On Expletives

  • Eric Haeberli
Part of the Studies in Natural Language and Linguistic Theory book series (SNLT, volume 54)


An important aspect of the analysis of‘XP-subject’ orders in the Germanic languages in chapter 4 was the assumption that in some languages subjects can remain below an adjunct XP because the highest subject position is filled by an empty expletive. This proposal raises the question what the status of expletives is within the framework proposed here. In particular, it is well known that the use of expletives can be subject to certain restrictions related for example to definiteness or to verb type (transitivity). The aim of this chapter is to explore how the properties of expletive constructions can be accounted for in terms of the theoretical proposal made in chapters 2 to 4.


Case Feature Multiple Attraction Nominal Element Inflectional Morphology Ergative Verb 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.


  1. 1.
    Given that V and T both have nominal features that must be checked, the idea would be that an expletive in the constructions in (1) is already inserted within the lowest VP and then moves up to finite TP (or possibly to CP in 1a). As for adjuncts like the one in clause-final position in (1c), I argued in chapter 2 that they are generally not accessible for categorial feature checking because they are generated in a position which is not a potential categorial checking position (cf. chapter, condition 21). Hence, the presence of an adjunct in (1c) does not have any influence on categorial feature checking and the insertion of an expletive is obligatory. Note finally also that the participle of the passivized main verb does not help for feature checking, either. In chapter, 1 argued that the feature matrix of participial T contains an interpretable N-feature. But an N-feature would not be sufficient for categorial feature checking because T-heads also contain an uninterpretable D-feature for which there would be no appropriate checker.Google Scholar
  2. 2.
    As pointed out in chapter, Afrikaans does not seem to meet the requirement in (3) as it has no inflectional morphology whatsoever but does not require the presence of expletives in sentences lacking arguments. What one may wonder is why a language like Afrikaans licenses empty expletives but for example Danish does not. The two languages share the property of not having agreement morphology and, in chapter 4, I argued that this morphological property has the consequence that no Agrs-head is present. Yet, the two languages are not entirely identical with respect to inflectional morphology. While Danish has an inflectional morpheme which is attached to the verb and which is generally analyzed as a tense morpheme, verbs in Afrikaans do not seem to bear any inflectional morphemes whatsoever (cf. chapter What could be argued then is that it is this property which accounts for the contrast between Afrikaans and Danish. More precisely, we could argue that a phonologically empty T can also have a phonologically empty [Spec, TP], and we therefore get constructions which lack overt expletives. As for languages with AgrsP, the idea would be that, as suggested in the text, Agrs has to be sufficiently rich to license an empty expletive. That morphologically impoverished Agrs requires the presence of an overt element can be argued to be related to the role of AgrsP as the projection licensing a value for Agrs. If the value is not or only weakly represented by agreement morphology, an overt nominal element has to be present, whereas an expletive element can be dropped if the agreement morphology is sufficiently rich to compensate for the absence of the expletive. As discussed briefly in Haeberli (2002b), certain issues arise with respect to what “sufficiently rich” means for the purposes of licensing empty expletives, but I will have to leave these aside here. Although the above proposals are fairly speculative, they would lead to a three-way distinction for expletives which resembles Huang’s (1984) classification with respect to referential pro-drop. Huang distinguishes “hot” langugages like English which have agreement morphology which is too weak to allow empty subjects, “medium” languages like Italian and Spanish which have rich agreement morphology which licenses empty subjects, and “cool” languages like Chinese or Japanese which have empty subjects but no verb-subject agreement at all. Adapting this classification to the syntax of expletives, we could say that there are “hot” languages like Danish, Swedish or West Flemish which have inflectional morphology which is too weak to license empty expletives, “medium” languages like German or Icelandic which have inflectional morphology which is rich enough to license empty expletives, and finally “cool” languages like Afrikaans which have empty expletives but no inflectional morphology whatsoever.Google Scholar
  3. 3.
    But cf. e.g. Birner and Ward (1998: chapter 3) and section 2.3 below for discussion of cases in which this restriction does not seem to hold.Google Scholar
  4. 4.
    As for the indefinite determiner et in (5a), we would have to assume that it occurs in a position below D (i.e. on a proxy head between D and the NP, or NP-adjoined). Cf. also chapter 2, fn. 43.Google Scholar
  5. 5.
    An anonymous reviewer raises the question whether the approach outlined here can rule out the occurrence of expletive-associate chains with objects as shown in the German example in (i). (i) * Er hat es gestern ein Buch gelesen German He has it yesterday a book read The idea of (i) is that the NP object ein Buch is merged in the object position of the verb together with the expletive es and the expletive then undergoes scrambling as the result of multiple attraction by D on T. Such a derivation can indeed be ruled out in terms of the proposals made in chapter 2. Given that the object in (i) would have to be an NP, it can check the N-features of V VP-internally. But, as proposed in chapter (cf. e.g. fn. 56), multiple attraction by D on T depends on the presence of an additional uninterpretable N-feature in T after checking of N on T by the subject. Since uninterpretable N on V is not available any more after V-movement to T, D on T cannot attract more than once in (i). The expletive could therefore not be separated from the object NP and the structure in (i) is ruled out.Google Scholar
  6. 6.
    In the discussion here I will focus on categorial feature checking. Checking of non-categorial features such as case or agreement is not crucial for the issues discussed at this point, but I will return to these types of checking in later sections.Google Scholar
  7. 7.
    The same variation is of course also possible with transitive verbs: (i) a. dan der vee mensen dienen boek gekocht een West Flemish that there many people (su) that book (DO) bought have ‘that many people have bought that book’ b. dan der dienen boek vee mensen gekocht een (expl.-DO-SU)Google Scholar
  8. 8.
    More precisely, the expletive must have moved on to AgrsP given that WF requires overt Agrs checking (cf. chapter However, the precise position of the expletive is not crucial for the points made below.Google Scholar
  9. 9.
    Den Dikken (1996:89) discusses WF data which are similar to those shown in (13) but provides no analysis of this variation. The data discussed by den Dikken combine the phenomenon found in (13) with the phenomenon of Verb Projection Raising (VPR) as illustrated in (i): (i) dan der zoun moeten [dienen boek meer Studenten kuopen] WFlemish that there should have-to that book (DO) more students (SU) buy ‘that more students should have to buy that book.’ It would go beyond the scope of this section to go into the complexities of VPR constructions. However, at first sight, (i) may suggest that reactivation of D and N onT in WF applies not only to finite T but also to non-finite T so that the reordering described in the text is also available within a non-finite structure. Yet, the same option does not seem to be available in VP-topicalization contexts. (ii) ??/* [Dienen boek vee Studenten gekocht] een der nie West Flemish This book (DO) many students (SU) bought have there not I have to leave it open here how the contrast between (i) and (ii) can be accounted for.Google Scholar
  10. 10.
    But cf. Jonas and Bobaljik (1993:93) for some potential exceptions in Icelandic. The status of these cases seems to be controversial however (cf. Jonas and Bobaljik 1993:93, fn. 41) and I will therefore leave them aside here.Google Scholar
  11. 11.
    Note however that (18) actually involves a double expletive-associate construction. Er and the IO form one expletive-associate chain, and, given the proposals made in chapter a non-overt expletive (in [Spec, AgrsP]) and the subject form another expletive-associate chain. The fact that only the latter expletive is non-overt can be accounted for in terms of condition (3) according to which a non-overt expletive must be licensed by agreement. This condition is only met by the expletive which forms a chain with the subject because it occurs in [Spec, AgrsP]. As for the expletive which forms a chain with the IO, it occurs in the outer [Spec, TP] and it is therefore realized overtly.Google Scholar
  12. 12.
    Within Chomsky’s system, it is not entirely clear what the exact syntactic status of Move F is. Chomsky seems to assume that a feature can move from a constituent and then adjoin like a head to another head. In terms of the proposal made here, feature movement is simply the transfer of features from one element to another one which occupies the same position. Thus, if we assume that in the feature matrix of a nominal element there is a slot for φ-features, then the proposal made in the text simply means that the φ-features of the feature matrix of the argument move to an empty φ-feature slot of the expletive.Google Scholar
  13. 13.
    This property does not seem to hold for all languages however. In French for example, associates cannot determine agreement (cf. e.g. Il reste/*restent quelques places — it remains(3sg)/remain(3pl) a-few places). In terms of the proposals made in the text, we may conclude then that expletives in French bear their own φ-features and that the presence of these features blocks feature transfer from the associate to the expletive. As for expletive constructions in German in which no associate is available (cf. section 1), we may assume that the expletive realizes a default φ-feature value and then licenses the 3sg agreement form of the verb.Google Scholar
  14. 14.
    The notion of “new information” mentioned by Breckenridge (1975) also plays an important role in Birner and Ward’s (1998) work. But they make a further distinction between “hearer-new” and “discourse-new”, the former being crucial in existential there-sentences and the latter in presentational there-sentences. Given for example the occurrence of proper names, which, typically, would be known to the hearer, it seems that “discourse-new” is a more important factor for the constructions in (21) to (25). This is also the conclusion reached by Birner and Ward (1998:274) for similar constructions in Yiddish. However, whether such a characterization is sufficient for German would have to beconfirmed on the basis of a detailed study of naturally occurring examples of this type.Google Scholar
  15. 15.
    There may be one exception to this observation however. As pointed out in chapter, certain ergative verbs license a definite subject within a topicalized VPin German. The best contexts for this option are contexts involving certain predicates which take an additional argument, as shown in the following example from chapter 3: (i) [Die Luft ausgegangen] ist dem Hans schon einmal German The air (su) run-out is the John (io) already once ‘John already ran out of air once before.’ If we assume that the fronted constituent corresponds to the non-finite TP, we have to conclude that agreement and NOM on finite T are checked by a non-overt expletive (as for the categorial features, they can be checked by the Dative argument, as proposed for Nominative-Dative inversion for Dutch). The question that arises then is how the non-overt expletive can move away from the associate to check case before checking agreement (NOM on finite T). What distinguishes a case like (i) from the cases discussed in the text is that the subject can be argued to check two case features in ergative constructions such as (i): a verbal case feature (ACC) and the case feature of finite T (NOM). Within the topicalized constituent in (i), the subject can check its categorial features and ACC. At this point in the derivation, the expletive and the subject still occur in the same position. It could be argued then that once the subject DP already has a case feature, it can, at least optionally, resist assignment of a second case feature and it is therefore only the expletive which is assigned NOM. Hence, the non-overt expletive moves out of the participial TP and we can obtain the order shown in (i). Thus, the idea would be that the expletive and the associate can be separated before agreement licensing with ergative verbs due to the presence of two case features which are checked within the subject chain. As for the morphophonological realization of NOM on the subject DP, it can be related to chain-internal transmission of the case feature, as proposed already in section 2.2 above. Cf. also chapter 6 (fn. 18) for some related proposals for Icelandic.Google Scholar
  16. 16.
    Another property of er/der insertion in Dutch and Frisian which, at least at first sight, seems difficult to explain in purely syntactic terms and which therefore could be argued to provide evidence for a pragmatic restriction is the fact that in transitive expletive constructions, the status of the object also has an influence on whether the occurrence of er/der is legitimate or not. Thus, the presence of a definite object already leads to marginality and the presence of an object pronoun is ruled out even if the subject is indefinite. This is shown in (i) (from Hoekstra 1991:64/65). (i) a. Der lêst ien in boek / ?dat boek / *it Frisian b. Er leest iemand een boek / ?dat boek / *het Dutch There reads somebody a book / that book / it’ somebody is reading a book/that book/it.’ This variation would follow from Bennis’ presuppositionality restriction (er/der is only present if there is no other presuppositional element).Google Scholar
  17. 17.
    Note however that such an account would only explain the Definiteness Effect. The phenomenon shown in (30) would require an independent explanation. Another issue the discussion in the text raises is the fact that English, a language with there-expletives, seems to allow definite subjects in there-constructions (cf. e.g. Birner and Ward 1998: chapter 3). At first sight, this option is surprising in terms of both the pragmatic and the syntactic approach discussed in the text. There should not be able to move away from a DP-associate because English has no AgrsP. One possibility to account for the situation in English would be to propose a structural analysis of there-constructions along the lines of the analysis of locative inversion constructions proposed in chapter 2. Thus, the idea would be that the associate inthere-constructions undergoes movement to an A’-position (cf. also Chomsky 2001:20) and that, since the associate then cannot be attracted to TP any more, there can check the categoriel features of T. Note finally that both approaches discussed in the text could be extended to West Flemish which, as observed in chapter (fn. 15), does not allow an overt expletive in AgrsP (or CP) with definite subjects. Hence, West Flemish behaves to a large extent like Dutch and Frisian with respect to the Definiteness Effect. Given that West Flemish uses a ‘there’-type expletive just like Dutch and Frisian, we can assume that either the pragmatic or the syntactic analysis outlined in the text for Dutch/Frisian can also account for the situation in WF.Google Scholar
  18. 18.
    A potential exception to this observation is the fact, pointed out by Vangsnes (1995:92, 1998), that so-called strong quantifiers such as ‘all’, ‘each’ etc. can co-occur with an expletive in Icelandic. This is shown in (i) (example from Vangsnes 1998:9). (i) bað hafa allir kettirnir étið mýs Icelandic It have all the-cats eaten mice ‘All the cats have eaten mice.’ Strong quantifiers have generally been treated like definite elements with respect to the occurrence in expletive-associate constructions. The occurrence of the strong quantifier ‘all’ in (i) is therefore rather surprising. One possibility to account for this property would be to assume that quantificational elements can occupy some quantificational (A’-)position and that once a subject occurs in this position an expletive checks higher features in the subject chain. Some evidence for such an approach comes from the distribution of quantified objects in Icelandic. Although Icelandic has the properties of a clear VO-language, quantified objects can precede the verb, at least for some speakers (cf. Svenonius 2000). An illustration with a strong quantifier is given in (ii) (from Svenonius 2000:263). (ii) % Strákarnir höfðu öllum steinunum hent í bílana Icelandic The-boys had all the-stones thrown in the-cars ‘The boys had thrown all the rocks at the cars.’ Since (ii) would be ungrammatical with a non-quantificational object, the grammaticality of (ii) suggests that the object has moved to a position which is only available for quantified elements, (i) could then be argued to illustrate a related phenomenon.Google Scholar
  19. 19.
    Yet, a more detailed investigation of Afrikaans may be necessary. Data discussed by Robbers (1997) suggest that the situation in Afrikaans is more complex. First of all, TECs do not seem to be entirely ruled out in Afrikaans. Robbers (1997:20) gives the following example: (i) ? Daar het iemand iets laat 1ê Afrikaans There has somebody something let lie’ somebody has left something behind.’ And secondly, the status of object movement out of the VP seems to be different in Afrikaans than in all the other Germanic languages discussed so far. As Robbers (1997:19) points out, in Afrikaans “an indefinite NP can always be scrambled, even when it receives an existential interpretation” This is in contrast to the languages discussed in the earlier chapters in which object movement is restricted to non-existential, i.e. specific objects. One way to account for this contrast would be to say that object movement in Afrikaans is of a different nature than in other Germanic languages, i.e. for example that it involves A’-movement rather than A-movement to TP.Google Scholar
  20. 20.
    The exact derivation of (36b) will be considered in more detail in chapter 6.3.Google Scholar
  21. 21.
    Admittedly, the connection here is not as close as between (34a) to (34c). Note however that this is an objection which one could also raise with respect to certain aspects of Bobaljik and Thráinsson’s (1998) system. For example, the presence of appropriate subject agreement morphology does not only entail the presence of AgrsP for them but also the presence of AgroP (cf. e.g. 1998:61 fn. 29). Given that morphological object agreement is generally not available on finite verbs in the Germanic languages (apart from Nominative object agreement in Icelandic) such a claim is not entirely unproblematic in a framework in which an attempt is made to relate the shape of the syntactic structure to morphological properties.Google Scholar
  22. 22.
    Note that proxy case projections above TP have the same effect as Agrs in the sense that they also create additional specifier positions in the inflectional domain. We may therefore assume that case can also favor a positive setting of parameter (8).Google Scholar
  23. 23.
    Within the system proposed here, (34d) and (34c) may not only be related through (34e) but also due to the fact that object movement (i.e. multiple attraction by D on T) is only possible if an unchecked N-feature of a V head triggers reactivation of D on T and hence if V has been moved to T (cf. also the discussion in chapter As for (34a) and (34b), it is not clear to me for the moment how they could be linked to (34d) in another way than via (34e).Google Scholar

Copyright information

© Springer Science+Business Media Dordrecht 2002

Authors and Affiliations

  • Eric Haeberli
    • 1
    • 2
  1. 1.University of ReadingUK
  2. 2.University of GenevaSwitzerland

Personalised recommendations