Some Implications of Number Agreement on COMP

Eric Hoekstra

F. Drijkoningen & K.Hengeveld (eds) Linguistics in the Netherlands 1993. Benjamins, Amsterdam, 61-68.

1. Formulating the problems

The phenomenon of complementiser agreement for number was first recognised as number agreement in Ganderheyden (1897), a supplement to Molema's (1887) dictionary of the dialect(s) of Groningen.1 Van Haeringen (1939), apparently independently, wrote an article which made the phenomenon generally known. It is illustrated below, with examples from South-Holland from Van Haeringen (1939):

(1) a. Ik weet niet ovve ze komme
I know not if-PL they come
b. Ik weet niet of ze komp
I know not if-SG she comes
c. ik/jij/hij komp - we/julle/ze komme
I/you/he come-SG - we/you/they come-PL
d. stof -stoffe ; lat - latte, etc.

cloth-SG cloth-PL lath-SG lath-PL

The complementiser is in the plural if the subject is plural. Note that the morphological expression of plurality on the complementiser is identical to the morphological expression of plurality on the noun and the verb, a fact, as will become clear, of some importance. The phenomenon was subsequently discussed in various influential articles (to mention a few, Van Haeringen 1958, Goeman 1979, Bennis & Haegeman 1983, Zwart 1992 and others).
The question to be answered here is: under what conditions can complementiser agreement for number become morphologically visible?
Interestingly, complementiser agreement for number, it seems, is only found in Germanic dialects. Although Romance dialects are well-studied, there does not seem to be a dialect with complementiser agreement for number only. The same applies to the Scandinavian languages. Nor do there seem to be non-Indo-European languages with complementiser agreement of this type (Ken Hale, p.c.). I will concentrate on dialects of Dutch. Here, complementiser agreement can be found in various parts of the country, such as Groningen (Ganderheyden 1897, De Vries 1940), Amsterdam (Ganderheyden 1897), North Holland and South Holland (Van Haeringen 1939), Zealand (see below) and various parts of Belgian Flanders and French Flanders (Vanacker 1949, with maps). Hence we ask ourselves:

(2) Why does complementiser agreement occur in a number of dialects
of Dutch, whereas it is at least rare (possibly non-existent)
in the other languages of the world?

This question is not addressed in the references cited, thus providing a raison d'être for this article.
In addition, we may ask ourselves whether complementiser agreement occurs arbitrarily in Dutch dialects or whether it correlates with other facts. To investigate this more specific question, we need to have a guiding hypothesis. I will adopt Van Haeringen's (1939:255) idea that there is a correlation (at least a precondition for the birth of complementiser agreement) between the morphological shape of the complementiser and the shape of the verb. This correlation is illustrated below:

(3) (Van Haeringen 1939:248-249)
a. Azze der zieke mense benne
if-PL there ill persons are-PL
b. We motte zien datteme klaar komme2
we must see that-PL-we ready get
(4) (Korstanje 1984:42.26-27, dialect of Wemeldinge, Zuid-Beveland)
a. Oons wazzen zo benauwd aan de schepen op de loop zouen gaen
we were so afraid that-PL the ships on the run would-PL go
b. Wieaan dat ezeid eaen motten dat me 's bewiezen
who-that-PL that said have-PL must it prove

Thus, the complementiser ends in schwa in those dialects in which the verb ends in schwa, and the complementiser ends in an /n/ in those dialects in which the verb ends in an /n/. However, the existence of this correlation still does not answer the question why complementiser agreement should occur in a number of Dutch dialects.
Furthermore, there is another indication that Van Haeringen's suggestion cannot be the whole story. The present tense form of the verb in several Zealand dialects (Goeree-Overflakkee, Walcheren, Tholen, St. Philipsland and Noord-Beveland) ends in schwa (Blancquaert & Tavernier-Vereecken 1944, and the references below). Nevertheless, these dialects display a systematic absence of complementiser agreement, as is clear from an investigation of written texts from this area. Thus there is no complementiser agreement in texts from Goeree-Overflakkee (Broecke-de Man, Soldaat-Poortvliet & Heerschap 1988), Walcheren (Broecke-de Man & A. Francke 1987), Tholen and St. Philipsland (Broecke-de Man & Krijger-Goedebuure 1986) and Noord-Beveland (Broecke-de Man & Eikenhout 1984). We can now ask ourselves the following question:

(5) Why is there no complementiser agreement in the northern
Zealand dialects mentioned above?

In order to answer the questions raised in this section, we must be more explicit about our assumptions about agreement and about the relation between COMP and AGRS.

2. Basic Idea: Nominal and Verbal Features

In the Minimalist Program (Chomsky 1992) and earlier (e.g. Chomsky 1981, but in the context of INFL), it is assumed that AGRS has both nominal and verbal features. Following Zwart (1992), a natural extension of Den Besten's original proposal in Chomsky's (1992) theory, I adopt the idea that there is abstract AGRS-to-C in Dutch in embedded clauses. Given that the nominal and verbal features are present on the head AGRS, and given that AGRS is bound by C, we expect the complementiser to also exhibit nominal and verbal features.
How can this complementiser agreement become morphologically visible? Variation between dat and datte, for example, may first have been phonological or lexical before datte got associated with plurality. Why can datte be syntactically interpreted as a plural? Given that COMP binds AGRS, and given that AGRS contains nominal and verbal features, it follows that datte can be interpreted as a plural only if the schwa designates both the nominal plural and the verbal plural.

3. Fleshing out the analysis

In order to become morphologically visible on C, the nominal plural must be homophonous to the verbal plural. This requirement can only be met if the verbal plural does not distinguish between first person, second person and third person. If the verbal paradigm has a person opposition in the plural (given that nouns do not have a person opposition in their inflectional paradigm), homophony fails by definition. The language-learning child will not be able to interpret a complementiser in schwa as a plural, if the schwa is not the unambiguous marker of the plural. Hence languages (or dialects) having a person opposition in the plural will fail to have complementiser agreement purely for number. This accounts for the cross-linguistic paucity of complementiser agreement, since a vast amount of languages has a person opposition in the verbal plural, such as most Romance dialects.
To sum, our account predicts that it is a precondition for complementiser agreement that the verbal paradigm has been leveled so as to obliterate person distinctions in the plural. Once there are no person distinctions in the verbal plural, nominal and verbal plural can be homophonous. We present this conclusion as a corollary about complementiser agreement:

(6) Corollary 1
It is a precondition for pure number agreement on COMP that
the verbal paradigm does not have a person opposition.

On the other hand, a language must at least have in its verbal morphology a distinction between singular and plural:

(7) Corollary 2
It is a precondition for number agreement on COMP that the verbal
paradigm have a morphological singular/plural opposition.

Corollary 2 correctly predicts that Scandinavian languages or Chinese dialects won't have any (visible) complementiser agreement:3 in those languages, the verbal plural is homophonous to the verbal singular. Thus the amount of languages of the world which don't have any visible complementiser agreement is further increased. Corollary 2 has an equivalent for nouns:

(8) Corollary 3
It is a precondition for complementiser agreement that the nominal
paradigm have a morphological singular/plural opposition.

Corollary 3 predicts that there is no complementiser agreement in languages which do not mark singular or plural on nouns, like some Creole languages do. As far as I know, such languages do not have complementiser agreement.
The remainder of this article is devoted to a presentation of evidence from the dialects of Dutch, and in particular of Zealand, in favour of the proposed analysis.

4. Complementiser agreement and homophony in the plural

Our analysis embodies two claims. First, complementiser agreement can only occur if the nominal plural is homophonous to the verbal plural. This is the case in the examples discussed above. Groningen: the verbal plural and the nominal plural are both morphologically expressed as a syllabic -n; correspondingly, complementiser agreement is found, expressed as a syllabic -n. North and South Holland: the verbal plural and the nominal plural are morphologically expressed as a schwa. Correspondingly, complementiser agreement is found in the morphological shape of a schwa. In Flanders, the verbal plural and the nominal plural are both expressed as -en. Correspondingly, complementiser agreement is found in -(e)n.4
The second prediction is that complementiser agreement is not found if the nominal plural is not homophonous to the verbal plural. This is the case in Frisian. The nominal plural ends in a syllabic -n whereas the verbal plural (in the present tense) ends in schwa. Thus it is correctly predicted that Frisian dialects will not have complementiser agreement, because there is no homophony.
Evidence also comes from Dutch dialects in which the nominal plural is not homophonous to the verbal plural. Comparison of maps (Blancquaert & Tavernier-Vereecken 1944, Ton Goeman p.c.) for nominal and verbal plurals teaches us that several Zealand dialects form the nominal plural in /en/ and the verbal plural in schwa. These areas include Goeree-Overflakkee (Van den Broecke-de Man, Soldaat-Poortvliet & Heerschap 1988), Walcheren (Van den Broecke-de Man & Francke 1987), and Noord-Beveland (Van den Broecke-de Man & Eikenhout 1984). Written texts from these areas were investigated, and no complementiser agreement was found.
Beveland is particularly interesting, since in Zuid-Beveland the contrast between verbal plural and nominal plural breaks down (Van den Broecke-de Man & Van Gilst 1985). More specifically, the verbal plural ends in /n/ in the eastern part of Zuid Beveland, e.g. in Kapelle and Wemeldinge. Thus the verbal plural and the nominal plural coincide. Correspondingly, we find complementiser agreement in /n/ in texts from this area, bearing out our hypothesis. Thus Wemeldinge has complementiser agreement, as illustrated in (4) above. Kapelle also has complementiser agreement, as shown below:

(9) Kapelle, Van den Broecke-de Man & Van Gilst (1985:99)
a. wat an ze d'r van vonnen
what that-PL they there of thought-PL
b. voe an ze beslooten om ...
before that-PL they decided to
c. az a j' zelf was
than that-SG you self were-SG

The correlation seems to hold at the level of individual informants. Thus an informant from Rilland does not have complementiser agreement (Van den Broecke-de Man & Van Gilst 1985: 196-200), even though Rilland is, roughly, in the area where we would expect this. But then it turns out that this informant employs verbal plurals in schwa, so that the required homophony is lacking.5
Surely, it cannot be an accident that we find complementiser agreement exactly in those Zealand dialects in which the verbal plural is homophonous to the nominal plural.

5. An Alternative

It might also be suggested that the nominal ending is irrelevant. Instead, what is required is merely that the verbal plural is stable across tenses. That is, if the verbal plural in the present is the same as the verbal plural in the past, then we can have complementiser agreement. This hypothesis accounts for the lack of complementiser agreement in Frisian by pointing out that the verbal present plural in Frisian ends in schwa whereas the verbal past plural ends in a syllabic -n.
However, this hypothesis cannot account for the fact that complementiser agreement is so rare. This hypothesis would expect complementiser agreement in, say, Romance languages to be possible, with a full paradigm for the complementiser. This hypothesis would thus fail to relate the phenomenon of complementiser agreement to the lack of a person opposition within the verbal paradigm.
Another disadvantage of this hypothesis is that it leads to a counterfactual prediction. It predicts the possibility of complementiser agreement wherever the verbal endings are the same in past and present, as far as plurality is concerned. This prediction is incorrect. In the dialects of Walcheren and Noord-Beveland, the verbal plural is in schwa in the present and the past. Yet we do not find any instance of a complementiser in schwa if there is a plural subject. Thus this hypothesis makes the wrong prediction. It is essential that the nominal plural is homophonous to the verbal plural.6

6. Conclusion

It cannot be an accident that we find a correlation between complementiser agreement and homophony of the verbal and nominal plural endings. This correlation supports our claim that complementiser agreement arises if there is homophony between the verbal plural and the nominal plural. Theoretically, this correlation supports the following three pillars of our analysis:

(10) a. There is an AGRS in Dutch
b. AGRS is bound by C in overt syntax
c. AGRS involves verbal and nominal features

This account explains why complementiser agreement occurs in Dutch (dialects) but not in a host of other dialects/languages of the world. Furthermore, it also accounts for the fact that pure number agreement on the complementiser must be homophonous to the verbal plural and to the nominal plural.
To close this paper on a historical note, I would like to point out that complementiser agreement is not a very recent phenomenon: even in the surviving written documents complementiser agreement for number is attested in South Holland as early as 1557 (Goeman 1993).


Albrecht, K. (1881) Die Leipziger Mundart. Arnoldische Buchhandlung, Leipzig.
Bennis, H. & L. Haegeman (1984) "On the status of agreement and relative clauses in West Flemish". In W. de Geest & Y. Putseys (eds.) Sentential Complementation. Foris, Dordrecht.
Blancquaert, E. & C. Tavernier-Vereecken (1944) "Apocope van slot-n na doffe e in het Nederlandsch". In H. Draye (ed.) Feestbundel H.J. van de Wijer, deel 2, 7-24. Instituut voor Vlaamsche Toponomie, Leuven.
Broecke-de Man, E.J. van den & L. Eikenhout (1984) Dialect op Noord-Beveland. Zeeuwsche Vereeniging voor Dialectonderzoek, Vlissingen. Drukkerij van der Peyl-Kruiningen.
Broecke-de Man, E.J. van den & J.A. van Gilst (1985) Dialect op Zuid-Beveland. Zeeuwsche Vereeniging voor Dialectonderzoek, Vlissingen. Drukkerij van der Peyl-Kruiningen.
Broecke-de Man, E.J. van den & A. Francke (1987) Dialect op Walcheren. Zeeuwsche Vereeniging voor Dialectonderzoek, Vlissingen. Drukkerij van der Peyl-Kruiningen.
Broecke-de Man, E.J. van den & D.K. Soldaat-Poortvliet en P. Heerschap (1988) Dialect op Goeree-Overflakkee. Zeeuwsche Vereeniging voor Dialectonderzoek, Vlissingen. Drukkerij van der Peyl-Kruiningen.
Chomsky, N. (1981) Lectures on Government and Binding. Foris, Dordrecht.
Chomsky, N. (1992) "A Minimalist Program for Linguistic Theory". MIT Occasional Papers in Linguistics 1.
Ganderheyden, A.A. (1897) Supplement op H. Molema's Woordenboek der Groningse Volkstaal. J.C. Mekel, Winsum.
Ginneken, J. van (1939) "De vervoeging der onderschikkende voegwoorden en voornaamwoorden". Onze Taaltuin 8, 1-11.
Goeman, T. (1979) "Comp-Agreement?" In W. Zonneveld (ed.) Linguistics in the Netherlands 1977-1979, 291-303.
Goeman, T. (1993) "Vroege attestaties van congruerende voegwoorden in de kuststreek van Zuid-Holland. Theoretische implicaties". Ms, P.J. Meertens Instituut.
Haeringen, C.B. van (1939) "Congruerende voegwoorden". Tijdschrift voor Nederlandsche Taal en Letterkunde 58, 161-176.
Haeringen, C.B. van (1958) "Vervoegde voegwoorden in het Oosten". In Gramarie, 309-318. Van Gorcum, Assen.
Hoekstra, E. (1992) "Generative Grammar meets Dialectal Variation". In S. Looyenga & D. Gilbers (eds.) Language and Cognition 2, 133-144. University of Groningen.
Korstanje, M.J. (1984) "Op weg van DiaLect naar Algemeen Nederlands. Een weg zonder einde." Deventer. Ms.
Molema, H. (1887) Woordenboek der Groningse Volkstaal in de 19e Eeuw. J.C. Mekel, Winsum.
Vanacker, V.F. (1949) "Over enkele meervoudsvormen van voegwoorden". Taal & Tongval 1, 32-45, 77-93, 108-112.
Zwart, C. Jan-W. (1992) "Verb-Movement and Complementiser Agreement". Talk presented at GLOW, Lisboa. Ms, University of Groningen.

1 I would like to thank Reineke Bok-Bennema, Hans den Besten, Marcel den Dikken, Ton Goeman, Riny Huybregts, Fred Weerman and Jan-Wouter Zwart, and the audience at the TIN-dag, for comments and/or discussion.

2 Datteme can only be 1P. Ze can be 3s (feminine) or 3P, but datteze can only be PL, according to Van Haeringen (1939).

3 As a reviewer points out, Scandinavian and Chinese can be said to have perfect complementiser agreement, there being no discrepancy between abstract and morphological/visible agreement due to the latter's absence.

4 The final consonant of the complementiser is often, but not always, elided, as in as/an, though asn is also found, see Vanacker (1949). Elision of schwa after vowel seems less surprising. The lack of complete morphological transparency might be taken as an indication that this type of complementiser agreement is relatively old.

5 Of course, geographical homogenuity is not what is at stake. Due to migration, for example, geographical correlations are dissolved.

6 Van Ginneken (1939) claims that complementiser agreement for singular/plural is also found in the German dialect of Leipzig (Albrecht 1881). This dialect does not have nominal plurals in "-n" generally, as we would predict. Hence it is a potential counterexample. Upon closer inspection the relevant complementiser agreement in "-n" turns out not to be related to a singular/plural contrast. The statement in (i) and the example in (ii), both from Van Ginneken's source, make this clear:
(i) .. vor du, Sie und wir sagt man obb'n statt ob..
"for you (2S), thou and we, people say
obb'n instead of ob"
(ii) obb'n de weg gehst
whether-2 you away go-2S (Albrecht 1881:27)
Du is a second person singular. Nevertheless, it triggers the suffix "-n" on the complementiser. This proves that the "-n" suffix does not correlate with a singular/plural distinction. It seems that the "-n" suffix is rather a remainder of person agreement.