Reflexive verbs and the restructuring of

clitic clusters

Christine Meklenborg Salvesen, University of Oslo

Abstract

In Old French, all clitic clusters containing objects observed the

order ACC–DAT. During the 15th and 16th centuries this order was

changed into DAT–ACC in cases where objects of the 1st and 2nd

person were involved. This change took place rather abruptly. In this

paper I will argue that increased use of reflexive forms provoked a

change in the order in these clitic clusters. More specifically, I will

argue that clusters involving 1st and 2nd person argumental clitics

form true clusters in Modern French (in the sense of Pescarini (2012)),

whereas they formed split clusters in the old language.

1 Introduction

In the history of the French language, the order of object clitics has changed.1

In the old language, the accusative clitic always preceded the dative one
(ACC–DAT). In Modern French, this is still the order for clitic clusters
1
I would like to thank the following for their comments to the present paper: Terje
Lohndal, Marios Mavrogiorgos, Marie Labelle, Hans Petter Helland, and Diego Pescarini.
Parts of the data were presented at the XX International Conference on Historical Lin-
guistics in Osaka, July 2011, and I am grateful for the feedback I got from the audience
there.

1
Reflexive verbs and the restructuring of clitic clusters 2

containing only 3rd person clitics, but in the cases where the 1st and 2nd
persons are involved, the dative precedes the accusative (DAT–ACC).2 The
difference between the two orders may be illustrated in (1-a) and (1-b).

(1) a. et il la me dona.
and he her.ACC me.DAT gave
‘and he gave her to me.’ (12th c., Graal, 6677)

b. Il me la donne.
he me.DAT her.ACC gives
‘He gives her to me.’ (Modern French)

This shift in the internal order of the clitic cluster is not restricted to French.
Similar changes have taken place in Northern and Central Italian dialects
(Aski, in press; Pescarini, 2012), Catalan, Occitan (Brusewitz, 1905), and
Provençal (see Wanner (1974) for a general overview). Unlike the other

languages, French has not undergone a complete change from ACC–DAT

to DAT–ACC as clusters involving only 3rd person still display the ACC–
DAT order. Nor did the changes take place at the same time in the languages
concerned. I will however not look at the other Romance languages in this

article, but focus on French.

I will argue that the change in the internal order of the clitic clusters
must be seen in relation to the expansion of the reflexive pronoun. As the
sequence subject + reflexive pronoun becomes more common, this leads to
a reanalysis which requires the form corresponding to the reflexive pronoun

to occur in first position. This in turn led to a change of the internal order of
clitic clusters. I base my analysis on large electronic corpora and exploit a
2
I use the term clitic in order to describe an element which does not have any phono-
logical independence and which needs to attach to a host. In Old French the host is always
a verb.
Reflexive verbs and the restructuring of clitic clusters 3

generative approach in my formal analysis.

In order to investigate the clitic clusters, I have used three different on-
line databases: the Corpus de la littérature médiévale (CLM) (Lalou et al.,

2001), Frantext, and Base de Français Médiéval (BMF).3 The CLM has
texts stretching from the 11th to the 16th centuries (approximately 900
texts, number of words unknown), whereas the BFM covers the period from
842 till the end of the 15th century (3,784,304 words). In addition, I’ve
used a subselection of the Frantext base for the clitics corpus (selection

of 5,862,989 words from the 16th century). References in the CLM and
the Frantext base refer to the date the manuscript was originally written,
whereas dating in the BFM refers to the manuscript in question.4
I have used the Frantext database and the CLM in order to establish a

database of clitic clusters involving the 1st and 2nd persons, stretching from
the 11th to the 16th century. This corpus counts a total of 5136 examples.
The article is organised as follows: First I briefly describe the difference
between Old and Modern French in section 2 before presenting previous

research in section 3. Section 4 presents the quantitative data, whereas sec-

tion 5 investigates the rise of the reflexive. In section 6 I give a theoretical
account of the changes, before concluding in section 7.

2 Two different systems

Pronominal objects are clitics in all stages of the French language, as far
back as we have written sources. In (2), taken from a 9th century poem,
3
CLM: http://www.classiques-garnier.com/numerique-bases/; Frantext:
www.frantext.fr; BMF: http:txm.bfm-corpus.orgbfm/
4
This is especially relevant in the oldest period, as the oldest documents have not been
preserved other than in copies. A text such as the Strasbourg oaths, will thus be labelled
9th century in the CLP base, but 10th century in the BFM.
Reflexive verbs and the restructuring of clitic clusters 4

the pronouns precede the verb, and the two are often written together as one
single word: lafaire ‘her-make’ instead of la faire. This is a clear graphical
reflex of the pronunciation of the time. The pronominal objects occur in the

same position they do in later Old French texts.

(2) Uoldrent la-faire diaule seruir

would her-make devil serve
‘They wanted to make her serve the devil.‘ (9th c., Eulalie, 4)

In Old French (OF) (9th – 13th c.) the accusative object pronoun preceded

the dative one: ACC–DAT. Thus, in (3), the accusative la ‘her’ precedes the
dative te ‘you’.5

(3) Pren la a feme, je la te vuel donner.

take her.ACC to woman I her.ACC you.DAT will give
‘Marry her, I give her to you.’ (12th c., Raoul)

The order ACC–DAT is observed also in cases where the clitic is a reflexive
pronoun. If the reflexive functions as a dative, it will follow the accusative
pronoun.

(4) onques ne le se pensa

never NEG it.ACC himself.DAT thought
‘He never thought about it.’ (13th c., MortArtu, II, 13)

The possible constructions in Old French are represented in figure 1. Clitics

from the first column always precede clitics from the second and are always
accusative. Clitics in the second column are either dative or oblique (1st

and 2nd persons). In a clitic cluster containing pronominal clitics from both
5
In Old French, the clitics preceded the modal auxiliary, contrary to Modern French,
where they cliticise on the infinitive. I will come back to this later in this section.
Reflexive verbs and the restructuring of clitic clusters 5

columns, the second clitic must be dative.

person 1 2
1st sg me ’me’
2nd sg te ’you’
3rd sg le li ’him’
3rd sg la li ’her’
3rd sg & pl se se reflexive pronoun
1st pl nos ’us’
2nd pl. vos ’you’
3rd pl. les lor ’them’

Figure 1: Clitic order in Old French

In addition to this ordering constraint, there is a ban on combinations

where a a 3rd person dative object is combined with a 1st and 2nd person
accusative object (Laenzlinger, 1993). In the literature this is known as
the Person–Case Constraint (PCC) (see among others Bonet (1991, 1994);
Boeckx (2000); Ormazabal and Romero (2007); Manzini (2012)). I have

specifically looked for examples that would violate this constraint, without
finding any. I thus conclude that the PCC holds also for Old French. There
are no examples where both the clitics are in the 1st or 2nd person either,
although such examples have been reported in the literature.

In Modern French the system is more complex, as clitics of the 1st and
2nd persons always precede clitics of the 3rd person, while 3rd person cl-
itics have kept the order ACC–DAT. Just as in Old French, the PCC is ob-
served, which essentially implies that clitics of the first column may not be
combined with clitics of the third. Emonds (1975) has explained this by

proposing that pronouns of the 1st and 3rd columns are base-generated in
the same position, and that the pronouns appearing in the first column have
Reflexive verbs and the restructuring of clitic clusters 6

been moved in front of ones in the second. We will come back to this line
of thought in section 6.

person ACC/DAT ACC DAT

1st sg me ’me’
2nd sg te ’you’
3rd sg le lui ’him’
3rd sg la lui ’her’
3rd sg & pl se reflexive pronoun
1st pl nous ’us’
2nd pl. vous ’you’
3rd pl. les leur ’them’

Figure 2: Clitic order in Modern French

In addition to the personal pronouns, there are different pronominal ad-

verbials that also function as clitics. I will come back to these in section
5.
The difference between the two systems becomes clear when we com-
pare examples from the 12th century with their modern counterparts.

(5) le me > me le

a. Se il ne le me tolent, ele ert en

if they NEG it.ACC me.DAT take-away she be.FUT in
vostre non
your name
‘If they don’t take it away from me, she [the city] shall carry
your name.’ (12th c., Antioche, 4549)

b. S’ ils ne me le volent pas ...

if they NEG me.DAT it.ACC steal NEG
‘If they don’t steal it away from me ...’ (Modern French)

(6) le te > te le

a. Jo le te cargerai
I it.ACC you.DAT charge.FUT
Reflexive verbs and the restructuring of clitic clusters 7

‘I will charge you with it [the task].’ (12th c., Antioche, 5366)
b. je te le charge
I you.DAT it.ACC charge
‘I charge you with it [the task].’ (Modern French)

(7) le nos > nous le

a. Damedex le nos dont.

Lord-God it.ACC us.DAT give
‘The Good Lord gives it to us.’ (12th c., Aspremont, 5366)

b. Dieu nous le donne.

God us.DAT it.ACC gives
‘God gives it to us.’ (Modern French)

(8) le vos > vous le

a. bien le vos doi conter

well it.ACC you.DAT shall tell
‘I shall tell it to you well.’ (12th c., Aliscans, 2942)
b. Je dois bien vous le conter
I shall well you.DAT it.ACC tell
‘I shall tell it to you well.’ (Modern French)

Note that nothing happens in clusters where both clitics are in the 3rd per-

son, as in (9) and (10).

(9) le li > le lui

a. Et cil le li dïent
et these it.ACC him.DAT say
‘And they tell it to him’ (12th c., Yvain, 5021)

b. et ils le lui dient

and they it.ACC him.DAT tell
‘And they tell it to him’ (Modern French)

(10) le lor > le leur

Reflexive verbs and the restructuring of clitic clusters 8

a. Ypomedon le leur desfent

Ipomedon it.ACC them.DAT deny
‘Ipomedon refuses it to them.’ (12th c., Thèbes, 6617)
b. Il le leur défend
he it.ACC them.DAT deny
‘He refuses it to them.’ (Modern French)

The clitics attach to verbs, much as they do in Modern French. In the oldest

texts, clitics obey the Tobler-Mussafia law, which means that they may not
occur in the first position of the clause (Hirschbühler and Labelle, 2000;
Labelle and Hirschbühler, 2005). In addition, clitics in the old language
cliticised to the modal auxiliary, whereas they cliticise to the infinitive in

Modern French.
In the following, I will not discuss the clusters’ position in relation to
the verb, but focus on the internal order of the clitic clusters.
To sum up: Pronominal objects are always clitics, and they clitisise to

the finite verb. In Old French clitic clusters are always ACC–DAT. In Mod-
ern French the order ACC–DAT is only observed if both the clitics are in
the 3rd person. All other cases are DAT–ACC. The PCC is observed in all
stages of the language.

The question is why this change took place. The system in Old French
was extremely simple, whereas the modern pronominal system is a lot more
complex. Why would a language exchange a simple system for a more
complex one? For a child learning the language there must be a strong
trigger for the system to change. The question is quite simple: What was

this trigger?
Reflexive verbs and the restructuring of clitic clusters 9

3 Previous research

The change in clitic clusters has intrigued researchers since the late 1800s.

Most acknowledge that more than one explanation must be evoked in order
to account for the change that took place, and why the 3rd person was not
affected. Generally researchers point to a combination of phonetic/prosodic
principles and syntactic/distributional ones in order to explain the change.

In the following section I describe the most influential explanations based

on some of the major researchers in the field.

3.1 Phonological explanations

Most researchers seem to agree that the shift has to do with phonological or
rhythmical changes. The first one to investigate the phonetic changes was
Jung (1887, cited by Brusewitz (1905) and de Kok (1985)), who claimed
that the pronouns could be divided into three groups according to their

sonority (‘Klangfülle’). Their organisation in the sonority hierarchy is as

follows (from little sonority to much sonority): me, te, se, le, la > les, nous,
vous > lui, leur. According to Jung, the language evolved into a system
where more heavy (sonorous) forms would be placed to the right, and lighter

ones to the left. According to this analysis, the cluster les me should be the
ones to change the first. Looking at my data, I find no evidence that would
support such a claim. There is no significant difference in the evolutionary
pace between me le-clusters and me les-clusters. An other issue is that it

is hard to grasp the notion of sonority or heaviness. According to Togeby

(1979), the final -s was lost some time in the 13th and 14th centuries, chang-
ing the pronunciation from [les] to [le], [nus] to [nu], and [vus] to [vu]. It is
Reflexive verbs and the restructuring of clitic clusters 10

hard to see any substantial difference between me, le, te, and se on one hand
and [le], [nu] and [vu] on the other.
Rhythm is the key-word for several other researchers. Meyer-Lübke

(1899: 801) suggests that there was a change in the rhythm of the language
which led to a pattern where two short syllables preceded a long, stressed
one. This would explain the order le lui, and also explain why le nous was
maintained longer than le me. It is, however, difficult to see how stress
could relate to the clitic order, as clitics – by definition – must be unstressed

(unless they receive secondary stress). It is equally difficult to see why such
a change would lead le me to change to me le, as le and me are simply a
minimal pair. Zink (1997) is also interested in stress, and he proposes that
a shift in the accent has played a major role. In the oldest texts, je + le ‘I +

him/it’ form an enclise: jol (11).

(11) Jo-l te dirrai tot veirement

I-it.ACC you.DAT say.FUT all truthfully
‘I will tell you it all truthfully.’ (12th c., Adam 145)

There is a syllable boundary between jol ‘I+it’ and te ‘you’. Towards 1300

this enclisis disappears, and the first clitic becomes proclitic on the second
one. This way the syllables are distorted: [jel] // [me] → [je] // [mle]. Zink
(1997) claims that [mle] (or [tle] in the case of te ‘you’) represented a pho-
netic challenge to the speakers, and a way to resolve this was to reorganise

the cluster: [mle] should be easier to pronounce than [lme]. This explana-
tion comes short of explaining what necessitated this shift: Why couldn’t
the speakers just pronounce the clusters [lem] or even [leme]? In fact, they
may have done so, as we do not know exactly how these clusters were pro-
Reflexive verbs and the restructuring of clitic clusters 11

nounced.
A related approach has been adopted by de Kok (1985). Kok suggests
that the clusters originally (prior to Old French) were enclitic to a preceding

element. The first change was the one from enclisis to proclisis in the tran-
sition to Old French. Then a syntactic principle came into play, requiring
that the direct object be closest to the verb (see below). In sum these two
changes provoked the change observed in the data. Culbertson (2009), who
doesn’t comment on the cluster order, also advocates an analysis where the

clitics are enclitic to the first element in the clause. It is however necessary
to note that in Old French, the pronominal subject is not obligatory, and any
XP may precede the verbal cluster. There are however no instances where
the clitic seems to cliticise to a preceding adverb or full DP subject, only to

pronominal subjects. The question of enclisis in the earliest texts is thus not
fully examined.

3.2 Syntactic explanations

The idea that the direct object is more tightly linked to the verb and that
this explains the ACC–DAT > DAT–ACC-shift has many adherents (Jung
(1889), cited by Zink (1997)), Meyer-Lübke (1899); Wanner (1974); de Kok

(1985). Meyer-Lübke (1899) suggests that the change in clitic order took
place after a shift from pragmatic to syntactic principles. In the earliest
times it was natural to have the accusative pronoun in the first position, as
this carried information already mentioned in the context. The dative pro-
noun would often refer to someone new, possibly accompanied by a gesture.

This way it was more natural to have it in second position. When the shift
took place, it was the grammatical relation between the verb and its comple-
Reflexive verbs and the restructuring of clitic clusters 12

ment that prevailed: as the direct object is more closely linked to the verb
than the indirect object, it is natural that it occupies the position closest to
the verb, specifically immediately in front of it in declaratives, immediately

after it in imperatives. In order to explain why there has been no change in

the cases where both clitics are of the 3rd person, Wanner (1974) points to
the the language’s oxytone intonation which ‘requires a ‘heavy’ element to
be placed last in a linear arrangement (...)’ (Wanner, 1974: 172). It could
be argued that the forms lui, lor, and leur are heavier than the others as they

are bi-moraic. The form li is however mono-moraic, and heaviness cannot

explain why it always comes last in a clitic cluster.6
The distributional approach begins with Brusewitz (1905: 32–35) who
pointed out that there is a connection between the emergence of new reflex-

ive verbs and the change of the internal order of clitics. The idea is that the
increasing number of constructions of the kind je me ‘I myself,’ tu te ‘you
yourself’ and il se ‘he himself’ led to the obligatory fronting of the reflexive
clitic in the clitic cluster, which in turn entailed the reorganising of ordinary

clitic clusters without reflexives. An argument in favour of this analysis is

that the restructuring took place only in the cases where the reflexive pro-
noun and the dative pronoun have the same form. In the 3rd person, where
the reflexive form is distinct (se ’himself/herself/themselves’), no restructur-
ing took place. This is the idea that I will explore in section 5. The strength

of my approach compared to previous research is that I combine empirical

data exploited from large data bases and use them in a formal analysis.
6
I thank Marie Labelle for pointing out the difference between bi-moraic and mono-
moraic constructions.
Reflexive verbs and the restructuring of clitic clusters 13

4 An abrupt change in the 15th century

In the literature, it is not uncommon to see the claim that the change in

the order of the clitic clusters in Old French was very slow (Wanner, 1974;
Zink, 1997). Their argument is that there are occurrences of the new clitic
order already in the 12th century, but that the change wasn’t carried through
until the 18th century. I will take the opposite stand: The clitic cluster order

was remarkably stable until the 15th century. As figure 3 shows, the use
of the new order is marginal from the 11th through the 14th centuries. It
suddenly increases in the 15th century, before it becomes the most common
clitic order in the 16th.

century DAT–ACC total

no. ratio
11th 0 0.0 27
12th 11 0.02 518
13th 12 0.01 1044
14th 20 0.02 1278
15th 158 0.19 849
16th 835 0.59 1420
Figure 3: The emergence of the new clitic order

4.1 Internal differences

Wanner (1974) provides time charts for the transition in both Old Italian and

Old Provençal. He shows that it is clusters which involve the 1st person sin-
gular that change first. His claim is however not supported for Old French,
which becomes clear when the data in figure 3 are broken down further, as
in figures 4 and 5. The p-values in the tables have been calculated using a

Fisher-test.
Reflexive verbs and the restructuring of clitic clusters 14

12th century 13th century 14th century

new total ratio new total ratio new total ratio
ME 3 155 0.019 6 239 0.025 4 279 0.014
TE 1 31 0.032 1 56 0.018 6 123 0.040
NOS 0 26 0 0 57 0 0 144 0
VOS 7 201 0.035 5 529 0.009 10 528 0.019
p=0.7441 p=0.2481 p<0.05
Figure 4: The pace of the change

15th century 16th century

new total ratio new total ratio
ME 68 200 0.34 304 401 0.758
TE 34 102 0.333 176 214 0.822
NOS 17 118 0.144 149 254 0.587
VOS 39 346 0.113 206 543 0.379
p < 0.05 p < 0.05
Figure 5: The pace of the change

There is no statistically significant difference between the four persons

in the 12th and the 13th centuries. In the 14th century, it is essentially the
lack of examples with the new word order in the nos-clusters that explains
the p-value. A Fisher-test based on the values for only me, te, and vos gives
a p-value > 0.05. The difference is thus not significant.

When we consider table 5, we see that the ratio of te-clusters in the 16th
century is higher than the ratio of me-clusters. Apparently this contradicts
the tendency found in Old Italian and Old Provençal. There is, however, at
no point a significant difference between the te-clusters and the me-clusters.

In the 16th century, the p-value is 0.5255, indicating that the difference in
ratio may be the result of pure chance, probably because there are a lot more
me-clusters than te-clusters in the corpus. If we compare me/te-clusters on
the one hand and nos/vos-clusters on the other, there is a significant differ-

ence between them in the 16th century (p-value < 0.05). We may thus claim
Reflexive verbs and the restructuring of clitic clusters 15

that clusters involving the 1st and 2nd person singular change before the 1st
and 2nd person plural, but we are not at liberty to claim that the clusters
involving the 1st person singular change prior to the 2nd.

We may thus conclude that the internal order of the clitic clusters was
quite stable until the 15th century, and the change towards the new order
took place within one or two generations. The data indicates that clusters
involving the 1st and 2nd persons singular started changing prior to clusters
involving the plural pronouns.

5 Rise of the reflexive

Brusewitz (1905) was the first to see a link between the increased use of

reflexive verbs and the change in the clitic clusters. In the earliest texts, the
position of the reflexive was determined by its syntactic role. In (12) the
reflexive se ‘himself’ follows the accusative le ‘it’.

(12) a. Cil le se tolt et cil

that-one.NOM it.ACC REFL.DAT lifted and that-one.NOM
le s’ emble
it.ACC himself.DAT stole
‘One lifted it up before him, and another one stole it.’

(13th c., Coinci, 1725)

b. onques ne le se pensa
never NEG it.ACC REFL.DAT thought
‘the thought never occurred to him’

(13th c., MortArtu, 14)

Consequently, the old system did not distinguish between form, only func-

tion. The function accusative always preceded the function dative, regard-
Reflexive verbs and the restructuring of clitic clusters 16

less of their form. In this way, Case dictated the position of clitics.
Brusewitz’ idea was that an increased use of reflexive forms led to a re-
analysis where the sequence subject pronoun + reflexive pronoun became

lexicalised, regardless of the syntactic function of the reflexive pronoun. In

(13) the reflexive occurs in first position of the cluster, immediately follow-
ing the subject.

(13) il me semble que je me le prescry

it me seems that I REFL.DAT it.ACC prescribe
‘It seems to me that I myself require it.’

(16th c., Mont-vanité, p 967)

As a next step, the clusters were reanalysed so that all forms of the pronouns
which have identical forms as the reflexive pronoun had to come first in the
cluster, also in the cases where they did not function as reflexive markers.
In the case of the 3rd person, the reflexive has a distinct form, so there was
no trigger for a restructuring of the clitic clusters involving the 3rd person.

The process is schematised in figure 6.

person REFLEXIVE ACC DAT

1st sg me me me ’me’
2nd sg te te te ’you’
3rd sg se le li ’him’
3rd sg se la li ’her’
1st pl nos nos nos ’us’
2nd pl. vos vos vos ’you’
3rd pl. se les lor ’them’

Figure 6: New clitic order

Zink (1997) claims that in order to prove this theory, it is necessary

to show that the clitic clusters first changed in the cases where one of the
Reflexive verbs and the restructuring of clitic clusters 17

pronouns is a reflexive. In the electronic corpora we have available, we

have not found that the first occurrences of the new clitic order appear with
reflexive verbs. Obviously, this may be one of the shortcomings of using a

corpus, namely that no matter how big a corpus, it only shows fractions of
actual or possible utterances. A different way of approaching this problem
is to say that the increase in the use of reflexive forms led to an analysis
where all forms which were identical to the reflexive form had to come first,
regardless of the type of verb. This is the line of reasoning I will follow. A

consequence of this is that I will not try to establish strong causality between
the two changes. I will rather show how the use of reflexive verbs became
more widespread and that this change happened immediately before, and
partly overlapping with, the change of the internal order of clitic clusters.

In the next section I will show that both the use of reflexive forms and
their frequency increased from the 12th to the 15th century. This increase
slightly precedes the change in the clitic clusters, and I will argue that the

two changes are linked.

5.1 Reflexive verbs

In order to test Brusewitz’ hypothesis, I searched for the sequence s’en ‘him-
self’/‘herself of-it’ from the late 11th and 12th centuries (group 1) and 15th
century (group 2) in the BFM base. This is a clear-cut instance of a reflex-
ive pronoun in front of an adverbial pronoun clitic. The reason I chose s’en

and not just se is that se may also be the hypothetic subordinator (‘if’) in
Old French, and in Middle French it may also be a particle in the left pe-
riphery (corresponding to si in the earliest texts). As a reflexive, se may be
both accusative and dative (with transitive verbs), and it may be an inherent
Reflexive verbs and the restructuring of clitic clusters 18

reflexive (with intransitive verbs).

En is what is described as a pronominal adverb in traditional French
grammar. Derived from Latin inde ‘from there,’ it may either be an adverbial

clitic meaning ‘from there’ (14-a) or an anaphoric pronoun clitic replacing

an indirect object introduced by the preposition de ‘of’ (14-b). As such, en
may appear with all kinds of verbs: transitive and intransitive.
In all stages of the language en must follow se in a clitic cluster (14).

(14) a. Vait s’ en li pople.

go REFL from-here the people
‘The folks leave.’ (12th c., Alexis, 601)

b. Il s’ en va
he REFL from-here go
‘He’s leaving.’ (Modern French)

5.1.1 Intransitive verbs

In order to assess the changes that have taken place, it is necessary to make
a distinction between the different types of verbs. As Perlmutter (1978)
pointed out, there are two different kinds of intransitive verbs, those whose
subject is generated as an internal argument and those whose argument is

generated as an external argument. The former verb is unaccusative (15-a),

the latter unergative (15-b).

(15) intranstive structures

a. unaccusative: [vP [V P NP Vo ]]

b. unergative: [vP NP [V P Vo ]]
Reflexive verbs and the restructuring of clitic clusters 19

Être ‘be’ is the most common auxiliary with unaccusative non-reflexive

verbs in French, but a number of them also take avoir ‘have.’ All reflex-
ive verbs take être ‘be’ as their auxiliary, but not all reflexive verbs are un-

accusative (Legendre and Sorace, 2003). In order to distinguish between

unergative and unaccusative reflexive verbs it is therefore necessary to con-
sider their behaviour in different environments. The difference between un-
accusative and unergative verbs becomes quite clear when one looks at par-
ticipial constructions (Helland, 2001; Legendre and Sorace, 2003). Whereas

the past participle of an unaccusative verb may enter into participial con-
structions such as free predicatives (16-a) or absolute constructions, and as
object predicatives, the participles of unergative verbs may not (16-b).

(16) a. Les Dupont partis, toute la famille se mit à table.

the Dupont left all the family REFL put at table
‘ When the Duponts had left, all the family gathered around
the table.’ (From Helland (2001))

b. *L’ ouvrier travaillé, il est fatigué.

the worker worked he is tired
‘The working man who had worked is tired.’

In addition to the purely syntactic criteria, I have taken semantics into ac-
count. This has been particulary important when distinguishing between
unaccusative and transitive verbs. Quite a number of verbs can be both
transitive and intransitive. Just like unaccusative verbs, transitive verbs can

occur in participial constructions. It is thus challenging to determine the

nature of a verb such as torner ‘turn.’ Used as a transitive verb, it signifies
‘turn,’ as in ‘turn the page.’ Used as an intransitive verb, it signifies ‘turn
back’ or ‘return.’
Reflexive verbs and the restructuring of clitic clusters 20

It is common to associate unaccusative and unergative verbs with cer-

tain semantic or aspectual features. Cennamo (1999), discussing intransi-
tive verbs in Late Latin, suggests that there is an unaccusativity hierarchy

and an unergative hierarchy. At top of the unaccusative hierarchy are the

change of state verbs, then come the change of location verbs, and in the
lowest position the state verbs. At the top of the unergative hierarchy, one
finds the mental process verbs, then follow speech act verbs, and finally
other activity verbs.

In the case of a verb such as torner ‘turn,’ it is thus the fact that the
verb signifies a change of location that has led me to label it unaccusative.
This way torner has (at least) two different listings in the lexicon: one as a
regular transitive verb, the other as an inherent reflexive unaccusative verb.

Discussing Italian data, Burzio notes that ‘inherently reflexive alternations

would result from alternations in the semantic role of the subject, varying
between an external agent, cause or experiencer (transitive), and an internal
one (inherent reflexive)’ (Burzio, 2008). (See also Levin and Rappaport Ho-

vav (1995) for a discussion of this kind of variation.)

In certain cases, such as with torner, I have added the reflexive se to
mark that there is a clear-cut change in meaning between the reflexive and
non-reflexive use of the verb in question. Intransitive verbs are listed in
figure 7; transitive verbs in figure 8.

In order to observe the diachronic change that has taken place, different
formatting has been used in the figures. Underlined verbs are only attested
in the 12th century; verbs in bold are attested both in the 12th and the 15th
centuries; verbs with no special formatting (other than italics) are only at-

tested in the 15th century. The spelling of the verbs and their translations
Reflexive verbs and the restructuring of clitic clusters 21

are based on Hindley et al. (2000).

Unaccusative verbs
afiler ‘flow’; aler ‘go’; atorner ‘turn’; avancier ‘go forward’; comovoir
‘move’; departir ‘leave’; embler ‘vanish’; entrer ‘enter’; eschapper
‘escape’; s’esloignier ‘move away’; espandre ‘spread’; fuir ‘escape’;
issir ‘leave’; mesaller ‘go astray’; partir‘leave’; raler ‘return’; refuir
‘run back’; se repairier ‘return’; retorner ‘return’; revenir‘come back’;
revoler ‘fly off’; se torner ‘turn’; se traire‘move’; venir ‘come’; voler
‘fly’.
Verbal periphrases
aler ‘go’ + gerundive, aler ‘go’ + infinitiv
Total: 25 different verbs (17 attested in the 12th century; 18 in the 15
th century). 1118 clauses (214 from the 12th century; 904 from the 15th
century)
Unergative verbs
abstiner ‘abstain’; apartenir ‘be related to’; apenser ‘think’; apercevoir
‘perceive’; cesser ‘cease,’ ‘end’; chevauchier ‘ride’; corrocier ‘become
angry’; courir ‘run’; deporter ‘entertain,’ ‘behave,’ ‘be patient’; de-
sesperer ‘despair’; desmouvoir ‘refrain from,’ ‘change’; disner ‘break
one’s fast’; doter ‘doubt,’ ‘fear’; entremetre ‘intervene’; escrier ‘cry
out’; esfreer ‘be alarmed’; esmovoir ‘move’; estordre ‘escape’; estre
‘be’; esveiller ‘awaken’; gabber ‘joke’; merveillier ‘marvel’; movoir
‘move’; pasmer ‘faint’; pener ‘suffer’; plaindre ‘complain’; recorder
‘remember’; repentir ‘repent’; resveillier ‘wake up again’; rire ‘laugh’;
saouler ‘become satisfied’; se trover ‘find oneself’; soucier ‘be wor-
ried’; taire ‘be silent’; tarder ‘delay’; vanter ‘boast’.
Total: 35 different verbs (12 attested in the 12th century; 29 in the 15
th century). 125 clauses (35 from the 12th century; 90 from the 15th
century)

Figure 7: Attested intransitive verbs

It is characteristic that a large number of unaccusative verbs appear with

a reflexive in Old French. The use of a reflexive seems to be paralleled

by the use of sibi ‘himself’ with unaccusative verbs in Late Latin (Cen-
namo, 1999). The most dominant unaccusative verb in the 12th century is
se tourner ‘return’ (17-a), which is later replaced by the verb se retourner.
The second most common unaccusative verb is aller ‘to go’ (17-b), which
Reflexive verbs and the restructuring of clitic clusters 22

is by far the most frequent in the 15th century. In the 15th century we also
find s’en aller used as a semi-auxiliary, in front of both gerundives (17-c)
and infinitives (17-d).

(17) Unaccusative verbs

a. Tut sul s’ en est Eufemien turnét

all alone REFL from-here is Euphemianus returned
‘Euphemianus returned home all by himself.’

(12th c., Alexis, 344)

b. Escrit la cartra tute de sei medisme, cum s’

wrote the charter all of him self how REFL
en alat e cum il s’en revint.
from-here went and how he REFL from-here came back
‘All by himself he wrote down the story about how he went
away and how he came back.’

(12th c., Alexis, 285)

c. Jouhanne s’ en va riant a sa damme.

Joanne REFL from-here goes laughing to her lady
‘Joanne is laughing as she approaches milady.’

(15th c., QJM, p 44)

d. et puis s’ en vont disner.

and then REFL from-here go dine
‘and then they go to dine.’ (15th c., Saintre, p 105)

There are about as many different unergative as unaccusative verbs with

the sequence s’en, but they are less frequent (18).

(18) Unergative verbs

a. Cent mille Francs s’ en pasment cuntre

hundred thousand French REFL from-here faint against
Reflexive verbs and the restructuring of clitic clusters 23

tere.
earth
‘A hundred thousand French fainted and fell to the ground.’

(12th c., Roland, 2932)

b. Chascun s’ en rit, disant: Dieu quelle joye!

everyone REFL from-here laughed saying God what joy
‘Everyone laughed saying: God, what a joy!"

(15th c., CDO, p.100, v.45)

The use of the reflexive with unaccusative verbs is significantly more

widespread than with unergative verbs. Even though there is no difference

in the number of verbs permitting a reflexive, the total number of clauses
with this structure is almost ten times higher with unaccusative verbs (1114)
than with unergative ones (120).

5.1.2 Transitive verbs

Distinguishing between the three types of verbs runs into methodological

problems when looking at the transitive verbs. The question is what hap-

pens with a transitive verb when it appears with a reflexive pronoun. Going
into a thorough debate on the analysis of reflexive pronouns is too extensive
for the present paper. Let me just briefly state that there is a fundamental
difference between intrinsic reflexives appearing on unaccusative verbs and

those appearing with transitive verbs. A long standing tradition within lin-
guistics has made the claim that a reflexive that appears with a transitive verb
absorbs a thematic role (see among others Reinhart and Reuland (1993) and
references in Labelle (2008)). A consequence of this line of thought would

be that a transitive verb becomes unergative as the the object role is absorbed
Reflexive verbs and the restructuring of clitic clusters 24

by the reflexive. A transitive verb such as doloser means ‘to mourn, lament,
bewail.’ As a reflexive verb it is intransitive with the meaning ‘to suffer, to
be in pain’ (Hindley et al., 2000). In Old French, the reflexive occurs in the

same position as the corresponding argumental clitic. This is clear when

one considers examples such as (4), repeated here as (19).

(19) onques ne le se pensa

never NEG it.ACC himself.DAT thought
‘He never thought about it.’ (13th c., MortArtu, II, 13)

I take this word order as an indication that the reflexive is in fact an argument
in the case of transitive verbs. To complicate matters, there are numerous
verbs which may have a transitive and an intransitive reading at the same
time, and it is not always immediately easy to tell what reading the reflexive
is derived from. These verbs are also listed under transitive in figure 9.

Figure 8 gives an indication as to the internal composition of the transitive

group. Comparing the data from the 12th century to those from the 15th,
we see that there is a substantial increase both in the number of clauses
attested with transitive verbs and in the number of verbs which appear in

such constructions. I build these tables on the listings in Hindley et al.

(2000).
Reflexive verbs and the restructuring of clitic clusters 25

Transitive verbs
acointier ‘make the acquaintance of’; adober ‘equip (for battle),’ ‘dub’;
afichier ‘lean against,’ ‘fasten to something’; afoler ‘become mad,’
‘die,’ ‘make mad’; aider ‘help’; aquiter ‘pay (debt),’ ‘justify,’ ‘aquit’;
armer ‘arm,’ ‘equip’; assembler ‘come together’; asseurer ‘have faith
in,’ ‘assure’; atendre ‘wait,’ ‘expect’; baignier ‘bath’; chargier ‘bur-
den,’ ‘weigh down’; chastiier ‘improve,’ ‘warn’; chaufer ‘heat’; clamer
‘call’; combatre ‘fight’; complaindre ‘complain’; conduire ‘guide’;
conseillier ‘counsel’; considerer ‘reflect’; constituir ‘appoint’; con-
tenter ‘satisfy’; couchier ‘lie down’; cuidier ‘believe’; defendre ‘de-
fend’; delivrer ‘free’; deschargier ‘unload’; desconforter ‘destroy,’
‘grieve’; descovrir ‘reveal’; desdeigner ‘disdain’; desdire ‘contradict’;
desjoindre ‘separate’; despechier ‘free’; dessaisir ‘disposess’; destolir
‘go away,’ ‘remove’; detenir ‘keep,’ ‘restrain’; deviser ‘talk,’ ‘sepa-
rate,’ ‘divide’; dire ‘say’; doloir ‘be in pain,’ ‘regret’; doloser ‘suf-
fer,’ ‘mourn’; doner ‘give’; donner garde ‘pay attention to’; don-
ner merveille ‘surprise’; ensivre ‘follow’; envoleper ‘envelope’; es-
bair ‘trouble’; esbatre ‘beat,’ ‘amuse’; eschiver ‘refrain from’; es-
cuser ‘apologise’; esforcier ‘seize’; esjoir ‘rejoice’; eslever ‘raise’;
esmerveillier ‘marvel’; estimer ‘estimate’; estrangier ‘remove’; faire
‘do’; falir ‘be lacking’; falloir ‘to need’; feindre ‘hesitate,’ ‘shape’;
garder ‘keep,’ ‘beware’; garir ‘cure’; joir (re-) ‘enjoy’; lever ‘raise’;
loer ‘praise’; malcontenter ‘not satisfy’; mesler ‘mingle’; metre ‘place’;
moquer ‘mock’; mostrer ‘show’; obligier ‘oblige’; parer ‘embellish’;
parjurer ‘break (one’s word)’; passer ‘manage,’ ‘pass over’; penser
‘think’; percoivre ‘perceive’; perdre ‘lose’; picquier ‘hack’; porchacier
‘seek,’ ‘provide’; porter ‘carry,’ ‘behave,’ ‘be pregnant’; prendre garde
‘watch out for’; purgier ‘purge’; raporter ‘bring back’; ravoir ‘have
again’; reconforter ‘comfort’; redouter ‘fear’; relever ‘raise up’; rendre
‘give back,’ ‘surrender’; retraire ‘withdraw’; revoir ‘see again’; saisir
‘seize’; sauver ‘save’; savoir ‘know’; se tirer ‘leave’; sentir ‘feel’;
sofrir ‘endure’; sordre ‘rise (up)’; sormonter ‘surpass’; tencier ‘quar-
rel’; tenir ‘keep’; vengier ‘avenge’
Verbal periphrases
faire ‘make’ + infinitive (causative construction).
Total: 100 different verbs (18 attested in the 12th century; 87 in the 15
th century). 262 clauses (19 from the 12th century; 243 from the 15th
century)

Figure 8: Attested transitive verbs

Reflexive verbs and the restructuring of clitic clusters 26

From the 12th to the 15th century there is a large expansion in the num-
ber of transitive verbs appearing with a reflexive. In some cases the reflexive
imposes a passive reading, as in (20-a). In other cases, the reflexive just in-

dicates co-reference with the subject (20-b).

(20) a. Plus de cent milie s’en adubent ensemble

more than hundred thousand S’EN dubbed together
‘More than a hundred thousand [knights] dubbed themselves
together.’

(12th c., Roland, 3000)

b. c’ est folie de s’ en donner trop grant courroux

it is madness of S’ EN give too big rage
‘It is madness to get too infuriated.’

(15th c., Mélusine, p 120)

To sum up: The results show that there is a strong increase in the use of
reflexive verbs from the 12th to the 15th century. The ratio of s’en compared
to the total of words in each century reveals an increase from 0.0423 % to

0.1492 %, which turns gives p-value < 0.05. The increase is thus statistically
significant.

12th century 15th century

type no. ratio no. ratio
unaccusative 214 0.799 900 0.729
unergative 33 0.123 87 0.07
transitive 21 0.078 248 0.20
total 258 1235
no. words 610,224 0.000423 827,692 0.001492
p-value < 0.05

Figure 9: The sequence s’en

Reflexive verbs and the restructuring of clitic clusters 27

Whereas unaccusative verbs account for 79.9 % of the examples in the

12th century, their amount shrinks to 72.9 % in the 15th. Transitive verbs, on
the other hand, only account for 7.8 % of the examples in the 12th century.

In the 15th century, as much as 20 % of the examples are with transitive

verbs. This is a statistically significant change (p-value < 0.05). This in-
crease is compatible with Brusewitz’ hypothesis, which states that there is
an increase in the use of reflexive verbs.
The use of reflexive pronouns is spreading to new verbs, but the increase

is stronger with transitive verbs than in any other group. Even though we
do not find an elevated number of s’en constructions with a verb such as
dire ‘say,’ we may still claim that the rise of the reflexive forms may have
induced a change in the clitic clusters. The word order subject–reflexive

became so frequent, also with intransitives, that this order was conceived of
as natural. Since there is no difference between the reflexive pronoun and
the object pronoun of the first and second persons, this induced a change in
the internal order of the clitic clusters. This is what I will return to in the

next section.

6 Changing the clitic clusters

We have seen that there was a substantial increase in the occurrences of the

reflexive pronoun se with transitive verbs, something which I interpret as

an expansion of the domain of the reflexive. From being first and foremost
associated with unaccusative verbs, the reflexive becomes more and more
common also with transitive verbs. The idea that I will explore in this sec-

tion is that this is the trigger for the internal change in the clitic clusters.
As the language users used the reflexive more often with transitive verbs,
Reflexive verbs and the restructuring of clitic clusters 28

and increasingly with a pronominal subject, the sequence pronominal sub-

ject + reflexive was conceived as the natural order. The consequence of this
change is that person, not case (or syntactic function) became the major

principle of clitic ordering in French. Even though the result relates to syn-
tax, there was a PF-related trigger that created an ambiguous input, which
in turn led to a reanalysis.
The exact position of clitics in clusters is not unproblematic. Kayne
(1994) points out that it is impossible to adjoin more than one clitic per

head as this would mean clitics would c-command each other. He suggests
two different solutions:

“(...) sequences of clitics must not be analyzed as successive adjunctions

to the same head but instead should be analyzed as involving either adjunc-
tions to distinct functional heads or as adjunction of one clitic to another, or
some combination thereof.” (Kayne, 1994: 21).

In other words: Clitics either occupy different phrases in some sort

of clitic field, as has been suggested by different scholars (see Sportiche

(1996); Kayne (2000); Poletto and Pollock (2004a,b); Manzini and Savoia
(2004)) – or they are incorporated into each other. Pescarini (2012) uses
this insight as he proposes two different kinds of clitic clusters. (21-a) is the

split cluster, where the clitics belong to different but adjacent heads. (21-b)
represents the incorporated clitics, which Pescarini refers to as true clusters.

(21) a. Split cluster: [clitic-ACC [clitic-DAT]]

b. True cluster: [clitic-DAT clitic-ACC [ clitic-DAT]]

The idea exemplified in (22-b) strongly resembles the one presented by

Reflexive verbs and the restructuring of clitic clusters 29

Emonds (1975) in a very different framework, but where the main idea was
that reflexives as well as 1st and 2nd person clitics were base-generated in
the same position as 3rd person dative clitics, in other words, in the position

following the accusative. In the transformational process, all clitics base-

generated in second position, with the noteworthy exception of 3rd person
clitics, are moved such that they appear in the first position in the cluster.
In addition to explaining the order in which clitics occur, Emonds also very
neatly explains the occurrence of the PPC in French.

In order to describe a change from a theoretical point of view, it is nec-

essary to know the starting point and the ending point of a process. We have
already seen (section 2) that clitics are structured in different ways in Old
and Modern French. In Old French, the clitic order is ACC – DAT. This

way clitics appear in the same order as full DP arguments. What kind of
syntactic structures do these word orders reflect? I will first look at Modern
French in section 6.1 before turning to Old French in section 6.2.

6.1 Clitics in Modern French

Person is the major determining factor when it comes to clitic ordering in

Modern French.7 Clitics of the 1st and 2nd persons always precede 3rd

person clitics, regardless of their syntactic role. In addition, the reflexive

pronoun also occurs in first position in a clitic cluster. Laenzlinger (1993)
suggests that is it necessary to take case-marking into account when explain-
ing the order in which French clitics occur. He distinguishes between case-
marked (Cl+K ) and non case-marked clitics (Cl−K ), exemplified in (22).
7
Of course, case comes into play when both clitics are in the 3rd person, but this is a
secondary principle. Note that 3rd person reflexive se also comes in first position in the
clusters.
Reflexive verbs and the restructuring of clitic clusters 30

(22) a. Cl+K : le, la, les, lui, leur

b. Cl−K : me, te, nous, vous, se

Clitics lacking case-marking do not distinguish between accusative and da-

tive, whereas clitics with case-marking do. Laenzlinger observes that +K
and -K clitics behave in different ways. A +K clitic needs to check its case,

which it does by morphological selection. This method of incorporation im-

plies that the clitic head and the verbal head are in a configuration of mutual
c-command; in other words: they are sisters. Because the verbal head and
the +K clitic must be in a configuration of mutual c-command, they are in
complementary distribution.

A -K clitic, on the other hand, has no case to check, and does not take
part in such a selectional process. Instead, a -K clitic is free to incorporate
into any host that carries a φ-feature.
A consequence of these observations is that a +K and a -K clitic may be

combined, but that two +K clitics may not. A clause such as me le ‘me him’
in (23) exhibit the combination of a Cl−K and a Cl+K , which is fine. The
o
two clitics are incorporated into the same head, T . This way, the clause in
(23) has the structure in (24).

(23) Elle me le donne.

she me.DAT it.ACC gives
‘She has given it to me.’

(24) [ TP Elle [ To me [ To le [ To donne]]]].

The reason why it is possible to have clusters combining two Cl+K clitics
(e.g. le lui ‘him him’) is that they are not cliticised to the same node. In (25)
Reflexive verbs and the restructuring of clitic clusters 31

o
the clitic lui ‘of-him’ is cliticised to the Aux , whereas le ‘him’ is cliticised
o
to T .

(25) Jean le lui a presenté.

John it.ACC him.DAT has presented
‘John has presented it to him.’ (from Laenzlinger (1993))

(26) [ TP Jean [ To le [[ To [ AgrOo [ Auxo lui [ Auxo a ]]]]]]] presenté.

Ever since Chomsky (1995) there has been no place for agreement projec-
tions in syntax, so the analysis in (26) is rendered somewhat obsolete. I will
however retain Laenzlinger’s main idea, namely that there are two differ-

ent kinds of clitic clusters: One where clitics are incorporated into the same
head, and one where they cliticise to different heads. Using Pescarini’s anal-
ysis in (21) I will suggest that clusters involving the 1st and 2nd person are
true clusters, whereas clusters involving the 3rd person are split clusters
(27).

(27) a. Split cluster: [le.ACC [lui.DAT]]

b. True cluster: [me.DAT le.ACC [ me.DAT]]

In the cases where the reflexive se is used, it is reasonable to assume that

se has been moved in front of the accusative, just as in (27).

(28) Il se le dis.
He REFL.DAT it.ACC say
‘He tells himself.’

(29) [se.DAT le.ACC [ se.DAT]]

Reflexive verbs and the restructuring of clitic clusters 32

As such, I analyse all clusters involving 1st and 2nd person clitics as
well as 3rd person reflexive clitics as true clusters, derived by the same
mechanism.

6.2 Clitics in Old French

What was the status of Old French clitic clusters? I will propose that they
were all split clusters, as Modern French le lui-clusters are still. There are

two reasons for assuming this. First and foremost because the le lui-cluster
is archaic in the sense that it has not been altered. If these clusters are
split in Modern French, they must be so in Old French as well. Pescarini
(2012) makes the same assumption when he analyses Old Italian clitics,

which have undergone a a change similar to the one that affected French
clitics.8 In Italian the shift also has phonological consequences in that the
leftmost vowel of the cluster changes from i to e (30).

(30) [la [mi]] → [ me la [ me ]]

The vowel does not change as the clitic order is reversed in French. How-
ever, in section 3 we saw that several scholars have suggested that there
was some sort of phonological change taking place in Old French simul-
taneously as the clitic cluster order started changing. Meyer-Lübke (1899)
suggested it was a rhytmical change, whereas Zink (1997) discusses what

he calls a change in accent. This change implies that the apparent enclitic
forms such as jol (jo + le, ‘I + it’) disappear (31).

(31) jol dis → je le dis

8
The major difference is that also 3rd person clusters are DAT–ACC in Modern Italian.
Reflexive verbs and the restructuring of clitic clusters 33

It is however crucial to note that in an example such as the one in (31),

the object must still be followed by the finite verb. This is the same ob-
servation as the one made for Modern Romanian by Miller and Monachesi
(2003: 100). In this language a clitic may be enclitic to the preceding el-

ement, also when the preceding element is a full DP. Interestingly, Miller
& Monachesi suggest that the state of affairs in Romanian is reminiscent of
the the cliticisation patterns in Old Romance.
Instead of viewing this as solely the change of a metric pattern (Meyer-

Lübke, 1899; Zink, 1997), it is also a change that may be directly linked to
a change in syntax, namely the change from a split to a true cluster.
To sum up: Clitics in Old French form split clusters, whereas clitics in
Modern French form either true clusters or split ones. The me le type is the
instantiation of a true cluster, whereas le me-clusters are split ones. We have

also seen that clusters involving the reflexive clitic se have the reflexive in
second position in the oldest texts when it functions as a dative (see example
(4) above). This is very interesting, as it shows that syntactic function is the
major principle of clitic placement. Other possible criteria, such as person

or reflexivehood, do not come into play.

6.2.1 The clitic field

In the following I will assume that clitics are heads, and that each clitic head
is associated with the head of a particular phrase. This way clitics appear
in what may be described as a clitic field, perhaps in the way described
by Sportiche (1996). I will further assume that clitics that are arguments

will occupy designated positions in this hierarchy, and that these positions
are defined by their syntactic status. This way there is a slot for accusative
Reflexive verbs and the restructuring of clitic clusters 34

clitics and one for dative clitics. I take this order to be universal. The change
from a system that is based on case (Old French) into one that is based on
person (Modern French) must then affect the operations that take place after

the clitics have been merged in their positions.9

Furthermore, I will argue that there is a designated slot for inherent re-
flexives, e.g. clitics that are reflexive and that do not correspond to any
argument. I take this position to be situated above the slots for accusative
and dative clitics.

A schematic representation of the three clitic positions in Old French is

presented in figure 10. Prior to any restructuring reflexive clitics occur in
Cl-1, accusative clitics in Cl-2 and dative clitics in Cl-3.

9
Also note that in the cases of true imperatives, when the clitics are enclitic to the finite
verb, they always occur in the order ACC–DAT.
Reflexive verbs and the restructuring of clitic clusters 35

Cl-1P

Cl-1’

Cl-1o Cl-2P

inherent Cl-2’
reflexive
clitic Cl-2o Cl-3P

direct Cl-3’
object
clitic Cl-3o

indirect
object
clitic

Figure 10: The clitic field in Old French (prior to change)

If what I have proposed is on the right track, namely that the increased
use of reflexives led to a reanalysis, we can account for this syntactically.
The input dictates that reflexives should occur in first position, which is

the only possible position for inherent clitics. As the language does not
overtly distinguish between the different functions of the clitics in the 1st
and 2nd persons, the analogy then imposes that all syncretic forms occur in
first position of the cluster. If, however, the syncretic clitic is a dative, it is
o
merged under Cl-3 and would occur in second position of the cluster.10 In
order to occur in the first position in the cluster, it has to move. This way
clitics of the 1st and 2nd persons aare moved to the first position in the linear
10 o
These clusters only contain two elements, so even though it is merged under Cl-3 , it
occurs in second position in the linear order.
Reflexive verbs and the restructuring of clitic clusters 36

order. Also dative se has to move as it is a reflexive form and should come
first in a cluster, regardless of its syntactic function.
The only way an indirect object may occur in front of a direct object

clitic is by moving to the same head and adjoin to the higher head. This is
indeed what happens, and we get a true cluster (figure 11).

Cl-2P

Cl-2’

Cl-2o Cl-3P

me le Cl-3’

Cl-3o

me

Figure 11: Derivation of a true cluster

This analysis implies that there is a major difference between Old and
Modern French when it comes to cluster types. While Old French only

displays split clusters, Modern French has both, as exemplified in (32). Note
that as reflexives are merged in a higher position, clusters in which they
occur must be split clusters.

(32) Modern French

a. Split clusters: Clusters involving reflexives, clusters involv-

ing 3rd person clitics.

b. True clusters: Clusters involving 1st and 2nd person argu-
mental clitics.
Reflexive verbs and the restructuring of clitic clusters 37

Thus, from a derivational point of view, what distinguishes Old French

from Modern French is that in Modern French clitics may move after they
have been merged. In Old French clitics remain in the position in which

they are merged. The trigger for the movement that started taking place in
the 15th century was the increased input of reflexive forms.

7 Conclusion

I have shown that the change in clitic clusters in French took place rather
quickly, in the 15th and 16th centuries. Clusters including 1st and 2nd per-
son clitics changed from the ACC–DAT order to the DAT–ACC order of
Modern French. In the case of 3rd person clitics, no change took place.

I have also shown that from the 11th–12th centuries to the 15th the use
of reflexive forms expanded, and that there was an increasing number of
transitive verbs which appeared with the reflexive pronoun se + the adverbial
pronoun en. At the same time, the relative proportion of clauses appearing

with the reflexive pronoun se increased. I have argued that it is the rise
of the reflexive pronoun which triggers the change in the clitic clusters in
French. As the frequency of the reflexive construction increased, there was
a reanalysis which led to a requirement that all reflexives should occur in
the first position of a cluster. As the clitics of the 1st and 2nd persons are

syncretic, the ambiguity of the input led to place these in the first position,
regardless of their syntactic function.
From a formal point of view, I have suggested that clitics in Old French
were split clusters, whereas clusters in Modern French involving 1st and

2nd person argumental clusters are true clusters. The change in the linear
organisation of the clitic clusters is the reflection of the transfer from a split
Reflexive verbs and the restructuring of clitic clusters 38

to a true cluster.

Appendix: Cited texts

Examples found in the online data bases

CLM = Corpus de la littérature médiévale: en langue d’oil des origines à

la fin du XVe siècle: prose narrative – poésie – théâtre (Lalou et al., 2001).
BFM = la Base de Français Médiéval.

Adam Anonymous: Le Jeu d’Adam (12 th c.), CLM

Alexis Anonymous: Vie de Saint Alexis (late 11th c., early 12th c.), BFM
Aliscans Anonymous: Aliscans (12 th c.), CLM
Antioche Anonymous: Chanson d’Antioche (12 th c.), CLM
Aspremont Anonymous: Chanson d’Aspremont (12 th c.), CLM
Béroul Béroul: Roman de Tristan (12 th c.), CLM
CDO Charles d’Orléans: Songe en complainte (15 th c.), BFM
Charrete Chrétien de Troyes: Chevalier de la charrete (12 th c.), CLM
Charroi Anonymous: Charroi de Nîmes (12 th c.), CLM
Coinci Gautier de Coinci: Miracles de Nostre Dame (13th c.), BFM
Couci Chastelain de Couci: Chansons attribuées au Chastelain de Couci
(12 th c.), CLM
Eulalie Anonymous: La Cantilène d’Eulalie (late 9th c.), CLM
Jouvencel Jean de Bueil: Le Jouvencel (15 th c.), BFM
Mélusine Jean d’Arras: Mélusine (15 th c.), BFM
Mont-Sebond Michel de Montaigne: Essais, t. 1, book 2, chapter 11:
‘Apoogie de Raimond Sebond,’ 1592, Frantext
Reflexive verbs and the restructuring of clitic clusters 39

Mont-vanité Michel de Montaigne: Essais, 2. 2, book 3, chapter 9: ‘De la

vanité,’ 1592, Frantext

Portigal Anonymous: Miracle de la femme et du roy de Portigal (before

1339), CLM

Raoul Anonymous: Raoul de Cambrai (12 th c.), CLM

Renart-I Anonymous: Le Roman de Renart (1175–1250), CLM

QJM Anonymous: La Chanson de Roland (written ca 1080, mauscript

from the 12 th c.), BFM

Saintré Antoine de la Sale: Jean de Saintré (15 th c.), BFM

Thèbes Anonymous: Le Roman de Thèbes (12 th c.), CLM

Vieil-T Anonymous: Mistére du Vieil Testament (15 th c.), CLM

Yvain Chrétien de Troyes: Yvain (12 th c.), CLM

QJM Anonymous: Quinze joies de mariage (15 th c.), BFM

Examples retrieved manually

MortArtu Anonymous: La Mort du roi Arthur (ca 1230), éd David Hult,

Paris 2009.

References

Janice M. Aski. Variable double object clitic order in medieval florentine.

In Fabian Alfie and Andrea Dini, editors, Accessus ad Auctores, Studies
in Honor of Christopher Kleinhenz. Medieval and Renaissance Texts and
Studies, Arizona, in press.
Cedric Boeckx. Quirky Agreement. Studia Linguistica, 54(3):354–380,
2000.
Eulàlia Bonet. Morphology after Syntax: Pronominal Clitics in Romance.
PhD thesis, MIT, Cambridge, Mass., 1991.
Reflexive verbs and the restructuring of clitic clusters 40

Eulàlia Bonet. The Person-Case Constraint: A Morphological Approach.

MIT Working papers in linguistics, 22:33–52, 1994.

Victor Brusewitz. Étude historique sur la syntaxe des pronoms personnels

dans la langue des Félibres. PhD thesis, Uppsala University, Stockholm,
1905.
Luigi Burzio. The anaphoric and pronominal system. Johns
Hopkins University, http://cogsci.jhu.edu/people/files/_pubs-
Burzio/ThePronominalSystem.pdf, accessed January 2013 2008.
Michela Cennamo. Late Latin Pleonastic Reflexives and the Unaccusative
Hypothesis. Transactions of the Philological Society, 97(1):103–150,
1999.
Noam Chomsky. The Minimalist Program. Current Studies in Linguistics
Series; 28. MIT Press, Cambridge, Mass., 1995.
Jennifer Culbertson. The status of old french clitics in the 12th century.
In Pascual José Masullo, Erin O’Rourke, and Chia-Hui Huang, editors,
Romance Linguistics 2007 Selected papers from the 37th Linguistic Sym-
posium on Romance Languages (LSRL), Pittsburgh, 15–18 March 2007,
number 304 in Current Issues in Linguistic Theory. John Benjamins, Am-
sterdam, 2009.
Ans de Kok. La place du pronom personnel régime conjoint en français;
une étude diachronique. Rodopi, Amsterdam, 1985.
Joseph Emonds. A transformational analysis of French clitics without pos-
itive output constraints. Linguistic Analysis, 1(1):3–24, 1975.
Hans Petter Helland. Le participe passé et l’hypothèse inaccusative. In Lan-
gage et référence, pages 255–264. Acta Universitatis Upsaliensis, Upp-
sala, 2001.
Alan Hindley, Frederick W. Langley, and Brian J. Levy. Old French–
[English Dictionary. Cambridge University Press, 2000.
Paul Hirschbühler and Marie Labelle. Evolving Tobler-Mussafia Effects in
the Placement of French Clitics. In Steven N. Dworkin and Dieter Wan-
ner, editors, New Approaches to Old Problems. Issues in Romance His-
torical Linguistics, number 210 in Current Issues in Linguistic Theory,
pages 165–182. John Benjamins, Amsterdam/Philadelphia, 2000.
Richard S. Kayne. The Antisymmetry of Syntax. Linguistic Inquiry Mono-
graphs; 25. MIT Press, Cambridge, Mass., 1994.
Richard S. Kayne. Parameters and universals, chapter Null Subjects and
Clitic Climbing. Oxford studies in comparative syntax. Oxford University
Press, Oxford; New York, 2000.
Marie Labelle. The French reflexive and reciprocal se. Natural Language
and Linguistic Theory, 26(4):833–876, 2008.
Reflexive verbs and the restructuring of clitic clusters 41

Marie Labelle and Paul Hirschbühler. Changes in clausal organization and

the position of clitics in Old French. In Montse Batllori, Maria-Lluisa
Hernanz, Carme Picallo, and Francesc Roca, editors, Grammaticalization
and Parametric Change, pages 60–71. Oxford University Press, Oxford,
2005.
Christopher Laenzlinger. A syntactic view of Romance pronominal se-
quences. Probus, pages 241–270, 1993.
E. Lalou, Dominique Boutet, and Elisabeth Gaucher. Corpus de la littéra-
ture médiévale: en langue d’oil des origines à la fin du XVe siècle: prose
narrative – poésie – théâtre. Champion electronique, Paris, 2001.
Géraldine Legendre and Antonella Sorace. Auxiliaires et intransitivité en
français et dans les langues romanes. In Danièle Godard, editor, Les
langues romanes. Problèmes de la phrase simple. CNRS éditions, Paris,
2003.
Beth Levin and Malka Rappaport Hovav. Unaccusativity: At the Syntax-
Lexical Semantics Interface. Number 26 in Linguistic Inquiry Mono-
graphs. MIT University Press, Cambridge, MA, 1995.
M. Rita Manzini. From Romance clitics to case: Split accusativity and the
Person-Case Constraint. In Irene Franco, Sara Lusini, and Saab Andrés,
editors, Romance Languages and Linguistic Theory 2010: Selected pa-
pers from ‘Going Romance’ Leiden 2010, pages 1–20. John Benjamins,
Amsterdam, 2012.
M. Rita Manzini and Leonardo M. Savoia. Clitics: Cooccurrence and Mu-
tual Exclusion Patterns. In Luigi Rizzi, editor, The Structure of CP and
IP: Oxford Studies in Comparative Syntax. Oxford University Press, Ox-
ford, 2004.
Wilhelm Meyer-Lübke. Grammatik der Romanischen Sprachen; Romanis-
che Syntax, volume 3. Fues Verlag, Leipzig, 1899.
Philip Miller and Paola Monachesi. Les pronoms clitiques dans les langues
romanes. In Danièle Godard, editor, Les langues romanes. Problèmes de
la phrase simple. CNRS éditions, Paris, 2003.
Jabier Ormazabal and Juan Romero. The Object Agreement Constraint.
Natural Language & Linguistic Theory, 25:315–347, 2007.
David M. Perlmutter. Impersonal passives and the unaccusative hypothesis.
Proceedings of the Annual Meeting of the Berkeley Linguistics Society, 4:
157–190, 1978.
Diego Pescarini. Prosodic restructuring and morphological opacity. the evo-
lution of Italo-Romance clitic clusters. lingBuzz/001477, 2012.
Cecilia Poletto and Jean-Yves Pollock. On the Left Periphery of Some Ro-
mance Wh-Questions. In Luigi Rizzi, editor, The Structure of CP and IP:
Oxford Studies in Comparative Syntax. Oxford University Press, Oxford,
2004a.
Reflexive verbs and the restructuring of clitic clusters 42

Cecilia Poletto and Jean-Yves Pollock. On Wh-Clitics and WH-doubling in

French and some North Eastern Italian Dialects. Probus, 16(2):241–272,
2004b.
Tanya Reinhart and Eric Reuland. Reflexivity. Linguistic Inquiry, 24:657–
720, 1993.
Donimique Sportiche. Clitic Constructions. In Johan Rooryck and Lau-
rie Ann Zaring, editors, Phrase structure and the lexicon, Studies in nat-
ural language and linguistic theory; vol. 33. Kluwer, Dordrecht, 1996.
Knud Togeby. Precis historique de grammaire francaise. Akademisk For-
lag, Copenhague, 1979.
Dieter Wanner. The Evolution of Romance Clitic Order. In R. Joe Camp-
bell, Mark G. Goldin, and Mary Clayton Wang, editors, Linguistic studies
in Romance languages : proceedings, pages 158–177. Georgetown Uni-
versity Press, Washington, 1974.
Gaston Zink. Morphosyntaxe du pronom personnel (non réfléchi) en moyen
français : (XIVe-XVe siècles). Librairie Droz, Genève, 1997.