Word order variation in German main clauses: A corpus analysis
Andrea Weber
Computational Linguistics
Saarland University
PO Box 151150
66041 Saarbr¨ucken, Germany
aweber@coli.uni-sb.de
Karin M¨uller
Informatics Institute
University of Amsterdam
Kruislaan 403
1098 SJ Amsterdam, The Netherlands
kmueller@science.uva.nl
Abstract
In this paper, we present empirical data from a
corpus study on the linear order of subjects and
objects in German main clauses. The aim was
to establish the validity of three well-known or-
dering constraints: given complements tend to
occur before new complements, definite before
indefinite, and pronoun before full noun phrase
complements. Frequencies of occurrences were
derived for subject-first and object-first sen-
tences from the German Negra corpus. While
all three constraints held on subject-first sen-
tences, results for object-first sentences varied.
Our findings suggest an influence of grammat-
ical functions on the ordering of verb comple-
ments.
1 Introduction
Word order variation has been described in
great detail by theoretical linguists. There is,
however, also an increasing interest in the topic
from both computational linguists and psy-
cholinguists (e.g. Kaiser and Trueswell (submit-
ted); Keller (2000); Kruiff and Duchier (2003);
Pechmann et al. (1996); R¨oder et al. (2000)).
Nevertheless, the empirical resources that re-
searchers can draw from for their studies are
still very limited, since only a few studies re-
port on the actual amount of word order vari-
ation (see Kempen and Harbusch (2004); Kurz
(2000)). This paper therefore presents a cor-
pus study on the linear order of subjects and
objects in German, and factors related to the
positioning of complements before or after the
verb. Our study is also new in that it looks
at main clauses rather than the Mittelfeld, for
which most ordering principles were originally
intended.
German is a language with a relatively free
word order in which the subject usually pre-
cedes the object, but can also follow it: In (1),
the subject “Turnverein Neur¨onnebeck” pre-
cedes the object “Fairneßpokal”; in (2) the same
object precedes the subject, without changing
the original meaning of the sentence.
(1) SVO
Der Turnverein Neur¨onnebeck gewann den
Fairneßpokal.
The club Neur¨onnebeck (NOM) won the fairness price
(ACC).
(2) OVS
Den Fairneßpokal gewann der Turnverein
Neur¨onnebeck.
The fairness price (ACC) won the club Neur¨onnebeck
(NOM).
For a German newspaper corpus, we inves-
tigate subject-verb-object (SVO) and object-
verb-subject (OVS) sequences, and examine the
extent to which certain ordering constraints in-
fluence the positioning of verb complements. In
particular, we investigate the validity of the
well-known constraints to place given before
new, definite before indefinite, and pronoun be-
fore full noun phrase (NP) complements (cf.
M¨uller (1999); Uszkoreit (1987)) using the Ne-
gra corpus (Brants et al., 1999).
2 Ordering Principles
In this section, we describe the three ordering
principles tested in this study. Whereas the first
principle attends to the contextual dependen-
cies of a sentence, the scope of the second and
the third principle is a sentence.
In scrambling languages, the position of verb
complements can reflect their connection to the
preceding context. In these languages, discourse
new information tends to occur towards the end
of a sentence, whereas discourse old informa-
tion is more likely to occur at the beginning
(cf. Birner and Ward (1998)). Thus, the in-
formation structure of a sentence in a scram-
bling language such as German, can reflect its fit
within a given discourse (Selkirk (1984); Steed-
man (1991))1: When an object precedes a sub-
ject, the object is likely to be given and the
1Different terms and concepts such as theme/rheme,
background/focus and given/new are used in the lit-
subject new; when a subject precedes an ob-
ject, the subject is likely to be given and the
object new, although the canonical subject-first
order is also expected when both complements
are either new or given. We establish for Ne-
gra how often both SVO and OVS main clauses
adhere to this basic order pattern.
German is also a language with definite and
indefinite articles. According to a second linear
ordering principle, definite NPs should tend to
precede indefinite NPs. We also presume, that
definiteness is correlated to the information sta-
tus of complements. As already Chafe (1976)
pointed out, indefiniteness often goes together
with newness, and definiteness with givenness
or newness. Thus, on the NP itself information
status can be partially encoded by the choice of
article. If the correlation with givenness drives
the positioning of definite and indefinite com-
plements, we should find for both SVO and
OVS sentences that definite complements tend
to precede indefinite ones. Another possibility
is, however, that definiteness is bound to gram-
matical functions (i.e. subjects are usually def-
inite). In that case, we would expect to find a
reversal of the ordering principle for definiteness
in OVS sentences.
A third common linear ordering principle
states that pronouns tend to precede full NPs.
Similarly to definiteness, the use of pronouns
can potentially encode information structure.
Almost by definition, pronouns refer to an an-
tecedent in the discourse and represent there-
fore given information (with the possible excep-
tion of indefinite pronouns). Whereas pronoun
complements usually represent given informa-
tion, full NP complements are not necessarily
new. Again, if the correlation with givenness
drives the positioning of pronominalized com-
plements, we should find that pronouns tend
to precede full NPs in both SVO and OVS
sentences. If, however, for example grammati-
cal functions determine which complements are
pronominalized, we might not find this ten-
dency.
3 Corpus Analysis
The Negra corpus (Skut et al., 1998) is an an-
notated collection of 20,602 sentences (355,096
erature to express information structure (for a recent
overview see Kruiff-Korbayov´a and Steedman (2003)).
Since annotations in the present study are based on sin-
gle referents rather than parts of sentences, we distin-
guish between given and new.
tokens) extracted from the German newspaper
Frankfurter Rundschau. The syntactic struc-
ture of sentences is represented in dependency
trees for which the nodes describe constituents
and the edges between the nodes are labeled
with grammatical functions expressing syntac-
tic relations. For our study, we choose the ’Penn
format’, a transformation of the original Negra
treebank, in which crossing edges and traces are
omitted.
3.1 Data extraction
Using a Perl program and the tree-search pro-
gram Tgrep2 (Rhode, 2002) all OVS sentences
in Negra are extracted by looking for object-
verb-subject sequences with the same depth of
embedding. Object and subject themselves as
well as the sentences in which the OVS structure
occur could be complex (see 3).
(3) OVS.
Den Satz von der Vergangenheit, die noch nicht ein-
mal vergangen sei, zitiert auch Peter R¨uhl in seinem
wie stets gescheiten Begleittext zur j¨ungsten CD des
Trompeters Frank Koglmann, einem der wichtigsten
Musiker des europ¨aischen Jazz.
The sentence (ACC) about the past that has not even
passed yet, cites also Peter R¨uhl (NOM) in his as al-
ways smart accompanying text to the latest CD of the
trumpet player Frank Koglmann, one of the most im-
portant musicians in European jazz.
Clausal objects with verbal constructions in
addition to direct and indirect questions are
manually omitted from the list after extraction.
A total of 625 OVS sentences are kept for anal-
ysis (3% of all sentences in Negra). Next, com-
parable 625 (out of 2773) SVO sentences are
chosen.2 Since the total number of SVO sen-
tences in Negra exceeds that of OVS sentences
notably, for practical reasons a subset of 625
SVO sentences is selected. Since the selection
is random, we assume that findings within the
subset are generally valid for SVO sentences. In
addition, for each selected OVS and SVO sen-
tence, the two immediately preceding sentences
in Negra are extracted and serve as context to
determine the information status (given or new)
of complements.
2We do not allow SVO sentences in which the ob-
ject is a reflexive pronoun (e.g. “er f¨urchtet sich”, he is
afraid), since reflexive pronouns are most unlikely to be
fronted. In fact, none of our OVS sentences contained
a fronted reflexive pronoun. Furthermore, reflexive pro-
nouns express coreference within a clause, whereas we
are interested in references across sentence boundaries.
3.2 Data Coding
For each extracted OVS and SVO sentence, the
authors annotate information status, definite-
ness, and pronominalization of its verb comple-
ments. For complex subjects and objects, an-
notations are based on the semantic head of the
complement (i.e. noun or pronoun if the se-
mantic head coincides with the syntactic head).
Definiteness of NPs is assigned by the deter-
miner and information status of NPs by the
noun. Whereas it seems obvious to base annota-
tions on the head of complex complements, the
decision is less clear in the case of more than one
equivalent NP within a complement (i.e. coor-
dinated NPs). Rather than annotating all NPs,
for comparability reasons, we base annotations
on the first NP only, whenever a complement
consists of a listing of more than one NP. Fur-
thermore, some SVO sentences contain both di-
rect and indirect objects, though none of our
OVS sentences do. For these SVO sentences,
only the direct object are considered for the
annotations. An exception are SVO sentences
with reflexive pronouns as indirect object, for
which annotations are based on the direct ob-
ject (1.6% of all SVO sentences).
Givenness. Two preceding context sen-
tences are used to determine whether verb com-
plements present new or given information.3 We
code complements as given if they present ac-
cessible information (Lambrecht, 1994). Access-
ability can either be textually or inferentially
provided. Textual accessability requires an ex-
plicit coreferential antecedent (i.e. the occur-
rence of the same lemma in the context). Infer-
entially accessible complements do not require
an explicit antecedent. Such inferables (Prince,
1981) are assumed to be activated via bridg-
ing inferences (Clark, 1977) that logical rela-
tions such as synonymy, hyponymy, and meron-
omy can provide. In such cases, shared gen-
eral knowledge of the relations between objects
and their components or attributes is assumed.
Whenever more specific knowledge is required
to establish such a relation, however, comple-
ments are considered to be new. The distinc-
tion between general and specific knowledge is
3A recent study by Morton (2000) showed for singu-
lar pronouns, that 98.7% of the times the antecedents
were available within two preceding sentences. His find-
ings are similar with those reported by Hobbs (1976).
Since the context in our study regularly introduces com-
plements, we believe that it gives an adequate picture
of the interaction between information status and word
order.
particularly hard to maintain, since the distinc-
tion is often clearly not binary. For instance,
geographic familiarity with the catchment area
of the Frankfurter Rundschau is considered spe-
cific knowledge: In (4), “Waldstadion” is one of
the local soccer stadiums in Frankfurt. Even
though many local readers of the Frankfurter
Rundschau will know this it can not be assumed
to be known by all readers of the newspaper,
and is therefore coded as new information.
(4) Frankfurt - Waldstadion
Moreover, when two entities X and Y of a po-
tentially larger group Z are considered equally
specific, Y is coded as new information after
X is mentioned in the context (see (5)): Here,
“Klassik” and “Jazz” are two examples of music
styles.
(5) Klassik - Jazz
classical music - jazz
A special case form constructions with “es”.
They are almost exclusively used impersonally
in sentences such as “Karten gibt es”, tickets
are available. We then annotate “es” as new
information.
Definiteness. For all complements, we an-
notate whether they are definite or indefinite.
We largely follow the classification suggested
by Prince (1992). Markers of definite comple-
ments are definite articles, demonstrative arti-
cles, possessive articles, personal pronouns, and
unmodified proper names. Markers of indefi-
nite complements are indefinite articles, zero ar-
ticles, quantifiers, and numerals. Note, that all
quantifiers are marked as indefinite even though
certain quantifiers like all and every have been
suggested in the literature to mark definite de-
scriptions. Furthermore, certain syntactically
indefinite DPs have been argued to be semanti-
cally definite and syntactically definite DPs to
be semantically indefinite. In our study, how-
ever, only formal syntactic properties are criti-
cal for the assignation of definiteness.
Pronominalization. For the annotation of
pronominalization, we check whether comple-
ments are realized as pronouns or full NPs.
4 Results
4.1 Givenness
In 74 cases, the antecedent of an anaphoric
complement occurs prior to our context win-
dow of two sentences. Sentences containing such
complements are excluded from the analysis of
givenness. In addition, 106 sentences with “es”-
complements are excluded for the analyses we
present in this paper. Table 1 shows the ob-
served orderings of given and new complements
for our set of SVO and OVS sentences.
(SVO)
second NP
given O new O
first
NP
given S 113 187
new S 88 175
(OVS)
second NP
given S new S
first
NP
given O 96 144
new O 134 170
Table 1: Frequency of subject (S) and object (O)
pairs ordered by givenness.
In SVO sentences, given subjects precede new
objects more often (187 times) than new sub-
jects precede given objects (88 times). This
tendency is in compliance with the linear or-
dering principle for information structure, even
though the principle is not strictly obeyed as
the 88 cases of new-before-given complements
show. In OVS sentences, both orders occur
about equally often. Given objects precede new
subjects 144 times, and new objects precede
given subjects 134 times. A chi-square test
confirms a significant interaction between sen-
tence type (SVO, OVS) and ordering (given-
before-new, new-before-given; χ2(1) = 14.44,
p < .001). Thus, in contrast with for ex-
ample Finnish (Kaiser and Trueswell, submit-
ted), information structure seems not to be en-
coded in German OVS sentences, in the sense
that fronted given objects do not cue upcom-
ing new subjects for language perceivers. Ob-
viously, factors other than givenness must have
influenced the fronting of objects as is also ap-
parent by the frequent occurrence of OVS sen-
tences with given-before-given (96 times) and
new-before-new (170 times) ordering of comple-
ments. If not, the canonical SVO order would
be expected.
We want to point out a second way of look-
ing at the results, one that involves a more lan-
guage producer-oriented view. Discourse con-
text defines the information status of comple-
ments. Supposing now that a subject has been
introduced in a context, but not an object, we
can check which sentence structure occurs more
often. We find more SVO (187 times) than OVS
sentences (134 times). On the other hand, when
the object of a sentence is given, but the sub-
ject new, we find more OVS (144) than SVO
(88) sentences (χ2(1) = 21.45, p < .001). In-
formation status of complements seems to have
influenced the choice of word order. However,
this interpretation must be taken with caution.
First, we only look at a subset of all SVO sen-
tences of Negra. Second, at least to a cer-
tain degree language producers can not only
choose word order but also the grammatical
function (subject or object) of discourse refer-
ents. The assignment of grammatical functions
to constituents is assumed to happen during
the functional stage of grammatical encoding in
sentence production; only at a later positional
stage the linear order is determined (e.g. Bock
and Levelt (1994)).
4.2 Definiteness
One hundred and six sentences containing “es”-
complements as well as four sentences in which
the object or subject is a citation are excluded
from the analysis. Table 2 shows the ob-
served orderings of definite and indefinite com-
plements.
(SVO)
second NP
def O indef O
first
NP
def S 237 242
indef S 50 62
(OVS)
second NP
def S indef S
first
NP
def O 286 48
indef O 190 25
Table 2: Frequency of subject (S) and object (O)
pairs ordered by definiteness.
In SVO sentences, definite NPs precede indef-
inite NPs 242 times but the reverse ordering oc-
curs only 50 times. Thus, the basic order of defi-
nite before indefinite NPs is largely met for SVO
sentences. For OVS sentences, however, the
preference to place definite NPs before indefi-
nite NPs is reversed. Only 48 times precede def-
inite NPs indefinite NPs, but 190 times precede
indefinite NPs definite NPs (χ2(1) = 205.58,
p < .001). Thus, the ordering principle for def-
initeness is violated in OVS sentences. Rather,
the results suggest a strong correlation between
grammatical function and definiteness 4: Sub-
jects are more often definite and objects indef-
inite, regardless of sentence type. Considering
all four ordering possibilities (see Table 2), how-
ever, this tendency is much stronger for subjects
than objects.
Definiteness and givenness. Not un-
expectedly, definiteness is significantly corre-
lated with givenness for all complements in
both sentence types (all p-values in Pearsons
tests < .01). At a closer look, indefinite NPs
represent more often new information (72%),
whereas definite NPs present given information
(52%) as often as new information. This re-
sult matches corpus studies in other languages,
which found that indefiniteness entails newness
whereas definiteness can entail both givenness
and newness (see e.g. Fraurud (1990)).
Definiteness, givenness, and word or-
der. We are also interested in whether the po-
sitioning of a complement before or after the
verb is influenced by its information status. Is
a definite complement, for example, more likely
to occur before the verb if it also is given? Ta-
ble 3 shows the number of occurrences of both
definite-indefinite and indefinite-definite orders,
split by the information status of the comple-
ments. For both sentence types (SVO and
OVS), neither the positioning of the definite
complement nor the positioning of the indefinite
complement is affected by information status
(in chi-square tests all p-values > .3). Thus, in
SVO sentences, definite subjects precede indefi-
nite objects more often than indefinite subjects
precede definite objects, regardless of whether
subjects and objects present given or new in-
formation. Similarly, in OVS sentences, indefi-
nite objects precede definite subjects more often
than the reverse, regardless of the information
status of the complements.
4.3 Pronominalization
As with definiteness, we exclude 106 sentences
containing “es”-complements. Table 4 shows
the observed orderings of pronoun and full NP
complements. In SVO sentences, 95 times a
pronoun precedes a full NP, whereas a full NP
precedes a pronoun only 33 times. Thus, as
with givenness and definiteness before, the ba-
sic order of pronoun complements before full
4Since we only look at sentences in active and not in
passive voice, subjects in our sentences are always agents
and objects patients. We can therefore not exclude the
possibility that thematic roles rather than grammatical
functions drive determiner choice.
(SVO)
def given def new indef given indef new
def<indef 147 95 71 171
indef<def 26 24 16 34
(OVS)
def given def new indef given indef new
def<indef 26 22 15 33
indef<def 85 105 57 133
Table 3: Linear order frequency of definite-
indefinite pairs for given and new complements.
NP complements is largely met for SVO sen-
tences. In OVS sentences pronouns precede full
NPs 76 times, but the reverse order also occurs
91 times (χ2(1) = 23.35, p < .001). Interest-
ingly, our results differ from what Kempen and
Harbusch (2004) found for subordinate clauses
in Negra. In adverbial and complement OVS
clauses, they found that full NP objects never
precede pronominalized subjects and translated
this findings into a rigid rule schema. They
argued that only strong conceptual influences
such as topic/focus relations could override the
ordering pattern. Such influences would then be
more likely to play a role in main clauses, as we
test them, than in subordinate clauses, since we
do observe full NP objects preceding pronom-
inalized subjects. However, this conclusion is
based on a relatively small set of sentences and
needs to be verified in a corpus larger than Ne-
gra.
(SVO)
second NP
pro O full O
first
NP
pro S 12 95
full S 33 450
(OVS)
second NP
pro S full S
first
NP
pro O 20 76
full O 91 366
Table 4: Frequency of subject (S) and object (O)
pairs ordered by pronominalization.
Pronominalization and givenness. As
expected, pronominalization is highly corre-
lated with givenness for all complements in
both sentence types (all p-values in Pearsons
tests < .001). Almost by definition, pronouns
are given, except for a few cases in which the
referent of a pronoun follows rather than pre-
cedes it within the same sentence. On the other
hand, clearly not all given complements are pro-
nouns. In fact, only 33% of all given subjects
and 26% of all given objects are pronouns.
Pronominalization, givenness, and
word order. Table 5 shows the number of
occurrences of both pronoun-full NP and full
NP-pronoun orders, split by the information
status of the complements. For both SVO and
OVS sentences, the positioning of the pronoun
is affected by its information status (for SVO:
χ2(1) = 16.92, p < .001; for OVS: χ2(1) = 4.76,
p < .03). Thus, the givenness of the pronoun
significantly increases the likelihood for this
complement to precede the other complement.
No such effect is found for full NPs.
(SVO)
pro given pro new full given full new
pro<full 92 3 28 67
full<pro 23 10 14 19
(OVS)
pro given pro new full given full new
pro<full 69 7 30 46
full<pro 70 21 37 54
Table 5: Linear order frequency of pronoun-full NP
pairs for given and new complements.
5 Discussion and Conclusion
We present in this paper a corpus-based study
on the linearization of subjects and objects in
German main clauses. We examined the ex-
tent to which the parameters of givenness, def-
initeness, and pronominalization influence the
ordering of verb complements in German SVO
and OVS sentences.
In general, our corpus data only support
the validity of the ordering principles for SVO
sentences: given subjects indeed tend to pre-
cede new objects, definite subjects indefinite
objects, and pronominalized subjects full NP
objects. However, clearly none of the order-
ing constraints is absolute since the reversed or-
ders were also observed, just not as often. For
OVS sentences, our results differed. None of the
three basic order patterns for givenness, defnite-
ness, and pronominalization was confirmed: For
givenness and pronominalization both orders
(given-new and new-given; pronoun-full NP and
full NP-pronoun) occurred about equally often.
For definiteness the basic order preference was
reversed (more indefinite-definite than definite-
indefinite orders). The fact that in OVS sen-
tences indefinite objects preceded definite sub-
jects more often than the converse, suggests
that grammatical functions rather than the lin-
ear structure of sentences influence the choice
of word order (i.e. subjects are more likely
to be definite, regardless of word order). Fur-
thermore, even though both definiteness and
pronominalization were correlated with given-
ness, only the positioning of pronouns before or
after the verb was additionally influenced by its
information status. Definiteness did not inter-
act with information structure as we would have
expected.
In sum, our data indicate that linearization
principles are soft constraints, and that a com-
bination of principles rather than one primary
constraint impact the choice of word order. The
fact, that for canonical SVO sentences cases of
reversed basic order patterns (e.g. new sub-
ject preceded given object) were observed for
all three tested linearization principles, suggests
that ordering constraints other than the ones
under investigation here influence the lineariza-
tion of complements. Indeed, in the literature
a range of such parameters has been suggested,
including animacy and length of complements
(Dietrich and Nice (in press); Hawkins (1994)).
A recent Negra corpus study by Kempen and
Harbusch (2004), confirmed the direct influence
of animacy on linearization in German subordi-
nate clauses. On the other hand, a Negra cor-
pus study by Kurz (2000) found no influence of
length on the ordering of subject and object.
6 Acknowledgements
We thank Stefan Baumann, Matthew Crocker,
and Amit Dubey for helpful comments on an
earlier version of this paper. This research
was supported by SFB 378 “ALPHA”, awarded
by the German Research Council and by the
Pioneer Project “Computing with Meaning”,
funded by the Netherlands Organization for Sci-
entific Research.

References
Betty Birner and Geoffrey Ward. 1998. Infor-
mation status and noncanonical word order
in English. John Benjamins, Amsterdam.
Kay Bock and Wilhelm Levelt. 1994. Lan-
guage production: Grammatical encoding. In
M. Gernsbacher, editor, Handbook of Psy-
cholinguistics. Academic Press, San Diego.
Thorsten Brants, Wojciech Skut, and Hans
Uszkoreit. 1999. Syntactic annotation of a
German newspaper corpus. In Proceedings of
the ATALA Treebank Workshop, pages 69–76,
Paris, France.
Wallace Chafe. 1976. Givenness, contrastive-
ness, definiteness, subjects, topics, and point
of view. In C. Li, editor, Subject and Topic,
pages 25–55. Academic Press, New York, NY.
Herbert Clark. 1977. Bridging. In P.N.
Johnson-Laird and P.C. Wason, editors,
Thinking: Readings in Cognitive Science,
pages 411–420. Cambridge University Press,
London, New York.
Rainer Dietrich and Kathy Nice. in press.
Belebtheit, Agentivit¨at und inkrementelle
Satzproduktion. In C. Habel and T. Pech-
mann, editors, Sprachproduktion. West-
deutscher Verlag, Wiesbaden.
Kari Fraurud. 1990. Definiteness and the pro-
cessing of NPs in natural discourse. Journal
of Semantics, 7:395–433.
John Hawkins. 1994. A Performance Theory of
Order and Constituency. CUP, Cambridge.
Jerry R. Hobbs. 1976. Pronoun Resolution.
Research report 76-1, Department of Com-
puter Sciences, City College, City University
of New York, August.
Elsie Kaiser and John Trueswell. submitted.
Role of discourse context in the processing of
a flexible word-order language.
Frank Keller. 2000. Gradience in grammar: Ex-
perimental and computational aspects of de-
grees of grammaticality. Ph.D. thesis, Univer-
sity of Edinburgh.
Gerard Kempen and Karin Harbusch. 2004.
How flexible is constituent order in the mid-
field of German subordinate clauses? A
corpus study revealing unexpected rigidity.
In Proceedings of the International Confer-
ence on Linguistic Evidence, pages 81–85,
T¨ubingen, Germany.
Geert-Jan Kruiff and Denys Duchier. 2003. In-
formation structure in Topological Depen-
dency Grammar. In Proceedings EACL 2003,
Budapest, Hungary.
Ivana Kruiff-Korbayov´a and Mark Steedman.
2003. Discourse and information structure.
Journal of Logic, Language and Information,
12:249–259.
Daniela Kurz. 2000. A statistical account on
word order variation in German. In Linguis-
tically Annotated Corpora LINC-2000, Work-
shop at COLING, Luxembourg.
Knud Lambrecht. 1994. Information structure
and sentence form. CUP, Cambridge.
Thomas S. Morton. 2000. Coreference for NLP
Applications. In Proceedings of ACL 2000.
Gereon M¨uller. 1999. Optimality, marked-
ness, and word order in German. Linguistics,
37:777–818.
Thomas Pechmann, Hans Uszkoreit, Johannes
Engelkamp, and Dieter Zerbst. 1996. Wort-
stellung im deutschen Mittelfeld. Linguistis-
che Theorie und psycholinguistische Evidenz.
In C. Habel, S. Kanngiesser, and G. Rickheit,
editors, Perspektiven der Kognitiven Linguis-
tik. Modelle und Methoden, pages 257–299.
Westdeutscher Verlag, Opladen.
Ellen Prince. 1981. Toward a taxonomy of
given-new information. In P. Cole, editor,
Radical Pragmatics, pages 223–255. Aca-
demic Press, New York, NY.
Ellen Prince. 1992. The ZPG letter: sub-
jects, definiteness, and information status. In
S. Thompson and W. Mann, editors, Dis-
course description: Diverse analyses of a
fund-raising text, pages 295–325. John Ben-
jamins.
Douglas Rhode, 2002. Tgrep2 User Manual.
Version 1.06.
Brigitte R¨oder, Tobias Schicke, Oliver Stock,
Gwen Heberer, Helen Neville, and Frank
R¨osler. 2000. Word order effects in Ger-
man sentences and German pseudo-word sen-
tences. Sprache und Kognition, 19(1/2):31–
37.
Elisabeth Selkirk. 1984. Phonology and Syntax.
MIT Press, Cambridge, MA.
Wojciech Skut, Thorsten Brants, Brigitte
Krenn, and Hans Uszkoreit. 1998. A linguis-
tically interpreted corpus of German newspa-
per text. In Proceedings of the ESSLLI Work-
shop on Recent Advances in Corpus Annota-
tion, pages 18–24, Saarbr¨ucken, Germany.
Mark Steedman. 1991. Structure and intona-
tion. Language, 67:262–296.
Hans Uszkoreit. 1987. Word Order and Con-
stituent Structure in German. Lecture Notes.
CSLI, Stanford University.
