March 14th, 2014

Lectures de linguistique exp´erimentale (Paris 7)

Split Intransitivity and Variable Auxiliary Selection in Old French Heather Burnett (Universit´e de Montr´eal)1 [email protected]

1

Introduction

Today’s topic: Auxiliary selection with intransitives in Old French (12-13th centuries). Auxiliary Selection in ‘Standard’ Modern French: Auxiliary selection in French can be divided into three main cases: 1. Reflexive verbs, which take the auxiliary ˆetre. • Intrinsic reflexives (ex. s’´evanouir ). • True reflexives (ex. se convaincre). • Middle (what Abeille and Godard (2002) call medio-passive) reflexives (ex. Ces livers se vendent bien). 2. Some intransitive verbs take the auxiliary ˆetre (aller, venire, entrer, sortir . . . ). 3. The rest of the verbs take the auxiliary avoir. (1)

Dr & Mrs P. Vandertramp verbs: devenir, revenir, monter, rester, sortir, passer, venir, aller, naˆıtre, descendre, entrer, retourner, tomber, rentrer, arriver, mourir, partir

Auxiliary Selection in Old French: • Auxiliary selection in Old French shows a different pattern than in Modern French (Foulet, 1919; Moignet, 1973; Nordahl, 1977; Buridant, 2000, among others). (2)

Some MF ˆetre verbs can take avoir. a. Tant ont al´ e qu’il vinrent a la chit de Navers (Aiol, 8176) ‘Ils ont tant chemin´e qu’ils atteignent la cit´e de Nevers.’ b. Tant a venu De la gent qu’ele a retenu (Rutebeuf, Grieches d’´et´e, 27-28) ‘Il est tellement venu de gens qu’elle a pris `a son service.’ Cited from Buridant (2000) (pp. 373-374)

(3)

Some MF avoir verbs can take estre. a. Mais tot li chevalier ensamble i sont coru por lui rescorre. (Vengeance Raguidel, 33) ‘Mais ensemble tous les chevaliers s’y sont pr´ecipit´es pour le secourir.’

1

Portions of this presentation present joint work with Michelle Troberg (U. Toronto)

1

March 14th, 2014

Lectures de linguistique exp´erimentale (Paris 7)

b.

a la terre est vol´ es. ‘Il a vol´e jusqu’`a la terre.’ Cited from Troberg and Burnett (2014)

(Roman d’Alexandre, br. 3, 92)

Questions to be investigated: 1. What factors (morpho-syntactic or semantic) determine auxiliary choice in intransitives in Old French? • To what extent is the pattern that we see in OF different from that of MF? • To what extent is the pattern that we see in OF similar to that of MF? 2. What is the time course of the change from the OF system to the MF system? • Did the change in the OF auxiliary system coincide with other changes in the language? 3. Can studying the history of split intransitivity shed light on the source of the distinctive Modern French pattern? • Auxiliary selection in MF shows a different pattern than that of other Romance (and Germanic) languages. Plan: 1. Split intransitivity cross-linguistically. 2. What does auxiliary selection in Modern French look like, really? 3. Auxiliary selection in Old French. 4. A Pilot Study 5. Conclusion.

2

Split Intransitivity Cross-Linguistically

It is well-known2 that auxiliary selection (for those languages that have an alternation) is subject to much variation along two dimensions: 1. Cross-linguistically: Which verbs are members of the ˆetre-taking class. 2. Within a single language: Which verbs exhibit variable auxiliary selection. Proposal characterizing both of these dimensions of variation: The Auxiliary Selection Hierarchy (Sorace, 2000; Legendre and Sorace, 2003, among others). • Across languages, verbs can be ordered with respected to certain properties of their semantic denotations. • This hierarchy of verb meanings has consequences for both the definition of the split in intransitives and the categoricity of the split. 2

See, for example, the discussion of these points in Sorace (2000)

2

March 14th, 2014

Lectures de linguistique exp´erimentale (Paris 7)

(Sorace, 2000, p.863):

From Legendre (2007):

• Change of location: Inherently telic verbe of change (go, arrive, leave. . . ). • Change of state: 1. (Not necessarily telic) change of location verbs (rise, descend ). 2. (Not necessarily telic) internally caused change of state verbs (become, wilt, decay. . . ). 3. Verbs of appearance (appear, disappear ) and happening (happen, occur ). 4. die and be born. • Continuation of Condition: stative verbs that have an ‘implicit change component in their semantics’ (stay, remain, last, survive) – Agentivity often plays a role in acceptability with this class of verbs. (4)

Il presidente `e/ha durato in carica due anni. The president is/has lasted in office two days. ‘The president lasted in office for two days.’

• Existence of state: stative verbs with no implicit change (be, exist, belong, seem, suffice, please. . . ) • Uncontrolled processes: weather verbs, waver, tremble, skid, catch-on, cough, sneeze, sweat. . . 3

March 14th, 2014

Lectures de linguistique exp´erimentale (Paris 7)

– In some languages, agentivity of the subject is important (higher use of have). • Controlled motional processes: Manner of motion verbs, i.e. implying a non-directed displacement of their single argument (run, walk, swim. . . ). – In some languages, telicity plays an important role in auxiliary choice for all manner of motion verbs (i.e. Dutch: swim, run etc.) or a subset of them (i.e. Italian: run, not swim). – In Italian, agentivity of the subject is also important. • Controlled non-motional affecting processes: abdicate, join in, yield, triumph. – In Italian, agentive subjects take have, while non-agentive subjects often take be. • Controlled, unemotional, non-affecting processes: verbs that denote agentive, normally non-motional processes that do not generally affect their subjects (work, play, talk . . . ). – (Sorace, 2000, p.874):In some languages (varieties of Italian), some speakers find these verbs not completely unacceptable with be with non-agentive subjects.

2.1

Summary

Generalizations we can draw from the ASH: Three linguistic factors have been argued to be important in cross-linguistic studies of split intransitivity: 1. Verb class. 2. Telicity. 3. Agentivity of the subject. Question: • How do these three factors (or others) affect auxiliary selection in Modern French?

3

Split Intransitivity in Modern French

Observation: We find both variation in the composition of the ˆetre class and auxiliary alternations in varieties of Modern French.

4

March 14th, 2014

3.1

Lectures de linguistique exp´erimentale (Paris 7)

Variation in the composition of the ˆ etre class

The definition of the set of intransitive verbs/constructions taking ˆetre is notoriously variable depending on population and time period. 1. Some varieties of French, such as Acadian, uniformly select aou`er 3 , despite having a class of ˆetre-taking intransitive verbs. (5)

From the Peronnet 85 corpus (Giancarli, 2011, p.210) a. I s’avont pr´epar´e b. L`a ¸ca fait coume l’autre fois/i s’avont rencontr´e c. Al avait toute la face pleine de sang/l`a/ a s’a regrich´ e contre la servante/ ¸ca venait la ‘fight’

(INF 1) (INF 1) (INF 4)

2. Variation in the class of ˆetre intransitives, according to grammarians (from Willis, 2000, p.81) (see also Tailleur (2007)):

• As we will see, the set of verbs that can take ˆetre in Medieval French is also different from those in Classical French.

3.2

Variation in auxiliary selection with ˆ etre verbs

• In dialectal French, even verbs that lie very high up on the ASH show variation between ˆetre and avoir (Canale and B´elanger, 1978; Sankoff and Thibault, 1977; Russo and Roberts, 1999; Willis, 2000; Renaud and Villeneuve, 2008). (6)

Je suis revenu–j’ai revenu a` seize ans, j’ai revenu a` Ottawa. Ottawa-Hull Corpus (Willis, 2000, p.)

(055/153)

3 Actually, the rate of avoir use with reflexive verbs is 97% in the Peronnet 85 corpus (according to Giancarli, 2011).

5

March 14th, 2014

Lectures de linguistique exp´erimentale (Paris 7)

Verbs allowing both auxiliaries in identical semantic contexts (from Willis, 2000, pp.5253):

3.2.1

Factors that determine auxiliary selection in Modern French

There has been a large amount of quantitative work done on variable auxiliary selection in Modern North American French. • The studies in this area investigate the patterns of variation within the set of verbs that are classified as ˆetre verbs by grammarians (i.e. the verbs in (1)). – A subset of the set of intransitives in the language. • Examples of composed tenses are coded for a variety of linguistic factors, and the results are analyzed using variable rule analysis (mostly using Goldvarb). • (Social factors in auxiliary selection are also studied). Common linguistic factors investigated: 1. Transitive use: Whether or not the intransitive verb independently has a transitive use. • In the varieties of Modern French studied, transitive non-reflexive verbs categorically take avoir (not included in the analysis). • Russo and Roberts (1999); Willis (2000); Renaud and Villeneuve (2008): Having a transitive use favours the use of avoir in intransitives. (7)

J’ai sorti les valises.

J’ai sorti de la maison.

6

March 14th, 2014

Lectures de linguistique exp´erimentale (Paris 7)

2. Reflexive use: Whether or not the intransitive verb independently has a reflexive use. • In the varieties studied, reflexive constructions categorically take ˆetre. • Willis (2000); Renaud and Villeneuve (2008): Having a reflexive use favours the use of ˆetre in non-reflexive intransitives. (8)

On s’en est revenus (OH corpus: 018/381)

Je suis revenu a` la maison.

3. Complement type: Locational complement vs no complement vs infinitival complement. • Willis (2000): Having an infinitival complement favours the use of ˆetre. 4. Frequency: • Sankoff and Thibault (1977); Russo and Roberts (1999): Within the set of pre-defined ˆetre verbs, frequency favours the use of ˆetre. 5. Existence of independent adjectival use: If a verb has an adjectival use, then avoir is favoured (Russo and Roberts, 1999; Willis, 2000). • (Not clear to me how adjectival status was determined. . . )

3.3

Summary

• Typologically, a number of factors are important in determining whether ˆetre or avoir will be used with intransitives. 1. Verb class. 2. Telicity. 3. Agentivity of subject. • Within varieties of Modern French, in addition to verb class, different factors determine whether ˆetre or avoir will be used with intransitives. 1. Transitive use. 2. Reflexive use. 3. Complement type. What role do the factors that influence auxiliary selection in Modern French play in split intransitivity in Old French? • Do factors like agentivity and telicity play a role in auxiliary selection in Old French (despite their absence in Modern French)?

7

March 14th, 2014

4

Lectures de linguistique exp´erimentale (Paris 7)

Variation in Auxiliary Selection in Old French

Although variable auxiliary selection has been frequently studied in varieties of Modern French (see above) and occasionally studied in Classical French (Tailleur, 2007; Rideout, 2011), there is surprisingly little work on this topic in Old French (12-13th century). • Important work by Buridant (2000) (following (I assume) Nordahl (1977)) on auxiliary selection with aller and venir.

4.1

Buridant (2000) on variable auxiliary selection

According to Buridant, auxiliary choice in Old French is conditioned by the following factors (p.374): 4.1.1

Reflexivity

• Buridant proposes that pronominal verbs appear with estre, but there are exceptions. • He cites a study by St´efanini (1962) which puts the rate of avoir with pronominal verbs in the north-east and anglo-norman at 3% (suggests the existence of variation). 4.1.2

Verb Class Les verbes intransitifs emploient l’auxiliaire avoir ou l’auxiliaire estre selon le mode d’action des verbes : les verbes perfectifs, portant en eux-mˆemes leur finalit´e et tendant vers un ´etat, emploient normalement l’auxiliaire estre, marquant l’au-del`a de la limite de tension, les verbes imperfectifs, pouvant exprimer un processus ind´efini, emploient normalement l’auxiliaire avoir.

• In the words of Sorace etc.: Change of state verbs select ˆetre, while processes select avoir. • As Buridant notes, many (most?) change of state verbs in Old French are morphologically complex. L’opposition perfectif vs. imperfectif s’observe bien dans les couples de verbes pr´efix´es inchoativement, conjugu´es avec estre, oppos´es aux verbes simples, conjugu´es avec avoir: endormir vs. dormir, esveillier vs. veillier, seoir vs. aseoir. Hypothesis: • Unlike in Modern French, where lexical class is crucial, perhaps in Old French auxiliary selection is not tied to lexical class, but rather to telicity more generally.

8

March 14th, 2014

4.1.3

Lectures de linguistique exp´erimentale (Paris 7)

Telicity

As in languages such as Italian and Dutch, according to to Buridant, telic interpretations favour ˆetre with controlled motional processes. Avec un sujet anim´e dans l’immense majorit´e des cas, la prise en compte de l’aboutissement du proc`es entraˆıne le verbe estre, soit le terme du mouvement pour les verbes de mouvement au sens large, la prise en compte de la dur´ee du proc`es entraˆıne le verbe avoir. Question: How can we tell atelic interpretations from a corpus? • One way: through the distribution of degree modifiers (see also Doetjes, 1997; Caudal and Nicolas, 2006). • “modifiers associated with open scales such as a lot accept atelic dynamic predications, but reject all types of telic predications.” (Caudal and Nicolas, 2006, p.5). (9)

a. Yannig walked a lot. b. #Yannig ran to the store a lot. c. #Yannig left a lot. d. *Yannig ate his pancake a lot. (Caudal and Nicolas, 2006, p.5) Pour ces verbes, cette dur´ee est le plus souvent marqu´ee par des adverbes quantitatifs positifs, assez, molt, petit, un poi, troi et surtout tant (. . . ) que, n´egatifs ne... gueres, ne pas gaires, ne... pas longu´ement, ne. . . mie granment, ne. . . pas granment, ou des compl´ements indiquant une mesure de distance ou de temps, trait´ee comme r´egime direct: arbales tee, le trait d’un arc, une archie, une huchie, une journee, une lieue, demi lieu´e quatre, .xii. . . lieues, quatre pas, piece, grant piece, une grant piece. • Within the context of our discussion, we can decompose Buridant’s comment into two parts: 1. Telicity: atelic VPs can appear with degree adverbs and favour the use of avoir. 2. Transitivity: Constructions with quantized mesure phrases (while (usually) telic) are transitive5 . – Transitivity favours the use of avoir.

4.1.4

Agentivity

• According to Buridant, “estre est constant dans les constructions impersonnelles avec aler.”

5

(10)

Je te dirai, fet li anemis, cornent il est al´e de li (Queste, 120, 10) (Je vais te dire, reprend le Diable, ce qui lui est arriv´e)

(11)

Si ne sai cornent ce est al´e (Je ne sais comment cela s’est produit)

Or are they? Syntactic status of the mesure phrase ‘direct object’ ?

9

(ibid., 147, 18)

March 14th, 2014

Lectures de linguistique exp´erimentale (Paris 7)

• “avec un sujet inanim´e, estre est quasiment constant. Ainsi avec aler : - afaire, chose (aler “se d´erouler, se passer”).” (12)

La chose est tant alee que. . . (MortArtu, 110, 23; 112, 41; 146, 47) (La chose est all´ee si loin/est arriv´ee a` un point tel que...)

• Impersonal constructions and verbs with inanimate subjects are all non-agentive constructions. • If agentivity is an important factor in auxiliary choice in Old French (as in Italian, for example, where agentivity favours have), then the pattern described by Buridant could be a reflection of state of affairs.

4.2

Summary

Based on the description of Buridant, we should expect the following factors to play a role in auxiliary selection in Old French: 1. Reflexivity (being in a reflexive construction favours estre). 2. Telicity (being in a telic construction favours estre). 3. Transitivity (being in a transitive construction favours avoir ). 4. Animacy/Agentivity (having an inanimate subject favours estre). Goals: • Test Buridant’s (and other’s) proposals by means of a large data set. • Investigate whether other factors (such as the ones that condition auxiliary selection in Modern French) are operative in Old French.

5

Pilot Study

A pilot study of auxiliary selection in Old French. • Based on the Textes de fran¸cais ancien (TFA) corpus (around 3 million words). • Textes primarily from the 12th and 13th centuries6 . First question: 200 years is a long time and we know that auxiliary selection changed in the history of French. • Is auxiliary selection within the complete Old French period homogeneous enough for it to make sense to group data from speakers generations apart together in the analysis? Test of temporal homogeneity: Consider the verb aler. 6

In our analyses, we set aside the very few texts that are dated later than 1300.

10

March 14th, 2014

Lectures de linguistique exp´erimentale (Paris 7)

• 356 occurrences as a past participle in the TFA, 93 (26%) of which appear with avoir and 263 with ˆetre. • In Modern French, aler is one of the verbs which shows the most categorical ˆetre selection. • Using Rbrul (see below), I tested to see whether the date of the text had any significant effect on auxiliary choice. – If the changes in French auxiliary selection had started by the end of the 13th century, we would expect to see some effect of the date of the text on the choice of auxiliary. – But we find no such effect. Conclusion: The Old French period constitutes a relatively homogeneous period with respect to auxiliary selection.

5.1

Linguistic Factors

• What linguistic factors influence auxiliary choice in Old French? 5.1.1

Verb Class

In many languages, ASH verb class plays a large role in split intransitivity: • Although factors such as telicity and agentivity may play a role in some languages with some verb classes, auxiliary selection with other verb classes is fixed. Which verb classes should we look at? • Within the variationist methodology, we only want to study syntactic/semantic contexts that show variation. – In Modern French, variation in auxiliary selection limited to those verbs that lie at the change of state point in the ASH hierarchy or higher, so studies of this phenomenon in the modern language only look at those verbs. • In Old French, many more verbs show variable auxiliary selection, including: Continuation of state: (13)

Durer a. Tant con li contes est durez. ‘He stayed with the count so much.’ b. trop a ice dur´e: ‘He stayed here too much.’

(LancR: 216) (Beroul: 128)

Motional processes: (14)

a. b.

Tant a coru et porchaci´e; ‘He ran and chased so much.’ Quant.xiiii. Fran¸cois i sont couru ‘When 14 Frenchmen ran there’ 11

(Renart7: 10) (Aiol: 99)

March 14th, 2014

(15)

a.

b. (16)

a. b.

Lectures de linguistique exp´erimentale (Paris 7)

Et quant il avoit tant vol´e que toz li monz le tenoit a merveille (Queste: 131) ‘And when he flew around so much that everyone marvelled at him.’ a la terre est vol´es. (Alexandre2: 292) ‘It flew to the ground.’ et maintes fois i ai sailli ‘And I jumped there many times.’ si est en pi´es saillis. ‘So he jumped to his feet.’

(Eracle: 121) (Alexandre3: 144)

Possible Conclusion: We need to look at all intransitive verbs except non-motional process. (!?!) • Today: A selection of verbs drawn from (most of) the ASH classes. Puzzle: Many of the OF verbs that we are looking at don’t seem to have the same meaning as their modern counterparts. • Auxiliary selection makes them vary from one class to another. Aler: In many cases, when it appears with avoir, it appears to have a meaning more similar to a motional process than a change of location verb: (17)

Tant a al´e qu’a li revient.

(Coinci4: 387)

paroistre/paroir: With estre, it appears to mean something like the Modern French apparaˆıtre ‘appear’7 . (18)

et esvellier ains que jor fust paru ‘and to wake up before it became day.’

(Moniage: 48)

• One option: we are dealing with two lexical items: one change of state verb, and one existence of state verb (this is done by Tailleur (2007)). • Another option: we are dealing with a single lexical item with one basic meaning, which changes depending on the morpho-syntactic/discursive context. • Since there was only 1 token of tomber, 2 tokens of sembler and 4 tokens of paroistre, I didn’t include the change of state and existence of state classes in the statistical analyses. 7 The few examples of paroistre/paroir that we have all take estre, even when there is a stative meaning. So maybe this verb simply selects estre. . .

(i)

molt li est bien paru. ‘It appeared very good to him.’

(Alexandre2: 98)

12

March 14th, 2014

5.1.2

Lectures de linguistique exp´erimentale (Paris 7)

Telicity

Telicity is a semantic notion; as such, at first glance it may appear very tricky to determine from a historical corpus. • Fortunately for us, telic interpretations are syntactically constructed (Verkuyl, 1972), and the morpho-syntactic patterns that give rise to possible/preferred telic interpretations are well known. – Claim: It is possible to establish the telicity of the VP in a particular example from looking at its morpho-syntactic context (and also its discursive context). Structures giving rise to telic interpretations with intransitives: 1. Goal locational complements (Bounded locative PPs, locative clitics, locative DPs). (19)

a. b. c.

E Bertram est al governail al´e. Et cil i sont al´e: li fieus Elie est cele part al´es:

(ChGuill: 121) (Alexandre3: 146) (Aiol: 290)

2. Infinitival clauses. (20)

s’i est culcher al´e.

(ChGuill: 63)

3. Bounded/quantized measure phrases. (21)

vostre mestre a grant saut sailli.

(Thebes 1: 162)

Structures giving rise to atelic interpretations with intransitives: 1. ‘Open scale’ degree adverbs (see above). 2. Durational adverbials (for three days). (22)

car la nef ot tote la nuit coru et tout le jor a plein voile. (Queste: 200) Because the ship sped on the whole night and the whole day with full sails.

3. Unbounded spatial PPs. (23)

a. b. c.

Droit vers Gales s’en sont al´e. (Beroul: 66) Parmi chest bois ai al´e traversant Plus de set jours (MoniageG2: 152) car par tout ont al´e; (Alexandre3: 167)

4. Directional particles. (24)

si est avant al´es

(MoniageR: 174)

Can we always tell the telicity of a VP? Very little syntactic context: 13

March 14th, 2014

(25)

a. b.

Lectures de linguistique exp´erimentale (Paris 7)

Tout sont al´e: ‘All of them left’ ? Al´e avez en male veue ‘You left in a bad light.’ ?

(Alexandre4: 334) (Renart7: 50)

• I have assumed that examples with directed motion verbs with no kinds of complements or modifiers are instances of achievements (i.e. telic), but this may be imposing modern judgements on historical data. . . 5.1.3

Reflexivity

• In Modern varieties of French, use of ˆetre (or avoir in the case of Acadian) in reflexive constructions in categorical. • Studies of Modern French that investigate the effects of reflexivity on auxiliary selection code verbs according to whether they appear independently in a reflexive construction (i.e. Willis, 2000). In our Old French corpus, we see variation: (26)

a. b.

S’ont tant al´e qu’il vindrent pres D’un fort Droit vers Gales s’en sont al´e.

(LionMsH: 93)8 (Beroul: 66)

• For each occurrence of an auxiliary, I coded whether it appeared in (i.e. (26) = yes) (27)

il sunt avant al´e.

5.1.4

Transitivity/Complement

(Becket: 138)

• What kind of ‘complement’ the verb has: locative, DP, infinitival or null. (28)

a. b. c. d.

E Bertram est al governail al´e. (ChGuill: 121) n’e¨ ussi´es mie demie lieue al´e que maintenant l’ot gari et san´e. ‘You hadn’t even gone half a league that it had cured him.’(MoniageR: 265) s’i est culcher al´e. (ChGuill: 63) Tout sont al´e: (Alexandre4: 334)

• Note that we find variation in Old French, even in transitive constructions. (29)

a. b.

5.1.5

anchois fuissi´es demie lieue al´es (MoniageR: 261) ‘Before you went half a league.’ n’e¨ ussi´es mie demie lieue al´e que maintenant l’ot gari et san´e. ‘You hadn’t even gone half a league that it had cured him.’(MoniageR: 265)

Agentivity

• Agentivity contrasts of the kind found with animate subjects with non-motional affecting processes in Italian are tricky to establish in a historical corpus. • Animacy is easier to code for and (I think) could give a decent approximation of the role of agentivity in the data. 8

Although this could be an instance of an elided sentence initial adverb si. I thank Patrick Caudal for bringing this possibility to my attention.

14

March 14th, 2014

5.2

Lectures de linguistique exp´erimentale (Paris 7)

Analyses

The data were analyzed using variable rule analysis. • What’s variable rule analysis? A set of statistical tests that are ‘standard’ in Variationist sociolinguistics, i.e. the ones that are included in David Sankoff’s original Varbrul program from the 70’s (see Johnson, 2009; Tagliamonte, 2012, for discussion). • The analyses were performed using the Rbrul software9 , which is an extension of Sankoff’s Varbrul software, run in an R environment. • Varbrul (and Rbrul by extension) performs stepwise logistic regression on multiple categorical predictors (factors/factor groups) using the maximum likelihood method. • The variable rule program returns an input value for the data set: the overall tendency of the dependent variable to surface in the data. • The variable rule program assigns factor weights or probabilities to each category of the factor groups included in the analysis using what is called contrast coding 10 – Factor weights from 0.51-1 indicate that the application value (in our case estre) is favoured at this value; factor weights from 0.49-0 indicate that the application value is disfavoured at this value; and a factor weight of 0.5 indicates that the application value is neither favoured nor disfavoured at this value. • Unlike Varbrul, Rbrul allows for variables and predictors/factors to be continuous and allows for the inclusion of random effects, and so has many advantages over the latest incarnation of Varbrul: Goldvarb LION (see Johnson, 2009, for discussion). • We included text as a random variable in the analyses of both the social and linguistic factors.

5.3

Results

5.4

Discussion

Complement: • As in Modern French, having an infinitival complement strongly favours estre. – In fact, estre is categorical with infinitival complements (so should be excluded from variable rule analysis?). • Also as in Modern French, being in a transitive construction favours avoir in Old French, but the restriction is not as strong as strong. 9

http://www.danielezrajohnson.com/rbrul.html Contrast coding compares the mean of the dependent variable for a given factor within a factor group to the overall mean of the dependent variable. 10

15

March 14th, 2014

Lectures de linguistique exp´erimentale (Paris 7)

– Unlike in Modern French, where the restriction is categorical, in Old French, we find variation. Telicity Telicity plays a huge role in OF, unlike in Modern French. • But the effect is not categorical. (30)

Tant sont al´e et cil venu Qu’il s’entred¨ıent lor salu. (Beroul:87) ‘These guys came and went so much that they threatened their safety.’

(31)

Car Diex nos a a bone rive Ariv´e ‘Because God has brought us to a good bank.’

(Coinci1: 148)

Reflexivity and Verb class : Reflexivity and verb class may play a role in Old French beyond telicity. • We need more data to determine these things. • Interactions between verb class and others factors (animacy, telicity etc.)? Animacy: • Unlike in some languages, and what is suggested by Buridant (and others), we find no animacy (and perhaps therefore no agentivity) effect in the data. (32)

5.5

Comment les choses unt al´e. ‘How things went.’

(Estoire: 59)

Summary

In some respects, the phenomenon of auxiliary selection in Old French shows many similarities to its counterpart in Modern French. • Infinitival complements categorically select ˆetre/estre throughout history. • Transitivity and reflexivity play a role (matter of a difference of degree). Main difference: The role of telicity. • VP telicity is an important determinant of auxiliary selection in Old French, and this is no longer the case in Modern French.

6

Conclusion

A new empirical study of split intransitivity in Old French. Questions investigated today: 1. What factors (morpho-syntactic or semantic) determine auxiliary choice in intransitives in Old French? 16

March 14th, 2014

Lectures de linguistique exp´erimentale (Paris 7)

• To what extent is the pattern that we see in OF different from that of MF? – Telicity plays an important role in OF, no longer so in MF. • To what extent is the pattern that we see in OF similar to that of MF? – Complement choice, transitivity and reflexivity all affect auxiliary choice in OF (as in MF). Questions not investigated today: 1. What is the time course of the change from the OF system to the MF system? • Did the change in the OF auxiliary system coincide with other changes in the language? 2. Can studying the history of split intransitivity shed light on the source of the distinctive Modern French pattern?

6.1

Resultative secondary predication in Old French

Observation: • It’s not just that, with all manner of motion (‘controlled motional process’) verbs, auxiliary selection coded the (a)telic distinction, and then both atelic and telic VPs started taking avoir. – This happened with a certain subclass of MM verbs: courir, sauter/saillir . . . (33)

a. b.

Modern French: J’ai couru a` la maison. Pseudo-Old French: Je suis couru a` la maison.

• For the majority of manner of motion verbs, the telic interpretations simply disappeared. (34)

a. b.

Modern French: # Il a vol´e a` la terre. Old French: a la terre est vol´es

(Alexandre2: 92)

Burnett and Troberg (2012); Troberg and Burnett (2014): The loss of directional telic interpretations of manner of motion verbs is just one small aspect of the general loss of resultative secondary predication constructions in the language. • Old and Middle French had verb particle constructions. (35)

a. b. c.

le mers reportoit le nef ariere (Clari, p.74, de Dufresne et al. (2003)) reporter ariere le bouter jus (Cent Nouvelle Nouvelles, c.1456-1467, 452, du DMF2009) bouter jus Et toutevoies recort il sus au serpent (Qgraal, p.94, cit´e de Burnett and Tremblay (2009)) recourir sus

• Old and Middle French had adjectival resultative secondary predication constructions. 17

March 14th, 2014

(36)

a. b. c. d. e.

Lectures de linguistique exp´erimentale (Paris 7)

Et le despoillirent tout nuz. (La Passion d’Autun, 106; (DMF2009)) d´ epouiller nu Que tricherie abat jus plate. (Pizan. [1400]. Livre de la mutacion (DMF2009).) abatre plat Que mort l’a abatu et cravent´e. (Anon. [1210 (?)], Aiol. p.39 (TFA)) abatre et cravanter mort Li rois se taisi tout quois. (Froissart, 846.18593; (MCVF)) se taire coi tute quarree la fendi. (Marie de France. [1160-70]. Lais. p.183 (TFA)) fendre carr´ e

• Burnett and Troberg (2013): All these constructions were lost at the end of the Middle French period (beginning of the 16th century). Hypothesis: • Auxiliary selection in Old French was driven primarily by telicity, which, in the case of motion verbs was a product of productive resultative secondary predication in the language. • When Old French lost productive resultative secondary predication, the ˆetre/avoir alternation could no longer encode a telicity alternation (because there was none–or it was very limited). – Some verbs like aller, which mostly appeared in telic constructions, became (strongly) associated with ˆetre. – Some verbs like voler, chevaucher etc., which mostly appeared in atelic constructions, became (strongly) associated with avoir. • This process created the lexical verb-class effects that we know and love in French spoken today. Obviously we need to do a lot more work to test these hypotheses. . .

References Abeille, A. and Godard, D. (2002). The syntactic structure of French auxiliaries. Language, 78:404–452. Buridant, C. (2000). Grammaire nouvelle de l’ancien fran¸cais. Sedes, Paris. Burnett, H. and Tremblay, M. (2009). Variable-behaviour Ps and the location of PATH in Old French. In Aboh, E. and Quer, J., editors, Romance languages and linguistic theory 2007, pages 25–50. John Benjamins, Amsterdam. Burnett, H. and Troberg, M. (2012). L’´evolution des constructions r´esultatives dans l’histoire du fran¸cais. In Colloque international Le fran¸cais en diachronie (DIACHRO6), KULeuven. Burnett, H. and Troberg, M. (2013). The shape of change at the syntax-semantics interface: New evidence from Old French. Diachronic Generative Syntax Conference (DiGs). 18

March 14th, 2014

Lectures de linguistique exp´erimentale (Paris 7)

Canale, M. and B´elanger, M. (1978). Analogical leveling of the auxiliary ˆetre in Ontarian French. In ner, M. S., editor, Studies in Romance Linguistics, pages 41–61. Georgetown University Press, Washington DC. Caudal, P. and Nicolas, D. (2006). Types of degrees and types of event structures. In Maienborn, C. and W¨ollenstein, A., editors, Event arguments in syntax, semantics and discourse. Niemeyer, T¨ ubingen. Doetjes, J. (1997). Quantifiers and Selection. PhD thesis, Leiden University. Dufresne, M., Dupuis, F., and Tremblay, M. (2003). Preverbs and particles in Old French. Yearbook of Morphology, 2003:33–60. Foulet, L. (1919). Petite syntaxe de l’ancien fran¸cais. Champion, Paris. ´ Giancarli, P.-D. (2011). Les auxiliaries ˆetre et avoir: Etude compar´ee corse, anglais, fran¸cais. Presses universitaires de Rennes, Rennes. Johnson, D. (2009). Getting off the GoldVarb standard: Introducing RBrul for mixedeffects variable rule analysis. Language and Linguistics compass, 3:159–383. Legendre, G. (2007). On the typology of auxiliary selection. Lingua, 117:1522–1540. Legendre, G. and Sorace, A. (2003). Auxiliaires et intransitivit´e en fran¸cais et dans les langues romanes. In Godard, D., editor, Les langues romanes: probl`emes de la phrase simple, pages 185–233. CNRS editions. Moignet, G. (1973). Grammaire de l’ancien fran¸cais. Persee, Paris. Nordahl, H. (1977). Assez avez al´e: estre et avoir comme auxiliaires du verbe aler en ancien fran¸cais. Revue Romane, 12. Renaud, C. and Villeneuve, A.-J. (2008). L’alternance des auxiliaires avoir et ˆetre a` chicoutimi-jonqui`ere. Les fran¸cais d’ici. Rideout, D. (2011). Auxiliary selection in 16th century french: Imposing norms in the face of change. In Johanson, S. and Hazenberg, E., editors, Memorial University occasional papers in linguistics, pages 1–17. Memorial University. Russo, M. and Roberts, J. (1999). Linguistic change in endangered dialects: The case of alternation between avoir and ˆetre in vermont french. Language Variation and Change, 11:67–85. Sankoff, G. and Thibault, P. (1977). L’alternance entre les auxiliaires avoir et ˆetre en fran¸cais parl´e a` montr´eal. Langue fran¸caise, 34:81–108. Sorace, A. (2000). Gradients in auxiliary selection with intransitive verbs. Language, 76:860–890. St´efanini, J. (1962). La voix pronominale en ancien et moyen fran¸cais. Presses Orphys, Aix en Provence. Tagliamonte, S. (2012). Variationist Sociolinguistics: Change, Observation, Interpretation. Wiley-Blackwell, West Sussex. 19

March 14th, 2014

Lectures de linguistique exp´erimentale (Paris 7)

Tailleur, S. (2007). Avoir et ˆetre: une ancienne rivalit´e? ´etude de la s´election de l’auxiliaire de parfait en fran¸cais classique. Master’s thesis, University of Ottawa. Troberg, M. and Burnett, H. (2014). Le pr´edicat r´esultatif adjectival en fran¸cais m´edi´eval. Lingvisticae Investigationes, accepted. Verkuyl, H. (1972). On the compositional nature of the aspects. Springer, Dordrecht. Willis, L. (2000). ˆetre ou ne plus ˆetre: Auxiliary alternation in Ottawa-Hull French. Master’s thesis, University of Ottawa.

20

March 14th, 2014

Study Canale et al. (1978)

Lectures de linguistique exp´erimentale (Paris 7)

# tokens 533

Sudbury & Welland

Sankoff & Thibault (1980) Montr´eal

2650

% avoir (range) 3%-aller

Significant ling. factors None

86%-rester

(Tested: adjectival use frequency transitive use morph. properties) frequency

0.7%-aller 90%-passer

Sankoff & Thibault (1997)

Russo & Roberts (1999) Vermont Willis (2000) Ottawa-Hull

Renaud & Villeneuve (2008) Chicoutimi-Jonqui`ere

453

24%-venir 90%-retourner 8-d´ec´eder 98%-commencer

2000

704

8%-revenir 78%-passer

subject pronoun animacy of subject4 predicate type intervening element aux. tense transitive use frequency transitive use adjectival use intervening material type of complement reflexive use (Tested: frequency conjugation class shape of onset subject pronoun animacy of subject predicate type verb polarity auxiliary tense proximity in the past) verb subject pronoun transitive use reflexive use verb tense

Table 1: Previous studies on variable auxiliary selection in Modern French

21

March 14th, 2014

Lectures de linguistique exp´erimentale (Paris 7)

Verb class Change of location Change of state Continuation of condition Existence of state Uncontrolled processes Controlled motional processes

Verb aler, arriver (tomber) durer (paroistre, sembler) cheminer, chevaucher, courir marcher, nager, troter, voler

Controlled non-motional processes Table 2: Verbs classes for pilot study

Verb aller arriver cheminer chevaucher courir durer marcher nager saillir troter voler total

# avoir 93 2 1 24 10 39 1 5 5 1 1 182

# estre 263 54 0 0 16 1 0 0 61 0 5 400

Total 356 56 1 24 26 40 1 5 66 1 6 582

Table 3: Auxiliary selection by verb in the TFA

22

March 14th, 2014

Lectures de linguistique exp´erimentale (Paris 7)

Input Total N

0.69 582 factor weight

tokens

% estre

Telicity*** telic atelic Range

0.87 0.13 76

385 193

94 16

Complement*** infinitival locative null DP Range

>0.99 0.08 0.01 0.002 99

41 306 204 37

1 94 29 35

Reflexivity** Reflexive Non-reflexive Range

0.76 0.24 56

55 533

98 66

Verb Class*** Change of location Motional process Continuation of state Range

0.70 0.56 0.23 47

412 130 40

77 63 3

***: p < 0.001; **: p = 0.0301, ***: p = 0.0362 Table 4: Significant linguistic factors governing estre with intransitives in Old French

23