• No results found

Syntax overview - mainly (but not exclusively) generative

N/A
N/A
Protected

Academic year: 2021

Share "Syntax overview - mainly (but not exclusively) generative"

Copied!
66
0
0

Loading.... (view fulltext now)

Full text

(1)

Syntax overview -

mainly (but not exclusively) generative

Tuesday, December 15, 2020, 9:30-16:00

Basis 1 – Introduction to linguistic research

PhD programme in Language, Linguistics, Communication and Cognition, Aarhus University

Sten Vikner

Dept. of English, School of Communication & Culture, Aarhus University, DK-8000 Århus C, Denmark

Room 330, Building 1481, Tel. 8716 2639

sten.vikner@cc.au.dk ● www.hum.au.dk/engelsk/engsv/ ● http://au.dk/en/sten.vikner@cc

Wer fremde Sprachen nicht kennt, weiß nichts von seiner eigenen

If you do not know foreign languages, you do not know anything about your own.

(Goethe 1833, Maximen und Reflexionen)

Methodisk Forsøg til en Fuldstændig Dansk Syntax

Hvoraf kan tages Anledning, ey al-ene til en ræt og tydelig Construction i vort Sprog,

men og til des nøyere Indsigt i alle andre

Methodical Attempt at a Complete Danish Syntax,

on which can be based not just a correct and clear construction of our own language, but also a better insight into all others

(the full title of Høysgaard's 1752 Danish grammar)

This is the first of three hand-outs that we will (or might) discuss today:

Syntax overview

The others are:

Comparison with Diderichsen's (1946) analysis of Danish clause structure Formal and functional approaches to linguistics

(66 pp.)

(2)

Contents

1. Introduction to generative linguistics ... 3

1.1 Two crucial questions ... 3

1.2 Insufficiency of the input ... 4

1.3 Universal grammar ... 5

1.4 Comparative linguistics ... 6

1.5 Competence and Performance ... 6

2. Introduction to syntax ... 8 2.1 Constituents ... 8 2.2 Constituency tests ... 9 2.2.1 Movement ... 10 2.2.2 Substitution ... 12 2.2.3 Deletion ... 13 2.3 Subcategorisation ... 14

2.4 Arguments versus adjuncts ... 17

3. The phrase structure of VP ... 18

4. The phrase structure of NP ... 25

5. The phrase structure of DP ... 31

5.1 Introducing the DP-analysis ... 31

5.2 Differences between NP and DP ... 32

5.3 Parallels between pronouns and (other) determiners ... 33

5.4 Possessive constructions (DP-spec-analysis vs GP-analysis) ... 35

5.5 D° is often empty ... 38

6. The phrase structure of IP and V°-to-I° movement ... 40

6.1 VP and IP in Danish ... 40

6.2 Overview: V°-to-I° in the Germanic SVO-languages ... 42

6.3 V°-to-I° with modern English auxiliaries ... 44

6.4 V°-to-I° in German? I no longer think so ... 48

7. The phrase structure of CP and V2 ... 51

7.1 Embedded clauses ... 51

7.2 Verb second (V2) ... 52

7.3 Movement of the finite verb (conclusion to sect. 6 & 7) ... 57

8. Passivisation ... 58

8.1 Introduction ... 58

8.2 Case and θ-roles in passive clauses ... 59

8.3 Arguments/evidence for passivisation as movement ... 62

8.3.1 Argument 1: Subcategorisation ... 62

8.3.2 Argument 2: Idiomatic expressions ... 62

(3)

1.

Introduction to generative linguistics

1.1

Two crucial questions

(1) a. What does the linguistic knowledge of the first language(s) look like in the brain? b. How did it get there?

The answer to (1)b will set certain limits to the possible answers to (1)a:

(2) Innate Acquired

a. 100% – The entire grammar is innate, nothing is acquired b. – 100% Nothing is innate, everything is acquired

c. Part Part Some of the grammar is innate and some is acquired

If EVERYTHING was innate, i.e. (2)a, then we should all speak exactly the same language, just like we all have exactly one liver, two kidneys, two arms and ten fingers, not two livers, one kidney, ten arms and two fingers (or any other combination). In other words, any kind of variation between languages is a problem for such a hypothesis (which admittedly has never been seriously suggested): (3) English boy = German Junge = French garçon = Italian ragazzo

= Swedish pojke = Danish dreng = Norwegian gutt = Icelandic strákur

If NOTHING was innate, i.e. (2)b, then there ought to be a lot more variation between languages than there actually is. Although about 8000 natural languages are spoken in the world today, they are not all that different, cf. the existence of linguistic universals, which are properties common to all languages. One example of a universal is that a name cannot be coreferent with a preceding pronoun (cf. e.g. Haegeman and Guéron 1999, 374–76). The examples below are from Vikner (1999a): (4) English: a. He thinks [that John is rich] NO COREFERENCE (He ≠ John)

b. John thinks [that he is rich] POSSIBLE COREFERENCE

(5) Danish: a. Han tror [at Johan er rig] NO COREFERENCE

b. Johan tror [at han er rig] POSSIBLE COREFERENCE

(6) Icelandic: a. Hann heldur [að Jóhann sé ríkur] NO COREFERENCE

b. Jóhann heldur [að hann sé ríkur] POSSIBLE COREFERENCE

(7) German: a. Er glaubt [dass Johann reich ist] NO COREFERENCE

b. Johann glaubt [dass er reich ist] POSSIBLE COREFERENCE

(8) French: a. Il croit [que Jean est riche] NO COREFERENCE

b. Jean croit [qu’ il est riche] POSSIBLE COREFERENCE

(9) Welsh: a. Mae o ’n gwybod [ bod Sion yn gyfoethog] NO COREF.

b. Mae Sion yn gwybod [ei fod o ’n gyfoethog] POSSIBLE COREF.

(10) Hungarian: a. Azt gondolja [hogy János gazdag] NO COREFERENCE

(4)

(11) Basque: a. [ Jon aberatsa dela] uste du NO COREFERENCE

b. Jonek [ aberatsa dela] uste du POSSIBLE COREFERENCE

(12) Turkish: a. [ Hasanın zengin olduğunu] sanıyor NO COREFERENCE

b. Hasan [ zengin olduğunu] sanıyor POSSIBLE COREF.

(13) Swahili: a. Yeye anafikiri [John ni tajiri] NO COREFERENCE

b. John anafikiri [yeye ni tajiri] POSSIBLE COREFERENCE

If NOTHING was innate, i.e. (2)b, then it would also be difficult to account for why language acquisition can take place so quickly and so successfully. If nothing is innate, then the child has to acquire everything from scratch.

When you have to learn something from scratch, like math or chess or roller-skating or speaking a foreign language, then the result is not always very impressive. But people who speak English as their first language vary much much more with respect to how well they play chess or how well they speak French than with respect to how well they speak English.

1.2

Insufficiency of the input

If NOTHING was innate, i.e. (2)b, then it would be an open question how the child gets from 0% to 100%. The answer is, of course, that the child's knowledge must come from the input, i.e. from what the child hears. But what the child hears is not that much, and it is not always all that good either:

Not enough

The child cannot possibly have heard all the sentences (or all the sentence types) that it can produce or understand. The time it takes for a child to acquire its mother tongue is simply not long enough.

Not good enough

On top of this, the sentences that the child actually hears are not all well-formed.

Negative input plays no role at all

The insufficiency of the input is even more striking when it is taken into consideration that negative input (information that something is not possible) seems to play no role at all: (14) Child: Want other one spoon, Daddy.

Father: You mean you want "the other spoon". Child: Yes, I want other one spoon, please, Daddy. Father: Can you say "the other spoon"?

Child: Other - one - spoon Father: Say "other".

Child: Other Father: "Spoon". Child: Spoon.

Father: "Other – spoon".

Child: Other - spoon. Now give me other one spoon?

(Fromkin, Rodman, and Hyams 2003, 345) (15) Child: I taked a cookie.

Parent: I Oh, you mean you "took" a cookie.

(5)

(16) Child: Ved du hvor meget jeg drikkede? Parent: Nej, hvor meget drak du?

Child: Først drikkede jeg en hel kop te og så drikkede jeg et glas juice, og så .. Parent: Drak du så meget?

Child: Ja, så meget drikkede jeg.

(Occurred on 14.3.2000, at age 5. Child now says only drak, never drikkede.) How can these children (and all other children) then ever learn that something is not possible in their language? E.g. how could children find out that (17)a/(18)a are impossible with coreference although (17)b/(18)b are possible?

(17) English: a. He thinks [that John is rich] NO COREFERENCE (He ≠ John)

b. John thinks [that he is rich] POSSIBLE COREFERENCE

(18) Danish: a. Han tror [at Johan er rig] NO COREFERENCE

b. Johan tror [at han er rig] POSSIBLE COREFERENCE

To a generative linguist like me, the answer is that children do not need to learn this, in some sense they already know it (or rather, they "know" some of it). Therefore I am not the only linguist to be in favour of (2)c:

(19) Some of the linguistic knowledge is innate (the so-called Universal Grammar)

(this is the reason for universals and for the speed and success rate of language acquisition)

and some of it has to be acquired

(this is the reason for those differences that actually exist between the languages of the world)

1.3

Universal grammar

Universal grammar (UG) is the innate part of our linguistic knowledge, and many (though not all) of the properties that all languages share can be derived from it. UG is the reason that there are linguistic universals (in many or most cases, but not all), and UG is also the reason that language acquisition can be as quick and as successful as it is.

The language that is acquired is of course also determined by other things than UG: The input (the sentences that the child hears) also plays a decisive role. The input determines all those properties that vary between languages, i.e. the properties that are not determined by UG. If the input played no role at all, all humans would speak the same language, not just now but also 2000 years ago and 2000 years from now.

The interaction between UG and input may be compared to the one between hardware and software. The child does not learn its native language (or the grammar of its native language), the child

acquires it, or better: The grammar of the native language develops in the brain in interaction between

the organs of the brain and the input that these organs are exposed to (see e.g. Pinker 1994). In other words, the theory of UG is two things at once:

(20) a. UG is a theory of grammar across all natural languages, and

(6)

It is by trying to be both at the same time that the theory of UG is particularly interesting:

The amazing speed and ease with which children acquire their mother tongue is even more amazing when the immense number of different possible mother tongues is kept in mind.

1.4

Comparative linguistics

The idea is thus that comparative linguistics should strive to account for as many (surface) differences between different languages as possible by deriving them from as few underlying (theoretical)

differences as possible.

This is desirable not only for the inherent theoretical elegance in being able to unite separate surface phenomena under one generalisation, but also because it may provide the other half of the account of first language acquisition: The fewer (underlying) differences there are between languages, the less data the child will have to encounter in order to be able to choose between possible alternative

grammars, and the fewer data that have to be encountered to acquire any given language, the better is our account for the speed of first language acquisition.

This also makes it possible to determine which aspects of a given language, e.g. Danish, also exist in other languages and which aspects are specific to the individual language. Such a typological

perspective not only makes it possible to establish a number of typological correspondences and predictions (e.g. of the kind "only languages which have X will have Y too"), but also to explain and justify these theoretically.

By comparing both different languages (e.g. Danish vs. English) and different stages of the same language (e.g. Middle English vs. modern English), it becomes possible to decide what constitutes possible (and impossible) types of language variation.

1.5

Competence and Performance

(21) Competence ≈ Grammar

≈ Abstract ≈ Sentence ≈ Grammaticality Performance Use of the grammar

(Pragmatics) Concrete Utterance Acceptability

Performance covers all the non-linguistic factors, e.g. memory, concentration, tiredness, emotion, ...

Competence (grammar) is what tells a native speaker/hearer that he and John in (22) cannot refer to

the same person but that in (23), John and he could refer either to the same person or to two different persons (in other words, (23) has two readings, cf. section 6 above):

(22) En. He thinks [that John is rich]. (23) En. John thinks [that he is rich].

Pragmatics is what enables a hearer to find out exactly who he refers to in (22).

Consider a different example:

(24) En. Do you know what time it is?

Competence (grammar) is necessary to understand (24) as a yes/no-question, but on top of this, a

(7)

We (i.e. linguists, including students of linguistics) have to abstract away from performance factors when we want to say something about the linguistic knowledge that native speakers possess. Otherwise, we would also have to consider all sorts of irrelevant data, e.g. interrupted sentences, slips of the tongue, etc. The distinction between grammaticality and acceptability makes the following distinctions possible:

Whether an example is acceptable or not depends on what native speakers think about it, and this in turn depends on many things, only one of which is the grammar of English.

Whether a sentence is grammatical or not depends only on the grammar. If the sentence does not violate any rules of the grammar, it is grammatical.

Most examples are either both acceptable and grammatical or both unacceptable and ungrammatical, but we now also have room in the system for:

Acceptable but ungrammatical:

(25) a. It would mean so much to Paul and I if you'd let us through to see Santa Claus. b. Paul is the kind of guy who, when you get to know him, you really like him.

c. He fancies himself as an expert on a variety of subjects about which he knows nothing

about: linguistics, biology, comparative religion, politics...

(http://bigfootevidence.blogspot.dk/2012/06/robert-lindsay-and-bigfooters.html)

d. The key to the cabinets are missing.

Unacceptable but grammatical:

(26) a. This is a very, very, very, very, very, very, very, very, very, very, very, very, very, very, very, very, very, very, very, very, very, very, very, very, very, very, very, very, very, very, very, very, very, very, very, very, very, very, very, very, very, very, very good book. b. This is definitely not a book you must not fail not to read.

c. Ann thinks that Bill believes that Cathie has heard that Donald has said that Eric wishes that Frank liked that George told Helen that Ian likes Swiss chocolate.

d. The dog chased through the garden barked.

e. I have just bought an amazing, brilliant, colourful, definitive, excellent, fabulous, gigantic, helpful, impressive, knowledgeable, lavish, monumental, nifty, optimal, prestigious, quintessential, remarkable, stupendous, terrific, ultimative, valuable, and quite simply wonderful book on British birds.

f. If because when Mary came in John left Harry cried, I’d be surprised.

(8)

2.

Introduction to syntax

2.1

Constituents

This section presents some diagnostics for the identification of the basic units of clause structure, called constituents:

Constituent: One of the parts that form something (the chemical constituents of the liquid).

(Merriam-Webster’s Advanced Learner’s English Dictionary = Perrault 2008)

Constituency: Relation, especially in syntax, between a unit which is part of a larger unit and the whole of which it is part. E.g. the adjective phrase very friendly is a constituent of the noun phrase

very friendly people. (Concise Oxford Dictionary of Linguistics = Matthews 1997)

It is easy to find the smallest constituents (we already know the smallest units; they are the words), and it is easy to find the largest constituent (we already know it; it is typically the main clause). What is difficult in constituent analysis is to find the constituents in between, i.e. the constituents larger than the individual words and smaller than the main clause.

Take e.g. the expression French wine and cheese in (27): (27)

Is French wine a constituent in (27)a? In (27)b? Is wine and cheese a constituent in (27)a? In (27)b? As illustrated in (27), French and wine form a constituent, in (27)a, but not in (27)b where French and wine are parts of two separate constituents. Conversely, wine and and and cheese form a constituent, in (27)b, but not in (27)a, where wine is part of French wine while and cheese is not.

Might this be related to a difference in interpretation? In other words, if (28) was a sign at a conference or in a hotel, would it be clear which kind of cheese might be served?

(28) French wine and cheese will be served in the lounge.

As argued in Vikner & Vikner (2008), the ambiguity of a word like unlockable is linked to there being two structures, one for each of the two interpretations.

(29)

The same goes for ambiguous clauses: They may have more than one structure (e.g. S. Vikner 2019). There are two different possible syntactic constituent structures of French wine and cheese, and there could be up to five different constituent structures of the NP (the) Swiss chocolate toy factory:

a. b.

wine

French and cheese French wine and cheese

lock

un- -able un- lock -able

[un- [lock-able]] "cannot be locked"

[[un-lock] -able] "can be unlocked"

(9)

(30)

Are there really five different interpretations? What are they?

Exercise

Which of the structures in (30) is the most plausible constituent structure for the following? (31) En. a. European parliament election

b. American football player

c. local election campaign organiser d. old English syntax teacher

Interpretations corresponding to the structures in (30)a-e:

(32) En. a. A Swiss factory that is made of chocolate and that makes toys

b. A Swiss factory that makes chocolate toys

c. A factory that makes chocolate toys which are Swiss d. A factory that makes toys of Swiss chocolate

e. A factory that is made of Swiss chocolate and that makes toys

2.2

Constituency tests

One way to determine whether or not two or more words constitute a unit is to apply constituency

tests. If a string of words passes a constituency test, it may be assumed to be a constituent, whereas if

a string of words does not pass a constituency test, no such conclusion may be drawn: (33)

a.

Swiss chocolate toy factory Swiss ew

chocolate toy factory b.

c. d.

Swiss w

chocolate toy factory Swiss

w

chocolate toy factory

e.

Swiss chocolate toy factory

Apply constituency

test

(Pass)

(Fail)

The string is a constituent

The string is not a constituent, OR

(10)

In other words, for a string of words to pass a constituency test, two requirements have to be met: (34) To pass a constituency test,

a. the string of words must be a constituent (form a unit), AND

b. the test must be the right kind of test

There are three main types of tests: movement, substitution, and deletion.

We will use the following sentence as a base example for the various constituency tests:

2.2.1

Movement

Topicalisation

(Haegeman and Guéron 1999, 225; also called fronting or preposing, Aarts 2011, 317) (37) [xxx] … … … _____ … …

(38) En. a. [In Cork], the chef from LA bought a used car . b. [A used car], the chef from LA bought in Cork. c. * [Car in], the chef from LA bought a used Cork. d. * [In] the chef from LA bought a used car Cork. (39) Da. a. [I Køge] købte kokken fra Ærø en brugt bil

b. [En brugt bil] købte kokken fra Ærø i Køge c. * [Bil i] købte kokken fra Ærø en brugt Køge d. * [I] købte kokken fra Ærø en brugt bil Køge

What is wrong with (38)c is thus that car and in either do not form a constituent or that they form a constituent of the wrong kind. Given that car forms a constituent with a and used, as shown in (38)b, car cannot at the same time form a constituent with in.

What is wrong with (38)d is also that in either is not a constituent or that it is a constituent of the wrong kind. Given that in has to be a constituent on its own (as we have to assume that all individual words are constituents), in must be a constituent of the wrong kind for this test.

Clefting (Haegeman and Guéron 1999, 49; also called it-cleft, Aarts 2011, 331)

(40) It is/was [xxx] (who / that) … … … _____ … …

(41) En. a. It was [in Cork] that the chef from LA bought a used car . b. It was [a used car] that the chef from LA bought in Cork. c. It was [the chef from LA] who bought a used car in Cork. d. * It was [the chef from] who LA bought a used car in Cork. e. * It was [car in] that the chef from LA bought a used Cork. f. * It was [bought] that the chef from LA a used car in Cork. (35) En. The chef from LA bought a used car in Cork.

(11)

(42) Da. a. Det var [i Køge] kokken fra Ærø købte en brugt bil b. Det var [en brugt bil] kokken fra Ærø købte i Køge c. Det var [kokken fra Ærø] som købte en brugt i Køge d. * Det var [kokken fra] som Ærø købte en brugt bil i Køge e. * Det var [bil i] kokken fra Ærø købte en brugt Køge f. * Det var [købte] kokken fra Ærø en brugt bil i Køge

Pseudo-clefting (Haegeman and Guéron 1999, 50; also called wh-cleft, Aarts 2011, 333)

(43) What … _____ … … … is/was [xxx]

(44) En. a. Where the chef from LA bought a used car was [in Cork]. b. What the chef from LA bought in Cork was [a used car]. c. * What the chef from LA bought a used Cork was [car in]. d. * Where the chef from LA bought a used car Cork was [in]. (45) En. a. The chef from LA did buy a used car in Cork.

b. What the chef from LA did in Cork was [buy a used car].

c. What the chef from LA did was [buy a used car in Cork]. d. * What the chef from LA did a used car was [buy] [in Cork].

e. * What the chef from LA did a used car in Cork was [buy]. (46) Da. a. Der hvor kokken fra Ærø købte en brugt bil var [i Køge]

b. Det kokken fra Ærø købte i Køge var [en brugt bil] c. * Det kokken fra Ærø købte en brugt Køge var [bil i]

d. * Der hvor kokken fra Ærø købte en brugt bil Køge var [i] (47) Da. a. Kokken fra Ærø købte en brugt bil i Køge

b. Det kokken fra Ærø gjorde i Køge var [at købe en brugt bil]

c. Det kokken fra Ærø gjorde var [at købe en brugt bil i Køge] d. * Det kokken fra Ærø gjorde en brugt bil var [at købe] [i Køge]

e. * Det kokken fra Ærø gjorde en brugt bil i Køge var [at købe]

Note that in all of (45), the finite main verb bought (inflected for past tense) is 'split in two', i.e. substituted by a finite version of the auxiliary verb do (inflected for past tense) AND the infinitive of the main verb, buy. This does not only happen in pseudo-clefts, but also e.g. to emphasise the truth of the sentence, as in (45)a, and in questions and with negation.

Passivisation (Haegeman and Guéron 1999, 199; Aarts 2011, 93)

(48) Someone V-ed something Something was V-ed _____ (by someone)

(12)

(50) Da. a. ... fordi [en brugt bil] blev købt i Køge (af kokken fra Ærø) b. * ... fordi [bil i] blev købt en brugt Køge (af kokken fra Ærø) c. * ... fordi [bil] blev købt en brugt i Køge (af kokken fra Ærø) d. * ... fordi [i Køge] blev købt en brugt bil (af kokken fra Ærø)

This test only works with a constituent which is assigned accusative case by the verb (e.g. the object). This is why in Cork fails the test in (49)d, even though we know from (38)a that it is a constituent.

Heavy NP shift

(Haegeman and Guéron 1999, 52, 221; also called extraposition or postposing, Aarts 2011, 320) (51) … … _____ … … … [xxx]

This test applies only to 'heavy' ('long') NP constituents. Applying heavy NP shift to 'lighter' ('shorter') constituents is either not very good, (53), or completely ungrammatical, (54).

(52) En. a. She bought [an old, rusty, run-down Morris] in Cork.

b. She bought in Cork [an old, rusty, and completely

run-down Morris Mini].

(53) a. She bought [a used car] in Cork.

b. ?? She bought _________ in Cork [a used car].

(54) a. She bought [it] in Cork.

b. * She bought __ in Cork [it].

(55) Da. a. Han købte [en gammel, rusten og slidt bil] i Køge

b. ? Han købte i Køge [en gammel, rusten og slidt bil]

(56) Da. a. Han købte [en brugt bil] i Køge

b. ?? Han købte _________ i Køge [en brugt bil]

(57) Da. a. Han købte [den] i Køge

b. * Han købte __ i Køge [den]

2.2.2

Substitution

Substitution by a pronoun (e.g. he/him, she/her, they/them, it) or another proform (e.g. an adverb, here/there, or a verb phrase, such as did so):

(58) En. a. [ The chef from LA ] [ bought [ a used car ]] [ in Cork ]. b. [ She ] [ bought [ a used car ]] [ in Cork ]. c. [ The chef from LA ] [ bought [ it ]] [ in Cork ]. d. [ The chef from LA ] [ bought [ a used car ]] [ there ]. e. [ The chef from LA ] [ did so ] [ in Cork ]. f. * [ She LA ] [ bought [ a used car ]] [ in Cork ]. g. * [ The chef from LA ] [ bought [ a used car ]] [ there Cork ]. (59) Da. a. [ Kokken fra Ærø ] [ købte [ en brugt bil ] ] [ i Køge ]

(13)

Substitution by a wh-element with subsequent movement to the beginning of the clause: (60) En. a. … that [ the chef from LA ] [ bought [ a used car ]] [ in Cork ].

b. Who [ ] [ bought [ a used car ]] [ in Cork ]? c. What did [ the chef from LA ] [ buy [ ]] [ in Cork ]? d. Where did [ the chef from LA ] [ buy [ a used car ]] [ ]? e. * Who [ LA ] [ bought [ a used car ]] [ in Cork ]? f. * What did [ the chef from LA ] [ buy [ car ]] [ in Cork ]? g. * Where did [ the chef from LA ] [ buy [ a used car ]] [ Cork ]? (61) Da. a. … at [ kokken fra Ærø ] [ købte [ en brugt bil ] ] [ i Køge ]

b. Hvem købte [ ] [ [ en brugt bil ] ] [ i Køge ] ? c. Hvad købte [ kokken fra Ærø ] [ [ ] ] [ i Køge ] ? d. Hvor købte [ kokken fra Ærø ] [ [ en brugt bil ] ] [ ] ? e. * Hvem købte [ Ærø ] [ [ en brugt bil ] ] [ i Køge ] ? f. * Hvad købte [ kokken fra Ærø ] [ [ bil ] ] [ i Køge ] ? g. * Hvor købte [ kokken fra Ærø ] [ [ en brugt bil ] ] [ Køge ] ?

After something has been substituted by a wh-element, this wh-element has to move to the left edge of the clause, as in (60)b-d. Without this movement, we would have a so-called echo-question:

2.2.3

Deletion

Only constituents (and only non-arguments) may be deleted.

(64) En. a. [The chef [ from LA ]] [bought [a used car]] [ in Cork ] [ last week ]. b. [The chef [ ]] [bought [a used car]] [ in Cork ] [ last week ]. c. [The chef [ from LA ]] [bought [a used car]] [ in Cork ] [ ]. d. [The chef [ from LA ]] [bought [a used car]] [ ] [ last week ].

e. [The chef [ ]] [bought [a used car]] [ ] [ last week ]. (2 deletions) f. [The chef [ from LA ]] [bought [a used car]] [ ] [ ]. (2 deletions) g. * [The [ LA ]] [bought [a used car]] [ in Cork ] [ last week ].

(65) Da. a. [Den gamle kok [ fra Ærø ]] [købte [en brugt bil]] [ i Køge ] [ i går ] b. [Den gamle kok [ ]] [købte [en brugt bil]] [ i Køge ] [ i går ] c. [Den gamle kok [ fra Ærø ]] [købte [en brugt bil]] [ i Køge ] [ ] d. [Den gamle kok [ fra Ærø ]] [købte [en brugt bil]] [ ] [ i går ]

e. [Den gamle kok [ fra Ærø ]] [købte [en brugt bil]] [ ] [ ] (2 deletions) f. [Den gamle kok [ ]] [købte [en brugt bil]] [ ] [ i går ] (2 deletions) g. * [Den gamle [ Ærø ]] [købte [en brugt bil]] [ i Køge ] [ i går ]

As illustrated in (64)e-f, deletion is not the most reliable test because we cannot tell afterwards whether it has been applied once or more than once.

(14)

2.3

Subcategorisation

Some but not all elements are obligatory:

(66) a. The painter sold two pictures last week in London at a very low price. b. The painter sold two pictures last week in London.

c. The painter sold two pictures last week at a very low price. d. The painter sold two pictures in London at a very low price. e. * The painter sold last week in London at a very low price. Some but not all preposition phrases (PPs) may be left out:

NP = noun phrase, e.g. two pictures or the evil scientist, PP = preposition phrase, e.g. in the lab or to the museum (67) a. The evil scientist worked in the lab all night.

b. The evil scientist worked all night. (68) a. The painter gave two pictures to the museum.

b. * The painter gave two pictures.

Whether or not an element is obligatory (= must be present) or optional (= may be left out) depends on the verb:

(69) sell: requires an NP

work: does not require anything

give: requires both an NP and a PP (or alternatively two NPs)

The verb sell is a (mono-)transitive verb, work is intransitive, and give is ditransitive:

(70) SUBCATEGORISATION TRANSITIVITY

sell V, [ __ NP] (MONO-)TRANSITIVE

work V, [ __ ] INTRANSITIVE

give V, [ __ NP NP], V, [ __ NP PP] DITRANSITIVE

Because this is a way of defining subcategories of the category verb (V), these "requirements" of the verb are called the subcategorisation of the verb (or the subcategorisation frame of the verb). The subject is not relevant to the subcategorisation frame, because all (finite) verbs (in English)

require a subject (as in fact it says in the subject requirement in Haegeman and Guéron 1999, 126)

Whether there is a subject or not therefore cannot be used to differentiate between verb types. The elements relevant for subcategorisation all occur after the verb, inside the verb phrase (VP).

(15)

There are also subcategorisation types other than the ones above. Some verbs require a PP

complement rather than an NP complement, and are called "prepositional (intransitive)" verbs (In Aarts 2011, 139, 144–46, a verb must have an object (i.e. an NP complement) to count as transitive.): (71) listen: V, [ __ PP] (PREPOSITIONAL VERB (INTRANSITIVE))

(Aarts 2011, 101) (72) a. * Henry listens.

b. * Henry listens rock music. c. Henry listens to rock music.

Some verbs require an object that is either a noun phrase (an NP) or a sentence (S):

(73) know: V, [ __ NP], [ __ S] (TRANSITIVE VERB, but with a clausal object)

(74) a. * Henry knows.

b. Henry knows [NP French]. c. * Henry knows [PP in Paris].

d. Henry knows [S that Mary likes him].

In (74)d, the object of know is not an NP but an embedded clause (which is the same as a

subordinate clause, Aarts 2011, 8, 12).

The following illustrates the difference between main and embedded clauses: (75)

In (75), (a) is a main clause (it contains the entire string of words), and (b) and (c) are both embedded clauses;

(a) and (b) are both matrix clauses, because both contain at least one embedded clause (the word matrix comes from Latin and originally meant 'mother').

Embedded clauses have a function ("play a particular role") in their matrix clause. In (75), for

example, (b) is the object of (subcategorised for by) the verb thinks and (c) is the object of believes. Sometimes (but not always) embedded clauses are introduced by a complementisers, such as that or if or because (complementisers are also called subordinating conjunction).

In other words, if you have a main clause and an embedded clause, the relevant question to ask is which of them is part of the other one. Do NOT ask where one stops and the other begins, because the embedded clauses is part of the main clause. (The latter would be like asking where California stops and the United States of America begin, which is absurd because California is part of the USA.)

Henry thinks (that) Joe actually believes

(c) EMBEDDED (NON-MATRIX)

(that) Mary was abducted by aliens

(a) MAIN CLAUSE, MATRIX CLAUSE

(16)

As illustrated in (76) below, a complex sentence (i.e. a sentence with one or more embedded clauses) does not have to have the embedded clause(s) at the end (as is the case in (75) above).

In (76), (a) is the main clause and a matrix clause, because it contains (b), and (b) is an embedded clause. (76) shows that an embedded clause can also be the first part of its matrix clause.

(76)

Rather than using boxes as in (75) and (76), sentence structure is most often illustrated by means of square brackets, as in the corresponding (77) and (78):

(77) [a Henry thinks [b that Joe actually believes [c that Mary was abducted by aliens]]] = (75) (78) [a [b Since Mary was abducted by aliens] Joe has been afraid of spaceships] = (76)

Exercise

Identify and classify (as embedded, matrix, and/or main clauses) all the clauses in the following: (79) En. Because Jim thought that Alice had let him down, he would not talk to her.

(80) En. Peter announced that he had an accident in the car.

(81) En. If you leave now, you might get home before the rain starts.

Joe has been afraid of spaceships

(b) EMBEDDED (NON-MATRIX)

Since Mary was abducted by aliens

(17)

2.4

Arguments versus adjuncts

(82) a. The evil scientist gave the book to the museum. b. The evil scientist read the book in her office.

In (82)a to the museum is an argument, whereas in (82)b in her office in is an adjunct. The difference is in the strength of the thematic link to the main verb:

Arguments are very closely linked to the verb. This is also why they may be obligatory. Arguments get their meaning from the main verb, cf. that whether a particular argument like the student is an AGENT, a PATIENT/THEME, an EXPERIENCER, or a BENEFICIARY depends on the main

verb and the position of the argument in relation to this verb.

Adjuncts are less closely linked to the verb. This is also why they are both always optional and

always possible. Adjuncts do not get their meaning directly from the main verb, cf. that whether a

particular adjunct like in the office is an adjunct of TIME, of MANNER, of REASON, of PLACE, or of MODALITY depends much less on the verb than on the content of the adjunct itself.

Whether an argument like her son is impossible, optional, or obligatory thus depends on the verb (mention is monotransitive, give is ditransitive, and buy can be either mono- or ditransitive): (83) a. Ann mentioned a book. d. * Ann mentioned her son a book.

b. Ann bought a book. e. Ann bought her son a book. c. * Ann gave a book. f. Ann gave her son a book.

Adjuncts, on the other hand, are in principle always (syntactically) possible, cf. that all the examples

which are possible without the adjunct last week in (83) are also possible with this adjunct in (84): (84) a. Ann mentioned a book last week. d. * Ann mentioned her son a book last week.

b. Ann bought a book last week. e. Ann bought her son a book last week. c. * Ann gave a book last week. f. Ann gave her son a book last week. In other words, arguments depend on the verb for their meaning, whereas adjuncts do not. (Aarts 2011, 85, 90, 107 discusses subjects and complements, both of which are arguments.) (Aarts 2011, 102 discusses adjuncts; but e.g. Quirk et al. 1985, 475 uses the term “adverbial”.) In other words, optionality in itself is not a reliable indication that we are dealing with an adjunct. A comparison between (83)e and (84)e shows that last week is optional, and last week is an adjunct, whereas a comparison between (83)b and (83)e shows that her son is optional, and nevertheless we claimed that her son is an argument.

The crucial difference shown in (83) and (84) is that the argument her son is only optional with the verb buy, whereas it is obligatory with the verb give (and impossible with the verb mention). The adjunct last week, on the other hand, is optional with all three verbs.

(18)

3.

The phrase structure of VP

We will be working towards a general phrase structure, a structural pattern common to all phrases and clauses. This structural schema is the so-called X-bar structure in (85). XP is shorthand for a phrase of the category X, where X can be any of the syntactic categories, for example, V, N, P, Adj, Adv. (85)

(About the ordering: A modifier can be either the left or the right sister of XP. A specifier is always the left sister of X'. In Danish and English, the complement is always the right sister of X°. In e.g. Korean, the complement is always the left sister of X°. In e.g. German, it depends on the X°: Verbs follow their complements, nouns and prepositions precede their complements.) The X-bar structure obeys two general principles of constituent structure:

(86) a. Binary branching: A node has at most two daughters.

b. Headedness: Every phrase is a projection of a head.

The structural difference between complements and modifiers in (85):

(87) a. A phrase (XP) may be the sister of a X° (a head), and then it is a complement of this head, i.e. it is an argument of the head (e.g. an object), and not an adjunct.

b. A phrase (XP) may be the sister of an XP (a phrase), and then it is a modifier of this phrase, i.e. it is an adjunct of the phrase (e.g. an adverbial), and not an argument.

(88) HEAD (V°/N°) COMPLEMENT OF V°/N° MODIFIER OF VP/NP

a. Teaching English linguistics at Aarhus University (VP/CLAUSE)

b. Teachers of English linguistics from Aarhus University (NP/DP)

The first question regarding the VP (verb phrase) is exactly which words or constituents are part of it. In order to find this out, we can test by means of do so. It is possible to refer to a preceding VP without repeating it word for word; instead, do so can substitute for it:

(89) En. a. Henry will buy presents in Paris tomorrow and

Joe will do so , too.

b. Henry will buy presents in Paris tomorrow and

Joe will do so next week.

c. Henry will buy presents in Paris tomorrow and Joe will do so in London next week.

d. * Henry will buy presents in Paris tomorrow and Joe will do so books in London next week.

XP

specifier X'

XP modifier

complement

MINIMAL PROJECTION (HEAD)

MAXIMAL PROJECTIONS

(19)

(90) Da. a. Henrik vil købe julegaver i Paris i morgen og

det vil Joachim også.

b. Henrik vil købe julegaver i Paris i morgen og

det vil Joachim i næste uge.

c. Henrik vil købe julegaver i Paris i morgen og det vil Joachim i London i næste uge.

d. * Henrik vil købe julegaver i Paris i morgen og det vil Joachim bøger i London i næste uge. In the above, the words do so can replace:

1. V°-NP-PP-AdvP: buy presents in Paris tomorrow = (89)a/(90)a 2. V°-NP-PP: buy presents in Paris = (89)b/(90)b

3. V°-NP: buy presents = (89)c/(90)c

but not just: 4. V°: *buy = (89)d/(90)d

How can we capture the fact that three different constituents may be replaced by the same element in (89)a-c/(90)a a-c)? By assuming that the three constituents are of the same category, namely, VP: (91)

Henry will

What excludes substitution in (89)d/(90)d is that do so/det must substitute a VP, not just a verb: (92)

Sue might

(93)

Sue might

In (92), do so cannot substitute for call, but only for call her mother. This shows that call her mother is a constituent (a VP), and that if there is an object (her mother), it must be part of every VP-node. This again means that her mother = the sister of V° = a complement of V°, i.e. an argument. In (93), do so can substitute both for call and for call next week. This shows that call is a constituent (not really a surprise, as individual words are always constituents), that call next week is a constituent

VP VP VPbuy NP presents PP in Paris AdvP tomorrow VPcall NP

her mother (a. ... and Joe might do so too) (b. * ... and Joe might do so his father

VPcall NP next week VP

(a. ... and Joe might do so too)

(20)

(a VP), and that call and call next week are constituents of the same type. In (93), call is therefore not just a verb, but also a VP. This again means that next week = sister of VP = a modifier, i.e. an

adjunct.

This further captures the fact that when both are present, her mother must precede next week (and do so may then replace either call her mother or call her mother next week, but not just call):

(94)

Sue might

The same argument can be made based on pseudo-clefting: (95) a. Henry will buy presents in Paris tomorrow.

b. What Henry will do is buy presents in Paris tomorrow. c. What Henry will do tomorrow is buy presents in Paris.

d. What Henry will do in Paris tomorrow is buy presents. e. * What Henry will do presents in Paris tomorrow is buy.

(96) a. Henrik vil købe gaver i Paris i morgen.

b. Det Henrik vil gøre er at købe gaver i Paris i morgen. c. Det Henrik vil gøre i morgen er at købe gaver i Paris.

d. Det Henrik vil gøre i Paris i morgen er at købe gaver. e. * Det Henrik vil gøre gaver i Paris i morgen er at købe.

Notice that the impossibility of (95)e/(96)e (and of (89)d/(90)d) does not show that buy is not a constituent at all, it only shows that buy is not a constituent of the same kind as buy presents, buy presents in Paris, and buy presents in Paris tomorrow.

What excludes pseudo-clefting in (95)e/(96)e is that what must correspond to a phrase, not just a head: (97) a. Sue might call her mother.

b. What Sue might do is call her mother. c. * What Sue might do her mother is call.

(98) a. Sue might call next week.

b. What Sue might do is call next week. c. What Sue might do next week is call.

In (97), we cannot pseudo-cleft call, but only call her mother. This is because call her mother is a phrase (a VP), and if there is an object (her mother), it must be part of every VP-node. This again means that her mother = the sister of V° = a complement of V°, i.e. an argument.

VPcall NP her mother VP NP next week COMPLEMENT (SISTER OF V°)

(a. ... and Joe might do so too)

(b. ... and Joe might do so a week later) (c. * ... and Joe might do so his father a week later)

MODIFIER

(21)

In (98), we can pseudo-cleft both call and call next week. This is because call and call next week are constituents of the same type. In (98), call is therefore not just a verb, but also a VP. This again means that next week = sister of VP = a modifier, i.e. an adjunct.

We have thus seen that the main verb forms a constituent together with the object. This constituent does not include the subject, and it does not include any auxiliary verbs. We can therefore rule out both (99)a and (99)b as possible structures of a clause:

(99)

In other words, even though in e.g. (89)a/(95)a, Henry will buy presents in Paris tomorrow, it is the auxiliary verb form will and the main verb form buy and no other parts of (89)a/(95)a) that determine that the tense of (89)a/(95)a is future, this does not mean that will and buy form a constituent. As was shown in (89)a-d/(95)b-e, there are several VPs in Henry will buy presents in Paris tomorrow, but will buy is not one of them. In fact will and buy cannot possibly form a constituent in Henry will buy presents in Paris tomorrow, given that buy forms a constituent with presents.

I say this because there are two different views of VPs in the literature (see also Vikner 2016): The one I prefer: a VP consists of one verb and its complement and modifier(s), The one I don't like: a VP consists of all the verbs in a clause and nothing else:

(100) a.

They [

VP must

have

read] it.

H&S = Hjulmand and Schwarz (2017), ...

b.

They

must [

VP

have [

VP

read it] ]

.

here = generative syntax (and others)

(101)

VP

clause subject

aux main verb object verb clause object b. a. subject IP I' DP I° must VP V° have VP DP V° read b. here a. H&S St H pers pron S NP H pers pron DO NP V VP PreM mod →inf H v PreM perf →ptp it

They must have read

D' D°

They

(22)

(102) a. Analyses of English that have a VP for each verb1 (VP = verb + complement/ modifiers) include Aarts and Haegeman (2006, 126–33), Aarts (2001, 43, 104–11, 196–201; 2011, 66), Altenberg and Vago (2010, 126), Carnie (2013, 80). Fromkin et al. (2013, 85, 92),

Huddleston and Pullum (2005, 13), Hurford (1994, 93, 186), Johansson and Manninen (2012, 67, 85–88), Payne (2011, 200–203), and Radford (2009, 40)

b. Analyses of English that have only one VP per clause (VP = all and only verbs) include Andersen (2006, 60), Ballard (2013, 101–3), Biber et al. (1999, 99–100), Collins and Hollo (2010, 78), Crystal (2004, 114), Greenbaum (1996, 59), Hasselgård et al. (2012, 21, 164), McGregor (1997, 123; 2015, 113–14), Preisler (1997, 27, 76–79), Quirk et al. (1985, 96–97) and Thomas (1993). Also Bache (1996; 2014, 65) and Bache, Davenport, Dienhart and Larsen (1993, 74–75) only have one VP per clause, but they call it "verb group".

One point in favour of one VP per verb is that such VPs may contain more than just verbs. Of the various phrases in Hjulmand and Schwarz (2017), Greenbaum (1996), etc., the VP is special by not being able to contain any modifiers or complements and by consisting of words from one word class only: The two opposing views agree e.g. that NPs contain more than just nouns, AdjPs contain more than just adjectives, and PPs contain more than prepositions.

Another point in favour of one VP per verb is the one already discussed above, namely that only such VPs form constituents in the sense of the constituent tests discussed earlier: VPs may be fronted (this is often called "VP-preposing"), and such examples clearly show that VPs consist not only of verbs, but of verbs, complements, and modifiers. (103)a is from Greenbaum & Quirk (1990, 409), (103)b from Quirk, Greenbaum, Leech & Svartvik (1985, 125), and (103)c from the British National Corpus: (103) a. They have promised to [finish the work], and [finish it] they will ___.

b. Bill said that he would [win the match], and [win the match] he did ___.

c. It was, it was indeed, a virus so small that the eye of man had never [seen it before]. But [see it now] we did ___.

Notice how it is absolutely impossible for will finish to be a constituent in (103)a, as here [finish it] is shown to be a constituent, and finish could not possibly form one constituent with it, and at the same time form another constituent with will. This is parallel to the impossibility of one and the same Thursday being part of the month of October as well being part as of the month of November.

1 Although Chomsky (1957, 111) had a heterogeneous VP (VP → Verb + NP), it also had the homogeneous VP as a

constituent (Verb → Aux + V), and it was therefore not at all compatible with (100)b/(101)b, and neither was Chomsky (1965, 43). One of the first generative analyses to be compatible with (100)b/(101)b was Ross (1969).

St sentence

S subject

V verb

DO direct object NP noun phrase

H head (of a phrase)

PreM premodifier

v verb

IP inflection phrase

VP verb phrase

DP determiner phrase

(23)

In English and Danish, the order verb-complement is the only possible order: (104) En. a. Jack thinks that he has seen a unicorn

b. * Jack thinks that he has a unicorn seen

(105) Da. a. Jack tror at han har set en enhjørning b. * Jack tror at han har en enhjørning set The order between modifiers and the VP, on the other hand, may vary: (106) En. a. * Jack will eat another cheeseburger definitely

b. Jack will definitely eat another cheeseburger

(107) Da. a. * Jack vil spise endnu en cheeseburger bestemt b. Jack vil bestemt spise endnu en cheeseburger (108) En. a. Sarah will buy a car there

b. * Sarah will there buy a car

(109) Da. a. Sarah vil købe en bil der b. * Sarah vil der købe en bil (110) En. a. Sue will send an e-mail soon

b. Sue will soon send an e-mail

(111) Da. a. Sue vil skrive en e-mail snart b. Sue vil snart skrive en e-mail

Whether a modifier may precede or follow the VP seems to depend on which semantic class the modifier belongs to (e.g. locations seem to have to follow the VP). English VPs thus allow both (112)a and (112)b:

(112)

Although English, like Danish, does not allow the order complement-verb, as in (112)c-d, this does not mean that this order is universally excluded. It is perfectly possible, for example, in Old English, German, Latin, Japanese, Korean or Turkish.

(24)

(113) Ge. a. * Jack glaubt dass er gesehen ein Einhorn hat = (104) b. Jack glaubt dass er ein Einhorn gesehen hat

Exercise

Are the following languages V-O or O-V: Japanese, Afrikaans, Icelandic, and Hindi?

(115) Ja. a. Junko-ga hon-o yonda

Junko.NOM book.ACC read ("Junko read the book")

b. * Junko-ga yonda hon-o c. * Yonda Junko-ga hon-o

(116) Af. a. ... dat Jan 'n appel geëet het

... that Jan an apple eaten has (".. that Jan has eaten an apple") b. * ... dat Jan het geëet 'n appel

c. * ... dat Jan geëet 'n appel het

(117) Ic. a. ... að strákurinn hefur lesið bókina

... that boy-the.NOM has read book-the.ACC (".. that the has read the book")

b. * ... að strákurinn hefur bókina lesið c. * ... að strákurinn bókina lesið hefur

(118) Hi. a. kitaab paRhnaa anu-ko pasand hai

Book read Anu.DAT pleasant is ("Reading books pleases Anu")

b. * paRhnaa kitaab anu-ko pasand hai

(119)

(114)Ko. a. * Insu-nŭn sa-ass-da sae sigye hana-lŭl Insu.TOP buy.PRET.DECL new watch a.ACC

b. Insu-nŭn sae sigye hana-lŭl sa-ass-da

Insu.TOP new watch a.ACC buy.PRET.DECL

'Insu bought a new watch'

N° DP D° NP V° I° VP CP C° IP

(25)

4.

The phrase structure of NP

First, an important revision: Articles (etc.) are not seen as part of the NP (the noun phrase). Instead, what is often called an NP is now seen as a DP (a determiner phrase) (cf. also section 5 below). (120) BEFORE: FROM NOW ON:

a. [NP a blue chair] b. [DP a [NP blue chair]]

c. [NP a chair from Germany] d. [DP a [NP chair from Germany]]

The first question regarding NP is exactly which words or constituents are part of it. In order to find this out, we can test by means of one. It is possible to refer to a preceding NP without repeating it word for word, by substituting it with one, much like the do so test in the VP above:

(121) a. These blue chairs are cheaper than

the green ones over there. onesSUBSTITUTES [chairs]

b. These blue chairs are cheaper than

the ones over there. onesSUBSTITUTES [blue chairs]

(122) a. This chair from Germany is cheaper than

the one from France over there. oneSUBSTITUTES [chair]

b. This chair from Germany is cheaper than

the one over there. oneSUBSTITUTES [chair from Germany]

How can we capture the fact that two different constituents may be replaced by one and the same element? By assuming that the two constituents are of the same category, namely NP:

(123)

If one substitutes NPs (just like do so substitutes VPs, cf. (91)-(93) above), then this can be used as a test for complements of N° (which have to be part of NP) vs. modifiers of NP (which may be included in one instance of NP but excluded by another instance of NP):

(124) a. This teacher of linguistics with an American accent is younger than

that one.

The teacher of linguistics with an American accent is younger than

the one with a Danish accent.

b. * The teacher of linguistics with an American accent is younger than

the one of physics with a Danish accent.

(124) shows that of linguistics is a complement (it has to be part of what is substituted by one).

(124) also shows that with an American accent is a modifier (it may be but does not have to be part of substituted by one). (Remember that a complement is the sister of X°, whereas a modifier is the sister of XP.)

(26)

(125)

The

This also accounts for the following contrast:

(126) a. A teacher of linguistics with an American accent b. * A teacher with an American accent of linguistics

As a complement, of linguistics insists on being the sister of the N° teacher.

Because a complement is the sister of X°, whereas a modifier is the sister of XP, complements have to be closer to the X° than modifiers do (hence the ungrammaticality of (126)b), and therefore all XPs include the complement (hence the ungrammaticality of (124)b).

One of the analyses that have only an NP and no DP is Hjulmand and Schwarz (2017), which was also discussed in the previous section. Hjulmand and Schwarz (2017) also differ from the analysis

advocated here in a different way, namely in that they have no difference at all between complements and modifiers in the NP2, see (127)a and (128)a. Both complements and modifiers are here called "postmodifiers". In a generative analysis, on the other hand, e.g. of linguistics in a teacher of

lingustics will be seen as a complement of a noun, (127)b, whereas e.g. with an American accent in a teacher with an American accent will be seen as a modifier of an NP, (128)b.

(127)

NP noun phrase

D determiner DP determiner phrase

ia indefinite article NP noun phrase N "the nominal part" (H&S 2017:64) PP preposition phrase H head (of a phrase)

n noun PoM postmodifier

2 Also in e.g. Bache & Davidsen-Nielsen (1997, 343) and Bache (2014, 150, 154), there are no differences made between

complements and modifiers at the NP/DP-level

NPteacher PP of linguistics NP PP

with an American accent

(27)

(128)

In the generative analysis, the system is thus basically the same at the VP-level and at the DP-level: A

complement like of linguistics is the sister of an X°, (127)b, whereas a modifier like with a blue shirt

is the sister of an XP, (128)b.

As shown above, the differences w.r.t one-substitution, (124), can be argued to follow from the structural differences between (127)b and (128)b, but they cannot be argued to follow from the structural differences between (127)a and (128)a, as there aren't any.

Similarly, the ordering restrictions illustrated in (126)a and (126)b can be argued to follow from only one of the two PPs being the sister of teacher in (129)b. This is not possible in (129)a where both PPs are sisters of teacher:

(129)

In other words, by not having any difference between complements and modifiers at the NP/DP-level, approaches like Hjulmand and Schwarz (2017) fail to capture this systematic set of differences.

(28)

In English, Danish and German NPs, the complement follows the head (i.e. N°): (130) En. a. The brutal destruction of the city.

b. * The brutal of the city destruction.

(131) Da. a. Den brutale ødelæggelse af byen. b. * Den brutale af byen ødelæggelse.

(132) Ge. a. Die brutale Zerstörung der Stadt. b. * Die brutale der Stadt Zerstörung.

The distribution of modifiers shows a little more variation: PPs, (133), and relative clauses, (139), follow the NP they modify:

(133) En. a. That man with the long hair is my cousin.

b. * That with the long hair man is my cousin.

(134) Da. a. Den der mand med det lange hår er min fætter.

b. * Den der med det lange hår mand er min fætter.

(135) Ge. a. Dieser Mann mit den langen Haaren ist mein Cousin.

b. * Dieser mit den langen Haaren Mann ist mein Cousin.

(136) En. a. My friend in Cambridge has written a book. b. * My in Cambridge friend has written a book.

(137) Da. a. Min ven i Cambridge har skrevet en bog.

b. * Min i Cambridge ven har skrevet en bog.

(138) Ge. a. Mein Freund in Cambridge hat ein Buch geschrieben. b. * Mein in Cambridge Freund hat ein Buch geschrieben.

(139) En. a. My friend who is a physicist has written a book. b. * My who is a physicist friend has written a book.

(140) Da. a. Min ven som er fysiker har skrevet en bog.

b. * Min som er fysiker ven har skrevet en bog.

(29)

Adjectives, (142), and participles, (145), on the other hand, precede the NP they modify, even though at least in English, "complex" participles cannot precede their NP, (148).

(142) En. a. * The milk organic is not always the most expensive. b. The organic milk is not always the most expensive.

(143) Da. a. * Den mælk økologiske er ikke altid den dyreste. b. Den økologiske mælk er ikke altid den dyreste.

(144) Ge. a. * Die Milch ökologische ist nicht immer die teuerste. b. Die ökologische Milch ist nicht immer die teuerste.

(145) En. a. * The messages annoying kept appearing on the screen. b. The annoying messages kept appearing on the screen.

(146) Da. a. * De meddelelser irriterende blev ved med at dukke op på skærmen. b. De irriterende meddelelser blev ved med at dukke op på skærmen.

(147) Ge. a. * Die Botschaften nervigen tauchten immer wieder auf. b. Die nervigen Botschaften tauchten immer wieder auf.

(148) En. a. The students reading their e-mail weren't listening. b. * The reading their e-mail students weren't listening.

(149) Da. a. * De studerende (deres) e-mail-læsende hørte ikke efter. b. De (*deres) e-mail-læsende studerende hørte ikke efter.

(150) Ge. a. * Die Studenten ihre E-mail lesenden hörten nicht zu. b. Die ihre E-mail lesenden Studenten hörten nicht zu. (151)

Although English and Danish and German all have the order N°-complement, this order is not universal, in e.g. Korean, the noun must follow its complement:

(30)

Exercise

Are the following languages N°-complement or complement-N°: French, Turkish, Japanese, Icelandic and Hindi?

(154) Fr. a. la déscription de l'accident b. * de l'accident la déscription (155) Tu. a. * yazılma kaza -nın b. kaza -nın yazılma (156) Ja. a. * setsumei jiko -no b. jiko -no setsumei (157) Ic. a. lýsingin á slysinn b. * á slysinn lýsingin (158) Hi. a. * varnan ghaTnaa kaa b. ghaTnaa kaa varnan

description (of) accident (of) (of) accident (of) description ("the description of the accident")

(152) Ko. a. * Kŭ tosi-ŭi p'akoe DEF town.GEN destruction

b. P'akoe kŭ tosi-ŭi Destruction DEF town.GEN

'The destruction of the town' (153) Ko. a. * Kŭ sako-ŭi myosa

DEF accident.GEN description

b. Myosa kŭ sako-ŭi

description DEF accident.GEN

(31)

5.

The phrase structure of DP

5.1

Introducing the DP-analysis

In many treatments, including not just Chomsky (1981, 154), Radford (1988, 173–96) and the first three chapters of Haegeman & Guéron (1999), but also e.g. Jensen (2012; 2014), constituents such as [these green chairs], [the brutal destruction of the city], [a new hammer], [John's book] and [he] are taken to be NPs.

However, following the analysis in Abney (1987), in Chomsky (1995, 246, (8a)), as applied to

Scandinavian in e.g. Hellan (1986), Delsing (1993; 1998) and Julien (2005), I would like to argue that there are good reasons to assume that the NP does not include the determiner, cf. e.g. Vikner (2014b). The minimal NP thus consists only of the head (N°) followed by its complement. Modifiers (such as AdjPs or PPs) may be adjoined to the left or to the right of NP.

(159)

(160) En. a. a [NP [NP picture [PP of the old castle]] [PP in a silver frame] ] b. a [NP [AdjP small] [NP picture [PP of the old castle]] ]

Da. c. et [NP [NP billede [PP af det gamle slot]] [PP i en sølvramme ] d. et [NP [AdjP lille] [NP billede [PP af det gamle slot]] ]

Then what about the determiner, such as these in these green chairs, my in my favourite movie, or a in a small picture, (142)?

The determiner is analysed as the head (D°) of its own projection, a Determiner Phrase (DP). (This means that the typical subject or object of a clause is now a DP, rather than an NP.)

I thus suggest replacing (161)a/(162)a with (161)b/(162)b:

NP/Det-ANALYSIS DP-ANALYSIS

(161)

picture

complement of the old castle NP modifier small picture complement of the old castle

(32)

(162)

5.2

Differences between NP and DP

One argument for making a distinction between DP and NP is that NPs and DPs occur under

different circumstances, i.e. they have different distributions. This can be seen in cases where an

NP can occur without being part of a DP:

(163) En. a. [NP Nice man] though he is, my uncle can be a bit boring. NP b. * [DP A [NP nice man]] though he is, my uncle can be a bit boring. *DP

(164) En. a. * I met [NP niceman]. *NP

b. I met [DP a [NP niceman]]. DP

(165) En. a. * I consider my uncle [NP niceman]. *NP

b. I consider my uncle [DP a [NP niceman]]. DP

(166) Da. a. [NPStor dyreven] som han var, den lille Emil, lukkede han katten ud. NP b. * [DP En [NPstor dyreven ]] som han var, den lille Emil, lukkede han katten ud. *DP

(A) great animal-lover as he was, little Emil, he let the cat out

(167) Da. a. * Ida kender [NP stor dyreven]. *NP

b. Ida kender [DP en [NP stor dyreven]]. DP

(33)

(In the NP/Det-analysis, (163) and (166) would have to be only N' (N-bar), which would make them into something very rare, a context in which an X-bar category is possible but the corresponding XP impossible.)

Another difference between DP and NP is the fact that DPs may be substituted by pronouns like it or they/them, whereas NPs may be substituted by one/ones.

(168) En. [DP These beautiful pictures of the palace ] had been lost for a long time, but now

[DP they ] have turned up again.

(169) En. These [NP beautiful pictures of the palace ] are much less valuable than

those [NP ones ].

(170) En. These beautiful [NP pictures of the palace ] are much less valuable than

those awful [NP ones ].

(In the NP/Det-analysis, one would substitute an N-bar category only.)

5.3

Parallels between pronouns and (other) determiners

Another argument for distinguishing between DP and NP is that it allows an analysis of pronouns as a kind of determiner:

(171) OLD DISCARDED NP/Det-ANALYSIS

In the old NP-analysis, (171), pronouns were N°, whereas determiners were a category of their own. (172) NEW DP-ANALYSIS

In the (new) DP-analysis, (172), pronouns and other determiners are instances of D°. Taking pronouns to be D°s (with no sisters) gives a better account of a large number of elements across languages that occur both as pronouns and as articles/demonstratives, e.g.

In the DP-analysis, pronouns and other determiners are all instances of D°. Taking pronouns to be D°s (with no sisters) gives a better account of the large number of elements across languages that occur both as pronouns and as articles/demonstratives – not just in English but also in other languages: (173) a. En. that and that book

b. Da. den and den bog c. Ge. das and das Buch d. Fr. le and le livre b. a. NP that book Det that NP

I haven't read I haven't read

b. a. DP that NP N° book that DP

References

Related documents

This perception of a longer time to recoup the initial cost of clean diesel technology has been the key reason for this study of the total cost of ownership that compares

[r]

How the study was conducted The researchers used a 3-D global atmospheric download to predict how the radioactive material download move over earth and a health-effects model to see

In summary and taking into account the resonance characteristics of the ACUREX plant, the main contribution of this paper, is to improve a gain schedul- ing (GS) predictive

Quality: We measure quality (Q in our formal model) by observing the average number of citations received by a scientist for all the papers he or she published in a given

The bending strength tests of full size Norway spruce posts with a cross section of 66x66 mm demonstrated changes of the timber strength after heat treatment, such as a decrease

As the file detection test is the most requested test by users (and magazines) and also due to some business requirements (our file detection test results are used/required by

Given the small size of the summer school group for whom scores were provided, and the pilot nature of the intervention, this is not definitive evidence of a harmful impact