* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Download Contents - Utrecht University Repository
Compound (linguistics) wikipedia , lookup
Navajo grammar wikipedia , lookup
Junction Grammar wikipedia , lookup
Scottish Gaelic grammar wikipedia , lookup
Georgian grammar wikipedia , lookup
English clause syntax wikipedia , lookup
Macedonian grammar wikipedia , lookup
Kannada grammar wikipedia , lookup
Untranslatability wikipedia , lookup
Morphology (linguistics) wikipedia , lookup
Malay grammar wikipedia , lookup
Yiddish grammar wikipedia , lookup
Transformational grammar wikipedia , lookup
Esperanto grammar wikipedia , lookup
Latin syntax wikipedia , lookup
Turkish grammar wikipedia , lookup
Chinese grammar wikipedia , lookup
Icelandic grammar wikipedia , lookup
Portuguese grammar wikipedia , lookup
Lexical semantics wikipedia , lookup
Agglutination wikipedia , lookup
Spanish grammar wikipedia , lookup
Pipil grammar wikipedia , lookup
English grammar wikipedia , lookup
Checking out basic word order in Cantonese The basic word order of Cantonese and its place within the theory of Collective and Individual Checking proposed by Jan Koster Doctoral thesis by Frank Oosterom 2007 Contents 1 Introduction 3 2 Koster’s Theory 15 3 Cantonese word order 22 4 Implications of Cantonese word order facts 41 5 Samenvatting in het Nederlands 44 Appendix I: Other sentences in Cantonese 46 References 48 1 Introduction In the thesis I am about to present, I will investigate the theory proposed by Koster (1999), in which he argues for a analysis that can explain basic word order differences between English and Dutch. Not only does he give an account for the differences between these two specific languages; he also wants to solidify the hypothesis that the word order differences between languages across the world can be derived by the use of one simple parameter, namely the Pied Piping Parameter. This parameter, which will be explained later on in this thesis, is the sole responsible trigger for the parametrical differences between languages that have the basic word order of the verb preceding the object (VO languages), and languages that have the basic word order of the verb following the object (OV languages). The central question of this work is: Can the basic word order of Cantonese be explained adequately with the theory proposed by Koster (1999)? I shall examine three phenomena used to support the theory proposed by Koster and then apply it to the Cantonese situation. The results then will be either problematic or unproblematic for the theory. This in turn shall lead me to conclude in favour of Koster’s theory or make a proposal for improvement. 2 Before I address the question as to how this investigation has been undertaken, I will discuss the theoretical background on which Koster’s theory is based. There are two major theories that are the basis of Koster’s theory: Antisymmetry (Kayne, 1994) and Minimalism (Chomsky, 1995). We also have to take a look at two other major influential publication in the field of comparative syntax, namely Pollock (1989) and Cinque (1999). All these shall be discussed in the following sections. 1.1 Antisymmetry The principal idea behind the Antisymmetry Theory is the Linear Correspondence Axiom (LCA): Linear Correspondence Axiom X asymmetrically c-commands Y iff X c-commands Y and Y does not c-command X. This asymmetrical c-command relationship provides a parallel between linear and hierarchical order of phrases, which are said to be corresponding. Thus, when phrases have a different linear order they must have a different hierarchical order. Beside this axiom, Antisymmetry Theory assumes a differentiation between segment and category. See for instance (1). (1) P / \ M P | / \ Q R S | | | q r T | t In this tree structure M and P symmetrically c-command each other, giving the ordered pairs <q,r> and <r,q>, which would violate Antisymmetry. To make a structure like (1) convergent with Antisymmetry, Kayne opts to keep segments from c-commanding. This is done according to the following rule: (2) X c-commands Y iff X and Y are categories and X excludes Y and every category that dominates X also dominates Y. 3 Now M c-commands P, but P does not c-command M, resulting in only one order, namely <q,r>. This derives from the fact that P is no longer a category, but a segment of the higher P. Thus we see that specifiers are in fact examples of adjunction. Is it possible to get multiple adjunction in Antisymmetry? To answer this, see (3). (3) P / \ L P | / \ K M P | | k Q R S | | | q r T / \ | t While L is asymmetrically c-commanding Q in (3), Q is also a-symmetrically ccommanding K. The result is two orders of k and q, <k,q> and <q,k>m, which is a violation of the LCA. Therefore we can assume that to have multiple adjuncts is not possible in Antisymmetry. Kayne also comes to this conclusion, and states that in a structure like (3) P can have only one specifier. The consequences of this “linear order” theory are quite fundamental, as according to Kayne, the notion of precedence is at the basis of a number of phenomena. One of which concerns the universal word order. In Kayne’s view, all natural languages are VO at their deepest level, and therefore also head-initial. The way head-final structures can be derived is by assuming leftward movement of post-head material. Such movement would 4 then have to be overt. This goes against accounts that take OV as the core of all languages (cf. Haider 1997, 2004, 2005 among others). What part of Antisymmetry is adopted by Koster in his account of the word order differences between VO and OV languages will be examined in chapter 2. 1.2 Minimalism Starting with Chomsky’s publication “The Minimalist Program” (1994), there is a new approach to the problems a generative linguist is faced with every day. Not so much a theory, it claims to be a tool in the process that should be unravelling the mysterious components of language. Basically, Minimalism proposes only two levels of representation: Logic Form (LF) and Phonetic Form (PF), eliminating Deep and Surface Structure. This leaves the familiar representations that were previously associated with the semantic and phonological/phonetic part of the computational system. Next to these two representations, i.e. LF and PF, there are two components that give a language what it needs to construct sentences: a lexicon and a computational system. The lexicon is seen as a numeration of lexical items, with all their relevant properties (phonological, semantic and formal). The task of the computational system is to select items out of this lexicon and putting them into a derivation. Hereby we encounter the first two operations of the computational system: Select and Merge. The operation Merge is what we get when the system takes a pair of syntactic objects and replaces them by a single combined object. It is important to notice that the moment lexical items enter the derivation they become syntactic objects. Take for example the verb “give” and the noun “presents”. In the lexicon they are two distinct lexical items. The operation Select takes them out of the lexicon and places them in a derivational structure by connecting the verb and the noun together. The result is a new category, VP, and this VP becomes a new syntactic object, which can be connected to another item out of the lexicon by the same operation Merge. The crucial proposal in the Minimalist Program is the way sentences are composed by the computational system. The key word throughout 5 the computational process is “economy”1; economy of derivation and economy of representation. No longer do we have to start with tree structures and try to fit in the words we encounter in a sentence. Instead, all the words in a sentence are in the aforementioned lexicon, out of which we construct a structure bottom up, starting with the verb. Let us see how this would work for an actual sentence. Take sentence (5): Each verb has a number of features to be checked and in order to do so other words out of the lexicon need to be added to the structure. So the first step of deriving the structure of sentence (5) would be (6) and (7). (5) Pete gives presents to Lucy. (6) {Pete, give, presents, to, Lucy} -> lexicon; take out “give” and put it in a structure: V | give (7) the verb “give” needs a theme: {Pete, presents, to, Lucy} -> lexicon; Take out “presents” and put it in the structure: VP / \ V NP | | give presents By adding the NP “presents” to the structure the verb “give” has one of its features checked, namely the one that demands that is has a theme. The computational system must find a way to check all the features an element may have. If it doesn’t, the derivation crashes. Hence ungrammatical sentences are ruled out by the system. As one can see for sentence (5), the verb “gives” has more features to be checked than the requirement of having a theme. Furthermore, every element out of the enumeration that 1 Economy is also seen in other aspects of the Minimalistic approach. These aspects, such as length of derivation, length of links and procrastinate will not be pursued further here. For a detailed account see Chomsky (1994). 6 is added to the structure presents new features to be checked until all elements have their features checked. Again, anytime during the process when features fail to be checked, a crash is inevitable. At some point in the derivation a third operation is carried out: Spell Out. During this operation information will be filtered out which is no longer relevant for the eventual logical form of the sentence. What is left will be sent to the component governing the phonetic form of the sentence (PF). An important consequence of Spell Out is that after this operation is carried out, no more lexical items can be added to the derivation. This counts for both LF and PF. At Spell Out the formal features of every lexical item must be eliminated. These features are the ones that need to be checked during the computational process. Unless every feature is checked accordingly, the derivation cannot converge and therefore will crash. This feature checking process will be at the core of Koster’s explanation for the difference between VO and OV languages. 1.3 Koster’s theory Without going in-depth into the theory of Koster (this will be done in chapter 2), we need to have a general understanding of what the theory explains and how. Crucially, it takes one characteristic of languages, namely the feature checking of internal arguments. Internal arguments are those within the lowest verb phrase (VP). This essentially follows Chomsky’s Minimalism (1995 and subsequent publications). In order to explain why English and Dutch differ in a number of respects, he uses a parameter, dubbed the Pied Piping Parameter. Pied Piping Parameter - In Dutch, the internal arguments check their corresponding case features individually. - In English the internal arguments check their corresponding case collectively, after percolating them to the VP. The essence of the analysis is as follows. In Dutch, the internal arguments don’t have to stay inside the VP. This allows for more flexibility in word order than in English. English 7 has VPs that act as recipients of features, percolated to them by their internal arguments. The features then are checked collectively through movement to the relevant positions of the whole VP. According to Koster the Pied Piping Parameter predicts not just the differences found between English and Dutch, but between VO and OV languages in general. Three of the discussed differences in Koster’s article will be treated in this thesis and will be tested for Cantonese, which is not closely related to English or Dutch. Taking Cantonese as the test language shows the effectiveness of Koster’s theory, since it needs to be accurate for all world languages if it is placed in a “restrictive theory of Universal Grammar” (Koster, 1999, p. 39). In short, it needs to be able to predict the right features of every language since all languages should be derived in the same manner. The only possible differences reside in parametrical choices. I have chosen to focus on three differences addressed in Koster’s article, because they deal with the most basic word order characteristics of the languages discussed. This makes the results of an investigation into the Cantonese word order facts more informative when it comes to evaluating them in a comparative framework. It enables us to see the fundamental characteristics of a language, which are to be explained by a theory placed in the tradition of the Universal Base Hypothesis (Greenberg, 1963). Now I will discuss two other recent theories that can give an explanation for the word order differences found between VO and OV languages. After that, I will continue with a discussion of two influential publications, that can shed some light on the issue of this thesis. 1.4 Advanced Antisymmetry Barbiers (2000), like Koster, takes Antisymmetry (Kaynes, 1994) and Minimalism (Chomsky, 1995) as a starting point. The difference between Koster and Barbiers lies in the way they feel that verb movement triggers derivation. Essentially, verb movement is said to be short verb movement in Barbiers’ view. The main factor determining the parametrical differences in word order is the difference in the moment of spell out. VO languages like English have short verb movement before spell out. The second type has short verb movement after spell out, resulting in an OV language like Dutch. 8 Koster, on the other hand, takes the feature checking of the internal arguments as a starting point. These features percolate to the VP in one type of language and the movement of the whole VP results in a specific and word order. The second type of language show the internal arguments move separately out of the VP to check their features. This results in a word order that is freer. 1.5 Directionality Finally, a theory proposed by Haider (2004, 2005) shows another way of analyzing basic word order differences in languages. Directionality Theory is based on the following three constraints: 1 projections are endocentric 2 heads license their complements directionally (left or right) 3 projections are universally right branching Together, these constraints result in the two options of argument structure projection in languages. In VO languages, where you have head-initial structures, the direction of licensing (constraint 2) is rightward. Nothing strange happens when the first complement is merged with the verbal head, only that the complement directly follows the head, where in head-final structures it precedes the head. Complications arise when the second complement needs to be merged (as in double object constructions). Now, we cannot simply merge the V’, resulting from the first merger, with the second complement, for that would yield a left-branching structure, like is evident from (10)a. (10) ... V' / V' -> / \ Δ \ V -> Δ 0 9 In contrast, in head-final structures this would not lead to any problems, as can be seen in (10)b ((12)b in Haider 2005). There, the direction of licensing is left, and because the head is to the absolute right in the structure, all the merging of any number of complements results in a right branching structure. To overcome this violation, Haider claims a VP shell structure (as proposed by Chomsky, 1995) leads to a well-formed structure. The extra VP, merged with the head as the first complement, generates an additional complement position, as can be seen in (11) ((12)a in Haider 2005). (11) ... V' / \ V -> VP 0 / \ Δ V' / v 0 \ -> Δ Now, the first and the second complement are directionally licensed in a right branching structure (complying with constraint 2 and 3). The last constraint (projections are endocentric) is another way of saying: the licensing relation between the head and its complement(s) must be local. This means that any arguments that need to be licensed can only be licensed if they are in the local domain of the head, where the local domain refers to the adjacency of the argument to both the licensing head and the projection head. Looking at the structure in (11) we see that both arguments that need to be licensed can be licensed locally, because either the licensing head or the projection of the head can directionally do so; there are no intervening elements to block such a licensing. The only position where a local licensing fails is the specifier position of the VP. This highest position of the VP is outside the local licensing domain of the head, thereby in need of external licensing (this position is not represented in (11), but is a sister to the highest V’ in that structure)2. For the derivation of OV structures, we already saw that the right branching constraint does not put us into trouble when adding more than one complement to the structure. 2 Another way in which this locality constraint is instantiated is mutual c-command. See Haider (2005) for more details. 10 Also, the directionality of licensing is not in danger of failing, since it is leftward and always able to comply (either by its head or its subsequent projections). Finally, the locality constraint is satisfied, since the arguments are always adjacent to their licensing heads or projections of these licensing heads. The two directions of licensing and the three constraints provide us with the tools to distinguish between VO and OV languages. There are two fundamental differences between them: 1 head-initial projections are compact, meaning that there can be no intervening elements in argument structure projections; head-final projections do not have to be compact; 2 head-initial projections have one and only one argument position which has to be licensed externally, because it is outside the local domain of the verbal head; head-final projections do not have such an external argument position To summarize: for head-initial languages, the direction of licensing is rightward, thereby exemplifying the English situation in which the arguments generate to the right of the verb. In contrast, head-final languages, like Dutch and German, show the internal arguments left of the verb before any movement takes place. Now let's turn to two very influential publications in the field of comparitive syntax, namely Pollock (1989) and Cinque (1999). 1.6 Pollock (1989) In Pollock (1989) it is discussed how a number of differences between English and French can be explained with breaking down IP into several projections. The result is a skeletal projection along the way of (12). (12a) CP (12b) / \ CP / IP / \ TP \ / VP \ NegP / \ AgrP / 11 \ VP Instead of the more simple structure of (12a), Pollock proposes that IP actually exists of three separate projections like in (12b), each with their own features. The central idea of his article is that in English AgrP is not transparent to Θ-role assignment which allows for verb movement in situations where it isn't allowed in French. The reason why Pollock's article is interesting for our evaluation of Koster's theory is that it tells us something important about word order, especially when it comes to adverbs. As we shall see, Cantonese behaves the same like English in a lot of basic ways. So if Koster is on the right track, we should find that the same characteristics for English presented by Pollock hold up for Cantonese. With respect to word order Pollock tells us that lexical verbs in English can and therefore must move first to AgrP and then to TP. The movement from AgrP to TP is mandatory because in AgrP there can be no Θ-role assignment. In French, on the other hand, AgrP can assign Θ-roles in certain constructions, namely finite constructions, and therefore does not need verb movement from AgrP to TP in those kind of constructions. In order to see whether Cantonese works the same way as English in this respect, we can observe if lexical verbs in Cantonese can and must move to TP. To determine this we need to look at the order of lexical verbs in relation to NegP. Do they appear to the left of negative elements or do they have a mixed distribution in word order? The same can be asked about the relative order of adverbs and negative elements. Both questions are relevant, as they are presented as exemplifying Koster's theory. There are some that raise certain doubts to Pollock's findings, like Iatridou (1990) who finds that the existence of the AgrP is not so compelling, and Ouhalla (1990) who argues for a different approach to NegP. I will not divert into the intricacies of these articles, since they are not directly relevant to the central topic of this thesis, but it suffices to say that although Pollock (1989) is a widely respected article, the views in it are not shared by everyone in the field. Still, the breaking down of IP into separate projections enables us to look at differences in languages and determine where these differences can be located and explained. 1.6 Cinque (1999) When one enters the world of adverbs, Cinque's “Adverbs and Functional Heads, A CrossLinguistic Perspective” from 1999 can not be left out of the discussion. In this book, he convincingly shows the rigid hierarchy between adverbs, which he groups in 30 types. 12 Though his core data comes from Roman languages, he also includes several other languages of the world to support his idea that adverbs have their own projections and that they have a tight order between them. These languages include Norwegian, Hebrew, Malagasy (an Austronesian language) and even Mandarin Chinese. It's important to note that the adverb hierarchy is universal. The same can be said about the order of the morphemes that encode the different types of functional notions of the clause like mood, tense and aspect; it is both fixed and universal. Cinque's main achievement is to correlate these two tendencies in languages. In other words, they match consistently. The strict hierarchy provides an interesting showcase for Koster's theory. Since Koster argues that adverb placement is determined by the feature checking of NPs, he predicts that in English certain adverbs will always follow the verb where others can be either following or preceding the verb. Although he does not clearly define the class of adverbs that have different placements in languages, we can take a couple of adverbs, and see whether they behave like Koster predicts they do. In the final section of this chapter, I will lay down the set up of my research. 1.7 Research set up All these theories have a way of dealing with the parametrical deviations between languages when it comes to basic word orders. Still, not one is compatible with another, although Koster and Barbiers both draw from the same theoretical background, namely Antisymmetry and Minimalism. The main focus of the present investigation will be the testing of Koster’s theory in a language outside of its closely language family, namely Cantonese. The reason why Cantonese has been chosen as the language with which to check the discussed theory is that this language shows remarkable resemblance with English as to its rigidity in word order. This is especially noticeable in the reason behind the rigid word order. In both languages, the lack of overt case marking necessitates the use of word orders to determine the grammatical relations in a sentence. As will be shown, when investigating the phenomena, under discussion in Koster’s paper, for the Cantonese situation, the linguistic set up of both languages seems to throw some shadows on the generalizing analysis, which generalizing character makes the theory of Koster so attractive in the first place. 13 There are several ways to deal with the found discrepancies. The most obvious one is to dismiss the theory altogether, and try to find a new way to explain the discussed word order differences. Reasons why this is not the desirable thing to do will be presented in the chapter Implications of Cantonese word order facts. The path that will be taken here is to come up with an advanced version of Koster’s proposal. That way, the benefits of the theory will, hopefully, be preserved while the pitfalls will be dealt with in a convincing and well-grounded manner. To get to this point, chapter two will go into the intricacies of Koster’s theory of collective and individual feature checking. This will be the basis for the advanced version of the proposal Koster makes. This is followed by a chapter in which I investigate the way Cantonese behaves in the case of the phenomena discussed in Koster’s theory. Then, a chapter on the implications of the results in Cantonese will be presented. The thesis ends with a conclusion and recommendations for further research. 2 Koster’s theory Abstract In this chapter the theory that Koster proposes in “The word orders of English and Dutch – Collective vs. Individual Checking” will be examined more closely. Goal is to get a clear understanding of what the theory explains, how this is done and how I choose to test the validity of this theory about word order differences. 2.1 Koster’s account There are four main differences in word order between English and Dutch that are discussed in Koster (1999). They can be summarized as follows: 1 English is a VO language, while Dutch is an OV language 2 Dutch has (leftward) scrambling; English does not 3 In Dutch, all adverbs are positioned to the left of VP, in English only a subclass 4 The order of English adverbs shows paradoxes of scope In what follows, I will give a short presentation of the first three phenomena and how Koster analyzes them, which will later be examined in Cantonese. 14 2.1.1 English VO, Dutch OV One of the most fundamental phenomena studied in comparative linguistics is the difference between VO and OV languages. Essentially, it’s about the relative order in which the verb and object in a language are found. This difference is most evident in subordinate clauses, as can be seen from (1) and (2). (1)a Pete paints the fence. (1)b Pete has painted the fence. (2)a Piet schildert de schuur. Pete paints the fence. (2)b Piet heeft de schuur geschilderd. Pete has the fence painted. While sentences (1)a and (2)a are exactly the same in relative word order, a difference surfaces if we compare (1)b and (2)b. Not only does this variation say something about the order of verb and object, there is a whole range of phenomena that can be linked to this VO/OV-difference (see Greenberg (1963) for an extensive list of characteristics in the word order of languages which can be traced back to the VO/OV-setting of languages), which makes it all the more intriguing as to why languages, that seem so closely related as English and Dutch, can differ so fundamentally in their syntactic composition. Koster (1999) asks himself the same thing, and finds that by using refined version of the Kayne’s Antisymmetry Theory and Chomsky’s Minimalist Program, he can come up with an effective way of explaining the word order differences between English and Dutch. But before I get into this explanation, let us look at two other phenomena that can be described with one single device. 2.1.2 Scrambling or not? A well-known, though not totally uncontroversial, operation in certain languages has been called scrambling. A number of related facts are actually meant when talking about scrambling3. These facts are: 1 double object scrambling 2 scrambling across adverbs 3 Koster (1999) only talks about scrambling across adverbs. 15 3 focus scrambling The first type of scrambling (double object scrambling) is found in German. In this language, it is possible to have different orders of internal arguments while keeping the meaning of the sentences. The reason why this is possible is because of the case system, which is very rich in German, compared to a number of other Indo-European languages like English and Dutch. This rich case system enables the scrambled orders of double objects. Scrambling across adverbs can be seen in Dutch sentences (from Koster, 1999) like (4). (4)a Jan heeft gisteren het boek gelezen. John has yesterday the book read “John read the book yesterday” (4)b Jan heeft het boek gisteren gelezen. John has the book yesterday read (meaning the same as (4)a) Here, the object moves across the adverb “gisteren” (yesterday) to a higher focalized/topicalized position. This particular kind of scrambling is the kind of scrambling Koster argues can be analyzed with ease using the aforementioned device, which will be laid out later on in this article. For a treatment of focus scrambling see Neeleman (1994). Barbiers (1995) opts for even a fourth kind of scrambling (scrambling involving movement of a constituent to the specifier of an adverbial) 2.1.3 Adverb placement As we saw in the scrambling examples, the position of the adverb is not fixed in Dutch. Another variation from English is that all adverbs that must follow the verb in English, are on the left of the verb in Dutch, as can be shown by (5) and (6)4: (5)a Ze heeft waarschijnlijk Wim gezien. She has probably Bill seen “She probably saw Bill.” (5)b 4 Ze heeft Wim gezien, waarschijnlijk. Adjusted from Koster (1999). 16 She has Bill seen, probably “She probably saw Bill.” (6)a Hij heeft Marie overal gezien. He has Mary everywhere seen “He has seen Mary everywhere.” (6)b* Hij heeft Marie gezien overal. He has Mary seen everywhere “He has seen Mary everywhere.” Though (5)a can be grammatically changed into (5)b by placing “waarschijnlijk” in sentence-final position, (6)b is totally out. Comparing the Dutch sentences with English equivalents (7), (7)a She probably saw Bill. (7)b* She everywhere saw Bill. (7)c She saw Bill everywhere. we can observe that in English it is not possible for all adverbs to be at the left of the verb. It has to be the case that Dutch and English have the following distribution of adverbs, where ADV1 are to be associated with modal adverbs like “probably” and “certainly”, whereas ADV2 are time adverbs and manner adverbs, like “yesterday” and “well” respectively. Adverb placement in Dutch and English Dutch ADV1 ADV2 V English ADV1 V ADV2 The question arises how this distribution can be accounted for in a universal structure theory, where every language has to have the same underlying structure, following Kayne (1994). We will now look at the solution Koster brings forward. For a more elaborate treatment, I refer to Koster (1999). 2.1.4 Collective vs. Individual checking In order to account for the three phenomena just discussed, Koster combines the minimalist view of feature checking (Chomsky (1995) and subsequent publications) and 17 the concept of Pied Piping (as described in Koopman & Szabolcsi (1998); for more on Pied Piping see also Ross (1967)). The crux of Koster’s analysis concerns the checking of features of arguments inside the VP, namely direct and indirect object. This can be done in two ways: either the arguments move out of the VP and check their relevant features in corresponding Spec positions higher up the tree (SpecAccP for accusative case in Koster’s account), or the arguments “pass” their features on to the VP, and the VP as a whole moves upwards, to check the percolated features. The first kind of checking is dubbed Individual Checking, where the second one is called Collective Checking. Consequently, VO languages, like English, apply Collective Checking, while OV languages, like Dutch, apply Individual Checking. The contrast in feature checking shows a number of phenomena, of which we will examine the ones that are of concern for the present issue. Observe that Koster adopts the universal base structure hypothesis laid out by Kayne (1994), which says that every language has the same underlying structure, and argues for VO to be one of the fundamental aspects of that structure. Now, the arguments in both languages, starting all on the right of the verb, have the option of either moving out of the VP for feature checking, or percolating their features to the VP, and moving collectively. When the arguments can move individually out of the VP, scrambling orders are predicted as desired for OV languages, while in the case of collective checking the close relationship of verb and object is predicted, which is to be found in VO languages. The different orders can surface in OV languages, because of the individual checking process. In contrast, VO languages can’t have the different word orders. This is due to the “cage”-like behavior of the VP, combined with the fact that the verb in these languages originate left of its internal arguments. Since it is impossible for constituents to cross the verb in the structure with the whole VP moving to check the relevant case features, there is no way these constituents end up left of the verb before any focus or topic effects may set in. For adverb placement, one can say that, for both kinds of languages, the first class of adverbs (modal adverbs) is base-generated in Spec position between CP and TP. The second class (manner and time adverbs) will follow the verb in VO languages, because the VP moves to Spec TP (checking the percolated feature along the way), whereas they 18 precede the verb in OV languages, since arguments check their features individually, leaving the verb in its place. Thus, the contrast between VO and OV languages is accounted for. Now, let’s take a look again at the parameter which the feature checking story is based upon. (8) Pied Piping Parameter In Dutch and German the complements of V check their corresponding functional heads individually; in English, the complement features are percolated to VP, which checks the functional heads collectively To generalize the parameter for it to be applicable in the universal base structure hypothesis, I’ll assume (9). (9) Pied Piping Parameter (generalized) In OV languages, the complements of V check their corresponding functional heads individually. In VO languages, the complement features are percolated to VP, which checks the functional heads collectively. Now we have the tool to see whether Koster’s account of the word order differences between VO and OV languages is really explaining all it says to explain. How this is done will be laid out in the final section of this chapter. 2.2 Testing Koster’s Theory To test the theory of Collective and Individual Checking, I will take the Generalized Pied Piping Parameter and the three word order facts that exemplify fundamental differences between VO and OV languages. These word order facts are the ones that lead Koster to assume they can be explained by one simple “device”, i.e. the Pied Piping Parameter. By generalizing the parameter proposed by Koster, we can take the theory to the next level, namely to see whether this theory really does explain word order features of VO and OV languages around the world. 19 Although this may seem trivial it is crucial for a theory based on the Universal Base Hypothesis. If one is to come up with a theory that explains word order facts of VO and OV languages, it is inevitable that such a theory will explain word order facts as distant related from English, Dutch or German, as one can imagine. The only practical restriction may be the access to the most remotely situated languages, for which we have no grammar or native speaker available. So in order to examine the validity of Koster’s theory, I have taken the generalized Pied Piping Parameter and the three word order facts mentioned earlier. Together they will be investigated for a language that has on the one hand a close resemblance to English, namely the rigidity in word order, and at the same time is very distinct from English, which is seen in the topic-prominence character of the language. The language in question is the aforementioned Cantonese. After this investigation, which will be the topic of the next chapter, a chapter will follow that goes into the implications of the word order facts, described in chapter 3. In that chapter I will explain why Koster’s theory is not fully applicable to account for word order facts in Cantonese. As mentioned in chapter 1, I will not argue for the dismissal of the theory of Collective and Individual Checking, rather I will make recommendations that shall improve the strength of the theory. 20 3 Cantonese word order Abstract In this chapter, I will examine the basic word order facts of Cantonese, a language spoken by approximately 52 million people in mainland China and some 3 million more outside of China (predominantly Vietnam and Malaysia) 5. Cantonese, otherwise known as Yue or Guangzhou, is increasingly spoken by Chinese people outside its original context, which is the province of Guangzhou or Canton. This growth is due to the economic developments of the Southern provinces, where Canton is the biggest contributor to that development. As a result, people who want to do business with Canton and its neighboring provinces, including students from the major universities of China, are encouraged to learn Cantonese. One needs to keep in mind that this type of language learning essentially involves second language learning, but by motivating people to learn the language and the resulting everyday use of the language it is maintained in being an active language. This also means that there will be regional or dialectical differences, which is not surprising considering the large population of Cantonese speakers. In this thesis, when I talk about 5 Ethnologue (2005). 21 Cantonese, I shall be talking about the Hong Kong dialect of Cantonese, which is also the dialect of Cantonese that is most commonly known in the rest of the world. By far the largest of the Chinese language family, Mandarin is the first language for most Chinese, which can account for the attention for and multitude publications about Mandarin Chinese. It is not to say that this attention for Mandarin is wrong in any way, since Mandarin does have a lot of interesting features to be explored in a generative framework. However, in order to broaden the language pool from which we are to extract a generative picture of world languages, it is important to take into examination languages that we have not so much information on as Mandarin, English, Spanish, etc. Therefore, I have chosen to investigate the basic word order facts of Cantonese. After all, the basic word order of a language is such a fundamental characteristic of a language and has also so much effects on other properties of the language, it is crucial that we have a good understanding of the basic word order before we can address other properties of that language. That’s why I will give an account of the basic word order of Cantonese in this chapter. The chapter is divided into three main parts: the first part (section 3.1) will be about the relative order between the verb and the object, one of the fundamental characteristics of a language that we have discussed in chapter 1 and 2. For the second part (section 3.2), I will examine whether it is possible to have scrambling in Cantonese, as Soh (1998) proposes, essentially defending the views of Kung (1993) and Lin (1994). In the third part I discuss the position of adverbs. The position of adverbs proofs to be the most problematic category of word order facts for the theory of Koster. The problems and their consequences will be laid out in section 3.3 and subsequently in chapter 4. 3.1 Cantonese is VO To state that Cantonese is a VO language would not be controversial, but a little needs to be said about the general behavior of the language. First of all, just like the other languages in the Chinese language family, Cantonese is topic-prominent, where most Indo-European languages are subject-prominent. This means that in topic-prominent languages, the sentence structure is built around the relation topic-comment, whereas subject-prominent languages typically base their structure around the relation subject-predicate. This is a fundamental different approach, although they have similar results in most unmarked cases. Since the topic of a sentence need not necessarily be the subject of the sentence, there are a number of so-called 22 topicalization processes possible in topic-prominent language, such as Cantonese, which give it the appearance of a free word order language. And so it is with Cantonese, where the object can be topicalized in first position (see (1) 6) or second position (see (2)). (1) Gāmyaht ge sung today CL food ngóh yíhgīng máaih-jó I la. already buy-PFV PRT “I’ve already bought food for today.” (CL = classifier; PFV = perfective; PRT = particle) (2) Ngóh Yīnggwok I meih England not-yet heui-gwo. go-EXP “I haven’t been to England.” (EXP = experience) We can counterfeit this appearance of free word order by first observing that Cantonese (as most other Chinese languages) does not have morphological case marking. This makes the constituent order very important for understanding the grammatical relations in a sentence. Aside from that, the order VOS or any order in which the subject follows the verb is strongly out. Another thing must be said about topic prominence, namely that it's existence is not so uncontroversial as it may seem from the many uses of it in the analysis of especially Sino-Tibetan languages (see for instance Matthews & Yip 1994). A different way of looking at the word order variations is to say that they are caused by scrambling operations, like we will examine in section 3.2 for Cantonese. This kind of reasoning will put Koster's theory in problematic territory, because as we will see Cantonese should be placed in the category of VO languages, and therefore cannot have scrambling in Koster's view. I will go deeper into the matter of scrambling later on, but it's important to see that the dichotomy of topic vs. subject prominence may not be so convincing to base differences between languages on it. Apart from the dependency on word order for its grammatical relations, Cantonese has another characteristic in common with English: the closeness of verb and object. This is seen in the fact that no element can intervene between verb and object, except for 6 There are several ways to write Cantonese, I have attempted to use only one writing system, namely the Yale Romanization, but there are some places where the same words seem to differ in their literal translations. These are only due to regional differences and have no impact on the word order discussion presented here. 23 durational and frequency adverbs, just like in French. This exceptional or unusual behavior will be discussed more explicit in section 3.3. Matthews & Yip (1994) observe that there are two other kinds of elements that can intervene between the verb and object, namely aspect markers (like PFV and EXP in (1) and (2)) and verbal particles (not to be confused with sentence-final particles, like in (1) at the end of the clause). However, if these elements can be analyzed as being as closely related to the verb then they can be considered constituting the verb, next to the lexical root. To see the exact nature of aspect markers and verbal particles consider Matthews & Yip (1994) which state with respect to aspect markers that: 'Aspect markers are bound forms, behaving essentially as suffixes: in their functions as aspect markers, they may not be separated from the verb, which distinguishes them from verbal particles.' (p. 200) For sake of clarity, aspect markers and verbal particles have been differentiated from each other. The point is that aspect markers need not be seen as separate elements from the verb when examining the relative order of arguments. As to verbal particles, Matthews & Yip give the following definition: 'There is a rich range of verbal particles indicating notions such as result (effect on an object) and phase of action (beginning, continuing or ending). These particles are comparable in form and function to the particles of English phrasal verbs such as up, which denotes direction in pick up but completion in eat up.' (p.210-211) 'These particles are closely related to aspect in function, and are treated as aspect markers in some descriptive works such as Kwok (1972)' (p. 211) Here, it is very clear why verbal particles are also strongly related to the verb. If we look at subordinate clauses, where the main difference between English and Dutch is seen, we stumble upon a peculiar aspect of Cantonese. Since in Cantonese 24 subordination is realized by using a conjunction structure, like in (3) 7, where we see no fundamental differences between main and subordinate clause structure. (3a) Dōng, ngóhdeih jouh hohksāang gójahnsí, sáu-tàih dihnwá jung meih while phone still we do students that-time hand-carry not làuhhàhng popular “While we were students, mobile phones were not popular yet.” (3b) Lèih Táai yānwaih taai guih sóyíh Lee Mrs because too tired so móuh làih did-not come “Mrs Lee didn't come because she was too tired.” (3c) Dáng léih fāan dou làih until you back V-PRT come gójahnsí dī sung yíhgīng dung saai that-time CL food already cold all la PRT “By the time you get back the food will all be cold.” In (3a) and (3c) we observe a conjunction both at the start and at the end of the subordinate clause, in (3b) main clause and subordinate clause are next to each other, without there being a hierarchical higher position for the main clause. This typical behavior of Cantonese can be traced back to an intuition about topic-prominent languages. Since it is more important for sentence structure to know what the topic of the sentence is, and what is said about that topic, it is of lesser importance what the main clause is and what the embedded clause. This intuition is backed up by the constraint on subordinate clause that they must be topicalized, unless they are used as an afterthought. We can conclude that subordinate clauses are not marked by a subordinate structure. Still there is the question whether the relative order between verb and object is changed in any type of sentence in Cantonese. As can be observed in the examples in (3), but also in (4) and (5), this is not the case, the VO order stays in tact in Cantonese. (4) Keoi pee He tjoh poen su Mary has given the book Mary “He has given Mary the book.” (5) John yiwai John think keoi hoo chong ming he be very smart “John thinks he is smart.” 7 (3a) en (3b) are taken From Matthews & Yip (1994), p. 293, (3c) from the same work p. 295. 25 Remember that the notion of topic prominence is not uncontroversial. The other way to explain this behavior is to call it a result of scrambling operations. This, however, as mentioned before, will lead Koster to problems. In the next section I will discuss scrambling in more detail. 3.2 Scrambling in Cantonese Despite the possibility of object movement (triggered by topicalization as I mentioned in the last section), there is no way scrambling can be applied in Cantonese, neither for objects across other arguments (double object scrambling) nor objects scrambling across adverbs. An interesting case seems to be the behavior of durational and frequency adverbs in Chinese, as presented by Soh (1998). In her master thesis she argues for object scrambling in (Mandarin) Chinese, for which she gives the following evidence8: (4a) (4b) (5a) (5b) I have invited that person twice. wo qing-guo [OBJ na-ge ren] [DFP liang ci] I invite-PFV [OBJ that-CL person] [DFP two time] Twice, I have invited that person. Wo qing-guo [DFP liang ci] [OBJ na-ge ren] I invite-PFV [DFP two time] [OBJ that-CL person] I have visited USA twice. wo qu-guo [OBJ Meiguo] [DFP liang ci] I visit-PFV [OBJ USA] [DFP two time] Twice, I have visited USA wo qu-guo [DFP liang ci] [OBJ Meiguo] I visit-PFV [DFP two time] [OBJ USA] Soh observes that the order of the object and the frequency adverb seem to be under the influence of scrambling. However, there is something to be said against such an analysis. Take for example the Cantonese counterparts of (4)-(5): 8 (1) and (2) in Soh (1998). 26 (6a) I have invited that person twice. ngo ceng-gwo [OBJ go-go-jan] [DFP loeng-ci] 1SG invite-EXP [OBJ that-CL-person] [DFP two-time] (EXP = experiential aspect) (6b) Twice, I have invited that person. [DFP loeng-ci (la)], ngo ceng-gwo [OBJ go-go-jan] [DFP two-time PRT], 1SG invite-EXP [OBJ that-CL-person] (PRT = utterance particle) (7a) (7b) I have visited USA twice. ngo heoi-gwo [OBJ MeiGwok] [DFP loeng-ci] 1SG go-EXP [OBJ USA] [DFP two-time] Twice, I have visited USA. [DFP loeng-ci (la)], ngo heoi-gwo [OBJ MeiGwok] [DFP two-time (PRT)], 1SG go-EXP [OBJ USA] For Cantonese, the order in which the frequency adverb is before the object seems odd at best. Moreover, the subject and the verb are intervening, indicating a topicalizing operation, instead of being an example of a scrambling operation. It is striking, however, that albeit no instance of scrambling, it is possible for Mandarin to have two orders of the object and the frequency adverb (more generally the DFP, Durational and Frequency Adverb, as used by Soh). Mandarin is just as VO as Cantonese, so it just shows that Mandarin and Cantonese have underlying differences in their syntactic set up. This is regularly (see Matthews & Yip (1994) for instance) seen as superficial or subtle differences, but as will be apparent by above distinction between the two languages, it is quite fundamental in its occurence. Other examples that indicate there is no scrambling in Cantonese are (8) and (9). (8a) John sek guo [Mary] [leung ci]. “John kissed Mary often twice.” (8b) John sek guo [Mary] [chiew guo] [leung ci] “John kissed Mary twice often.” (9) Peter fan sam chi sik leung guo ping guo. “Peter eats three apples twice.” “Peter eats twice three apples.” In (8) we see that although “chiew guo” can intervene between “Mary” and “leung ci”, the order of the object and the DFP stays in tact. In (9) one and the same sentence is used to convey the meaning of the two English orders. 27 Another point of concern is the notion of object shift. It would be wrong to consider object topicalization as an instance of object shift, since it does not meet the basic requirements Haider (2005) gives in his treatment of the Scandinavian languages. Object shift is correlated with verb movement; in Cantonese the verb stays in situ. Also, the relative order of the arguments stays intact in the case of object shift, which is obviously not the case with the topicalization of objects in Cantonese. An interesting feature of the Cantonese double object structure is the meaning correlation with the order of the indirect and direct object. Typically, the direct object must be placed directly following the verb (as can be derived from the basic word order). Hence, we get (8)a and not (8)b. (8)a keoi bei-zo go-bun-syo 3sg give-PFV that-CL-book Mary bei-zo Mary go-bun-zo give-PFV Mary that-CL-book (8)b* keoi 3sg Mary “He has given the book to Mary.” The only way this order of indirect and direct object can be altered is when the direct object is particularly long (or heavy), like in (9). (9) Ngóh béi I léih géi chīn give you a-few thousand mān tùhngmàaih yāt jēung gēipiu. dollar air-ticket plus one CL “I’ll give you a few thousand dollars plus an air ticket.” Surprisingly, it is also possible for a number of verbs, to have two orders relating to two meanings of the verb. Among them are “lend”, “teach” and “punish”. See for example (10). (10)a Ngóh I je-jó yāt chīn mān kéuih. lend-PFV one thousand dollar him “I lent one thousand dollars to him.” (10)b Ngóh I je-jó kéuih yāt chīn mān. lend-PFV him one thousand dollar “I lent one thousand dollars from him.” 28 The question is whether the prepositions belonging to the verb have been incorporated into the verb. This would then explain the need for different ordering of the argument to distinguish between the two meanings of the verb. We can conclude, however, that we do not find any operation that resembles scrambling, like we do in Dutch or German. 3.3 Adverb placement in Cantonese As for the case of adverb placement, I already mentioned that durational and frequency adverbs are the only elements (putting aside the aspect markers and verbal particles issue) that can intervene between the otherwise rigidly connected verb and object. Now, we need to look more explicitly at the nature of those kinds of adverbs. First of all, let’s see how auxiliaries affect the word order in Cantonese. Auxiliaries typically precede the main verb but they behave like other verbs (serializing is also possible), with the exception of aspect markers or verbal particles, which they do not take. Functionally, auxiliaries like “have”,“can”, “will” and “would” act as modal verbs, preceding the verb. See for example (11). (11) Bīngo wúih who jūngyi nī would like júng yàhn ge jēk? this kind person PRT PRT “Who would like a person like that?” (12) Ngóh yìhgā I now yiu hąōi-wúi. need hold-meeting “I have to attend a meeting now.” Mostly, adverbs precede the main verb, and in case of an auxiliary, they precede the auxiliary, which is a striking difference with most Indo-European languages. It is however, also possible to have an adverb intervening between auxiliary and main verb, as in (13). (13) Dī syū yīnggōi CL book should jīkhāak wàahn. immediately return “We should return those books immediately.” Before we get to the analysis of these kinds of structures, we first need to say something general about adverbs in Cantonese. 29 As we have seen in English and Dutch, a simplified division can be made in the classes of adverbs, which show different distribution in VO and OV languages. In Cantonese we have a more detailed division of adverbs and behavior with respect to the order with the verb. Adverbs in Cantonese can be divided in predicate adverbs, which describe the manner of action and apply to the verb phrase. These adverbs typically follow the verb (see for details about the gam-construction Matthews & Yip (1994) p. 181). The second class of adverbs is sentential adverbs, which describe the circumstances of events and states, such as time and probability, and apply to the whole sentence or clause. These adverbs normally precede the verb in Cantonese. Observing the contrast with English, is that time adverbs and modal adverbs seem to fall in the same category in Cantonese, where in English there is a clear distinction in distribution of the two adverbs. Next to this discrepancy, there also seems to be a problematic characteristic of the aforementioned durational and frequency adverbs. These adverbs behave exceptional, because they must follow the verb in Cantonese, which makes Cantonese resembling French rather than English (see Pollock (1989) for a discussion about English and French word order). The problem resides in these two points: how can it be that in some cases, adverbs seem to be quite clearly divided (predicate adverb vs. sentential adverbs), while at the same time it is possible for certain adverbs to intervene between auxiliary and main verb (like “immediately” in (13)), but also for durational and frequency adverbs to follow the verb (essentially resembling the French situation) like in (14) and (15)? (14) Ngóh hohk-jó I baat lìhn learn-PFV eight years Yīngmán. English “I’ve been learning English for eight years.” (durational) (15) Ngóhdeih gin-gwo we géi chi see-EXP a-few times mihn. face “We’ve met a few times.” (frequency) 30 It seems that we need to get a closer look at the adverbs in Cantonese to determine the exact nature as to whether Cantonese can be considered a VO language like English, on the basis of the Pied Piping Parameter. As we have seen in the discussion about object scrambling and in the case mentioned above, durational and frequency adverbs seem to pose problems to the unified account of Koster’s theory. Though it is very marginal to have a DFP precede an object in Cantonese (6(b) and 7(b)), the fact that there is some option between the order of object and DFP in Mandarin poses serious problems to the theory proposed by Koster. In chapter 4 I will address this problem. Before I continue to chapter 4, where I will sum up the implications of the word order facts in Cantonese for the theory of Collective and Individual Checking, let’s see how Cantonese behaves with respect to the adverbs Koster uses to distinguish English and Dutch (and consequently VO and OV languages). 3.4 A closer look at adverbs in Cantonese (16)9 He probably worked, yesterday. Hij heeft [ADV1 waarschijnlijk] [ADV2 gisteren] [V gewerkt] he has [V worked] ADV1 (17a) 10 [ADV1 probably] ADV2 [ADV2 yesterday] V He has [ADV1 probably] [V worked], [ADV2 yesterday] ADV1 V (18a) ADV2 English ADV2 (17b)11*He has [ADV1 probably] [ADV2 yesterday] ADV1 Dutch [V worked] V He probably read a book. keoi [ADV1 honang] [V tai-zo] jat-bun-syu 3sg [ADV1 probably] [V read-PFV] one-CL-book ADV1 V Cantonese (18b) He read a book yesterday. [ADV2 camjat] keoi [V tai-zo] jat-bun-syu [ADV2 yesterday] 3sg [V read-PFV] one-CL-book ADV2 9 10 11 V (91)a in Koster (1999). (91)c in Koster (1999). (91)b in Koster (1999). 31 Cantonese (18c) He read a book yesterday. keoi [ADV2 camjat] [V tai-zo] jat-bun-syu 3sg [ADV2 yesterday] [V read-PFV] one-CL-book ADV2 V Cantonese The Cantonese sentences ((18a)-(18c)) show that both “probably” and “yesterday” precede the verb, unlike in English. It is self-evident, although these adverbs are not in the same sentence, that there is no way these adverbs can move to the right of the verb by some extraposition operation, because although leftward effects like topicalization are abundant in Cantonese, rightward extraposition is prohibited 12. Let's see how other kinds of adverbs behave in Cantonese, using the hierarchy Cinque (1999) proposes to determine categories of adverbs that precede and follow the verb. In order of high to low adverb, following Cinque's order and labeling of the adverbs, I shall examine the following adverbs: loosatgong “frankly” speech act mood adverb hoochoi “fortunately” evaluative mood adverb honang “probably” epistemic modal adverb camjat “yesterday” past tense adverb pingsi “usually” habitual aspect adverb hoodohsi “often” frequentative aspect adverb faai “fast”/“quickly” celerative aspect adverb yiking “no longer” terminative aspect adverb moeici/sengyat “always” perfective aspect adverb taandi “almost” prospective aspect adverb dunsi “briefly” durative aspect adverb ming/lo “well” voice adverb What I've done is put each adverb (I have also put in “probably” an “yesterday” for completeness) in all possible positions relative to the subject, verb, object and negation. Then I checked whether these orders can occur in Cantonese. These are the results of that investigation. 19a. Frankly we cannot see the birds. loosatgong 12 ngodei tai-mdoo dizheuk Also excluded in Kayne's Antisymmetry Theory. 32 frankly we see-NEG the birds 19b. We frankly cannot see the birds. Ngodei loosatgong tai-mdoo dizheuk we frankly see-NEG the birds 19c.* We can frankly not see the birds. 19d.* We cannot frankly see the birds. 19e.* We cannot see frankly the birds. 19f.* We cannot see the birds frankly. 20a. Fortunately, we cannot see the birds. Hoochoi, ngodei Fortunately, we tai-mdoo dizheuk see-NEG the birds. 20b. We fortunately cannot see the birds. Ngodei hoochoi tai-mdoo dizheuk We see-NEG the birds fortunately 20c.* We can fortunately not see the birds. 20d.* We cannot fortunately see the birds. 20e.* We cannot see fortunately the birds. 20f.* We cannot see the birds fortunately. 21a. Probably he read a book. honang keoi tai-zo probably he read-PFV jat-bun-syu one-CL-book 21b. He probably read a book. keoi honang tai-zo jat-bun-syu he probably read-PFV one-CL-book 21c.* He read probably a book. 21d.* He read a book probably. 22a. Probably the girl wrote a letter. honang go-neoi-zai se-zo jat-fung-seon probably that-CL-girl write-PFV one-CL-letter 22b. The girl probably wrote a letter. go-neoi-zai honang se-zo jat-fung-seon that-CL-girl probably write-PFV one-CL-letter 22c.* The girl wrote probably a letter. 22d.* The girl wrote a letter probably. 33 23a. Probably mother made pancakes. Honang mama zyu-zo baangik probably mother make-PFV pancakes. 23b. Mother probably made pancakes. mama honang zyu-zo baangik mother probably make-PFV pancakes 23c.* Mother made probably pancakes. 23d.* Mother made pancakes probably. 24a. Probably we cannot see the birds. Honang, ngodei tai-mdoo dizheuk probably, we see-NEG the birds 24b. We probably cannot see the birds. Ngodei honang tai-mdoo dizheuk We see-NEG the birds probably 24c.* We can probably not see the birds. 24d.* We cannot probably see the birds. 24e.* We cannot see probably the birds. 24f.* We cannot see the birds probably. 25a. Yesterday he read a book. camjat keoi tai-zo jat-bun-syu yesterday he read-PFV one-CL-book 25b. He yesterday read a book. keoi camjat tai-zo jat-bun-syu he yesterday read-PFV one-CL-book 25c. He read yesterday a book. [interpretation: “he read yesterday's book.”] keoi he tai-zo camjat read-PFV yesterday jat-bun-syu one-CL-book 25d.* He read a book yesterday. 26a. Yesterday the girl wrote a letter. camjat go-neoi-zai se-zo jat-fung-seon yesterday that-CL-girl write-PFV one-CL-letter 26b. The girl yesterday wrote a letter. go-neoi-zai camjat se-zo jat-fung-seon that-CL-girl yesterday write-PFV one-CL-letter 34 26c. The girl wrote yesterday a letter. [interpretation: “The girl wrote a letter from yesterday”] go-neoi-zai se-zo camjat jat-fung-seon that-CL-girl write-PFV yesterday one-CL-letter 26d.* The girl wrote a letter yesterday. 27a. Yesterday mother made pancakes. camjat mama zyu-zo baangik yesterday mother make-PFV pancakes 27b. Mother yesterday made pancakes. mama camjat zyu-zo baangik mother yesterday make-PFV pancakes 27c.* Mother made yesterday pancakes. 27d.* Mother made pancakes yesterday. 28a. Yesterday he has given the book to Mary. camjat keoi bei-zo go-bun-syo Mary yesterday he give-PFV the-CL-book Mary 28b. He yesterday has given the book to Mary. keoi camjat bei-zo go-bun-syo Mary he yesterday give-PFV the-CL-book Mary 28c. He has given yesterday the book to Mary. keoi bei-zo camjat Mary go-bun-syo he give-PFV yesterday Mary the-CL-book 28d.* He has given the book to Mary yesterday. 29a. Usually we cannot see the birds. pingsi13, ngodei tai-mdoo dizheuk usually, we see-NEG the birds 29b. We usually cannot see the birds. ngodei pingsi tai-mdoo dizheuk we usually see-NEG the birds 29c.* We can usually not see the birds. 29d.* We cannot usually see the birds. 29e.* We cannot see usually the birds. 13 “pingsi” can also be substituted by “kingseung”, “patlou”, “tongseung”. 35 29f.* We cannot see the birds usually. 30a. Often we cannot see the birds. hoodohsi ngodei tai-mdoo dizheuk often we the birds see-NEG 30b. We often cannot see the birds. ngodei hoodohsi tai-mdoo dizheuk we often see-NEG the birds 30c.* We can often not see the birds. 30d.* We cannot often see the birds. 30e.* We cannot see often the birds. 30f.* We cannot see the birds often. 31a. Fast he read the book. hou-faai tai-zo-bun-syu very-fast read-PFV-CL-book 31b. He read the book very/quite fast. tai-syu-tai-dak hou/gei-faai read-book-read-DAK very/quite-fast V-OBJ-V-DAK intensifier-ADJ 31c. He read the book very fast. *tai-syu-(hou)-faai read-book-(very)-fast 32a.* No longer we can(not) see the birds. 32b. We no longer can(not) see the birds. Ngodei yiking tai-mdoo dizheuk We no longer see-NEG the birds 32c.* We can no longer (not) see the birds. 32d.* We can(not) no longer see the birds. 32e.* We can(not) see no longer the birds. 32f.* We can(not) see the birds no longer. 33a.* Always we cannot see the birds. 33b. We always cannot see the birds. ngodei moei-ci do-tai-mdoo dizheuk we each-time NEG-see-NEG the birds ngodei seng-yat do-tai-mdoo dizheuk we all-day NEG-see-NEG the birds 36 33c.* We can always not see the birds. 33d.* We cannot always see the birds. 33e.* We cannot see always the birds. 33f.* We cannot see the birds always. 34a.* Almost we cannot see the birds. 34b. We almost cannot see the birds. ngodei taandi14 tai-mdoo dizheuk we almost see-NEG the birds 34c.* We can almost not see the birds. 34d.* We cannot almost see the birds. 34e.* We cannot see almost the birds. 34f.* We cannot see the birds almost. 35a.* Briefly we cannot see the birds. 35b. We briefly cannot see the birds. Ngodei dunsi tai-mdoo dizheuk We briefly see-NEG the birds 35c.* We can briefly not see the birds. 35d.* We cannot briefly see the birds. 35e.* We cannot see briefly the birds. 35f.* We cannot see the birds briefly. 36a.* Well he read the book. 36b.* He well read the book. 36c. He read well the book. keoi tai-ming-zo jat-bun-syu he read-understand-PFV one-CL-book 36d. He read the book well. keoi he tai-zo read-PFV jat-bun-syu lo one-CL-book well 37a.* Well the girl wrote a letter. 37b.* The girl well wrote a letter. 37c. The girl wrote well a letter. 14 go-neoi-zai se-hou-zo jat-fung-seon that-CL-girl write-well-PFV one-CL-letter “keefoe” is more formal. 37 37d. The girl wrote a letter well. go-neoi-zai se-zo jat-fung-seon lo that-CL-girl write-PFV one-CL-letter well 38a.* Well mother made pancakes. 38b.* Mother well made pancakes. 38c. Mother made well pancakes. mama zyu-hou-zo baangik mother make-well-PFV pancakes 38d. Mother made pancakes well. mama zyu-zo baangik lo mother make-PFV pancakes well From the above results the following generalization can be made: all adverbs down to “dunsi” (briefly – the 21st in Cinque's hierarchy of 30 types of adverbs) can only occur in preverbal position; only “ming/lo” (well – the 26th in Cinque's hierarchy) can and must occur postverbally. Some remarks have to be made though about sentence (25) and (26). For example, “camjat” (yesterday) seems to follow the verb in (25c) and (26c), but here the adverb does not act as modifier of the whole clause or sentence. Instead it modifies only the noun “book” and “letter” respectively, which is shown in the interpretation. This explains the postverbal occurrence. In (28) this line of reasoning cannot explain the postverbal occurrence of “camjat”, since it clearly refers to the whole sentence. This is the only counterexample that I have found to the hierarchical order from Cinque (1999), and I therefore conclude this as exceptional, rather than to dismiss the hierarchical ordering of adverbs, and it remains to be adequately dealt with. Another explanation needs to be sought for the postverbal position of “faai” in (31b). The unmarked position of this adverb is preverbal, as (31a) shows. At the same time, (31c) indicates that the postverbal occurrence of “faai” is very restricted. The adverb can only occur postverbally in a clause with a duplicated verb compound. This leads us to consider that the compounding takes up a position, which is normally reserved for adverbs, or at least makes that position inaccessible for adverbs. From these results we can conclude that Cantonese behaves like English with regard to adverb placement, namely most adverbs appear to the left of the verb and some appear to the right. Recall that in Koster's example of adverb placement, here 38 repeated as (39)15, “yesterday” is used as an afterthought to the clause before the comma. (39) He has probably worked, yesterday. In Cantonese it is also possible to place adverbs as an afterthought, like in (40) 16. (40) Taai gwai la. Hóu leng wo, daahnhaih. Too dear PRT very nice PRT however “It's too expensive. Very nice, though.” As to the problem DFPs pose for the theory, it is likely that these adverbs are in the same category as “well” and can perhaps be considered the way I analyzed the postverbal occurrence of “yesterday”, namely as modifying the object it precedes. Cantonese is thus similar to English as its division in two categories of adverbs. If one places the DFP in the object-modifier class of adverbs, just like “well”, it is possible to state that in both languages sentential adverbs are class 1 adverbs, while objectmodifier adverbs are class 2 adverbs. The only thing left to determine is the exact position of a class 2 adverb. In both languages it can be either left (yesterday) or right (ming/lo, faai) of the object; for the English equivalents of the discussed Cantonese examples, see (41)-(43). (41a) John read a book, yesterday. (41b)* John read yesterday a book. (41c) John reads yesterday's book. (42a) John read a book fast. (42b)* John read fast a book. (42c) John read a fast book. (43a) John read a book well. (43b)* John read well a book. (43c) John read a good book. If the adverb is on the right of the object, the interpretation must be sentential, whereas if it is on the left of the object, the only possible interpretation is as in the c-sentences in (41)-(43). The difference between English and Cantonese that in English this modifying 15 16 Koster (1999), sentence (91c) on p. 37. From Matthews & Yip (1994), p. 294. 39 relationship between the adverb and the object has to be marked in an adjectival morphology, whereas in Cantonese the adverb retains its adverbial morphology. In the next chapter I will show how these observations should be turned into an analysis that can have a place in the theory proposed by Koster. 4 Implications of Cantonese word order facts Abstract In the previous chapter I have unraveled some basic characteristics of Cantonese. This language behaves like Koster would predict with his theory of Collective and Individual Checking, although Koster's classification of adverbs needs to be made more precise. In this chapter I will summarize my findings and point to where the classification needs to be made. 4.1 Koster evaluated I will start with what Koster rightfully predicts for Cantonese. Cantonese is rigidly VO. Although topicalization is a fruitful operation, the main order is by far the one where the 40 verb precedes the object. This is also seen in subordinate clauses, like in (5) from chapter 3, repeated here as (1): (1) John tells me that he is smart. John tung ngo [V gong] [OBJ keoi hou cong-ming] John with me [V say] [OBJ 3SG very smart] Moreover, as I have shown in chapter 3, the verb and object are very strictly bonded together in the syntactic structure, so that they may very rarely be intervened by any material. This characteristic is just what Koster would predict for a VO language. Although this is quite trivial a statement, it is important to understand that a language could well be VO, while at the same time OV order may be found in another clause type. The only way the VO order can be changed in Cantonese is by ways of topicalization. This can move the object to the left of the verb. Though this may seem to undermine the rigidness of the language, one must keep in mind that topicalization is also available in English; nonetheless English is regarded as being a VO language. Secondly, I argue contra Kung (1993), Lin (1994) and Soh (1998) that there is no such a thing as object scrambling in Chinese. There is something mysterious about the order of an object and a DFP, which may seem as if we have a case of scrambling, but as I have shown in chapter 3, this scrambling account from Mandarin should extend to the closely related Cantonese. Since this is not the case, and since the “scrambling” cases presented by Soh (1998) do not necessitate a scrambling analysis, I conclude that there is no object scrambling in Cantonese or Mandarin. What is striking is the possibility of two orders in Mandarin between the object and the DFP. As I have argued in chapter 3, this exceptional behavior, as well as the behavior of other adverbs in English and Cantonese can be explained in a unified account of adverbs. This account, that will fit inside Koster's theory, will be discussed in the next section. 4.2 Adverb placement in Koster's theory Adverbs in the Universal Grammar are to be divided in two classes: class 1 preverbal adverbs class 2 postverbal adverbs 41 Class 2 adverbs can further be divided in pre- and postobject adverbs. If the adverb is on the right of the object, the interpretation must be sentential, whereas if it is on the left of the object, the only possible interpretation is object-modifying. All preverbal adverbs are sentential. This classification explains when an adverb is placed in what position, namely by ways of possible interpretations. In preverbal position, adverbs can only have a sentential interpretation, whereas postverbal adverbs can have either a sentential interpretation (in post-object position) or an object-modifying interpretation (in pre-object position). Adverbs that can have no interpretation in a certain position can never be in that position, whereas adverbs that can, determine their interpretation on their position. Some languages may also morphologically mark the adverb for being in object-modifying position like English, where others do not mark their adverbs for this. Cantonese is an example of such a language. To come back to the mysterious word order variations in Mandarin Chinese, as discussed by Soh (1998), she herself observes that there are differences in interpretation between sentences (2a) and (2b)17. (2a) wo qing-guo [quanbu de xuesheng] I invite-PFV [all DE student] [liang ci] [two time] 'I have invited all students twice.' (2b) (i) all students >> two times (ii) two times >> all students wo qing-guo [liang ci] [quanbu de xuesheng] I invite-PFV [two time] [all DE student] 'Twice, i have invited all students.' (i) ?*all students >> two times (ii) two times >> all student In Soh's words, 'when the object follows the DFP [like in (2b)], it can only be interpreted as having narrow scope...[and] when the object precedes the DFP [like in (2a)]... both group and distributive readings are available.' It would take further investigation to determine the precise role of the quantifier phrases in these constructions, but it stands without a doubt that the interpretation directs whether an adverb (be it a DFP or another) can occur pre- or post-object. This is not a question of scrambling procedures, but to what extent an interpretation is possible. 17 Soh's (7) and (8) on p. 37. 42 One problem for Koster is the supposed topicalization of adverbs like “immediately” which strands them between the auxiliary and the main verb. This must have some topicalization background, but I do not have the time or space to get into this matter in more detail, so this must be reserved for future research. 4.3 Conclusion and recommendations for further research I conclude that Cantonese behaves the way Koster (1999) predicts it does, with a more precise classification of adverbs. I have given this amendment to Koster's classification that will help his theory to explain the word orders of more world languages than just the ones that are closely related to each other. Still not everything can be accounted for with this classification. In particular the question why “immediately” can intervene between the auxiliary and the main verb remains a topic for future research. Another interesting question is the role of quantifier phrases with respect to the different possible interpretations. These possible interpretations direct the possibilities of adverbs in certain positions, and therefore require more investigation. As a final remark, let me point out that it would serve future case studies of Cantonese if a unified writing system of the language is used. This facilitates the analysis of translations by different native speakers. There tend to be regional differences in vocabulary, but these are mainly superficial and will only matter if these regional differences are the topic of one's study. 5 Samenvatting in het Nederlands In deze doctoraalscriptie heb ik de theorie van Jan Koster in zijn artikel “The word orders of English and Dutch, Collective vs. Individual Checking” uit (1999) getest op zijn toepasbaarheid voor het Kantonees. Het centrale idee van Koster's theorie is het verklaren van de verschillen tussen VO en OV talen d.m.v. het aannemen van een parameter, genaamd de Pied Piping Parameter. Deze parameter is gebaseerd op eerdere bevindingen van o.a. Vanden Wyngaerd (1989), Zwart (1993, 1994 en 1997), Kaan (1992) en Den Dikken (1996). Dankzij deze parameter weet Koster te verklaren waarom VO en OV talen op vier distincte manieren van elkaar verschillen. Drie van die vier verschillen worden in deze scriptie getest voor het Kantonees, omdat Koster tot nu toe alleen de verschillen tussen talen wist te verklaren die dicht aan elkaar gerelateerd zijn, namelijk het Engels en het Nederlands. 43 De drie verschillen zijn achtereenvolgens: 1. de volgorde van het werkwoord en het object (VO vs. OV), 2. scrambling of geen scrambling mogelijk, 3. plaatsing van adverbia (voor of na het werkwoord). Eerder genoemde parameter ziet er als volgt uit: Pied Piping Parameter – in het Nederlands checken de interne argumenten hun corresponderende features individueel. – in het Engels checken de interne argumenten hun corresponderende features colleftief, na deze te hebben doorgegeven aan de VP (verbale frase). De analyse verloopt als volgt: In het Nederlands hoeven de interne argumenten niet binnen de VP te blijven. Ze hebben meer bewegingsvrijheid dan in het Engels. In het Engels ontvangen VPs de features van hun interne argumenten. Deze features worden vervolgens gechecked door de gehele VP te verplaatsen naar de posities waar de features gechecked kunnen worden. Ik laat zien dat het Kantonees zich gedraagt zoals Koster (1999) dat voorspelt. De enige kanttekening maak ik bij de classificatie van adverbia, die naar mijn mening preciezer geformuleerd moet worden. Deze classificatie ziet er als volgt uit: Adverbia in de Universele Grammatica kunnen worden verdeeld in 2 klassen: klasse 1 preverbale adverbia klasse 2 postverbale adverbia Klasse 2 adverbia kunnen verder worden verdeeld in pre- and post-objecte adverbia. Als het postverbale adverbium rechts van het object staat kan de interpretatie alleen sententieel zijn, terwijl als het links van het object staat, het alleen een objectmodificerend kan zijn. Alle preverbale adverbia zijn sententieel. Deze classificatie verklaart wanneer een adverbium in welke positie geplaatst wordt, namelijk op grond van mogelijke interpretaties. In preverbale positie kunnen adverbia alleen een sententiele interpretatie krijgen; postverbale adverbia hebben de keuze tussen een sententiele interpretatie (in post-objecte positie) en een objectmodificerende interpretatie (in pre-objecte positie). 44 Sommige talen kunnen het adverbium ook morfologisch markeren voor het staan in een object-modificerende positie, zoals het Engels en het Nederlands. Andere talen markeren hun adverbia niet in zulke posities. Het Kantonees is een voorbeeld van zo'n taal. Appendix I: Other Cantonese sentences Below you will find the sentences from my investigation that were not mentioned in the chapters, together with their translation, but without direct glosses. 1. He has given Mary the book, yesterday. Keoi camyat pee tjoh poen su Mary. 2. John thinks he is smart. John yiwai keoi hoo chong ming. 3. John tells me he is smart. John tong ngo gong, whaa keoi hoo chong ming. 4. John tells me a good story. John gong hoo goe sie pee ngo teng. 5. I have invited that person twice. Ngo cheng guo keoi leung ci. 45 6. Twice, I have invited that person. Leung ci a, ngo yiking cheng guo keu leung ci. 7. I have visited China twice. Ngo heu guo tong kok leung ci. 8. Twice, I have visited China. Leung ci a, wo yiking heu guo tong kok leung ci. 9. I think that John likes to read a book. Ngo lam John seung tai yat poen su. 10. When you listen carefully, you can hear the birds sing in the garden. Yu guo lee sai sam die teng, lee tjouw woei teng dow tjeuk tjak kieuw. 11. Without saying anything, Peter ran to the station to catch the last train. Keoi yat geu waa do mow kong. Peter tjouw pauw heu fo ce tjaam zhui zui hou koh fo ce. 12. Not being interested in biology, Peter studies history. Keoi dui san mat hok mow hing cheu, Peter hai dok lik sie kaa. 13. Being not interested in biology, Peter studies history . Yanwai keoi dui san mat hok mow hing cheu, tjouw dok lik sie. 14. Working not much, Peter flunked. [to flunk = zakken voor een toets/tentamen.] Yan wai keoi m kan lik, soh yie sat paai. 15. Not working much, Peter flunked. Yan wai keoi m kan lik, soh yie sat paai. 16. John was not arrested. John mow peei kaw law. 17. John not being arrested, he drove home to eat dinner. John mow peei kaw law, keoi tjaa che faan ok kee sik maan chaan. 18. John is not happy. John m hoi sam. 19. John is not happy about his grades. John yan wai keoi die fan sow m hoi sam. 20. John likes not Mary. John m tjong yi mary. 21. I thought I wouldn’t be able to sleep in this room. I lam ngo mow ho nang hee li kaan fong fan dak tjeuk. 22. He had thought it unnecessary to take action concerning my letter. Keoi lam m seu yiew kwan yu ngo fong sun tjoo yam ho choo si. 23. He has said he did not wish to take action concerning my letter. Keoi waa keoi m hee mong kwan yu ngo fong sun tjoo yam ho choo si. 24. I believe that Tom very smart bought a book for his son. Ngo kok tak Tom hoo chong ming, sik maai poen su pee keoi tjai. [interpretation: “It is very smart that Tom bought a book for his son.”] 46 25. I believe that Tom bought a book very smart for his son. Ngo kok tak Tom yong hoo chong ming ko fong faat mai poen su pee keoi tjai. [interpretation: “Tom bought a book for his son in a very smart way.”] 26. He succeeds in stopping smoking. Keoi sing gong lieuw kaai yien. 27. He does not succeed in stopping smoking. Keoi m sing kong kaai yien. References Barbiers, S., 1995a, The Syntax of Interpretation, Ph.D. diss, Leiden University. Barbiers, S., 1995b, Another case of scrambling in Dutch, In: Linguistics in the Netherlands 12, p. 13-24, John Benjamins, Amsterdam. Barbiers, S., 2000, The right periphery in SOV languages: English and Dutch, In: P. Svenonius (ed.), The derivation of VO and OV, John Benjamins, Amsterdam, p. 181-218. Belletti, A., 1991, Generalized Verb Movement, Rosenberg, Torino. Chomsky, N., 1994, The Minimalist Program, MIT Press, Cambridge, Massachussetts. Cinque, G., 1999, Adverbs and functional heads, a cross-linguistic perspective, Oxford University Press, New York. 47 Costa, J., 1996, Adverb positioning and V-movement in English: some more evidence, In: Studia Linguistica 50:1. Dikken, M. den, 1996, The Minimal Links of Verb (Projection) Raising, In: W. Abraham, S. Epstein, H, Thráinsson, and J.-W. Zwart, Eds., Minimal Ideas, John Benjamins, Amsterdam. Gordon, Raymond G., Jr. (ed.), 2005, Ethnologue: Languages of the World, 15th Ed., Dallas, Tex.: SIL International. Online version: http://www.ethnologue.com/ . Greenberg, J.H., 1963, Some Universals of Language With Particular Reference to the Order of Meaningful Elements, In: Universals of Language, ed. J.H. Greenberg, p. 58-90. Haider, H., 1997, Extraposition, In: Rightward Movement, D. Beerman, D. LeBlanc & H. van Riemsdijk (eds.), p. 115-152, Amsterdam, John Benjamins. Haider, H, 2004, Pre- and postadverbials in OV and VO, In: Lingua, vol. 114, p. 779-807, Elsevier, Amsterdam. Haider, H., 2005, How to turn German into Icelandic – and derive the OV-VO contrasts, Journal of Comparative Germanic Linguistics 8: p. 1-53, Kluwer Academic Publishers. Haider, H, & I. Rosenberg, 2003, Scrambling – Non-triggered Chain formation in OV-Languages, In: Journal of Germanic Linguistics 15, p. 203-267. Haspelmath, M., H.J. Bibiko (et al.), 2005, The World Atlas of language structures, Oxford University Press, Oxford. Iatridou, S., 1990, About Agr(P), In: Linguistic Inquiry 21, 4, 551-577. Johnson, K., 1991, Object Positions, In: Natural Language and Linguistic Theory 9, p. 577-636. Kaan, E., 1992, A Minimalist Approach to Extraposition, MA Thesis, University of Groningen, Groningen. Kayne, R.S., 1994, The Antisymmetry of Syntax, MIT Press, Massachussetts. Koopman, H. & A. Szabolcsi, 1998, Verbal complexes, Ms., UCLA, Los Angeles. Koster, J., 1999, The Word Orders of English and Dutch: Collective vs. Individual Checking, In: W. Abraham (ed.), Groninger Arbeiten zur germanistischen Linguistik, University of Groningen, Groningen, 1999, p. 1-42, http://www.let.rug.nl/~koster/1999.htm . 48 Kung, H.I., 1993, The Mapping Hypothesis and postverbal structures in Mandarin Chinese, Ph.D Dissertation, University of Wisconsin-Madison. Larson, R., 1988, On the double object construction, In: Linguistic Inquiry 19, p. 335-397. Lin, J.-W., 1994, Object expletives, definiteness effect and scope interpretation, In: Proceedings of the North East Linguistic Society 24 (1), Merce Gonzalez (ed.), University of Massachussetts, Amherst. Matthews, S., & V. Yip, 1994, Cantonese: A comprehensive grammar, Routledge, London. Mulder, R. & R. Sybesma, 1992, Chinese is a VO language, In: Natural Language and Linguistic Theory, 10, p. 439-476, Reidel, Dordrecht. Neeleman, A., 1994, Complex predicates, diss., OTS Dissertation Series, Utrecht. Ouhalla, J., 1990, Sentential Negation, Relativised Minimality, and the Aspectual Status of Auxiliaries, In: The Linguistic Review 7, 183-231. Pollock, J.-I., 1989, Verb movement, Universal Grammar, and the structure of IP, In: Linguistic Inquiry 20, p. 365-424. Ross, J.R., 1967, Constraints on variables in syntax, doct. diss., MIT, published as Infinite syntax!, Ablex, Norwood, 1986. Soh, H. L., 1998, Object Scrambling in Chinese, MIT Working Papers in Linguistics, Massachussetts. Tai, J. H-Y, 1985, Temporal sequence and Chinese word order, In: J. Haiman, ed., Iconicity in syntax: Proceedings of a Symposium on Iconicity in Syntax, Stanford, June 24-6, 1983, John Benjamins, Amsterdam, p. 49-72. Vanden Wyngaeerd, G., 1989, Object Shift as an A-movement Rule, MIT Working Papers in Linguistics 11, 256-271. Zwart, J.-W., 1993, Dutch Syntax: A Minimalist Approach, PhD Dissertation, University Groningen, Groningen. Zwart, J.-W., 1994, Dutch is Head Initial, The Linguistic Review 11, 377-406. Zwart, J.-W., 1997, Morphosyntax of Verb Movement: A Minimalist Approach to The Syntax of Dutch, Kluwer, Dordrecht. 49 50