A quantitative model of word order and movement  
in English, Dutch and German complement constructions 
KARIN HARBUSCH 
Computer Science Department 
University of Koblenz-Landau 
PB 201602, 56016 Koblenz/DE 
harbusch@uni-koblenz.de 
GERARD KEMPEN 
Cognitive Psychology Unit 
Leiden University, and 
Max Planck Institute, Nijmegen/NL 
kempen@fsw.leidenuniv.nl 
 
Abstract We present a quantitative model of word 
order and movement constraints that enables a 
simple and uniform treatment of a seemingly het-
erogeneous collection of linear order phenomena 
in English, Dutch and German complement con-
structions (Wh-extraction, clause union, extraposi-
tion, verb clustering, particle movement, etc.). Un-
derlying the scheme are central assumptions of the 
psycholinguistically motivated Performance Gram-
mar (PG). Here we describe this formalism in de-
clarative terms based on typed feature unification. 
PG allows a homogenous treatment of both the 
within- and between-language variations of the or-
dering phenomena under discussion, which reduce 
to different settings of a small number of quantita-
tive parameters. 
1. Introduction 
We propose a quantitative model for expressing 
word order and movement constraints that enables 
a simple and uniform treatment of a heterogeneous 
collection of linear ordering phenomena in English, 
Dutch and German complement structures. Under-
lying the scheme are central tenets of the psy-
cholinguistically motivated Performance Grammar 
(PG) formalism, in particular the assumption that 
linear order is realized at a late stage of the gram-
matical encoding process. The model is described 
here in declarative terms based on typed feature 
unification. We show that both the within- and be-
tween-language variations of the ordering phe-
nomena under scrutiny reduce to differences be-
tween a few numerical parameters. 
The paper is organized as follows. In Section 2, 
we sketch PG's hierarchical structures. Section 3, 
the kernel of the paper, describes the linearization 
and movement model. In Section 4, we turn to cen-
tral word order phenomena in the three target lan-
guages. Section 5, finally, contains some conclu-
sions. 
2.    Hierarchical structure in PG 
PG's hierarchical structures consist of unordered 
trees ('mobiles') composed out of elementary build-
ing blocks called lexical frames. These are 3-tiered 
mobiles assembled from branches called segments. 
The top layer of a frame consists of a single 
phrasal node (the 'root'; e.g. Sentence, Noun 
Phrase, ADJectival Phrase, Prepositional Phrase), 
which is connected to one or more functional 
nodes in the second layer (e.g., SUBJect, HeaD). 
At most one exemplar of a functional node is al-
lowed in the same frame. Every functional node 
dominates exactly one phrasal node ('foot') in the 
third layer, except for HD which immediately 
dominates a lexical (part-of-speech) node. Each 
lexical frame is 'anchored' to exactly one lexical 
item: a lemma (printed below the lexical node serv-
ing as the frame's HeaD). A lexical frame encodes 
the word category (part of speech), subcategoriza-
tion features, and morphological diacritics (person, 
gender, case, etc.) of its lexical anchor (cf. the 
elementary trees of Tree Adjoining Grammar 
(TAG; e.g. Joshi & Schabes, 1997). 
Associated with every categorial node (i.e., 
lexical or phrasal node) is a feature matrix, which 
includes two types of features: agreement features 
(not to be discussed here; see Kempen & Harbusch, 
forthcoming) and topological features. The latter 
play a central role in the linear ordering mecha-
nism. Typed feature unification of topological fea-
tures takes place whenever a phrasal foot node of a 
lexical frame is replaced (substituted for) by a lexi-
cal frame. Substitution is PG's sole composition 
operation. Substitution involves unification of the 
feature matrices that are associated with the substi-
tuted phrasal foot node and the root node of the 
substituting lexical frame. Substitution gives rise to 
the derivation tree of a well-formed syntactic struc-
ture iff the phrasal foot node of all obligatory seg-
ments of each lexical frame successfully unifies 
with the root of another frame. The tree in Figure 1 
is well-formed because the MODifier segments are 
not obligatory. 
NP
HD
pro
we
S
S UBJ
NP
HD
v
hate
DOB J
NP
NP
HD
n
Dana
NP
HD
n
Kim
S
S UBJ
NP
HD
v
know
CMP
S
MOD
A P |PP| S
MOD
A P |PP| S
 
Figure 1. Simplified lexical frames underlying the 
sentences We know Dana hates Kim and Kim we know 
Dana hates (example from Sag & Wasow,1999). Or-
der of branches is arbitrary. Filled circles denote sub-
stitution. (The feature matrices unified as part of the 
substitution operations are not shown.) 
3. Linear structure in PG 
The above-mentioned topological features are as-
sociated with the phrasal root nodes of lexical 
frames. Their value is a feature matrix specifying a 
'topology', that is, a one-dimensional array of left-
to-right slots. In this paper we will only be con-
cerned with topological features associated with S-
nodes. They serve to assign a left-to-right order to 
the segments (branches) of verb frames (i.e. lexical 
frames specifying the major constituents of 
clauses). On the basis of empirical-linguistic ar-
guments (which we cannot discuss here), we pro-
pose that S-topologies of English, Dutch and Ger-
man contain exactly nine slots: 
E F1 F2 F3 M1 M2 M3 M4 E1 E2 
D/G F1 M1 M2 M3 M4 M5 M6 E1 E2 
The slots labeled Fi make up the Forefield (from 
Ger. Vorfeld); the Mj slots belong to the Midfield 
(Mittelfeld); the Ek's define the Endfield (Nachfeld; 
terms adapted from traditional German grammar; 
cf. Kathol, 2000). Table 1 illustrates which clause 
constituents select which slot as their 'landing site'. 
Notice, in particular, that the placement conditions 
refer not only to the grammatical function fulfilled 
by a constituent but also to its shape. For instance, 
while the Direct Object takes M3 as its default 
landing site, it selects F1 if it is a Wh-phrase or 
carries focus, and M2 if it is a personal pronoun 
(it). In terms of Figure 1, if Kim carries focus, it 
may occupy slot F1 of the topology associated with 
the complement clause headed by hate. 
Table 1. Examples of topology slot fillers (English). 
MODifier constituents are not shown. Precedence be-
tween constituents landing in the same slot is marked 
by "<". 
Slot Filler 
F1 Declarative main clause: Topic, Focus (one 
constituent only) 
Interrogative main clause: Wh-constituent.  
Complement clause: Wh-constituent (includ-
ing CoMPlementizeR whether/if) 
F2 Complement clause: CoMPLementizeR that 
F3 Subject (iff non-Wh) 
M1 Pre-INFin. to < HeaD (oblig.) < PaRTicle 
M2 Direct OBJect (iff personal pronoun) 
Interrogative main clause: SUBJect (iff 
     non-Wh); SUBJ < DOBJ 
M3 Indirect OBJect < Direct OBJect (non-Wh) 
M4 PaRTicle  
E1 Non-finite CoMPlement of 'Verb Raiser' 
E2 Non-finite CoMP of 'VP Extraposition' verb 
Finite CoMPlement clause 
How is the Direct Object NP Kim 'extracted' 
from the subordinate clause and 'moved' into the 
main clause? Movement of phrases between 
clauses is due to lateral topology sharing. If a sen-
tence contains more than one verb, each of the verb 
frames concerned instantiates its own topology. 
This applies to verbs of any type, whether main, 
auxiliary or copula. In such cases, the topologies 
are allowed to share identically labeled lateral (i.e. 
left- and/or right-peripheral) slots, conditionally 
upon several restrictions to be explained shortly. 
After two slots have been shared, they are no 
longer distinguishable; in fact, they are the same 
object. In the example of Figure 1, the embedded 
topology shares its F1 slot with the F1 slot of the 
matrix clause. This is indicated by the dashed bor-
ders of the bottom F1 slot: 
F1  F3 M1     E2 
•
 
 we know     
•
 
↑        ⇑ 
Kim  Dana hates      
In sentence generation, the overt surface order 
of a sentence is determined by a Read-out module 
that traverses the hierarchy of topologies in left-to-
right, depth-first manner. Any lexical item it 'sees' 
in a slot, is appended to the output string. E.g., Kim 
is seen while the Reader scans the matrix topology 
rather than during its traversal of the embedded to-
pology. See Figure 2 for the ordered tree corre-
sponding to Kim we know Dana hates
1
. 
S
SUBJ
NP
we
HD
v
know
CMP
S
SUBJ
NP
Dana
HD
v
hate
DOBJ
NP
 Kim
F1 F3 M1 E2
F3 M1
 
Figure 2. Fronting of Direct Object NP Kim due to pro-
motion (cf. Figure 1). Rectangles represent (part of) the 
topologies associated with the verb frames. 
The number of lateral slots an embedded topol-
ogy shares with its upstairs neighbor is determined 
by the parameters LS (left-peripherally shared area) 
and RS (right-hand share). The two laterally shared 
areas are separated by a non-shared central area. 
The latter includes at least the slot occupied by the 
HeaD of the lexical frame (i.e., the verb) and usu-
ally additional slots. The language-specific pa-
rameters LS and RS are defined in the lexical en-
tries of complement-taking verbs, and dictate how 
(part of) the feature structure associated with the 
foot of S-CMP-S segments gets instantiated. For 
instance, the lexical entry for know (Figure 1) 
states that LS=1 if the complement clause is finite 
and declarative. This causes the two S-nodes of the 
CoMPlement segment to share one left-peripheral 
slot, i.e. F1. If the complement happens to be inter-
rogative (as in We know who Dana hates), LS=0, 
implying that the F1 slots do not share their con-
tents and who cannot 'escape' from its clause. 
In the remainder of this Section we present a 
rule system for lateral topology that is couched in a 
typed feature logic and uses HPSG terminology. 
The system deals with a broad variety of movement 
phenomena in English, Dutch and German. 
We define a clausal topology as a list of slot 
types serving as the value of the topology ("TPL") 
feature associated with S-nodes: 
  S [TPL 〈F1t,F2t,F3t,M1t,M2t,M3t,M4t,E1t,E2t〉] 
                                           
1
The value of a TPL feature may be a disjunctive set of 
alternative topologies rather than a single topology. 
See the CMP-S node of Figure 3 for an example. 
As for syntactic parsing, in Harbusch & Kempen 
(2000) we describe a modified ID/LP parser that can 
compute all alternative hierarchical PG structures li-
censed by an input string. We show that such a parser 
can fill the slots of the topologies associated with any 
such structure in polynomial time.  
for English, and 
  S [TPL 〈F1t,M1t,M2t,M3t,M4t,M5t,M6t,E1t,E2t〉] 
for Dutch and German. Slot types are defined as 
attributes that take as value a non-branching list of 
lemmas or phrases (e.g. SUBJect-NP, CoMPle-
ment-S or HeaD-v). They are initialized with the 
value empty list, denoted by "〈〉" (e.g., [
F1t 
F1 〈〉]). 
Lists of segments can be combined by the ap-
pend operation, represented by the symbol " ⊕". 
The expression "L1 ⊕L2" represents the list com-
posed of the members of L1 followed by the mem-
bers of L2. We assume that L2 is non-empty. If L1 
is the empty list, "L1 ⊕L2" evaluates to L2. Slot 
types may impose constraints on the cardinality 
(number of members) of the list serving as its 
value. Cardinality constraints are expressed as sub-
scripts of the value list. E.g., the subscript "c=1" in 
[
F1t 
F1 〈〉
c=1
]
 states that the list serving as F1's value 
should contain exactly one member. Cardinality 
constraints are checked after all constituents that 
need a place have been appended. 
Depending on the values of sharing parameters 
LS and RS, the list can be divided into a left area 
(comprising zero or more slot types), the central 
area (which includes at least one slot for the HeaD 
verb), and the right area (possibly empty). Topol-
ogy sharing is licensed exclusively to the lateral 
areas. LS and RS are set to zero by default; this ap-
plies to the topologies of main clauses and adver-
bial subclauses. The root S of a complement clause 
obtains its sharing parameter values from the foot 
of the S-CMP-S segment belonging to the lexical 
frame of its governing verb. For example, the lexi-
cal entry for know states that the complement of 
this verb should be instantiated with LS=1 if the 
clause type (CTYP) of the complement is declara-
tive. This causes the first member of the topologies 
associated with the S-nodes to receive a corefer-
ence tag (indicated by boxed numbers): 
S TPL  1 F1,F2,...,E2
[]
CMP
S
TPL 1 F1,F2,...E2
CTYP Decl
 
 
 
 
 
 
 
If, as in the example of Figure 1, know's comple-
ment is indeed declarative, the foot of the comple-
ment segment can successfully unify with the root 
of the hate frame. As a consequence, the F1 slot of 
the complement clause is the same object as the F1 
slot of the main clause, and any fillers will seem to 
have moved up one level in the clause hierarchy: 
 
S 
TPL 1 F1,F2,...E2
CTYP Decl
 
 
 
 
 
 
S 
TPL F1,F2,...,E2
CTYP Decl
 
 
 
 
 
 
  
⇒ S 
TPL 1 F1,F2,...,E2
CTYP Decl
 
 
 
 
 
 
 
Filling a slot also involves coreference tags. For 
example, the HeaDs of English verb frames obtain 
their position in the local topology by looking up 
the slot associated with the coreference tag: 
  
 S  TPL F1,...,M1 ⊕ 1 ,...,E2
[ ]
HD
 v  1 LEMMA hate
[]
 
The information associated with the foot node of 
the HeaD segment will now be appended to the 
current content, if any, of slot M1. The same 
mechanism serves to allocate the finite comple-
ment clause (or rather its root S-node) to slot E2 of 
the matrix clause: 
   S   TPL 1 F1,...,E1, E2 ⊕ 2
[]
CMP
   S   2
TPL 1 F1,...,E1, E2
CTYP Decl_finite
 
 
 
 
 
 
 
Other clause constituents receive their landing site 
(cf. Table 1) in a similar manner. Figure 2 depicts 
the configuration after Fronting of NP Kim. 
Figure 3 below includes a paraphrase where the 
focus on Kim is stressed prosodically rather than 
by Fronting. This is indicated by the disjunctive set 
carrying the tag 4 . In sentence generation, the 
Read-out module selects one alternative, presuma-
bly in response to pragmatic an other context fac-
tors. In parsing mode, one or the alternatives is 
ruled out because it does not match word order in 
the input string. 
The formalism defined so far yields unordered 
hierarchical structures. However, the values of the 
TPL features enable the derivation of ordered out-
put strings of lexical items. As indicated above in 
connection with Figure 2, we assume that this task 
can be delegated to a simple Read-out module that 
traverses the clause hierarchy in a depth-first man-
ner and processes the topologies from left to right
2
. 
If a slot is empty, the Reader jumps to the next slot. 
If a slot contains a lexical item, it is appended to 
                                           
2
 A slot may contain more than one phrase (e.g., Direct 
and Indirect OBJect in slot M3; cf. Table 1). We assume 
they have been ordered as part of the append operation, 
according to the sorting rule associated with the slot. 
the current output string and tagged as already 
processed. It follows that, if a slot happens to be 
shared with a lower topology, its contents are only 
processed at the higher clause level, i.e., undergo 
promotion. 
4. Linearization of complement clauses in 
English, Dutch and German 
The PG formalism developed above provides a 
simple quantitative linearization method capturing 
both within-clause and between-clause phenomena. 
The assignment of constituents to topology slots 
(including, e.g., scrambling in Dutch and German) 
has been dealt with in Kempen & Harbusch (in 
press; forthcoming). In the present paper we focus 
on promotion in complement constructions — a 
domain where the three target languages exhibit 
rather dissimilar ordering patterns. We highlight 
the fact that PG enables highly similar treatments 
of them, differing only with respect to the settings 
of some quantitative parameters. 
The movement (promotion) phenomena at issue 
here depend primarily on the values assigned to 
sharing parameters LS and RS in five different 
types of complement clauses. These settings are 
shown Table 2. They are imported from the lexi-
con and control the instantiation of the TPL feature 
of the root S-node of the complement. We begin 
with some illustrations from English. 
Table 2. Size of the left- and right-peripheral shared 
topology areas (LS and RS) in diverse complement 
constructions. 
Clause type English Dutch German 
Interrogative LS=0 
RS=0 
LS=0 
RS=1 
LS=0 
RS=1 
Declarative & Finite LS=1 
RS=0 
LS=1 
RS=1 
LS=1 
RS=1 
Decl. & Non-Finite, 
    VP Extraposition 
LS=3 
RS=0 
LS=1 
RS=1 
LS=1 
RS=1 
Decl. & Non-Finite, 
      Verb Raising 
LS=3 
RS=0 
LS=4:6 
RS=1 
LS=5 
RS=1 
Decl. & Non-Finite, 
   Third Construction 
   n.a. 
LS=1:6 
RS=1 
LS=1:6 
RS=1 
The non-finite complements of do and have in 
sentence (1) below are both declarative. (Cf. the 
paraphrase "For which person x is it the case that I 
have to call x", which highlights the scope of who.) 
It follows that LS=3 in both complements. Notice 
that do is treated as a 'Verb Raiser', have (in have 
to) as a VP Extraposition verb. 
Figure 3. Analysis of Kim we 
know Dana hates (cf. Figure 
1) and We know Dana hates 
Kim. The versions correspond 
to different options of the 
topology  value associated 
with the CoMPlement 
(indicated within curly 
brackets). Empty slots are not 
shown in the TPL features. 
 
(1) Who do I have to call? 
F1   M1    E1 E2 
•
   do I   
•
  
↑       ⇑  
   have     
•
 
↑        ⇑ 
who   to call      
In example (2), the lower clause is finite and de-
clarative —cf. the paraphrase “For which person x 
is it the case that you said that John saw x”. The 
scope of who exceeds its ‘own’ clause and includes 
the matrix clause. In (3), on the other hand, the 
scope of the interrogative pronoun does not include 
the main clause (“I know for which person x it is 
the case  that John saw x”). Therefore, the com-
plement is interrogative and cannot share its F1 
slot with that of the main clause. 
 (2) Who did you say John saw?  
F1  F3 M1 M2   E1 E2 
•
   did you   
•
  
↑       ⇑  
   say     
•
 
↑        ⇑ 
who  John saw      
 (3) I know who John saw 
F1  F3 M1     E2 
  I know     
•
 
        ⇑ 
who  John saw      
The system predicts 'island effects' as in (4).  
(4) a.  Who did you claim that you saw? 
 b.*Who did you make the claim that you saw? 
The lexical frame of the verb claim includes an 
S-CMP-S segment identical to that of know above 
(repeated here for convenience): 
   S   TPL 1 F1,...,E1, E2 ⊕ 2
[]
CMP
   S   2
TPL 1 F1,...,E1, E2
CTYP Decl_finite
 
 
 
 
 
 
 
The feature matrices of root and foot nodes of this 
segment both specify a TPL feature referencing the 
slot F1. This enables insertion of coreference tag 
2  and thus promotion of the filler of slot F1. 
However, the complement segment of the noun 
claim is rooting in an NP node, which cannot have 
a TPL feature with type F1t. 
  
  NP   ...
[]
CMP
   S   
TPL 1 F1
CTYP Decl_finite
 
 
 
 
 
 
 
So, tag 
  
1
 is meaningless here, ruling out promo-
tion in (4b). 
Turning now to Dutch, we first refer to Table 3, 
which specifies some important landing sites for 
major clause constituents. Because of the similarity 
with German, we combine the two languages. First, 
we illustrate question formation. 
Dutch interrogative main clauses feature Sub-
ject-Verb inversion without the equivalent of do-
insertion (cf. 5). 
(5) a. M1 Zag M2 je M3 dat? 
    saw      you      that 
   ‘Did you see that?’ 
 b. F1 Wie M1 zag M3 dat? 
     who      saw       that 
    ‘Who saw that?’ 
 c. F1 Wat M1 zagen M2 ze? 
   ‘What did they see?’ 
Because the complement in (6) is interrogative, 
the sharing rule in Table 2 prohibits left-peripheral 
sharing (LS=0). 
(6)  Zij vroeg of ik hem kende 
  She asked whether I him knew 
 ‘She asked whether I knew him’ 
F1 M1 M2 M3   M6  E2 
zij vroeg       
•
         ⇑ 
of  ik hem   kende   
Table 3. Examples of topology slot fillers (Dutch and 
German). Precedence between constituents landing in 
the same slot is marked by "<". 
Slot Filler 
F1 
Declarative main clause: SUBJect, Topic or 
Focus (one constituent only)  
Interrogative main clause: Wh-constituent, 
including Du. of  and Ger. ob 'whether' 
Complement clause: Wh-constituent 
M1 
Main clause: HeaD verb 
Complement clause: CoMPLementizeR 
     dat/om (Du.), dass (Ger.) 
M2 
Subject NP (iff non-Wh), Direct Object 
    (iff personal pronoun) 
M3 
Direct < Indirect OBJect (iff non-Wh) 
M4 
PaRTicle (Du. only) 
M5 
Non-finite CoMPlement of Verb Raiser 
M6 
Subordinate clause: 
  Du.: Pre-INFinitive te < HeaD verb 
  Ger.: PaRTicle < Pre-INFinitive zu < HeaD 
E1 
Non-finite Complement of Verb Raiser (Du. 
only) 
E2 
Non-finite CoMP of VP Extraposition verb 
Finite Complement 
The subordinate clause in (7) features clause 
union, causing the auxiliary zal to intervene be-
tween   the  Direct hem the latter's governor bellen. 
The left-peripheral sharing area may vary between 
4 and 6 slots (LS=4:6). Because hem lands in M3, 
i.e. in the shared area, it is promoted. The remain-
der of the lower topology, including the HeaD bel-
len itself, occupies E1 — one of the options of the 
complement of a Verb Raiser. The other option, 
with the complement in M5 (giving bellen zal) is 
also allowed. 
(7) ...dat ik hem zal bellen 
    that I him will phone 
 '...that I will phone him' 
 M1 M2  M3   M6 E1  
 dat ik 
•
   zal 
•
  
   ↑    ⇑  
   hem   bellen   
Sentence (8) illustrates the treatment of 'particle 
hopping'. The positions marked by "
∧
" are gram-
matical alternatives to the particle (op) position 
mentioned in the example; no other positions are 
allowed. Given LS=4:6 for complements of Verb 
Raisers, it follows that hem is obligatorily pro-
moted into the higher topology: 
 (8) ...dat ik hem 
∧
 zou 
∧
 hebben op gebeld 
    that I  him  would    have  up called 
 '...that I would have called him up' 
 M1 M2 M3 M4  M6 E1  
 dat ik 
•
   zou 
•
  
   ↑    ⇑  
      hebben 
•
  
   ↑    ⇑  
   hem op  gebeld   
However, sharing of the fifth slot (M4) is optional. 
If this option is realized in the middle topology, the 
order zou op hebben gebeld ensues. If, in addition, 
the middle topology shares M4 with its governor, 
the string comes out as op zou hebben gebeld. 
The treatment of cross-serial dependencies is 
exemplified in (9). In order to deal with this con-
struction, we need to make an additional assump-
tion about the order of constituents that land in the 
same slot but originate from different levels in the 
clause hierarchy. We stipulate that constituents 
from more deeply embedded clauses trail behind 
constituents belonging to higher clauses. This or-
dering can be determined locally within a slot if we 
equip every constituents in the hierarchy with a 
numerical 'clause depth' index (for instance, a Gorn 
number; Gorn, 1967). Given this index (not shown 
in the topology diagram accompanying (9)), the 
order hem de fiets results. 
 (9)  ... dat ik hem de fiets wil helpen maken 
       that I   him the bike want-to help repair 
  '... that I want to help him to repair the bike' 
 M1 M2 M3   M6 E1  
 dat ik 
•
 
•
   wil 
•
  
   ↑  ↑    ⇑  
   hem   helpen 
•
  
       ↑    ⇑  
   de fiets   maken   
We now turn to German, concentrating on 
structures usually labeled "VP Extraposition" (10) 
and "Third Construction" (11). 
(10)  ... dass er uns zwingt es zu tun 
      that  he us  forces  it  to  do 
  '... that he forces us to do it' 
 M1 M2   M3   M6  E2 
 dass er uns   zwingt  
•
 
        ⇑ 
  es    zu tun   
(11) a. ... dass er uns verspricht es zu tun 
      that  he  us   promises  it  to do 
  '... that he promises us to do it' 
 b.  ... dass er es uns zu tun verspricht 
 M1 M2 M3  M5 M6   
 dass er
•
 uns   
•
 verspricht   
     ↑   ⇑    
      es    zu tun   
 c.  ...dass er uns es zu tun verspricht 
 M1 M2 M3  M5 M6   
 dass er uns   
•
 verspricht   
     ⇑     
     es    zu tun   
 d.  ...dass er es uns verspricht zu tun 
 M1 M2 M3   M6  E2
 dass er
•
 uns   verspricht  
•
 
     ↑      ⇑ 
    es    zu tun   
 e. ? ...dass er uns es verspricht zu tun 
 M1 M2 M3   M6  E2 
 dass er uns
•
   verspricht  
•
 
         ↑     ⇑ 
     es   zu tun   
The verb zwingen allows its complement to share 
slot F1 only (LS=1). This prevents promotion of 
the Direct OBJect es. Third Construction verbs like 
versprechen allow a great deal of variation in the 
size of the left-peripherally shared topology area 
(LS=1:6), thereby licensing optional promotion of 
es. However, since es is a personal pronoun, it only 
takes M2 as its landing site (see Table 3). The lat-
ter constraint is violated in (11e). 
5. Discussion 
We have shown that the introduction of topologies 
with a fixed number of slots, in conjunction with 
cross-clause lateral topology sharing, enables a 
simple treatment of word order and movement 
(promotion) in complement structures of the three 
target languages. The considerable within- and be-
tween-language variation typical of these construc-
tions could be analyzed as resulting from different 
settings of a small number of quantitative parame-
ters, in particular the size of shared areas. We 
claim that our approach is conducive to theoretical 
parsimony (and, presumably, computational effi-
ciency). For instance, HPSG-style treatments of 
Wh-movement and Clause Union typically invoke 
very different types of mechanisms (e.g., the 
SLASH or GAP feature for WH-movement, and 
argument composition for Clause Union; cf. Sag & 
Wasow, o.c., and Kathol o.c.). 
  
Elsewhere we have provided a more fine-
grained  discussion of our approach and its psycho-
linguistic motivation (Kempen & Harbusch, in 
press; forthcoming). Future study is needed to find 
out whether the PG approach generalizes to other 
languages. 
Finally, we refer to the PG sentence generator 
for Dutch which was implemented by Camiel van 
Breugel. It covers the ordering phenomena de-
scribed here and in Kempen & Harbusch (forth-
coming) and runs under Java-enabled Internet 
browsers (www.liacs.nl/~cvbreuge/pgw). Vosse & 
Kempen (2000) describe a computational model of 
human syntactic parsing based on a PG-like for-
malism. 

References 

GORN, S. (1967). Explicit Definition and Linguistic 
Dominoes. Systems and Computer Science. To-
ronto: University of Toronto Press. 

HARBUSCH, K. & KEMPEN, G. (2000). Complexity 
of linear order computation in Performance 
Grammar, TAG and HPSG. In: Proceedings of  
Fifth International Workshop on Tree Adjoining 
Grammars and Related Formalisms (TAG+5), 
University of Paris 7, May 2000. 

JOSHI, A.K. & SCHABES, Y. (1997). Tree Adjoin-
ing Grammars. In: Rozenberg, G. & Salomaa, A. 
(Eds.), Handbook of formal languages (Vol. 3). 
Berlin: Springer. 

KATHOL, A. (2000). Linear Syntax. New York: 
Oxford University Press. 

KEMPEN, G. & HARBUSCH, K. (in press). Word 
order scrambling as a consequence of incre-
mental sentence production. In: Haertl, H., Olsen, 
S. & Tappe, H. (Eds.), The syntax-semantics in-
terface: Linguistic structures and processes. Ber-
lin: De Gruyter. 

KEMPEN, G. & HARBUSCH, K. (forthcoming). 
Dutch and German verb clusters in Performance 
Grammar. In: Seuren, P. & Kempen, G. (Eds.), 
Verb clusters in Dutch and German. Amsterdam: 
Benjamins. 

SAG, I.A. & WASOW, T. (1999). Syntactic theory: 
a formal introduction. Stanford CA: CSLI Publi-
cations. 

VOSSE, T. & KEMPEN, G. (2000). Syntactic struc-
ture assembly in human parsing: A computational 
model based on competitive inhibition and a 
lexicalist grammar. Cognition, 75, 105-143. 
