Outline The residue of syntactic change: Syntactic Change Partial - - PDF document

1 The residue of syntactic change: Partial pro-drop in Old English

Emily Coppess (University of Chicago) and Acrisio Pires (University of Michigan) LSA Summer Institute – Ann Arbor Diachronic Syntax Workshop June 29-30th, 2013

1

Outline

• Syntactic Change
• Partial pro-drop Languages
• Referential Null Subjects in OE
• Null and Overt Expletives in Old English
• That-trace effect (and V-S inversion?)
• Loss of referential null subjects in OE

2

Syntactic Change

• The path of change between any two given stages in the syntax of a language

reflects a principled shift between at least two fully structured grammatical systems, represented as I-grammars (mental grammars) of different individuals: Lg1 Grammar > Lg1 output > Acquisition/learning > Lg2 Grammar A syntactic domain may undergo multiple shifts across generations, reflecting different levels of stability/instability along the way Null subject (NS) grammar > Non null-subject grammar (all or nothing?) NS grammar > > partial-NS grammar 1..> > partial-NS grammar 2…>> non null-subject grammar

3

Null subject vs. Non-null subject languages

• Agreement (phi-features) on Tense are uninterpretable, and are assigned a value by a nominal

argument (Chomsky 2001, Agree and phi-feature/Case valuation) Consistent NS languages):

• “Agreement-based” full null-subject languages (- They require a D(eterminer)-feature in

Inflection/Tense, to allow a null pronoun (phi-P) to be referential (Holmberg, 2005)

• They only have null expletives (in Holmberg’s approach)
• Topic-drop full null-subject languages: a (null) provides reference to the null subject

(Chinese long distance topics, Germanic (matrix) topics)

• Non-NS languages:
• Lack of agreement and D-feature in Inflection/Tense head.
• No topic binding of null subjects
• no null subjects in finite clauses (except imperatives)
• vert expletives.
• Partial pro-drop languages:
• No D-feature in I/T

. Null pronoun (phi-P) must be bound by higher DP (or be generic)

• presence of null subjects is restricted.
• vert expletives are allowed (sometimes required)

4

Gelderen (2013)

1) “Old English is a genuine pro-drop language, although the system is in decline.” p. 271; (see also Mitchell 1985:628-634; Traugott 1992; Gelderen 2000). 2) Elly van Gelderen argues that verbal agreement with subject is linked to licensing of pro- drop in Old English. 3) An aboutness-shift topic licenses the null subject (Frascarelli 2007, Sigurðsson 2011) Our focus: 1) Evidence that an agreement based proposal faces problems:

• Highly restricted referential null subjects appear mostly in subjunctive clauses.
• Restriction also to main clauses (see Coppess 2011, also Walkden 2011, 2012).

2) Further loss in distribution of null subjects throughout the OE period. 3) Null expletive distribution remains largely stable.

6

The Corpus: YCOE

York-Toronto-Helsinki Corpus of Parsed Old English (YCOE, Taylor et. al 2003). Separation of texts into (Coppess 2011):

• Early Period → 300-950CE → 26 texts
• Middle Period → 950-1000CE → 34 texts
• Late Period → 1000-1100CE → 20 texts
• Unidentified → n/a → 20 texts

Notes on Corpus Searches (Coppess 2011):

• searches done with Corpus Search (Upenn), and restricted to Early and Late Period
• texts without a clear/consistent date were excluded from analysis (unidentified)

YCOE: corpus of approximately 1.5 million words, including 100 texts, with a total of 110,136 tokens (a token is a segment of parsed words).

7

2 Early period texts

Early Period Texts (300 - 950 CE) filenames Text Names coalex.o23 Alexander's Letter to Aristotle coblick.o23 Blickling Homilies coboeth.o2 Boethius' Consolation of Philosophy cobede.o2 Bede's History of the English Church cochad.o24 Saint Chad cocura.o2 Cura Pastoralis codocu2.o12 Documents 2 (O1/O2) codocu3.o23 Documents 3 (O2/O3) codocu1.o1 Documents 1 (O1) codocu2.o2 Documents 2 (O2) codocu4.o24 Documents 4 (O2/O4) cogregdC.o24 Gregory's Dialogues colaece.o2 Leechdoms colacnu.o23 Lacnunga colawafint.o2 Alfred's Introduction to Laws colawine.ox2 Laws, Ine colawaf.o2 Laws, Alfred comart2 Martyrology, II comart3 Martyrology, III comarvel.o23 Marvels of the East comart1 Martyrology, I coorosiu.o2 Orosius coprefsolilo Preface to Augustine's Soliloquies coquadru.o23 Pseudo-Apuleius, Medicina de quadrupedibus coprefcura.o2 Preface to the Cura Pastoralis cosolilo

• St. Augustine's Soliloquies

8

Late Period Texts

Late Period Texts (1000 - 1100 CE) filenames Text names caelhom.o3 Ælfric, Supplemental Homilies coapollo.o3 Apollonius of Tyre cobyrhtf.o3 Byrhtferth's Manual cocanedgX Canons of Edgar (X) coducu3.o3 Documents 3 (O3) coinspolX Wulfstan's Institute of Polity (X) colaw1cn.o3 Laws, Cnut I colaw2cn.o3 Laws, Cnut II colaw5atr.o3 Laws, Æthelred V colaw6atr.o3 Laws, Æthelred VI colawnorthu.o3 Northumbra Preosta Lagu colawwllad.o4 Laws, Williams I, Lad colsigef.o3 Ælfric's Letter to Sigefyrth colsigewZ.o34 Ælfric's Letter to Sigeweard (Z) colwgeat Ælfric's Letter to Wulfgeat colwstan1.o3 Ælfric's Letter to Wulfstan I colwstan2.o3 Ælfric's Letter to Wulfstan II conicodA Gospel of Nicodemuc (A) covinsal Vindicta Salvatoris cowulf.o34 Wulfstan's Homilies

9

Overall Distribution of Null Subjects

Referential NS Expletive NS Elided NS Total Subjects Periods Hits % Hits % Hits % Hits Early (<950) 1151 3.77% 1668 5.46% 6062 19.8% 30547 Late (>1000) 206 1.95% 439 4.15% 2027 19.2% 10568

Percentage of each type of null subject out of total subjects.

• Null subjects are still present, but with very low distribution.
• Distribution of null expletives over total number of subjects is misleading.
• Elided subjects correspond to subjects of second conjuncts in coordinated

clauses, which we excluded from the counts of true RefNSs

• Referential null subjects show further restrictions in their distribution.

10

Elided subjects in coordinated clause

(1) and him bebead þas word (CONj and) (NP-NOM Ø) (NP-DAT him) (VBDI bebead) (NP-ACC þas word) and Ø him entrusted the word ‘and entrusted the word to him.’ (2) and het hine sendan of ðæs folces meniu men to ðam lande, þær and Ø commanded him to send of the people’s company men to the land, there ‘and commanded him to send men of the people’s army to the land, there’ þær hi to sceoldon, and sceawian þæt land,… there they to should go, and Ø to.behold the land,… ‘they should go to there, and behold the land,…’ (Ælfric Supplemental Homilies,+AHom_21:140.3149)

• In addition to conjunction phrases, there are elided subjects associated with infinitival

structures.

• Both types of elided subjects with conjunction and infinitival phrases are constructions that

appear in non-pro drop languages as well.

• These elided subjects were excluded from the evidence for a NSL system.

14

Distribution of Null Subjects by type

Referential Subjects Expletive Subjects Elided Subjects Null Total % Null Total % Null Total* % Early (<950) 1151 22362 5.15 1668 2123 78.6 6062 28424 21.3 Late (>1000) 206 8013 2.57 439 528 83.1 2027 10040 20.2

• Even in contexts restricted to possible referential null subjects context, there is still

a very low distribution of referential null subjects.

• Expletive null subjects are clearly very productive throughout the OE period, and

do no show an indication of overall decrease in use.

*Conjoined subjects calculated over total number of overt subjects (% not relevant here). Percentage of the each type of null subject with respect to the specific environment where their overt counterpart occurs.

15

Distribution of Referential NS (matrix vs. subordinate)

Matrix Clause Subordinate Clause Total Periods Hits % Hits % Hits Early (<950) 995 86.4% 156 13.6% 1151 Late (>1000) 187 90.8% 19 9.22% 87

• Most of the referential null subjects appear in matrix clauses – they are

not syntactically dependent on a higher syntactic antecedent (similar

• bservations by Walkden 2011, Gelderen 2013).
• There is a slight shift from the Early period to the Late period where

more of the null subjects are appearing in the matrix clause.

16

3

Distribution of Referential NS: subjunctive clauses

Subjunctive Clauses Not Subjunctive Clauses Total Subjects Periods Hits % Hits % Hits Early (<950) 830 72.1% 321 27.9% 1151 Late (>1000) 178 86.4% 28 13.6% 206

• Referential null subjects seem to be found primarily in subjunctive clauses.
• The count could potentially be higher, since this chart only counts verbs

explicitly marked as subjunctive (no ambiguous forms).

• The restriction to subjunctive clauses increases throughout the OE period.

17

Subjunctives (in main clauses) in OE

Faulkner 1902:42:

The subjunctive as the mood of desire occurs in [main] clauses to express a wish or a command. “The 3rd person, singular and plural, of the present subjunctive is used regularly as the representative of the 3rd person of the imperative. The ME translation is ‘let’ with the infinitive.”

18

Subjunctives in OE

Fischer & der Wurff 2006: 143-4: “The subjunctive and the modal verbs could used to express three types of modality in OE, usually referred to as deontic, dynamic and epistemic modality.” “The dynamic and deontic use of the modals in main clauses is well established in OE. This is not true for the epistemic use of modals.” “Instead of an epistemic modal verb we do find a subjunctive used even in the main clause in OE. […] It is likely that here the subjunctive lasted longer because the modal auxiliaries had not yet developed a clear epistemic function in OE.”: (1) …He gymde þy laes his agenra þearfa & wenunga hine sylfne forlete [SUBJ] …He heeded the less his own needs and probably him self neglected ‘…he cared about his own needs less and less and probably neglected himself.’ (GD2 (C)3.106.10)

20

Distribution of Referential NS (in Subjunctive Clauses)

Matrix Clause Subordinate Clause Total Periods Hits % Hits % Hits Early (<950) 767 92.4% 63 7.6% 830 Late (>1000) 167 93.8% 11 6.2% 178

• Null referential null subjects in subjunctive clauses occurs mostly in

matrix clauses.

21

Results: Loss of RefNSs in OE

• Referential null subjects are for the most part lost by end of OE period (overall

distibution below 2.7%, with 13 out of 20 texts with distribution below 2%.

• Six out of twenty texts (less than 7% of total data set for Late Period) have a more

conservative distribution of RefNSs, ranging between 7.9% and 69.6%.

• In fact, an analysis by text genre indicates that only four out of eighteen different genres

show a distribution of RefNSs above 5% over the entire OE period: Ecclesiastical laws, medical handbooks, laws and rules.

• These are also text genres that tend to show more conservative grammatical features.

22

Expletive Subjects

• Old English has a quasi-argument hit/it which turns into the overt

pure expletive seen in Middle English.

• This quasi-argument appears in constructions that use overt expletives

in Modern English

• impersonal constructions
• expressions of time, space, and distance
• Hit/it could be considered an overt expletive. They can appear in the

same contexts as the null expletive subjects.

24

4

Distribution of Expletive Subjects

Overt Expletives Subjects Null Expletive Subjects Total Expletive Subjects Periods Hits % Hits % Hits Early (<950) 455 21.4% 1668 78.6% 2123 Late (>1000) 89 16.9% 439 83.1% 528

• Null expletive subjects are still the more preferred construction throughout OE.
• However, overt expletives have a robust presence.

25

Null subjects and That-trace Effect

• In a classical proposal for a null subject parameter, Rizzi (1982, 1986) argued

that in Agreement-based null subject languages a cluster of properties followed:

• Referential (and expletive) null subjects licensed by agreement;
• Post-verbal subjects in overt syntax (VP-internally);
• Lack of that-trace effects (subject wh-extraction possible over overt complementizers).
• The that-trace effect is argued not to be enforced at the beginning of Middle

English (overt that is not entirely blocked in subject wh-extraction/relative structures).

• Old English shows signs of a lack of that-trace effect as well

Relative clauses allowed that+(subject) gap constructions

• However, the change introducing the that-trace effect might have originated

earlier in the Old English period: Zero+gap constructions allowed in OE.

27

Wh-extraction in the OE corpus

• Early Period:
• Number of wh-questions: 1001
• Number of subject extractions: 206
• Non-relative subject extractions: 116
• Late Period:
• Number of wh-questions: 274
• Number of subject extractions: 68
• Non-relative subject extractions: 32
• Of the non-relative subject extractions constructions

(in questions) in these two periods, all of them use zero complementizer+gap construction.

28

S-V Inversion

Inverted (XP-V-SU) Non-inverted (XP-SU-V) % uninverted Hits Hits Percentage Early (<950) 127 94 42.5% Middle+Late (>950) 187 62 24.9%

• Haeberli 2002: distribution of overt subject-verb order (this does not include all
• f the texts represented in previous tables).
• Inversion mainly happens with full DPs in Old English.
• In Haeberli’s data, there is an increase in V-S inversion from the Early to

Middle/Late period.

• At least during this period, there seems to be an independent development

from the path of loss of RefNSs.

29

Results: Loss of Referential NSs in OE

1) Referential null subjects have very restricted overall distribution in Old English. They are for the most part lost by end of OE period (overall distribution below 2.7%). 2) Remaining occurrences of referential null subjects are further restricted by the fact that they appear mostly in subjunctive clauses (86,4% of occurrences by Late Period > 1000) Lack of agreement: Subjunctives lack person distinction in OE (no person distinction; sg./pl. distinction only) Contra Gelderen (2013): who argues that verbal agreement with subject is linked to licensing of pro-drop in Old English. 3) Null subjects are also mostly restricted to main clauses - 90.8% by Late Period (see also Walkden 2011, 2012). Compatible with a topic drop analysis (see also Gelderen 2013), given (increasing) main clause restriction. Matrix distribution is even more significant for subjunctive clauses (93.8% in matrix by Late period). Restriction may be explained by matrix clause distribution of directive speech (if used e.g. as imperative) This is especially the case in a scenario in which Ref NSs are lost.

30

Results: Partial Null Subject Grammar of OE

4) Residue occurrence of of RefNSs falls below 2.7% by Late Period in OE, with 13 out of 20 texts with distribution below 2%). Six out of twenty texts (less than 7% of total data set for Late Period) that sill show some productive use of of RefNSs to also correspond to more formulaic, conservative text genres. 5) The system remains mostly stable regarding expletive null subjects throughout OE (83.1% of total expletive subjects in Late Period). 6) OE shows a split behavior. Overall it is a partial null subject language only with respect to null expletives, at the end of the OE period.

31