A Landslide in the Field of Deontic Modals:
The Rise of Polar moeten and hoeven in (Middle) Dutch
Thesis Research Master’s in Linguistics Anna Marlijn Meijer
Supervisors: Prof. dr. F.P. Weerman
Prof. dr. H.H. Zeijlstra July 28, 2014
Abstract
This paper argues that the present day Holland Dutch universal modal moeten was reanalysed as a positive polarity item (PPI) (an item taking scope over negation), in the mid sixteenth century, due to a shift in the modal landscape. Around this time, the Middle Dutch existential deontic negative polar modal dorven was lost. This loss caused a landslide in the field of deontic modal verbs. The result of this landslide was the reanalysis of the neutral modal moeten into a PPI (as defined in Iatridou and Zeijlstra 2013); presumably, moeten started to outscope negation in order to fill a lexical gap. This change seems to have lead, on its turn, to the rise of the neutral universal modal hoeven (which is a negative polar item (NPI) in Modern Dutch). Strikingly, this suggests a rather different diachronic pathway than was assumed up and till now for these modals in the literature (see Van der Wouden 1996). The present study thus provides a new and more coherent pathway, and thereby describes the circumstances under which items are reanalysed as polar items.
1
Introduction
Universal modals differ with respect to scoping below or above negation. The Holland Dutch deontic universal modal moeten (‘must’), for instance, scopes above negation, see (1).
(1) Jan Jan moet must niet neg slapen sleep ‘Jan must not sleep’
2 > ¬
The sentence above thus states that it is necessary for Jan to not go to sleep (which is also true for the English translation given). The German counterpart of moeten, m¨ussen (‘must’), on the other hand, scopes below negation, see (2).
‘Jan need not sleep’ ¬ >2
Sentence (2) states that it is not necessary for Jan to sleep. According to Iatridou and Zeijlstra (2013), the positive polarity-hood of moeten explains this difference. Since moeten is a positive polarity item (PPI), it cannot be interpreted below negation. Its German cognate m¨ussen is a neutral item, that can therefore occur in the scope of negation. Historical data suggests that moeten was more similar to m¨ussen in the past, as it scoped below negation as well. The Middle Dutch sentence below illustrates this.
(3) Context: Jesus is talking to the saint Lutgardis. a. dat that gi you nit neg noede worry En neg motet must doen. do want because ic I v you dis this Geue good orlof permit ‘You do not have to worry because I favour you with the good’
¬ >2 (Affligem 1265-70)
This apparent change in moeten is not surprising. Since moeten and m¨ussen stem from the same verb, it is only natural to assume that their semantics was similar once. Furthermore, German seems to have been subject to fewer changes than English or Dutch in general (Weerman 2007); this makes it not unlikely that the usage of m¨ussen is more related to the usage of moeten in an older phase of Dutch than the present day usage. However, it is not clear why this is the case. The present study aims to find an explanation for this difference and revolves around the following research questions:
Research Question 1
1a Why did Holland Dutch moeten become a PPI? 1b When did Holland Dutch moeten become a PPI?
It is presumable that the rise of polar moeten is related to a change in other modals. Interestingly, in Middle Dutch a modal shift is reported, involving two modals. First, the Holland Dutch modal hoeven (‘need’) is reported to have emerged in this period. Second, the Middle Dutch modal dorven (‘can’, ‘need’) was lost. This modal is rather unusual, as it displays variation in modal force. This means that it can both surface as an existential modal (‘can’) and an universal modal (‘need’) - a property that was only reported in non-European languages, such as Lillooet Salish (see Rullmann et al. 2008). Besides that, the Woordenboek der Nederlandsche Taal (2011) (henceforth: WNT) notes that dorven was a negative polarity item (NPI), which means that it could only occur in the scope of a negative element. However, apart from that little is known about dorven. Concerning hoeven, it is known that it developed out of the lexical verb (be)hoeven (‘to be in need of’) (WNT 2011). Thereafter, it grammaticalized into an NPI (which happened at the latest in the early nineteenth century, according to Van der Wouden 1996 and Van Baal 2013). The rise of polar hoeven is ascribed to the loss of the negative polar modal dorven. Both modals namely scoped below negation with the meaning ‘need’, which may
The rise of the polar modal hoeven could in principle also easily be connected to the PPI-hood of moeten. The Holland Dutch universal modal hoeven namely takes scope below negation (just like m¨ussen and Middle Dutch moeten), see (4):
(4) Jan Jan hoeft needs *(niet) neg te to slapen sleep ‘Jan need not sleep’
¬ >2
Moreover, sentence (4) would be ungrammatical without a negation, since hoeven is an NPI in Modern Dutch. As the usage of moeten in Middle Dutch is very similar to that of present day Dutch hoeven, it appears likely that there is a diachronical correlation between the rise of modal NPI hoeven and PPI moeten. In Middle Dutch, moeten was a universal modal that took scope below negation. This is exactly what hoeven does in Modern Dutch. Thus, in order to investigate what happened with these modals in Middle Dutch, one should look at the historical data. In addition, in order to examine the correlations, the role the loss of the modal NPI dorven should be taken into account too. However, its role might not be as transparant as suggested in the literature, as this modal was flexible in expressing modal meaning. This gives rise to the follow up research questions below.
Research Question 2
2a Is there a diachronical correlation between the changes in moeten, hoeven and dorven? 2b Is the rise of polar moeten the result from changes in the modal landscape of Holland
Middle Dutch?
The present study thus aims to verify if there is a diachronical and perhaps even a causal correlation between the changes signaled in the semantics of the three modals indeed, or whether these were all independent developments. Thereby, its goal is to provide an answer to the question under what circumstances (modal) verbs are reanalysed as polar items. In particular, it aims to explain when and why moeten became a PPI.
The paper is set up as follows. Section 2 first briefly elaborates on the semantics of modal verbs. Second, it sketches the Middle Dutch modal landscape in more detail. It furthermore discusses the hypotheses that result from the data and the changes reported. Section 3 considers the corpus study that was conducted in order to answer the questions asked. The results are discussed in section 4. Section 5 concludes.
2
Changes in the Modal Landscape
The current section aims to provide more insight into the modals dealt with. In order to gain this, it first briefly considers the semantics of modals. Second, it discusses the Middle Dutch verbs moeten, dorven and hoeven in more detail and examines the changes undergone by these verbs. Attention is specifically payed to their changes with respect to taking scope above or below negation negation. Third, it considers the hypotheses and corresponding predictions that arise on the basis of what is known of the modals and the changes observed in them.
2.1 The Semantics of Modals
By expressing modals speakers can indicate whether they consider some proposition possible or necessary (Von Fintel 2006). These possibilities or necessities are always from a certain perspective; for instance one’s believes or hopes. The present study only considers deontic modality, which concerns a set of laws or rules. Consequently, it deals with what is allowed or obligatory, depending on the modal. Possibility modals (like ‘may’) represent existential force (and allowance); necessity modals (like ‘must’) universal force (and obligation), see (5).
(5) a. Sue may be home
b. Sue must be home (taken from Von Fintel 2006)
Kratzer’s framework of modality is often considered the standard work within the field of theo-retical linguistics (Portner 2009: 47, Von Fintel 2006: 22). Her work is grounded on modal logic and possible world theories. In these theories, the universe as we know it is taken to be the actual world (Portner 2009: 21). Possible worlds, then, are worlds that deviate from the actual world in great or minor details. One can think of a set of worlds in which certain wishes or believes are satisfied. In case of deontic modality, this set of worlds would describe the relevant set of laws or rules. Modals can now be understood as operators quantifying over this set of worlds. Thereby, they indicate which worlds (or scenarios) are (im)possible or (un)necessary. The relation between the evaluation world and domain of possible worlds is characterized by the so-called accessibility relation (Von Fintel 2006: 22).
Possibility modals, such as may in (6), yield an accessibility relation in which at least one of the possible worlds of some set, characterizing for instance one’s believes, is accessible. It thus corresponds to the existential quantifier. Figure 1 illustrates this. In this model, w1 is the actual world; w2, w3 and w4 are possible worlds. In worlds w2 and w3, John parks his car in this street. In w4, he does not do so. Now, we see that both w2 and w4 are accessible; so according to the speaker of (6), John has the possibility to park his car in this street (or to not do so).
(6) [In view of the laws,] John may park in this street
(Park in this street (j)) ≡ ∃w (R(w1, w) ∧ Park in this street (j)w)
w1 w2
w3
Park in this street (j)
w4 Park in this street (j)
Not park in this street (j)
Figure 1: Existential accessibility relation
Necessity modals, like must in (7), correspond to the universal quantifier: they indicate that all accessible worlds are, in case of (7), in accordance with the law. Figure 2 shows this. Now,
(7) [In view of the law,] John must go to jail
2 (Go to jail (j)) ≡ ∀w (R(w1, w) → Go to jail (j)w) w1 w2 Go to jail (j)
w3
w4 Go to jail (j) Not go to jail (j)
Figure 2: Universal accessibility relation
For more information on accessibility relations and the selection of the domain of possible worlds, the reader is referred to Portner (2009).
2.2 Modals From Early Middle to Modern Holland Dutch
This subsection deals with the changes in the Middle Dutch (1250-1550) modal landscape. For each of the modals moeten, dorven and hoeven, it briefly sketches the usages from Early Middle Dutch (1200-1300) till Modern Holland Dutch (1500-present).
2.2.1 Moeten : from neutral to polar modal
The earliest findings of moeten stem from the beginning of the thirteenth century, according to the Middelnederlandsch Woordenboek (2011) (henceforth: MNW). In these occurrences in Early Middle Dutch, moeten could both express possibility and necessity (MNW 2011). This variation is usually taken to indicate a shift in meaning from a possibility to a necessity modal (as opposed to inherent variation in force found in Salish modals, see Rullmann et al. 2008). Examples of these meanings are given in (8) and (9) respectively. (All examples are taken from the (Early) Middle Dutch entry for moeten in the MNW. Unfortunately, no contexts for the sentences are provided there.)
(8) her sir .Ridder. knight in in dit this lant country moeti may.he willecomen welcome sin be
‘[This] sir [the] knight, in this country he may be welcome’ (anonymous 1276-1300)
(9) Nu Now salic shall.I moten must saen immediatly beginnen start Dat that ander other deel part uan of desen these werke works
‘Now I will have to start immediately with the rest of these works’ (Affligem 1265-70)
There is no literature on moeten’s change from a possibility to a necessity modal. Nevertheless, the Middle Dutch entry in the MNW lists no examples of this usage after 1480 (although most examples date from before the fourteenth century). Interestingly, moeten’s Old and Middle English counterpart, mote, appears to have behaved similarly. Yanovich (2013) argues that there is a phase in which mote is ambiguous in force. This phase is preceded by a stage in which it is a genuine variable force modal (like the Salish modals). After the ambiguity phase, mote is reanalysed into an actual necessity modal - like moeten. The possibility readings of mote are already relative rare in the fifteenth century (Yanovich 2013: 139). The same ambiguous
Nonetheless, the present study focusses only on the necessity usage of moeten. As was already shown in the introduction, in this usage moeten took scope under negation. The example is repeated in (10).
(10) Context: Jesus is talking to the saint Lutgardis. a. dat that gi you nit neg noede worry En neg motet must doen. do want because ic I v you dis this Geue good orlof permit ‘You do not have to worry because I favour you with the good’
¬ >2 (Affligem 1265-70)
In (10), Jesus tells his listener that he has no need to worry, since he will be favoured with the good. The interpretation shows that moeten must be interpreted below negation. According to the MNW, moeten could take scope under negation throughout Early Middle Dutch and Middle Dutch. However, the introduction already stated that in present day Holland Dutch moeten always outscopes negative operators, such as a plain negation niet (‘not’), see (11).
(11) Jan Jan moet must niet neg slapen sleep ‘Jan must not sleep’ 2 > ¬, *¬ > 2
In (11), the narrow scope interpretation of moeten, present in (10), is not available. It states that in all accessible worlds, Jan does not sleep. We can therefore conclude that something has changed in moeten from the transition from Middle Dutch to Modern Dutch.
For German m¨ussen, section 1 stated that it functions more like the Middle Dutch variety of moeten, which scoped below negation. This difference in scope taking in German and Dutch can be further illustrated by the examples in (12). The sentence in (12-a) shows that niet moeten vertrekken (‘must not leave’) in Dutch cannot be followed up by a phrase with a possibility modal taking the same VP-complement (in this case ‘to leave’). This namely is a contradiction, as (12-a) first states that Jan is not allowed to leave and then that he is. In German, however, this is felitious (as shown in (12-b)). In my view, this provides crucial evidence for the difference between the modals.
(12) a. *Jan Jan moet must niet not vertrekken, leave, maar but hij he mag may wel prt vertrekken leave ‘Jan mustn’t leave, but he may’
2(¬ Leave (j)) ∧ 3 (Leave (j)) ≡ ⊥ b. Jan Jan muss must nicht not abfahren, leave aber but er he darf may abfahren leave ‘Jan doesn’t have to leave, but he may’ ¬(2 Leave (j)) ∧ 3 (Leave (j))
(13) Niemand no.one moet must vertrekken leave ‘No one must leave’ 2 > ¬ > ∃, * ¬ > ∃ > 2
The only interpretation available for (13) is one in which there is necessarily not a person, such that s/he leaves. Yet, there are weaker negative operators under which moeten can take scope, see (14). (14) Weinig Few studenten students moeten must vertrekken leave ‘Few students must leave’
few > 2, 2 > few
The quantifier weinig (‘few’) can outscope moeten in specific contexts. This makes moeten a PPI of medium strength (Iatridou and Zeijlstra 2013). It can namely outscope some kind of ‘inherent’ negative operators (such as ‘no’ and ‘nobody’), which are also called anti-additive operators.1 A plain negation is anti-additive as well, but such a negation like not is much stronger than a regular anti-additive operator. Plain negations are therefore called anti-morphic (Van der Wouden 1994).2 Yet, moeten can take scope below a non-anti-additive operator such as weinig (‘few’) and does therefore not resist all negative environments. For an elaborate proof of moeten’s PPI-hood, the reader is referred to Iatridou and Zeijlstra (2013).
2.2.2 Dorven : from polar modal to null
As already mentioned in the introduction, dorven could vary in modal force. This means that dorven surfaced with existential as well as universal modal force - the accessibility relation it corresponded to therefore yielded that either at least one possible world or all possible worlds of the domain of worlds were accessible. However, the default reading seems to be existential (as will be illustrated below). Furthermore, the MNW states that dorven was a negative polarity item (NPI) in (Early) Middle Dutch. Consequently, this verb must always co-occur with a negative element. The sentences below illustrate resulting meanings.
(15) Ganse Health ne neg darftu may.you niit not vermiden. avoid Dune you.neg salt shall dat that niit not vergeten. forget ‘You may not avoid [good] health. You must not forget that.’
¬ >3 (anonymous 1253)
In (15), dorven takes scope below negation. The sentence states that the addressee may not forget that a good health is important. This means that in all the accessible worlds, the addressee does not forget this. The scopal relations are similar in sentence (16), but in (16) dorven can be translated as a universal quantifier. This proposition poses that the man spoken of had no need to complain; there are thus accessible worlds in which the man does not complain.
1
A function f is anti-additive if f (X ∪ Y ) = f (X) ∩ f (Y )
(16) Mar but dat that ne neg darf need hi he clagen complain nit not Die that goede good man. man want because hi he uerschit ends Jn in goeden good pointe moment/position
‘But the good man need not complain, as he ends up in a fortunate position’
¬ >23 (Affligem 1265-70)
Nevertheless, one could also interpret dorven in (16) as an existential modal. The translation ‘the good man may not complain’, seems felicitous in this minimal context as well. This shows that it can often be difficult to distinguish between the two readings. However, there are also occurrences of dorven that only allow for a universal reading. Take for instance (17):
(17) Context: Men are riding out to fight the Swiss, fortunately, the Swiss are found (nearby) in the valley. The author comments:
a. mer but si they en neg dorsten need niet not diep deep int in.the lant country riden ride om around de the hoge high geberchten mountains die that daerinne there.in leggen lay
‘They need not ride deep into the country all around the high mountains in it’
In the example above, it is clear that it was unnecessary for the men to continue riding and looking for the Swiss, since they had found them already. Yet, they were allowed to ride on if they had wanted to - it was just not necessary to do so. Therefore, the example in (17) shows that dorven can get a pure universal reading, although it is a bit rarer than the existential one. Sentences (15)-(16) suggest that dorven had to scope below a plain negation. However, the example in (18) shows that dorven could also grammatically occur under a negative indefinite nember (‘never’). (18) die that knape young.man sprac spoke om around dat that orloge war en neg dorfdi may nember never weder+keren return ‘The young man said that because of the war, he could never return’
¬ > ∃ >3 (anonymous 1260-1280)
Furthermore, other data suggests that dorven was also grammatical in questions, see (19):
(19) Context: The author discusses Jesus and his relation with Maria. a. Nu Now machmen may.man vragen, ask hoe how dorste could Maria Maria Magdalena Magdalena dit this hooft head besalven, put.ointment.on die that een a sonderssche sinner was, was
‘One may ask how Maria Magdalena was allowed put ointment on is had, as she
was a sinner.’ (Dirc van Delf 1404: 428)
On the basis of the data, dorven appears to be a weak NPI - this means that it is allowed to occur in any kind of negative contexts: in strong ones (as in (16)) and in weak ones (as in
(19)) (Van der Wouden 1994). Nevertheless, due to the low number of sources and of course the lack of native speakers, the specific constraints imposed by the NPI-hood of dorven cannot be investigated in more detail. In spite of this, the data remain particularly interesting, as they show that deontic existential modals can in fact be polar. Such items were not attested before; therefore, Iatridou and Zeijlstra (2013) doubt that such items could exist at all. The data presented here strongly suggest that this is actually the case.
According to Van der Wouden (1996) and the WNT, dorven lost ground in the first half of the seventeenth century. Both sources state that this is due to the homophony of dorven and dorren (‘to dare’) in the past tense paradigm. These paradigms are shown below.
Person/number dorven dorren
1sg dorste dorste(n) 2sg dorst(e)s dorst(e)s 3sg dorste/dervede do(e)rs(ten) 1pl dorsten -2pl dorst dorst 3pl dorsten dorsten/dursten
Table 1: Past tense paradigm of dorven and dorren
The paradigms in table 1 appear to be homophonous indeed. Section 4 shows that the corpus study confirms that there was indeed lexical competition between the verbs in table 1. However, it also shows that there was a third verb involved in the competition for the paradigm of dorven.
2.2.3 Behoeven : from lexical verb to polar modal
Compared to the other two modals, the diachrony of hoeven is relatively well studied. This modal grammaticalized out of the lexical verb behoeven (‘to be in need of’), which is still found in Modern Dutch (albeit it is a bit archaic).4 The verb behoeven was never an NPI. Its usage is illustrated in (20). (20) Hi he gaf gave hem him al all dies that hem he.dat behoevede needed
‘He gave him all that he needed’ (anonymous 1340-1360)
However, the modal hoeven is an NPI and must therefore co-occur with a negative element. The examples in (21) show that the usage of hoeven is only felicitous if the negative operators weinig (‘few’), niemand (‘no one’) or niet (‘not’) are present in the same phrase.
(21) a. *(Weinig) few kinderen kids hoeven need te to slapen sleep ‘Few kids need sleep’
f ew >2 b. *(Niemand) no.one hoeft needs te to slapen sleep ‘None needs sleep’
c. Jan Jan hoeft needs *(niet) not te to slapen sleep ‘Jan needs not sleep’ ¬ >2
Van Baal (2013) argues that hoeven was not always in NPI, despite the fact that the verb was already used predominantly in negative clauses from the seventeenth century on. She reports that 79% of its occurrences from 1600 till 1700 are with anti-additive operators. Yet, in this period, some occurrences of hoeven are in positive clauses. An example is given in (22):
(22) Ghij You hoeft need wel prt spijs, food en and drank. drink
‘You need food and drink’ (Spiegel 1614)
Van Baal found sentences like (22) until the nineteenth century. Consequently, she therefore concludes that hoeven was reanalysed as an NPI around this period.
These diachronic data could thus give a clue as to when the change in moeten as well as the loss of dorven may have occurred, provided the changes are related. How the three modals and their changes could have been dependent on the changes in one another is described in the next subsection.
2.3 Diachronic Hypotheses & Predictions
The previous subsection dealt with the three relevant modals of (Middle) Dutch, their behaviour with respect to negation and the changes in this behaviour. It illustrated, first, how moeten changed from a neutral into a positive polar modal; second, that negative polar dorven was mostly an existential (and sometimes a universal) modal and was presumably lost before the rise of hoeven; and finally, third, how hoeven changed from a lexical into a negative polar universal modal. The goal of this subsection is, to draw up concrete hypotheses that combine these changes and thereby constitute an answer to the two research questions asked in section 1. Research question 1 asked why and when moeten became a PPI. Follow up question 2 asked whether this development was related to the changes in dorven and hoeven. It seems very natural that due to one change in the modal landscape, other modals changed as well. It is most likely that modals such as moeten and hoeven were reanalysed, as a learner’s response to the loss of another modal meaning, such as the meanings expressed by dorven. Such changes enable the speakers of a language to express the various meanings that arise from the interaction of modality and negation (Van der Wouden 1996). To disentangle the chain of modal events, the following discusses the meanings involved in the chain and the equivalences between the three modals - before and after their changes. Thereafter, it can be examined what kinds of the lexical gaps the modals might have left after changing and the lexical gaps they might have filled by changing.
modals and their Modern Dutch usages, we see that both moeten and hoeven, combined with a negation, are equivalent to one of the usages of dorven. On the one hand, modern moeten in combination with a negative operator namely expresses that something is necessarily not the case: 2 > ¬. This is logically equivalent to saying that something is not possible: ¬ > 35, which is precisely what dorven could convey. On the other hand, hoeven combined with a negation expresses that something is not necessarily the case: ¬ > 2. This corresponds to the other meaning of dorven. Crucially, there is one further semantic relation: that of moeten and hoeven. As was already mentioned, Middle Dutch moeten combined with a negation is equivalent to hoeven in the same configuration in Modern Dutch - both are universal modals scoping below negation. Note, in addition, that the Middle Dutch usage of moeten is also equivalent to the second meaning of dorven (¬ >2).
Based on the facts presented above, this study proposes the following diachronic pathway from Middle Dutch to Modern Dutch, which involves dorven, moeten and hoeven. It starts with the lexical competition of dorven and non-modal verbs, due to which dorven could not be acquired as a modal verb anymore. The most salient meaning that was lost, was ‘not possible’ (¬ >3), since dorven was mostly used as an existential verb. This lexical gap could be filled by moeten, if it took scope above negation and expressed ‘necessarily not’ (2 > ¬). In this view, the loss of dorven thus caused the reanalysis of moeten. At the same time, when acquiring its new meaning moeten probably lost the ability to scope below negation. Therefore, this change created a lexical gap as well (of the meaning ‘not necessary’ (¬ >2)). However, this gap could be filled by a reanalysis of the lexical verb behoeven; the meaning of this verb was compatible with that of a universal modal (as the English counterpart of hoeven, need, also illustrates). This new modal could scope below negation and thus fill the lexical void. As moeten was already an established modal verb in sentences without negation, hoeven was presumably used mostly in the sentence with a negative element. Therefore, it was most likely reanalysed as an NPI over time.
Note, that this hypothesis goes against the claims by Van der Wouden and Van Baal, who suggest that hoeven filled the lexical gap left by dorven. Yet, on the basis of the data presented in section 2.2.2, we can conclude that this proposal is not compatible with the data. The preferred interpretation of dorven is an existential reading (¬ > 3) and not a universal one (¬ > 2). Therefore, it is rather difficult to argue that a universal modal like hoeven took its place, as hoeven can only account for a very small number of dorven’s occurrences. Moreover, this makes hoeven’s role as a successor of moeten all the more likely. Middle Dutch moeten namely expressed precisely the meaning that hoeven expresses nowadays. The rise of hoeven would thus be unexpected if moeten could still be used in this meaning. In case moeten had already changed into a polar before dorven and hoeven did, the proposal by Van der Wouden seems more presumable. Nonetheless, he would still have to account for the fact that hoeven can only be a partial successor of dorven. This is highly problematic, in view of the empirical evidence.
The present study thus hypothesizes that the scenario sketched above describes what
hap-5The De Morgan’s laws state the following equivalences: ¬ >3 ≡ 2 > ¬ (There is no possibility ≡ necessarily not)
pened to the modal landscape in the transition from Middle to Modern Dutch. This hypothesis clearly explains the PPI-hood of moeten. In addition, the hypothesis appears to fit the data and reported changes best. The table below summarizes the relevant semantic changes (indicated by arrows) and their suggested correlations with one another. In this table, stage I denotes the point of departure, namely, the phase in Middle Dutch in which the modals have not undergone change yet. Stage V describes the situation of Modern Dutch: both moeten and hoeven are polar modals and dorven has been long lost. The transition from stage I to II is set in process by the loss of dorven. During and after stage II, learners of Dutch could not analyse dorven as a modal any longer. Consequently, a lexical gap arose. In stage III, moeten starts to fill this gap and eventually becomes a PPI. As a result, a new lexical gap came into being which is filled by hoeven in stage IV. In stage V, hoeven is reanalysed again, as an NPI.
Stage I Stage II Stage III Stage IV Stage V
NPI dorven loss of dorven ¬ >3
lexical gap: ¬ >3 ≡2 > ¬
Neutral moeten PPI moeten PPI moeten PPI moeten
¬ >2 → 2 > ¬ 2 > ¬ 2 > ¬
lexical gap: ¬ >2 ≡3 > ¬
Behoeven (neutral) (be)hoeven NPI hoeven
(lexical verb) → ¬ >2 ¬ >2
Table 2: Proposed diachronic stages and the corresponding modal meanings and changes
The data gathered by Van Baal (2013) provides a good empirical starting point for the present study. As mentioned above, she poses that hoeven became an NPI around 1800; this is thus when stage V commences. From her study, we also know that hoeven already occurred as a universal modal scoping below negation around 1600 although there were still some occurrences of hoeven in positive contexts (see section 2.2.3). Stage IV must have thus started before the seventeenth century.
Below, concrete hypotheses for each of the modal changes and their correlations with the other changes, as worked out in table 2, are drawn up.
Hypotheses
1 Due to decreasing evidence in the language input, learners of Middle Holland Dutch were unable to analyse dorven as a modal any longer. Consequently, its modal meanings were lost.
2 Moeten became a polar item, due to the lexical gap that arose when the predominant modal meaning of the NPI dorven (¬ >3) was lost.
3 The lexical verb behoeven was reanalysed as a modal, after the reanalysis of moeten. Thus filling the lexical gap left by moeten scoping below negation (¬ >2).
Predictions
1a The modal meanings of dorven occur less often toward the end of the sixteenth century (which is, according to Van Baal’s (2013) corpus study, most likely to have been the starting period of stage IV, in table 2, in which hoeven becomes a modal - the other changes must then thus have already been carried through). 1b In the same period of time, lexical meanings start to dominate the paradigm of
dorven.
2a The changes in dorven and moeten occur in the same period.
2b Dorven originally occurred more often with existential modal force, than with universal modal force.
2c Moeten occurs under the scope of a negative operator less often and start outscoping negation more often. (Presumably, this new semantics is accom-panied by a more expliciet context first.)
3a The changes in moeten and hoeven occur in the same period of time.
3b The first modal occurrences of hoeven are expected to occur in the scope of negative operators almost, but not completely, exclusively.
The next subsection elaborates on the corpus study that was conducted in order to verify the predictions made.
3
Corpus Study
This section deals with the corpus study employed to verify the hypotheses above. Since the positive polarity hood of moeten is central to this study, the next subsection first discusses a very general sample survey that was conducted in order to historically locate the change in moeten in Holland Dutch. Second, it deals with the more elaborate diachronic study of moeten, dorven and hoeven in the time span fixed for the change in moeten by the general sample study.
3.1 Sample survey
As discussed in the previous section, the rise of polar hoeven and moeten is expected to be related. Therefore, the prediction is that moeten became polar at the lastest before the nine-teenth century (Van Baal 2013). In order to verify this, occurrences of moeten were gathered from 1500 till 1800 (all texts used are listed in Appendix 1), from old Holland Dutch texts available at the Digitale Bibliotheek Nederlandse Letteren (DBNL). In each century, about 75 occurrences from approximately three periods in time (spread out as much as possible) were collected, resulting in a total of 219 occurrences (see Appendix 2). For each of the sentences in which moeten occurred in the same phrase as an anti-additive or anti-morphic negative opera-tor, it was examined whether the modal outscoped negation or whether the negation outscoped the modal. Mostly, this could be inferred from the context (as shown in section 2.2.1). For some occurrences, however, it was difficult to get one distinct reading - these were kept apart. The results are shown in the table below. As the table shows, for the sixteenth century, only occurrences after 1559 are taken into account.6
Year ¬ >2 2 > ¬ unclear ¬&2 no ¬&2 Total 1559-65 2 0 2 51 55 1591 0 0 1 18 19 1619-24 0 0 0 30 30 1649-61 0 1 0 27 28 1690 0 2 0 18 20 1713 0 2 0 17 19 1750 0 1 0 19 20 1777-91 1 1 0 36 38 Total 2 7 3 216 219
Table 3: Results of the first sample survey of the usage(s) of moeten
The table lists the investigated time spans in the first column. The second column shows the occurrences where moeten is outscoped by negation. The third column shows occurrences in which moeten outscopes negation. Occurrences of moeten and a negation in which the scopal relations are unclear are tallied in the fourth column. Finally, the sixth column yields the occurrences of moeten without a negative operator and the seventh column shows the total number of occurrences found.
On the basis of the data given, it appears very likely that a change occurred before 1600, as moeten seems to scope over negation increasingly more often in the later centuries. In the first two rows, the results for the sixteenth century show that moeten co-occurs with a negation five times, out of which it is outscoped by negation twice. In the seventeenth century three co-occurrences are found: in all three cases moeten outscopes negation. In the final century, moeten was found with negation five times: four times taking scope over negation four and once under. The latter occurrence is shown in (23).
(23) Context: Neeltje has agreed to marry Krelis, who may or may not have died at war shortly after. She is then proposed to by Teeuwis, to whom she feels obliged to say yes, but she still loves Krelis. Her father then utters:
a. Neen, no neen no Neeltje! Neeltje gij you moet must Teeuwis Teeuws niet not trouwen. marry
‘No, no Neeltje! You do not have to marry Teeuwis’ (Brender ´a Brandis 1791)
Interestingly, (23) appears to be an occurrence of contrastive negation (since the alternative to marrying Teeuwis namely, marrying Krelis is in the air). This is one of the contexts in which PPI’s are allowed to scope under negation, according to Iatridou and Zeijlstra (2013: 10).
A second sample survey into the sixteenth and seventeenth century reduplicated the results found for the first pre study. The texts used here are listed in Appendix 3. The results for the second study are shown in the next table.7 In this sample study, only occurrences of moeten with negation are taken into account (for all occurrences see Appendix 4).
Table 4 clearly confirms the conclusions drawn from the first study: something in the usage of
Year ¬ >2 2 > ¬ unclear Total
1508-25 4 5 2 11
1580-1600 1 3 2 6
1600-45 0 5 4 9
1645-89 0 18 2 20
Table 4: Results from the second sample survey
moeten has changed around the end of the sixteenth century and the beginning of the seven-teenth. No occurrences of moeten outscoping negation seem to be found after the beginning of the seventeenth century. However, moeten appears to outscope negation as early as of 1508 on. Yet, all five occurrences showing this scope-relation in the first time span are from one author (named Govert van Ghemen). This raises the question whether this is a real break from the pattern or an idiosyncracy. The next section shows it is most likely the latter. The differ-ences between the number of occurrdiffer-ences of negation outscoping moeten in the two centuries is significant according to Fisher’s exact test in R (p = 0.004514).
To gain more insight into the change of moeten and its relation to the changes in other modals, more texts of Middle Holland Dutch from the fifteenth and sixteenth century were examined. The next subsection discusses the set up corpus of this periods and the results found for moeten, hoeven and dorven.
3.2 Corpus Results
For this corpus the texts used for the sample surveys above were complemented with few more texts from 1508 till 1548 (from the DBNL). For the second period of the century, the cor-pus mainly consisted of texts by the Amsterdam humanist Coornhert and Lauris Janszoon, a playwright from Haarlem. The texts by both authors are available on the DBNL as well. In addition, the corpus was complemented by official papers from municipals in Holland from 1450 till 1589. These texts, from the corpus Compilatiecorpus Historisch Nederlands 1.0 (CHN), are hosted by the Meertens Institute (at www.diachronie.nl). A full list of all texts used is found in appendix 5.
Of course, the relevant criteria, methods and search terms differed for each of the modals. These are discussed separately in the next subsections.
3.2.1 Results for moeten
As discussed, for moeten the scopal relations of the modal and a clausemate negative operator are relevant. This could be investigated by analysing the relevant phrase in its context, as shown in section 2.2.1. Two types of negative operators were taken into account, anti-morphic and anti-additive operators. These are the operators that present day PPI moeten cannot take scope under (see section 2.2.1). The relevant search terms were collected via the WNT, which lists the full paradigm of Middle Dutch moeten (shown in Appendix 6).
In this sub study, interrogatives were not taken into account for two reasons. First, following Krifka (2012), in some questions with a negation (so-called negated polarity questions) the
contain a modal, there would thus be no interaction between the negation and the modal despite appearances. Second, it is not entirely clear what effect questions have on PPI’s, since they constitute environments that can license weak NPI’s. Another type of sentences that was excluded was similar to the one in (24):
(24) ... dat that daer there niement no.one in in wonen live en neg mach, may hi he en neg moste must.pst doef deaf warden become ‘that no one may live there, or he would become deaf’ (Dirc van Delf 1404: 41)
As the glossing of (24) indicates, moste is preceded by the neg-particle ‘en’. One could interpret it thus as a phrase in which negation and moeten are in some scopal relation. However, as the translation points out, this is not the case. The phrase containing moste gives a reason for the negative order uttered in the first part of the sentence. In order to make the meaning of (24) coherent, we must assume that the second negation only surfaces to indicate the relation with the preceding negative statement.
The results from the corpus study are listed in the table below. In this table, results for interaction of moeten both kinds of negative operators are collapsed. In total 1546 occurrences of moeten were examined (see Appendices 7-21). However, 65 were questions and 15 similar to (24). The final count was thus 1466. The same rows and columns are employed as in the tables above. The numbers in each column indicate how many occurrences were tallied in that period of time. If that number is followed up by a number n and a question mark, this indicates that there were n occurrences which seemed to be ordered most appropriately in the column they are in, however, both the second reader and me were not as sure about these occurrences as about the other ones. The percentages are relative to the entire number of occurrences found in the corresponding period.
Year ¬ >2 % 2 > ¬ % Total2 & ¬ % Total
1404 4 4,1% 1 1% 68 6,1% 98 1450-90 8 (2?) 57,1% 0 - 8 57,1% 14 1511-17 6 3,6% 4 2,5% 10 6% 166 1525-39 7 (1?) 19,4% 0 - 7 19,4% 36 1548 0 - 0 - 0 - 48 1559-64 1 0,4% 10 (3?) 3,7% 11 4,1% 271 1575-79 1 0,2% 11 (1? ) 2,1% 12 2,3% 516 1585-92 0 - 7 3% 7 3% 236 1602 0 - 0 - 0 - 81 Total 27 33 61 1466
Table 5: Results of the corpus study into moeten
Looking at table 5, several observations must be made. First and foremost, we see that after 1579, no occurrences of moeten scoping under negation are found. Second, before 1559, a small number occurrences of moeten outscoping negation are tallied; from 1559 on this number rises. The four occurrences found of moeten outscoping negation in 1511-17 are from one author. Third, the number of times moeten then co-occurs with a negation seems to decrease. Until
1548, the between 5,4% and 57,1% of the occurrences of moeten are in a negative environment. Looking at the timespans after 1548, we see the percentages getting lower: from 4,1% to 2,3%. The percentages in the timespans 1404 and 1511-17 are also relatively low. However, Fisher’s exact test in R shows the difference in the number of occurrences of moeten with negation before 1559 and after it is still significant (p = 2.53e-06).
Table 6 is a more extensive version of table 5: it shows the results of moeten and the different negative environments separately. The results for anti-additive operators (like nobody) and anti-morphic operators (like not ) are thus kept apart. This table shows that, only taking into account the occurrences with an anti-additive operator, the border is of the change is 1564 (and not 1575).
Year am > 2 % 2 > am % aa > 2 % 2 > aa % Total 2 & ¬ % Total
1404 2 2% 1 1% 2 2% 0 - 6 6,1% 98 1450-90 2 (1?) 14,3% 0 - 6 (1?) 42,9% 0 - 8 57,1% 14 1511-17 1 0,6% 1 0,6% 5 3% 3 1,8% 10 6% 166 1525-39 1 2,8% 0 - 6 (1?) 16,7% 0 - 7 19,4% 36 1548 0 - 0 - 0 - 0 - 0 - 48 1559-64 0 - 6 (2?) 2,2% 1 0,4% 4 (1?) 1,5% 11 4,1% 271 1575-79 1 0,2% 7 1,4% 0 - 4 (1?) 0,8% 12 2,3% 516 1585-92 0 - 4 1,7% 0 - 3 1,3% 7 3% 236 1602 0 - 0 - 0 - 0 - 0 - 81 Total 7 19 20 14 61 1466
Table 6: Results of the corpus study into moeten, with anti-morpic and anti-additive operators in separate columns
In order to see if the difference in the number of occurrences of moeten with a negation is also found between languages with a PPI modal moeten and a neutral moeten, a small sample study was conducted, for present day English, Dutch and German, as well as Middle Dutch (from 1300-1400). For English the British National Corpus was used, for Dutch the Corpus Gesproken Nederlands and for German the Digitale W¨orterbuch der deutschen Sprache des 20. Jahrhunderts. Middle Dutch was examined with the CD-rom Middelnederlands. For each language, 100 sentences containing must, moeten or m¨ussen were investigated. In order to keep the study feasible, no scopal relation were examined. It yielded the following results:
Language 2 & ¬ Total 2
English 3 100
Dutch 3 100
German 10 100
Middle Dutch 21 100
Table 7: Occurrences of moeten cognates with negation
Table 7 clearly shows that the universal modal occurs more often with a negation in Middle Dutch and German. The difference between Middle Dutch and present day Dutch/English are significant, as shown by Fisher’s exact test in R (p = 0.0001149). The difference between German and Dutch is strictly speaking not significant (p = 0.08179). The difference between German and Middle Dutch is significant (p = 0.04945).
3.2.2 Results for dorven
Just like for moeten, the entire paradigm of dorven was investigated in the corpus. The paradigm is listed in Appendix 26. For dorven, it was examined whether its occurrences were, first, modal or lexical; and, second, how the modal occurrences behaved with respect to negation. The results for both criteria are discussed in that order. Thereafter, the lexical occurrences of dorven are discussed in more detail.
The corpus yielded 257 occurrences of dorven (see Appendices 27-32). The results are shown in table 8. This table only distinguishes between modal meanings (in the second column) and lexical meanings (in the third column). The occurrences from which the meaning was unclear are kept apart under this category in the fourth column.
Year Modal meaning % Lexical meaning % Unclear % Total
1404 3 33,3% 3 33,3% 3 33,3% 9 15179 34 58,6% 16 27,6% 8 13,8% 58 1525 6 50% 4 33,3% 2 16,7% 12 1548 0 0% 0 0% 0 0% 0 1559-65 9 21,4% 26 61,9% 7 16,7% 42 1575-79 4 5,2% 70 90,9% 3 3,9% 77 1585-92 3 11,1% 22 81,5% 2 7,4% 27 1602-14 2 6,25% 24 75% 6 18,75% 32 Total 61 161 27 257
Table 8: Results from study into dorven: lexical vs. modal meanings
Table 8 shows that the modal usages of dorven dominate until 1548. Yet, the lexical meanings seem to be present from the first texts on. From 1559 on, the lexical usages seem to become more prominent. In the time span 1559-1564, 61,9% of the occurrences is non-modal. The percentage remains over 75% in the next periods. The difference between the number of non-modal occurrences in the periods before and after 1559 (not taking into account the unclear cases) is highly significant according to Fisher’s exact test (p = 9.853e-16).
In addition, table 9 zooms in on the modal occurrences found. This table distinguishes between the different forces found in the modal occurrences. In the first column, the publication years of the texts are listed again. In the second column, the existential occurrences are tallied. In the third column, this is done for the universal occurrences. The fourth column lists the total. This overview confirms what was already said in section 2.2.2, namely that the most salient modal meaning was the existential one. In total 61 occurrences were found, out of 54 are existential.
Year 3 % 2 % Total 1404 2 66,7% 1 33,3% 3 1517 34 100% 0 - 34 1525 6 100% 0 - 6 1548 0 - 0 - 0 1559-65 4 44,4% 5 55,6% 9 1575-79 3 75% 1 25% 4 1585-92 3 100% 0 - 3 1602-14 2 100% 0 - 2 Total 54 7 61
Table 9: Results of corpus study into dorven: the modal occurrences
Furthermore, it was investigated whether or not the modal occurrences appeared in the scope of a negative operator, to confirm whether dorven was an existential modal NPI indeed. Table 10 shows that almost all of the 54 occurrences were in some kind of negative environment. No less than 50 were in the scope of a negative operator. Out of these 50 occurrences, 39 were in an anti-morphic context and thus involved a plain negation, such as not. Only two occurrences of dorven in questions were found. Furthermore, there are two occurrences of dorven without a negative operator - these are both found after 1575.
Year AM>3 % AA>3 % No ¬ % Total
1404 1 50% 0 - 1i 50% 2 1517 25 73,5% 9 26,5% 0 - 34 1525 6 100% 0 - 0 - 6 1559-65 2 50% 1 25% 1i 25% 4 1575-79 2 66,7% 0 - 1 33,3% 3 1585-92 2 66,7% 0 - 1 33,3% 3 1602-14 1 50% 1 50% 0 - 2 Total 39 11 4 54
Table 10: Occurrences of existential dorven in a negative environment (‘i’ indicates interrogative)
The seven universal occurrences of dorven were all in a negative environment. Three co-occurred with a plain negation. The remaining four were in the scope of an anti-additive operator.
A remaining question concerning dorven then is, which lexical occurrences took over the verbal paradigm. As listed in WNT, the verb dorven was expected to be in lexical competition with durven (‘to dare’). Occurrences with this meaning were thus set apart. (25) provides an example of such an occurrence:
(25) waer where doer through Joachim Joachim hem refl seer very schaemde was.embarrassed ende and en neg durste dare mit with sijn his ghebuyren happenings niet neg thuus home comen come
‘Because of that, Joachim was very embarrassed and did not dare to come home with
his happenings’ (Arent Willemsz 1525: 178)
Furthermore, a reasonably large amount of occurrences was found in the corpus with the lexical meaning ‘to lack’, ‘to disclaim’. These are from the verb derven, which was also homophonous with dorven. The next sentence provides an example of this lexical verb:
(26) Zonde Sin doet does godlycke divine blydschappe happiness derven lack ende and droefheyd sadness hebben. have
‘Sin makes [one] lack happiness and makes [one] sad’ (Coornhert 1585: 456)
Table 11 shows the different usages of dorven once again. Now, the emphasis lies on the different lexical usages the verb could have. The modal occurrences are taken together in the second column of this table (under ‘3/2’). The third column, then, tallies the lexical occurrences that bear the meaning ‘to dare’. Some occurrences were borderline cases between the existential modal and lexical ‘to dare’ meaning. These are tallied in the fourth column. The fifth column lists the occurrences with the second lexical meaning ‘to lack’/‘to disclaim’. In the sixth column, occurrences with an unclear meaning were collected - these were not reducible to any of the meanings mentioned above. Finally, the seventh column lists the total.
Year 3/2 % dare % 3/dare % lack % ? % Total
1404 3 33,3% 2 25% 3 37,5% 1 12,5% 0 - 9 1517 34 58,6% 16 27,6% 8 13,8% 0 - 0 - 58 1525 6 50% 3 23,1% 2 15,4% 1 7,7% 0 - 12 1548 0 - 0 - 0 - 0 - 0 - 0 1559-65 9 21,4% 20 47,6% 7 16,7% 6 14,3% 0 - 42 1575-79 4 5,2% 7 9,1% 2 2,6% 63 81,8% 1 1,3% 77 1585-92 3 11,1% 0 - 1 3,7% 22 81,5% 1 3,7% 27 1602-14 2 6,25% 17 53,1% 4 12,5% 7 21,9% 2 6,25% 32 Total 61 65 27 100 4 257
Table 11: Results from study into dorven: the different lexical meanings
From table 11, it becomes clear that dorren (‘to dare’) was homophonous with the modal dorven from the earliest texts on. The number of its occurrences remain relatively constant. The same cannot be said of the lexical verb derven (‘to lack’/‘to disclaim’). This verb was found twice before 1559, but 95 times after this year. It thus seems to win ground after 1559. This table clearly shows that the rise of the latter verb correlates with the decreased usage of modal dorven.
The results show that the change in dorven occurred in roughly the same period as the change in moeten. Both modals undergo change in the second half of the sixteenth century.
3.2.3 Results for hoeven
For the verb hoeven, only verbal occurrences of either behoeven or hoeven were taken into account. The criteria relevant for the present study were whether the usages were modal and if a negation was involved. As for the first criterion, the standard for using a verb as a modal verb, was defined as taking a verb phrase as a complement instead of a noun phrase. The different complements are illustrated in (27-a) and (27-b) respectively.
(27) a. Jan Jan behoeft needs [V P te to slapen] sleep
Concerning the second criterion, the same negative operators as taken into account for moeten were deemed relevant, as the assumption is that hoeven is the successor of moeten.
The corpus yielded 202 occurrences of hoeven (see appendices 22-25). Several texts from after the year 1602, were added for this sub study to raise to total amount of occurrences for the final period. The final count was therefore 222 occurrences. The results are be discussed in view of the criteria mentioned above. First, the rising modal usage of hoeven is dealt with; second, the interaction of this modal property with negation. The table below shows the results for the first criterion. The first column shows the publication year of the texts used. The second columns tallies the modal occurrences. The third column does that for the lexical usages. The fourth column lists the total.
Year Modal usage % Lexical usage % Total
1404 0 0% 12 100% 12 1450-1516 0 0% 0 0% 0 1517-43 0 0% 20 0% 20 1559-65 6 28,6% 15 71,4% 21 1575-79 9 12,7% 62 87,3% 71 1585-92 8 10,3% 70 87,6% 78 1602-14 10 50% 10 50% 20 Total 33 189 222
Table 12: Results of corpus study into (be)hoeven: lexical vs. modal occurrences
Table 12 shows no modal occurrences of behoeven were found before 1559. From that year on, different percentages of modal usages are found - varying from 10,3% to 50%. The increase from 1559 on is found to be significant by Fisher’s exact test in R (p = 0.005974).
In table 13 the criteria relevant for hoeven are taken together. This table thus shows how often the modal occurrences co-occur with a negative operator in comparison to the lexical occurrences. Like in table 12, the first column shows the publication year of the texts and the second column the modal usages of hoeven. The third column indicates how many of those modal usages were accompanied by a negative operator (anti-morphic or anti-additive) - this column is thus a subset of the second column. Thereafter, the fourth column shows how many occurrences of hoeven were in total found with a negative operator. This column combines the lexical and modal occurrences. Finally, the total number of occurrences of hoeven is listed.
Year Modal usage % ¬ > Modal usage % Total hoeven & ¬ % Total
1404 0 0% 0 0% 4 33,3% 12 1450-1516 0 0% 0 0% 0 0% 0 1517-43 0 0% 0 0% 4 25% 20 1559-65 6 28,6% 5 23,8% 12 57,1% 21 1575-79 9 12,7% 7 9,9% 15 21,1% 71 1585-92 8 10,3% 6 7,7% 17 21,8% 78 1602-14 10 50% 8 40% 8 40% 20 Total 33 26 72 222
Table 13: Results of corpus study into (be)hoeven
Interestingly, this table shows that most modal occurrences seem to be accompanied by a nega-tive operator. In total, 33 modal occurrences were found, of which 26 in a neganega-tive environment
usages and non-negative modal usages compared to the negative and non-negative non-modal usages is highly significant (p = 6.786e-10). This increase of the modal usage of hoeven under a negative operator thus strongly correlates with the change found in moeten, as the latter seemed to lose the ability to scope under negation in the same period of time.
The next section further discusses the results with reference to the questions asked in the introduction and the relevant literature.
4
Discussion
The results obtained in the corpus study are discussed in this section. First, this is done accordingly to the research questions drawn up in section one. That means that first the rise of polar moeten is discussed. Second, the section turns to the correlation between, on the one hand, the rise of polar moeten and the loss of dorven and, on the other hand, the rise of polar moeten and the rise of modal hoeven. The second part of this section then discusses what caused the modal landslide that consisted of the changes elaborated on above. It furthermore elaborates on the cross-linguistic differences and implications of the present study, especially with respect to the modal paradigms in English and German.
4.1 The Rise of Polar moeten
The first research question asked why and when moeten changed from a neutral modal into a positive polar modal. The results of moeten in the pre studies (in section 3.1) and the final results (in table 5) clearly show that the rise of polar moeten dates back to in the mid-sixteenth century. Table 5 shows that 81% of the occurrences of moeten with a negative operator from 1404 till 1559 are with the neutral scope relation. In the period from 1559 till 1602, out of all occurrences of moeten with a negation 88,9% is positive polar. The fact that four out of five non-neutral occurrences before 1548 are from one author, makes it difficult to say if the first decades of the sixteenth century were a run-up to the actual change or not. It might be that this author indeed showed normal behaviour and the change started in around 1510 to 1525. Yet, not taking into account this one writer, we might say it is possible that it started around 1539 at the earliest (as no data was found between 1539 and 1559). The obtaining of PPI-hood by moeten then seems to have been completed in at most forty years, as no neutral occurrences of moeten are found after 1579. More data from the period 1539 to 1559 should point out whether the PPI-occurrences of moeten in the period before 1525 consists of idiosyncratic behaviour of some author or if this way of uttering moeten was already relatively normal in that period. However, in both scenarios, the change came about relatively quick.
The figure below shows the fall of neutral moeten (in red) and the rise of positive polar moeten (in green) in one graph. The coordinates used are the percentages gathered in section 3, for each of the time spans. The reader should thus note that only the coordinates in the plot provide valid information. The lines drawn between dots are just tools for the interpretation.
1420 1440 1460 1480 1500 1520 1540 1560 1580 1600 0 20 40 60 Year P ercen tage ¬ >moeten moeten> ¬
Figure 3: Occurrences of neutral and PPI moeten (in percentages)
In the results section it was already pointed out that after the change, the occurrences of moeten with a negation (irrespective of scope) decrease significantly. The graph shows this even clearer. The question why that is the case is difficult the answer, but this phenomenon is not specific to universal modals. Also in other domains of quantification, speakers tend to avoid constructions in which universal quantifiers outscope negation. Consider the Dutch temporal adverb altijd (‘always’) in (28). In (28-a), altijd takes scope over negation; in (28-b) it scopes below negation - the former is highly marked whilst the latter is fine.
(28) a. ??Jan Jan gaat goes altijd always niet neg voor before twaalf twelve uur hour slapen sleep Intended: ‘Jan never sleeps before twelve o’clock’ b. Jan Jan gaat goes niet neg altijd always voor before twaalf twelve uur hour slapen sleep ‘Jan doesn’t always go to bed before twelve o’clock’
It might not seem illogical to hypothesize that the cognitive task to assign a negative value to every moment t in the temporal domain is more difficult than to assign one negative value to some moment. However, I am not familiar with research on the computability of different scopal relations and will leave this matter for others. Yet, this does raise the question why learners would reanalyse moeten as a PPI, if its usage with a negation is not preferable at all. This question will be touched upon later in the discussion, when dealing with the correlation between the loss of dorven and the rise of polar moeten. Then, naturally, the question why moeten became polar will also be touched upon.
4.2 The Correlation Between dorven, moeten and hoeven
The second research question asked about the correlation between the changes in moeten, hoeven and dorven. Section 2.3 hypothesized that moeten started to outscope negation at the same time, dorven’s existential modal meaning was lost (hypothesis 2). Furthermore, it posed that there is a correlation between the loss moeten’s ability to scope below negation and the rise of the modal hoeven (hypothesis 3). First and foremost, the data presented in 3 showed that there
between 1539 and 1559. This subsection deals with the proposed correlation between dorven and moeten first and second with the correlation between moeten and hoeven.
As elaborated on above, moeten started to take scope above negation after 1559. Section 3.2.2 showed that dorven’s paradigm less often bears existential modal meanings after 1565; the lexical meanings start to win ground after 1559. This loss of the modal and increase of the lexical meanings was predicted in section 2.3. Moreover, we see that there seems to be some kind of correlation between the loss of dorven and the rise of polar moeten, as both changes occur in the same decades. The figure below shows the change in moeten, as also shown in figure 3 above, and additionally illustrates the loss of dorven’s existential modal meaning.
1420 1440 1460 1480 1500 1520 1540 1560 1580 1600 0 20 40 60 Year P ercen tage ¬ >moeten moeten> ¬ ¬ >3 dorven
Figure 4: Results of moeten and dorven
Yet, the data in figure 4 is somewhat puzzling. In fact, the modal usage of dorven appears to decrease. At the same time, moeten becomes polar. However, modal dorven is not entirely lost by the end of the sixteenth century, although moeten already could express the same meaning, when combined with a negation. The question thus is, whether polar moeten really arose due to the loss of dorven, as it is hardly used with a negation after its change. An argument in favour of this stand point would be the fact that both dorven below negation (¬ > 3) and moeten above negation (2 > ¬) are logically equivalent and both make strong semantic claims (Hoeksema 1998, Jackson 1995). The former claims that states that there is no such world in which property P holds. The latter says that in all of the accessible worlds, it property P does not hold. Both claims are falsified if one counterexample is found (which would consist of a world in which P holds). This makes it likely that the loss of dorven and the emerging polarity-hood of moeten are somehow related, as it enabled speakers to make such strong claims. However, the behaviour of both modals after 1559 does raise the question if there was not another modal player on the field that would complete the current overview. It is for instance possible that moeten was not the best candidate after all to express2 > ¬ (or for that matter ¬ > 3), since we seem to dislike universal quantifiers outscoping negation. For that reason, speakers of Dutch might have hung on to dorven for a while - putting up with the fact that it might need some extra context to distinguish between the different meanings. Then, as soon as
A modal expressing a deontic and existential modal meaning (¬ >3) in present day Dutch is mogen. Strikingly, only the fourth WNT entry on this Middle Dutch verb lists a modal meaning. This meaning is circumstantial and the modal force is existential (so it could for instance express that something had ‘a physical possibility’). Thereafter it says mogen might also have a ethical (thus: deontic) modal meaning (‘to have a moral possibility’). Then, in the seventh entry, mogen is argued to be somewhat similar to several other modals. Crucially, these are dorven, dorren and moeten. One could thus stipulate that the deontic meaning of mogen was not quite available yet when dorven was put under pressure, whereas moeten was already used deontically. If this is right, the question should be asked, why moeten could adapt to this situation, if mogen could not. One could also hypothesize that mogen was a PPI as well and could therefore not take over dorven’s meaning, as it could not scope under negation. If so, it should have first lost its polar properties before it was able to express ¬ > 3. I do not wish to make any claims concerning mogen and the crucial role it might have played in the field of changing modals. Yet, I do think that the data given on this modal in the WNT is intriguing and strongly suggests that in order to obtain a more complete view, this modal should have been taken into account in the present study as well. Unfortunately, this was not possible within its time span. For now, the only conclusion that can be drawn from the data on moeten and dorven, is that their changes seem to be related and most likely, the loss of dorven triggered the reanalysis of moeten - yet, it is unclear if moeten is the only or most suitable successor to dorven.
Turning to the third hypothesis, on the correlation of moeten and hoeven, the results section showed that the correlation between the rise of polar moeten and modal hoeven was clearer than the correlation between the former and dorven, because the data on both of these modals was more unambiguous. Section 3.2.3 sketched a picture in which hoeven clearly rose after 1559 -the period in which moeten starts to scope over negation. The figure below combines results from both corpus studies in one graph. It illustrates the changes in moeten as above and also shows the modal usages of hoeven, in which it scopes below an anti-additive or anti-morphic operator. 1420 1440 1460 1480 1500 1520 1540 1560 1580 1600 0 20 40 60 Year P e rc en tage ¬ >2 hoeven ¬ >moeten moeten> ¬
Figure 5: Moeten scoping below and above negation and modal hoeven scoping under negation
of moeten scoping below negation are found. From the year 1560 on, already more than 20% of the usages of hoeven is modal and in a negative context. Section 3.2.3 showed that relatively few modal usages are in positive contexts. This is a strong suggestion that the first usages of the hoeven in such contexts, were uttered to fill the lexical gap left by the change in moeten. Had the first modal usages of hoeven not been in such a overwhelming number of negative contexts, one could argue that hoeven gave rise to the change in moeten. However, based on the current data, that does not seem tenable. It thus appears to be the case that hoeven acquired modal properties to be able to express the meaning not necessary (¬ >2). Unfortunately, data fails from the crucial period again. More information on the years between 1539 and 1559 could strengthen the claim posed here.
In sum, the results unequivocally illustrate that the changes dealt with occur in the same time span: at the end of the sixteenth century, modal dorven appears to be past its prime and moeten and hoeven have started to show distinct behaviour with respect to negation. The correlation between the rise of polar moeten and modal hoeven is very clear. This cannot be said of the correlation between dorven and moeten. Figure 6 shows the three developments in one graph. 1420 1440 1460 1480 1500 1520 1540 1560 1580 1600 0 20 40 60 Year P e rc en tage ¬ >2 hoeven ¬ >moeten moeten> ¬ ¬ >3 dorven
Figure 6: Results of the three sub studies in one graph
On the basis of the data, it is clear that the three developments are not independent, but all part of a landslide in the modal paradigm of Middle Holland Dutch. The scattered data on dorven and the little use of PPI moeten with a negation strongly suggests that the picture sketched in figure 6 does not provide a complete overview yet. Perhaps more modals were involved (such as for instance mogen, as suggested above). Additionally, the number of occurrences of modals found with a negation were generally rather low - especially concerning moeten. More solid results would be gained if more data could be garthered, especially from the period 1539 to 1559. Unfortunately, it was not possible for the present study to gather and examine more data. However, from the picture sketched above we can still conclude that moeten became a PPI due to this modal landslide and that this also resulted in the rise of modal hoeven.
4.3 What caused the modal landslide?
In order to come to know how this modal landslide came about, we must look into the lexical competitors of dorven. As mentioned, Van der Wouden (1996) and the WNT suggested that the competition arose due to the homophony of dorven and dorren (‘to dare’). However, the table in 3.2.2 suggests there is an add-on. Dorven namely seemed to be homophonous with dorren from the date of the earliest text of the corpus study on. Although this might have eventually led to a decrease in and the final loss of dorven, the corpus study shows there was another competitor on the lexical market, namely derven (‘to lack’, ‘disclaim’). From 1575 on, hits of the verbal paradigm of dorven, significantly more often seem to actually be occurrences of this lexical verb. The data strongly suggests that the new lexical competitor of dorven was the last drop the modal could bear.
Looking further into the etymology of derven, the Etymologiebank (Van der Sijs 2010) does not point at any specific origin of this word, as it has ancestors in Old Norse, Old High Ger-man and Old English (a.o.) (see http://www.etymologiebank.nl/trefwoord/derven). Yet, in-terestingly, the WNT provides 30 examples; no less than 20 are from Flemish authors and/or publishers, 6 are from the southern parts of the Netherlands (Brabant and Limburg) and the resulting 4 are from Germany (1) or Groningen (2) or unknown location - yet, none with a known location is from North or South Holland. This strongly suggest that derven was of southern origin and entered the Holland Dutch dialect due to language contact. Unfortunately, not much is known of language contact in the early sixteenth century. Bagwell et al. (2008) do state that for instance the population of Amsterdam grew from 14.000 till 30.000 in the first half of the sixteenth century, and from 30.000 till 65.000 in the second half (Bagwell et al. 2008: 8). These numbers become even more impressive if one takes into account the fact that the mortality rates were fairly high - about 12.000 natives from Amsterdam died between 1550 and 1600. Data for the first fifty years of this century are not given. And unfortunately, the origins of the immigrants in the early sixteenth century do not appear to be well described either. It is thus unclear which group of immigrants may have affected the native speakers of Holland Dutch, such that the verb derven was acquired by the latter group. Nevertheless, the idea that a language contact situation was at the root of the change, is appealing as it would explain the sudden march of a lexical verb like derven.
Interestingly, this effect of language contact might fit into the cross-linguistic pattern as well. In the literature, it is often argued that of the three most spoken West German languages, Dutch is situated between English and German (Van Haeringen 1956, H¨unig et al. 2005, Weerman 2007, H¨unig 2010). For instance, looking at the verb paradigms, one can see that German is the richest language and English the poorest, since English has 2 verbal endings, Dutch 3 and German no less than 5 (Weerman 2007). Weerman (2007) argues that the pattern also holds for the phenomenon of arbitrary reference - this linguistic feature is still in use in German, but has been long lost in English and is decreasing in Dutch (ibidem: 32-33). Weerman ascribes this to language contact: English and Dutch suffered from more language contact than German, and may therefore have undergone more changes (ibidem: 19). One can wonder if the pattern found for the universal modal verbs also fits within this pattern. Nothing conclusive can be said
