T he basic stru ctu re fo r a free relative like that in I d o n ’t like w ho yo u ju s t m e t is repeated here:
(59) [d [d < w ho> ] [c [d <w ho> ] you ju s t m et r]]
Section 4 d ealt w ith the question o f how such a structure is in terpreted at LF, and in p articu lar w ith the ro le o f the tw o copies o f the wh operator. In this section, I consider the o ther interface, and in particu lar the question o f w hich o f the tw o copies o f the wh operators is p ronounced.
Since the tw o co p ies o f the relative operator are adjacent, it seem s difficult to find evidence con cern in g w hich one is pronounced. N evertheless, there are som e revealing facts from the ex trap o sitio n o f headed relative clauses th at w ere first discussed by G roos and van R ie m sd ijk 1981. H eaded relative clauses in D utch and G erm an need not alw ays be ad jacen t to th e head th at they m odify. T he exam ples in (60a,b) show that they can be extraposed. H o w e v er extraposition o f the head noun along w ith the relative clause is not p o ssib le (cf. (60c)).
(60) a. D e r H ans h a t [d das G eld [c das e r gestohlen hat]] zuriickgegeben T he H ans has the m oney that he stolen has returned
‘H ans retu rn ed the m oney that he has sto le n .’
b. D er H ans h at [d das G eld t] zuriickgegeben [c das er gestohlen hat] c. * D e r H ans h a t t zuriickgegeben [d das G eld [c das e r g estohlen hat]]
Since free relatives h av e been analysed here in term s o f a structure that parallels the structure o f head ed rela tiv e constructions, there are tw o options. If the pronounced copy
o f the o p e ra to r is the o ne in the head position, it is expected that the sequence you-just- m e t in (59) occurs in the ex traposed position, leaving the op erato r w as behind in object p o sitio n . A ltern ativ ely , i f the copy o f the operator in the sp ecifier o f C position is p ro n o u n ce d , it is e x p e cte d th at the w hole string w h o -you-just-m et is extraposed.
I f w e c o n sid e r a n o th e r aspect o f the extraposition o f headed relatives, w e get a c le a r pred ictio n c o n c ern in g w hich o f the tw o possibilities should be realised. As noted in c h a p te r 2, relativ e clauses can extrapose only w hen they contain (overt) m aterial at th e c o m p le m e n tiz e r level. T hus, (61) show s that a relative clau se can occur w ith or w ith o u t the c o m p le m e n tiz e r that w hen it is adjacent to the noun it m odifies, b ut w hen the rela tiv e clause is e x tra p o sed the com plem entizer is obligatory. S im ilarly, in (62) the rela tiv e o p erato r w ho is optional w here the relative is ad jacen t to the head noun, but o b lig a to ry w hen the relativ e clause is extraposed. (The ex am ples are from D oherty 1993:59.)
(61) a. A m oose (that) B ill shot at appeared. b. A m oose ap p eared th at Bill shot at. c. * A m oose ap p eared B ill shot at.
(62) a. T h e m an (w ho) B ill knew arrived yesterday. b. T h e m an arrived yesterday w ho B ill knew. c. * T he m an arrived yesterday Bill knew.
If e x trap o sitio n is only com patible w ith clauses that contain overt m aterial at the c o m p le m e n tiz e r level, there is a clear prediction for ex trap o sitio n o f free relatives. If the relativ e o p erato r w ere pronounced in the head p osition, this w ould leave the ex tra p o sed clausal c o n stitu e n t w ithout any overt m aterial in its c o m p lem en tizer level. It is then ex p ected th at free relatives can only be extraposed if the copy o f the operator inside the relative c lau se is pronounced.
T his is indeed the p attern that is found. The ex traposition pattern o f related free relatives is illu strated in the follow ing exaijiples (G roos and van R iem sd ijk 1981). In
" E x a m p le s c o m p a r a b le to ( 6 2 c ) are gram m atical in E nglish ; H e r e tu r n e d y e s te r d a y th e m o n e y th a t h e h a d s to le n f r o m th e b a n k . T h is d iffe r e n c e b e tw e en E n glish and G erm an is d u e to the fa c t that E n g lish a llo w s h e a v y -N P sh ift rather fr e e ly , w h ile G erm an d o e s not. B ut s e e b e lo w .
S e e H a w k in s 2 0 0 1 fo r s o m e r e le v a n t d isc u s sio n fro m a p r o c e s sin g p e r s p e c tiv e .
(63a) the free rela tiv e is in th e canonical object position. (63b) show s th at the com plete free relativ e c o n stru c tio n can be extraposed in a sim ilar w ay to a relativ e clause in a headed relativ e co n stru ctio n . H ow ever, as (63c) show s, ex trap o sitio n o f the sequence er gesto h len hat, leav in g b e h in d the relative operator w as is not possible.
(63) a. D e r H ans h at [d w as e r gestohlen hat] zuriickgegeben
T h e H ans has w h at he stolen has returned ‘H an s retu rn ed w hat he has sto le n .’
b. D e r H ans hat t zuriickgegeben [d was er gestohlen hat] cf. (62c) c. * D e r H ans hat [d w as t] zuriickgegeben [ e r gestohlen hat] cf. (62b)
As G roos and van R ie m sd ijk note this pattern is problem atic fo r a p roposal like that o f B resnan and G rim sh aw 1978 in w hich the relative o p erator does not occu r as a co n stitu en t o f the rela tiv e clause.
(64) [d w as [c O P er t gestohlen hat]]
G iven such a structure, it m ight be argued that extrap o sitio n in (63a) is the result o f heavy N P shift. W h ile heavy N P shift in G erm an is not as easily av ailab le as in E nglish, the follow ing ex am p les show th at it does occur:
(65) a. D er iran isch e P ra sid e n t M oham m ed C hatam i dagegen v eru rteilte the Iranian p resid e n t M . C. in-contrast cond em n ed sc h arf die Z e rsto ru n g h istorischer B uddha-S tatuen als u n islam isch. sharp the d e stru c tio n historical B. statues as u n -islam ic
‘In contrast, the Iranian president Mohammed Chatami sharply comdemned the destruction of the historic Buddha statues as un-islamic.’ {Berliner Z 6 March 2001)
b. D ie U S A veru rteilten sch arf die von den albanischen E xtrem isten the U S A co n d e m e n d sharply the o ( the A lbanian ex trem ists ausgehende G ew alt.
em erging v io le n c e {Tagesspiegel 25.03.2001)
‘T h e U S A c o n d e m n e d sharply the violence caused by A lab an ian e x tre m ists.’ 193
c. A n der H altestelle L o th rin g er Stras se sind eben eingestiegen in die at the stop L o th rin g er Street w ere ju s t entered in the 4 v ier L eute, zw ei altliche Frauen, ein bekiim m erter ein fach er “4 ” fo u r p eo p le tw o elderly w om en a w orried sim ple M an n u n d ein Ju n g e m it einer M iitze und O hrenklappe.
M an and a b o y w ith a cap and ear m u ff (A lfr e d D u b lin , B e rlin A le x a n d e r p la tz p .4 2 ) ‘A t the stop L o th rin g e r Street, four people entered the tram on Line 4: tw o elderly w om en, a w orried sim ple m an, and a boy w ith a cap and ear m u ffs.’
d. A m 25. S e p tem b e r verschied plotzlich an ein em H erzsch lag m ein o n -the 25'^ S e p tem b e r passed-aw ay suddenly at a h eart-attack m y in n ig g elieb ter M ann, unser lieber Vater, Sohn, B ruder, S ch w ag er und d early -lo v ed m an o ur dear father son brother bro th er-in -law and O nkel M ax RUst im noch nicht vollendeten A lte r von 55 Jahren. U ncle M ax R iist in-the n ot yet com pleted age o f 55 years
‘O n S e p tem b e r 25, passed aw ay suddenly with a h eart attack at less than 55 years o f age. M ax Riist, m y beloved husband, o u r dear father, son, brother, b ro th er-in -law , and u n c le .’ (A lfred D oblin, B erlin A le xa n d e rp la tz p.42)
e. W ir k annten nich t sein unerhortes H aupt, darin die A ugenapfel reiften. W e knew n o t his fantastic head, in-w hich the eye-apples ripened ‘W e n ever knew his fantastic head in w hich the eye-apples rip e n e d .’ (R ain er M a ria R ilke, ‘A rchaischer Torso A p o llo s’)
A p ro b le m for such a proposal is that, as noted by G roos and van R iem sd ijk 1981, heavy N P shift, even though is not im possible, only seem s to o c c u r in stylistically m ark ed contexts. T hus the phrase verurteilte{e) s c h a r f m (65a,b) is very typical o f new s reports. T h e exam ple in (65b) occurs in the context o f the ob itu ary in (65c), and indeed the use o f H eavy N P sh ift is fairly com m on an this genre. F inally, (65e) is from a poem . T h ese restrictio n s are at odds w ith the lack o f stylistic m ark ed n ess in structures in v o lv in g ex traposed free relatives.
M o re o v e r, it is not clear how an analysis like that in (64) could explain the u n g ram m atic lity o f (63c). If the relative operator is in the p o sition occu p ied by the head no u n in a h ead ed relativ e construction, there seem s to be no obvious reason that w ould b lo c k ex tra p o sitio n o f the relativ e clause that m odifies it.
T h e ex trap o sitio n facts d iscussed by G roos and van R iem sd ijk 1981 then provide a stro n g arg u m en t ag ain st an analysis o f free relatives along the lines o f (64), but, as sh o w n above, they are fully in line w ith the present proposal.
7 C o n c lu s io n
T h is c h a p te r dealt w ith the analysis o f free relatives. T he p ro posed analysis incorporates tw o fea tu re s o f ea rlie r analyses. In particular, it preserves the structural sim ilarity b etw een free relatives and headed relative constructions o f the C om p analysis (G roos and van R ie m sd ijk 1981), w hile at the sam e tim e it inherits the ability o f the H ead analysis (B resnan and G rim shaw 1978) to account for m atch in g effects in a straig h tfo rw ard way. M oreover, the m ajor counter-argum ents to these earlier analyses (m atch in g effects and ex traposition data, respectively) also find natural accounts. T he claim th at (copies of) the relative operator occurs both inside and outsid e the relative clau se in a free relativ e construction also led to an explanation o f the req u irem en t on op erato rs th at in tro d u ce free relative that they contain a sem antic restriction.
I f th e s u g g e s tio n in n o te 8 a b o v e that the fa c t that o n ly o n e o f th e tw o c o p ie s o f the o p erator is p r o n o u n c e d is d u e to a c o n str a in t a g a in st h a p lo lo g y is on the right track, it has im p lic a tio n s fo r the a n a ly sis o f e x tra p o sitio n . T h e co n stra in t a g a in st h a p lo lo g y p resu m a b ly h o ld s at the P F in terface. T h is m ea n s that it can o n ly e x te n d to the e x tra p o sitio n fa c ts i f ex tra p o sitio n o c c u r s s o m e tim e du rin g s p e ll-o u t a fter th e h a p lo lo g y p o lic e has a lre a d y left.
Ch a p t e r Fiv e C o n c lu s io n
In Chapter 1 o f this dissertation, I developed a new model o f phrase structure, whose major ingredients can be stated as in (1) to (3). In chapters 2, 3, and 4, I investigated the predictions o f different aspects o f the proposed theory in different empirical domains. In this chapter, I give a b rief overview over how some o f the analytical results follow from w hich o f the assum ptions, and talk a little bit about the implications o f these proposals.
(1) M aterial that is not in the deepest position within its chain can head a structure (cf. A ckem a et al 1993, Neelem an and Weerman 1999).
(2) The concept o f categorial projection is rejected (Brody 1997a, 2000). (3) Phrase structure does not express linear order, only hierarchical structure.
The assumption in ( I) allows for movement that extends clause structure. Since e.g. verb m ovem ent varies across languages, this entails that clause structure is not universal. This in turn implies that the structure o f a particular language must be leam able. This has the consequence that clause structure cannot contain any (lexical) material that remains consistently w ithout overt content (except for the possible effect o f universal semantic conditions). This conclusion directly rules out a whole range o f otherwise plausible analyses. Thus, the derivation o f Vikner’s generalisation and m ost o f the proposals in chapter 2 rely crucially on the idea that there is no empty counterpart to complementizers like that (except as the result o f leamable, language-specific deletion rules such as the doubly-filled-Com p Filter). Similarly, the derivation o f the generalisation that VSO languages must have (overt) preverbal particles would not be possible if empty heads were freely available. B eyond that it becomes possible to derive something like the generalisation proposed by Fukui and Takano 1998:54f that “a functional head H enters into feature checking only if H lacks phonetic content.” In terms o f present assumptions, such a conclusion follows automatically, namely that all derived heads, which are by definition heads w ithout independent phonetic content, require their specifier position to be filled.
The assum ption in (2) allows for the formulation o f head chains that forms the basis o f the analysis o f the interaction between verb movement and selection in chapter 2. (2) is also crucial in m aking it possible to generalise over structures involving verb movement and structures involving adjunction to their highest position in that chapter. W ithout such a generalisation, the identical effect o f the two on the distribution o f complementizers would be difficult to explain, (2) is also a central ingredient o f the derivation o f V2 effects in chapter 3.
The implications o f (3) are somewhat broader (and possibly deeper) than those o f (1) and (2), even though (3) is in some form assumed in most standard assum ption o f the three. (3) is what m akes it possible to represent phrase structure in terms o f sets o f treelets.