64

Click here to load reader

Movement and Deletion in Old English

Embed Size (px)

Citation preview

Page 1: Movement and Deletion in Old English

Movement and Deletion in Old EnglishAuthor(s): Cynthia AllenSource: Linguistic Inquiry, Vol. 11, No. 2 (Spring, 1980), pp. 261-323Published by: The MIT PressStable URL: http://www.jstor.org/stable/4178162 .

Accessed: 16/06/2014 01:30

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .http://www.jstor.org/page/info/about/policies/terms.jsp

.JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range ofcontent in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new formsof scholarship. For more information about JSTOR, please contact [email protected].

.

The MIT Press is collaborating with JSTOR to digitize, preserve and extend access to Linguistic Inquiry.

http://www.jstor.org

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 2: Movement and Deletion in Old English

Linguistic Inquiry Volume 11 Number 2 (Spring, 1980) 261-323.

Cynthia Allen Movement and Deletion in Old English*

0. Introduction

Chomsky (1977) proposes that if a construction has the following properties, it involves Wh Movement:

(1) a. It leaves a gap. b. Where there is a bridge,1 there is an apparent violation of Subjacency,

the Propositional Island (Tensed-S) Condition, and the Specified Subject Condition.

c. It observes the Complex NP Constraint. d. It observes the Wh-Island Constraints.

Noting that several constructions traditionally analyzed as involving controlled deletion over a variable show the properties (la-d), Chomsky suggests that perhaps it would be possible to eliminate deletion over a variable altogether. All constructions appearing to involve such deletion and showing properties (la-d) would then actually involve Wh Movement, with subject deletion of the wh-pronoun. To avoid circumlo- cutions, I will refer to such Wh Movement and later deletion as "ghost" Wh Movement, although the processes of movement and deletion are of course separate.

Nothing in Chomsky's system of conditions rules out the possibility of unbounded deletion, but Chomsky and Lasnik (1977) have proposed a theory of grammar by which such deletion is impossible. In this theory, the only deletion rules are quite restricted local ones which apply after the transformations of "core" grammar, such as Wh Movement. Chomsky and Lasnik further assume that transformations are all unordered,

* I wish to express my thanks to Joan Bresnan, S. J. Keyser, and Lisa Selkirk for comments on an earlier version of this article, and to Avery Andrews and three anonymous LI readers for helpful comments on this version. I would also like to thank Joan Maling for information and discussions about the Germanic languages, and Jean Lowenstamm, Anke de Rooij, and Elizabeth Patz for their help with Yiddish, Dutch, and German, respectively. All errors are of course my own. Finally, I wish to acknowledge my debt to Jane Grimshaw, whose work on Middle English drew my attention to the importance of the OE facts discussed here.

1 A bridge permits extraction from a tensed clause. Chomsky says that it is unclear what particular property of the matrix VP causes it to be a bridge.

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 3: Movement and Deletion in Old English

262 CYNTHIA ALLEN

optional, and cannot express contextual dependencies. They suggest (p. 443) that "the consequences of ordering, obligatoriness, and contextual dependency can be captured in terms of surface filters."

In this article I will present some facts from Old English (OE) which indicate a need for unbounded deletion rules showing configuration (1). I am using Old English here because in that stage of our language there was a clear-cut difference with respect to preposition stranding between constructions apparently involving deletion and those clearly involving movement. A subset of these facts has been discussed in Allen (1976) and Bresnan (1976a). Similar data for Middle English (ME) are presented in Grimshaw (1975). Chomsky and Lasnik (1977) argue that these facts could easily be accommodated within their theory by the use of surface filters and are in any case irrelevant. I will examine here a wider range of constructions in OE and show that these data are relevant to the question of whether unbounded deletion is needed, and furthermore cannot be easily accommodated within Chomsky and Lasnik's system. We will see that the filters proposed by Chomsky and Lasnik do not deal adequately with the facts. I will also point out the inadequacies of a different approach within the ghost Wh Movement framework proposed under the pseudonym of Jan Vat (1978).

1. On the Reliability of the OE Texts

The difficulties inherent in drawing theoretical conclusions from a dead language are obvious, so I will attempt here to allay misgivings before discussing particular OE constructions.

It should be noted first that the OE corpus of prose writings is very large. For a complete list of the texts examined, see the appendix.

Second, it should be emphasized that while many OE writings are either translations of Latin works or works based on Latin sources, investigation of various OE construc- tions has shown Latin influence to be slight.2 Systematic divergences between the Latin and OE syntax are found in all these translations. In particular, Latin relative clauses employing relative pronouns and exhibiting "pied piping" are frequently rendered in the English translation by an indeclinable relative particle and a stranded preposition, even though the Latin construction was also possible in English. We also find no significant difference in the constructions under consideration here between translations of Latin and original prose works .3

We must also ask how much the written style diverged from the spoken language.

2 Claims of strong Latin influence in OE syntax have usually come from people who have looked very little at the syntax. Those who have studied particular constructions dispute the claim of strong Latin influence. See for example Zeitlin (1908) on translations of the Latin infinitive with accusative subject; Wende (1915, 20-24) on the independence of OE word order in translations from that of the Latin originals; and Bacquet (1962). See also Brown's (1970) arguments that there was little Latin influence even in the syntax of one of the most latinate OE translations.

3While the translations do not entirely escape Latin influence, instances of such interference stick out fairly dramatically, and at any rate are not found in the particular constructions discussed here.

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 4: Movement and Deletion in Old English

MOVEMENT AND DELETION IN OLD ENGLISH

This is a difficult question, but I am confident that written Old English differed less from the spoken language than did the writings of later periods when there had been more time for literary conventions for English to develop. One important consideration is that the OE writers were writing either for those who could not read at all (as with the sermons of Aelfric and Wulfstan, which were to be read aloud to illiterate parishion- ers) or for those who could read English, but not Latin. We may assume that writings for the illiterate would be as close to ordinary speech as possible, even though the style might be more elevated.4 This is particularly true because Aelfric and Wulfstan were extremely concerned about saving people's souls by imparting to them a fundamental understanding of Christianity. Aelfric frequently mentions that he is writing for under- standing, not elegance.

To summarize, the OE prose texts are extensive enough to permit us to draw certain conclusions about their syntax with confidence, and their syntax appears to be representative of that of the spoken language. We must still, however, face the problem of the lack of negative evidence. It is true that we cannot be absolutely certain that given constructions were not possible in Old English. However, the severity of this problem varies with the construction, and each case must be considered individually. If the corpus is rich in a certain construction, but all the texts fail to show a hypothesized variation of that construction (which is no more complicated than the attested examples), it is reasonable to conclude that the hypothesized variation was not possible. This conclusion is strengthened if we find that all the texts of a later period suddenly do show the variation under consideration. This is precisely the situation we find in Old and Late Middle English with respect to preposition stranding in questions and relatives using pronouns.5

A particular aspect of the problem of negative data is the question of whether it is possible to determine anything about the behavior of Old English with respect to Chomsky's proposed conditions on movement. This is a crucial question, as the argu- ments presented here depend on the assumption that certain types of OE relatives, like their modern counterparts, showed configuration (1).

Chomsky and Lasnik (1977, 498) state that "no one has the slightest idea" whether OE relative clauses and questions were governed by the Complex NP and Wh-Island

4 We find that writings for the literate tend to have more complex sentences, but no difference in the basic syntax; in particular, the pied piping facts are the same in the two styles.

' The difference between these two periods cannot be due to a change in literary convention, as the superficial differences in the OE constructions allowing preposition stranding and those not permitting it are in some cases too subtle to be distinguished by a mere convention. The literary conventions I know of are based on surface structure and are applied with little sensitivity to the differences in various constructions. A typical example is the literary prejudice against preposition stranding which began in the seventeenth century. Dryden, for example, went so far as to "correct" all instances of preposition stranding in his essay

Of Dramatic Poesy". He applied his rule blindly, not permitting stranded prepositions even in that-relatives, which have always allowed stranding. The results of this sort of rule are very different from the facts we find in the OE period. At any rate, I know of no evidence that the OE writers followed such conventions. The desire to standardize the language (and to ape Latin) came later.

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 5: Movement and Deletion in Old English

264 CYNTHIA ALLEN

Constraints since no speakers are available. They argue that the lack of counterexamples to the constraints in the texts does not indicate that the constraints held, because "the relevant constructions are quite rare." I assume that by this they mean that these structures are complicated and therefore cannot be expected to occur frequently. How- ever, violations of the Complex NP Constraint such as *Who do vou doubt the fact that Bill saw? are no more complex, syntactically or semantically, than extractions out of tensed clauses more than one sentence down, such as Who do You doubt that Bill said he saw? There is no reason why sentences of the first type should be rarer than those of the second in a language permitting both,6 but in the OE texts, although we find no violations of the Complex NP or Wh-Island Constraints,7 we do find several extractions out of doubly embedded tensed clauses:8

(2) & geliefe6 6aet he swelc sie swelce he gehier6 6aet his olicceras and believes that he such is as he hears that his flatterers saecgea6 6xet he sie say that he is 'and believes that he is such as he hears that his flatterers say that he is _

CP Sweet p.110.11 (3) Ac is wolde witan hu de &uhte be 6acm monnum 6 wit

but I would know how thee-dat. seemed about the men that we aer cwaedon 6et unc 6uhte oaet weron wildiorum gelicran 6one earlier said that us seemed that were wild-beasts like-er than monnu. men 'But I would like to know how it seemed to you about the men that we said earlier (that) were more like wild beasts than men.'

Boeth.XXVIII.5 p. 122.13

6 If these sentences are indeed rare in the languages allowing them, the question of how the language learner knows that they are possible in his language must be considered.

7The only violation of either of these constraints that I have recorded has a returning pronoun:

(i) . . .6ot nas an hid landes innan Englklande 6a5t he nyste hwa heo hefde that not-was one hide land-gen. within England that he not-knew who it had

'that there was not one hide of land within England that he did not know who had it' Sax.Chron. 1086

It is well known that returning pronouns improve island violations. The fact that such a pronoun was needed in this example is a further bit of evidence that OE was subject to the Wh-Island Constraints.

8 Notice the violation of the Fixed Subject Constraint in (3). Such violations were extremely common in OE relatives and questions. This is of interest because OE had no pronoun-drop rule, and Chomsky and Lasnik's proposed explanation for the Fixed Subject Constraint relies crucially on the hypothesis that all languages not subject to this constraint have a pronoun-drop rule. Two modern Germanic languages which do not obey the Fixed Subject Constraint but lack a pronoun-drop rule are Icelandic and the majority dialect of Dutch. For a discussion of the Icelandic facts, see Maling and Zaenen (1978b).

It was erroneously reported in Allen (1977) that Norwegian was not subject to this constraint. Icelandic, not Norwegian, was intended there.

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 6: Movement and Deletion in Old English

MOVEMENT AND DELETION IN OLD ENGLISH

Such examples show that the lack of violations of the Complex NP and Wh-Island Constraints cannot be due to a lack of sufficiently complex sentences in the texts.

Finally, consider the correlation found by Maling (1978) between Old Norse texts and some facts in the modern Scandinavian languages. In these modem languages, it is possible to relativize, but not to question, out of indirect questions, and it is impossible either to relativize or to question out of a complex NP. These facts present problems for Chomsky's system. Maling also found that precisely the same situation is indicated by the Old Norse texts. The fact that the same sorts of violations of the Wh-Island Constraint are found in these texts as are possible in the modem language, while no violations of the Complex NP Constraint are found, strongly suggests that we may expect to determine the basic facts about conditions on movement from a large written corpus.

To summarize, because of the extent and reliability of the OE texts, I have complete confidence in the facts presented here, which were gathered by means of a thorough examination of all the texts in the appendix. Naturally, no one will wish to accept or reject a grammatical theory solely on facts from a dead language. But there is evidence from living languages for the need for unbounded deletion rules, and against treating all constructions with configuration (1) as involving movement. For details, see de Rijk (1972), Bresnan (1975; 1976a), Maling (1978), and Bresnan and Grimshaw (1978), among others. This evidence from living languages means that we can give more credence to the OE facts than we would if they were the only known counterexamples to Chomsky's theories. When the data are as clear as those for Old English, facts from dead languages should surely be used to augment our knowledge of universal grammar.

2. The OE Facts

Now let us turn to the evidence of the texts concerning the nature of certain rules in Old English. In the earlier published literature, only questions and a couple of types of relatives have been discussed. Here we will look at a wider range of constructions, including Complement Object Deletion structures, topicalization, and a wider array of relative constructions. We will see that, in all constructions in which there was an item on the surface which had clearly been moved,9 pied piping of prepositions was obliga- tory, while preposition stranding is found in those in which the affected item was deleted. These facts fit easily into a theory making a distinction between movement and unbounded deletion, but we will see in sections 3 and 4 that these facts present problems for any ghost Wh Movement approach.

2.1. Headed Relatives

We will first consider the behavior of prepositions in relative clauses with heads (as opposed to "free" relatives). I will first present the basic facts about the three major

9 I will assume that an item has clearly been moved if it appears "out of place" in the sentence and furthermore exhibits the case marking we would expect from its logical underlying position.

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 7: Movement and Deletion in Old English

266 CYNTHIA ALLEN

types of headed relatives, followed by a discussion of how each type is to be analyzed. We will then consider a less common fourth type, and also infinitival relatives, before moving on to free relatives. It will be seen that in each type of relative in which a moved pronoun is to be found at the surface, pied piping was obligatory, while in relatives in which no such pronoun is to be found, preposition stranding occurred.

2.1.1. oe Relatives The most common type of 0 E relative employed the indeclinable relative complementizer &e.10 These relatives were analogous to Modern English that- relatives. That be was not a relative pronoun is shown by the fact that it did not decline for case, number, and gender, as did all other OE pronouns (including relative pro- nouns), nor could it be preceded by a preposition of which it was the object, unlike the true OE relative pronouns.

&- could be employed regardless of the role of either the head of the relative or the relativized item in their respective clauses.'1 The following are a few typical examples:

(4) Gemyne he ows yfeles & he worhte. remember he the-gen. evil-gen. that he wrought

'Let him remember the evil that he wrought.' CP Sweet p.25.3

(5) ... .oxt ealle his geferan & him mid eodon sceoldon sealmas leornian that all his companions that him with went should psalms learn

'that all his companions that went with him should learn psalms' Alc.S .XXVI.77

(6) Se halga froforgast min Faxder asent on minum naman he eow the Holy Ghost that my Father sends in my name he you

taec6 ealle 6ing. teaches all things

'The Holy Ghost that my Father will send to you in my name, he will teach you all things.'

Alc.P.X. 11 Relativization out of tensed subordinate clauses was extremely common:'2

10 According to the OED, &e was "apparently an unstressed worn-down case or derivative formation from the stem &a of that demonstrative and relative pronoun." The complementizer that was also originally a demonstrative pronoun.

"' The claim of Bever and Langendoen (1971, 441) that "in relative clauses introduced solely by the relative particle &e, the shared nominal could be deleted only if it was the subject of the relative clause," is incorrect, as indicated by examples (4) and (6), along with many others. This error has also been pointed out by Bresnan (1976a).

12 There was in fact more opportunity for extractions out of tensed clauses in Old English than in Modern English, because the use of the infinitive was formerly much more restricted than it is now, and tensed clauses were used in many situations where an infinitival clause would be used today. (7) is such a case.

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 8: Movement and Deletion in Old English

MOVEMENT AND DELETION IN OLD ENGLISH

(7) 5is is se rihta geleafa&e gehwylcum men gebyre6 6it he wel this is the correct belief that each-dat. man-dat. behooves that he well

gehealde & geleste. hold and perform

'This is the correct belief that it behooves every man that he hold __ and perform well.'

Blickling p. III. 12 (8) Ne meaht Ou da 6rowunge gelettan, e Faxder wolde & geteohod

not mightest thou the suffering hinder that Father would and determined

haefde 6aet ic for mancynnes halo ge6rowian sceolde. had that I for mankind's salvation suffer should

'You cannot prevent the suffering that the Father desired and had determined that I should suffer __ for the salvation of mankind.'

Ver.I.42 (9) . . .6axt he ongiet his agene unnytte 6eawas & geoohtas be

that he perceive his own unprofitable morals and ideas that

wen is 6at he xr hafde probable is that he earlier had

'that he perceive his own unprofitable morals and ideas which it is probable that he had __ earlier'

CP Sweet p.258.14

In be-relatives, if the relativized item was the object of a preposition, that prepo- sition was always "stranded". The surface position of the stranded preposition was nearly always directly before the verb of the relative clause (or else before ne 'not' or to, which were inseparable from the verb):13

(10) Ac he sylf asmeade 6a up-ahefednysse te he turh ahreas. but he self devised the presumption that he through fell

'But he himself thought up the presumption that he fell through (i.e., on account of which he fell).'

Alc.Th.Vol. 1 p. 192.17 (1 1) Seo gesih6 &e we god myd geseon scylon is angyt.

the sight that we God with see shall is understanding

'The sight that we shall see God with __ is understanding.' Sol. p.67.6

13 However, a few examples with the preposition after the verb are found in OE, and this order became the norm in the thirteenth century.

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 9: Movement and Deletion in Old English

268 CYNTHIA ALLEN

(12) ... .on &ere ylcan byrig be crist on 6rowode in the same city that Christ in suffered

in the same city that Christ suffered in _

Alc .S .XXVII .9

Such examples are to be found on nearly every page of any OE text. It is important to establish that these are truly stranded prepositions, and not cases of verbs with prepositional prefixes. Such compound verbs were common in Old English; a few examples are ofer-&-on 'to excel' (lit. 'over-perform'), in-gan 'to enter' (lit. 'in-go'), and iurh-stingan 'to stab through, pierce' (lit. 'through-thrust'). It is clear in these cases

(and many others) that the particle preceding the verb is not a true preposition, because it always precedes the verb, while true prepositions may follow the verb if their objects are not deleted, and these particular preposition + verb combinations occur even when there is no understood missing object of the "preposition":

(13) and in &es gesi6es hus ineode and into the companion's house in-went land entered into the companion's house'

Bede V.4 p.394.27 (14) . . .oat he huru hine selfne ne ;burhstinge mid 6y sweorde

that he at least himself not through-sting with the sword unryhthaemedes fornication-gen. 'that he at least not stab himself with the sword of fornication'

CP Sweet p.313.8

On the other hand, the preposition + verb combinations in (10) through (12) are found only in constructions in which there is an understood missing object of the preposition, as in &-relatives. Thus, we see that not all P+Verb combinations are true cases of stranded prepositions. I have been careful to use as examples of preposition stranding only those particular combinations which are not found as compound verbs.14

2.1.2. Se-Relatives The second major type of OE relative used a demonstrative pro- noun as a relative pronoun, a common usage in the Germanic languages. This type is usually known as the "se'"-relative, after the masculine nominative singular form of the pronoun. The following are typical examples of se-relatives:

14 I have not relied here only on my own observations about what verbs appear with particles in cases where there is no possibility of preposition stranding. The Bosworth and Toller OE dictionary shows consid- erable sensitivity to the difference between compound verbs and stranded prepositions before verbs. I have not used any combinations listed in this dictionary as compound verbs in my examples of preposition stranding.

A very helpful study of OE prepositions which also is sensitive to this question is Wende (1915).

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 10: Movement and Deletion in Old English

MOVEMENT AND DELETION IN OLD ENGLISH

(15) Ac ge onfo6 &em mTgene Halges Gastes but you receive the-n.d.s. power-n.d.s. Holy-m.g.s. Ghost-m.g.s. se cyme6 ofor eow. who-m.n.s. comes over you 'But you shall receive the power of the Holy Ghost, who will come over you.,

Blickling p. 1 19. 1 1 (16) Her feng to Dearne rice Osric, &one Paulinus er

here succeeded to Deira kingdom Osric whom-m.a.s. Paulinus earlier

gefullode. baptized

'In this year, Osric, whom Paulinus had earlier baptized, succeeded to the kingdom of Deira.'

Sax.Chron.634 (17) 6a man ofsloh 6es Caseres gerefan se was Labienus

then one killed the emperor's tribune-m.a.s. who-m.n.s. was Labienus gehaten. called

'Then someone killed the emperor's tribune, who was called Labienus.' Sax.Chron.Prologue

The case of the relative (demonstrative) pronoun was regularly that required by the role of the relativized item in the lower clause, rather than that of the head, when these differed. There are, however, a very few examples in which the relative pronoun has been "attracted" into the case of the head NP.

As with i-relatives, it was possible to relativize out of a tensed subordinate clause:

(18) Ic seolfa cube sumne bro6ar, &rne ic wolde 6at ic I self knew some brother-m.a.s. whom-m.a.s. I would that I nafre cube. never knew

'I myself knew one brother whom I wish that I never knew.' Bede XV p.442.9

(19) Eode da to sumum maessepreoste, from &em he gewende it went then to a masspriest from whom-m.d.s. he thought that him halu weg aeteawed beon meahte. him salvation way shown be might 'He went then to a masspriest, by whom he thought that the way to salvation might be shown to him.'

Bede IV.25 p.350.16

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 11: Movement and Deletion in Old English

270 CYNTHIA ALLEN

(20) Syx dagas synd on iam gebyra6 &et man wyrce. six days are on which is fitting that one work 'There are six days on which it is fitting that one work.'

St.Luke 880

Unlike the 6e-relatives, se-relatives did not permit preposition stranding. Rather, pied piping was obligatory:

(21) . . .6at he us 6ingige wi6 6one Heofonlican Cyning, for &es that he us intercede with the Heavenly King for whose

naman he 6rowode. name he suffered 'that he intercede for us with the Heavenly King, for whose name he suffered'

Alc.Th. Vol.1 p.434.35 (22) Gehyr 6u arfesta God mine stefne, mid Yxre ic earm to

hear thou merciful God my voice-f.a.s. with which-f.d.s. I poor to 6e cleopie. thee cry 'Hear thou, merciful God, my voice, with which I, poor one, cry to thee.'

Blickling p.89. 13 (23) and het getimbrian medomlic hus, on &et nxnig

and ordered to build small house-n.a.s. in which-n.a.s. no wer nefde ingang man not-had admittance 'and ordered a small house built, in which no man had admittance'

Mart. p.106.5

I have found no counterexamples to the generalization that preposition stranding was impossible in OE se-relatives. My findings here agree with those of Wende (1915), who explicitly states the same conclusion. Visser (1963) also gives no counterexamples to this rule. The only exception to obligatory pied piping of prepositions in this con- struction is the systematic stranding in relatives with the locative pronoun 3'r, which will be discussed in section 2.6.

2.1.3. Se oe-Relatives The third type of OE relative appears to be a combination of the first two, exhibiting both the demonstrative pronoun and the indeclinable be. The following are examples of this "se &e" type:15

15 One difference between se- and se &-relatives is that the latter show more frequent attraction to the case of the head. This is probably because these relatives were so similar to those with demonstrative pronouns as heads, discussed in section 2.2.1.1. The two types are sometimes indistinguishable.

I have found no clear examples of apparent violations of Subjacency, the Tensed-S Condition, or the Specified Subject Condition with the se &e-relatives. I do not believe in this case that this indicates that such apparent violations were not possible, because the se &-relative was considerably rarer than the other two types. It should be noted that there are many fewer apparent violations of the conditions with se-relatives

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 12: Movement and Deletion in Old English

MOVEMENT AND DELETION IN OLD ENGLISH

(24) Ic wat wytodlice but ge seca6 bone Halend xone I know truly that you seek the-m.a.s. Savior-m.a.s. whom-m.a.s.

&e on rode ahangen was. that on cross hanged was

'I know truly that you seek the Savior who(m) (that) was hanged on the cross.

St.Mat. 1766 (25) Ure Drihten ararde anes ealdormannes dohtor, seo

our Lord raised an alderman's daughter-f.a.s. who-f.n.s. &? kxg deed. that lay dead

'Our Lord raised an alderman's daughter, who (that) lay dead.' Alc.P.VI. 176

(26) swa swa Aaron wes, se arwuroa bisceop, &one as Aaron was the-m.n.s. worthy-m.n.s. bishop-m.n.s. whom-m.a.s.

3e God sylf geceas that God self chose

'as Aaron was, the worthy bishop, whom (that) God himself chose' Alc.P.XX.243

As with the plain se-relatives, when the relativized item in a se &-relative was the object of a prepositional phrase, pied piping was obligatory:16

(27) Eala 6u wundorlice rod, on ;ere &e crist wolde orowian. hail thou wonderful cross on which-f.d.s. that Christ would suffer

'Hail, thou wonderful cross, on which (that) Christ deigned to suffer.' Alc.S.XXVII. 1 15

than with &-, probably because &- was by far more common. The structural ambiguity of the se &e-relatives with the demonstrative-headed ones further reduces the chances of finding apparent violations, as I have not considered ambiguous examples.

16 In the texts on which he based his study, Wende (1915) found only one possible counterexample to obligatory pied piping in se- and se &-relatives. This example is apparently due to scribal error, as a later manuscript changes this sentence, using a &-relative and conforming to the rule.

Wende also found three more possible counterexamples in Gregory's Dialogues, a text which he did not use to base his statistics on. He found this text to be exceptional in other constructions as well, and concluded that it was truly different from the texts of his study. At any rate, all three examples are dubious. In one, the stranded preposition is inne, which Bosworth and Toller list as being only an adverb. In another, te is supposedly stranded before stande, but Bosworth and Toller list ctstandan as a compound verb. The third example appears in a different form in a second manuscript, which does not violate the restriction on preposition stranding. Thus, no convincing failures of pied piping in se- or se &-relatives have been found.

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 13: Movement and Deletion in Old English

272 CYNTHIA ALLEN

(28) Wa 6am men &urh bone 5e by6 mannes woe the-m.d.s. man-m.d.s. through whom-m.a.s. that is man's

sunu belhwed. son betrayed

'Woe to the man through whom (that) the son of man is betrayed.' St.Mat. 1561

2.1.4. Analyzing the Relatives It seems clear that se-relatives should be analyzed as involving movement of the demonstrative relative pronoun. Otherwise, we would expect the case of these pronouns always to be determined by the upper clause, instead of by that of the relativized NP, and we would not expect pied piping of prepositions to be possible. A movement analysis explains the fact that these relative pronouns follow prepositions only when a preposition could be generated in the lower clause.

By Chomsky's system, se-relatives would involve movement into Comp, since they exhibit apparent violations of the Specified Subject, Subjacency, and Tensed-S Condi- tions. If we reject Chomsky's system, we may analyze these relatives as involving movement to the front of S, before Comp.

The se &e-relatives are probably to be analyzed exactly the same way as the se type, except that the complementizer &e is deleted (or not spelled out) in the se type. One might be tempted to analyze the &e of se &e as merely a part of the pronoun, but in genitive relatives in which the pronoun is not the final part of the relativized phrase, the 3e appears separated from the pronoun:

(29) and &5et he hine betehte 6am heofonlican Gode, and that he him commit the-m.d.s. heavenly-m.d.s. God-m.d.s. &irh 5as mihte 5e he afligde 6aera he6enra godas through whose power that he drove out the heathens' gods

and that he commit him to the heavenly God, through whose power (that) he drove out the heathens' gods'

Alc.P.XXI.636

Thus, 5e is clearly a complementizer here. For the be-relatives, we must determine whether movement has first taken place,

and then deletion of the pronoun (in or before Comp), or whether deletion alone has occurred. In the latter event, is it deletion under identity with the head, or a free deletion of the relative pronoun?

We can easily satisfy ourselves that these relatives are not formed by free deletion of a pronoun. First, OE had no general rule of pronoun deletion, as noted by Bresnan (1976a). Second, in OE it was possible to delete a whole prepositional phrase by relativization, just in case the relativized prepositional phrase was identical with a

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 14: Movement and Deletion in Old English

MOVEMENT AND DELETION IN OLD ENGLISH

prepositional phrase containing the head NP:

(30) ... .oonne he gesiho 6dt mod on em ilcan unryhtan willan 6e he bi6 when he sees the mind in the same evil desires that he is (in)

CP Sweet p.415.25 (31) ludas se swicola hraoe eode to 3am arleasum ehterum 3e

Judas the traitor quickly went to the impious persecutors that he ar gesprxc. he earlier spoke (to)

Alc.Th. Vol.11 p.246.7 (32) and he was eft-cyrrende &irh &one ylcan sii(at &es

and he was returning through the same path the-gen. westenes &e he er 6yder becom wilderness-gen. that he earlier thither came .and he was returning by the same path in the wilderness that he had come there (by) earlier'

Alc.S.XXIIIb.641

In each of these cases, we must supply a preposition in the relative clause identical with the preposition of which the head is the object. It should be noted that prepositions were not omitted as freely in OE as in languages relying more heavily on case, such as Latin. Prepositions like on and 5urh in particular could not generally be omitted.17 It is only in these &e-relative clauses having an identical preposition in the head that they disappear. Clearly, a simple pronoun deletion rule could not delete the whole PP. If no movement was involved in these relatives, the only solution is a rule of unbounded deletion under identity, with identity being defined on either an NP or a PP head:

(33) Proposed Relative Deletion 18

(P) NP g[Wl x[(P) Pro] W2] [+dem]

1 2 3 4 5 6 => 1 2 3 p ) 6

Finally, it is well known that languages with free pronoun deletion in relatives tend not to obey the Complex NP and Wh-Island Constraints. But no violations of these constraints are found among the extremely numerous OE &-relatives.

2.1.5. oxt-Relatives We have seen that in OE 5e was the ordinary complementizer for

17 See Quirk and Wrenn (1957, 66). 18 (P) means that the structural description may be met whether there is a preposition or not. For this

use of parentheses in transformations, see Bresnan (1976b). It is unnecessary to mention as part of the rule that I = 4 and 2 = 5, since recoverability of deletion

demands this identity.

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 15: Movement and Deletion in Old English

274 CYNTHIA ALLEN

relative clauses. However, even in OE 3it 'that' was occasionally used in relatives.19 Since we find some examples of preposition stranding with set, we must establish that &et in these relatives was a complementizer, rather than a relative pronoun, if we wish to maintain our generalization that pied piping was obligatory when a moved item appeared on the surface. While it is possible that &et was a relative pronoun, since it was homophonous with the neuter nominative and accusative singular demonstrative pronoun, it is equally possible that it was a complementizer, since Yxt was used as a complementizer in other situations, especially in indirect discourse and purpose clauses, as in Modern English. Various instances of this usage can be seen in the examples in this article, such as in (2) and (3).

Now let us consider whether all instances of &et as relative marker can be analyzed as pronouns. If this is so, we expect to find &et used only to replace relativized items which should be nominative or accusative neuter singulars. But consider the following examples:

(34) oa for he foro bi 6tem screfe &et he oninnan was. then went he forth by the-n.d.s. cave-n.d.s. that he within was

'Then he1 passed by the cave that he2 was in .'

CP Sweet p. 197.13 (35) Ond owt seolfe weter, &et heo da baan mid 6wogon, guton

and that same water-n.a.s. that they the bones with washed poured-out in enne ende 6are cirican. in one end the-gen. church

'And poured out the same water that they had washed the bones with at one end of the church.'

Bede 111. 11 p. 184.3 (36) Nu is se tima &t 6eos woruld is gemtencged mid

now is the time-m.n.s. that this world is confused with meanigfealdan mane. many evils

Wulf.V.24 (37) .. .hwtr hie landes hxfdon &et hie mehten an gewician

. where they lands-n.a.p. had that they might on camp 'where they had lands that they might camp on _

Oros. p.80.8

&et cannot be a pronoun in these examples. Consider the stranded preposition oninnan in (34). Concerning this example, Bacquet (1962) noted that if &et were a

19 This usage was quite restricted in OE. It was most frequently found in the following types of relatives: (i) those with a neuter head, (ii) those on temporal heads, and (iii) those with eall 'all' as their head.

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 16: Movement and Deletion in Old English

MOVEMENT AND DELETION IN OLD ENGLISH

pronoun, it should be in the dative case, since oninnan always took a dative object. There is no possibility of case attraction here, since the head of the relative is also in the dative case, while yet, if it is a pronoun, is either nominative or accusative. Similarly, in (35), Yxt should be in the dative case as the object of mid. In (36) and (37), there is a lack of agreement between the head and &wt. In (36), the head is masculine, but the pronoun 6zt can only be neuter. In (37), there is no disagreement in gender, but the head is plural, while the pronoun &et could only be singular. Examples like (36) and (37) are quite common.

While these examples cannot be explained if &et was a pronoun in these cases, since such lack of agreement in number, gender, and case is not found in other se- relatives, there is no problem if we treat &et as an undeclinable relative complementizer, a less common alternative to &e. Furthermore, if &et were a pronoun, we would have a problem accounting for its ability to strand prepositions within any theory, as far as I can see. We would be left with an uninteresting exception to the otherwise valid rule that se-relatives did not allow preposition stranding.

This is not to say that 3xt was never a relative pronoun. When its form is consistent with the head and the role of the relativized item, it is ambiguous between a pronoun and a complementizer.20 The important point is that it was not always a pronoun.

2.1.6. Infinitival Relatives Infinitival relatives in OE were quite similar to their modern counterparts:

(38) Ic hebbe mete to etenne done 6e ge nyton. I have food to eat that you do not know

Alc.P.V.72 (39) . . .cwxe6 &t he naefde 3xtfeoh him to alenenne

said that he not-had the money him to lend

'said that he did not have the money to lend him' Alc.Th.Vol.II p. 178.2

Unlike the Modern English infinitival relative, however, the OE type did not permit the use of a relative pronoun. There are no examples like a friend with whom to talk until Chaucer's period. This being the case, preposition stranding was obligatory in these OE relatives when the relativized item was the object of a preposition. Such examples are very numerous:

20 If &et was sometimes a true relative pronoun, we would expect some cases of pied piping with it, and indeed a few examples are found, as in (23). The scarcity of such examples is no doubt due to the fact that Yet was always neuter, and either nominative or accusative. Nominative case was never used with preposi- tions, and fewer prepositions took accusative than dative. Thus there was little opportunity for pied piping with Mwt.

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 17: Movement and Deletion in Old English

276 CYNTHIA ALLEN

(40) Gif &er 6onne sie gierd mid to Oreageanne, sie &er eac stef mid if there then is rod with to beat be there also staff with to wre6ianne. to support 'If there is a rod to beat with __, let there also be a staff to support with

CP Sweet p.126.1 (41) da gebohton hie enne ecer mid 6am feo tygel-wyrhtena on to

then bought they an acre with the money potters in to bebyrgeanne ael6eodige men. bury strange men 'Then they bought with the money a potters' field to bury strangers in .'

St. Mat. 1660 (42) Oeah he nu nanwuht elles nabbe ymbe to sorgienne

though he now nothing else not-have about to worry 'though he now have nothing else to worry about _

Boeth.XL.l p.24.15

Like the he-relatives, this type must be analyzed as involving either movement and then deletion, or controlled deletion alone.

2.2. Free Relatives

Now let us consider some types of "free" relatives. It will be seen that while some of these relatives were headless, other types had pronominal heads.

2.2.1. Pronominally Headed Relatives 2.2.1.1. Demonstrative Heads OE had a type of relative roughly equivalent to the modern type he who lives in a glass house. . . These relatives merit attention here because of their superficial similarity to, but fundamental difference from, the se 3e- relatives. The following are typical examples:

(43) 6et is, oaet man for-gife,&m &e wi6 hine gegylte that is that one forgive him-dat. that against him sins 'that is, that one2 forgive him, who sins against him2'

Ver.III. 170 (44) Micele mare miht ys menn to gescippene 6onne to araerenne5one

much greater ability is men to create than to raise him e xr wes.

that earlier was 'It is a much greater ability to create men, than to raise up him who existed earlier.'

Alc.P.VI. 122

It might be suggested that these relatives involve movement of the demonstrative

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 18: Movement and Deletion in Old English

MOVEMENT AND DELETION IN OLD ENGLISH

pronoun, as with the se be type. However, we saw that case attraction of the relative pronoun was optional in the se 3e-relatives. In these free relatives, on the other hand, the case marking of the pronoun is always that required by the upper clause. In (43), for example, the demonstrative pronoun is in the dative case as the object offor-gife, while its role in the subordinate ciause is that of subject, which would require nominative case. Thus, it is unlikely that the pronoun originates in the lower clause.

The most straightforward analysis of this type of relative is that it is simply a 5e- relative with a demonstrative pronoun head. This analysis accounts for the case marking of the pronouns, since the head of a relative naturally had the case marking of its role in the upper clause. By this analysis, we would predict preposition stranding in these relatives, and indeed such examples are numerous:

(45) foroan de we nabba6 5a & he on Orowade because that we not-have that-f.a.s.that he on suffered

'because we do not have that which he suffered on i

Alc.Th.Vol.II p.306.22 (46) . . .6axt he wolde mancynn ahreddan Ourh &one 3e he ealle

that he would mankind redeem through him-acc. that he all

gesceafta mid geworhte creatures with wrought

'that he would redeem mankind through him who he wrought all things with

Alc.Th.Vol.I p. 192.21 (47) & he tobryst done & he onuppan fyl1

and it pulverizes him-acc. that it upon falls

land it pulverizes him who it falls upon St.Mat. 1249

There are no examples of pied piping in these relatives. On the other hand, the pronoun may be the object of a preposition in the upper clause, as in (46), in which case it naturally has the case required by that preposition. These facts are predicted if the pronoun is actually the head of the relative. If there is any movement in these relatives, therefore, it must be ghost Wh Movement, as opposed to the movement of the dem- onstrative pronoun in se &-relatives. The other possibility is that these relatives involve unbounded deletion. Unbounded, because relativization could take place out of tensed clauses:

(48) Ne hafast Ou aenige mihte wi6 me butan 6a &e ic wille &et not hast thou any power against me except that-f.a.s. that I will that

ou habbe. thou have

'You do not have any power against me except that which I desire that you have.'

Ver.1.237

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 19: Movement and Deletion in Old English

278 CYNTHIA ALLEN

(49) For&em sceal &es modes lxce aer tilian&1?s be he wen6 therefore shall the spirit's physician first tend that-gen. that he thinks 6xet done mon er mage gebrengan on faerwyrde. that the man-acc. first may bring into perdition 'Therefore shall the physician of the spirit first tend that which he think (that)

may first bring the man into perdition.' CP Sweet p.457.10

(50) And he ahef6 hine sylfne ofer ealle ca be ha6ene men cwadon aet and he raises him self over all those that heathen men said that godas beon sceoldon. gods be should 'And he raises himself over all those that heathen men said (that) were gods.'

Wulf.N. p.197.16

2.2.1.2. Wh Heads The ordinary OE relative pronouns were demonstrative pronouns. However, the interrogative pronouns were used as indefinite pronouns, meaning 'any- one', 'someone', 'anything', etc. This fact made the interrogative pronouns natural candidates for use in indefinite free relatives. In this function, the wh-pronoun was both preceded and followed by swa, which meant 'so' or 'as'. Swa was also used as a determiner with quantifiers, as in Modern English so tall, as tall. The following are examples of the use of interrogative pronouns in free relatives:

(51) Sooes ic 6e sylle swa hwat swa 6u me byddest. truly I thee give so what-acc. as you me ask 'Truly I will give you whatever you ask of me.'

St.Mark 290 (52) Feder and moder moton heora bearn to swa hwylcum crwfte

father and mother must their child to so which-dat. occupation gedon swa him leofost by6. put as him most pleasing is 'Father and mother must put their child to whatever occupation is most pleasing to him.'

Alc.P.XIX.54 (53) 6onne mihte we micle 6y e6 ge6olian swa hwet

then might we much the easier endure so what-nom./acc. earfozSnessa swa us on become. hardships-g.p. as us on came 'Then might we much the more easily endure whatever hardships came on us.'

Boeth.X. p.23.12

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 20: Movement and Deletion in Old English

MOVEMENT AND DELETION IN OLD ENGLISH

The wh-pronouns in this construction have the case required by their roles in the main clauses. For example, in (51), if the pronoun took its case from the lower clause, we would expect genitive marking, because this is what the verb byddan required of the object requested (as in (59) below), and so on. These relatives appear to be exactly parallel to the demonstrative-headed type. The second swa appears to be a comple- mentizer like &e, rather than part of the pronoun, since it can appear separated from the pronoun, as in (52) and (53).21

If the wh-pronoun is the head in these relatives, it is no surprise to find preposition stranding in this construction:22

(54) Se oe raoost com on bone mere aefter 6es weteres styrunge he that quickest came into the lake after the-gen. water's stirring wearO gehxeled fram swa hwilcere untrumnysse swa he on was. was healed from so which-dat. infirmity as he in was 'He who came most quickly into the lake after the stirring of the water was healed from whatever infirmity he was in _ (i.e. from whatever disease he had).'

St.John 226 (55) And heo gefret softnysse o66e sarnysse, swa hwA?er swa heo on bid.

and it feels softness or pain so which-acc. as it in is 'And it feels softness or pain, whichever it is in .'

Alc.P.XI.218

No examples of pied piping are to be found. Therefore, we are led to conclude that the wh-pronouns in this construction are the heads of the relatives. Bresnan and Grim- shaw (1978) have argued convincingly for the same analysis of the descendants of these relatives in Modem English, and furthermore, that such relatives must involve controlled unbounded deletion.

2.2.2. Free Relatives with Movement There are many examples of free OE relatives which at first glance seem to contradict the claims just made about pronominally-headed relatives. Some of these examples involve wh-pronouns, and some, demonstratives. It will be demonstrated, however, that there is strong evidence for analyzing these appar- ent counterexamples as being structurally quite different from the types just discussed.

2.2.2. 1. Free Relatives with Wh Movement In section 2.2.1.2 I stated that no examples

21 The first swa (probably some sort of determiner) apparently selected for the complementizer. This is similar to the restriction in Modern English that relatives with such in the head take as:

(i) Such snakes {a we saw on Black Mountain were uninteresting. 22 It should be noted that examples like these are rare. I believe that this is because relatives with wh-

heads were themselves much rarer than those with demonstrative heads. These few examples, however, are significant when combined with the fact that no examples of pied piping are found in this type of relative. Thus there is reason to believe that preposition stranding was obligatory in this construction.

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 21: Movement and Deletion in Old English

280 CYNTHIA ALLEN

of pied piping were to be found in wh-headed relatives. But apparent counterexamples to this claim are numerous:

(56) And to swa hwilcere leode swa we cumao, we cunnon 65ere gereord. and to so which-dat. people-dat. as we come we know their language And whatever people we come to, we know their language.'

Alc.Th. Vol.11 p.474.2 (57) Ond 3urh swa hwelces bene swa he gehaled sy, 6isses geleafa

and through so which-gen. prayer as he healed is his belief & wyrcnis seo lefed God onfegne. and works be believed God acceptable 'And whoever's1 prayer he2 is healed through, let his, belief and works be believed acceptable to God.'

Bede 11.2. p.98.3 1 (58) Ond on swa hwelcre stowe swa min orowunge awriten sy ond man

and in so which-dat. place as my passion written is and one da mersige, afyrr ou, drihten from &ere stowe blindness. it celebrates drive you Lord from that place blindness 'And whatever place my passion is written in and is celebrated, drive, 0 Lord, blindness from that place.'

Mart. p. 1 16.8

However, closer examination reveals that in these relatives, the wh-pronoun is not the head. Rather, it is moved to its surface position. Notice that in (56) through (58), the relative clauses are "left-dislocated" to the front of the sentence. Furthermore, note the resumptive pronoun in the main clauses (in the case of (58), a full NP, rather than a pronoun). We do not find left-dislocation with resumptive pronouns in (51) through (55). Rather, these relatives are "in place" in their main clauses.23 It turns out that in every single case of pied piping in indefinite wh-relatives, we find this left-dislocation with resumptive pronoun. Furthermore, we find that just when left-dislocation with resumptive pronoun is present, the case of the wh-pronoun is that dictated by the lower clause, as in the following examples:

(59) And swa hwces swa hie rihtlice bidda6 for 6inum naman & and so what-gen. as they rightly ask for thy name and for oinum gearningum hig hyt onfo6. for thy merit they it receive 'And whatever they ask rightly, for your name and your merit, they receive it.'

30E p.74.4 23 Naturally, the pronominally-headed relatives had a certain amount of freedom as to where they

occurred in the sentence, as did all NPs. For example, they might be topicalized. The topicalized relatives differ from the left-dislocated ones in that the former have no returning pronouns. The case of the pronoun head is that of its role in the upper clause, as we would expect.

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 22: Movement and Deletion in Old English

MOVEMENT AND DELETION IN OLD ENGLISH

(60) Foroon oe Drihten cwx6 to him oaet swa hwylcne swa he on eor6an because that Lord said to him that so which-acc. as he on earth gebunde, oet se wore on heofonum gebunden. bound that he-nom. was in heaven bound 'Because the Lord said to him that whoever he bound on earth, he would be bound in heaven.'

Blickling p.49. 15

On the other hand, when dislocation and the resumptive pronoun are lacking, the case marking is dictated by the upper clause, as in (51) through (55).

Both the case marking and pied piping facts are easily explained if we assume that the left dislocated type involves movement of the wh-pronoun. To make things clear, I present here the proposed derivation for (56). The deep structure is (61):

(61)

NP S

Comp NP VP

swa NP VP we V NP

we V PP cunnon 6aere gereord

cuma6 P NP

to swa hwilcere leode

For reasons which need not concern us here, I assume that the "idislocated" clause is generated in its surface position, rather than being moved there transformationally. It is also not crucial here whether the dislocated clause is dominated by NP and the main clause by another S.

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 23: Movement and Deletion in Old English

282 CYNTHIA ALLEN

The surface structure is derived by the application of Relative Movement, the exact formulation of which is unimportant here:

(62) S

NP S

NP VP

PP S we V NP

P NP C p S cunnon oere gereord

to swa hwilcere swa NP VP leode

we V

cumao

2.2.2.2. Free Relatives with Demonstrative Movement As we might expect there is a type of demonstrative free relative parallel to the wh-relatives just discussed, with the pronoun in the case required by the lower clause and the relative left-dislocated with a resumptive pronoun:

(63) And one fr 6u nu haefst, nis se 6in wer. and him-acc. that thou now hast not-is he-nom. your husband 'And him who you now have, he is not your husband.'

Alc.P.V.37 (64) &5e his leasungum gelyfao, am he arao.

those-nom. that his lies believe those-dat. he honors 'Those that believe his lies, them he honors.'

Alc.Th.Vol.I p.6.5

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 24: Movement and Deletion in Old English

MOVEMENT AND DELETION IN OLD ENGLISH

It seems clear that these relatives are to be analyzed in the same way as the free relatives with Wh Movement.24

To conclude our discussion of free relatives, we see that they pose no problem for the generalization that preposition stranding was not possible in non-ghost movement rules. One final type of OE relative, the one with a locative pronoun, remains, but we will defer discussion of it until section 2.6.

2.3. Complement Object Deletion

A construction closely related to the infinitival relative is that exemplified by Modern English Mary is pretty to look at. Lasnik and Fiengo (1974) have argued that the Modern English construction is formed by a rule of Complement Object Deletion, which removes the object in the adjectival infinitival clause under identity to a noun in the main clause. I will refer to this construction as CO deletion in spite of the fact that it might actually involve ghost Wh Movement or an interpretive rule.

As we would expect, preposition stranding was possible in this construction in OE :25

(65) Waos seo wunung &aer swy6e wynsum on to wicenne. was the dwelling there very pleasant in to camp 'The dwelling there was very pleasant to camp in _.'

Alc.S.XXX.3 15

24 This being so, we would expect to find pied piping in these relatives. However, I have only found one such example:

(i) Ofer 3one &- ou gesyhst nyoer stigende gast and ofer hyne wuniende, oet ys se (e over him-acc. that you see descending spirit and over him remaining that is he that fullao on halgum gaste. baptizes in Holy Ghost 'The one that you see a spirit descending over and remaining on him, that is the one who baptizes with the Holy Ghost.'

St.John 41 This example cannot be taken too seriously, because it is a literal translation from Latin (super quem

videris spiritum descendentum.. . ). However, I do not believe that the lack of convincing examples of pied piping indicates that it was impossible in this construction, because this type of relative was quite rare with a relativized item anything other than the subject of the relative. When the target was an object in the lower clause, it was much more common to use a demonstrative-headed relative. Also, there are no examples of preposition stranding in this relative, so no counterexamples to our claim.

25 Interestingly, although examples of preposition stranding in this construction with adjectives like wynsum, wundorlic, etc., are common, I have found no examples of stranding with ea&e 'easy' or earfo&? difficult', as in Mary is easy to talk to. Sentences of this sort have traditionally been analyzed as being

derived by Tough Movement" from it is easy to talk to Mary, and so forth. Lasnik and Fiengo, however, suggest that "Tough Movement" is just another case of CO Deletion, and there is no movement. The fact that preposition stranding is not found in this construction in OE, although examples of the construction without stranding are common enough, suggests that the traditional view may be correct, at least for OE. If movement of an NP, rather than deletion, were involved with adjectives like earfo&-, the lack of stranding would follow, assuming a prohibition against movement out of PP. The fact that pied piping is also not found in this construction may be explained by the fact that Tough Movement applied only to NP.

I am indebted to Joan Maling for pointing out the lack of stranding in this construction to me.

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 25: Movement and Deletion in Old English

284 CYNTHIA ALLEN

(66) Seo menniscnes is wundorlic ymbe to smeagenne. the incarnation is wonderful about to think

'The incarnation is wonderful to think about Wulf.VI. 142

(67) Heo waes swi6e egeslic on to beseonne. it was very terrible on to look

it was very terrible to look on .'

Alc.P.XXI.536

As in Modern English, no pronoun was possible in CO deletion, so pied piping was not possible. This construction is presumably to be analyzed along the lines of the Ae- relatives and infinitival relatives.

2.4. Questions

Wh-questions in OE were basically the same as in Modern English, with two major differences. First, do was not used as an empty tense and number carrier; any verb could invert with the subject. Second, pied piping was obligatory in questions. I have found no counterexamples to this rule, and neither did Wende (1915). Visser (1963, 406), however, claims:

When an interrogative sentence or dependent clause opens with whom or what the prepo- sition has end-position. The putting the preposition before these pronouns has always been less usual.

This statement is simply false. The earliest examples of preposition stranding in questions I have seen are from the thirteenth century, and there are only two examples even then, as opposed to many examples with pied piping. It is significant that the earliest example Visser gives to illustrate his point is from Shakespeare. In another place, he gives one of the two thirteenth-century examples alluded to above. As Visser is usually careful to give examples from all periods, his failure to give even one example of this construction from before the thirteenth century is suspicious. In light of his failure to provide any evidence for his claim and the explicit statement of Wende, who made a particular study of prepositions in OE, that he found no examples of preposition stranding in OE questions, Visser's claim must be rejected.26

Although pied piping in indirect questions is stilted in Modern English, we find pied

26 We see, therefore, that the assertion of Chomsky and Lasnik in footnote 122 that "Visser also cites data in support of this conclusion (p. 401)" is misleading, as these data are not from Old or Early Middle English.

I would like to state here that I do not wish to detract from the general value of Visser's excellent history. It is perhaps inevitable that a few errors would creep in with a project of such scope, and such mistakes are very rare in Visser's book. The fact that even the best of histories of English contain some factual errors makes it important for one to do one's own checking on the facts, whenever possible.

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 26: Movement and Deletion in Old English

MOVEMENT AND DELETION IN OLD ENGLISH

piping in both direct and indirect questions in OE:

(68) To hwwm locige ic buton to 6om ea6modum? to whom look I but to the humble

'To whom shall I look but to the humble?' CP Sweet p.299.19

(69) And gif ic on his naman adrafe deofla of mannum, on hwes naman and if I in his name drive-out devils from men in whose name

adrafao eowre suna 6onne? drive-out your sons then

'And if I drive devils out in his name, in whose name do your sons drive them out then?'

Alc.P.IV.25 (70) Ic nat ful geare ymb hwat 6u giet tweost.

I not-know full well about what thou yet doubtest 'I do not know full well about what you still doubt.'

Boeth.V.3 p. 12.26 (71) For&em sio halige gesomnung ourh gesceadwisnesse gesiho &

because the holy assembly through sagacity sees and

ongieta6 of hwwm xelc costung cyme6. understands from whom each temptation comes

CP Sweet p.64.24

There are numerous apparent violations of the Tensed-S, Specified Subject, and Subjacency Conditions:

(72) Hwelc wite wene we 6at se felaspraecea scyle habban? what punishment think we that the loquacious shall have

'What punishment do we think that the loquacious shall have?' CP Sweet p.281.14

(73) And gehyrst mid hwilcum crlefte he is ge6uht 6aet he untrumnysse gehale. and hearest with which skill he is thought that he illness heals

'And you will hear with what skill he is thought to heal illness.' Alc.Th.Vol.I p.460.23

(74) and. . . hi ealle befran, hwwt hi wendon 6at he wore and them all asked what they thought that he was land asked all of them what they thought that he was'

Angl.Hom.XIII. 1 15

It should be noted here that we find no examples in the OE texts of questions on infinitival phrases such as I don't know what to do.27 The first examples of this sort are

27 This fact was also noted by Visser (1963, 976).

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 27: Movement and Deletion in Old English

286 CYNTHIA ALLEN

from the thirteenth century, and they do not become common until the early fourteenth century. Thus, the question of pied piping in infinitival questions does not arise.

2.5. Topicalization

Now let us consider the behavior of prepositions in another construction involving movement, namely topicalization.28 Topicalization was quite common in Old English. The following are a few typical examples, including ones in which the topicalized item has been removed from a tensed subordinate clause, giving apparent violations of the Tensed-S, Specified Subject, and Subjacency Conditions:29

(75) Micele ing abaedon ba meran apostolas zet 6am halgan Faeder great things asked the great apostles at the holy Father aefter oaws Hxlendes upstige. after the Savior's ascension 'Great things the great apostles asked of the holy Father after the Savior's ascension.'

Alc.P.VIII.73 (76) Twa rng ic oe gehet oact ic 6e wolde gelestan.

two things I thee promised that I thee would perform 'Two things I promised you that I would perform for you.'

Sol. p.70.9 (77) To &iem sodum geswl3um ic tiohige 6but ic 6e lade.

to the true happiness-dat. I intend that I thee lead 'To the true happiness I intend to lead you.'

Boeth.XXII.2 p.51.12

There are many examples which at first glance suggest that prepositions could be freely stranded in this construction:

(78) And me com oar-rihte to godes encgel mid rode. and me came directly to God's angel with cross 'And to me one of God's angels came directly with a cross.'

Alc.S.VII.356 (79) & him 6a si66an se feondscipe wxs betweonum weaxende.

and them then afterwards the enmity was between growing 'And between them afterwards the enmity was growing.'

Oros. p.232.26

28 I will assume that topicalization moves the topic to the front of S. However, Chomsky (1977) has proposed that Modern English topicalization involves Wh Movement, with the topic generated in its surface position. This analysis will be considered in section 3.

29 The fact that items could be topicalized out of a tensed clause shows that topicalization was not just a case of scrambling", as OE did not permit scrambling out of tensed clauses.

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 28: Movement and Deletion in Old English

MOVEMENT AND DELETION IN OLD ENGLISH

(80) And him man gebrohte Oa to fela bedridan menn. and him one brought then to many bedridden men 'And to him people then brought many bedridden men.'

Alc.P.XVII. 13

Notice that in each of these examples, the topic is a pronoun, rather than a full NP. Examples of preposition stranding with topicalized pronouns are extremely common in OE literature, while preposition stranding with topicalized full noun phrases is extremely rare.30 To understand why pronouns allowed preposition stranding in this construction, it is necessary to take a closer look at the behavior of pronominal objects of prepositions in simple OE sentences.

In OE, pronominal objects of prepositions could generally invert with their prep- ositions:

(81) And hi ne dorston him fore gebiddan. and they not dared him for pray 'And they dared not pray for him.'

Alc.P.XIX.226 (82) gif ic eowfram ne fare

if I you from not go 'if I do not go from you'

Alc.P.VII.46 (83) ... and cwa6 him to

and said him to 'and said to him'

Alc. S.II .235

This optional process of inversion was quite common with pronominal objects of prepositions, although it was extremely rare with full NPs, at least in prose.31

In the above examples, the inverted preposition and object are still together and might form a constituent. However, it was also quite common for a constituent or group of constituents to intervene between an inverted pronoun and the preposition of which it was the object:

(84) 6a wendon hi me heora bec to. then turned they me their backs to 'Then they turned their backs to me.'

Boeth.II p.8.11

30 Wende found only five examples outside of Gregory's Dialogues, all of which he felt to be somewhat dubious for various reasons.

31 Wende found only two cases of inverted prepositions with full NP objects in the prose texts of his study. Inversion with full NPs was much more common in poetry, especially with polysyllabic nouns.

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 29: Movement and Deletion in Old English

288 CYNTHIA ALLEN

(85) oo heora wif him sendon xrendracan cefter until their wives them sent messengers after 'until their wives sent messengers after them'

Oros. p.44.20 (86) 06 6is ic sprac 6e lioelice to.

until this I spoke thee meekly to 'Until this I spoke to thee meekly.'

Alc.Th.Vol.I p.594.8 (87) 6a for 6are ceorunge sende him God to byrnende naddren.

then for the murmuring sent them God to burning adders 'Then God sent burning adders to them for the murmuring.'

Alc.P.XX.3 14

Nothing ever intervened between a preposition and its object when the object followed the preposition (that is, when inversion did not take place). The inverted PP, then, behaved as a nonconstituent with respect to whatever "scrambling" process gave OE its fairly free word order.32 Space does not allow me to present examples of all the different positions of pronoun and preposition in the inverted form and the constituents which could intervene between them, but as far as I can determine, there is no regular position for the pronoun or the preposition after inversion, nor is there any rule about what may intervene between the two.33 The only restriction I have found is that when the pronoun and its preposition are separated, the preposition cannot precede the pronoun in the sentence. There are no examples of to. . .him, etc. Since I can find no specific position for the pronoun or the preposition, or any specific rule moving con- stituents between them, I will assume that the surface orders with constituents inter- vening between the two are the result of scrambling.

Scrambling in OE did not generally permit the breaking up of prepositional phrases or simple noun phrases.34 It seems clear that there was an intimate link between the ability of pronominal objects of prepositions to scramble freely and the ability of

32 The freedom of OE word order should not be exaggerated, but it was much less restricted than that of Modern English. For details, see Bacquet (1962), Gardner (1971), Pillsbury (1967), Shores (1971), Sprockel (1973), and Wulfing (1901), among others.

33 Van Riemsdijk (1978, 290) claims that:

. . if we discard the effects of topicalization, the (OE version of) the verb second rule, and the greater mobility of the subject NP, the surface position in all of Allen's [1977/CLA] examples. . . is the same as in Dutch: to the right of the subject. But the fact is that any object NP can appear in the position where these pronouns appear, and split

prepositional phrases occur after verbs which are not in second position, as in (86) and (87), so there is no evidence for a specific position of inverted pronouns.

34 However, it was possible to extrapose whole genitive noun phrases, as in: (i) 6et 6u hlaford beo oara aehta and min

that you lord be the-gen. treasure-gen. and my that you be lord of the treasure and of me'

Alc.S.II. 159

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 30: Movement and Deletion in Old English

MOVEMENT AND DELETION IN OLD ENGLISH

pronouns to invert with the prepositions of which they were the objects. The most straightforward explanation seems to me to be that the inversion of a preposition and its pronominal object resulted in a nonconstituent. Either the inversion process itself broke up the PP, or else there were two processes: an optional inversion of prepositions and their pronominal objects, and an optional rule making inverted prepositional phrases into nonconstituents.

A couple of considerations argue against viewing the inversion and breaking up as one process. Notice that if a prepositional phrase were first topicalized and then inverted and broken up, we would have to allow the inverted preposition to scramble back into the sentence to form sentences like (78) through (80). Little is known about the nature of scrambling, but it seems likely that scrambling rules should not be permitted to affect items which have been moved to the front of the sentence (or into Comp) by specific movement rules. Support for such a restriction comes from the fact that relativized and questioned items do not scramble back into the sentence in OE.35 On the other hand, if inversion and breaking up occur together before topicalization, we cannot account for such examples as (88), in which the inverted PP is topicalized as a unit:

(88) Hym to genealkhton his leorningcnyhtas. him to neared his disciples 'His disciples came near to him.'

St.Mat.810

Second, we will see in section 3.2 that in ME, there was still inversion of prepo- sitions and their pronominal objects, but no more splitting of them. This suggests that inversion and splitting were two processes in OE. I will assume, therefore, that there were two rules involved. The first, which I will call PP Inversion, optionally inverted a preposition and its pronominal object:

(89) PP Inversion

W1 pp [P NP ] W2 [+pro] l +persj

1 2 3 4 #> 1 3 2 4

An alternative way of formulating this rule would be to postulate that the pronoun moved leftward, leaving a trace:

35 This would appear to be a universal restriction on scrambling. For example, Avery Andrews informs me that in Ancient Greek, NPs which had been raised or passivized could be further scrambled, but relativized and questioned ones could not. It seems probable that there is a general restriction that constituents which have been moved to the front of S (or into Comp) cannot be scrambled. The reason behind this is presumably that scrambling would undo the effects of rules which move things to particular positions. On the other hand, because of case marking, scrambling would not remove the evidence for Passivization and Raising.

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 31: Movement and Deletion in Old English

290 CYNTHIA ALLEN

(90) PP Inversion (one trace theory version)36

W, pA[P NP ] W2 E +pro] I+pers j

1 2 3 4 4 1 3#pp[2 t 4

A third possibility is that the preposition moved rightward. Which particular for- mulation of the rule is correct is not crucial here. The consequences of the various formulations within the surface filter framework will be discussed in section 3.

The second rule readjusts the constituent structure of an inverted PP, so that (9Ia) becomes (91b):

(91) a. pp[NP P] b. NP PP[P]

This rule, which merely involves a change in bracketing, cannot be formulated as a transformation, unless as one involving the movement of the NP out of the PP, for which there is no evidence. It is possible that such an extraposition of NP is involved, but it seems more likely that this was a readjustment rule.37 I will refer to this read- justment as PP Split.

If we assume that both PP Inversion and PP Split preceded topicalization, we can account for the fact that sometimes the inverted PP was topicalized as a unit, while at other times only the pronominal object was topicalized, stranding the preposition. Because PP Split was optional, the inverted PP could either be a constituent or not at the time of topicalization.

By this analysis, the facts about preposition stranding in topicalization follow directly from the fact that only pronouns generally underwent PP Inversion (and could therefore undergo PP Split).38 Assuming a general prohibition against movement out of PP,39 preposition stranding would only be possible in topicalization when the topic was

36 Such a trace theory version of this rule is proposed for Dutch by Van Riemsdijk (1976). Van Riemsdijk's formulation Chomsky-adjoins the NP to PP, but it could be sister-adjoined to the P.

37 A similar rule to readjust the bracketing of certain NPs is proposed by Chomsky (1977). This rule breaks up a picture of John, for example, into an NP and a PP. Chomsky notes that this could be an extraposition rule, but is probably a readjustment rule.

38 Wende (1915) found Gregory's Dialogues to be exceptional in containing several examples of topical- ized full noun phrases with preposition stranding. However, this text also contains an unusual number of inverted prepositions with full NP objects. It seems clear that the greater freedom of stranding in this text is due directly to the fewer restrictions on PP Inversion, which is what our account would predict.

The very few examples of preposition stranding in topicalization in the other texts (if they are true examples) can be accounted for when we consider that in these texts we find a very few examples of inverted PP with full NP objects. The fact that preposition stranding in topicalization in a given text is proportional to the frequency of inversion in simple sentences supports our conclusion that preposition stranding was not allowed by topicalization per se, but was the result of PP Inversion.

39 It is possible that some rules must be formulated as specifically removing NP from PP. For example, it may be that PP Split should be formulated as a leftward extraposition of NP from PP, in which case it would violate an absolute prohibition against movement out of PP. In this event, it may be that the restriction

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 32: Movement and Deletion in Old English

MOVEMENT AND DELETION IN OLD ENGLISH

no longer a constituent with the preposition of which it was the object. Whatever the precise formulation of the rules involved, it is clear that preposition stranding with pronouns in topicalization was a side effect of the inversion of PP, rather than a feature of topicalization itself.

PP Inversion (89) has been formulated to apply only to personal pronouns. This is because demonstrative and interrogative pronouns were not subject to this permuta- tion.40 It follows, then, by our analysis, that no preposition stranding is found in se- and se &e-relatives or in questions, since, without inversion, there was no splitting.

2.6. 6a5r-Relatives

The final construction to be considered is the locative relative using 3er. &er was a demonstrative locative pronoun, and as such was used in locative relative clauses instead of its interrogative counterpart hwaer:

(92) oat wes sio stow, Yr man nytenum hira andlifan sealde. that was the place there one beasts-dat. their food gave 'That was the place where one gave beasts their food.'

Ver.V. 157 (93) And after 6issum se haliga Andreas het cyrican getimbrian on &ere

and after this the holy Andrew ordered church build on the stowe &er se swer stod. place where the column had stood 'And after this the holy Andrew ordered a church to be built on the place where the column had stood.'

Blickling p.247.27

When &er was the object of a preposition, it was possible for it to be relativized, stranding the preposition.

should be on "incidental" movement out of PP. That is, a rule may be specifically formulated to move NP out of PP, but NP may not be moved out of PP by a rule whose specific function is not to remove NP from PP, but to move NP to the front of S, etc., and for which PP is not an obligatory context.

40 Wende found only four examples of demonstrative pronouns preceding the prepositions of which they were the objects, as opposed to 484 instances of inverted PP with personal pronouns (including those with splitting). It is not surprising that a few such examples should appear with demonstrative pronouns, since these were sometimes used as personal pronouns. The important fact here is that in none of these four examples is the demonstrative used in a relative function.

Wende found no examples of inversion with (nonlocative) interrogative pronouns. I have found one: (i) .. .ac Oonne hi hwawmfro hweorfende bio6

but when they whom from departing are 'but when they are departing from anyone'

Boeth.VII.2 p. 16.14 Since the pronoun is used here in an indefinite capacity, it cannot be considered a true example of

inversion with an interrogative pronoun.

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 33: Movement and Deletion in Old English

292 CYNTHIA ALLEN

(94) And se stede aefre sy66an wes temtig, Yxr heo wr on stod. and the place ever after was empty where it earlier on stood 'And the place where it had earlier stood (on) was empty ever after.'

Angl.Hom.V.86 (95) ... .6et se dael (axre ciricean ne meg habban 6one hrof er

that the part the-gen. church-gen. not may have the roof where 6tes haelendes fotlastas sindon under the Savior's footprints are under 'that the part of the church where the Savior's footprints are under may not have a roof

Mart. p.76.11 (96) and on o6erre w2s an gewrit, &er wTron on awriten ealra 6ara

and in another was a letter where were in written all the ricestena monna noman richest men's names

'and in another was a letter where all the richest men's names were written (in)'

Oros. p.258.14

On a superficial examination, &er-relatives appear to present counterexamples to the claim that preposition stranding was not possible in movement rules. However, a closer examination reveals the preposition stranding in these relatives to follow directly from the behavior of locative pronouns in simple sentences. It is also important to note that, if it were a counterexample to this claim, preposition stranding in &er-relatives would be a systematic deviation from the rule, which must be accounted for by any analysis. It is not enough to show that one group of pronouns is an exception to the generalization, since the lack of stranding in other se-relatives, and hence the asymmetry between be- and ordinary se-relatives, remains.

Let us now see how locative pronouns behaved in simple sentences. Like personal pronouns, OE locative pronouns could invert with prepositions of which they were the objects:

(97) ealle Oe &erbinnan4l waron all that there within were 'all that were within that place'

Oros. p.200.16 (98) ... .6ot crestene menn &erto faran magan

that Christian men thereto go may 'that Christian men may go (to) there'

Wulf.XVIII.35 41 Although binnan was historically a compound on innan, which could be either a preposition or an

adverb, binnan is listed in the Bosworth and Toller dictionary as being only a preposition.

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 34: Movement and Deletion in Old English

MOVEMENT AND DELETION IN OLD ENGLISH

(99) Awyrtwala graedignysse of 6inre heortan, and aplanta 3eron root-up greediness from thy heart and plant therein 6a so6an lufe. the true love

Alc.Th.Vol.II p.410.1

I propose a rule of Locative Inversion, similar to PP Inversion, to permute a preposition and its demonstrative locative pronominal object:

(100) Locative Inversion

W1 PAP NP ] W2 +pro -wh +loc

2 3 4 = 3 2 4

This rule could be collapsed with PP Inversion, but I will treat them as separate rules here. The feature [-wh] is included because locative interrogative pronouns, unlike their demonstrative counterparts, did not undergo Locative Inversion in OE. Whereto, wherefore, etc., were not possible until the thirteenth century.42

After Locative Inversion, the inverted prepositional phrases can be broken up by PP Split and separate in simple sentences:

(101) . . . 6t Ercol se ent &er was to gefaren that Hercules the giant there was to gone

'that Hercules the giant had gone there' Oros. p. 132. 10

42 Wende found only one example of inversion with hwa-r: (i) Hwer to beoo 6as geendebyrde?

'where to are these annexed 'What are these annexed to?'

Alc.Th.Vol.1 p.344.29 Hwer does not have locative meaning here. It appears that Locative Inversion has taken place here

because Locative Replacement, discussed below, has exceptionally applied. The only other possible example of inversion with hwwr I know of is the following:

(ii) 6at tacn nugyt is orgyte on 6es sxs stae, hwter &era wigwaegna hweol on the marks still are perceived on the sea's shore where the chariots' wheels on gongende wwron. going were 'the marks are still perceived on the sea shore on which the chariots' wheels were passing.'

Oros. p.38.34 Here PP Split has apparently also applied. But I do not find this example convincing, because it is quite

possible that ongangan was a compound verb meaning 'to enter', since there was a noun ongang 'entrance'. Supposing these two examples to be true examples of Locative Inversion, they still cannot be considered

serious evidence that Locative Inversion applied to wh-pronouns. When only two examples from more than two centuries of texts are considered proof of the grammaticality of a construction, historical linguistics is on shaky ground indeed.

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 35: Movement and Deletion in Old English

294 CYNTHIA ALLEN

(102) . . .&et hie Yer mehten betst fri6 binnan habban that they there might best security within have

'that they might have the best security in there' Oros. p.116.6

Given these facts, it is no surprise to find preposition stranding in topicalization with these pronouns:

(103) . . .&et &er was butan seo swa6u on that there was but the mark on

'that only the mark was (on) there' Mart. p.102.22

(104) & &er 6onne befeolle on o66e oxa oMe esol and there then fell in either ox or ass 'and either an ox or an ass fell in there'

CP Sweet p.459.24

If Locative Inversion and PP Split apply before Relative Movement, there is no problem accounting for preposition stranding in locative relatives, since &er and its preposition need not be a constituent.43

43 Maling (1978) argues that stranding was obligatory in the &Tr-relatives. Vat (1978), however, disagrees with this conclusion, although he has no OE examples to back up his belief. His conclusion that pied piping was possible in &er-relatives is based on two things. First, E. Anklam (1908) "quotes examples of &Tr- relatives with both stranded and pied-piped prepositions (postpositions) in the period between the years 1000 and 1200." But in fact the only two such examples given by Anklam are from the early thirteenth century, or possibly the very end of the twelfth.

Second, Vat claims that the absence of pied piping examples in the syntactic studies of the period before 1000 proves nothing, because "with the sole exception of Wende (1915), &er-relatives have never been distinguished as such." He adds, "We consider that culling the older literature-this time on &Tr-relatives- might prove worthwhile." Vat ignores the fact that the older texts have indeed been culled for 3er-relatives by Allen (1977), where he could have found one of the examples he was looking for. I have in fact found exactly three possible examples of pied piping with &Tr-relatives in OE:

(i) . . .he was on Simones huse 6ds lic roweres, 6erin geat 6aet wif 6a deorwyr6an he was in Simon's house the leper's wherein poured the woman the precious

smerenesse on his heafod. ointment on his head 'He was in the house of Simon the leper, wherein the woman poured the precious ointment on his head.'

Blickling p.73.2 (ii) He lufode forhaefednysse on mettum and on drence, and wwccan on syndrigum

he loved abstinence in food and in drink and vigils in solitary gebedum; 3xr-to-eacan he 6rowade singallice untrumnysse. prayers where besides (i.e. besides which) he suffered incessant infirmities

Alc.Th.Vol.II p.120.5 (iii) ... into anre byrig, &Tr binnan wos swi6e smeoe feld and brad

into a city wherein was (a) very smooth field and (a) broad Alc.Th.Vol.1I p.352

The first of these is from the tenth century, and the last two are from the late tenth or early eleventh. All of these examples are ambiguous, since &er could be a demonstrative pronoun in each case. Supposing

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 36: Movement and Deletion in Old English

MOVEMENT AND DELETION IN OLD ENGLISH

There are two more rules involving locative pronouns which need to be mentioned. First, sometimes locative pronouns which participated in Locative Inversion had not a locative meaning, but rather that of a nonanimate demonstrative pronoun:

(105) Nu wylle we eow geopenian 6xt andgit 3xrto. now will we you open the meaning thereto 'Now we will open the meaning to that for you.'

Alc.P.XIII.35 (106) oat hig to lyt &erymbe oenceao

that they too little thereabout think 'that they think too little about that'

Wulf.N. p.273.1

To account for this fact, we can postulate a rule changing an inanimate demonstrative pronoun into a locative pronoun.

(107) Locative Replacement

W1 Pp[P NP ] W2 +pro 1 +dem -anim

1 2 3 4 = 1 2 [+loc] 4

Such rules are common in the Germanic languages. Note that this rule applies only to inanimates. In fact, we never find &er referring to an animate, nor is &er used as a relative pronoun with animate heads.

It is possible that Locative Replacement should be formulated to effect not only the replacement, but also the locative inversion, in one rule.

To see that Locative Replacement was optional in OE, compare the two parts of (108):

(108) ... .6at hi waere beon 6bs cwydes 65e &erefter that they attentive be the-gen. sentence-gen. that thereafter

gecweden is. He cwx6, se witega, efter dam:... said is he said the prophet after that 'that they be attentive to the sentence which is said after that. He said, the prophet, after that:. .

Wulf.VI. 16

these to be true instances, the paucity of examples still leaves us in doubt as to their significance. I will assume that it was an idiosyncrasy of PP Split that it was obligatory, or nearly so, with relative cxr.

Unable to supply the desired examples from OE, Vat gives examples of pied piping with &er from the early thirteenth century. The dangers of such a practice are obvious, since English syntax had changed considerably by this time. Indeed, coming after the nearly total lack of examples in OE, the sudden flood of examples of Locative Inversion with hwer and pied piping with Locative Inversion in all early thirteenth- century texts strongly indicates that a change has occurred.

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 37: Movement and Deletion in Old English

296 CYNTHIA ALLEN

Like true locative pronouns, these locatives created by Locative Replacement were available to PP Split; we also find preposition stranding in relative clauses using &er which were underlyingly nonlocative, along with cases of &er. . .P in simple sentences:

(109) Be 6aem 6u meaht ongietan 6bt 6u &er nane myrh6e on naefdest. by that you may understand that you there no joy in not-had 'By that, you may know that you had no joy in that.'

Boeth.VII. 1 p. 15.11 (110) ... .for6i Oat he wolde us to his rice gebringan &er we to

because that he would us to his kingdom bring where we to gesceapene waeron created were 'because he would bring us to his kingdom, to (i.e. for) which we were created'

Alc.Th.Vol.II p.6.26

The final rule we need to consider concerning locative pronouns is illustrated in the following examples:

(Ill) Was HTesten 6a &er cumen mid his herge. was Haesten then there come with his army 'Then Haesten came there with his army.'

Sax. Chron.894 (112) gif&ar man an ban findeo unforberned

if there one one bone finds unburned 'if one finds one unburned bone there'

Oros. p.21.12

In these examples, the locative pronoun is understood as the object of a preposition, although there is no preposition on the surface. In (111) the understood preposition is to, while in (112) it is aet. The combinations at 3er, to &er are not found in OE.

These facts are similar to Modern English data concerning locative pronominal objects of certain prepositions. For example, we can say John went to that place, but not John went to there. To account for these facts, and similar ones, Katz and Postal (1964) proposed a rule deleting a preposition immediately preceding its pronominal locative object. I will adopt this rule of Preposition Deletion for OE also."

OE Preposition Deletion must be allowed to apply after Locative Inversion, or else all instances of &er to and &er tt would be wrongly ruled out.

44 Perhaps a better way of dealing with these facts, and the PP Inversion facts, would be to generate inverted prepositional phrases in the base, with subcategorization determining which pronouns occur before their prepositions. Some sort of rule of PP Split, however, is still needed, and none of the arguments in this article would be affected by a change to this approach.

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 38: Movement and Deletion in Old English

MOVEMENT AND DELETION IN OLD ENGLISH

To summarize, we see that preposition stranding in acr-relatives presents no prob- lem for the hypothesis that pied piping was obligatory in movement rules in OE. Our analysis predicts that we find stranding in der-relatives, but not hwer-questions, since hwa-r did not undergo Locative Inversion, a fact completely independent of our analysis of relatives.

Another analysis of OE locative relatives has been proposed by Vat (1978). By Vat's analysis, inverted locative pronouns could be moved without their prepositions because Locative Inversion moved the pronoun into an "escape hatch" within PP. Vat attempts to extend this analysis to account for preposition stranding in &e-relatives. I will discuss this analysis in section 4.

2.7. Summarv of the Facts

Before we go on to consider how the surface filters proposed by Chomsky and Lasnik would deal with the data presented here, let us briefly review the facts themselves. We have found that in all constructions in which nothing which had been moved was overt on the surface, prepositions were stranded. These constructions included &e-relatives, pronominally-headed relatives, infinitival relatives, and CO deletion. In constructions in which something remaining on the surface had clearly been moved, we found pied piping to be obligatory. These included se- and se be-relatives, free relatives involving movement, questions, and topicalization. The only two apparent exceptions to this rule, topicalization of pronouns and der-relatives, were found to follow directly from inde- pendently needed rules of inversion and splitting up of certain types of prepositional phrases. Some rules of this sort would appear to be necessary under any theory, since the exceptions to pied piping are quite systematic.

3. The Filters

Chomsky and Lasnik propose that the restrictions on preposition stranding might be accounted for by either a local or a nonlocal filter. Let us consider the local filter first.

3. !. The Local Filter

Chomsky and Lasnik's proposed local fllter works in the following way: all languages are assumed to have a rule marking the objects of prepositions with a feature indicating that the NP is the object of a preposition. Their formulation of this rule, which I will refer to as P Marking, is as follows:

(113) wh-NP --[P ]/P

Some languages, such as OE, are also hypothesized to have a local surface filter

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 39: Movement and Deletion in Old English

298 CYNTHIA ALLEN

ruling out a [P ] pronoun in the complementizer when that pronoun is not preceded by a preposition:

(114) Local Filter (Chomsky and Lasnik)

*comp[wh> P 1 ] .. I

To see exactly how this filter works, let us consider some derivations. First, a &e- relative with preposition stranding:

(115) 6aet port-geat getacna6 sum lichamlic andgit &e menn Surh syngia6. that town-gate betokens a bodily sense that men through sin

'That gate of the town betokens a bodily sense that men sin through .'

Alc.Th.Vol.I p.492.13

The underlying structure of the relative clause and its head are as follows:

(116) NP

NP S

sum lichamlic Comp S andgit

NP VP

menn PP V

P NP syngiao

ourh &it

On the S cycle, the relative pronoun &et will be marked [P ] as the object of a preposition. As a relative pronoun, it is also marked [+wh]. Wh Movement then applies, placing the relative pronoun in the complementizer:

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 40: Movement and Deletion in Old English

MOVEMENT AND DELETION IN OLD ENGLISH

(117) S

Comp S

6xet, P NP VP

menn PP V

P syngia6

6urh

The pronoun in Comp is then deleted, and 3e is spelled out. There is no violation of the filter, since the [P ] pronoun has been moved before the surface, where the filter applies.

Instead of moving only the relative pronoun, it would have been possible in this structure to move the whole PP, giving sum lichamlic andgit 3urh &et (&e) menn syngia5, which was also grammatical. However, other things being equal, it should be possible to delete the relative pronoun in Comp when the preposition is moved, giving the ungrammatical *sum lichamlic andgit 3urh &e menn svngia&. Chomsky and Lasnik claim that such a deletion would be ruled out by the A-over-A Principle holding on the rule deleting any category in Comp. Whether this particular formulation of the A-over-A Principle is correct or not is an open question.45 If it is not, we will need another filter to rule out stranded prepositions in Comp. By the nonfilter approach, however, the impossibility of these prepositions' appearing by themselves in Comp is of course an automatic consequence of assuming that relatives without surface pronouns involve deletion in place, rather than in Comp.

Now let us consider a case where the filter would apply. The example must of course be hypothetical, since it would be ungrammatical. Following are a hypothetical example of preposition stranding in a question and the grammatical attested example on which it is based:

45 Such a formulation would make the A-over-A Principle apply to all nodes meeting the structural description of a rule, rather than only to nodes of the same category. For some problems with this approach for OE, see Allen (1977, 9.1.1.3).

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 41: Movement and Deletion in Old English

300 CYNTHIA ALLEN

(118) a. *Hw. twwost 6u flu vmbe? b. Ymbehwettweost Ou nu?

about what doubtest thou now 'About what do you doubt now?'

Sol. p.85.16 The deep structure of both of the sentences would be (119):

(119)

Comp S

Q NP VP

Ou V Adv P

tweost nu P NP

ymbe hwxet First the wh-pronoun is marked [P ]. Since Chomsky and Lasnik assume pied

piping to be optional, it is possible to move the NP hwet by itself, giving (1 18a), which after Subject-Verb Inversion would give this structure:

(120) S

hwt, P V NP VP

twa-ost 6u Adv PP

nu P

ymbe

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 42: Movement and Deletion in Old English

MOVEMENT AND DELETION IN OLD ENGLISH

The surface filter rules this derivation out. If, on the other hand, the PP is moved, the filter does not apply, and we have the grammatical (1 18b).

Chomsky and Lasnik admit that the rule of P Marking is ad hoc, but claim that it is no more so than hypothesizing two sources for relatives. However, it is not clear why postulating two sources for different types of relatives is ad hoc when those two types exhibit different properties. Maling (1978) has shown a need for two types of relatives in some of the Scandinavian languages. On the other hand, there is no inde- pendent motivation for a rule like P Marking. Note that if this rule were assumed to be language-specific, it would clearly be undesirable, since there is a complete lack of evidence for it which is not bound to Chomsky and Lasnik's theory.

What is the precise nature of P Marking? Chomsky and Lasnik are vague on this point, saying this rule "can be regarded simply as a general convention, invariant across languages" (p. 497). They do not specify at what level of structure this convention is to apply, but one would hope that a rule like this would be part of the base. If the rule of P Marking can be shown to be ordered between two transformations in a language, its status as a universal rule becomes dubious. But in OE, P Marking must be allowed to apply before Wh Movement, but after Locative Inversion. We will now see why this is so.

Consider again the problem of preposition stranding in &er-relatives. Under the filter approach, &er would be marked [P ] like other objects of prepositions. This means that preposition stranding should never occur in 3er-relatives, which is false. There are three plausible approaches to dealing with this problem. First, we might try amending the filter by adding the feature [-loc] to the pronoun in the filter. This would prevent the filter from applying to locative pronouns, and so permit stranding in locative relatives. However, this solution will not work, because locative questions, unlike locative relatives, did not permit preposition stranding, as we have seen. A filter with the feature [-loc] would predict that they did.46 Even if features could be made to make the filter work, such a solution would be ad hoc, as it would give no explanation for the fact that preposition stranding with pronouns occurred only when there could be inver- sion.

The second possibility is that inverted locative phrases are generated as such in the base, thereby escaping P Marking. However, it is not clear that inverted locative objects should escape P Marking. If P Marking is to be a universal convention, it should presumably apply to postpositional as well as to prepositional languages, meaning that it should not be formulated in terms of linear order. Second, this approach makes the wrong historical predictions. If the ability of &er to strand prepositions is a direct result

46 That is, assuming both hweer and &r were generated as objects of P. Angle brackets could possibly be used to express the fact that if the pronoun was [+locative], the filter

only applied if it was also [+Q]. However, this would fail to capture the generalization that only those pronouns undergoing Locative Inversion could strand prepositions, and would also greatly reduce the uni- versality of the filter.

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 43: Movement and Deletion in Old English

302 CYNTHIA ALLEN

of its inverted position in the base, we would expect that when hwaer also became subject to inversion, it also could strand prepositions, not being affected by P Marking. This is not the case, however, as we shall see in section 3.2.

The third possible solution is to permit P Marking to apply after Locative Inversion. In this way, &er which has undergone Locative Inversion would not be marked [P ] and so would not offend the filter. The rule deleting prepositions before locative pro- nouns must also be permitted to apply before P Marking. Otherwise, questions with hwcer and relatives with &er would always be marked [P ] before the preposition deleted.

While this approach will work, having to permit P Marking to follow Locative Inversion and Preposition Deletion is an unpleasant result, as one would hope that a supposedly universal "convention" of this sort would necessarily apply before all transformations. Also, note that this approach would also assume P Marking to be sensitive to linear order, an impossible attribute for a universal rule.

We see, then, that all the possible ways for the local filter to account for &er- relatives run into severe problems. It is also worth noting that P Marking violates a couple of Chomsky and Lasnik's own proposed restrictions on the form of grammar. They say (p. 433):

Our hypothesis, then, is that the consequences of ordering, obligatoriness, and contextual dependency can be captured in terms of surface filters, something that surely need not be the case in principle; and further, that these properties can be expressed in a natural way at this level.

However, P Marking is of course both obligatory and contextually dependent (on P). This would perhaps be all right if it could be argued to apply before all transformations, but it seems that it cannot.

Finally, notice that Locative Inversion must be permitted to apply before Relative Movement. Since Locative Inversion is presumably a rule of the sort which Chomsky and Lasnik would classify as "stylistic", this ordering is contrary to their hypothesis that such rules always apply after the "core" rules, such as Wh Movement.

Another problem arises when we reconsider topicalization in light of the local filter. We saw that preposition stranding in topicalization was possible only with pronouns, which could invert with their prepositions and separate from them in simple sentences.

Chomsky (1977), noting that topicalization in English shows configuration (1), has proposed a new analysis of this construction to accommodate it to his theory that all constructions showing this behavior involve Wh Movement. Since OE topicalization also apparently violated the Specified Subject, Tensed-S, and Subjacency Conditions, but not the Complex NP or Wh-Island Constraints, it must be analyzed in the same way as the Modern English rule.

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 44: Movement and Deletion in Old English

MOVEMENT AND DELETION IN OLD ENGLISH

In Chomsky's analysis, a phrase which appears to have been moved to the front of a sentence by topicalization is actually generated in a special topic position before S, under a new node S. A wh-pronoun is generated in the place where the topic appears to have originated in the sentence, and is then moved into the complementizer and later deleted. Under Chomsky's approach, the deep structure for (121) would be (122).

(121) Ac 6as 6ing ic sprec to eow. but these things I said to you

Alc.P.IX. 18

(122) S

Top S

Oas 6ing Comp S

NP VP

ic V NP P

spraec [+pro] P NP +wh

to eow

The wh-pronoun would then be moved into Comp and later deleted. The new node S is necessary because it is possible to topicalize within embedded sentences:

(123) Harold retorted that his subscription to Linguistic Anguish, he would never cancel.

According to Chomsky, such sentences have the following structure:

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 45: Movement and Deletion in Old English

304 CYNTHIA ALLEN

(124) S

/ ~~~~~S

Comp S

that Top S

Topicalization within subordinate clauses is found quite frequently in the OE texts, so Chomsky's analysis of topicalization is presumably to be extended to OE:

(125) . . .oe wende &aet him ne mihte nan werod wi6standan who thought that him not might no army withstand

'who thought that him, no army could withstand' Alc.S .XXVI .29

(126) foroam de him nan man oone godcundan geleafan ne tehte because him no man the divine faith nottaught 'because him, no one taught the divine faith'

Alc.S.XXX. 12

If OE topicalization is to be analyzed in the way suggested for Modern English by Chomsky, the local filter cannot account for the fact that pied piping was obligatory in topicalization except with pronouns. Consider a hypothetical ungrammatical OE sen- tence with topicalization and preposition stranding, along with the attested example upon which it is based:

(127) a. 1&es lichaman life, 6e langsum beon ne meg, swinca6 menn swi6efor. b. For &es lichaman life, 6e langsum beon ne mxg, swinca6

for the body's life that long be not may toil menn swioe. men greatly 'For the life of the body which cannot last long, men toil greatly.'

Alc.P.VI. 145

In deriving (127b), the PP appearing at the beginning of the sentence is generated in the topic slot, after which the preposition and wh-pronoun are deleted. The problem here is that there appears to be no way to prevent just the NP &es lichaman life from being generated in the topic slot, with a corresponding wh-pronoun as the object of a preposition within the sentence:

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 46: Movement and Deletion in Old English

MOVEMENT AND DELETION IN OLD ENGLISH

(128) s

Top S

otes lichaman Comp S life

NP VP

menn V Adv PP

swinca6 swi6e P NP

for [+pro _+wh

The wh-pronoun would be marked [P ] by P Marking and moved into the complementizer, but it would be deleted before the filter applied. In this case, there would be no violation of the filter, and the ungrammatical (127a) should result.47 The only way the filter could rule out stranding in topicalization by this analysis would be if it applied before the deletion of the wh-pronoun.48 But this of course is impossible, since if the pronoun deletion rule followed the filter, preposition stranding should be ruled out in &-relatives and other constructions supposedly involving ghost Wh Move- ment, as well as in the constructions in which the pronoun remains.

If the local filter is to be made to account for the topicalization facts, Chomsky's analysis of topicalization must be rejected in favor of one in which the topic itself, rather than a wh-pronoun, moves. Then preposition stranding with pronouns could be dealt with by ordering PP Inversion before P Marking. But where exactly would the topic move to? If Chomsky's explanation for this construction's exhibiting configuration (1) is to be maintained, the topic must be moved into Comp. But it clearly cannot be

47 The same problem arises under a theory in which topicalization involves no movement, but only deletion.

48 However, Chomsky and Lasnik specifically reject presurface filters as adding too much power to the grammar.

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 47: Movement and Deletion in Old English

306 CYNTHIA ALLEN

moved into the same slot in Comp as the relative pronouns, because these occur before the complementizer 3e, while topics in subordinate clauses always occur after the complementizer 35t. To move the topic into Comp, we would need to postulate an otherwise unmotivated three-place Comp:

(129) S

[+wh] {oe } Top

With such an analysis of Comp, however, the motivation for moving topics into Comp evaporates. If the topic does not occupy the same slot as the relative pronoun, why should topicalization block extraction of other items from the clause, such as relative pronouns?

To conclude our discussion of the local filter: we have seen that, to work at all, this filter must be combined with some of the very same specific restrictions on trans- formations which local filters are intended to help replace. The filter does not allow us to dispense with the theory of grammar in which stylistic rules may be ordered before rules like Wh Movement. It further necessitates abandoning the Wh Movement analysis of topicalization, which would undermine Chomsky's system of conditions, and there- fore the motivation for a surface filter in this case.

3.2. The Nonlocal Filter

Since the local filter proposed by Chomsky and Lasnik (and as far as I can see, any other local filter) cannot deal adequately with the facts presented here, we are forced, within Chomsky and Lasnik's theory, to resort to a nonlocal filter. This is an undesirable result, since such a filter is a very powerful device, and one which is not known to be independently necessary. One would hope that surface filters (if they are needed at all) could be limited to local ones, as this would result in a more restrictive theory and fewer possible analyses. Keeping this in mind, let us consider the nonlocal filter proposed by Chomsky and Lasnik (1977, 497) as a possible solution to the OE facts:

(130) Nonlocal Filter

*[wh- . . . P t], where t is the trace of wh

It should first be noted that this filter must be revised in two minor ways if it is to work at all. First, we need brackets around P t to show that they constitute a preposi- tional phrase. This is because in Old English, as in Modern English, there were intran- sitive prepositions or particles:49

49 Strang's (1970, 275) claim that verb + particle combinations were practically unknown in OE is an exaggeration. Her statement is based on Kennedy's (1920) study of these combinations in English. In the first

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 48: Movement and Deletion in Old English

MOVEMENT AND DELETION IN OLD ENGLISH

(131) da eode Petrus ut & byterlice weop. then went Peter out and bitterly wept 'Then Peter went out and bitterly wept.'

St.Luke 1383 (132) Oa ahof Paulus up his heafod.

then raised Paul up his head 'Then Paul raised up his head.'

Blickling p. 187.36

As it stands, the filter would wrongly rule out the possibility of relativizing, topi- calizing, or questioning an NP which happened to follow one of these prepositions.

Second, since the stranded preposition nearly always preceded the verb in OE, we need to allow for material after the trace, giving this revised filter:

(133) Revised Nonlocal Filter

*[wh-. . pp[P t] . . . ], where t is the trace of wh

This filter works in the following way: when an NP is moved, it leaves a trace, which is indexed and bound to the NP of which it is the trace. This filter prevents any wh-word from binding a trace in a prepositional phrase.

Let us briefly consider the theoretical status of such a filter. We have already noted that nonlocal filters add great power to the grammar. This might be justified if such filters greatly reduced the power of the theory in other ways. Chomsky and Lasnik claim that filters restrict the power of the grammar by eliminating rule ordering, obli- gatoriness, contextual dependency, and rules over a variable. However, as noted by Bresnan (1976a), the power taken away from one part of the grammar is simply shifted to another. Bresnan notes that the filter needed for the OE facts makes crucial use of variables and labeled brackets, which are devices which Chomsky has argued should not be used in transformations. Filters by their very nature must of course also make use of contextual dependency and obligatoriness. It is not at all clear that shifting these devices from the transformational component to a filter component creates a more restrictive theory, and to permit nonlocal filters would be to undermine the restrictive- ness supposedly gained by allowing local filters. Therefore, the nonlocal filter should be rejected on general theoretical grounds.

Apart from its general undesirability, it is clear that this filter runs into the same

300 lines of Beowulf, Kennedy found only five examples of a verb used with a separate adverbial modifier, as opposed to twenty-five examples of verbs with inseparable prefixes. But any construction which appears five times in only 300 lines of OE poetry cannot be considered rare. It is true that verbs with inseparable prefixes were much more common than verb + particle combinations, but this does not mean the latter construction was rare, even though it is more common in Modern English than it formerly was.

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 49: Movement and Deletion in Old English

308 CYNTHIA ALLEN

problem with topicalization as the local filter. Assuming Chomsky's analysis of topi- calization, it would be possible to generate a simple NP in the topic position, corre- sponding to a wh-word within a PP in the sentence, and move the wh-word into the complementizer. Since the wh-pronoun disappears before the filter applies, there should be no violation of the filter.

It is, of course, impossible to revise the filter as (134), since this would wrongly rule out preposition stranding in &e-relatives, etc.

(134) * * .. PP t]

It is also useless to revise (133) by substituting NP for wh, since the topic NP does not bind the trace, being generated in its surface position, and in any event this move would wrongly rule out preposition stranding with topicalized pronouns.

Such a nonlocal filter could handle the preposition stranding facts only if we abandoned the hypothesis that topicalization involves Wh Movement, and postulated instead that the topic itself moves into the complementizer (or to the front of the sentence), as discussed above. Then the nonlocal filter would have to be revised as (135).

(135) * .NP. . .pp[P t]. . . , where t is the trace of NP

Aside from the difficulties that this account of topicalization raises within Chom- sky's theory, we are still left with two problems: why preposition stranding was possible in topicalization when the topic was a pronoun, and why it was possible in &er-relatives. We saw that the local filter could deal with these facts by permitting P Marking to apply after Locative Inversion and PP Inversion. The nonlocal filter, however, does not have recourse to this solution since it does not use P Marking.

Whether or not the nonlocal filter can account for preposition stranding in the cases where inversion takes place depends on the precise formulation of PP and Locative Inversion. If these rules involve movement of the NP only, the NP should leave a trace. Then if the pronoun is topicalized, it should violate the filter, since the pronoun would bind a trace after P. Let us consider a sample derivation:

(136) and him come owt leoht to ourh paules lare and him came that light to through Paul's teaching 'and him, the light came to through Paul's teaching'

Alc. S.XXIX. 18

Assuming that topics move into Comp, the underlying structure would be (137):

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 50: Movement and Deletion in Old English

MOVEMENT AND DELETION IN OLD ENGLISH

(137) 5

Comp S

Top NP VP

6xt leoht V PP PP

com P NP P NP

to him 6urh paules lare

If PP Inversion moves the NP to the left, after it applies we would have the following structure:50

(138) S

Comp

Top NP VP

&aet leoht V NP PP PP

com him P NP P NP

to t 6urh paules lare

50 I am assuming here a version of PP Inversion in which the pronoun is removed from the PP, but this is immaterial here.

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 51: Movement and Deletion in Old English

310 CYNTHIA ALLEN

Topicalization would then move him into Comp, leaving a trace before PP. This trace causes no problems, but the trace after P should cause the filter wrongly to rule out the sentence. The same problem would arise in locative relatives.

A second possible analysis of the inversion rules would be that they involve a permutation, an operation Chomsky and Lasnik do not allow in transformations. By this analysis, there would be no trace, and no violation of the filter.51

The third possibility is that it is the P, rather than the pronominal NP, which moves. Again, there would be no trace, and therefore no violation of the filter.

By either of the last two analyses, the stranding of personal and demonstrative locative pronouns follows automatically from their participation in inversions which leave no traces, rather than from any ability of inverted prepositional phrases to break up into two constituents. This might seem a good result, as it eliminates PP Split and connects inversion of PP and stranding very intimately.

However, there is historical evidence that it is a mistake to consider preposition stranding an automatic result of inversion. To see this, let us take a brief look at the history of locative pronouns in Middle English.

By the end of the first quarter of the thirteenth century, hwer (OE hwwer) was in common use as a relative pronoun, as the interrogative pronouns were now vying with the demonstratives as relative pronouns. By this time, Locative Inversion and Locative Replacement had generalized to locative wh-pronouns, both interrogative and relative:

(139) Hare confort & hare delit, hwerin is hit al meast but i fleshes their comfort and their delight wherein is it all mostly but in flesh's ful6e o6er in worldes vanite. filth or in world's vanity 'Their comfort and their delight, what is it mostly in, but in the filth of the flesh or the vanity of the world?'

H.M. p.39.1 (140) Ne don no 6ing ne ne siggen hwar &iruh hire silence muwe beon

not do no thing nor not say where through her silence might be i sturbed. disturbed 'Do not do or say anything through which her silence might be disturbed.'

A.Riwle p.194.32 (141) . . .6tt in 6e haues alle 6ing hwerfore mon ah beo

that in thee hast all things wherefore one ought be luuewuroe to o6er loveworthy to other 'that in you have all things for which a person ought to be worthy of love to another'

Wohunge 621 51 If PP Inversion is assumed to leave no trace, however, some problems might arise with the semantic

interpretation of the pronoun as the object of the preposition.

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 52: Movement and Deletion in Old English

MOVEMENT AND DELETION IN OLD ENGLISH

What was the behavior of hwer at this time with respect to preposition stranding? If we assume, as the nonlocal filter approach demands, that preposition stranding with Uer was simply due to Locative Inversion, we would predict that stranding now became possible with hwer. But although relatives and questions with hwer P, such as (139) through (141), are common in the thirteenth-century texts, I have been unable to find any examples of hwer. . .P.

Fortunately, we can compare relative and interrogative hwer with relative &er (OE &xr), because although relative se had by this time been replaced by the wh-pronouns, the locative and temporal demonstrative pronouns were more resistant, and in the thirteenth (and fourteenth) century we find examples of &-r-relatives along with hwer. And although we find no preposition stranding with hwer, we do find it frequently with 6er in relatives, as well as occasionally in simple sentences:

(142) Nes nawt iteiet to oe treo oer he deide upon. not-was nothing tied to the tree where he died upon 'Was there nothing tied to the tree that he died on.'

St.Kat. 1 186 (143) 6o ure louerd wolde bisenchen sodome &-r loth his freond

when our Lord would destroy Sodom where Lot his friend wunede inne dwelled in 'when our Lord would destroy Sodom, where Lot his friend dwelled'

A.Riwle p. 187.1 (144) Sire ich hit dude for uuel 6ah 2xer ne come nan of.

Sire I it did for evil though there not came none of 'Sire, I did it for evil, though none came of it.'

A.Wisse p. 164.16 (145) Hwat weole o6er hwat wunne se 3er eauer of cume, to deore hit

what wealth or what pleasure as there ever of comes too dearly it beo6 aboht. is bought 'Whatever wealth or pleasure comes of it, too dearly it is bought.'

H.M. p.37.13

These facts are quite mysterious if we assume that preposition stranding was a direct result of Locative Inversion's not leaving a trace. By this analysis, both 3er and hwer should allow preposition stranding in ME, because both underwent Locative Inversion. The facts present no problem, however, for the assumption that the restric- tions on OE preposition stranding were due to a prohibition against moving out of PP, if we assume the existence of a PP Split rule. By this analysis, we can account for the historical development in the following way: although Locative Inversion and Locative Replacement generalized to hwer in the thirteenth century, PP Split became more

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 53: Movement and Deletion in Old English

312 CYNTHIA ALLEN

restrictive, adding the feature [-wh].52 Assuming this development, we can account for the lack of stranding in hwer-questions and relatives, since hwer always formed a constituent with its preposition.

The next stage of development was the loss of PP Split altogether. In the early fourteenth century, we find many examples of inverted locative prepositional phrases, but very few with anything intervening between the locative object and the preposition. At this time, preposition stranding was beginning to appear sporadically in wh-relatives and questions in some dialects, but was still rare or nonexistent in most texts. In the texts in which such stranding was unusual or completely absent, hwer-questions and relatives and &r-relatives followed the same pattern as other questions and relatives. The Ayenbite of Inwit, a Kentish manuscript written in 1340, is such a text. In this long work we find abundant examples of hwer P in questions and relatives and &er P in relatives, but only three possible examples of PP Split.53 There are no examples of preposition stranding in ordinary wh-relatives or questions (although preposition strand- ing was still of course possible in that-relatives, CO deletion, etc.). Similarly, Chaucer's works ordinarily show stranding in wh-relatives and questions only when the preposition in question is inne ;54 this is also the only preposition we find stranded in there- and where-relatives and in where-questions. It is clear, therefore, that the possibility of stranding prepositions in &er-relatives in OE was the result of PP Split, rather than just of Locative Inversion. But with the local filter, PP Split cannot be a factor-whether the PP is split up after Locative Inversion or not, there still should be no violation of the filter if the inverted locative pronoun itself is moved away, because there is no trace.

To summarize, there are difficulties with all three possible analyses of Locative Inversion within the nonlocal filter framework. If Locative Inversion is analyzed as a movement of the NP, the resulting trace should cause a violation of the filter. If the rule involves a simple permutation or a movement of the preposition, there is no trace, so the OE facts can be accounted for, but the later difference between &er and hwer cannot.

As a final consideration against filters as a way of dealing with pied piping facts, let us consider whether they could account in a natural way for the differences in

52 It is important to note, however, that there was no change in the data concerning PP Split, since splitting with &1r and nonsplitting with hwaer had always been the rule. The change was in the language learners' interpretation of the data, brought about by the new ability of wh-locatives to participate in Locative Inversion.

It is somewhat puzzling that PP Split, unlike Locative Inversion, failed to generalize to wh-locatives, but this is perhaps not so surprising when we consider that English word order was becoming more restrictive at this time, and that PP Split was on its way to disappearing.

53 All three of these are dubious, as they involve verb + preposition combinations which are possibly verb + particle, such as come in. I have found only one example of PP Split with &er in simple sentences in the Ayenbite, also dubious because it involves go in, and no examples with personal pronouns.

54 This fact was noticed by Grimshaw (1975). A very few examples of preposition stranding with prepositions other than inne occur in Chaucer's works. It appears that in this dialect, stranding was beginning to make its way into wh-relatives and questions, but was not really acceptable yet. I do not know why inne should have been the first preposition to strand freely.

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 54: Movement and Deletion in Old English

MOVEMENT AND DELETION IN OLD ENGLISH

preposition stranding among the world's languages. If the filter approach is correct, one would hope to be able to account for these differences by the presence or absence of the same filter.

I know of three basic types of languages with respect to preposition stranding. Each of the three is represented by at least one of the Germanic languages. The first type allows free preposition stranding. In this category are Modern English, Icelandic, Swedish, Norwegian, and Danish. The second type allows stranding only when no moved item remains at the surface. We have seen that OE was of this sort. The final type allows no preposition stranding at all. Dutch,55 Standard German, and Yiddish are of this type. Yiddish is particularly interesting because, unlike the other nonstranding Germanic languages, it has a relative involving an indeclinable complementizer. Let us look briefly at the Yiddish facts and see how the filter approach would account for them.

Yiddish has a complementizer vos which is similar to English that:

(146) Es iz a glik vos er redt khinezish. it is a good (thing) that he speaks Chinese

Vos is also used as a relative complementizer:

(147) Der yid vos ikh ze iz basheftikt. the man that I see is busy

Lowenstamm (1977) has argued convincingly that relative vos cannot be a pronoun. Yiddish does have a real relative pronoun, which always exhibits pied piping:

(148) a. *Dos is der yid vemen ikh hob geret mit. this is the man whom I have spoken with

b. Dos is der yid mit vemen ikh hob geret.

Pied piping is likewise obligatory in questions and topicalization:

(149) a. *Vemen hostu geret mit? whom have you spoken with

b. Mit vemen hostu geret? (150) a. *Im farloz ikh zikh oyf.

him depend I on 'Him, I depend on.'

b. Oyf im farloz ikh zikh.

So far, the facts are similar to those of OE. But in Yiddish, preposition stranding is not possible in vos-relatives (even though pied piping is also forbidden):

(151) *Dos is der yid vos ikh hob geret mit.

55 However, stranding is possible in Dutch with locative (inverted) wh-words, as in OE.

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 55: Movement and Deletion in Old English

314 CYNTHIA ALLEN

CO deletion is likewise impossible on the object of a preposition:

(152) *Zi iz interesant tsu redn mit. she is interesting to talk with

Clearly, neither of the surface filters proposed by Chomsky and Lasnik for OE can account for these facts, since the pronoun which is assumed to have been moved is deleted before the filters could apply. There is, of course, another filter which could handle these Yiddish facts:

(153) *pp[P t]

Assuming that vos-relatives and CO deletion in Yiddish involve movement, this filter will rule out preposition stranding in these constructions as well as in those involving overt movement. There are two objections to this filter, however. First, Chomsky and Lasnik note that all the filters they discuss have to do with the comple- mentizer system, and suggest that it might be possible to restrict the theory so that surface filters can only affect phrases in Comp. Without such a restriction, the surface filter becomes very powerful. Filter (153) does not apply to elements in Comp, neces- sitating a more powerful theory of filters.

Second, this filter does not give a unified approach to preposition stranding in the world's languages. The Yiddish filter is quite different from either of those proposed for OE. On the other hand, a theory allowing controlled deletion, but no filters, easily explains the Yiddish data. Within such a theory, we would assume a universal constraint to the effect that a language learner assumes that his language permits preposition stranding in neither deletion nor movement contexts, unless he hears evidence to the contrary. If the learner hears numerous instances of stranding in deletion constructions, but not in movement, he assumes that his language has a weaker form of the restriction: preposition stranding is permitted only in deletion. If numerous examples of preposition stranding occur in both types of construction,56 he realizes that his language lacks the restriction altogether. Thus, the theory allowing deletions gives a unified treatment of the different languages by means of modified forms of the same universal principle, instead of very different filters for different languages.

4. A Nonfilter Movement-Only Approach

A Wh Movement approach for all OE relativization which does not involve the filters discussed above has been proposed under the pseudonym of Jan Vat (1978). We will see here that while Vat's analysis might generate all the grammatical relatives for OE

56 There must be some sort of threshold frequency of examples, to rule out the possibility that the language learner will reject the restriction on the basis of a few ungrammatical examples (unless the restriction can be shown to correlate directly to another feature of the language). But such a threshold is presumably also needed in a theory with nonuniversal filters.

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 56: Movement and Deletion in Old English

MOVEMENT AND DELETION IN OLD ENGLISH

and none of the ungrammatical (although there are serious problems even here), it fails to account for the facts of OE topicalization.

Vat's approach is as follows: we may assume (contrary to Chomsky and Lasnik) that movement out of PP is generally universally prohibited. But in OE and Modern Dutch,57 there is an escape hatch in PP through which pronouns may be extracted- precisely the positions in PP in which we find inverted &er (and, in OE, optionally the personal pronouns). These pronouns can make use of this escape hatch by virtue of the feature [+R] which they carry, while [-R] pronouns, such as demonstrative and wh- pronouns, cannot, as only [+R] pronouns undergo inversion into the escape position. Thus, stranding in 3er-relatives follows directly from the ability of the [+R] &er to move into the [+R] position before the preposition.

While this analysis would account for the 3er-relatives, what about the &e-relatives with stranding? Concerning these, Vat says (1978, 71 1):

Given the fact that fier could strand its preposition, and given the fact that there was a rule deleting p-pronouns [i.e. demonstrative pronouns/CLA] in COMP. . it does not come as a surprise that there is ample evidence for pattern (52c).

The pattern (52c) to which Vat refers is that of preposition stranding in &-relatives. Vat does not explicitly spell out his analysis of &e-relatives, but this statement clearly implies an analysis by which &er is the pronoun involved in these relatives, which is moved into Comp and then deleted. But clearly, &r cannot be the pronoun involved in &-relatives, because 3xr and &e were used with different types of heads. While there are no examples of 3er-relatives with animate heads, there are many examples of &e-

relatives with such heads. Thus, it is clearly a mistake to equate the stranding in i-

relatives with the stranding in ber-relatives. There remains the possibility that a different [+R] pronoun was involved in &-

relatives. The only other such candidates are the personal pronouns. There is no evidence that the personal pronouns underlie the i-relatives; in fact, there is evidence against such an analysis. By this analysis, we would expect complete symmetry between stranding with personal pronouns and stranding in &?-relatives. But this is not the case. While stranding is found with all prepositions in i-relatives, we do not find it with certain prepositions and personal pronouns, apparently because of idiosyncrasies of PP

57 The Dutch facts are briefly as follows: preposition stranding is generally impossible in all constructions. However, a nonhuman pronominal object of P is replaced by a locative pronoun, which inverts with P. Just those pronouns which invert, to which Vat assigns the feature [+R], can be extracted by Topicalization, Wh Movement, etc.

While the Dutch facts are very similar to those of OE, a few differences between the two should be noticed. First, in Dutch, unlike OE, Locative Replacement is obligatory. Second, different pronouns invert in the two languages. In Dutch, all locative pronouns invert with P, while in OE the wh-locatives did not. Also, OE, unlike Dutch, permitted personal pronouns to invert. Finally, in Dutch there is a specific position (directly after the subject) into which pronouns move, while in OE no such position can be found, as noted earlier.

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 57: Movement and Deletion in Old English

316 CYNTHIA ALLEN

Inversion and PP Split. For example, Wende (1915) found no examples of &rh following its pronominal object in simple sentences, but did find examples of it stranded in &e- relatives.58 Similarly, while Wende gives numerous examples of in stranded in &e-rela- tives, he finds no examples of in following a personal pronoun. Thus, the idiosyncratic restrictions on PP Inversion are not reflected in the i-relatives, contrary to the require- ments of this theory.

The only other possible solution is that the [+R] pronoun of ie-relatives was an abstract one. But surely we would prefer a theory which did not require such an abstract pronoun, which must have the same features as the demonstrative pronouns except for the difference in the [R] feature. It is also difficult to see how hypothesizing two different kinds of relatives with two slightly different pronouns is less ad hoc than postulating two types differing in movement vs. deletion.

Vat's analysis faces other severe difficulties. First, consider the topicalization facts presented earlier, which Vat does not mention, although they are a central argument of Allen (1977), whose analysis of preposition stranding Vat attacks.59 Vat's analysis runs into a problem similar to that of the filter approach with topicalization. As Vat wishes to explain all constructions showing configuration (1) as involving Wh Movement, he is apparently committed to a Wh Movement analysis of topicalization. But consider again the discussion of (127). We saw that there was no reason that an NP topic could not be generated in the topic position, while a related PP was generated within the S. Within Vat's framework, the pronoun moved in topicalization (by Wh Movement) would pre- sumably be the same [+R] one of &e-relatives, since it always deletes. But if this pronoun is [+R], topicalization should always be able to strand prepositions, since the pronoun could invert into the escape hatch position.60

The only recourse is to accept an analysis of topicalization in which the topic itself moves, which we have seen is inimical to the assumption that Wh Movement underlies all constructions with configuration (1). But without this assumption, there is no moti- vation for a movement analysis of &-relatives.

Second, Vat's analysis relies on the assumption that stranding with pronouns which invert is an automatic result of the inversion. Indeed, Vat considers it a virtue of his analysis that it makes this connection. But we have seen that this assumption cannot be right. If the [+R] position was an escape hatch for &-r in Middle English, why was it

58 I have also found some more examples of stranded Warh in &-relatives in texts which Wende did not include in his study, but no examples of Y*rh inverted with (or separate from) a personal pronoun object. It should be noted that the construction 3urh + personal pronoun was by no means rare, so the lack of inversion cannot be due to lack of opportunity.

59 Indeed, Vat implies that my only reason for preferring a deletion analysis of stranding constructions is that this approach provides a unified explanation for the pied piping facts in the different Germanic languages, as discussed above (although he makes no mention of the reasons why my approach accounts for these facts in a more unified way than Chomsky and Lasnik's). But I consider this uniform explanation to be a pleasant side-effect of my approach, rather than a central argument. Vat does not mention any of the more central arguments.

60 Conversely, if the pronoun is [-RI, stranding should never be possible in topicalization, even with personal pronouns and locatives.

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 58: Movement and Deletion in Old English

MOVEMENT AND DELETION IN OLD ENGLISH

not for hwer? On the other hand, if we assume that the stranding with the inverted pronouns was due to a rule of PP Split which broke off the constituent PP, there is no problem-hwer simply did not undergo PP Split. It seems that inversion was a necessary, but not a sufficient, condition for stranding (other than stranding by deletion, of course).

To conclude, it seems that Vat's analysis cannot reach the level of descriptive adequacy within his own theoretical framework.

5. Conclusion

We have seen that there is every reason to believe that all OE relative clauses were subject to the Complex NP and Wh-Island Constraints, but allowed systematic apparent violations of the Subjacency, Tensed-S, and Specified Subject Conditions of the same sort that Modern English allows. This being so, the relatives in which no pronoun appears on the surface (the &-relatives) must be analyzed as involving Wh Movement, given Chomsky's system of conditions on movement. However, all constructions in which a moved item was overt did not allow preposition stranding, while those which must be analyzed within Chomsky's theory as involving ghost Wh Movement did permit stranding. These facts suggest a possible need for two types of rules exhibiting config- uration (1): controlled unbounded deletion and movement. The former type of rule being impossible in Chomsky's system, Chomsky and Lasnik propose that the OE facts could be dealt with by either a local or a nonlocal filter.

We have seen that the local filter offers no advantages over the nonfilter approach, other than making the OE data consistent with Chomsky's explanation for configuration (1), since it cannot dispense with rule obligatoriness or contextual dependence, and furthermore must make use of a rule of P Marking not motivated by any independent considerations. The local filter is also inconsistent with Chomsky's Wh Movement analysis of topicalization, an analysis which is necessary if the proposed explanation for configuration (1) is to be maintained.

The nonlocal filter is an undesirable addition to the power of grammatical theory. This filter is also inconsistent with the Wh Movement analysis of topicalization. It is furthermore incapable of giving a reasonable account of the changes in preposition stranding with locative pronouns which took place in the thirteenth and fourteenth centuries. We conclude therefore that the filters do not give a more principled account of the OE facts than the assumption that OE had a set of unbounded deletion rules.

We finally discussed Vat's (1978) proposal that stranding in 5e-relatives was related to stranding in &er-relatives. We saw that his analysis also ran aground on the topical- ization facts, and that furthermore it is impossible to identify the pronoun supposedly underlying the &e-relatives.

The major objection which has been raised against unbounded controlled deletion is that it cannot in principle explain the adherence to configuration (1) of constructions assumed to involve such deletion. But Bresnan and Grimshaw (1978) have argued that Chomsky's constraints should be reinterpreted as conditions on syntactic binding, rather

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 59: Movement and Deletion in Old English

318 CYNTHIA ALLEN

than on movement. They show that this reinterpretation can explain the similarity of behavior among constructions clearly involving movement, those apparently involving controlled deletion, and those which could not possibly have movement. On the other hand, if the conditions apply only to movement, we cannot explain why constructions in certain languages which cannot be argued to involve ghost Wh Movement obey the conditions.

At the very least, I hope to have established that the OE facts are of synchronic interest, and that it is possible to learn a great deal, including basic facts about the behavior of the major constructions with respect to Chomsky's conditions, from texts when they are of the length and quality of those studied here.

These facts are also of great diachronic interest, as they represent the first in-depth study (to my knowledge) within a generative framework of the OE constructions con- sidered here. It is important to firmly establish the facts of these constructions as a preliminary to a satisfactory history of English syntax. Such a history will hopefully give evidence bearing on the best synchronic theory, as the best theory of syntax should be able to deal with syntactic change in the most satisfactory way.

Appendix: Abbreviations and Texts Examined

A. Abbreviations

A.Riwle = Ancrene Riwle, ed. Day. Page and line. A.Wisse = Ancrene Wisse, ed.Tolkien. Page and line. Alc.P. = Homilies of Aelfric, ed. Pope. Roman numerals = homily no., arabic = line

no. Alc.S. = Aelfric's Lives of Saints, ed. Skeat. Alc.Th. = Homilies of the Anglo-Saxon Church, ed. Thorpe. Volume, page, and line. Angl.Hom. = Angelsachsische Homilien und Heiligenleben, ed. Assman. Bede = Old English Version of Bede's Ecclesiastical History, ed. Miller. Section,

subsection, page, and line. Blickling = Blickling Homilies of the Tenth Century, ed. Morris. Boeth. = King Alfred's Old English Version of Boethius, ed. Sedgefield. Section,

subsection, page, and line. CP Sweet = King Alfred's West-Saxon Version of Gregory's "Pastoral Care", ed.

Sweet. H.M. = Hali Mei&enhad, ed. Cockayne. Line no. Mart. = Old English Martyrology, ed. Herzfield. 3 OE = Three Old English Prose Texts, ed. Rypins. Oros. = King Alfred's Orosius, ed. Sweet. Sax.Chron. = Two of the Saxon Chronicles, ed. Plummer. By year of entry. Sol. = King Alfred's Version of St. Augustine's "Soliloquies", ed. Carnicelli. St.John = Gospel of St. John, in the West-Saxon Gospels, ed. Grunberg. Line no.

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 60: Movement and Deletion in Old English

MOVEMENT AND DELETION IN OLD ENGLISH

St.Jul. = Liflade ant te Passion of Seinte Juliene, ed. D'Ardenne. Line no. St.Kat. = Life of Saint Katherine, ed. Einenkel. Line no. St.Luke = Gospel of Saint Luke, in the West-Saxon Gospels. Line no. St.Mark = Gospel of St. Mark, in the West-Saxon Gospels.Line no. St.Mat. = Gospel of St. Matthew, in the West-Saxon Gospels. Line no. Ver. = Die Vercelli Homilien, ed. Forster. Homily and line no. Wohunge = Wohunge of ure Lauerd, ed. Thompson. Line no. Wulf. = Homilies of Wulfstan, ed. Bethrum. Homily no. and line. Wulf.N. = Wulfstan, ed. Napier.

B. OE Prose Texts Examined

The texts are listed by title, rather than by editor. For a fuller account of the manuscripts, see Allen (1977). See also Karlberg (1954) for a useful summary of Old and Middle English texts by dialect, and Ker (1957) for detailed information about all the extant OE manuscripts. (EETS = Early English Text Society)

Aelfric's "De Temporibus Anni", ed. H. Henel, EETS Vol. 213, Humphrey Milford, London, 1942.

Aelfric's Grammatik und Glossar, ed. J. Zupitza, 1880, rpt. Max Niehaus Verlag, Berlin, 1966.

Aelfric's Lives of Saints, ed. W. Skeat, EETS Vols. 76, 82, 94, and 114, University Press, Oxford, 1881-1900.

Angelsaichsische Homilien und Heiligenleben, ed. B. Assman, Bibliothek der angelsachsischen Prosa, Vol. 3, 1889, rpt. Wissenschaftliche Buchgesellschaft, 1964.

"Das Angelsachsische Prosa-Leben des hl. Guthlac", ed. P. Gonser, Anglistlice For- schungen 27 (1909).

Die Angelsdchsischen Prosabearbeitungen der Benediktinerregel, ed. A. Schroer, 1885- 1888, rpt. Wissenschaftliche Buchgesellschaft, 1964.

Be Domes Daege, ed. J. Lumby, EETS Vol. 65, N. Trubner, London, 1876. The Blickling Homilies of the Tenth Century, ed. R. Morris, EETS Vols. 58, 63, and 73,

N. Trubner, Oxford, 1880. Canons of Edgar, ed. R. Fowler, EETS Vol. 266, Oxford University Press, London,

1972. Exameron Anglice or the Old English Hexameron, ed. S. J. Crawford, Wissenschaftliche

Buchgesellschaft, Darmstadt, 1921. Homilies of Aelfric: A Supplementary Collection, ed. J. Pope, EETS Vols. 259 and 260,

University Press, Oxford, 1967-1968. The Homilies of the Anglo-Saxon Church, ed. B. Thorpe, 2 vols., Aelfric Society, 1844,

rpt. Johnson Reprint Corp., New York, 1971. The Homilies of Wulfstan, ed. D. Bethrum, Clarendon Press, Oxford, 1957.

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 61: Movement and Deletion in Old English

320 CYNTHIA ALLEN

King Alfred's Old English Version of Boethius, ed. W. Sedgefield, 1899, rpt. Wissen- schaftliche Buchgesellschaft, Darmstadt, 1968.

King Alfred's Orosius, ed. H. Sweet, EETS Vol. 79, N. Trubner, London, 1883. King Alfred's Version of St. Augustine's Soliloquies, ed. T. Carnicelli, Harvard Uni-

versity Press, Cambridge, Massachusetts, 1969. King Alfred's West-Saxon Version of Gregory's Pastoral Care, ed. H. Sweet, EETS

Vols. 45 and 50, N. Trubner, London, 1871-1872. An Old English Martyrology, ed. J. Herzfield, EETS Vol. 116, Kegan Paul, Trench,

Trubner, and Co., London, 1900. The Old English Version of Bede's Ecclesiastical History of the English People, ed. T.

Miller, EETS Vols. 95 and 96, N. Trubner, London, 1890-1891. The Old English Version of the Heptateuch, ed. S. J. Crawford, EETS Vol. 160, Oxford

University Press, London, 1922. The Oldest English Texts, ed. H. Sweet, EETS Vol. 83, N. Trubner, London, 1885, rpt.

1938. Three Old English Prose Texts, ed. S. Rypins, EETS Vol. 161, Oxford University Press,

London, 1924. Two of the Saxon Chronicles, ed. C. Plummer, based on an edition by John Earle,

Clarendon Press, Oxford, 1924. Die Vercelli-Homilien, ed. M. Forster, Bibliothek der angelsachsischen Prosa, Vol. 12,

1932, rpt. Wissenschaftliche Buchgesellschaft, Darmstadt, 1964. The West-Saxon Gospels, ed. M. Grunberg, Scheltem and Holkemaa NV, Amsterdam,

1967. Wulfstan. Sammlung der ihm zugeschriebenen Homilien nebst Untersuchungen uber

ihre Echtheit, ed. A. Napier, 1883, rpt. Weidman, Berlin, 1966.

C. Middle English Texts

Only the texts mentioned in this article are listed here. For a full account of the texts examined, see Allen (1977). For a guide to Middle English texts, see the Manual of the Middle English group of the Modern Language Association of America.

Ancrene Wisse, ed. J. R. R. Tolkien, EETS Vol. 249, Oxford University Press, London, 1962.

Dan Michel's Ayenbite of Inwit, ed. R. Morris, EETS Vol. 23, N. Trubner, London, 1867.

The English Text of the Ancrene Riwle, ed. M. Day, EETS Vol. 225, Oxford University Press, London, 1952.

Hali Mei&nhad, ed. 0. Cockayne, EETS Vol. 18, N. Trubner, London, 1866. Re-edited by F. J. Furnivall, 1922.

The Life of Saint Katherine, ed. E. Einenkel, EETS Vol. 80, N. Trubner, London, 1884.

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 62: Movement and Deletion in Old English

MOVEMENT AND DELETION IN OLD ENGLISH

&? Liflade ant te Passion of Seinte Juliene, ed. S. R. T. 0. d'Ardenne, EETS Vol. 248, Oxford University Press, London, 1961.

&e Wohunge of ure Lauerd, ed. W. Thompson, EETS Vol. 241, Oxford University Press, London, 1958.

References

Allen, C. (1975) "Case Marking and the NP Cycle in Old English," Linguistic Analysis 1, 389- 401.

Allen, C. (1976) "The Origins of the Wh-that Construction in English," unpublished paper, University of Massachusetts, Amherst.

Allen, C. (1977) Topics in Diachronic English Syntax, Doctoral dissertation, University of Mas- sachusetts, Amherst.

Andrew, S. 0. (1936) "Relative and Demonstrative Pronouns in Old English," Language 12, 283-293.

Andrews, A. (1975) Studies in the Syntax of Relative and Comparative Clauses, Doctoral disser- tation, MIT, Cambridge, Massachusetts.

Anklam, E. (1908) Das englische Relativ im 11. und 12. Jahrhundert, Mayer & Muller, Berlin. Bach, E. and G. Horn (1976) "Remarks on 'Conditions on Transformations'," Linguistic Inquiry

7, 265-361. Bacquet, P. (1962) La Structure de la Phrase Verbale d l'Epoque Alfredienne, Les Belles Lettres,

Paris. Bever, T. G. and D. T. Langendoen (1971) "A Dynamic Model of the Evolution of Language,"

Linguistic Inquiry 2, 433-463. Bresnan, J. (1972) Theory of Complementation in English Syntax, Doctoral dissertation, MIT,

Cambridge, Massachusetts. Bresnan, J. (1973) "Syntax of the Comparative Clause Construction in English," Linguistic

Inquiry 4, 273-343. Bresnan, J. (1975) "Comparative Deletion and Constraints on Transformations," Linguistic Anal-

ysis 1, 25-74. Bresnan, J. (1976a) "'Evidence for a Theory of Unbounded Transformations," Linguistic Analysis

2, 253-393. Bresnan, J. (1976b) "On the Form and Functioning of Transformations," Linguistic Inquiry 7, 3-

40. Bresnan, J. and J. Grimshaw (1978) "'The Syntax of Free Relatives in English," Linguistic Inquiry

9, 331-391. Brown, W. Jr. (1970) A Syntax of King Alfred's Pastoral Care, Mouton, The Hague. Callaway, M. Jr. (1913) The Infinitive in Anglo-Saxon, Carnegie Institution of Washington Pub-

lication no. 169, University Press, Cambridge, Massachusetts. Carlton, C. (1970) Descriptive Syntax of the Old English Charters, Mouton, The Hague. Chomsky, N. (1973) "Conditions on Transformations," in S. R. Anderson and P. Kiparsky, eds.,

A Festschrift for Morris Halle, Holt, Rinehart and Winston, New York. Chomsky, N. (1976) "Conditions on Rules of Grammar," Linguistic Analysis 2, 303-351. Chomsky, N. (1977) "On Wh-Movement," in P. W. Culicover, T. Wasow, and A. Akmajian,

eds., Formal Syntax, Academic Press, New York. Chomsky, N. and H. Lasnik (1977) "Filters and Control," Linguistic Inquiry 8, 425-504. Curme, G. (1912) "A History of the English Relative Constructions," Journal of English and

Germanic Philology 11, 10-29, 190-204, and 355-380.

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 63: Movement and Deletion in Old English

322 CYNTHIA ALLEN

Einenkel, E. (1916) Geschichte der englischen Sprache II: Historische Syntax, Karl von Trubner, Strassburg.

Fiengo, R. (1974) Semantic Conditions on Surface Structure,Doctoral dissertation, MIT, Cam- bridge, Massachusetts.

Gardner, F. (1971) An Analysis of Syntactic Patterns of Old English, Mouton, The Hague. Goldman, S. H. (1970) Basic and Marked Sentence Patterns in the Vercelli Homilies, Doctoral

dissertation, University of Wisconsin, Madison, Wisconsin. Grimshaw, J. (1975) "Evidence for Relativization by Deletion in Chaucerian Middle English," in

E. Kaisse and J. Hankamer, eds., Proceedings of the Fifth Annual Meeting of the North Eastern Linguistic Society, Harvard University, Cambridge, Massachusetts.

Grimshaw, J. (1977) Wh Constructions and the Theory of Grammar, Doctoral dissertation, Uni- versity of Massachusetts, Amherst.

Harrison, T. (1892) The Separable Prefixes in Anglo-Saxon, rpt. 1970 by McGrath Publishing Company, College Park, Maryland.

Jespersen, 0. (1909-1949) A Modern English Grammar on Historical Principles, 7 vols., George Allen and Unwin, London.

Johnson, J. (1975) A Transformational Analysis of the Syntax of Aelfric's Lives of Saints, Mouton, The Hague.

Karlberg, G. (1954). The English Interrogative Pronouns, Almquist and Wiksell, Stockholm. Katz, J. and P. Postal (1964) An Integrated Theory of Linguistic Descriptions, MIT Press,

Cambridge, Massachusetts. Kennedy, A. (1920) "The Modern English Verb-Adverb Combination," Stanford University

Publications: Language and Literature 1. Ker, N. R. (1957) Catalogue of Manuscripts Containing Anglo-Saxon, Clarendon Press, Oxford. Keyser, S. J. (1975) "A Partial History of the Relative Clause in English," in J. Grimshaw, ed.,

Papers in the History and Structure of English, University of Massachusetts Occasional Papers in Linguistics no. 1, Amherst, Massachusetts.

Klima, E. (1965) Studies in Diachronic Transformational Syntax, Doctoral dissertation, Harvard University, Cambridge, Massachusetts.

Koch, E. (1897) The English Relative Pronouns. A Critical Essay, Druck von Perlmutter, Berlin. Lasnik, H. and R. Fiengo (1974) "Complement Object Deletion," Linguistic Inquiry 5, 535-571. Lowenstamm, J. (1977) "Relative Clauses in Yiddish: A Case for Movement," Linguistic Analysis

3, 197-216. McIntosh, A. (1947) "The Relative Pronouns &e and Yxt in Early Middle English," English and

Germanic Studies 1, 73-87. Maling, J. (1978) "An Asymmetry with Respect to Wh Movement," Linguistic Inquiry 9, 75-89. Maling, J. and A. Zaenen (1977) "Filters and Traces in Icelandic," paper presented at the eighth

annual meeting of the North Eastern Linguistic Society, University of Massachusetts, Amherst.

Maling, J. and A. Zaenen (1978a) "The Nonuniversality of a Surface Filter," Linguistic Inquiry 9, 475-497.

Maling, J. and A. Zaenen (1978b) "Germanic Word Order and the Format of Surface Filters," unpublished revised version of a paper presented at the Amsterdam Colloquium, April 7, 1978.

Middle English Group of the Modern Language Association of America, (1967) A Manual of the Writings in Middle English 1050-1500, 5 vols., a rewriting and expansion of Wells's Manual, The Connecticut Academy of Arts and Sciences, New Haven, Connecticut.

Mustanoja, T. (1960) A Middle English Syntax, Societe Neophilologique, Helsinki.

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions

Page 64: Movement and Deletion in Old English

MOVEMENT AND DELETION IN OLD ENGLISH

Ogden, M., C. Palmer, and R. McKelvey (1954) A Bibliography of Middle English Texts, Uni- versity of Michigan Press, Ann Arbor, Michigan.

Oxford English Dictionary (1933) James A. Murray et al., Clarendon Press, Oxford. Pillsbury, P. (1967) Descriptive Analysis of Discourse in Late West Saxon Texts, Mouton, The

Hague. Poutsma, H. (1904-1928) A Grammar of Late Modern English, in 2 parts, Noordhoff, Groningen. Quirk, R. and C. L. Wrenn (1957) An Old English Grammar, 2nd ed., Holt, Rinehart and Winston,

New York. Riemsdijk, H. van (1976) "Extraction from PP and the Head Constraint," unpublished manuscript,

University of Amsterdam. Riemsdijk, H. van (1978) A Case Study in Syntactic Markedness, Peter de Ridder Press, Lisse. Rijk, R. de (1972) Studies in Basque Syntax: Relative Clauses, Doctoral dissertation, MIT,

Cambridge, Massachusetts. Ross, J. R. (1967) Constraints on Variables in Syntax, Doctoral dissertation, MIT, Cambridge,

Massachusetts. Shores, D. (1971) A Descriptive Syntax of the Peterborough Chronicle from 1122 to 1154, Mouton,

The Hague. Sprockel, C. (1973) The Language of the Parker Chronicle, Martinus Nijhoff, The Hague. Strang, B. M. H. (1970) A History of English, Methuen, London. Toller, T. N., ed. (1898) An Anglo-Saxon Dictionary Based on the Manuscript Collections of the

Late Joseph Bosworth, rpt. Pitman Press, Bath, 1973. Traugott, E. C. (1972) A History of English Syntax, Holt, Rinehart and Winston, New York. Vat, J. (pseudonym) (1978) "On Footnote 2: Evidence for the Pronominal Status of 3xr in Old

English Relatives," Linguistic Inquiry 9, 695-716. Visser, F. (1963-1973) An Historical Syntax of the English Language, Brill, Leiden. Wende, F. (1915) Uber die nachgestellten Prdpositionen im Angelsdchsischen, Palaestra LXX,

Mayer & Muller, Berlin. Wulfing, E. (1901) Die Syniax in den Werken Alfreds des Grossen, P. Hansteins Verlag, Bonn. Zeitlin, J. (1908) The Accusative with Infinitive and some Kindred Constructions in English,

Columbia University Press, New York.

Linguistics, SGS Australian National University Canberra, A. C. T. Australia 2600

This content downloaded from 62.122.72.154 on Mon, 16 Jun 2014 01:30:55 AMAll use subject to JSTOR Terms and Conditions