Вы находитесь на странице: 1из 36

Modals in the Germanic languages1 Tanja Mortelmans, Kasper Boye and Johan van der Auwera

1. Introduction: the Germanic modals In contrast to e.g. Romance (see Cornillie et al., this volume) and Slavonic (see Hansen 2004, this volume) modals, the Germanic modals can be claimed to stand out morphologically, syntactically and semantically such that they can be said to build a grammatical paradigm. From a morphosyntactic point of view, the central members of the modal verb class are preterite-presents, on the one hand, and are followed by a bare infinitive, on the other. Semantically, the Germanic modals can be characterized by the fact that they express both root (i.e. non-epistemic) and epistemic modal meanings2. This implies neither that the class of modal verbs can be easily delineated there are peripheral candidates striving for admittance (e.g. brauchen in German, behve in Danish) nor that the class of modal auxiliaries or expressions is diachronically stable (cf. Krugs (2000) discussion of emerging modals in English). Moreover, it is well-known that there are differences between the modals in the different Germanic languages, both with respect to individual modal items within the paradigm of a particular language (the German modal wollen, for instance, seems to be less grammaticalized than the other modals of its paradigm, see section 3) and crosslinguistically with respect to entire paradigms (such that the entire modal verb paradigm in one language can be said to be grammaticalized to a larger extent than in another; the English modal verb paradigm is a case in point). In the following, we will compare the various core or central modals in the Germanic languages by means of Lehmanns parameters of grammaticalization As these parameters concern three aspects of grammaticalization (weight, cohesion and variability), which are further differentiated along a syntagmatic and paradigmatic axis, we want to establish whether the grammaticalization of the Germanic modals involves the various processes to the same extent.
Consequently, if we want to measure the degree to which a sign is grammaticalized, we will determine its degree of autonomy. This has three principal aspects. First, in order to be autonomous, a sign must have a certain weight, a property which renders it distinct from the members of its class and endows it with prominence in the syntagm. Second, autonomy decreases to the extent that a sign systematically contracts certain relations with other signs; the factor inherent in such relations which detracts from autonomy will be called cohesion. Third, a sign is the more autonomous the more variability it enjoys; this means a momentary mobility or shiftability with respect to other signs (Lehmann 2002: 1089)

Whereas a signs weight (both in terms of its paradigmatic weight (integrity) and its syntagmatic weight (structural scope) and the degree of paradigmatic and syntagmatic variability are assumed to decrease with increasing grammaticalization, a signs paradigmaticity (its degree of paradigmatic cohesion) and its degree of bondedness (its syntagmatic cohesion) are expected to rise with grammaticalization. In the following table, an overview is given of the six processes leading to an increased degree of grammaticalization.

Tanja Mortelmans, Kasper Boye & Johan van der Auwera

Table 1. Lehmanns parameters and grammaticalization processes (on the basis of Lehmann 1995: 1255) Parameter integrity Weak grammaticalization bundle of semantic features; possibly polysyllabic item participate loosely in semantic field free choice of items according to communicative intentions item relates to constituent of arbitrary complexity item is independently juxtaposed item can be shifted around freely Process attrition Strong grammaticalization few semantic features; oligo- or monosegmental small, tightly integrated paradigm choice systematically constrained, use largely obligatory item modifies word or stem item is affix or even phonological feature of carrier item occupies fixed slot

paradigmaticity paradigmatic variability structural scope bondedness

paradigmatization obligatorification

condensation coalescence

syntagmatic variability

fixation

In our study of the Germanic languages, we will focus on the central modals in five Germanic languages3 only: English, Dutch, German, Danish and Icelandic. The highly grammaticalized English modals the grammatical properties of which have been described in great detail in numerous studies (for instance, Quirk et al. 1985, Biber et al. 1999) will function as a starting point and standard of comparison; we will offer more detailed accounts of the central modals in Dutch and German (as representatives of the West Germanic languages) and of Danish and Icelandic (representing the North Germanic or Scandinavian languages), an inventory of which is given in table 2. As stated already, we consider as central modals those that are preterite-presents (including will and its counterparts), on the one hand, and combine with a bare infinitive, on the other. As high text frequency is a typical characteristic of grammatical items (see Bybee 2003: 603), we also assume that central modals being the most strongly grammaticalized ones are frequent.
Table 2. The central modals in five Germanic languages ENG can shall may will5 must / / / (ought) DUT kunnen zullen mogen willen moeten / / / GER knnen sollen mgen wollen mssen drfen / / DAN kunne skulle mtte ville / (turde)6 (burde) / ICE (kunna)4 skulu mega (vilja) / (urfa) munu (eiga)

Not only will we restrict ourselves to mainly five Germanic languages, we will also leave out of account the parameters of (formal) integrity, bondedness and syntagmatic variability. Formal integrity7 and bondedness (syntagmatic cohesion)8 seem to be less relevant for the modals in the languages under account. Decrease of syntagmatic variability, which manifests itself through increasing fixation, can be traced back to general constraints regarding word order in the various languages rather than to individual properties of the modal verbs themselves.9 Our focus will therefore lie on the parameters of paradigmatic variability, paradigmaticity, and structural scope. As far as these remaining parameters are concerned, it has been observed

Modals in the Germanic languages

that English does not formally distinguish between non-epistemic and epistemic meanings. The English central modals show a very high degree of paradigmaticity, on the one hand, and a low degree of paradigmatic variability and structural scope, on the other, and this irrespective of whether they express non-epistemic or epistemic meanings (see Goossens 1985 for a similar claim). On the semantic side, the central modals do not seem to be able to express lexical or premodal meanings like to like, to want or to know anymore, which points to the fact that there has been a lot of semantic attrition. Our starting point will therefore be the high degree of grammaticalization of the entire set of English modals. In fact, four morphosyntactic features can be distinguished that set the central modals clearly apart from other auxiliary verbs and lexical verbs (Krug 2000: 44). The English modals have no s form in the present tense (*he cans), i.e. they make no person or number distinctions, they do not have non-finite forms (*canning, *to can), they show abnormal time reference to the extent that past tense forms do not normally express past time and they can only be combined with a bare infinitive (He must be there, *He must it). These features are indicative of the decrease of intraparadigmatic variability (as each modal has either only one (must) or two (may/might) forms at its disposal, its formal variability is low) and structural scope (as the English modals can only modify another infinitive) and the increase of paradigmaticity (the set of English modals is limited and formally highly similar). These features, however, do not hold for the German, Dutch or Danish modals: they all have non-finite forms, for instance, their past tense forms do refer to past time, and they allow for complementation patterns other than the bare infinitive. However, a closer look at the modals in other Germanic languages reveals that non-finite uses are not always sanctioned; more particularly, it seems that the non-English Germanic modals in clear contrast to the English ones formally distinguish between non-epistemic and epistemic modality, whereby the latter is generally more strongly grammaticalized than the former.10 In the following sections, we will address the relevant parameters for the Germanic languages under consideration. We will first present the inventory of the central modals in the respective languages, have a look at the semantic range of the central modals (whether or not they still express non-modal11 meanings, for instance) and then apply the various parameters to the members of the modal paradigm, in order to establish differences between individual members of the paradigm. We will thereby focus on three questions:
(a) if and to what extent is the distinction between non-epistemic and epistemic meanings relevant for the grammaticalization degree of the various modals in the languages under consideration? (b) are the central modals in the Germanic languages under consideration as a group grammaticalized to the same extent, that is: are there languages in which the central modals overall seem to have acquired a higher degree of grammaticalization in comparison to the central modals in another language? (c) are there differences between the individual items of a languages central modal paradigm regarding degree of grammaticalization?

Before embarking on our analysis of the modals in the other Germanic languages, we will offer a brief description of the range of meanings expressed by the modals, the aim of which is to clarify a number of semantic concepts. Modal meanings can be classified in various ways. A crucial distinction for this paper is the one between non-epistemic and epistemic modality, whereby the latter concerns an indication of the estimation, typically, but not necessarily by the speaker, of the chances that the state of affairs expressed in the clause applies in the world (Nuyts 2006: 6), as in (1) Someone is knocking at the door. That will be John.

As counterparts of will prominently figure in table 2, we will include verbs of volition in our account, whereby it should be clear that these counterparts have not reached the high degree

Tanja Mortelmans, Kasper Boye & Johan van der Auwera

of grammaticalization of English will, which tends to express epistemic meanings (prediction and predictability) considerably more often than volitional ones (see Coates 1983: 171). An adequate classification might distinguish between dynamic and deontic meanings (in the non-epistemic realm of modality). Dynamic modality is typically taken to include capacities/abilities/potentials and needs/necessities [] inherent to the first argument participant (Nuyts 2006). Nuyts, in accordance with Palmer (2001), also includes cases in which the potential or necessity must be ascribed to (i.e. is inherent in) the situation as such (and is not dependent on a particular participant), as in (2). (2) a. b. The book need not be in the library. It can also be on my desk. John cannot be in Spain. Ive just seen him in the grocery store.

As will appear in the discussion to come, this type of dynamic modality (termed circumstantial dynamic (Palmer 2001), situational dynamic (Nuyts 2006), non-deontic participant-external (van der Auwera and Plungian (1998)) is hard to distinguish from genuine epistemic modality. Finally, deontic modality offers an indication of the degree of moral desirability of the state of affairs, as in (3). (3) We should be thankful for what he has done for us, so we must find a way to show our gratitude to him.

Whereas deontic modality will not figure prominently in our discussion (in fact, most modals seem to be able to express deontic meanings, without this having important repercussions on their grammatical behavior), we will have to introduce the category of evidential modality or evidentiality, which refers to the source of evidence the speaker has for his or her statement (De Haan 2006: 57). Evidentiality concerns the expression of direct and indirect sources of information, and in the latter category another subdivision is made between inferentials and hearsays or reportatives. Interestingly, it will appear that hearsay evidentiality (the speaker was not a witness but obtained knowledge about the action from another person, De Haan 2006: 57) is generally expressed by a member of the core modals in the Germanic languages under discussion (e.g. sollen in German, zou (and moeten, see De Haan 2001) in Dutch, skulle in Danish and munu in Icelandic) with the English modals as an important exception (Old English sculan has been attested with evidential meaning (Traugott 1989: 41), which it has lost, however). In the following sections, we will present the central modals paradigms for Dutch (section 2), German (section 3), Danish (section 4) and Icelandic (section 5). For each section, a general introductory overview will be given, followed by a detailed discussion of Lehmanns parameters of paradigmaticity, paradigmatic variability and structural scope for the modals in each language.

2.

The Dutch modals

The group of Dutch core modals has only five members: kunnen (can), zullen (shall), mogen (may), moeten (must) and willen (will). Often, the verb hoeven a negative polarity item the meaning of which can be compared to that of need in English or brauchen in German is mentioned as a sixth modal verb (see for instance the list of modal verbs in the Dutch reference grammar Algemene Nederlandse Spraakkunst (Haeseryn et al. 1997: 984)). In contrast to the other ones, however, hoeven is not a preterite-present and it is followed by a te-infinitive instead of a bare one (although hoeven sometimes takes a bare infinitive as well, see Haeseryn et al. 1997: 970973). It should also be noted that the use of (niet) hoeven (expressing lack of necessity) is typical of Northern Dutch, whereas in Southern Dutch (Flemish), moeten in combination with the negation marker niet is more frequently used instead of niet hoeven (Diepeveen et al. 2006: 14ff.).

Modals in the Germanic languages

(4)

Dat hoef je that need.PRS you You neednt do that.

niet not

te doen. to do.INF

The verb durven, meaning to dare, to have the courage, has left the group of modal verbs (unlike its cognate in German and Danish, for instance, where drfen viz. turde expresses permission and probability). Note, however, that Dutch durven has developed an auxilary use, typically expressing a frequentative meaning. This use is particularly common in Southern Dutch (Ronny Boogaart, p.c.). Typically, the verb combines with a te-infinitive, but uses without te are attested as well (Haeseryn et al. 1997: 1010). (5) Dan durft het wel eens fout then dare.PRS it well once wrong Then it sometimes goes wrong. (te) lopen. (to) go.INF

In contrast to the English modals, the Dutch modals have a full present and past tense paradigm, and they allow non-finite forms (infinitive, past participle (except for zullen), and present participles for kunnen and willen (kunnend, willend)). Imperatives and passives, however, are generally ruled out (Haeseryn et al. 1997: 94100), although an imperative for willen is mentioned (wil-wilt; (Haeseryn et al 1997: 98). More details will be given in section 2.1. (6) a. Hij/Zij moet/moeten het doen. he/they must.PRS.SG/PL it do.INF He/they must do it. b. Hij moest het doen. he must.PST it do.INF He had to do it. c. Hi heeft het moeten doen. He have.PRS it must.INF do.INF Hes had to do it. d. Hij heeft het gemoeten/gekund/gemogen/gewild. he have.PRS it must.PP/can.PP/may.PP/will.PP Hes had to/hes been able to/he has been allowed to/he has wanted it.

One of the striking features of the Dutch modals is the fact that they express a wide range of non-modal meanings. These can be found not only for willen to want and kunnen to know, the cognates of which in other languages also often have premodal meanings, but also for mogen and moeten, both of which express (dis)like, sympathy and antipathy (Barbiers 2002). (7) a. Hij moet haar niet. he must.PRS her not He doesnt like her. b. Ik mag geen soep.12 I may.PRS no soup I dont like soup. c. Ik mag hem wel. I may.PRS him well I like him.

Moreover, in some (southern) varieties of Dutch, moeten can also mean to owe (Diepeveen et al 2006: 100f.).

Tanja Mortelmans, Kasper Boye & Johan van der Auwera

(8)

Hoeveel moet ik how much must.PRS I How much do I owe you?

u? you

In combination with a nominal object designating a language, the verb kunnen can occur with the meaning to know.13 (9) Hij kan Frans. he can.PRS French He knows French.

The verb kunnen in combination with the preposition tegen can also express (physical or psychological) aversion. (10) Ik kan niet tegen I can.PRS not against I cant stand cats. katten. cats

As far as epistemic meanings are concerned, all core modals are claimed to have developed them (Haeseryn et al. 1997: 984). It should be noted, however, that a number of these epistemic meanings are not as strongly developed as in English. Nuyts (2001: 187; cp. also van der Auwera 1999: 60), for instance, claims that mogen has lost its epistemic meaning in present-day Dutch and that the epistemic use of kunnen is highly instable (most cases of epistemic modality are in fact vague between dynamic and epistemic modality). Moreover, Nuyts doesnt mention an epistemic meaning for Dutch willen. The only two central modals that in Nuytss view can express epistemic or evidential modality are moeten (11a)) and zullen (whose past tense has reportative meaning, see (11c)). (11) a. Ze moet oponthoud gehad hebben. she must.PRS delay have.PP have.INF She must have been delayed. (otherwise she wouldnt be an hour late) b. Ze zal wel oponthoud gehad hebben. she shall.PRS well delay have.PP have.INF She will have been delayed.(I gather) c. Ze zou oponthoud gehad hebben. she shall.PST delay have.PP have.INF Shes said to have been delayed. In the following subsections, we will apply the parameters of paradigmaticity (2.1.), paradigmatic variability (2.2) and structural scope (2.3) to the Dutch modals.

2.1.

Paradigmaticity

On the whole, i.e. irrespective of whether the Dutch modals invite an epistemic or a nonepistemic reading, the degree of paradigmaticity has increased over time (see Birkmann 1987: 375). Important elements in this respect are the size of the paradigm, on the one hand, and its formal homogeneity, on the other. Like English, Dutch has a relatively small paradigm of core modals, whose members (which can be characterized as preterite-presents) can all take bare infinitives. Still, one could claim that paradigmaticity (in particular: the formal similarity between the members of the modal paradigm) is much more outspoken in English than in Dutch. In English, the individual members of the modal paradigm have fewer forms at their disposal (in the present tense there is only one form (can, no *cans), and formal variation is restricted to vowel change and addition of a dental suffix in the case of the past forms of the

Modals in the Germanic languages

modals (will-would, can-could). For must, there is no alternation between a present and a past form anymore. (12) He said he must / had to go back. In Dutch, however, there is clearly more formal variation. A number of the Dutch modals (kunnen, mogen, zullen) show vowel alternation between singular and plural (e.g. ik kan wij kunnen), which is one of the typical characteristics of the preterite-presents (Birkmann 1987: 55).
Table 3. Formal variation in the present-tense paradigm of the Dutch modals kunnen present kan kan / kunt kan kunnen willen present wil wil / wilt wil willen mogen present mag mag mag mogen zullen present zal zal / zult zal zullen moeten present moet moet moet moeten

ik (1st sg) jij (2nd sg) hij/zij (3rd sg) wij/jullie/zij (pl)

Moreover, in the singular paradigm of kunnen, willen and zullen, the first and third persons are identical, whereas the second person has two options at its disposal.14 Note that the use of the forms without -t increases the homogeneity of the singular paradigm, theres only one remaining singular form. This points to an ongoing tendency towards simplification or regularization of the present tense paradigm. Moreover, the loss of the forms with -t makes the modal verbs different from ordinary lexical verbs, which have -t in both the second and the third person singular (jij drinkt, hij drinkt vs. jij kan, hij kan). As far as the past tense is concerned, we observe that the modal willen has two past tense forms at its disposal: wilde and wou, with the latter being the more recent one (see table 4). The other Dutch modals only have one preterite form (moest, kon, zou, mocht).
Table 4. Formal variation in the past tense paradigm of Dutch willen willen past wilde (more frequent)/ wou15 wilde / wou wilde /wou wilden / (substandard) wou(d)en

ik (1st sg) jij (2nd sg) hij/zij (3rd sg) wij/zij/jullie (pl)

Another factor, this time contributing to the modals paradigmaticity, is the fact that in combination with an infinitive, the expected past participle of the modal (in perfect and pluperfect constructions) is replaced or substitued by the infinitive form (the so-called infinitivus pro participio or IPP-construction16). The use of the past participle (gekund, gemoeten, etc.) is therefore restricted to cases in which the modal verb has a non-verbal complement. It should be noted, however, that Dutch IPP-constructions occur with a relatively wide range of verbs, i.e. they are not restricted to the modal verbs, but also occur with inchoatives and durative verbs, for instance (het is beginnen te regenen, lit. it is start (IPP) to rain (INF), it has started to rain) (see Ponten 1973 for more details). In German, as we will see, there has been a narrowing down of verbs with IPP to the extent that in presentday German, this category is more or less restricted to the modal verbs (together with brauchen) and verbs of perception. As the Dutch modals share the IPP-parameter with a wider range of verbs, this factor contributes less to the Dutch modals paradigmaticity than in German, where IPP is much more restricted and in a way typical of the modals.

Tanja Mortelmans, Kasper Boye & Johan van der Auwera

(13) a. Hij heeft het moeten (*gemoeten) zeggen. he have.PRS it must.INF (*PP) say.INF Hes had to say it. b. Dat heeft hij nooit gemoeten. that have.PRS he never must.PP Hes never had to do this. Another characteristic of Dutch modals is the fact that in compound past tenses, either the auxiliary selected by the modal (hebben) or the auxiliary selected by the infinitive complement (zijn) can be used. The latter case is indicative of higher grammaticalization, as the modal verb loses its typical properties, as it no longer controls the selection of hebben as a perfect auxiliary (for a discussion of the parameters that favour either hebben or zijn, see Hofmans 1980ab). So, modals in combination with infinitives typically taking zijn (like gaan (to go) in example (14a)), can also take zijn (next to hebben) as a perfect auxiliary. Note that this phenomenon is typical of Dutch: German, Danish and Icelandic allow only of a havecognate in the perfect (14b) (but see Cornillie et al. (this volume) for a similar observation pertaining to modal verbs in Italian). (14) a. Hij is naar huis He be.PRS to-home Hes had to go home. Er hat/*ist nach Hause He have.PRS to home Hes had to go home. moeten must.INF gehen go.INF gaan. go.INF mssen. must.INF

b.

Strangely enough, the perfect auxiliary zijn can also be used when the modal (without an infinitive) is combined with a directional phrase (as in example (15) naar de garage to the garage). This is difficult to account for in grammaticalization terms, as one would expect less grammaticalized behaviour in this (clearly non-epistemic, dynamic) environment. (15) Onze auto is/ heeft vaak naar de garage Our car be.PRS/ have.PRS often to the garage Our car has had to be taken to the garage often. gemoeten. must.PP

2.2.

Paradigmatic variability

Paradigmatic variability concerns the freedom with which the language user chooses a sign. It is high when the speakers freedom is hardly restricted, so that she can pick out forms of the paradigm according to her own communicative needs; paradigmatic variability is restricted to the extent that the speakers options are narrowed down. As far as this parameter is concerned, Dutch seems to make a clear distinction between non-epistemic and epistemic modality, to the extent that non-finites are (more or less) excluded in epistemic uses. This affects the degree of intraparadigmatic variability, i.e. the selection of alternatives which are in opposition (Lehmann 2002: 138): with epistemic uses, only finite forms can be chosen. Lets have a look at the following examples. In the IPP-constructions (16a) and (16b), the infinitives kunnen and moeten, respectively, only have a non-epistemic reading. The same holds for the past participles gemoeten and gekund in (16c), which get a deontic reading. (16) a. Hij heeft het kunnen zeggen. he have.PRS it can.INF say.INF He has been able to say it.

Modals in the Germanic languages

b. Ze heeft het moeten lezen in de krant. she have.PRS it must.INF read.INF in the newspaper She had to read it in the newspaper. c. Dit had beter gekund en gemoeten. this have.PST better can.PP and must.PP It could and should have been better. One important counterexample has to be mentioned, however. In combination with zou (formally a past tense of zullen), a formal infinitive kunnen appears that can get a conditional epistemic reading, as in (17a) (see Van Ostaeyen and Nuyts 2004: 108). Moreover, the complement infinitive after kunnen can be omitted with zou so that we get an epistemic meaning expressed by a nonfinite form of a modal (kunnen) and complemented by a that-clause (17b). It should be stressed that this seems to be possible only with kunnen. (17) a. Het zou kunnen zijn it shall.PST can.INF be.INF Jan could be at home. b. Het zou kunnen dat it shall.PST can.INF that Jan could be at home. dat that Jan Jan thuis is. at home be.PRS

Jan Jan

thuis is. at home be.PRS

2.3.

Structural scope

As far as structural scope is concerned (involving the loss of syntagmatic weight), Dutch modals show an increased preference for infinitival complements in the epistemic realm. In their dynamic and deontic uses, however, they often occur without infinitive complements. The Dutch core modals have a clear liking for intransitive uses and allow for a wide range of non-infinitive arguments, which clearly distinguishes them from German and Danish modals. In Dutch, non-epistemic modals are still compatible with directional prepositional phrases (18a) and adverbs (18b), which used to be the case for English modals up to the Early Modern English period as well (see Plank 1984: 325). Moreover, the modal verb willen (in its nonepistemic use) is compatible with a that-clause (in object-function). (18) a. Hij mag/kan/moet/wil naar huis. he may.PRS/can.PRS/must.PRS/will.PRS to home He may/can/must/wants to go home. b. Alles moet (mag, kan) weg. everything must (may, can) away Everything must/may/can be sold. c. Ze wil dat hij naar huis komt. she will.PRS that he to home come.PRS She wants him to come home. Dutch non-epistemic modals can also be combined with past participles, expressing passive meanings. Note that the passive auxiliary (worden) can be added, without difference in meaning. (19) a. Dat moet gezegd! that must.PRS say.PP That must be said! b. De hele gang moest geschilderd. the entire corridor must.PST paint.PP The entire corridor had to be painted.

10 Tanja Mortelmans, Kasper Boye & Johan van der Auwera

Dutch modals also allow for nominal objects and adjectives. As shown in Barbiers (2002), combinations with adjectives are only possible when the adjective denotes a value on a bounded scale, i.e. a value that is either zero or one.17 This is the reason why a Dutch modal like moeten can be combined with adjectives like vol (full), leeg (empty) and dood (dead), but not with adjectives like lang (long) or oud (old). (20) a. Hij moet en zal een snoepje. he must.PRS and shall.PRS a candy He insists on having a candy. b. Deze fles moet vol/leeg. this bottle must.PRS full/empty This bottle must be filled/be emptied. c. Moet dat nu echt kapot? must.PRS that now really broke Do you really want to destroy it? Finally, the Dutch modals can also occur in intransitive structures.18 Note that these sentences are not really felt to be elliptical, as it is often very difficult to decide what has been left out. (21) a. Moet dat nu echt? must.PRS that now really Is this really necessary? b. Kan dat? can.PRS it Can it be? Van Ostaeyen and Nuyts (2004: 4346) observe that for kunnen, the frequency of such mainverb or main-verb like uses has increased in present-day Dutch. In early modern Dutch, these uses were rare. Van Ostaeyen and Nuyts (2004: 46) hint at the idea that a process of formal degrammaticalization (at least as far as this particular parameter is concerned) is going on here. It should also be stressed that in this main-verb like use, the meaning of the verb is not at all restricted to dynamic modality; in example (22), the meaning of kunnen is clearly deontic. (22) [] wat u doet kan helemaal niet, een klooster bouwen op het grootste Joodse kerkhof ter wereld. (Van Ostaeyen and Nuyts 2004: 45) what youre doing is not done (lit. can not i.e. is not allowed), building a monastery at the spot of the biggest Jewish cemetry in the world.

The fact that Dutch allows this main-verb like deontic use is also in line with the observation that in Dutch, the modal source (i.e. the entity in which the modal relation originates) can appear as an onstage participant introduced by the preposition van. (23) Hij mag/moet dat niet van mij. He may/must that not from me He is not allowed to do that/he neednt do that, as far as Im concerned. Note also that the epistemic use of zou kunnen (without zijn) as in example (17b) above seems to suggest that intransitive uses of the modals can even express epistemic modality. 2.4. Summary: the Dutch modals

The Dutch modals do not seem to score particularly high with respect to the three grammaticalization parameters. As far as paradigmaticity is concerned, we have observed that

Modals in the Germanic languages 11

there is considerable variation within the paradigm, i.e. the Dutch modals do not behave in a particularly homogeneous way. There are double forms in the present paradigm of zullen, willen and kunnen; three modals show up vowel alternation between singular and plural (kunnen, mogen, zullen), two others dont (moeten, willen). The past tense is built irregularly as well (especially past tense moest stands out) and has double forms in the case of willen (wilde/wou). The past participle either ends in a dental suffix (gekund, gewild) or in en (gemoeten, gemogen). Paradigmaticity seems to increase for the epistemic modals, as only three modals seem to have developed a well-established epistemic reading (moeten, zullen and to a lesser extent kunnen). Moreover, participles of the modals are not compatible with an epistemic reading, which increases paradigmaticity as well. Typical of the Dutch modals is the fact that the perfect auxiliary zijn can occur next to hebben in composed past tenses, i.e. zijn can be selected by the infinitive complement of the modal.
Table 5. Dutch modals: paradigmaticity paradigmaticity [+] Class of central modals consists of only five members. Dutch epistemic modals build an even smaller paradigm (no epistemic meanings for mogen and willen). IPP-constructions restrict the use of the past participle. Perfect auxiliary zijn (next to hebben) in perfect constructions with modal verb and infinitive taking zijn.

[-]

For each modal, there are present, past, infinitive and participle (both present and past) forms (only zullen lacks participle forms). Irregular formation of the various forms (past tense, past participle) decreases formal homogeneity of the paradigm. Formal variation (double forms) in the present/past tense paradigm of willen, zullen and kunnen indicates lack of formal stability.

The distinction between non-epistemic and epistemic meanings is particularly relevant if we consider the parameters of paradigmatic variability and structural scope. Both are on the decrease when the modals are used epistemically, as the epistemic use tends to rule out nonfinite forms of the modals (with some exceptions) and epistemic modals normally require an infinitive complement, thus restricting their structural scope. It should be stressed that nonepistemic uses of the Dutch modals are compatible with a particularly wide range of arguments, i.e. the Dutch modals generally possess a high degree of structural scope, which distinguishes them as we will see from the modals in German and Danish. Dutch modals are compatible with past participles, adjectives, noun phrases, complement clauses and pronouns.
Table 6. Dutch modals: paradigmatic variability and structural scope paradigmatic variability [+] Non-epistemic modals occur in finite and nonfinite forms (participles, infinitives). [-] Epistemic modals favour finite forms (with some exceptions).

structural scope Non-epistemic modals occur in intransitive structures and can be combined with nouns, prepositional phrases, adjectives, participles and infinitives. Their structural scope is rather wide. Epistemic modals clearly prefer infinitive complements. Their structural scope is rather narrow.

In some cases, there seems to be an imperfect match between epistemic meanings and the expected increased degree of grammaticalization, on the one hand, and non-epistemic meanings and the expected lesser degree of grammaticalization, on the other. We find epistemic meanings with intransitive kunnen, as in (24), repeated from (17b).

12 Tanja Mortelmans, Kasper Boye & Johan van der Auwera

(24) Het zou kunnen dat it shall.PST can.INF that Jan could be at home.

Jan Jan

thuis is. at home be.PRS

Conversely, main verb uses allow for zijn as a perfect auxiliary (as if an ellipted infinitive (verb of movement) governed the selection of the auxiliary), as in (25). (25) Hij is naar het ziekenhuis he be.PRS to the hospital He had to go to the hospital. gemoeten. must.PP

From an intralinguistic perspective, moeten and zullen seem to be the most strongly grammaticalized modals in Dutch, whereas willen (and to a somewhat lesser extent) kunnen are at the other end of the grammaticalization spectrum, i.e. they behave still more strongly like full verbs. The latter verbs dispose of an almost complete conjugation paradigm (with present participles and even imperative forms for willen), show up formal instability (range of alternatives) and allow for the widest range of complementation patterns.

3. The German modals The paradigm of German modals consists of six preterite-present19 members: knnen, sollen, mssen, mgen, drfen and wollen. They can occur in the present and past indicative, in the present and past conjunctive (labelled Konjunktiv I and Konjunktiv II, respectively), and in perfect tense forms (in which case the so-called Ersatzinfinitiv is used; see also section 2.1 on IPP-constructions in Dutch). (26) a. Er muss/musste/msste/msse he must.PRS.IND/PST.IND/PST.CONJ/PRS.CONJ b. Er hat nach Hause fahren mssen. he have.PRS to-home drive must.INF Hes had to drive home. c. Er hat nach Hause gemusst. he have.PRS to-home must.PP He had to go home. nach Hause. to home

Strictly speaking, the core modals paradigm is more elaborate than in the case of the Dutch modals, which have no synthetic conjunctive forms. This, however, seems to be due to the general morphological complexity of the German verb system, in which the category of mood is still productive. Formally and functionally, the negative polarity verb brauchen is attracted towards the category of German core modals (Duden 2005: 465, Askedal 1998). Note that the verbs does not belong to the group of preterite-presents and is normally combined with a to-infinitive. First, its past conjunctive often gets Umlaut (bruchte instead of the regular form brauchte, which is analogous to the past conjunctive forms msste, knnte, mchte and drfte); second, a present tense form brauch (without t, thus resembling the other t-less forms muss, kann, darf etc.) is used both in the first and third person singular. Note that this form is not accepted as standard German. A third indication of brauchens integration in the modal verb paradigm is the loss of the infinitival marker zu, illustrated in (27):

(27)

Er braucht es he need.PRS it He neednt do it.

nicht (zu) not (to)

tun. do.INF

Modals in the Germanic languages 13

Another verb whose membership in the category of German modals is somewhat unclear, is the future auxiliary werden. In contrast to English (will/shall), Dutch (zullen) or Danish (skulle), none of the German core modals expresses futurity. The verb werden has stepped in here. It does not belong to the group of preterite-presents, but patterns with a bare infinitive in present-day German. Originally meaning something like turn, werden develops inchoative uses (mainly in combination with a present participle) in Old High German. The grammaticalization of werden with a bare infinitive into a future auxiliary takes place in the Middle High German period, in which it came to replace earlier future uses expressed by the predecessors of the modals sollen, wollen and mssen (see for an account on the history of werden, Kotin 2003: 151163 and 166175). Apart from syntactic arguments, werdens inclusion in the group of German modals can also be argued for with functional arguments, as it has developed an epistemic use as well, as in (28) for a discussion of the grammaticalisation status of werden in present-day German, see e.g. Mortelmans (2004). (28) Er wird jetzt wohl da sein. he become.PRS now well there be.INF He will be there by now. As far as the semantics of the German modals is concerned, it can be noted that the modals express a wide range of meanings, from lexical to dynamic, deontic and epistemic ones. The German modals mgen (both indicative mag like, want and past conjunctive mchte would like to have), wollen want and knnen know express non-modal meanings: (29) a. Ich mag keine Suppe. I may.PRS no soup I dont like soup. b. Ich mchte ein Kind. I may.CONJ.PST a child I would like to have a child. c. Ich will ein Kind. I will.PRS a child I want a child. d. Er kann das Gedicht auswendig. he can.PRS the poem by heart He knows the poem by heart. No other lexical meanings can be discerned for the German modals, though, which means that their range of non-modal meanings is a bit lower than in Dutch, where moeten can both mean like (see example (30a) repeated from (7a) above) and owe (see example (30b) repeated from (8) above). Note that the German counterpart of Dutch moeten meaning to owe is the verb sollen rather than mssen. This sollen is found in main verb like use until the 18th century (Diewald 1999: 322), as exemplified in example (30c). In present-day German, however, the infinitive seems to be obligatory. (30) a. Ik moet haar niet. I must.PRS her not I dont like her. b. Hoeveel moet ik je? how much must.PRS I you What do I owe you? c. Was soll ich Ihnen fr den Tee? (Goethe) what shall.PRS I you for the tea What do I owe you for the tee?

14 Tanja Mortelmans, Kasper Boye & Johan van der Auwera

Every German modal has developed an epistemic meaning; in the case of sollen and wollen, this is of the evidential kind, expressing reportative modality (comparable to zou in Dutch, skulle in Danish and munu in Icelandic). The frequency of these epistemic meanings, however, is rather low. Diewalds admittedly rather restricted corpus of present day German material (Diewald 1999: 217) yields frequencies between 15.8 % (for drfte) and 0.0 % (for reportative wollen). This might be indicative of the fact that the epistemic variant is not as well-entrenched in German as it is in English (where the epistemic use of the modals is much more frequent, compare e.g. Abraham 2002: 8). Due to a lack of comparable analyses, it is rather difficult to judge German and Dutch in this respect.

3.1.

Paradigmaticity

The German core modals built a relatively concise paradigm with a rather strong formal homogeneity. All the members are preterite-presents (or, in the case of wollen, are modelled on them); five out of six verbs show the original vowel alternation between singular and plural in the present tense paradigm. Only sollen has <o> everywhere. No double forms as in Dutch seem to exist. The past tense is built by means of the dental suffix te, which is attached to the stem. The past particple ends in t.
Table 7. Present, past and past participle forms of the central modals in German present tense ich, er muss ich, er kann ich, er will ich, er mag ich, er darf ich, er soll wir mssen wir knnen wir wollen wir mgen wir drfen wir sollen past tense musste konnte wollte mochte durfte sollte past participle gemusst gekonnt gewollt gemocht gedurft gesollt

Unlike Dutch, no tendencies towards further simplification or regularization of the paradigm can be observed; the present-day German paradigm seems to be more stable than the Dutch one. With some modals, epistemic readings are only compatible with particular forms; this holds for past conjunctive drfte and present indicative mag (this formal restriction could also be considered as an example of the decrease of paradigmatic variability, see section 3.2.). (31) a. Sie drfte (*darf) nicht mehr ganz nchtern she may.PST.CONJ not anymore completely sober Shes probably not completely sober anymore. b. Mag sein, dass das in der Schweiz wohl may.PRS be.INF that that in Switzerland indeed Its very well possible that this functions in Switzerland. sein. be.INF funktioniert. function.PRS

It has already been noted that both German and Dutch have the IPP-construction. In German, however, IPP mainly pops up with the modals and brauchen need, with the causative verb lassen make, and with the perception verbs sehen see, hren hear, fhlen feel and spren feel (Duden 2005: 473). The latter verbs of perception can take a past participle as well. In Dutch, the category of verbs taking IPP is much larger, so that the IPP-parameter adds to the paradigmaticity of the modals in German to a larger extent than in Dutch, where IPP does not single out the central modals.

3.2.

Paradigmatic variability

Modals in the Germanic languages 15

As was the case for Dutch, epistemic uses of the modals have to be finite, i.e. EMV [epistemic modal verbs] cannot occur in non-finite environments i.e. no epistemic reading emerges in embedded constructions (Abraham 2001: 11). So, the perfect tense of mssen in the following example can only express dynamic modality and does not invite an epistemic interpretation. (32) Er hat das Buch bis Dienstag lesen mssen. he have.PRS the book till Tuesday read.INF must.INF He had to read the book by Tuesday. (example taken from Diewald 1999) In Dutch, zou (past tense form of zullen) can be combined with an infinitive of a modal to express epistemic meaning. The combination of German wrde with the infinitive of a modal (*? wrde sein knnen in 33b), however, is awkward, especially when the modal has epistemic meaning. (33) a. Het zou kunnen it shall.PST can.INF b. Es knnte sein, it can.PST.CONJ be.INF It could be that hell come. dat that dass that hij he er he komt. come.PRS kommt. (*Es wrde sein knnen, ) come.PRS

However, some authors state that there are exceptions to this rule. Reis (2001: 295), for instance, claims that non-finite forms do allow an epistemic interpretation in irrealis conditional constructions and in attributive infinitival complements like the ones in (34); according to Reis, therefore, the non-finite gap is not absolute. For Reis, semantic factors are responsible for the fact that non-finites are generally not compatible with epistemic meaning, but the general incompability is not a grammatical characteristic of the German epistemic modals as such. (34) a. Nach allem, was ich wei, htte er dann zu Hause sein mssen. after everything that I know have.PST.CONJ he then home be must.INF On the basis of what I know, he should have been home by then. b. Nach allem, was ich wei, htte er da noch in Prag sein knnen. after everything that I know have.PST.CONJ he then still in Prague be can On the basis of what I know, he could have been in Prague at the time. c. Der Gedanke, sich tuschen zu mssen, drngte sich auf. The thought be mistaken.INF to.must.INF forced itself on The idea that he was wrong forced itself upon him

3.3.

Structural scope

In their non-epistemic use, the structural scope of the German modals is considerably higher than in English, as the German modals (just like the Dutch ones) can still be combined with directional prepositional phrases20 and adverbs. (35) a. Er kann nach Hause. he can.PRS to-home He can go home. b. Alles muss weg. everything must.PRS gone Everything must go.

16 Tanja Mortelmans, Kasper Boye & Johan van der Auwera

Still, there has been considerably more condensation in German than in Dutch, as a number of structures without a complement infinitive are ruled out in German. Put differently, the complexity of constituents a German modal can combine with is considerably lower than in Dutch, so that its structural scope is smaller than the structural scope of a corresponding Dutch modal. Intransitive-uses (i.e. without an infinitive, see (36ab), uses with a past participle (36c) and uses with adjectival complements (36d) are clearly out for the German modals. (36) a. ?* Darf (kann) das? (But: Darf (kann) das sein?) Mag (kan) dat? may (can) that Is it permitted/possible? b. ?* Muss das nun wirklich? (But: Muss das nun wirklich sein?) Moet dat nu echt? must it now really Is this really necessary? c. * Es muss gesagt, dass (but: Es muss gesagt werden, dass...) Het moet gezegd, dat it must said that It must be said. d. * Der Stuhl darf wei. De stoel mag wit. the chair may white The chair may be painted white. German modals can only be combined with nominal objects in the case of knnen, wollen and mgen (see examples under (29) above), and with complement clauses introduced by dass in the case of wollen and mgen (see example (37a), taken from Diewald 1999: 55). Contrary to Dutch, uses of mssen and sollen with an NP (as in (37b)) are generally impossible. (37) a. Er mag, dass du ihm he may.PRS that you him He wants you to help him. b. * Er muss und soll he must.PRS and shall.PRS He insists on having a candy. hilfst. help.PRS einen Bonbon. a candy

The German modals can be combined with pronominal objects like das and es, as in the following example (see for a discussion also hlschlger 1989: 6568). (38) Er kann es, er muss es, er he can.PRS it, he must.PRS it he Hes able to do it, he must do it, he may do it. darf may.PRS es. it

3.4.

Summary: the German modals

It seems that the German modals score higher with respect to the grammaticalization parameters than the Dutch ones, especially with regard to the parameter of structural scope. The German modals more often take an infinitive also in their non-epistemic uses than the Dutch ones (compare German das kann sein with Dutch dat kan). Uses in which the modal is combined with a past participle (*das muss gesagt) or an adjective (*er muss tot) are ruled out in German, and German modals resist the combination with nominal objects to a greater extent than Dutch ones (German *er muss einen Bonbon vs. Dutch hij moet een snoepje).

Modals in the Germanic languages 17

Table 8. German modals: structural scope structural scope [+] [-] German modals can be combined with noun German modals show a strong tendency phrases and prepositional phrases in non- towards infinitival complements; combinations epistemic use only. with past participles and adjectives are ruled out, intransitive structures do not seem to occur either. German epistemic modals share an outspoken preference for infinitive complements.

As far as paradigmaticity is concerned, the German paradigm is more complex than the Dutch one (allowing for synthetic conjunctive forms, for instance), on the one hand, and formally more stable and regular than the Dutch one, on the other. These two observations point into different directions as far as grammaticalization is concerned: for German, the higher complexity (i.e. the wide array of formal realizations of a particular modal like muss, msste, msse, mssen, gemusst etc.) could be interpreted as signalling low paradigmaticity (which would be symptomatic of low grammaticalization), whereas the lack of formal stability in Dutch leads to an increase of formal variants (e.g. wou/wilde), hence to a more complex paradigm, which again can be taken to signal low grammaticalization. As the formal complexity of the German modal paradigm is completely in line with the greater inflectional possibilities of the German verb in general, we are inclined to think that it is the second element, i.e. the Dutch lack of formal stability and internal regularity in contrast to the stability and regularity of the German modal paradigm, which should be stressed as more significant in the comparison of both languages.
Table 9. German modals: paradigmaticity paradigmaticity [-] [+] German modals build a relatively small paradigm. The German paradigm is homogeneous and fairly regular (past tense = stem + -te, all past participles end in t). There is no formal variation (double forms) in German. IPP-constructions are obligatory. The category of verbs that obligatorily take IPP is restricted to the modal verbs (with brauchen) and heien, lassen and sehen.

The paradigmatic variability of the German modals clearly decreases when they are used epistemically: with only a few exceptions (see Reis 2001), the epistemic use is restricted to finite forms of the modals.

18 Tanja Mortelmans, Kasper Boye & Johan van der Auwera

Table 10. German modals: paradigmatic variability paradigmatic variability [+] [-] In German (as in Dutch), non-epistemic modals In German (as in Dutch) epistemic modals occur in finite and non-finite forms (participles, favour finite forms of the modals (with a few infinitives). exceptions).

From a semantic point of view, all of the German modals have developed an epistemic or evidential reading. In contrast, the existence of strongly entrenched epistemic uses for every modal in Dutch can be doubted (see Nuyts 2001).

4. The Danish modals The Danish paradigm of modal preterite-presents consists of six members: ville will (cf. however footnote 5), skulle must, mtte21 may/must, kunne can, burde22 ought to and turde dare.23 All members can be constructed with a bare infinitive. (39) Hun vil/skal/m/kan/br/tr svmme. she will/must/may/can/ought.to/dare.to.PRS swim.INF She wants to/must/may/can/ought to/dares to swim.

One member, turde, can also be constructed with an infinitive marked by at (to). (40) Han tr ikke (at) he dare.PRS not (to) He does not dare to swim. svmme. swim.INF

In this respect, turde is similar to the verbs behve need and gide feel like, be bothered (to) that are normally included among the Danish modals, although they are not preteritepresents (e.g. Skyum-Nielsen 1971, Brandt 1999, Boye 2001, Jensen 2005). (41) a. Butleren behver ikke (at) vre morderen. butler.DEF need.PRS not (to) be.INF killer.DEF The butler doesnt have to be the murderer. b. Jeg gider ikke (at) lse den bog. I is.bothered.PRS not (to) read.INF that book I cant be bothered to read that book.

Among the six modal preterite-presents, kunne, skulle, ville and mtte may be taken to constitute a core paradigm. First, they are the only modals that cannot be represented by gre (do) (see Brandt 1999: 17). Compare: (42) a. Jeg tr /behver ikke se den film gr I dare/need.PRS not watch.INF that movie do I dare/need not watch that movie do you? b. Jeg kan/vil/skal/m ikke se den film *gr I can/will/must/may not watch that movie do I cannot/will not/must not/may not watch that movie do you? du? you du? you

Second, kunne, skulle, ville and mtte are by far the most frequent of the modal verbs (Brandt 1999: 27).

Modals in the Germanic languages 19

Some of the modals have additional non-modal meanings. Among the preterite-presents, kunne like its German cognate can still mean know how to, as in (43). (43) Han kan en sang. he can.PRS a song He knows a song.

The modals kunne, burde and skulle are all regularly used epistemically. So is mtte with necessity meaning, in contrast to mtte with possibility meaning (e.g. Boye 2001: 5154). Note that epistemic skulle (44b) has an evidential interpretation, just like its cognates in German (sollen) and Dutch (zou). (44) a. Butleren kan/ br/m vre morderen. butler.DEF can/ought/must.PRS be.INF killer.DEF The butler may/should/must be the murderer. b. Butleren skal vre morderen. butler.DEF must/shall.PRS be.INF killer.DEF The butler is the murderer, it is said. The modal turde also seems to have epistemic uses. However, these are found mainly in idiomatized expressions such as (45). (45) Det tr vre overfldigt it dare.PRS be.INF superfluous It should be unnecessary to mention that at nvne to mention.INF at. that

As for the modal ville, it is disputed whether it has an epistemic variant or not (e.g. DavidsenNielsen 1990: 161163, Brandt 1999: 6062, Boye 2001: 5864). In any case, epistemic uses of ville as in (46) would be very rare. (46) Hun vil vre i Danmark she will.PRS be.INF in Denmark She will be in Denmark right now. lige nu. right now

As has already been noted, mtte can be used epistemically only with necessity meaning, not with possibility meaning. As it is the case with other Germanic modals, the past tense forms of the Danish modals are regularly used as past-tense hypotheticals rather than with temporal meaning. So, in (47), past tense kunne weakens the strength of the epistemic judgement. (47) Butleren kunne vre morderen. butler.DEF can.PST be killer.DEF The butler could be the murderer.

4.1.

Paradigmaticity

The most salient feature of the Danish modals (preterite-presents as well as non-preteritepresents) is the ability to co-occur with a bare infinitive. This feature is what defines the paradigm of modals, and it has defined the paradigm all the way back to Common Nordic (Old Norse). For some individual modals, however, a diachronic change in paradigmaticity can be registered, which manifests itself in the gradual loss and acquisition, respectively, of the infinitival marker at (to) before the infinitives accompanying the modals. Thus, while it seems that turde is currently moving out of the paradigm (i.e. it seems that it is gradually acquiring at before its accompanying infinitive (see example (48)), behve seems to be

20 Tanja Mortelmans, Kasper Boye & Johan van der Auwera

entering the paradigm (i.e. it seems that behve is losing at before its accompanying infinitive (see example (49)) (Hansen 1977). This mirrors the behaviour of brauchen in present-day German, which is said to increasingly lose the infinitival marker zu (see section 3.1.). (48) a. Han tr ikke danse. (old construction) he dares not dance.INF b. Han tr ikke at danse (new construction) he dares not to dance.INF He doesnt dare to dance. (49) a. Han behver ikke danse. (new construction) he needs not dance b. Han behver ikke at danse.(old construction) he needs not to dance He doesnt need to dance. While the modal paradigm can be defined in terms of the infinitival marker, the Danish modals show a few other peculiarities as well. Notably, the inflection of Danish modals is reduced in comparison to non-modal verbs such as e.g. svmme (swim). In the words of Brandt (1999: 69), modals practically only occur in four forms (here illustrated with kunne): infinitive (kunne), present (kan), past (kunne), and past participle (kunnet) (in which case they always select have (have) rather than vre (be) as an auxiliary; cf. section 2.1. on Dutch). Among the other verb forms found in Danish, present participles (kunnende) and passives (kunnes) of some of the modals24 can be found, but all modals lack an imperative form. In fact, it might be argued that the inventory of modal forms is even smaller. First, the infinitives of preterite-present modals are formally identical to the pasts, both in spoken and written language. Second, for some speakers, the past participles of preterite-present modals are formally identical to the infinitives (cf. section 2.1. on the infinitivus pro participio or IPPconstruction) and thus the pasts. This latter convergence of forms can be registered even in the written language. Thus, rather than the correct (50), (51) is frequently encountered.25 (50) Han har kunnet svmme. he has can.PP swim He has been able to swim. Han har kunne svmme. he has can.INF swim He has been able to swim.

(51)

This means that morphologically the preterite-present modals might be taken to move diachronically towards having only two regularly used forms: present (kan/m/skal/vil/tr/br) and non-present/non-finite (kunne/mtte/skulle/ville/turde/burde).

4.2.

Paradigmatic variability

While all the Danish modals co-occur with non-modal verbs (in the form of bare infinitives) in the same clause, the use of non-modal verbs is not bound up with a systematic choice between modals. Thus, Danish modals just like the Dutch and German ones are not obligatory in any useful sense of the term. As mentioned in the previous section, Danish modals do not have as many forms at their disposal as regular main verbs. They show a defective conjugation pattern. For epistemic modals, this pattern is even more defective than for non-epistemic modals. Notably, epistemic modals lack participles (Brandt 1999: 96), and - as was observed for Dutch and German -

Modals in the Germanic languages 21

infinitive modals can only rarely be used epistemically. The following example, however, would be a case in point. (52) Han synes at mtte vre morderen. he seems to must.INF be murderer.DEF It seems he must be the murderer.

Accordingly, a modal infinitive may occur with an infinitival periphrastic blive-passive in its scope. Whereas the morphological -s-passive is compatible only with a non-epistemic reading of the modals in the scope of which it occurs (53b), the blive-passive typically goes with an epistemic reading (Lauridsen and Lauridsen 1989: 248). (53) a. Han synes at kunne blive valgt i morgen. (blive-passive) he seems to can.INF be.INF elected tomorrow It seems that he may (epistemically) be elected tomorrow. b. Han synes at kunne vlges i morgen. (-s-passive) he seems to can.INF elect.PASS tomorrow It seems he has the (dynamic or deontic) possibility to be elected tomorrow.

4.3.

Structural scope

In their non-epistemic uses some of the modals do not require an infinitive (see Brandt 1999 and Boye 2001 and 2005 for details). Among the preterite-presents, only kunne readily occurs with a noun phrase. (54) Han kan en sang. he can a song He knows a song. A similar construction is found with ville. With ville, however, the construction has a ceremonious, archaic, or as Brandt (1999: 74) puts it, slogan-like flavour. (55) Vi vil frihed! (Brandt 1999: 74) we will freedom We want freedom!

With a pronoun or a general noun phrase which can be interpreted as having infinitival meaning (Brandt 199: 7071), the construction is possible with other modals as well (cf. Davidsen-Nielsen 1990: 21). (56) a. Hun skal/m/tr ikke noget. she must/may/dares not anything She has no plans./She is not allowed to do anything./She does not dare to do anything. Hun kan/vil/m/skal en masse. she can/will/may/must a lot of things. She can/wants to/may/must do a lot of things.

b.

In their non-epistemic uses, all Danish modals except turde, kunne and mtte with possibility meaning readily co-occur with directional adverbials (but not adjectives) (cf. Boye 2001 and 2005 for a cognitive analysis which captures the distribution of directional adverbials with modals). (57) Hun skal hjem.

22 Tanja Mortelmans, Kasper Boye & Johan van der Auwera

she must.PRS home She has to go home. With ville (will) even finite complement clauses may appear. This behaviour unites ville with its German (wollen) and Dutch (willen) cognates. (58) Kongen vil at den sorte prins forlader king.DEF will.PRS that the black prince leave.PRS The king wants that the black prince leaves the country. landet. country.DEF

With epistemic modals, however, all these constructions are impossible. Epistemic modals always require a bare infinitive. (59) a. Det kan vre hun it can.PRS be she It may be she has gone. Det kan vre hun it can.PRS be she She may not have gone. Hun kan vre get. er get. is gone ikke er get. not is gone.

b.

c.

she
d.

can.PRS

be

gone

She may have gone. Hun kan ikke vre get. she can not be gone She cannot have gone.

Omission of the infinitive is impossible (in contrast to the Dutch construction (60b)): (60) a. *Det kan hun er get. it can she is gone b. Het kan dat ze weg is. it can.PRS that she gone is She may have gone. Thus, while in their non-epistemic uses the Danish modals have a quite wide structural scope, in their epistemic uses their structural scope is narrow. It should be mentioned, finally, that both non-epistemic (61ab) and epistemic (61c) modals may co-occur with a subject only. In that case, however, a complement is represented by ellipsis, or the meaning of a complement is retrievable from the context. (61) a. Hun skal g nu. Nej, hun skal ikke. she must.PRS go now. no she must not She must leave now. No, she mustnt/neednt. b. Jeg kan ikke lngere. I can.PRS no longer I cannot do that (obvious from the context) any more. c. Butleren m vre morderen. M han? butler.DEF must be murderer.DEF must he The butler must be the murderer. Really?

4.4.

The Danish modals: summary

The epistemic and the non-epistemic uses of Danish modal verbs do not show differences with respect to degree of paradigmaticity. However, in their epistemic use the modals have a

Modals in the Germanic languages 23

much more restricted structural scope than in their non-epistemic use. As mentioned, this may be taken to suggest that epistemic modals are more grammaticalized than their non-epistemic counterparts (alternatively, one might speculate that the explanation is semantic, but unrelated to grammaticalization; it is generally accepted that epistemic modals in contrast to nonepistemic ones take a proposition i.e. something with a truth value in their scope). In support of this, two facts may be noted: First, unlike non-epistemic Danish modals, epistemic modals do not occur as participles (Brandt 1999: 96) which affects their paradigmatic variability - and second, unlike infinitives of non-epistemic Danish modals, infinitives of epistemic modals cannot occur as complements of other verbs (Brandt 1999: 136137); however, they can occur in complements of other verbs. The main findings for Danish are summarized in table 11.
Table 11. Danish modals: paradigmaticity, paradigmatic variability and structural scope paradigmaticity [-] [+] The Danish paradigm of core modals is relatively small (only 4 core members: kunne, skulle, ville and mtte). The inventory of inflectional forms each modal has at its disposal is quite restricted as well: the past tense form is identical to the infinitive and, for some speakers, even to the past participle.

paradigmatic variability [+] [-] Non-epistemic modals occur in finite and non In general, Danish modals have less forms at finite forms (participles, infinitives). their disposal than the modals in both German and Non-epistemic modals allow passive forms. Dutch (which hangs together with the strongly reduced inflectional paradigm in Danish) Epistemic modals prefer finite forms of the modals (with some exceptions). Epistemic modals do not allow passive forms. structural scope [+] [-] Structural scope is wide with non-epistemic Epistemic modals always take infinitive modals. complements, the structural scope is therefore Non-epistemic modals combine with nounnarrow. phrases and directional adverbs, and ville even with finite complement clauses. The co-occurrence with directional phrases seems to be somewhat restricted.

5. The Icelandic modals Icelandic has seven modal preterite-presents: vilja will, skulu shall, will, mega may, must, munu will, eiga ought (to), kunna can, may and urfa need (Einarsson 1945: 164167, Thrinsson and Vikner 1995, Kress 1982: 245). In contrast to three non-modal preterite-presents, vita know, unna love and muna remember, the modal preteritepresents all express meanings within the necessity-possibility range, most, if not all, of them have both epistemic and non-epistemic uses (Thrinsson and Vikner 1995), and they lack an imperative (Einarsson 1945: 103). Interestingly, only four preterite-present modals combine with a bare infinitive: vilja, skulu, munu and mega. Among these, mega is unique in being

24 Tanja Mortelmans, Kasper Boye & Johan van der Auwera

capable of expressing both possibility (see 62c) and necessity (in 62d). In this sense, mega is comparable to Danish mtte (see section 4). (62) a. Hann vill lera mlvsindi.26 he will learn.INF linguistics He wants to study linguistics. b. skalt fara. you shall go.INF You have to go. c. Hn m alveg taka minn bl. she may well take.INF my car It is fine that she takes my car. d. Han mtti lesa upp og lra betur. he must.PST read.INF up and learn better He had to read everything once more and learn it better. e. etta mun vera rtt. this shall/will be.INF right This will be right/This is probably right. The three remaining modal preterite-presents, eiga ought (to), kunna can, may, and urfa need, all combine with an a-marked infinitive; they can express both non-epistemic and epistemic meanings. (63) a. a tti it ought.PST.SUBJ. It ought to be enough. b. Han kan ekki he can.PRS not He cannot swim. c. Vi urfum ekki we need.1.PL. not We need not go. a vera to be.INF a synda. to swim.INF a fara. to go.INF ng. enough

As skulu, vilja, mega and munu are the only preterite-present modals that combine with a bare infinitive, they may be considered the central modals of Icelandic. It should be noted that the morphological status of vilja is somewhat unclear in present-day Icelandic. Birkmann (1987: 377) states that vilja is a weak verb with irregularities in the 2nd and 3rd person singular of the present indicative and therefore excludes it from his list of preterite-present modals (which thus contains six instead of seven members). However, vilja is undoubtedly one of the more frequent modals which in combination with the fact that it combines with a bare infinitive warrants its inclusion in this list of central modals. As in Dutch, German and Danish, some of the Icelandic modals have additional nonmodal meanings. Like its Danish cognate kunne, for instance, kunna has the meaning know, know how to besides its modal meaning of possibility. This meaning is expressed when the verb takes nominal complements as in (64ab) note that in (64b) ensku is a feminine accusative noun, rather than an adjective: as an adjective English translates into the adjective enskur. Similar to its German, Dutch and Danish cognates, vilja can also express the nonmodal meaning want (in 64c (provided to us by Johanna Bardal)). (64) a. Hann kann fyrsta boori. he can.PRS first commandment He knows the first commandment. b. Hann kunni ensku. He can.PST English He knew English.

Modals in the Germanic languages 25

c. Hann vill essa bk. He will.PRS this book He wants this book. As for the central modals mega, munu and skulu, they all have clearcut epistemic uses. Notably, the modal mega (unlike its Danish cognate mtte; cf. section 4) can be used epistemically both with its possibility (65a) and its necessity (65b) meaning, and munu seems to be capable of expressing both epistemic necessity (65c) and a distinct evidential meaning of report (65d). (65) a. a m vel vera. . that may.PRS well be.INF That is quite possible. b. a m rigna miki. it must.PRS then rain.INF much Then it must be raining a lot. c. etta mun vera rtt. this shall/will.PRS be.INF right This is probably right. d. a mun hafa veri ms there shall/will.PRS have.INF been.PP mouse There is said to have been a mouse in the bathtub. e. a skal hafa veri ms there shall.PRS have.INF been.PP mouse in I am sure there has been a mouse in the bathtub.

bakerinu. in bathtub.DEF bakerinu. bathtub.DEF

As for the fourth central modal, vilja, Thrinsson and Vikner (1995) seem to think that it has an epistemic use as well, and they give (66) as an example. (66) Mig vill sennilega Me.ACC will.PRS probably I will probably need money. vanta need peninga. money

However, it is not obvious to us that vilja in (66) is indeed epistemic. In fact, (66) is judged odd by Ellert Thor Jhannson (p.c.), and in a sentence like the following (provided by Ellert Thor Jhannson) where vilja cannot be read with future meaning, only a non-epistemic reading is possible: (67) Hann vill vera skrifstofunni sinni nna. he will be.INF in office his now He wants to be in his office now. (* It must be the case that he is in his office now.) Epistemic Icelandic modals differ from non-epistemic modals in some interesting respects. Epistemic modals allow of non-argument subjects (see examples 68ab), and they cannot be pseudo-clefted. According to Thrinsson and Vikner, the former explains why only epistemic modals can occur with quirky (accusative or dative) subjects, as in (68a) where vanta requires an accusative subject (Harald rather than nominative Haraldur) and in (68b) where lika requires a dative subject (Haraldi) (examples are taken from Thrinsson and Vikner 1995: 60). (68) a. Harald vill oft vanta peninga. Harald.ACC will.PRS frequently lack money Harald frequently tends to lack money.

26 Tanja Mortelmans, Kasper Boye & Johan van der Auwera

b. Haraldi tlar a lka vel Stuttgart. Harald.DAT intends to like well in Stuttgart It looks like Harald will like it in Stuttgart. Note that the non-epistemic readings Harald frequently wants... (for 68a) and Harald intends (for 68b) are excluded.

5.1 Paradigmaticity As mentioned above, there are seven modal preterite-present modals in Icelandic: kunna, urfa, munu, skulu, mega, eiga, and vilja. Within this group, a morphosyntactic paradigm of four modals, munu, skulu, mega, and vilja, can be delimited by the fact that these verbs occur with a bare infinitive rather than with a marked infinitive. In fact, one might subsequently distinguish a core paradigm consisting of munu and skulu only. In contrast to the rest of the modal preterite-presents these two modals lack preterite indicatives27 and present as well as past participles. Moreover, in contrast to the other modal preterite-presents, as well as in contrast to all other Icelandic verbs, they have infinitives ending in -u, which are in fact preterite infinitives (a preterite form of the infinitive vilja (vilju) is rarely found as well) (Einarsson 1945: 95, 103, 160, Hammerich 1960: 53). Apart from these properties, the Icelandic central modals are similar to other regular verbs: They are conjugated for number and person, as well as for tense (present vs. past) and mood (indicative vs. subjunctive), and they have infinitives and (apart from skulu and munu) present and past participles. (69) a hefi mtt rigna mjg miki til ess a stflan it had.CONJ may.PP rain.INF very much to it to dam It would have to rain a lot for the dam to break. brysti. break

On the whole, therefore, the Icelandic modal verbs still show a fairly complete conjugation paradigm. They conform to Thrinsson and Vikners generalization over Scandinavian modal verbs:
Modal verbs show subject-verb agreement in those Scandinavian languages that have subject-verb agreement in general, i.e. in Faroese [...] and Icelandic, and not in those where the finite verb never shows any kind of agreement with the subject (nor with anything else, i.e. Danish, Norwegian, and Swedish). (Thrinsson and Vikner 1995: 53)

It should also be noted that a number of non-preterite-presents complement the category of modal verbs in Icelandic: geta can, vera must, hljta must and tla intend.28 Whereas geta is combined with a past participle, the other ones take an a-marked infinitive. Interestingly, these verbs express both non-epistemic (ac) and epistemic meanings (bd), as the following examples with vera (70ab) and geta (70cd) show. (70) a. Hann verur a fara. he must to go He must go. b. a var a vera. that must.PST to be.INF That must have been so./That must have been the reason. c. Hann getur komi. He can come.PP He is able to come.

Modals in the Germanic languages 27

d. a getur veri. it can be.PP It may be. We can therefore conclude that the process whereby the original preterite-presents develop into a modal paradigm which is clearly distinct from all other verbs has not taken place to a high extent in Icelandic. The survival of the non-modal preterite-presents supports this conclusion as well as the fact that the modal preterite-presents behave in a less systematic way as far as their complementation patterns are concerned (allowing for both bare and amarked infinitives in epistemic and non-epistemic uses). Moreover, other verbs with different morphological and syntactic characteristics have entered into the modal verb category of Icelandic.29 We will return to this issue in section 5.4. 5.2. Paradigmatic variability

Icelandic modal verbs still show a fairly complete conjugation paradigm (cf. above). If we disregard the lack of modal imperatives, which probably has semantic reasons, the Icelandic modals do not in general show a decrease of paradigmatic variability in the sense that there are less verb forms at disposal the only exceptions here are munu and skulu. If we concentrate on the verbs expressing epistemic modality, however, we observe again that epistemic forms are typically finite. As a rule, epistemic modal infinitives cannot be combined with non-epistemic modal finites (example from Thrinsson and Vikner 1995: 78). The following example, in which the first occurrence of a kunna has an intended epistemic reading, is therefore unacceptable. (71) *Hann he verur must a kunna a kunna to can.INF to can.INF synda. swim

Nevertheless, Thrinsson and Vikner argue that in both Danish and Icelandic an epistemic modal finite can combine with an epistemic modal infinitive; they provide the following examples for Danish (72ab) and Icelandic (73ab) (Thrinsson and Vikner 1995: 7677): (72) a. Det m kunne st p en side. it must can.INF stand on one page It must be possible to fit it onto one page. b. Der vil let kunne g noget there will easily can.INF go something It will easily be possible that something goes wrong. (73) a. a mun vilja rigni mean i eru it will want rain.INF while you are It will tend to rain while you are there. b. Strkana tla i a vilja reka boys.DEF intended to want.INF drift to It looked like the boys tended to drift ashore. ar. there land. land

galt. wrong

As for Danish, however, Brandt (1999: 130132) argues convincingly that both (72a) and (72b) involve non-epistemic modal infinitives (of the situational kind). Double epistemic modals are not possible in Danish. As for Icelandic, the examples seem dubious as well. For both (73a) and (73b), it can be claimed that vilja has a future rather than an epistemic reading. While double epistemic modals may be excluded in both Danish and Icelandic, however, epistemic modal infinitives are found (although only rarely) in both languages (cf. section 4.2.). The requirement is that the finite verb is epistemic (an epistemic raising verb, in fact) but non-modal. Thrinsson and Vikner provide Icelandic examples of this, but they are not

28 Tanja Mortelmans, Kasper Boye & Johan van der Auwera

quite convincing. (74a) and (74b) (provided by Ellert Johansson, p.c.) are better examples, and (74c) (repeated from section 4.2) is a good Danish example. (74) a. Mr virist kunna a hafa rignt. me.DAT seems can.INF to have.INF rain.PP It seems to me that it may have been raining. b. Hann hltur a vera a vera Islandi. he seems to must.INF to be.INF on Iceland It seems that he must be in Iceland. c. Han synes at mtte vre morderen. he seems to must.INF be murderer.DEF It seems that he must be the murderer. Both finite and non-finite modals are used epistemically and non-epistemically, then. 5.3. Structural scope

Non-epistemic Icelandic modals are found combining with nouns and directional adverbials, rather than with infinitives. (75) a. Hann kann fyrsta boori. he can first commandment He knows the first commandment. b. g vill heim. I want home I want to go home. Just like its cognates in Danish, Dutch and German, the modal vilja even combines with a finite complement clause. (76) Hann vill a u he will that you He wants that you come. komir. come

With epistemic modals, however, all these constructions seem to be impossible. Epistemic modals require an infinitive (which may however be marked by a) or a past participle (in the case of the non-preterite-present geta).

5.4.

The Icelandic modals: summary

It could be argued that there is a highly grammaticalized category of four core modals (munu, skulu, mega and vilja) in Icelandic that share all the relevant characteristics of the core modals that we have used as delimiting criteria so far: they are preterite-present polyfunctional modals taking a bare infinitive. One could even, as was argued in section 5.1., go one step further and claim that only two modals (munu and skulu) are at the heart of this category, on behalf of the fact that their formal inventory is heavily reduced. Following this line of thinking, the degree of grammaticalization of the Icelandic core modal category would be rather high, as it consists of only two members. However, the category of Icelandic core modals (whether it consists of either two or four members) seems to distinguish itself crucially from the core modal categories in the other languages to the extent that the differences with respect to the non-core modals are of a more gradual nature. Thus, we do not only find four modal polyfunctional preterite-presents taking a bare infinitive, but also three modal polyfunctional preterite-presents taking an a-marked infinitive30 (eiga ought (to), kunna can, may, and urfa need), and a number of modal polyfunctional auxiliaries that

Modals in the Germanic languages 29

are non-preterite-presents (e.g. geta can, vera must, hljta must), most of which are combined with an a-marked infinitive (geta needs a past participle). Moreover, the Icelandic core modals share the preterite-present feature with three main verbs vita know, unna love and muna remember. Thus, we would suggest that the delimiting criteria in Icelandic do not have the same degree of saliency as in the other Germanic languages, as the criteria are shared by a considerable number of other auxiliary (and even main) verbs as well. So, although a clearly delimitable group of highly grammaticalized core modals can be isolated, it remains doubtful whether this paradigm is as highly paradigmatized as in the other Germanic languages, as it shares its main (formal as well as semantic) features with a considerable number of other verbs.
Table 12. Icelandic modals: paradigmaticity paradigmaticity [-] - Core modals share features with other modal and main verbs: (1) preterite-presents modals with bare infinitive: munu, skulu, mega, vilja (2) preterite-presents modals with a-infinitive: eiga, kunna, urfa (3) non-preterite-presents modals with a-infinitive: vera, hljta, tla (4) non-preterite-present modal with past participle: geta (5) other modals: taka, fara, (6) preterite-present main verbs: vita, unna, muna

[+] Small paradigm consisting of either two (munu, skulu) or four core modals (munu, skulu, mega, vilja).

As far as paradigmatic variability is concerned, we observe the same tendencies as in the other Germanic languages: with epistemic readings, finite forms of the modals are clearly preferred, although epistemic uses with infinitives (in the scope of finite epistemic raising verbs) can be found. It could be argued that paradigmatic variability is thus a bit lower than with the non-epistemic modals, as they can freely occur in all kinds of nonfinite constructions. Finally, epistemic modals have a much more restricted structural scope than non-epistemic ones, since they only combine with verbal complements (which, if one also takes the non core modals into account) are considerable more varied than in the other Germanic languages). As was the case for the other Germanic languages, this may be taken to suggest that epistemic modals are more grammaticalized than non-epistemic ones.
Table 13. Icelandic modals: paradigmatic variability and structural scope paradigmatic variability [+] [-] Non-epistemic modals occur in finite and Epistemic modals prefer finite forms of the nonfinite forms (participles, infinitives). modals, but do not exclude nonfinite forms. structural scope Non-epistemic modals have a wide structural scope. Epistemic modals only take verbal complements.

6. General conclusions In so far as grammaticalization is conceived of in accordance with Lehmanns parameters, we can set up a grammaticalization cline, starting from the highly grammaticalized English modals and ending up with the Icelandic ones (due to their lower degree of paradigmaticity). The Danish modals with their rather restricted inflectional paradigm (mainly two forms,

30 Tanja Mortelmans, Kasper Boye & Johan van der Auwera

present-finite and non-present non-finite, in an extreme formulation) share many similarities with the English modals, whereas the German and Dutch modals hover somewhere inbetween. For the latter, the structural scope is remarkably wide, as the Dutch modals allow for a very broad range of complements and often occur in what we have labelled intransitive use, i.e. without a complement (which however does not necessarily give rise to an elliptic analysis). So, although the grammar of Dutch is often intermediate between those of German and English (Van Haeringen 1956, Hning et al. 2006), this does not seem to be the case for the grammar of the modals. The distinction between epistemic and non-epistemic modality is relevant for all four languages: the epistemic modals are clearly more grammaticalized than the non-epistemic ones, although the often cited reluctance against nonfinite modals in epistemic use (the epistemic non-finiteness gap, see e.g. Abraham (2001)) had to be relativized for every language. In fact, we would suggest that this reluctance against non-finites in epistemic use is more bound up with semantic reasons than with purely grammatical ones. As far as the question regarding the usefulness of applying Lehmanns parameters to this particular group of verbs is concerned, we have to provide a two-sided answer. First of all, the parameters can be said to function as eye-openers and classification tools at the same time. For this clearly delimitable group of verbs, the parameters function as torch lights, leading one to observations that might otherwise pass unnoticed. Moreover, seemingly arbitrary facts concerning the form and behaviour of the core modals can be integrated in a larger framework and understood accordingly. This immediately brings us to the second point we would like to raise: these parameters should not be applied in isolation, i.e. without taking the entire structure of the verbal system of a particular language in account. In languages with a strongly developed inflectional system, the modal verbs can be expected to reflect this, i.e. they can be expected to have more inflectional categories at their disposal than modals in languages where these categories do not exist (anymore). Whether this automatically leads to a lesser degree of grammaticalization, however, is doubtful and can only be decided after carefully comparing the item with ordinary verbs in its own language with respect to the other grammaticalization parameters. By the same token, the fact that a particular language doesnt generally mark subject-verb agreement (e.g. Danish), will have its bearing on the formal make-up of the Danish class of core modals as well: it can be expected that Danish modals wont be marked for subject-verb agreement either, and this could be interpreted as an indication of a higher degree of paradigmaticy (as the number of formal options within the paradigm decreases). From a cross-linguistic perspective, however, it would be wrong to claim on the basis of this observation alone that Danish modals score higher with respect to paradigmaticity than their German or Dutch counterparts, which do mark subject-verb agreement (and hence have more forms at their disposal). The former also implies that Lehmanns conception of grammaticalization does not in itself constitute an ideal means for measuring the degree of grammaticalization: what is a sign of grammaticalization according to Lehmanns conception, may in fact occasionally be due to factors non-related to grammaticalization as such. General semantic factors, for instance, or the overall structure of a particular language can highly influence the linguistic appearance of particular meanings and meaning categories, without however being immediately tied to grammaticalization per se.

Notes
Thanks are due to the University of Antwerp with financial help to Tanja Mortelmans and Johan van der Auwera (GOA grant 2003-2006). Kasper Boye wishes to thank The Carlsberg Foundation for a grant which made his work on the paper possible, and Johan van der Auwera is also grateful to the University of Antwerp, Princeton University and the Flanders Research Foundation for supporting a sabbatical leave. Finally, we thank Jhanna Bardal and Ellert Thor Jhannsson for expert help on
1

Modals in the Germanic languages 31

Icelandic, Karen Margrethe Pedersen for expert help on Danish dialects and Bjrn Hansen and Ferdinand de Haan for their editorial patience. 2 Many languages use different markers to express non-epistemic and epistemic possibility and necessity (van der Auwera and Ammann 2005: 310314). 3 For more information on the modals in Norwegian, see Eide (2005); for Swedish, see Telemann et al. (1999); for Faroese, see Thrinsson et al. (2004); for Afrikaans, see Donaldson (1993), for Pennsylvania German, see Burridge (1997, 1998), for Yiddish, see Hansen/Birzer (in prep), for Low German, see Mortelmans (2007). 4 The Icelandic modals kunna, eiga and urfa are added between brackets, as they combine with an amarked infinitive only (which incidentally also holds for ought in the English group, which doesnt combine with a bare infinitive either). Vilja is added between brackets as well, as it does not conform to the characteristics of a preterite-present. Following Birkmann (1987: 377), modal vilja is a weak verb with some formal irregularities in the 2nd and 3rd person of the present indicative (syntactically, vilja behaves as a central modal, combining with a bare infinitive though). 5 The modal will and its cognates in the other Germanic languages are not real preterite-presents, as they go back to original optative forms with present meaning (Birkmann 1987: 116). In many Germanic languages, the will-cognates have been formally integrated in the preterite-present paradigm, (with the exception of Icelandic). 6 The Danish modals turde and burde are both considered as non-central modals, the former because of the fact that it can also take an at-marked infinitive (see section 2.3.1), the latter because of its rather low frequency. Note also that in spite of the fact that they formally behave like preterite-presents in present-day Danish, their ancestors did not belong to this category (Birkmann 1987: 381). 7 The realization of will as ll counts as an instance of attrition (decrease of formal integrity). It should be noted that the Danish modals ville and skulle do show attrition, as their pronunciation involves deletion of the final consonant (and schwa). 8 The bondedness of a sign is defined as the intimacy with which it is connected with another sign to which it bears a syntagmatic relation (Lehmann 2002: 131). The English modals fuse with negation markers (cant, mustnt) and pronouns (Ill), which obviously increases their degree of bondedness. In general, this parameter does not seem to be very relevant for the Germanic modals, as they all function as syntagmatically independent items, at least in written language (but see note 18). 9 This does not imply that the central modals are thought to be completely regular or even uninteresting as far as word order is concerned. German and Dutch, for instance, differ with respect to the position of the modal in constructions with double infinitives (see also section 2.1). Compare Dutch hij heeft het niet kunnen zeggen (Vaux Infmod Inf) with Er hat es nicht sagen knnen (Vaux Inf Infmod) He hasnt been able to say it. Moreover, there are interesting word order differences in subclauses. For the following example with epistemic moeten, Dutch allows three possibilities dat hij gezegd moet hebben/moet gezegd hebben/moet hebben gezegd), whereas German only allows the word order [PP Inf VMOD] dass er gesagt haben soll, with the modal verb in final position. 10 For the German modals, this is claimed by Diewald (1997: 19-27). 11 Note that it is not always that easy to distinguish between non-modal and modal (non-epistemic) meanings. It is quite unclear for instance whether German knnen meaning know should be considered to express a non-modal (lexical) meaning or a modal (dynamic) one in the sentence Sie kann Deutsch (sprechen). 12 The use of mogen for edible entities is restricted to Southern Dutch (Diepeveen et al. 2006: 98). More frequently, the lexical verb lusten is used in these contexts. Conversely, the use of mogen with persons (expressing feelings of sympathy) is restricted to Northern Dutch (Diepeveen et al. 2006). 13 Although this is generally frowned upon and characterized as Southern Dutch, the modal kunnen pops up both in Southern and Northern Dutch corpora with a comparable frequency (Diepeveen et al. 2006: 39-40). 14 The forms without t (which are the younger ones) are considered to be informal in Northern Dutch, which generally prefers the forms with t. In Southern Dutch, however, no such distinction is made; here the forms without t are typical of spoken language (Haeseryn et al 1997: 94100). 15 Whereas wou is mainly used in spoken language, the form wilde is more typical of written language (Renkema 1995: 103); the plural form wou(d)en is considered as highly informal spoken language (Smets and Van Belle 1997: 138).

32 Tanja Mortelmans, Kasper Boye & Johan van der Auwera

IPP-constructions are mainly characteristic of the West Germanic languages (but see section 2.3.1 for an interesting exception in Danish); they can be found in German, Afrikaans, Swiss German and Dutch (Schmid 2005:12). 17 The condition that the modal has to denote a value on a bounded scale when it combines with a nominal or adjective phrase does not hold for cases in which the modal has a lexical reading (see ex. (7a-b-c) above). 18 In fact, the possibility of such intransitive uses can also be interpreted as an instance of less syntagmatic cohesion (see Cornillie et al., this volume). 19 The preterite present wissen governs an infinitive with zu, in which case it expresses an ability-like meaning. Epistemic meanings of wissen do not exist, though. 20 Following hlschlger (1989: 64), Diewald (1999) considers this use as elliptic. 21 The modal mtte (which is related to English may and Dutch mogen) may be regarded as multifunctional, as it expresses both possibility and necessity. Note, however, that epistemic mtte can only express (epistemic) necessity. 22 Counterparts of burde which is absent in the West-Germanic languages are also found among the Swedish (bra) and the Norwegian (burde) modals. 23 The modal turde has its origin in a fusion between two distinct verbs: one which is cognate with German drfen and which meant need, must, and another which meant dare. In table 2, turde is represented as a cognate of drfen. 24 Synthetic passives are least uncommon for ville (villes) and kunne (kunnes) (Brandt 1999: 69). Also skulles and mttes are marginally acceptable. 25 As for the formal identity between past participles and infinitives (and pasts) of preterite-present modals, it might be claimed that it is phonetically (and orthographically) conditioned only, residing in the omission of final -t from the conservative participles (e.g. kunnet) to yield forms identical to the infinitives (e.g. kunne). Such omission is occasionally found also with regular main verbs. However, the convergence of past participles and infinitives (and pasts) of preterite-present modals is found in several Danish dialects (e.g. Seelandic) where such a simple phonetic explanation is inadequate. In the case in point, the difference between the conservative participles and the infinitives (and pasts) is marked by presence vs. absence of std (the special Danish glottal stop) in addition to presence vs. absence of final -t. This suggests that the convergence resides in the disappearance of the past participle as a unique form, rather than in a phonetic omission (Karen Margrethe Pedersen, p.c.). 26 Unless it is noted, examples come from either Thrinsson and Vikner 1995: 55-57, or from Einarsson 1945: 161 and 164167. 27 For munu there is the form mundi, which is however not used with past-time meaning. It is accordingly classified in Einarsson (1945: 103) as a preterite subjunctive side by side with myndi. 28 Birkmann (1987: 377) also mentions taka and fara. The verb geta has counterparts in English and Danish, too; for a study of these acquisitive modals, modals deriving from a lexical meaning to acquire, see van der Auwera, Kehayov and Vittrant (in print). 29 On the basis of a small-scale investigation (a comparison of a small parallel corpus), Hrafnbjargarson (2007) shows that the other Scandinavian languages (Danish, Norwegian, Swedish and Faroese) make use of a limited and much more similar inventory of modals than Icelandic. So, Norwegian kunne translates as kunne or f in Faroese, kunne or ville in Danish or kunna and f in Swedish. The Icelandic translation contains no less than four alternatives (geta, kunna, mega, skulu). Similarly, whereas Norwegian skulle is rendered by skula or kunna in Faroese, ville or skulle in Danish and skola in Swedish, Icelandic translates the occurrences by means of skulu, munu, eiga and urfa. The same holds for the translation of Norwegian mtte and ville: Icelandic has four alternatives for the former (hljta, vera, mega and urfa) and three for the latter (munu, vilja, tla) whereas none of the other languages lists more than two alternatives for either verb. 30 Note that at-marked infinitives are found with some Danish modals as well, but only on a nonepistemic reading. In Icelandic, however, these complements occur with both non-epistemic and epistemic uses of the modals.

16

References
Abraham, Werner

Modals in the Germanic languages 33

Modals: toward explaining the epistemic non-finiteness gap. In Modalitt und Modalverben im Deutschen, Reimar Mller and Marga Reis (eds.), 736. (Linguistische Berichte Sonderheft 9.) Hamburg: Helmut Buske Verlag. 2002 Modal verbs: Epistemics in German and English. In Modality and its Interaction with the Verbal System, Sjef Barbiers, Frits Beukema, and Wim van der Wurff (eds.), 19 50. (Linguistik Aktuell/Linguistics Today 47.) Amsterdam/Philadelphia: John Benjamins. Askedal, John Ole 1998 brauchen mit Infinitiv. Aspekte der Auxiliarisierung. In Jahrbuch der ungarischen Germanistik 1997, Antal Mdl and Gunther Dietz (eds.), 5368. Budapest: Gesellschaft Ungarischer Germanisten; Bonn: Deutscher Akademischer Austauschdienst. Barbiers, Sjef 2002 Modality and Polarity. In Modality and its Interaction with the Verbal System, Sjef Barbiers, Frits Beukema, and Wim van der Wurff (eds.), 5173. (Linguistik Aktuell/Linguistics Today 47.) Amsterdam/Philadelphia: John Benjamins. Biber, Douglas, Stig Johansson, Geoffrey Leech, Susan Conrad, and Edward Finegan 1999 The Longman grammar of spoken and written English. London: Longman. Birkmann, Thomas 1987 Prteritoprsentia. Morphologische Entwicklungen einer Sonderklasse in den altgermanischen Sprachen. (Linguistische Arbeiten 188.) Tbingen: Niemeyer. Boye, Kasper 2001 The force-dynamic core meaning of Danish modal verbs. Acta Linguistica Hafniensia 33: 1966. 2005 Modality and the concept of force-dynamic potential. In Modality. Studies in form and function, Alex Klinge and Henrik Heg Mller (eds.), 4980. London: Equinox. Brandt, Sren 1999 Modal Verbs in Danish. (Travaux du Cercle Linguistique de Copenhague XXX). Copenhagen: The Linguistic Circle of Copenhagen. Burridge, Kate 1997 On the trail of the Conestoga modal: recent movements of modal auxiliaries in Pennsylvania German. In Languages and lives. Essays in honor of Werner Enninger, James R. Dow and Michele Wolff (eds.), 728. New York: Peter Lang. 1998 From Modal Auxiliary to Lexical Verb: The Curious Case of Pennsylvania German Wotte. In Historical Linguistics 1995. Vol II: Germanic Linguistics, Richard M. Hogg and Linda van Bergen (eds.), 1933. (Current Issues in Linguistic Theory 162.) Amsterdam: Benjamins. Bybee, Joan 2003 Mechanisms of Change in Grammaticization: The Role of Frequency. In The Handbook of Historical Linguistics, Brian D. Joseph and Richard D. Janda (eds.), 602623. Malden MA, Oxford: Blackwell. Coates, Jenifer 1983 The semantics of the modal auxiliaries. London: Croom Helm. Davidsen-Nielsen, Niels 1990 Tense and Mood in English. A Comparison with Danish. Berlin: Mouton de Gruyter. De Haan, Ferdinand 2001 The relation between modality and evidentiality. In Modalitt und Modalverben im Deutschen, Reimar Mller and Marga Reis (eds.), 201216. (Linguistische Berichte Sonderheft 9.) Hamburg: Helmut Buske Verlag. 2006 Typological approaches to modality. In The expression of modality, William Frawley (ed.), 2769. Berlin/New York: Mouton de Gruyter. Diepeveen Janneke, Ronny Boogaart, Jenneke Brantjes, Pieter Byloo, Theo Janssen and Jan Nuyts 2006 Modale uitdrukkingen in Belgisch-Nederlands en Nederlands-Nederlands: corpusonderzoek en enqute. Amsterdam, Mnster: Stichting Neerlandistiek VU Amsterdam, Nodus Publikationen. Diewald, Gabriele 1997 Grammatikalisierung. Eine Einfhrung in Sein und Werden grammatischer Formen. Tbingen: Niemeyer.

2001

34 Tanja Mortelmans, Kasper Boye & Johan van der Auwera

1999

Die Modalverben im Deutschen. Grammatikalisierung und Polyfunktionalitt. (Reihe Germanistische Linguistik 208.) Tbingen: Niemeyer.

Donaldson, Bruce 1993 A grammar of Afrikaans. (Mouton Grammar Library 8.) Berlin: de Gruyter. Duden 2005 Die Grammatik. 7., vllig neu erarbeitete und erweiterte Auflage. Herausgegeben von der Dudenredaktion. Mannheim, Leipzig, Wien, Zrich: Dudenverlag. Eide, Kristin Melum 2005 Norwegian modals. ( Studies in generative grammar 74.) Berlin: Mouton de Gruyter. Einarsson, Stefn 1945 Icelandic Grammar, Texts, Glossary. Baltimore: The John Hopkins Press. Goossens, Louis 1985 Differentiating the English modals in Functional Grammar. In Antwerp Studies in Functional Grammar, Jan Nuyts (ed.), 4971. Antwerp: U of Antwerp. Haeseryn, Walter, K. Romijn, G. Geerts, J. de Rooij and M.C. van den Toorn 1997 Algemene Nederlandse Spraakkunst. 2nd edition. Groningen: Nijhoff. Hammerich, L.L. 1960 ber die Modalverben der neugermanischen Sprachen. Zeitschrift fr deutsche Wortforschung 16: 4770. Hansen, Bjrn 2004 Modals and the boundaries of grammaticalization: The case of Russian, Polish and Serbian-Croatian. In What makes Grammaticalization? A look from its Fringes and its Components, Walter Bisang, Nikolaus P. Himmelmann and Bjrn Wiemer (eds.), 245270. Berlin, New York: Mouton de Gruyter. Hansen, Bjrn and Sandra Birzer In prep. The Yiddish modal system between Germanic and Slavonic. A case study on the borrowability of modals. In Grammaticalization and language contact, Bjrn Wiemer, Bernhard Wlchli and Hansen, Bjrn, (eds.). Hansen, Erik 1977 Behver vi at? Nyt fra Sprognvnet 19: 14. Copenhagen: The Danish Language Council. Hofmans, Mark 1980a Hebben of zijn: Een enquete naar het gebruik van hebben of zijn in de konstruktie Thww + Mhww + HWW in Nederland en Vlaanderen. Rapport dactivits de linstitut de phonetique 14: 83129. 1980b Hebben of zijn en de deverbalisering van de modale werkwoorden in het Nederlands. In Linguistics in Belgium/Lingustiek in Belgi/Linguistique en Belgique 5, Dominicy, Marc (ed.), 81109. Brussel: Didier Hatier. Hrafnbjargarson, Gunnar Hrafn 2007 Generating a lexicon of Scandinavian modals. Nordlyd 34: 183199. Hning, Matthias, Ulrike Vogl, Ton van der Wouden and Arie Verhagen (eds.) 2006 Nederlands tussen Duits en Engels. Handelingen van de workshop op 30-9 en 1-102005 aan de Freie Universitt Berlin. Leiden: SNL. Jensen, Torben J. 2005 Irrealitetsmarkrer i dansk talesprog. En korpuslingvistisk undersgelse af hviskonstruktioner og modalverber i medarbejdersamtaler og sociolingvistiske interview. Ph.D. Thesis. Copenhagen: University of Copenhagen. Kotin, Michail L. 2003 Die werden-Perspektive und die werden-Periphrasen im Deutschen. (Danziger Beitrge zur Germanistik 6.) Frankfurt/M.: Peter Lang. Kress, Bruno 1982 Islndische Grammatik. Munich: Max Hueber Verlag. Krug, Manfred 2000 Emerging English Modals. A Corpus-based Study of Grammaticalization. Berlin, New York: Mouton de Gruyter. Lauridsen, Karen M., and Ole Lauridsen

Modals in the Germanic languages 35

Modalverber med passiv infinitiv i moderne dansk. 2. Mde om Udforskningen af Dansk Sprog, 239-249. Lehmann, Christian 1995 Synsemantika. In Syntax. Ein internationales Handbuch zeitgenssischer Forschung. 2. Halbbd, Joachim Jacobs, Arnim von Stechow, Wolfgang Sternefeld and Theo Vennemann (eds.), 12511266. Berlin, New York: de Gruyter 2002 Thoughts on grammaticalization. Second, revised edition. (Arbeitspapiere des Seminars fr Sprachwissenschaft der Universitt Erfurt 9.) Erfurt: Seminar fr Sprachwissenschaft der Universitt. Mortelmans, Tanja 2004 Grammatikalisierung und Subjektivierung: Traugott und Langacker revisited. Zeitschrift fr germanistische Linguistik: deutsche Sprache in Gegenwart und Geschichte 32: 188209. 2007 Modalverben im Niederdeutschen: Ansatz zu einem Vergleich mit dem Modalverbbestand im Deutschen und im Niederlndischen. Niederdeutsches Wort: Beitrge zur niederdeutschen Philologie, 47/48: 135148. Nuyts, Jan 2001 Epistemic modality, Language and Conceptualization. Amsterdam: John Benjamins. 2006 Modality: Overview and linguistic issues. In The expression of modality, William Frawley (ed.), 126. Berlin/New York: Mouton de Gruyter. Nuyts, Jan, Pieter Byloo and Janneke Diepeveen 2005 On deontic modality, directivity, and mood. A case study of Dutch mogen and moeten. Antwerp Papers In Linguistics 110. Antwerp: University of Antwerp. hlschlger, Gnther 1989 Zur Syntax und Semantik der Modalverben des Deutschen. (Linguistische Arbeiten 144.) Tbingen: Niemeyer Palmer, Frank R. 2001 Mood and modality. 2nd edition. Cambridge: Cambridge University Press. Plank, Frans 1984 The modals story retold. Studies in language 8.3: 305364. Ponten, Peter Jan 1973 Der Ersatz- oder Scheininfinitiv. Ein Problem aus der deutschen und niederlndischen Syntax. Wirkendes Wort 23.2: 7385. Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech and Jan Svartvik 1984 A comprehensive grammar of the English language. London: Longman. Reis, Marga 2001 Bilden Modalverben im Deutschen eine syntaktische Klasse? In Modalitt und Modalverben im Deutschen, Reimar Mller and Marga Reis (eds.), 287318. (Linguistische Berichte Sonderheft 9.) Hamburg: Helmut Buske Verlag. Renkema, Jan 1995 Schrijfwijzer (3rd ed.). Den Haag: Sdu. Schmid, Tanja 2005 Infinitival Syntax. Infinitivus Pro Participio as a repair strategy. (Linguistik Aktuell/Linguistics Today 79.) Amsterdam: John Benjamins. Skyum-Nielsen, Peder 1971 Modalverberne I Nudansk. Unpublished prize essay. Smedts, Willy and William Van Belle 1997 Taalboek Nederlands. Kapellen: Pelckmans. Telemann, Ulf, Staffan Hellberg and Erik Andersson 1999 Svenska Akademiens Grammatik. 14. Stockholm: Norstedts Ordbok. Thrinsson, Hskuldur and Sten Vikner 1995 Modals and Double Modals in the Scandinavian Languages. Working Papers in Scandinavian Syntax 55:5188. Dept. of Scandinavian Linguistics, Lund University. Thrinsson, Hskuldur, Hjalmar P. Petersen, Jgvan Lon Jacobsen and Zakaris Svabo Hansen 2004 Faroese. An Overview and Reference Grammar. Trshavn : Froya Frdkaparfelag. Traugott, Elizabeth Closs 1989 On the rise of epistemic meanings in English: An example of subjectification in semantic change. Language 65,1:3155.

1989

36 Tanja Mortelmans, Kasper Boye & Johan van der Auwera

van der Auwera, Johan 1999 On the semantic and pragmatic polyfunctionality of modal verbs, in The semantics/pragmatics interface from different points of view, Ken Turner (ed.), 4964. Oxford: Elsevier. van der Auwera, Johan and Andreas Ammann 2005 Overlap between situational and epistemic modal marking. In The World Atlas of Language Structures, Martin Haspelmath, Matthew Dryer, David Gil and Bernard Comrie (eds.), 310314. Oxford: Oxford University Press. van der Auwera, Johan and Vladimir Plungian 1998 Modalitys semantic map. Linguistic Typology 2: 79124. van der Auwera, Johan, Petar Kehayov, and Alice Vittrant In print. Acquisitive modals. In Cross-linguistic Studies of Tense, Aspect, and Modality, Lotte Hogeweg, Helen De Hoop and Andrej Malchukov (eds.). Amsterdam: Benjamins. Van Haeringen, C.B. 1956 Nederlands tussen Duits en Engels. Den Haag: Servire. Van Ostaeyen, Gert and Jan Nuyts 2004 De diachronie van kunnen. Antwerp Papers in Linguistics 109. Antwerp: U of Antwerp.

Вам также может понравиться