven & misteren 277 The form nede þ expresses a strong internal necessity for the agonists: the authority who bids them leave also suggests that they should search everything that they should not lack, that is, all that is ‘ necessarily necessary ’ for them. If it is necessarily necessary, the force expressed is strong, and if it refers to something they must have for themselves, the origin of such a force is internal. This clear example of strong internal force expressed by nede n v.2 also illustrates the use of the adverb nedly , ‘ necessarily, ’ which is very frequent since Old English in combination with þ urfan and will also be ve ry common with ME bihoven . In sentence (4.73) it also serves the purpose of reinforcing the meaning of the verb. As an example of what I have called strong inner-physiologic al force, witness (4.74): (4.74) And þ us schalt þ ou knowe when þ in hors nede þ to be I-lete blod. ± if he and thus shall you know when your horse needs to be let blood if he be ranke of blod he wol gnappe himself & rubbe him a- Ñ ens þ e be strongly-sm elly of blood he will gnaw himself & rub him against the walle þ at he stonde þ bi. wall that he stands by ‘ And thus you shall know when your horse needs to let blood. If it is strongly smelly of blood, it will gnaw itself and rub itself against the wall that it stands by. ’ (614 helsinki\cmhorses) The other two examples of strong internal inner-physiological force expressed by neden v.2 are also related to blood-letting or venesection, which, according to the OED ( s.v. phlebot omy n. 1), is a medical practice used therapeutically. When a person or an animal, as in this case, suffers some kind of illness, it may be necessary for them to have a vein cut so that blood flows. This practice must have been quite common to judge from the OED ( s.v. blood n. I 1 (d)), which includes a series of quotations of different periods of English, as well as an explanation of the passive use of this expre ssion, as is the case in sentence (4.74). Therefore, it seems fairly evident that in this context neden v.2 expresses strong internal force, and the origin of the force is not on th e agonist ’ s self, but on the agonist ’ s body. Hence its characterization as inner-physiological force. Moving on to the non-affirmative examples of neden v.2, we observe in Table 4.17 that in most of the cases it expresses absence of force, as in (4.75): (4.75) What wirshippe shulde we wynne therby? what worship should we win thereby Chapter 4. Middle English thurven (& durren ), bethurven, neden, bihoven & misteren 284 The noun phrase al þæ o ð ere , ‘ all the others, ’ is the direct object of nedden and its only complement. This syntactic pattern occurs three times with active ne den v.1. On some occasions the sentence also specifies the type of imposition which is inflicted on the agonist. As mentioned, the i mposition may have the form of a to -infinitival clause, or a to -prepositional phrase, as seen in (4.83) and (4.84) respectively: (4.83) Willfulnesse lette ðþ e mannes shrift. þ at þ inche ð uuel þ at man him wilfulness obstructs the man ’ s confession who thinks much that man him wile neden his sinnes to forleten . will compel his sins to forsake ‘ Wilfulness obstructs the m an's confession, who thinks (m uch) that he is compelled to forsake (i.e. deny) his sins. ’ (3,242 helsinki\cmtrinit) (4.84) Sume wene ð bien sacleas of ð essere senne, for ð an ð em e nett some think be innocent of these sins because man / one compel hem to ð an a ð e. them to the oath ‘ Some think they are in nocent of these sins, because th ey are compelled to the oath/curse. ’ (1,252 helsinki\cmvices1) In sentence (4.83) the imposition inflicted on the agonist, him , is expressed by the to -infinitival clause his sinnes to forleten . This is the most frequent type of complementation of neden v.1 in active sentences, since it occurs in four out of the nine cases. This is an interesting innovation of Middle English, because this construction was the least common one for active neodian in Old English (cf. Table 3.31), which pre ferred that -clauses to all other types of complementation. The ME syntactic complementation pattern of neden v. 1, therefore, comes closer to the types of complements found in Present-Day English with semantically similar verbs such as compel or force . Sentence (4.84), in turn, illustrates the possibility for neden v.1 to be complemented by a to -prepositional phrase, to ð an a ð e , a type of complementation which was also productive in Old English. As far as the passive instances of neden v.1 are concerned, in all six cases the verb is complemented by a to -infinitival clause: (4.85) And so Crist suffrede more # freely þ an Baptist or o þ re martires, but he and so Christ suffered m ore freely than Baptist or other martyrs but he [Document text truncated for crawler view.]