VERB ORDER, OBJECT POSITION AND INFORMATION STATUS IN OLD ENGLISH
ANN TAYLOR, SUSAN PINTZUK
University of York
Abstract
This paper demonstrates that the distribution of objects differs in VAux and AuxV
clauses in Old English in three ways: post-verbal objects are less frequent in VAux
clauses than in AuxV clauses; discourse/performance factors have a greater effect in
VAux clauses than in AuxV clauses; and non-referential objects do not appear postverbally in VAux clauses but do in AuxV clauses. These results cannot be reconciled
with a single grammar in which object position is predominantly dependent on
discourse/performance factors, as has previously been proposed. Rather, they show that
this is the case only for VAux clauses. In AuxV clauses, object position is partly
dependent on discourse factors and partly fixed syntactically. The primary implication of
our results for syntactic analyses of OE is that there need to be two ways to derive postverbal objects, one which involves discourse/performance factors and one which does
not, and that these two derivations are associated with verb order.
1. Introduction
The correct analysis of the Old English (OE) object-verb/verb-object (OV/VO) alternation is
a long-standing and highly-disputed issue in the diachronic syntax literature (see Taylor and
van der Wurff 2005 for some views). The situation is complicated by the fact that in addition
to the OV/VO variation, there is variation in the order of finite and non-finite verbs. Thus in
clauses with a finite auxiliary, a non-finite main verb and an object, there are two seemingly
independent alternations: VAux vs. AuxV order and OV vs. VO order. The result is the four
orders illustrated in (1).1 Notice that the examples in (1) are from a single text, Ælfric’s Lives
of Saints, and this text is not exceptional: all four orders occur in almost all OE texts of any
length.2 The variation cannot therefore be attributed to variation within the community, with
some OE speakers categorically producing one constituent order, e.g. OV, while other
speakers categorically produce the alternate order, e.g. VO. Instead, the variation is at the
level of the individual, with all speakers producing all four orders, and therefore all variants
must be derivable within the grammar(s) of individual speakers.
(1)
a. O V Aux
gif heo þæt bysmor forberan wolde
if she that disgrace tolerate would
‘if she would tolerate that disgrace’
(coaelive,+ALS_[Eugenia]:185.305)
1
We omit two logical possibilities: 1) V-O-Aux clauses, because they do not exist (Biberauer et al 2008,
Pintzuk 2005); and 2) O-Aux-V clauses, because in these clauses the position of the object may be influenced by
factors different from those affecting the OV/VO alternation. We leave O-Aux-V clauses for future research. In
the examples, the finite auxiliary is underlined, the non-finite main verb is italicized and the object is bolded.
2
Only two texts in our dataset with more than 15 tokens of the types illustrated in (1) lack any of the variants.
The Holy Rood Tree and The West-Saxon Gospels with 32 and 20 relevant tokens, respectively, both lack
VAuxO order.
York Papers in Linguistics Series 2
© The Author, 2012
ISSN 1758-0315
Issue 12a
Ann Taylor and Susan Pintzuk
30
b. V Aux O
þæt he friðian
wolde þa leasan wudewan
that he make-peace-with would the false widow
‘that he would make peace with the false widow’
(coaelive,+ALS_[Eugenia]:209.315)
c. Aux O V
þurh þa heo sceal hyre scippend understandan
through which it must its creator understand
‘through which it must understand its creator’
(coaelive,+ALS_[Christmas]:157.125)
d. Aux V O
swa þæt heo bið forloren þam ecan life
so that it is lost
the eternal life
‘so that it is lost to the eternal life’
(coaelive,+ALS_[Christmas]:144.117)
Syntactically, the most parsimonious approach to analysing object position in OE is to
disregard the order of the finite auxiliary verb and the non-finite lexical verb and to derive all
post-verbal objects in the same way, as has been proposed in, for example, Roberts (1997),
van Kemenade (1987), van der Wurff (1997). This view is based on the reasonable
assumption that the two alternations are independent, and thus that verb order is not a factor
in determining verb-object order. In analyses of this type, the source of the variation in object
position is, rather, attributed to discourse/performance factors such as new information focus.
In this paper we show that this simple view is not supported by the evidence. When we widen
our view of the OV/VO data to include the interaction of discourse/performance factors with
the order of finite and non-finite verbs, we see a very different picture. There is clear
evidence that the distribution of pre- vs. post-verbal objects differs systematically in VAux
and AuxV clauses. Although this appears to support the non-independence of verb order and
verb-object order, we argue instead that the two alternations are indeed syntactically
independent, but that each alternation is associated with a change in progress: from head-final
to head-initial TP (VAux to AuxV) and head-final to head-initial VP (OV to VO),
respectively. Crucially, these two changes are linked in such a way as to create the
appearance in the synchronic data of non-independence. We show that by adopting a model
of syntactic change in which speakers have access to more than one grammar during the
course of change, a conservative "outgoing" grammar and an innovative "incoming" grammar
(Kroch 2003), we can account for the empirical facts, while maintaining syntactic
independence of the headedness parameter settings in TP and VP.
2. Data and terminology
The data for this study are taken from the York-Penn-Helsinki Parsed Corpus of Old English
Prose (Taylor et al 2003, henceforth the YCOE). The dataset includes clauses with finite
auxiliaries, non-finite main verbs and nominal objects from all texts (duplicate manuscripts
excluded) included in the YCOE with the exception of Ælfric's Supplemental Homilies,
Wulfstan's Homilies, the Vercelli Homilies and the Institutes of Polity. There are 85 texts in
31
Verb Order, Object Position and Information Status in Old English
total in our dataset, although some texts are represented by very small numbers of tokens.
The sample of VAux clauses is exhaustive; for AuxV clauses, approximately one-third of the
total cases from the included texts (every third token) were collected. The main dataset
contains 1525 tokens in total, 692 VAux clauses and 833 AuxV clauses. The data are further
restricted by the exclusion of the following types:
Clauses with movement of the finite auxiliary to C, since these clauses are ambiguous
between underlying AuxV and underlying VAux structure and order. As discussed in
much previous work, e.g. Pintzuk 1999, most so-called V2 clauses in Old English involve
verb movement to T; it is only in exceptional clause types (e.g. direct questions, clauses
beginning with þa/þonne ‘then’, verb-initial clauses) that the finite verb moves to C, as in
(2).
AuxV clauses with the object before the Aux (OAuxV order), as in (3) below. As noted in
footnote 1, the position of the object in these clauses may be influenced by factors
different from those affecting the OV/VO alternation.
VAux main clauses with non-overt subjects, or with the object before the subject, as in
(4), to avoid potential cases of topicalisation.
Pronominal (personal and demonstrative) objects, since pronominal objects usually
appear pre-verbally, particularly in VAux clauses (Pintzuk 2005). Pronominal objects are
optionally clitics or weak pronouns in OE (Pintzuk 1996, Wallenberg 2009), and
frequently move to a high position in the clause structure. Since their syntax is different
from non-pronominal objects, they are not considered here.3
Quantified objects, since it has been shown (Pintzuk and Taylor 2006) that quantified
objects exhibit special syntactic behaviour compared to non-quantified objects.
A few additional cases where the information status of the object is unclear.
Non-referential objects (negative and semantically incorporated) are excluded from the
main dataset and discussed separately in section 3.2.
•
•
•
•
•
•
•
(2)
(3)
Verb movement to C
a.
hu mæg ic þæt ece
lif gebycgan?
how may I the eternal life purchase
‘how may I purchase the eternal life?’
(coaelive,+ALS[Ash_Wed]:124.2766)
b.
Þa wolde an ðæra preosta wytan þa synne
Then desired one of-the priests know the sin
‘Then one of the priests wanted to know the sin’
(coaelive,+ALS_[Basil]:657.921)
Aux V
þæt hi mine þeawas magon him secgan
that they my customs may him tell
‘that they might tell him my customs’
(coaelive,+ALS[Agnes]:313.1932)
(4)
S Aux V
and
þam deadan þu ne miht eft lif forgifan
and (to) the dead you NEG can again life give
‘and you cannot give life again to the dead’
3
Note that pronouns are used in section 4.1 as a diagnostic for head-initial structure.
Ann Taylor and Susan Pintzuk
32
(coaelive,+ALS_[Cecilia]:327.7304)
The data are therefore restricted to the clause types shown in (1) above and (5) through
(8) below. VAux clauses are those with the non-finite main verb followed by the finite
auxiliary. In these clauses, the two verbal elements are always adjacent, as shown in (7) and
(8), since the word order V-XP-Aux does not occur in OE regardless of the type of XP; see
footnote 1. AuxV clauses are those with the auxiliary followed by the main verb; in these
clauses, the verbal elements are not necessarily adjacent, since constituents of all types (not
only complements but also adjuncts) can appear between the two verbs, as shown in (5)
below. VO clauses are those with the object after both verbal elements and not necessarily
adjacent to the second one, as shown in (6) for AuxV clauses and (7) for VAux clauses. OV
clauses are those with the object before the non-finite main verb. For VAux clauses, this is …
O (…) V Aux order, as in (8). For AuxV clauses, this is … Aux (…) O (…) V, as shown in
(5).
(5)
(6)
(7)
Aux XP O V
a.
and Crist wolde on his tocyme clænnysse aræran
and Christ desired at his coming chastity
establish
‘And Christ wanted to establish chastity at his coming’
(coaelive,+ALS[Peter's_Chair]:213.2414)
b.
Aux O XP V
Martianus hæfde his sunu ær
befæst
to woruldlicre lare
Martin had his son before committed to worldly
instruction
and to uðwitegunge
and to philosophy
‘Martin had committed his son before this to worldly instruction and philosophy’
(coaelive,+ALS_[Julian_and_Basilissa]:184.1049)
Aux ( … ) V O XP
a.
ac þæt halige godspell hæfð oferswiðod swylcera gedwolena andgit
but the holy gospel has surpassed such
heretics
understanding
foroft
very-often
‘but the holy gospel has very often surpassed the understanding of such heretics’
(coaelive,+ALS_[Christmas]:7.7)
b.
Aux (…) V XP O
He sceal habban eac mæssereaf
He must have also mass-vestments
‘He must also have vestments for the mass ...’
(colwsigeXa,+ALet_1_[Wulfsige_Xa]:55.68)
V Aux O XP
a.
þeah ðe heo secgan cunne
sum ðincg þurh
deofol
although she tell
may-be-able-to some thing through devil
‘although she may be able to tell something through the devil’
(coaelive,+ALS_[Auguries]:124.3586)
b.
V Aux XP O
þæt hi awræccan ne magon mid heora wodlican plegan ænige galnysse
that they arouse
NEG may with their foolish sport any
lust
on me
in me
‘that they may not arouse any lust in me with their foolish sport’
(coaelive,+ALS_[Chrysanthus]:62.7371)
33
(8)
Verb Order, Object Position and Information Status in Old English
XP O V Aux
a.
hu heo æfre wæras wissian sceolde
how she ever men direct should
‘how she ever should direct men’
(coaelive,+ALS_[Eugenia]:121.261)
b.
O XP V Aux
forþan þe hi þa anlicnyssa æfre tobrecan dorston
because they the images
ever break
dared
‘because they ever dared to break the images’
(coaelive,+ALS_[Sebastian]:237.1354)
3. The distribution of objects in VAux and AuxV clauses
In this section we present data on the distribution of objects in OE VAux and AuxV clauses.
After a look at the overall picture, we turn in Section 3.1 to three particular factors which
have been associated in the literature with surface word order variation: information status,
case and grammatical weight. Under the null hypothesis (that there is no association between
verb order and object position), we would expect to find these constraints applying equally in
both types of clauses. In section 3.2, we then consider two types of objects omitted from the
main analysis due to their non-referential status, negative objects and incorporated objects,
and show that their distribution also differs significantly in VAux vs. AuxV clauses.
Table 1 presents the overall frequency of VO order in VAux and AuxV clauses; the
texts are listed by the frequency of AuxV (vs. VAux) order, which is a rough indication of
chronology; see Taylor and Pintzuk 2012a for discussion. Table 1 shows that overall VO
order is more frequent in AuxV than in VAux clauses, not only in aggregate, but also for each
text. If all post-verbal objects are derived by the same discourse/performance constraints, this
is an unexpected result. While some variation is inevitable given the nature of the data, if
AuxV and VAux clauses do not differ with respect to the derivation of object position, we
would expect the frequency of VO to be closer in value overall; in addition, text by text, we
would expect to see variation in frequency in both directions, rather than always in the
direction of VO.
34
Ann Taylor and Susan Pintzuk
Text
VAux
AuxV
N
%VO
N
%VO
Orosius
66
4.5
47
31.9
Bede
58
6.9
46
10.9
Boethius
74
8.1
49
53.1
Cura Pastoralis
51
21.6
72
55.6
Catholic Homilies I
49
10.2
95
47.4
Catholic Homilies II
42
7.1
80
46.3
Lives of Saints
33
45.5
91
62.6
Gregory’s Dialogues (C)
36
27.8
66
68.2
total
409
13.9
546
49.5
sample total
692
12.6
833
47.1
Table 1: Frequency of VO order by verb order in texts with more than 100 tokens4,5
1.1. Factors affecting object position (Note: should be 3.1)
1.1.1. Information status (Note: should be 3.1.1)
There is a vast literature addressing the assignment of information status to entities within
utterances (cf. Bech 2001, Bies 1996, Hinterhölzl 2009, van Kemenade and Los 2006,
Petrova 2009 and Sapp 2006, among others, for recent work on discourse factors in the older
West Germanic languages). Although details differ, there is general agreement that
information status correlates with position in the clause: given elements favour an early
position and newer elements a later one (Bech 2001, Pintzuk and Taylor 2006, among many
others). We have followed the common line in studies of this sort (e.g. Arnold et al. 2000,
Bech 2001, Gries 2003) and used a binary distinction of given vs. new. We divide the data
into given and new entities primarily on the basis of insights drawn from the work of Birner
(2006) building on Prince (1981) and Gundel et al. (1993), as follows. Examples are given in
Error! Reference source not found. and Error! Reference source not found., with the
object to be assigned information status in boldface.
Our new category includes:
4
The difference between VAux and AuxV clauses is significant at the 0.001 level for both totals and for all
texts, except for Bede and Lives of Saints, where the difference is not significant.
5
Recall that VO order in VAux clauses means … V Aux (…) O, while VO order in AuxV clauses means
… Aux V (…) O.
35
•
•
•
(9)
Verb Order, Object Position and Information Status in Old English
referentially new objects (in the sense of Gundel et al. 1993)
new discourse referents (in the sense of Karttunen 1976): a short-term discourse referent
that fails to establish a permanent referent; it is or could be referred back to with a
pronoun within a limited domain but not beyond it. E.g. If a man buys a donkey, he will
beat it. #It [the donkey] is old.
“bridging” inferable (Birner 2006):6 the referent is anchored to an already mentioned
entity but the referent is not accessible without the anchor. Bridging inferables are
frequently cases of alienable possession.
Object types coded as new
a.
referentially new
Be ðam sagað Sanctus Arculfus
about that said Saint Arculfus
þæt he gesawe medmicle cirican butan Bethlem þære ceastre
that he saw
little
church outside Bethlehem the city
‘St. Arculfus says about this that he saw a little church outside the city of
Bethlehem’
(Mart_5_[Kotzor].Se30_A.3.1906)
b.
new discourse referent
Ðeah þe hwa
wille her on life habban gode dagas, he ne mæg hi her
Yet
whoever will here in life have good days, he NEG can them here
findan
find
‘Yet whoever will have good days here in life, he cannot find them here.’
(coaelive,+ALS[Ash_Wed]:82.2748)
c.
bridging inferable
ðæt is, ðæt ic sette minne renbogan on wolcnum
that is, that I set my
rainbow in clouds
‘that is, that I set my rainbow in the clouds’
(cootest,Gen.9.13.388)
Our given category includes:
•
•
•
•
previously mentioned entities
shared/cultural knowledge
situationally evoked entities: the referent forms part of the situation of the discourse, and
is therefore ‘known’ to both the speaker and the hearer.
“elaborating” inferables (Birner 2006): the referent is anchored to an already mentioned
referent and the referent is accessible without the anchor. Elaborating inferables are
frequently body parts or other cases of inalienable possession.
(10) Object types coded as given
a.
previously mentioned
& of ðæs treowes wæstme þe
is on middan neorxnawange, God bebead us,
and of the tree’s
fruit
which is in middle paradise
God bid us
ðæt we ne æton, ne we ðæt treow ne hrepodon ði læs ðe we swelton.
that we NEG eat nor we the tree NEG touched lest
we die
6
Birner (2006) draws a distinction between bridging and elaborating inferences (see discussion below) and
shows that in Modern English the two types of inferences are subject to different constraints with respect to
preposing and postposing. An anonymous reviewer informs us that Eastern Armenian distinguishes the two
types of inferences morphosyntactically (Hodgson 2012).
36
Ann Taylor and Susan Pintzuk
b.
c.
d.
‘and of the fruit of the tree which is in the middle of Paradise, God bid us that we
may not eat, nor may we touch the tree lest we die.’
(cootest,Gen:3.3.123)
shared/cultural knowledge
Ond þeah þe wærgcweodole Godes rice
gesittan ne mægen
And although those-that-curse God's kingdom occupy NEG may
‘And although those that curse may not occupy God's kingdom ...’
(cobede,Bede_4:27.356.26.3595)
situationally evoked
[Context: Katherine has been matched against fifty debaters.]
ÞV qð ha keiser nauest nawt þis strif rihtwisliche idealet
You said she Emperor NEG-have not this contest fairly
matched
‘Emperor, she said, you have not matched this contest fairly …’
(CMKATHE,30.187)
elaborating inferable
Healdað mine bebodu & mine domas
Keep my commands and my judgements
ðæt ge libbon
eower lif butan ælcere sorhge.
so-that you may-live your life without any sorrow
‘Keep my commands and my judgements so you may live your life without any
sorrow’ (Lev:25.18.3864)
Our results, given in Table 2, show the expected effect in both VAux and AuxV clauses, i.e.,
new objects favour post-verbal position, but the difference between given and new is much
greater, and only significant, for VAux clauses.7 In Tables 2 through 5, shaded cells indicate
results that are not statistically significant.
VAux
AuxV
Info Status
N
%VO
N
%VO
given
507
9.7
507
45.0
new
185
20.5
326
50.3
VAux: χ2 = 14.6, p-value = 0.0005
AuxV: χ2 = 2.27, p-value = 0.132
Table 2: Frequency of VO order in VAux vs. AuxV clauses by information status
7
An anonymous reviewer points out that if OE is a typical West Germanic language, not all pre-verbal objects
in VAux clauses are expected to be given: while scrambled objects (like the one in (6c)) will be given, in situ
objects that remain adjacent to the non-finite verb will be new. However, OE is not a typical West Germanic
language: complements and adjuncts can appear post-verbally. There are thus (at least) three positions to
investigate: post-verbal, pre-verbal adjacent to the verb and pre-verbal scrambled out of the VP. We have
combined the latter two positions, leaving a more detailed investigation to future research. Note that Biberauer
and Roberts 2008 also assume that post-verbal position in OE is for focused objects, while pre-verbal position
(whether or not adjacent to the verb) is for defocused objects.
37
1.1.2.
Verb Order, Object Position and Information Status in Old English
Case (Note: should be 3.1.2)
Dunbar (1979: 175) gives the hierarchy in (11) for different constituents according to their
likelihood of appearing post-verbally in OHG.
(11) lower post-verbal --------------------------------------------------------- higher post-verbal
[subject < pred. nominal] < direct object < genitive/indirect object [< prep. phrase]
He notes that this hierarchy is the opposite of Kuno’s accessibility hierarchy (Kuno 1976:
427) in which elements higher on the hierarchy are less likely to be topical/themes than those
farther down. Subjects are the most likely constituents to be topics, and therefore the least
likely to appear post-verbally. Complements of prepositions are the least likely to be topics,
and therefore the most likely to appear post-verbally. Objects in general are less topical than
subjects, and indirect objects are less topical than direct objects; this means that direct objects
are less likely than indirect objects to appear post-verbally. Topicality is of course related to
information status: topics represent given information, while non-topics may represent new
information. On this basis, Dunbar (1979) attributes the case effect to information status:
accusative is the most common case for direct objects, and therefore we expect accusatives to
occur in post-verbal position less frequently than datives, the most common case for indirect
objects.8
As shown in Table 3, our data do indeed reflect the difference between accusatives and
datives as expected, although the difference is not significant. The most extreme effect,
however, is with genitives, which appear post-verbally more frequently than either
accusatives or datives. The reason for this is not clear. Neither Dunbar nor Burridge discuss
genitives separately from datives, treating them both as “indirect objects”, but in our data the
difference between datives and genitives is larger than the difference between accusatives and
datives. Whatever the source of this difference, the effect is greater, and only significant, in
VAux clauses, just as with information status.
VAux
Case
AuxV
N
%VO
accusative
517
dative
genitive
N
%VO
9.9
644
46.4
133
14.3
140
47.9
42
40.5
49
53.1
VAux: χ2 = 33.6, p < 0.0005 [acc vs. dat χ2 = 2.15, p = 0.142]
AuxV: χ2 = 0.847 , p = 0.655
Table 3: Frequency of VO order in VAux vs. AuxV clauses by case
8
Burridge (1993: 106) finds almost the same hierarchy for Middle Dutch. Although her order of DO/IO is
reversed from that of Dunbar, she attributes the reversal to her inclusion of pronouns in the counts: almost all of
the Middle Dutch indirect objects are pronouns, which are highly topical and therefore show a lower tendency
than direct objects (presumably a mixture of pronouns and full DPs) to appear in post-verbal position.
38
Ann Taylor and Susan Pintzuk
1.1.3.
Grammatical weight (Note: should be 3.1.3)
The grammatical weight (length/heaviness) of constituents has frequently been shown to
correlate with clause position cross-linguistically, with longer/heavier elements favouring
later position (Arnold et al 2000, Gries 2003, Pintzuk and Taylor 2006, Siewierska 1993,
Taylor and Pintzuk 2012b, among many others). Measures of grammatical weight may be
categorical (e.g. a constituent is heavy if it contains more than one phonological phrase) or
continuous (e.g. the number of words); among the continuous measures that have been used
are number of words, nodes, maximal nodes, syllables or graphemes. Exactly how
grammatical weight is measured seems to make little difference to the result, and no single
method or unit has been unanimously adopted by researchers. Wasow (1997:85) lists eight
different proposals found in the literature; he concludes (p. 120) that continuous measures are
better than categorical ones. Szmrecsányi (2004) compares three measures – word counts,
node counts and an index of syntactic complexity – and finds that all three are highly
correlated. Here we measure weight in the number of open class words.9 Table 4 shows that
for each additional open class word, an object is more likely to appear post-verbally in both
AuxV and VAux clauses. As with the other factors considered, this factor thus shows an
effect in both clause types, but here the effect is significant for both.
VAux
AuxV
Words
N
%VO
N
%VO
1
453
5.3
442
34.8
2
197
25.9
313
56.4
3
28
35.7
53
74.1
4+
14
14.310
25
88.5
VAux: p-value = 3.499e-14 [Fisher's exact ] AuxV: p-value = 1.614e-14 [Fisher's exact]
Table 4: Frequency of VO order in VAux vs. AuxV clauses by length of object
(in open class words)
1.1.4.
Strength of effect (Note: should be 3.1.4)
In the previous section we saw that, on the basis of frequencies alone, information status, case
and grammatical weight all show an effect in the expected direction in both AuxV and VAux
clauses. The former two effects, however, are significant only for VAux clauses, which
implies that their effect should be less strong in AuxV clauses. We can test this prediction
9
Demonstratives are often used to signal definiteness in OE, but the language does not have an indefinite
determiner that is used with any consistency. If our unit of measurement included all words, both open and
closed class, definite DPs would frequently be heavier than indefinite DPs simply due to the inclusion of a
determiner. For this reason we use open-class words only.
10
The low frequency of VO order for objects of 4+ words in VAux clauses is unexpected. We note the
relatively low N for this category; nevertheless we have no explanation for the low frequency.
39
Verb Order, Object Position and Information Status in Old English
using multivariate analysis to assess the strength of each effect while simultaneously taking
all other effects into account.11 The results are given in Table 5. Shaded cells indicate nonsignificant results.
Variable
Level
VAux
Log odds
grammatical
weight
per additional open-class
word
info status
case
AuxV
Odds ratio
Log odds
Odds ratio
0.99
2.68
0.89
2.43
given
-0.45
0.64
-0.04
0.96
new
0.45
1.57
0.04
1.04
accusative
-1.06
0.35
-0.22
0.81
dative
-0.38
0.68
-0.17
0.85
1.45
4.25
0.38
1.47
genitive
VAux: C = 0.88; AuxV: C = 0.78
Table 5: Results of multivariate analysis, effects in logodds and odds ratios
The results from the multivariate analysis confirm the picture given by the frequencies, but
the log odds and odds ratios also indicate that the effect of grammatical weight is (slightly)
weaker in AuxV clauses, a fact not evident from the frequencies alone. We can see this by
comparing the log odds for grammatical weight for the two clause types: it is significant for
both, but nevertheless slightly lower in AuxV than in VAux clauses (0.89 vs. 0.99). In terms
of odds ratios,12 each additional open class word added to an object in VAux clauses
increases the odds of VO order by somewhat over two and a half times (2.68), while in AuxV
clauses it increases the odds by slightly under two and a half times (2.43). The other two
factors, information status and case, are significant only in VAux clauses, as we saw
previously with the frequencies. As with the frequencies, we can see that the effect is in the
same (expected) direction in both clause types, but again the log odds show that the effects
are stronger in VAux than in AuxV clauses. Thus for VAux clauses, the distance between
information status given and new on the log odds scale is 0.90 (0.45+0.45), while in AuxV
clauses it is only 0.08 (0.04+0.04). In terms of odds ratios, in comparison to the mean, being
new increases the odds of an object appearing in post-verbal position in VAux clauses by
57% (1.57), but only by 4% (1.04) in AuxV clauses. The same is true of the case effect. The
distance between accusative and genitive on the log odds scale in VAux clauses is 2.51
(1.06+1.45), while in AuxV clauses it is 0.60 (0.22+0.38). In comparison to the mean, being
genitive increases the odds of an object appearing post-verbally by 4.25 times in VAux
11
The analysis was carried out using Rbrul (http://www.ling.upenn.edu/~johnson4/Rbrul_manual.html),
including clause type (main, that-complement, adverbial, relative) as a factor and with text as a random effect,
using sum contrasts (the Rbrul default). The coefficients indicate the increased likelihood of a post-verbal object
given the factor level (e.g. given, new, etc.) against the average case, measured on a logarithmic scale. The
numbers, therefore, have no straightforward real-world interpretation, but the scale is linear, so the absolute size
of the effects is comparable.
12
The odds ratio is the probability of success over the probability of failure (p / (1 – p)). Transforming the log
odds into odds ratios makes them easier to interpret. The odds ratio can be calculated by raising the constant e to
the log odds power, e.g. for grammatical weight in VAux clauses, e to the power of .985 = 2.678.
40
Ann Taylor and Susan Pintzuk
clauses, but by only 1.5 times (1.47) in AuxV clauses; in contrast, being accusative decreases
the odds by about 65% (0.35) in VAux clauses, but only by about 20% in AuxV clauses
(0.81).13 Note finally that the fit for this model is better for VAux clauses than for AuxV
clauses: C, an index of concordance between the predicted probability and the observed
response,14 is 0.88 for the VAux model and 0.78 for the AuxV model.
1.2. Non-referential objects (Note: should be 3.2)
1.1.5.
Negative objects (Note: should be 3.2.1)
Negative objects, as in (12), were excluded from the analysis above due to their nonreferential status. Their distribution, given in Table 6, shows a sharp contrast between VAux
and AuxV clauses: negative objects never appear post-verbally in VAux clauses (cf. also
Pintzuk 2005). In contrast, non-negative objects, as we saw in Table 1, occur post-verbally at
approximately 13% and 47% in VAux and AuxV clauses, respectively.
(12) a.
b.
heo nan ðing elles don ne mihte
she no thing else do NEG might
‘she might do nothing else’
(cocathom2,+ACHom_II,_42:316.180.7143)
þæt nan cristen mon ne moste habban nænne his sunderfolgeþa
that no Christian man NEG might have none (of) his official-teachings
‘that no Christian man might have any of his official teachings’
(coorosiu,Or_6:31.150.14.3217)
VAux
AuxV
Object type
N
%VO
N
%VO
negative
70
0.0
107
14.0
692
12.6
833
47.1
non-negative
Table 6: Frequency of post-verbal negative objects and non-negative objects
in VAux and AuxV clauses
1.1.6.
Semantically incorporated objects (Note: should be 3.2.2)
Semantically incorporated objects, as in (13), were also excluded from the analysis as they
are known not to respond to information status factors in the same way as referential objects;
13
It is unclear whether the relative weakness of the information status effect with respect to case, and
particularly heaviness, is a real effect or simply reflects the difficulty of coding this factor in comparison to
heaviness and case, which are much more straightforward.
14
C ranges from 0.5 (random) to 1 (perfect fit). Values of 0.8 or above indicate the model has real predictive
value.
41
Verb Order, Object Position and Information Status in Old English
in particular, they do not exhibit continuity of identity with previously or subsequently
mentioned noun phrases (du Bois 1980). 15
(13) a.
b.
Ac gif ic deað þrowian sceal, leofre me is, þæt ...
but if I death suffer shall, dearer me is that ...
‘But if I suffer death, it is dearer to me that ...’
(cobede,Bede_2.9.128.6.1213)
and wiscton þæt hi moston wite
þrowian for Criste
and wished that they might punishment suffer for Christ
‘and wished that they might suffer punishment for Christ’
(coaelive,+ALS_[Chrysanthus].216.7456)
We might expect a difference for these objects similar to that for negatives, but in this
case the data are less clear. As shown in Table 7, in VAux clauses, 2 out of the 32 cases of
incorporated objects (6.3%) are post-verbal where we would expect zero. However, both
these cases occur in Ælfric's Lives of Saints. For reasons which are not entirely clear, but are
possibly related to its metrical properties, this text often behaves differently from others
(including Ælfric's own non-metrical Catholic Homilies) with respect to object positioning
(cf. Taylor 2006). If we omit this text, then we find the same distribution for incorporated as
for negative objects: no post-verbal objects in VAux clauses as opposed to 15% in AuxV
clauses. This is very close to the difference that we see with negative objects in Table 6 (0%
vs. 17.5%). The difference in the frequency of incorporated vs. non-incorporated objects in
AuxV clauses is also instructive. There is no weight distinction here since the data
(incorporated and non-incorporated) have been restricted to cases of one open-class word,
and yet non-incorporated objects appear post-verbally about twice as often as incorporated
ones.
15
As far as we know, there is no completely objective way to determine whether an object is incorporated in
OE. The clearest cases are those in which the object is a single bare noun, it is not later referred to, and the
verb+object can be paraphrased with a verbal idea alone: e.g. suffer death = die. Another indication of
incorporation in this sense is the continued use of the same bare noun (or NP) in subsequent mentions.
42
Ann Taylor and Susan Pintzuk
VAux
AuxV
Object type
Texts
N
%VO
incorporated
All texts
32
LoS
LoS omitted
non-incorporated All texts
LoS
LoS omitted
N
%VO
6.3
45
15.6
5
40.0
5
20.0
27
0.0
40
15.0
463
5.6
451
34.8
14
35.7
55
47.3
449
4.7
396
33.1
Table 7: Frequency of post-verbal position for incorporated and non-incorporated objects
(one open-class word only) in VAux and AuxV clauses
1.3. Summary (Note: should be 3.3)
In summary, we have seen that VAux and AuxV clauses differ on three dimensions:
1. Post-verbal objects are less frequent overall in VAux clauses than in AuxV clauses.
2. Factors associated with end position – weight, information status and case – have a
greater effect, to varying degrees, in VAux clauses than in AuxV clauses.
3. Non-referential objects (negative and semantically incorporated) do not freely appear
post-verbally in VAux clauses but do so in AuxV clauses.
4. Analysis
On the face of it, it is hard to reconcile the differences between VAux and AuxV clauses with
respect to object position discussed in section 3 with a syntactic analysis in which all postverbal objects are triggered by the same discourse/performance constraints. In this section we
suggest a different analysis for the distribution of objects in VAux and AuxV. The analysis
we present retains the independence of the two syntactic alternations (head-final/initial TP
and head-final/initial VP), but nevertheless is able to account for the differences in the
distribution of objects between the clause types.
1.4. Change in progress (Note: should be 4.1)
The first point to note is that the four word orders under consideration here (illustrated in (1)
above) are associated not only with different discourse/performance constraints, but also with
two independent but linked changes that occur starting in the (pre-)OE period. The first
change is the replacement of VAux order by AuxV order, i.e. the TP changes from head-final
to head-initial. This change is a prerequisite for the second one, the replacement of OV by
VO order, i.e. the VP changes from head-final to head-initial. As stated in footnote 1, a headfinal IP cannot select a head-initial VP, i.e. the structure [IP [VP V O ] Aux ] is never generated
or derived. From this it follows that the change from VAux to AuxV is a necessary
precondition for the change from OV to VO. By precondition we do not mean that the first
43
Verb Order, Object Position and Information Status in Old English
change has to be completed before the second change begins. On the contrary, the time spans
of the two changes overlap almost completely: as noted above, all four orders can be found in
any OE text of any length. Nevertheless, there is some evidence that the variation in
headedness in the TP starts earlier than the variation in headedness in the VP. The earliest
extant OE text, the narrative poem Beowulf, was uniformly OV with variation in the
headedness of TP (Pintzuk 1999, Pintzuk and Kroch 1989), and the prose Laws of Æthelbert
(composition dated to approximately 600 CE) are consistently surface V-final, even in main
clauses, according to Oliver (2002: 31).16 Based on the Germanic ancestors of OE and on the
diachronic trends, it is reasonable to assume that the earliest stage of OE was uniformly headfinal in both the VP and the TP, with variation in surface object position (OVAux vs.
VAuxO) triggered by discourse constraints (cf. Pintzuk and Kroch 1989). Moreover, VAux
order disappears before OV order. While both survive into Middle English, the frequencies of
VAux order are already very low in Early Middle English (Kroch and Taylor 2000 find
approximately 10% surface INFL-final clauses in the texts included in the PPCME2 period
M1 (1150-1250)), while OV order continues with some frequency throughout Middle English
until finally disappearing from prose in the course of the 16th century (van der Wurff 1997,
1999; Pintzuk & Taylor 2006). A very rough timeline is given in Error! Reference source
not found.: head-final TPs and VPs are used in periods (a) through (d), head-initial TPs are
first used in period (b), while head-initial VPs are first used in period (c).
(14)
(a)
(b)
(c)
(d)
(e)
pre-OE
Beowulf
first AuxV-pro17
early ME
late ME
orders
c. 1250
16th cent.
(f)
PDE18
head-final TP-------------------------------------------------------head-initial TP--------------------------------------------------------------------head-final VP-------------------------------------------------------head-initial VP------------------------------------------------This scenario produces a somewhat complicated situation, illustrated in Table 8, in which are
shown three distinct grammatical systems with their base order and derived VO order.19
16
This text is fairly short and in a very particular genre, so it is not clear how seriously to take its consistency.
It does differ from later texts in the same genre on the position of the verb, however, indicating that it is not just
a genre-specific feature.
17
Since pronouns do not postpose in OE (Pintzuk 1999), we can “date” the appearance of the head-initial VP
from the first appearance of post-verbal pronouns. Six of the 17 texts dated as pre-950 in our corpus which
contain examples of AuxV clauses with pronominal objects have some AuxV-pro order: Blickling Homilies
(2/66), Boethius' Consolation of Philosophy (2/137), Cura Pastoralis (1/150), Gregory's Dialogues C (3/145),
Laws of Ine (1/10), and Augustine's Soliloquies (3/59). In the remaining texts – Alexander's Letter to Aristotle,
Bede's History of the English Church, Life of Saint Chad, early parts of Chronicle A, Documents 2, Leechdoms,
Laws of Alfred, Martyrology II/III, Marvels of the East and Orosius – all pronominal objects appear before the
non-finite verb.
18
The third stage, illustrated as extending to PDE, is actually not quite the PDE grammar, because it allows
preposing of negative and quantified objects from post-verbal to pre-verbal position (Pintzuk and Taylor 2006),
as in Modern Icelandic (Rögnvaldsson 1987, Svenonius 2000). This option was lost in late ME.
19
Although G1/G2/G3 are labelled “grammatical systems”, it is not the case that there are major differences
between them. If we think of parameters (like the headedness parameter) as features on functional heads, then
the differences between the grammars are simply differences in the features on T and V: T in G1 selects a VP
44
Ann Taylor and Susan Pintzuk
VP
base
order
derived
order
characteristics of
post-verbal
objects (idealized)
System
Periods used
TP
G1
(a) through (d)
head-final
G2
(b) through (d) head-initial head-final AuxOV AuxVO new and/or heavy
G3
(c) through (f) head-initial head-initial AuxVO AuxVO
head-final OVAux VAuxO new and/or heavy
no discourserelated effects
Table 8: Three grammatical systems
During the period in which both G1 and G2 are used (from the time of Beowulf through early
ME), the use of G2 increases in frequency at the expense of G1; in other words, the TP
changes from head-final to head-initial. During the period in which both G2 and G3 are used
(from the OE period through the late ME period), the use of G3 increases in frequency at the
expense of G2; in other words, the VP changes from head-final to head-initial. The output of
G1 can be distinguished from the output of G2 and G3 by the surface order of the verbs, as all
VAux clauses belong to G1. However, G2 and G3 cannot be reliably distinguished by surface
order. AuxOV order must belong to G2, but AuxVO can belong to either G2 or G3,
depending on how it is derived. If generated by G2, it should be associated with particular
discourse constraints; but if generated by G3, it should show no discourse-related effects. In
theory, therefore, it ought to be possible to distinguish G2 AuxVO from G3 AuxVO, but of
course in practice discourse constraints are neither categorical nor detectable with enough
certainty to apply to individual cases. What we can reliably distinguish is cases of G1, in
which all VO order is hypothesized to be derived from an OV base, and a mixed bag of G2
and G3 cases, some with VO order derived from an OV base and some from a VO base.
In an idealized world, all post-verbal objects in G1 are focussed (new) and/or heavy; the same
is true of post-verbal objects in G2. G3 is like PDE, however, where all objects are postverbal regardless of information status or weight; i.e., they do not exhibit discourse-related
effects. The expected outcome of this scenario in which AuxVO clauses belong to either G2
or G3 is that, when treated as a group, some of these post-verbal objects will exhibit
discourse-related effects and some won't, and thus there will appear to be a weakening of the
discourse-related effects in comparison to the uniform VAuxO class.
A straightforward way to model the variation described in Table 8 is by invoking a
headedness parameter for TP and VP. Simplified versions of the relevant derivations, with
the minimal amount of structure and movement shown, are given in (15).20 T and V take VP
complements. Finite verbs (V1) move obligatorily to T, regardless of the position of T, while
non-finite verbs (V2) remain in situ. DP objects may remain in their merged position, or they
may postpose to post-verbal position (i.e. adjoin to the right periphery of TP), presumably
because of focus or heaviness.
complement to its left, while T in G2/3 selects a VP complement to its right. Similarly, V in G1/2 selects a DP
complement to its left, while V in G3 selects a DP complement to its right. Thus the competition between G1
and G2 in periods (b) through (d) is simply a competition between two functional heads T that differ in a single
feature (directionality), while the competition between G2 and G3 in periods (c) and (d) is a competition
between two functional heads V. See Kroch 1994 for further discussion.
20
These structures are not intended to reflect assumptions about the nature of auxiliary verbs in OE, i.e. whether
they appear in monoclausal or biclausal structures.
45
Verb Order, Object Position and Information Status in Old English
(15) a.
OVAux: head-final TP and VP
c.
b.
AuxOV: head-initial TP, head-final VP d.
VAuxO: head-final TP and VP
with postposition of O
AuxVO: head-initial TP, headfinal VP, with postposition of O
The difference between a and b and between c and d is the assignment of focus: focussed
objects move out of the VP and right-adjoin to TP. However, AuxVO clauses also have a
second derivation: all projections can simply be head-initial, as shown in (16):
Ann Taylor and Susan Pintzuk
46
(16) AuxVO, derivation 2: head-initial TP and VP
In summary, an account with two post-verbal object positions in AuxV but not VAux
clauses, such as Pintzuk and Taylor (2006), predicts differences not only between OV and
VO clauses but also between VAuxO and AuxVO clauses, while accounts with a single postverbal object position can capture only the first difference. In the discussion that follows, we
will use the head-initial vs. head-final terminology for ease of explanation; but it should be
clear that an anti-symmetry analysis such as Biberauer and Roberts 2008 or Wallenberg 2009,
incorporating pied-piping and focus movement and providing two post-verbal object
positions, would work equally well.21
1.5. Two post-verbal object positions (Note: should be 4.2)
In this section we demonstrate how the empirical facts described in Section 3 can be
accounted for under an analysis in which there are two post-verbal object positions which
differ with respect to discourse-related effects.
As we saw in Table 1, post-verbal objects are more frequent in AuxV than in VAux clauses,
both in aggregate and text by text. Under the reasonable assumption that post-verbal objects
triggered by discourse/performance constraints occur at approximately the same rate in G1
(VAux) and G2 (AuxV) clauses, the “extra” AuxVO tokens shown in Table 1 are the result of
the addition of the G3 base-generated AuxVO tokens to the G2 derived ones. In other words,
surface AuxVO tokens are derivationally ambiguous, comprising both base-generated postverbal objects and those derived by focus movement, whereas VAuxO tokens are all derived
by movement, as illustrated in Table 9.
21
Note, however, that the analysis of Biberauer and Roberts 2008 is problematic with respect to data coverage:
it predicts constituent orders that do not occur, and does not derive all of the attested constituent orders of Old
English.
47
Verb Order, Object Position and Information Status in Old English
Grammar
Surface
Order
Focus Movement?
Information Status of Object
(Idealized)
1
G1 (VAux, OV)
O V Aux
no
old
2
G1 (VAux, OV)
V Aux O
yes
new
3
G2 (AuxV, OV)
Aux O V
no
old
4
G2 (AuxV, OV)
Aux V O
yes
new
5
G3 (AuxV, VO)
Aux V O
no
old or new
Table 9: Grammars and their outputs
Table 5 above shows that the effects of information status, case and grammatical weight are
less strong in AuxV clauses than in VAux clauses, although in the case of weight, the effect
is quite small. Burridge attributes both the weight and the case effects to information status.
As noted above, she attributes the case effect to likelihood of different types of object being
topical; as the more topical, accusatives are less likely to represent new information and thus
less likely to appear post-verbally than datives. In her view (p. 107; cf. also Niv 1992)
heaviness is also simply a reflection of information status, as newer constituents require more
explanation and thus are generally longer. However, in more recent work, Arnold et al. 2000
and Gries 2003, as well as Taylor and Pintzuk 2012b have argued that heaviness has an effect
independent of information status, although the relationship between the two is
complicated.22 Given this, and given the smaller difference in the effect of weight between
VAux and AuxV clauses as compared to information status and case, it is possible that
weight represents a completely separate effect (e.g. processing) from information status and
case, and should be treated separately, although we will not pursue this point here.
Nevertheless, even though it is not entirely clear if/how these three effects are related, the
difference in the clause types is clear. Whatever is behind these effects, whether it is a single
movement related to new information focus or not, the effects are less strong in AuxV than
VAux clauses. Again, we suggest that this follows from the mix of G2 and G3 structures
underlying AuxVO order; see Table 9 above. As some (unknown) percentage of AuxV
clauses (the ones generated by G3) have post-verbal objects that are not necessarily triggered
by discourse/performance-related effects, they are not affected by these factors, thus diluting
the effects.
Finally, we saw in section 3.2 that non-referential objects (negative and incorporated) do not
appear freely in post-verbal position in VAux clauses. In general, incorporated objects are
short, and it seems reasonable to assume that they cannot easily be focussed.23 The same
cannot be said of negatives, however. Rather, negatives appear to be barred from post-verbal
position for an entirely different reason. Given that OE is a negative concord (NC) language,
in which negative constituents do not cancel each other out but rather combine to express a
single negation, one possibility is to adapt the analysis of Haeberli and Haegeman (1995), and
22
In all studies that we are aware of, the effect of length, however measured, is consistently stronger than that
of discourse factors. It thus may be that length reflects (partially) a different underlying factor, perhaps
processing cost, or it may simply be that length is easier to operationalise than discourse factors such as
information status, and thus there is more “noise” masking the effect in the latter case.
23
The apparent exceptions in Lives of Saints require further investigation.
Ann Taylor and Susan Pintzuk
48
assume that rightward movement in OE creates scope islands.24 Thus postposed negative
objects in VAux clauses cannot undergo NC, while the configuration of base-generated postverbal objects in AuxV clauses permits NC. This analysis is further supported by the absence
of post-verbal negative constituents in clauses with finite main verbs and two or more heavy
constituents in pre-verbal position. These clauses must have head-final TPs, and the
postposition of the negative constituent blocks NC. In contrast, negative constituents do
appear post-verbally in clauses with the finite main verb in second position; these clauses
may have either head-final or head-initial VPs.
Whatever the reason for the inability of negatives to postpose, the difference between
VAux and AuxV clauses on this point is clear. If non-referential objects cannot move to postverbal position, as indicated by their distribution in VAux clauses, then all AuxV clauses with
post-verbal non-referential objects necessarily belong to G3. We also saw in Table 7 that in
AuxV clauses, the frequency of post-verbal non-incorporated objects is approximately twice
as high as the frequency of incorporated ones (33% vs. 15% for all texts). For negatives
(Table 6) the difference appears greater (47% non-negative to 14% negative), but note that
this table includes negatives of any length. If we restrict the negatives to one open-class word
to match the incorporated objects in Table 7, the difference is 35% non-negative to 15%
negative, virtually identical to the frequency for incorporated objects. This distribution can be
accounted for under the assumption that for referential objects (non-incorporated and nonnegative) our AuxVO sample includes both base-generated post-verbal objects and those
derived by focus movement, while all the non-referential (incorporated and negative) objects
in post-verbal position are base-generated by the G3 grammar. This suggests, in addition, that
the frequency of underlying VO order in OE overall is around 15%.
5. Conclusions
In this paper we set out to demonstrate that the distribution of objects differs in VAux and
AuxV clauses in OE and to consider the implications for syntactic analyses of object position
in OE. Our results show that the distribution of objects differs in VAux and AuxV clauses in
three ways that may reasonably be interpreted as related to discourse/performance factors:
1) post-verbal objects are less frequent in VAux clauses than in AuxV clauses; 2) factors
associated with end position – information status, case and grammatical weight – have a
greater effect in VAux clauses than in AuxV clauses; and 3) non-referential objects do not
freely appear post-verbally in VAux clauses but do so in AuxV clauses. These results cannot
be reconciled with a single grammar in which object position is predominantly dependent on
discourse/performance factors. Rather, they show that only in VAux clauses (G1) is object
position dependent on these factors. In AuxV clauses (G2 and G3), object position is partly
dependent on discourse factors (G2) and partly fixed syntactically (G3).
The primary implication of our results for syntactic analyses of OE is that there need to be
two ways to derive post-verbal objects, one which involves focus (and/or other
discourse/performance factors) and one which does not, and that these two derivations are
associated with verb order: in VAux clauses only the G1(discourse dependent) derivation is
available, while in AuxV clauses both G2 (discourse dependent) and G3 (non-discourse
24
Note however that Haeberli and Haegeman assume that verb raising and verb projection raising, analyzed as
rightward movement of (a projection of) V, block NC, while postposition of DPs and PPs permit NC in OE.
Haegeman and Zanuttini 1996:138 claim that extraposition of PPs in West Flemish blocks NC, but apparently
only in the case where the postposed constituent is preceded by an intonation break and interpreted as an
afterthought.
49
Verb Order, Object Position and Information Status in Old English
dependent) derivations are available. We argue that this is the result of an ongoing change in
the syntax of objects during the Old English period: the existence of non-focussed objects in
post-verbal position in AuxV clauses is an innovation associated with the incoming G3
grammar.
Acknowledgements
This paper was presented at the 12th Diachronic Generative Syntax Conference (DiGS12) in
Cambridge, July 2010. We thank the audience for comments. We also thank two anonymous
reviewers for very helpful suggestions. All errors and omissions remain our own
responsibility.
References
ARNOLD, JENNIFER E., LOSONGCO, ANTHONY, WASOW, THOMAS & GINSTROM, RYAN. 2000.
Heaviness vs. Newness: The Effects of Structural Complexity and Discourse Status on
Constituent Ordering. Language 76.128-55.
BECH, KRISTIN. 2001. Word Order Patterns in Old and Middle English: A Syntactic and
Pragmatic Study. University of Bergen, Department of English PhD.
BIBERAUER, THERESA, HOLMBERG, ANDERS & ROBERTS, IAN. 2008. Disharmonic word orders
and the final-over-final constraint. In Antonietta Bisetto & Francesco E. Barbieri (eds.),
Proceedings of the XXXIII Incontro di Grammatica Generativa, 86-105. Bologna:
Universita di Bologna.
BIBERAUER, THERESA & ROBERTS, IAN. 2008. Cascading parameter changes: internally
driven change in Middle and Early Modern English. In T. Eythórsson (ed.),
Grammatical Change and Linguistic Theory: the Rosendal Papers. Amsterdam: John
Benjamins, 79–113.
BIES, ANN. 1996. Syntax and discourse factors in Early New High German: Evidence for
verb-final word order. MA dissertation, University of Pennsylvania.
BIRNER, BETTY. 2006. Inferential relations and noncanonical word order. In Betty Birner &
Gregory Ward (eds.), Drawing the Boundaries of Meaning: Neo-Gricean Studies in
Pragmatics and Semantics in Honor of Laurence R. Horn, 31-51.
Amsterdam/Philadelphia: Benjamins.
BURRIDGE, KATE. 1993. Syntactic change in Germanic. Amsterdam Studies in the Theory and
History of Linguistic Science, Volume 89. Amsterdam: John Benjamins.
DU BOIS, JOHN W. 1980. Beyond definiteness: The trace of identity in discourse. In Wallace
L. Chafe (ed.), The Pear Stories: Cognitive, Cultural, and Linguistic Aspects of
Narrative Production, 203-274. Norwood, New Jersey: Ablex Publishing Corporation.
DUNBAR, RICHARD. 1979. Discourse Pragmatics and Subordinate Clause Word Order in
German: An Explanation of Related Main Clause Phenomena in German and English
Clauses. PhD dissertation. University of Wisonsin, Madison.
GRIES, STEFAN TH. 2003. Multifactorial Analysis in Corpus Linguistics: a Study of Particle
Placement. London: Continuum.
Ann Taylor and Susan Pintzuk
50
GUNDEL, JEANETTE, HEDBERG, NANCY & ZACHARSKI, RON. 1993. Cognitive status and the
form of referring expressions in discourse. Language 69(2). 274-307.
HAEBERLI, ERIC & HAEGEMAN, LILIANE. 1995. Clause structure in Old English: Evidence
from negative concord. Journal of Linguistics 31. 81-108.
HAEGEMAN, LILIANE & ZANUTTINI, RAFFAELLA. 1996. Negative concord in West Flemish. In
Adriana Belletti & Luigi Rizzi (eds.), Parameters and Functional Heads. Oxford:
Oxford University Press. 117-179.
HINTERHÖLZL, ROLAND. 2009. The role of information structure in word order variation and
word order change. In Roland Hinterhölzl & Svetlana Petrova (eds.), Information
Structure and Language Change: New Approaches to Word Order Variation in
Germanic, 45-66. Berlin: Mouton de Gruyter.
HODGSON, K. 2012. Definiteness and Case in Eastern Armenian. Unpublished final-year
undergraduate dissertation: University of Cambridge.
KARTTUNEN, LAURI. 1976. Discourse referents. In J.D. McCawley (ed.), Syntax and
Semantics 7: Notes from the Linguistic Underground, 363-385. New York: Academic
Press.
KROCH, ANTHONY S. 1994. Morphosyntactic variation. In Katharine Beals, Jeannette Denton,
Robert Knippen, Lynette Melnar, Hisami Suzuki & Erica Zeinfeld (eds.), Proceedings
of the 30th annual meeting of the Chicago Linguistics Society, vol 2, pp. 180-201.
Chicago: Chicago Linguistics Society.
KROCH, ANTHONY S. 2003. Syntactic change. In Mark Baltin & Chris Collins (eds.),
Handbook of Contemporary Syntactic Theory, Oxford: Blackwell. 699-729.
KROCH, ANTHONY S. & TAYLOR, ANN. 2000. Verb-complement order in Middle English. In
Susan Pintzuk, George Tsoulas & Anthony Warner (eds.), Diachronic Syntax: Models
and Mechanisms, 132-163. Oxford: Oxford University Press.
KUNO, SUSUMU. 1976. Subject, theme, and the speaker’s empathy. In Charles N. Li (ed.),
Subject and Topic, 417-44. New York: Academic Press.
NIV, MICHAEL. 1992. Right association revisited. Paper presented at the Thirtieth Annual
Meeting of the Association for Computational Linguistics.
OLIVER, LISI. 2002. The beginnings of English Law. Toronto: University of Toronto Press.
PETROVA, SVETLANA. 2009. Information structure and word order variation in the Old High
German Tatian. In Roland Hinterhölzl & Svetlana Petrova (eds.), Information Structure
and Language Change: New Approaches to Word Order Variation in Germanic, 251280. Berlin: Mouton de Gruyter.
PINTZUK, SUSAN. 1996. Cliticization in Old English. In Aaron L. Halpern & Arnold Zwicky
(eds.), Approaching Second: Second Position Clitics and Related Phenomena, 366-400.
Stanford, CA: CSLI Press.
PINTZUK, SUSAN. 1999. Phrase Structures in Competition: Variation and Change in Old
English Word Order. New York: Garland.
PINTZUK, SUSAN. 2005. Arguments against a universal base: evidence from Old English.
English Language and Linguistics 9(1). 115-138.
PINTZUK, SUSAN & KROCH, ANTHONY S. 1989. The rightward movement of complements and
adjuncts in the Old English of Beowulf. Language Variation and Change 1. 115-143.
PINTZUK, SUSAN & TAYLOR, ANN. 2006. The loss of OV order in the history of English. In
Ans van Kemenade & Bettelou Los (eds.), The Handbook of the History of English,
249-278. Oxford: Blackwell.
PRINCE, ELLEN. 1981. Toward a taxonomy of given-new information. In Peter Cole, (ed.),
Radical Pragmatics, 223-255. New York: Academic Press.
51
Verb Order, Object Position and Information Status in Old English
ROBERTS, IAN. 1997. Directionality and word order change in the history of English. In Ans
van Kemenade & Nigel Vincent (eds.), Parameters of Morphosyntactic Change, 397426. Cambridge: Cambridge University Press.
RÖGNVALDSSON, EIRÍKUR. 1987. On word order in Icelandic. In Proceedings of the Seventh
Biennial Conference of Teachers of Scandinavian Studies in Great Britain and
Northern Ireland, 33-49. London: University College London.
SAPP, CHRISTOPHER. 2006. Verb Order in Subordinate Clauses from Early New High
German to Modern German. PhD dissertation, Indiana University, Bloomington.
SIEWIERSKA, ANNA. 1993. Syntactic weight vs information structure and word order variation
in Polish. Journal of Linguistics 29(2). 233-265.
SVENONIUS, PETER. 2000. Quantifier movement in Icelandic. In Peter Svenonius, (ed.), The
Derivation of VO and OV, 255-292. Amsterdam: John Benjamins.
SZMRECSÁNYI, BENEDIKT. 2004. On operationalizing syntactic complexity. In Gérard
Purnelle, Cédrick Fairon & Anne Dister (eds.), Le poids des mots. Proceedings of the
7th International Conference on Textual Data Statistical Analysis, 1032-1039.
Louvain-la-Neuve: Presses universitaires de Louvain.
TAYLOR, ANN. 2006. Rhythmic prose and its effect on verb-object order. Paper presented at
the York-Holland Symposium on Historical English Syntax (SHES). York, UK.
TAYLOR, ANN, WARNER, ANTHONY, PINTZUK, SUSAN & BETHS, FRANK. 2003. The YorkToronto-Helsinki Parsed Corpus of Old English Prose (wwwusers.york.ac.uk/~lang22/YCOE/YcoeHome.htm). Electronic texts and manual
available from the Oxford Text Archive.
TAYLOR, A. & PINTZUK, S. 2012a. The interaction of syntactic change and information status
effects in the change from OV to VO in English. Catalan Journal of Linguistics 10.
1-24.
TAYLOR, A. & S. PINTZUK. 2012b. Rethinking the OV/VO alternation in Old English: the
effect of complexity, grammatical weight, and information status. In Terttu Nevalainen
and Elizabeth Traugott (eds.). The Oxford Handbook of the History of English. New
York: Oxford University Press. 1199-1213.
TAYLOR, ANN & VAN DER WURFF, WIM. 2005. Special issue on aspects of OV and VO order
in the history of English. English Language and Linguistics 9(1). 1-4.
VAN DER WURFF, WIM. 1997. Deriving object-verb order in late Middle English. Linguistics
33. 485-509.
VAN DER WURFF, WIM. 1999. Objects and verbs in modern Icelandic and fifteenth-century
English: A word order parallel and its causes. Lingua 109. 237-265.
VAN KEMENADE, ANS. 1987. Syntactic Case and Morphological Case in the History of
English. Dordrecht: Foris.
VAN KEMENADE, ANS & LOS, BETTELOU. 2006. Discourse adverbs and clausal syntax in Old
and Middle English. In Ans van Kemenade & Bettelou Los (eds.), The Handbook of the
History of English, 224-248. Oxford: Blackwell.
WALLENBERG, JOEL. 2009. Antisymmetry and the Conservation of C-command: Scrambling
and Phrase Structure in Synchronic and Diachronic Perspective. PhD dissertation,
University of Pennsylvania.
WASOW, THOMAS. 1997. Remarks on grammatical weight. Language Variation and Change
9. 81-105.
Ann Taylor and Susan Pintzuk
Ann Taylor, Susan Pintzuk
Department of Language and Linguistic Science
University of York
Heslington
York
YO10 5DD
email: [email protected], [email protected]
52