On Koopman's Generalisation. Remarks on verb clusters in Old Frisian and Old English
In this article I would like to investigate the validity of what I will refer to as Koopman’s Generalisation for Old Frisian, a generalisation saying that head-final verb clusters are not broken up. I will categorise verb order in the verbal cluster with the help of numbers. To illustrate, consider the following example sentences from Modern Dutch and Modern Frisian:
(1a) Modern Dutch
omdat ze de volgorde wilden-1 vastleggen-2 verb order: 12
because they the order wanted-1 fix-2
‘because they wanted to fix the order’
(1b) Modern Frisian
om’t se de folchoarder fêstlizze-2 woenen-1 verb order: 21
because they the order fix-2 wanted-1
‘because they wanted to fix the order’
The numbers express selection restrictions. “Willen” selects “vastleggen”, and not vice versa, hence “willen” is assigned a number that is exactly 1 less than “vastleggen”. In this way, we can classify verb orders in the verb phrase. See the following examples of verb phrases consisting of three verbs:
(2a) Modern Dutch
wie weet waarom Huntelaar niet geselecteerd-3 mocht-1 worden-2 order: 312
who knows why Huntelaar not selected might be
‘who knows why Huntelaar was not allowed to be selected’
(2b) Modern Frisian
wa wit wêrom’t Huntelaar net selektearre-3 wurde-2 mocht-1 order: 321
who knows why Huntelaar not selected might be
Thus the numbers reflect selection (and c-command) relations. For the sake of simplicity, I will restrict myself in this article to verb clusters of which the matrix verb is a perfect participle.
If there are three verbs, then logically speaking six orders are possible. I will illustrate this with the example (2a) from Modern Dutch:
3 12 wie weet waarom Huntelaar niet geselecteerd-3 mocht-1 worden-2
3 21 wie weet waarom Huntelaar niet geselecteerd-3 worden-2 mocht-1 *
2 13 wie weet waarom Huntelaar niet worden-2 mocht-1 geselecteerd-3 *
2 31 wie weet waarom Huntelaar niet worden-2 geselecteerd-3 mocht-1 *
1 23 wie weet waarom Huntelaar niet mocht-1 worden-2 geselecteerd-3
1 32 wie weet waarom Huntelaar niet mocht-1 geselecteerd-3 worden-2
Of the six possible orders, only 312 and 123 are fully acceptable in Modern Dutch.
The orders 321, 213 and 231 are unacceptable, as indicated by the asterisks following the sentences. In contrast, the only order to be acceptable in Modern Frisian is 321. This much being said, we are now in a position to investigate verb order in the verb phrase in Old English and Old Frisian.
2. Koopman’s Generalisation introduced and refined
Koopman (1990: 38-74) investigated the word order of the verbs in the verb phrase in Old English on the basis of the concordance of Healey & Venezky (1980); on Old English syntax, see Mitchell (1985), Fischer, Van Kemenade, Koopman & Van der Wurff (2000). All in all, clusters of three verbs turned out to have one of the following three formats in Old English:
(3a) V1 = modal, V2 = have, V3 = Past Participle (e.g. ‘wolde habban genumen’)
(3b) V1 = modal, V2 = be, V3 = Present Participle (e.g. ‘wolde beon nimende’)
(3c) V1 = modal, V2 = beon, weoardhan, wesan, V3 = Past Participle
(e.g. ‘wolde beon genumen’).
Koopman limits himself to verb clusters of the (3c)-type, though he claims that the other two types do not differ in their relevant properties as far as word order is concerned. Neither does he study verb clusters involving the infinitival marker ‘to’ (on the infinitival marker ‘to’ in Old English, see Los 2000).
Koopman arrived at a remarkable generalisation in the course of his research. He first noted that verb order was quite free in Old English. Consider the following example sentences from Koopman:
(4) thaet cristes dheowdom ne sceal beon geneadad 1 23
that Christ’s service not may be forced
‘that Christ’s service may not be forced’
(5) thaet thaet festen sceolde abrocven bion 1 32
that that fortress should destroyed be
‘that that fortress was to be destroyed’
(6) thaet heo haligra gemeted beon meahte 3 21
thaet it holier considered be could
‘that it could be considered holier’
All three orders are frequent in Old English, relatively speaking. Then there are two more orders possible, although they are represented by a handful of examples only.
(7) aer hit geended mehte beon 3 12 (8 examples only)
before it ended could be
‘before it could be ended’
(8) & swa dheah ... heo beon maeg ongyten sodh martyr 2 13 (5 examples only)
and nevertheless ... he be can considered true martyr
‘and nevertheless he can be considered a true martyr’
The order 23 1 is absent in Old English. Koopman concludes, basing himself in part on arguments from interlinear glosses (which avoid that order when it occurs in the Latin text), that it must have been unacceptable in Old English. However that may be, Koopman makes another observation on these word orders, which is what interests us here.
As is well known, verb clusters may be broken up by non-verbal material. To exemplify, the verb cluster in Dutch may be broken up by a verbal particle (Bennis 1992: 39):
(9) dat hij dat probleem
that he that problem
(a) op moet hebben willen kunnen lossen
up must-1 have-2 wanted-3 can-4 solve-5
(b) moet op hebben willen kunnen lossen
(c) moet hebben op willen kunnen lossen
(d) moet hebben willen op kunnen lossen
(e) moet hebben willen kunnen op lossen
(up) must (up) have (up) want (up) could (up) solve
‘that he must have wanted to be able to solve that problem’
For some speakers, (b, c and d) are less acceptable than (a) and (e). Standard Dutch allows only particles to break up the cluster. Modern West-Flemish dialects allow all sorts of non-verbal material to break up the verb cluster (Haegeman 1992). Koopman investigated whether the Old English verb cluster could be broken up or not. It turns out that the order 123 could be broken up in Old English, much the same as in Modern West Flemish:
(10) Old English 123
(a) thaet he eft mage aet sumon saele beon geclaensod 1 X 2 3
that he afterwards may at some time be purified
‘that he may afterwards be purified at some time’
(b) thaet hi ne sceoldon beon to swidhe afyrhte 1 2 X 3
that they not might be too strongly terrified
‘that they might not be too terrified’
Notice that the break-up may occur either between 1 and 2, or between 2 and 3. Consider next the break-up possibilities for the cluster 132:
thaet he moste sumum rican menn to bearne geteald beon 1 X 3 2
that he might some powerful man as child accounted be
‘that he might be accounted the child of some powerful man’
Break-up only occurs between 1 and 32, but never between 3 and 2. The order 321 is never broken up. The rare order 312 is not broken up either. The other rare order, 213, is represented with 5 examples, four of which are broken up. An example is given below:
(12) 2 1 X 3
thaet dhu wurdhan scealt mid urum swurdum ofslagen mid him
that you become must with our swords killed with him
‘that you must be killed with our swords together with him’
The break occurs between 1 and 3, never between 2 and 1. The following table summarises where the break-up may occur in Old English:
1 X 2
2 X 3
1 X 3
This table sums up all the facts presented by Koopman. It is clear from the table what the generalisation is. We make Koopman’s generalisation explicit, as below:
Note that break-up possibilities for 31 can hardly be tested, since the order 231 does not occur at all, and the order 312 is rare. The order 312 is represented with only 8 examples, but it never exhibits break-up, neither between 3 and 1 nor between 1 and 2. Given the table, we might expect break-up to occur only between 1 and 2. The facts from Modern Dutch cited in (9) provide further support for (our version of) Koopman’s Generalisation.
There is one, and only one, systematic exception to the claim that a head-final order (or, descending) cannot be broken up: the negative particle ‘ne’ can be found in such a cluster. A study of negation is outside the scope of this paper (compare also note 9).
3. Verb-verb order in Old Frisian
will go on to show that Old Frisian verb clusters are in line with the version
of Koopman’s Generalisation proposed here. Bor’s (1971: 52-66) chapter on verb
clusters is almost exclusively concerned with verb clusters of two verbs. His
grammar is based on Skeltana Riucht, a text of
less than 900 lines. Interestingly, clusters of two verbs tend to be
(14) dat dae schelten ... unden bannes bigonnen habbe schillet 321
that the sheriff ... with the court session begun have shall
‘that the sheriff will have begun the court session’
De Haan (2001: 627), in his overview of Old Frisian syntax, draws attention to a relevant observation of Van der Meer. Van der Meer (1990) studied a small number of text from Jus Municipale Frisonum (Buma & Ebel 1977). Although he limited himself to verb clusters of two verbs, he did also study the break-up effects. He notes (Van der Meer 1990: 321, 325) that the order 2 X 1 is systematically absent, but his theoretical bias prevents him from ascribing due importance to that fact. All in all, Van der Meer may be said to have noticed the validity of what we refer to as Koopman's generalisation for two-verb clusters. Unfortunately, no tagged corpus of Old Frisian is available for research on the internet. However, part of the Old Frisian texts is available in an untagged format on the site of the ‘Deutsches Rechtswörterbuch’ (http://www.rzuser.uni-heidelberg.de/~cd2/drw/frameset.htm). This text archive is the largest Old Frisian text archive currently accessible on the world-wide web. It contains the following Old Frisian texts, based on the edition of Buma & Ebel (1963-77):
· Brokmer Recht
· Emsiger Recht
· Fivelgo Recht
· Hunsingo Recht
· Rüstringer Recht
· Westerlauwers Recht
I collected some 60 2-verb sentences with the modal mota (‘must’) or willa (‘want’). The variation I encountered is equivalent to what Bor and Van der Meer reported for Old Frisian. Thus I found the following three types:
(a) dat hi onder dae swirde mote thinghia
that he under the sword may prosecute
‘that he is allowed to prosecute under the sword’
[dass er sich unter dem Schwerte ausbedingen darf,]
(WesterlauwersR. I 398 )
(b) hit ne se thet <hit on reddian ondert den se iefta thet> ti clagere then forma eth wille layna.
it NEG be that it in sheriff’s presence done be or that the plaintiff the first oath wants reject
‘unless it is done in the sheriff's presence or if the plaintiff wants to reject the first oath’
[es sei denn, daß es [die Tat] in Gegenwart der Redjeven verübt sei oder der Kläger den ersten Eid (des Beklagten ohne dessen Widerrede) ablehnen wolle.]
(FivelgoR. 198 )
(a) soe di fria Fresa ti stride thinghie wille ende dy oera da vta habba wille,
if the free Frisian to battle prosecute wants and the other the non-residence have wants
‘If the free Frisian wants to prosecute to a twosome battle and the other claims non-residence’
[Wenn der freie Friese auf einen Zweikampf klagen will und der andere die Einrede der Nichtansässigkeit erheben möchte]
(WesterlauwersR. I 100 
(b) Ac ne skel thi side nawet wese, thetter enge quade liude ange mon ruogia mota
also NEG shall the custom nought be, that any evil people any man sue may
befta tha hellega sinetha.
without the holy court of justice
‘In addition, the custom will be abolished whereby any evil people may prosecute anybody without the cooperation of the holy court of justice’.
[Ferner soll die Sitte abgeschafft sein, dass irgendwelche böse Leute jemanden ohne Mitwirkung des heiligen Sendgerichtes anklagen dürfen.]
(a) Alder feder and moder hira dochter .... and hia ther thenne wille mit vnriuchte onspreka,
if father and mother their daughter .. and they it [her property] want with unjustice confiscate
‘if father and mother their daughter .. and they unjustly want to confiscate it’
[Wenn Vater und Mutter ihre Tochter ... und sie dann das (Gut der Tochter) zu Unrecht anfordern wollen]
(EmsigerR. 32 )
(b) Sa ist alra Brocmonna kere, thet ma nene freta ne mote thene liudafrethe vriewa;
So ist aller Brokmänner Küre, dass man keinem Friedlosen das dem Volke gebührende Friedensgeld erlassen darf;
(BrokmerR. 78 par 123 (= Rq. par 133))
The order 2X1 does not occur at all, which supports Koopman's Generalisation for Old Frisian. Consider also the frequency of the three orders in embedded clauses in my mini-corpus:
(18) Embedded clauses
(a) order 21 26 occurrences
(b) order 12 6 occurrences
(c) order 1x2 4 occurrences
This is in keeping with Bor's research findings about Skeltana Riucht. Bor also found that the order 21 was by far the most common. Comparison of Old English with Old Frisian thus yields two results:
We may wonder why there should be this difference between Old Frisian and Old English. It is generally assumed that the Indo-European ancestral language was head-final. If correct, then some outside factor (for example, language contact) must be held responsible for the head-initial bias of Old English as far as verb clusters are concerned.
Clusters of three verbs in Old English, and presumbly in Old Frisian as well, always begin with a modal. In contrast, Modern Dutch and Modern Frisian normally feature three-verb clusters beginning with the counterpart of to have or to be. An example of the former is given below:
(19a) Omdat hij had willen komen 123
because he had want-Inf come
‘because he had wanted to come’
(19b) Omdat er komme wold hie 321
because he come want-PfP had
‘because he had wanted to come’
(19c) V1 = have / be, V2 = modal, V3 = main verb
Note that Modern Dutch, like Modern German, exhibit the IPP-effect (Infinitivus-pro-Participio), with a so-called ‘Ersatzinfinitiv’ (replacement infinitive) where one would have expected a past participle. Modern Frisian does not have the IPP-effect.
Interestingly, three word clusters beginning with a finite form of ‘have’ or ‘be’ are radically absent from Old English. Correspondingly, the IPP-effect is (trivially) absent from Old English as well. Thus we have two types of clusters:
(20a) A-construction: 1 = have/be, 2 = modal, 3 = main: Modern Dutch, Modern German
(20b) B-construction: 1 = modal, 2 = have/be, 3 = main: Old English, Old and Middle Frisian, Modern English
I will follow Jarich Hoekstra (1997: 47) in referring to the type exemplified in (19) as the A-construction. The B-construction is exemplified by Koopman’s Old English examples, and by Modern English examples like “he could have come”.
The A-construction crucially depends on the presence of modal infinitives. Modern English does not have modal infinitives (*to can, *to must, and so on). Correspondingly, it does not have the A-construction. The B-construction does not depend on the presence of modal infinitives. Finite forms of modals, which do exist in Modern English, suffice to build a B-construction of three verbs. Old English is like Modern English in only exhibiting the B-construction. We may thus wonder whether Old English had modal infinitives.
In Frisian, the B-construction gradually went out of use in the 20th century. Jarich Hoekstra (1990, 1997) shows that it was still common in the 19th century. Old Frisian, like Old English, exclusively features the B-construction. Did Old Frisian have modal infinitives at all? Buma’s (1996) dictionary on the manuscript Jus does not contain any infinitival form of the modal auxiliary ‘motan’. We may now wonder to what extent Old English and Old Frisian had modal infinitives. And what was the role of the (disappearance of the) subjunctive? For future research, we note the task of investigating the syntax and semantics of the modal infinitive in Old Frisian in relation to the subjunctive; these questions are outside the scope of this paper, though we consider it useful to have formulated them.
The following is the central conclusion of our paper:
Head-initial verb clusters can in principle be broken up, whereas head-final verb cluster cannot be broken up (but see note 9). This generalisation covers facts of Old English, Old Frisian, Modern English, Modern Dutch and Modern Frisian. Interestingly, when I investigated the wild variation in verb clusters in the present-day dialects of Dutch (Hoekstra 1997), I noted that Koopman’s generalisation was also supported in those dialects. Thus we seem to witness an asymmetry between head-final and head-initial verb clusters, in accordance with Kayne’s asymmetry hypothesis (Kayne 1994), as implemented in Hoekstra (1997). I would like to round up with some bold claims which have the status of research questions:
Bennis, H. (1992). ‘Long Head Movement: The Position of Particles in the Verbal Cluster in Dutch’. In: R. Bok-Bennema & R. van Hout (eds.), Linguistics in the Netherlands. Amsterdam, 1992. 37-47.
Blom, E. and E. Hoekstra (1996). ‘IPP en werkwoordsvolgorde in het Achterhoeks’. Taal en Tongval 48, 72-83.
Bor, A. (1971). Word-groups
in the language of the Skeltana Riucht. A syntactic analysis with occasional
lexicological observations; followed by an inquiry into its punctuation and the
possibility of the influence on the text of spoken language.
Bor, A. (1990). ‘The use of
the negative adverbs ne and nawet in Old Frisian’.
In R. Bremmer, G. van der Meer and O. Vries (eds.), Aspects
of Old Frisian Philology.
Buma, W.J. (1996). Vollständiges
Wörterbuch zum westerlauwersschen Jus Municipale Frisonum.
Buma, W.J. & W. Ebel (1963-77). Eds., Altfriesische Rechtsquellen. Texte & Übersetzungen. Vol. 1: Das Rüstringer Recht; 2: Das Brokmer Recht; 3: Das Emsiger Recht; 4: Das Hunsingoer Recht; 5: Das Fivelgoer Recht; 6: Westerlauwerssches Recht. Jus municipale Frisonum. 2 vols. Göttingen.
Coetsem, F. van (1988). Loan
phonology and the two transfer types in language contact.
Fischer, O., A.
van Kemenade, W. Koopman & W. van der Wurff (2000).
of Early English.
Haan, G.J. de (2001). ‘Syntax of Old Frisian’. In N. Århammar, V.F. Faltings, J.F. Hoekstra, O. Vries, A.G.H. Walker and O. Wilts (eds), Handbuch des Friesischen. Handbook of Frisian Studies. Tübingen, 626-636.
(1992). Theory and
description in generative syntax. A case study in West Flemish.
Healey, A. & R. Venezky (1980). A
microfiche concordance to Old English. Publications of the Dictionary of Old
Hoekstra, E. (1997). ‘Analysing Linear Asymmetries in the Verb Clusters of Dutch and Frisian and their Dialects’. In D. Beerman, D. LeBlanc en H. Van Riemsdijk (eds), Rightward Movement. Amsterdam, 153-169.
Hoekstra, E. and W. Taanman (1996). ‘Een West-Friese gradatie van het Infinitivus-pro-Participio effect’. Nederlandse Taalkunde 1, 13-25.
Hoekstra, J. (1990). ‘Trijetiidwurdkonstruksjes’. It Beaken 52, 59-95.
Hoekstra, J. (1997). The
syntax of infinitives in Frisian.
Kayne, R. (1994). The
Antisymmetry of Syntax.
Koopman, W. (1990). Word
in Old English. With special reference to the Verb Phrase.
Los, B. (2000) Infinitival complementation in Old and Middle English. Dissertation, Vrije Universiteit Amsterdam.
G. van der (1990).
position of Old Frisian verbs and pronouns’.
In R. Bremmer, G. van der Meer and O. Vries, Aspects of Old Frisian
Mitchell, B. (1985). Old English Syntax. Oxford.
Wolf, H. (1987). ‘Default-morphological forms in Interference Frisian’. Groninger Arbeiten zur Germanistischen Linguistiek 40, 249-256.
 I would like to thank for comments
This sentence (order 132) is ungrammatical for some speakers of Dutch from The
Netherlands, although it is generally acceptable for speakers of Dutch from
 By “Modern Frisian”, we mean the variety of Frisian such as it is spoken by older native speakers of Frisian. This is also the type of Frisian that is described in grammars like for example Tiersma (1985). In so-called Interference Frisian (Wolf 1987), much more variation is encountered, due to imperfect language acquisition of Frisian, and subsequent interference from Dutch.
 Koopman considers the orders 3 12 (8 examples) and 2 13 (5 examples) to have been acceptable in Old English. He then construes a grammatical theory in which those orders are marked, in the sense that they require the application of marked rules of grammar. We would like to present a speculation here, and suggest that the marginal orders are the result of language contact between Old English and Latin. That is, some monks, who were particularly fluent in Latin, would be bilingual. As is well-known from language contact theory (Van Coetsem 1988), bilingual speakers do not keep their grammar separate; instead, the grammars start acting upon one another, causing changes. It is also well-known that such changes lead to the creation of sentences which may be unacceptable in either of the original grammars. To exemplify, the effect of Modern Dutch grammar on Modern Frisian grammar is such that sentences are created in so-called Interference Frisian which is are acceptable neither in Dutch nor in Frisian (see Wolf, Reitsma). Pursuing this speculation is beyond the scope of this article. Note though, if correct, it might well explain the existence of those marginal word orders.
 Given the generalisation to be proposed, we would expect that break-up in the order 312, if it occurred, would occur between 1 and 2, but not between 3 and 1. It is possible that 31 must be excluded from the scope of the generalisation, as there is no selection relation between 3 and 1 (only between 1 and 2, and 2 and 3).
 Particles are not ideal for testing break-up possibilities, since they obey a mysterious restriction, namely, they may not occur to the right of the matrix verb, that is, they are never extraposed. On the whole, it seems that even in Flemish clusters, which are heavily broken up, the relative word order restrictions of the matrix verb and its dependents (direct object NP, indirect object NP, particle, and so on) is always respected. I fail to see how syntactic theory can capture this generalisation (assuming it to be correct).
 Koopman is not explicit about the
scope of his generalisation. He notes that the order 321 is never broken up,
but he does not extend this generalisation to the other orders, where they are
valid for the sequence
 The German translations are from Buma & Ebel (1963-1977). They can also be found in the database of the Deutsches Rechtswörterbuch.
 Unless X is the negative marker ‘ne’. It would be interesting to study this negative marker in its own right (see Bor 1990 on Old Frisian), and to compare Old Frisian and Old English on this point.