Syntactic Effects of Morphological Case

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


As discussed in the previous chapter, the distributional properties of nominal arguments have to a large extent been related to the theoretical concept of abstract’ Case within the generative framework. However, the notion of case already played a central role in many traditional discussions of the distribution of nominal constituents. What was relevant in these discussions is the notion of morphological case. It has generally been observed that the presence of a morphological case system allows nominal constituents to occur in a relatively free order within the clause. One source for this observation is the fact that the loss of a morphological case system and the loss of free word order generally seem to be closely linked in diachronic developments. For example, Sapir (1921:168) talks about “the drift toward the abolition of most case distinctions and the correlative drift toward position as an all-important grammatical method” (italics mine). Similar points are made by Meillet (1921) and Jespersen (1922). Meillet (1921:9) observes:


Word Order Case Feature Thematic Role Possessive Pronoun Nominal Element 
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.
    “The relations among words and the subtleties of meaning expressed by cases are expressed by other means: by the word order which tends to become fixed, as opposed to free as it used to be, and by particular words: prepositions, conjunctions, articles.”Google Scholar
  2. 2.
    Jespersen then argues that the development of a fixed word order is the cause of the disappearance of flexional endings. However, the analysis proposed in this chapter would rather imply that fixed word order is the effect of the loss of flexional endings. I will not go into the details of Jespersen’s proposal here. Note however that at least for the history of English Jespersen’s account does not seem to be correct since it is not the case that word order became rigid before the loss of case morphology (cf. Haeberli 1999: chapter 8.2.1 for discussion). Furthermore in Jespersen’s view of cause and effect, it would be difficult to explain why word order ever should become more rigid.Google Scholar
  3. 3.
    The distribution of objects with respect to the verb (SOV vs. SVO) has sometimes also been related to case morphology (cf. e.g. Greenberg 1963). 1 will leave this issue aside here for the moment and return to it in section 4 of this chapter.Google Scholar
  4. 4.
    An alternative way to analyze scrambling (and object movement in Icelandic) would be to say that the occurrence of objects before or after an adjunct is due to different base generation options for the adverb. Thus, the variation in (1) would occur because the adverb can be generated above or below the object. Here, I will adopt a movement analysis of (1) for two main reasons. First, from a purely theoretical point of view, in a framework such as the Minimalist Program in which movement is an option anyway, it seems natural to use this option for the analysis of word order variation. Base generation can then be treated in very simple and restrictive terms. Furthermore, in a framework which makes use of movement, it seems to be easier to express cross-linguistic variation as variation with respect to whether a movement process occurs or not rather than as variation with respect to where certain elements can be base generated (cf. also chapter 4.3.1). The second main reason for adopting a movement analysis is that there are certain empirical problems that arise for base generation approaches to scrambling. Several of them will be discussed in the context of German scrambling in section 2.4.2 below. For the time being, let us simply mention one point which is related to some issues discussed in chapter 2. In a system in which scrambling is simply a matter of inserting an adjunct above or below an object, the object does not seem to be directly involved in the mechanism giving rise to variation and we might therefore not expect that scrambling imposes any constraints on the object. However, this expectation is not borne out. As in Icelandic, we can find restrictions on the interpretation of objects preceding adjuncts in Dutch and German (cf. e.g. de Hoop 1992, Diesing 1992, 1996, Meinunger 1995). Such objects generally can only have a specific interpretation. It is not clear what the source of this restriction would be within a base generation approach. In terms of a movement approach, interpretational restrictions can be accounted for by assuming that, as proposed in chapter 2, only a particular type of object can undergo movement. For additional observations concerning the movement/base generation distinction, cf. section 2.4.2 below.Google Scholar
  5. 5.
    Zwart (1993a: 132) points out that for many speakers there is one context in which the order of arguments in a double object construction is not entirely rigid. With ditransitive verbs containing a particle, the objects can occur in both orders according to these speakers. (i) a. dat Jan Marie het boek terug gegeven heeft Dutch that Jan Mary (10) the book (DO) back given has ‘that John gave Mary the book back.’ b. dat Jan het boek Marie terug gegeven heeft (DO-IO) In addition, DO-IO orders also seem to improve with marked stress patterns, e.g. by stressing the IO (cf. Zwart 1993a:132) or by stressing one of the elements in a verbal cluster (J.W. Zwart, p.c.). Note however that this option seems to affect only the order of objects but not the subject-object order. I will have to leave it open here how the variation in (i) can be accounted for (but cf. Haeberli 1999:140, fn. 30 for some speculations).Google Scholar
  6. 6.
    Cf. also Haider (1993:202) for additional examples exhibiting contrasts between topicalization and scrambling and for the same conclusion, i.e. that these contrasts are problematic for an A’-movement analysis of scrambling.Google Scholar
  7. 12.
    Alternatively, it could be argued that an empty expletive is inserted in IP for satisfying the EPP and for allowing a subject of a transitive verb to remain VP-internally in German. However, such an approach does not seem to be more promising, either. On the basis of data from Icelandic, Bobaljik and Jonas (1996) argue that the subject of a transitive verb always leaves the VP although Icelandic can also be argued to have non-overt expletives. Thus, there would have to be a contrast between German and Icelandic and in order to account for this contrast, we would have to parametrize the use of non-overt expletives. But it would again be unclear what the motivation for the different parameter settings would be.Google Scholar
  8. 13.
    Cf. also fn. 4 above for a similar observation in the context of the distribution of objects with respect to adjuncts.Google Scholar
  9. 14.
    Note however that A-movement approaches have not always been based on the presence of AgroP. Thus, Fanselow (1990) treats adjunction of scrambled elements as A-movement. However, it is not clear in his system why this type of movement is of the A-type. Similarly, Haider (1993) uses VP-positions for scrambled elements. Yet, he does not seem to provide an explanation as to why an argument should ever occur in a VP-related A-position which does not correspond to its underlying position.Google Scholar
  10. 15.
    But cf. Diesing (1997) for a proposal according to which the parallelism is due to a semantic condition which applies to both A-and A’-object-movement.Google Scholar
  11. 16.
    This observation can be illustrated with example (21b), repeated here in (i), which shows that participles cannot be scrambled: (i) ?* dass Hans [Maria geküsst][i sicher ti hatte German that John Mary kissed certainly had [continued on next page] Regardless of whether we analyze the moved constituent simply as a VP or alternatively as a functional projection (i.e. as a [-D(+N), +T(-V)] category, cf. chapter, it does not have the adequate categorial features for being attracted by D on T. Hence scrambling is ruled out. It seems plausible that D is also missing in the structure of secondary predicates (cf. example 20) but it would lead us too far afield here to consider potential analyses of secondary predication within the framework adopted here.Google Scholar
  12. 17.
    One potential option that I exlude here is that the incorporated P-head in V1 counts as a feature checker for N on V2 after V1-to-V2movement. This option can be excluded under the assumption that feature percolation with incorporated heads is restricted. For incorporated P to count as a feature checker, its categorial features would have to percolate within the complex V-head (cf. feature checking under immediate dominance). So far I have used feature percolation in two main contexts: (i) percolation of a head’s features within its own projection (cf. chapter 2.4.2); (ii) percolation of an incorporated head’s feature if this feature acts as an attractor in combination with the host head (cf. overt object movement, chapter 2, section 5.2.10, fh. 59). For our purposes the second context is important. As pointed out in fn. 59 of chapter 2, when deleted D on T attracts an object DP in combination with N-features on V, the N-features on V have to percolate within TP so that they can attract the object and establish a checking relation with it. This percolation process could therefore be argued to be related to the combination of features on T (T and deleted D) with features contained in the complex T-head (V and uninterpretable N on V). Given this observation, we can propose now that categorial feature percolation is restricted to contexts in which the features of the host are related to the features of the incorporated head. In the case of P-incorporation into V, the categorial features of the P-head are not related to V in any way since the two feature matrices are entirely distinct. That categorial feature percolation must be restricted in such a context is plausible because it would be undesirable to say that for example the N-feature of an incorporated P head automatically percolates within the VP and thereby gives the VP the status of a nominal element. Similarly, for example N-incorporation into V should also not turn the VP into a nominal element through feature percolation. Hence, it would be plausible to assume that percolation of the categorial features of an incorporated head is generally blocked except for contexts like those found with multiple attraction to TP and that therefore an incorporated P-head cannot count as a feature checker for V2 in the context discussed in the text.Google Scholar
  13. 18.
    Cf. also chapter 2, fn. 20, for another aspect of this proposal, i.e. checking of the V-feature of an NP complement of D.Google Scholar
  14. 19.
    As throughout this chapter, I am focusing on non-pronominal elements here and P-stranding is impossible with non-pronominal complements of P. However, when elements like Dutch er (‘there’), i.e. elements that have been referred to as R-pronouns (cf. van Riemsdijk 1978), occur as complements of P, they do allow P-stranding. R-pronouns therefore can undergo scrambling as shown in (i) (fromvan Riemsdijk 1978:192). (i) Ik had er niet op gerekend Dutch I had there not on counted ‘I had not counted on it’ This option seems to be related to the syntax of pronominal elements and it is therefore not crucial for our purposes here.Google Scholar
  15. 20.
    As a matter of fact, in contrast to wh-movement, the preposition cannot move along in this case: (i) * About John was talked One possibility for ruling out (i) would be to say that a PP does not have the relevant features for determining agreement on the verb and that it therefore cannot move to subject position when there is no alternative element which could establish the agreement relation (contrary to for example the locative inversion construction where agreement can be established by the postverbal argument). Hence, the P has to be stranded so that a DP can determine agreement on the verb. The same problem does not arise with PP scrambling since in that case no agreement relation has to be established. Note also that the analysis proposed here suggests that the prepositions of locative elements are not incorporated in English. Otherwise, we would expect the complement of a locative preposition to move out of the PP, a possibility which is ungrammatical: (ii) * [The room]i walked (John) [into ti] (John) This fact can be explained in terms of the structure of locative inversion constructions discussed in chapter 2 fn. 38. There, I proposed, following Hoekstra and Mulder (1990), that locative inversion constructions may involve a V-head taking a small clause in which the DP is the argument of the PP. The relevant structure is shown in (iii). (iii) [vp V [sc DP PP] The structure in (iii) has a desirable consequence for the analysis of (ii). In terms of (iii), the head of the locative PP does not play a role for categorial feature checking on V. There is therefore no reason for P to incorporate into V in order to check N on V. But if a locative P as in (ii) is not incorporated into V due to the structure in (iii), it is impossible for a DP to be attracted to TP out of the PP because the PP does not become transparent for DP-attraction.Google Scholar
  16. 21.
    One additional issue that may arise in terms of the analysis proposed in the text is related to the syntax of adjuncts. The relevant data are given in (i) (ia from Vanden Wyngaerd 1989, ib from Collins and Thráinsson 1996:409). (i) a. dat ik morgen de directeur over die zaak in Kopenhagen ontmoet Dutch that I tomorrow the director on this matter in Copenhagen meet b. Hann lanar Maríu líklega pessa bók aldrei Icelandic He lends Mary probably this book never In the examples in (i), we have a direct object which occurs to the left of a VP-peripheral adjunct and which therefore has moved out of the VP. In terms of the analysis proposed here, this means that the direct object has moved to the lowest [Spec, TP], Assuming that the argument that precedes the DO in (i) occurs in a higher [Spec, TP], the word order in (i) seems to suggest that the adjunct that intervenes between the two arguments also occurs within TP. This would mean then that adjuncts are not only licensed in independent specifier positions, as suggested in chapter, but also within multiple specifier configurations. In other words, we may have to conclude that a head such as T which licenses multiple attraction and hence multiple specifiers in languages like Dutch, German and Icelandic can also make one or several specifiers available for adjunct licensing. [continued on next page] There may be an alternative analysis of the cases in (i) however. For the VP structure, I have been assuming that more than one projection can bear categorial features and that the number of V projections depends on the semantic properties of the verb (i.e. on the number of thematic roles that are assigned). For example a ditransitive verb has been represented in terms of three VP-shells due to the semantic properties of the verb (three thematic roles). What could be argued then by analogy is that the inflectional domain does not consist of a single projection containing categorial features, either, but that more than one projection bears such features. Thus, one such projection hosts tense, as assumed so far, whereas another (lower) projection with the same categorial features could host another semantically contentful feature such as aspect for example. Such an approach would require a more abstract representation of functional features (e.g. [-F(N), +F(V);-N, +V] for inflectional projections in the verbal domain, where F(N)/F(V) stands for nominal or verbal functional features). The result is that for example in (ib) all arguments first are attracted to the lower inflectional projection through multiple attraction. But the higher inflectional projection then only attracts the subject and the IO. Finally, the subject moves on to CP. The adjunct aldrei thus can occur in a proxy projection between VP and the lower inflectional projection and the adjunct liklega can occur in a proxy projection between the higher inflectional projection and the lower one. In terms of such an analysis, adjuncts would not have to be inserted into multiple specifier structures. Cf. also chapter 6 fn. 4 for another context (agreement) in which an approach in terms of multiple inflectional projections may have desirable consequences. However, for the general points made in the text, the presence of additional inflectional projections would not be directly relevant, and I will therefore not pursue this option in more detail.Google Scholar
  17. 22.
    As for the distinction between possessive pronouns and Nominative/Accusative pronouns, we may assume that it is not a matter of case but of word class (cf. also Hudson 1995:385ff). Alternatively, along the lines of proposals made in chapter 2 (fh. 49), possessive pronouns could be argued to be the result of the combination of a pronoun with a possessive morpheme which reflects the presence of a dummy licensing P and they therefore would not have to be defined in terms of a distinct case.Google Scholar
  18. 23.
    As for LF interpretability, the situation is a bit more complex. Certain cases like Dative or Genitive could be argued to have at least some content that is interpreted by LF, but for cases like Nominative or Accusative it seems less plausible to assume that they play a role at LF. Since for our purposes it is sufficient to establish that a feature is interpretable at least for one of the two interfaces and since the case features postulated here clearly play a role at PF for the realization of morphological case, I will not pursue the issue of the LF-interpretability of case here.Google Scholar
  19. 24.
    The approach to the syntactic representation of case discussed in the text is similar to the approach to the syntactic representation of agreement outlined by Bobaljik (1995), Thráinsson (1996), Bobaljik and Thráinsson (1998) (cf. chapter 1.3). These authors argue that there is a close relation between the syntactic representation of e.g. Agrs and properties of agreement morphology. Cf. chapter 4 for a similar proposal within the framework outlined here. [continued on next page] The analysis proposed here also has some similarities with proposals made by Neeleman (1994) and Weerman (1997) (cf. also Neeleman and Weerman 1999). Both authors suggest that case distinctions like Nominative, Accusative or Dative are only available in the grammar of languages which have a rich morphological case system. However, in both frameworks (cf. Neeleman 1994:420ff., Weerman 1997:452), word order freedom has to be derived by associating thematic roles to case features, an option that I have argued against in section 2.4.2. In addition, abstract Case is still present to some extent in the two approaches. Neeleman (1994:419) still seems to assume some kind of abstract Case for languages without case morphology. As for the system proposed by Weerman (1997), it is more in line with the goal pursued here because it eliminates the notion of abstract Case. However, the elimination of abstract Case is mainly obtained by introducing an abstract element of a different type, i.e. an empty functional head in the DP-structure, and this functional head can be argued to have the same shortcoming as the concept of abstract Case in the sense that its presence simply has to be stipulated in languages without case morphology.Google Scholar
  20. 25.
    An apparent problem for this analysis is what has been referred to as “long distance scrambling”, i.e. scrambling out of certain non-finite subordinate clauses. This issue will be discussed in section 2.8.3.Google Scholar
  21. 26.
    Note that the A’-movement type of scrambling which can be found in Dutch and which depends on a particular intonational pattern (“focus scrambling”, cf. section 2.4.1, example 35) is possible with PPs (J.W. Zwart, p.c.): (i) dat op ZULK.E opmerkingen Jan natuurlijk NAUWELIJKS kon reageren Dutch that on such remarks John of-course hardly can react [contd. next page] As argued in section 2.4.1, such cases must be distinguished from the type of scrambling found in German and they are therefore not directly relevant for our purposes.Google Scholar
  22. 27.
    Given the proposals made in chapter 2 and in section 2.6 of this chapter, the object is an NP due to its indefiniteness/non-specificity. However, this choice is not crucial for the points made below. If the object was a DP, it could be assumed that it is in its base position and only undergoes non-overt movement or alternatively that it moves overtly but only within the non-finite TP.Google Scholar
  23. 28.
    In line with the standard assumption made within head-final approaches to the syntax of Dutch, I will assume here that the verb right-adjoins to te in I. But cf. Zwart (1993:98ff.) for some problems this analysis may raise.Google Scholar
  24. 29.
    Given that I will assume in chapter 4 that adjunction to functional projections is generally not possible, there are two main options for deriving extraposition. Either the extraposed constituent is moved to a right-branching specifier of a proxy category above the inflectional head occupied by the verb, or it occupies an X’-adjoined position within the projection whose head is occupied by the verb.Google Scholar
  25. 30.
    Note that for both (69a) and (69b) there would be alternative word orders. In (69a), the non-finite TP in brackets could precede the finite verb whereas in (69b) the infinitival TP could occur to the right of the finite verb. The derivations for the two options in (69) are basically identical. The only difference is whether the non-finite TP has undergone extraposition or not. In (69a), extraposition has taken place, but in (69b) it has not.Google Scholar
  26. 31.
    Note that these movement processes within the infinitival clause are both non-cyclic, but, as discussed in chapter 2.4.4 and, this option is available within the framework adopted here.Google Scholar
  27. 32.
    I will use this term throughout this section for referring to the construction illustrated in (74). However, as we will see, the analysis proposed here suggests that topicalization does not necessarily involve simply a VP but, at least in some cases, the entire non-finite TP.Google Scholar
  28. 33.
    It is interesting to note though that the occurrence of indefinite subjects (i.e. NPs) in VP-topicalization contexts with transitive verbs seems to be grammatical in West Flemish when an overt expletive functions as a feature checker in the matrix domain (L. Haegeman, p.c.). (i) [vee Studenten dienen boek gekocht] een der nie West Flemish many students (SU) this book (DO) bought have there(expl) not ‘There aren’t many students who have bought this book.’ One possibility for analyzing this option would be in terms of topicalization of a constituent which is larger than the participial TP, i.e. for example of the auxiliary VP, This option is available only for NP-subjects because a DP would be attracted to finite TP (i.e. the position occupied by the expletive in (i)). Yet, the availability of the option in (i) seems to vary cross-linguistically. Data discussed by Haider (1993:153) suggest that the equivalent of (i) can also be found in German. (ii) a. ? [Linguisten Langusten gespeist] haben da wohl noch nie German Linguists (su) lobsters (DO) eaten have there(loc.) probably still never ‘Linguists have probably never eaten any lobsters there.’ b. ? [Ein Aussenseiter ein Derby gewonnen] hat da wohl noch nie An outsider (su) a derby (DO) won has there(loc) probably still never ‘An outsider has probably never won a derby there.’ However, Haider considers (ii) as marginal, and to other speakers the examples in (ii) sound much more degraded or entirely ungrammatical. Thus, the grammatical option which derives (i) in West Flemish seems to be more marked or even impossible in German. At present, it is not clear to me how this contrast can be explained. Cf. also Alexiadou and Anagnostopoulou (2001) for other instances in which subjects of transitive verbs seem to occupy low positions. I have to leave it for future research to determine how these cases could be analyzed within the framework proposed here.Google Scholar
  29. 34.
    It has sometimes been suggested that the construction illustrated in (79) is subject to a definiteness restriction. However, as pointed out for example by Bayer and Kornfilt (1994:38) and von Stechow (1990:192) and as shown in (79c/d), definite subjects can occur in fronted constituents. This suggests that the definiteness restriction is not a rigid syntactic constraint. Cf. also chapter 5.2.3 for some additional observations on definiteness restrictions in German (or the absence thereof).Google Scholar
  30. 35.
    The result would then be roughly as proposed in GB terms within frameworks such as those outlined by Safir (1985), den Besten (1985) or Belletti (1988), in which Nominative is realized on a lower element although it does not occupy a position in which Nominative would normally be assigned.Google Scholar
  31. 36.
    Note however that the claim should probably not be as strong as to say that the analysis of word order freedom in German holds for any type of argument order variation in a language with a rich morphological case system. Cf. e.g. Kiss (1994) for a discussion of word order freedom in Hungarian which suggests that syntactic properties such as binding found with arguments occurring in various orders are substantially different in this language from those found in languages such as German, Korean or Japanese and that therefore word order freedom in Hungarian may require a different analysis.Google Scholar
  32. 37.
    One major difference between Korean and German concerns long distance scrambling. As discussed in sections 2.3.1 and 2.8.3, long scrambling in German is only possible out of certain infinitival clauses. In Korean however, elements can also be scrambled out of finite clauses. There has been some controversy as to whether long distance scrambling out of finite clauses has the same syntactic status as short scrambling. Thus, it has been argued for example by Saito (1992) for Japanese that long distance scrambling is a different syntactic phenomenon than short scrambling. On the other hand, Lee (1993) argues for Korean that the two phenomena should be treated identically. If Lee’s proposal is correct, the main option to account for the difference between Korean and German within the framework proposed here would be to assume that a clause union process through head incorporation, as described in section 2.2.3 for German infinitival complements, is also possible with finite clauses in Korean.Google Scholar
  33. 38.
    In the literature on Korean, word order freedom has of course also been related to the presence of case morphology. Cf. e.g. O’Grady (1991:2): “This ‘free word order’ option is made possible in part by the existence of the case particles” Cf. also Kim (1995) for an example within the Minimalist framework. However, Kim’s analysis does not provide a genuine theoretical explanation for the correlation between word order freedom and case morphology but it is simply based on a stipulation concerning the effect of case morphology on optionality of feature strength (cf. 1995:51, fh.l4).Google Scholar
  34. 39.
    Note however that Lee (1993:89f.) discusses some examples which suggest that at least in certain discourse contexts scrambling with two Nominative arguments is acceptable. Two options come to mind to deal with this observation. First, as suggested to me by Liliane Haegeman (p.c.), it could be argued that inversion of the Nominative arguments involves a different type of movement, i.e. an A’-movement to a topic position for example. Alternatively it could be argued that, at least as a marginal option, a second proxy category for NOM checking can be created. The presence of a second proxy category would derive the inverted order of arguments because the lower head first attracts the higher (i.e. closest) Nominative argument and the second proxy category attracts the lower argument. Hence, the order of arguments is inverted. This option could be argued to be marked because a structure with a single proxy category and hence an arguably more economical structure would be available. Note that this marked option can be related to the fact that there are two distinct NOM features. The same option would not be available in cases where attraction of identical elements is triggered by the same feature as in the case of multiple attraction by D on T where I have argued that it is the same D-feature on T which is reactivated and thereby triggers multiple attraction. Note finally that a similar phenomenon as that found in Korean can also be observed in German. There are some rare ditransitive verbs such as lehren (‘to teach’) or abfragen (‘to test’) which take two Accusative objects as shown in (i) (ia and iia from Neeleman 1994:421). (i) a. dass der Lehrer die Schüler diese Sprache lehrt German that the-NOM teacher the-ACC students this-ACC language teaches ‘that the teacher teaches this language to the students.’ b. Der Lehrer hat den Schüler die unregelmässigen Verben abgefragt The-NOM teacher has the-ACC student the-ACC irregular verbs tested ‘The teacher tested the student on the irregular verbs.’ With these verbs, scrambling generally leads to bad results, too. (ii) a. ?* dass der Lehrer diese Sprache die Schüler lehrt that the-NOM teacher this-ACC language the-ACC students teaches b.?* Der Lehrer hat [die unregelmässigen Verben]i den Schüler ti abgefragt The-NOM teacher has the-ACC irregular verbs the-ACC student tested Again, the contrast between (ii) and cases involving scrambling of e.g. DAT IOs (cf. 2, 55) can be related to attraction by an identical feature (i.e. ACC). However, an anonymous reviewer points out that the examples in (ii) become acceptable if the lowest DP receives focus stress. This observation could be related to a proposal that I will make in chapter 6 in the context of word order in Icelandic. There, 1 propose that case checking can be done non-overtly by focused nomináis. For (ii), this means that the unstressed object overtly moves to a case checking position above TP, whereas the stressed object remains under TP at the surface, thereby giving rise to inversion of the two objects.Google Scholar
  35. 40.
    That (89b) is not judged as entirely ungrammatical may be due to some interference of the fact that the string in (89b) can actually get a grammatical interpretation. Topic markers can also be dropped in Korean and the DO in (89b) could therefore be interpreted as a topic rather than as a scrambled element. However, a topic reading generally goes together with a certain prosodie pattern. As Lee (1993:92) points out, it is therefore only with a clear intonation break between the object and the subject that (89b) would be acceptable.Google Scholar
  36. 41.
    The omission of case morphemes in Korean is of course a phenomenon which is substantially different from the occasional lack of case morphology discussed for German in section 2.8.1. There, I argued that nominal elements such as proper names which do not show any morphological case information nevertheless bear a case feature. But these elements simply cannot show any morphological case information due to their morphological properties. In other words, there is no actual optionality of case in German in the sense that the same element could occur with a case morpheme in one example and without a case morpheme in another example. Full DPs which can show case distinctions always realize them in German, proper names never show case distinctions (except for Genitive). And given this lack of actual optionality, there is no reason to assume that there is optionality as to whether case features occur in German. In Korean however, there is genuine optionality and, as the contrast in (89) shows, this optionality seems to be syntactically encoded.Google Scholar
  37. 42.
    The examples given here are based on my own dialect which is spoken in the East of Switzerland. Cf. e.g. also Cooper (1995:12ff.) on scrambling and case in Zurich Swiss German.Google Scholar
  38. 43.
    In Haeberli (2001), the setting of (91) is related to the occurrence of case morphology on both pronouns and full NPs/DPs. Apart from the reasons discussed in section 2.7 and in this section (i.e. the motivations for postulating case features), an additional reason for pursuing the alternative approach summarized in (92) is that, as I will show in chapter 4, it can be extended to capture aspects of the relation between morphology and syntactic representation in the context of agreement as well. The discussion in chapter 4 will also show that, apart from the factor mentioned in (92), there may be an additional motivation for postulating case features at least in some languages (cf. chapter 4, fn. 10).Google Scholar
  39. 44.
    Similarly, this proposal also means that a language would not be expected to change from a negative setting of (92) to a positive setting in the absence of any evidence for a rich morphological case system.Google Scholar
  40. 45.
    As for the fact that the case feature on A is generally of the type DAT or GEN, but not ACC or NOM, it can be related to the distinct status of the various features within the morphological case system adopted here. As suggested earlier, NOM and ACC can be argued to be default case features for T and V, i.e. for the categorial feature matrices [-D(-N), +T(+V)] and [-N, +V]. A having neither of these categorial feature specifications, it is associated with one of the remaining cases, i.e. DAT or GEN.Google Scholar
  41. 46.
    It should be pointed out though that the correlation between the occurrence of nominal complements of adjectives and the presence of a morphological case system may not be as strong as the one concerning word order order freedom. As observed for example by Platzack (1982) or Weerman (1997), nominal complements of adjectives can also be found in languages without a morphological case system (ia from Platzack 1982:41, ib from Weerman 1997:438). (i) a. Han var trogen sin hustru Swedish He was faithful his wife ‘He was faithful to his wife.’ b. Hij is het Frans machtig Dutch He is the French able ‘He is able to speak French.’ However, in both Swedish and Dutch, there is only a small set of adjectives of the type shown in (i) and the large majority of adjectives do not license nominal complements. This suggests that (i) is not a productive option as opposed to German where we can find a considerable number of adjectives which license nominal complements (cf. van Riemsdijk 1983). Similarly, in English, we can find at least one adjective which seems to license a nominal complement, namely worth (e.g. This book is worth a lot.) The status of (i) and worth in English could be analyzed as follows. Dutch, English or Swedish are languages which, in earlier stages of their history, had a rich morphological case system and which licensed nominal complements of adjectives more productively. The occurrence of some nominal complements of adjectives in present-day Dutch, English or Swedish could therefore be argued to be a relic of an option which used to be productive at an earlier stage of these languages. As for the formal analysis of this residue within the framework adopted here, there are two main possibilities. First of all, we could assume that the relevant adjectives license an empty prepositional head and that, as a result, nominal complements of adjectives are nominal only at the surface. Alternatively, it could be argued that a few adjectives have maintained the ability to establish a checking relation with their complement in languages like Dutch, English or Swedish. After the loss of case features in the history of these languages, a different (entirely abstract) feature would thus have taken over the function of initiating a feature checking relation with the complement. As a consequence of this feature checking relation, these adjectives still allow their complement to select their V-feature as a feature checker. The conclusion that we reach is similar to the one reached by Platzack (1982) for Swedish, i.e. that the option which licenses nominal complements of adjectives makes “the Swedish grammar marked to a certain extent” (1982:53). In our analysis, the markedness is related to abstract elements whose presence within the syntactic representation is not independently required (i.e. an empty P or an abstract attractor). Note that this marked option can be acquired by the language learner because there is positive evidence for its existence (i.e. the occurrence of adjectives with a nominal complement in the input). If this positive evidence did not exist, there would be no motivation for introducing any abstract elements related to A, and languages without nominal complements of adjectives are therefore not expected to develop this option over time. The empty P/abstract attractor option is thus restricted to the scenario sketched here, i.e. to contexts in which a historical residue is integrated into the grammar.Google Scholar
  42. 47.
    For example in the next chapter it will be argued that verbal agreement can also trigger the occurrence of proxy projections above TP. This could account for why Dutch also has OV order (i.e. remnant movement to a projection between CP and TP) despite the absence of case checking projections.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