Académique Documents
Professionnel Documents
Culture Documents
coordinate construction
Moreno Mitrovic
1 introduction
This paper investigates the word order change in Indo-European (IE)
coordinate construction. Across the entire IE family, two morphosyntactic patterns of coordination are found as Agbayani and Golston (2010)
have investigated most recently. In one type of coordinate construction,
the coordinator occupies the enclitic (peninitial, or second) position with
respect to the internal (second) coordinand (1a), while in another type,
the coordinator is initially placed between any two, or more, coordinands
(1b), as the the minimal representative pair from Homeric Greek shows
in (1). Diachronically, the change from the two competing structures
with peninitial and initial positions to the initial type is uniform across
the IE board.
(1)
a. .
eia te pteroenta
aspidas eukuklous lais
shields round
pelt
and feathered
The round shields and fluttering targets.
(Homer, Iliad, book M: l. 426)
b.
o polemoio
kes emi ka anti
there go and meet battle
Go thither, and confront the war.
(Homer, Iliad, book M: l. 368)
Moreno Mitrovi
c( )
Jesus College
Cambridge cb5 8bl
mm821@cam.ac.uk http://mitrovic.co
Mitrovi
c
mel
for a binary-branching model of coordinate syntax go back to Blu
(1914) with subsequent substantiation from Bloomfield (1933), Bach (1964),
Chomsky (1965), Dik (1968), Dougherty (1969), Gazdar et al. (1985), Goodall
(1987) and Muadz (1991), and many others in the last two decades. Following Kayne (1994), we will assume that coordinators are heads, merging an internal argument (coordinand) as its complement, and adjoining an external argument (coordinand) in its specifier, as per (2).
(2)
&P
.
.
XP
.
coordinand
.
1
.
&.
YP
.
.
.coordinator
coordinand
.
2
Assuming a binary branching structure for coordination, which corresponds to the representation in (2), den Dikken (2006) argues that exponents such as and and or do not in fact occupy the coordinator-head position as indicated in (2) but are rather phrasal subsets of the coordinator
projection, with their origins in the internal coordinand. The actual coordinator head, independent of conjunction and/or disjunction which
0
&P
.
.
XP
.
J.
coordinand
.
1
.
andP
.
.
0
coordinator
.
and
.
(silent)
.
.
.and
YP
.
coordinand
.
2
The core motivation for den Dikkens postulation of the silent presence
0
Mitrovi
c
(6)
(7)
either. . . .
XP
.
JP.
.
0
(. . .) either
. ...
J.
.
.
YP
.
or ... .
Employing (in his words, the abtsract head) J , den Dikkens account
covers and explains not only the either...or coordinate constructions but
also the whether...or and both...and, which are unified under the structural
umbrella of JP structure. den Dikken (2006: 58) takes the head introduc0
abstract J instead.
An updated and enriched structure of (2) that den Dikken puts forth
and I assume here is therefore the following.
(8)
.
.
JP.
.
P
.
coordinand
.
1
.
.
J.
.
.
.
0
.
.
P
.
coordinand
.
2
his account rests on J not being lexicalised. I take it as a reasonable hypothesis that there may be languages, which overtly realise this junctional component of coordination. In 1.2.2, empirical justification for
(8) is provided. The following section will show that IE syntax of coordination was of the same type.
1.2.2 Lexicalised J: Avar, Hungarian, South Slavonic
There are empirical arguments substantiating the fine-grained (double-headed) structure for coordination (3). Our structure for coordination supposes there are three heads involved (a J and two s). Mutatis
mutandis, the theory predicts that there may be languages that realise all
three (J+/) heads simultaneously.
In this subsection, we consider contemporary languages, which show
evidence for the split coordination structure, i.e. two coordinator positions.
Southeastern Macedoni-
Mitrovi
c
(9) [
Roska] i [
Ivan]
0
0
0
( ) R
J ( ) I
Roska and Ivan.
(10)
[i Roska]
[i Ivan]
0
0
0
R
(J ) I
both Roska and Ivan.
(11)
[
Roska] i [i Ivan]
0
0 0
( ) R
J I
Roska and also Ivan.
heads (two and a J ) without an explicit counterexpectational (butlike) morpheme. SerBo-Croatian, as reported in (13), also allows three
coordinate morphemes per two conjuncts but the J head is adversative,
unlike (12).
(13)
[i Mujo] a
[i Haso]
0
0
0
M
J .but H
Not only Mujo but also Haso.
Beyond Slavonic (and Indo-European), we also find triadic
s) Mari is
Kati is (e
K
J
M
the prediction of subphrasal-status of complement to J . This in fact obtains and the gi-phrasea Pexhibits additive (focal) semantics. The
following shows the strings and (generalised) structures of such Ps in
Avar.
(16) Dida [gyeb gi] lala
I
know this
I [even/also know] this
(17) [Dida gi] gyeb lala
I
know this
[Even I/I too] know this
Aside from the polysyndetic type (18), Avar also allows an English-like
construction with a conjunction marker placed between the two coordi0
keto gi
hve gi
cat (J) dog
cat and dog
This novel data was provided by Ramazanov (p.c.) and Mukhtarova (p.c.).
Mitrovi
c
It is the possibility of co-occurring realisations of the two types of positions that Avar allows which is typologically novel and, for our purposes,
most intriguing. The last type 20 shows a union of phonological realisations in 18 and 19 and the triadic exponency of conjunction. In this
construction type, both heads as well as J are realised simultaneously.
There is currently no alternative syntactic model of coordination, which
could explain the third (20) option of co-occurring realisation of coordination markers without further stipulations. Our fine-grained system
(8), however, can not only handle (20) without any problem, it even predicts its existence. Equipped with these theoretical and empirical observations, we now turn to the core component of this paper and investigate
the syntax of coordination in IE.
2 indo-european
Having motivates fine-grained J- complex for coordinate construction,
both theoretically and empirically, we now address the central concern
of this paper, the IE coordinate construction. The existence of two types
of construction with respect to the pen/initial positioning of the coordinator does not only correlate with
(i) the alternation in linear placement of coordinator but also
(ii) the very morphological structure of the the two types of coordinators heading pen/initial constructions.
In the following two subsections, we take each of the two (i, ii) properties in turn.
2.1 Alternation in linear placement
We start our discussion with a diachronic perspective on IE syntax of
coordination, which shows linear alternation in coordinator placement.
The earliest IE languages show that there existed two syntactic types of
coordinate structures. One in which the coordinator occupies the initial, and another in which the coordinator occupies the peninitial position with respect to the internal coordinand. Klein (1985a, 1985b) has
shown for R
. gvedic, and Agbayani and Golston (2010) for IE more generally, that the alternation between initial and peninitial placements of
the coordinator patterns with the category of the coordinands, whereby
the peninitial (enclitic) coordinators generally cannot coordinate clauses
which the initial coordinators can.
The following pairs of initial (a) and peninitial (b) coordinate configurations from Sanskrit, Greek, and Latin exhibit the alternation in linear
placement of the coordinating particle.
(21) Homeric Greek:
a. .
aspidas eukuklous lais
eia te pteroenta
shields round
pelt
and feathered
The round shields and fluttering targets.
(Hom., Il., M: l.
426)
b.
polemoio
kes emi kai antio
there go and meet battle
Go thither, and confront the war.
ajinvasu
va
am
. v
strength-bestowing
Vayu-Indra-and rush.2.dl rich
Vayu and Indra, rich in spoil, rush (hither).
(R. gveda,
1.002.5 )
b.
rs. i
sy
dvis. a
h.
pa
ta
a uta
save.imp.2.sg this and enmity
Save us from this and enmity.
tem que
b. vam samu
life safety
and
10
Mitrovi
c
The syntactic duality of the double placement of the coordinator extends beyond the three classically representative IE languages above. It
is clear from these pairs of examples that IE had prepositive (a) and a
postpositive (b) series of coordinators. We could distinguish the two types
of configurations by positing that the peninitially placed (enclitic) coordinator induces some form of movement, either syntactically or postsyntactically, but that the difference lies only in the linearisation of the
surface placement of the coordinator. Let us now turn to briefly sketching the empirical facts surrounding this taxonomy of two types of coordinators in IE.
Old Avestan, just like R
. gvedic, distinguishes between initial uta and
enclitic ca:
(24)
Old Avestan:
a. atu ldzam
amQuruh
uta mazd
a
huruma
and wisdom.m.sg.gen increase.m.sg.nom
amoah
esoar
arag
haoma
raose
gara
haoma.m.sg.voc grow.2.subj.mid mountain.sg.m.loc
itiap
paiti
toward
11
b. mVZUY
OiibiEa
Aruha
z m
ahur
a
yu
a
eibiio
you.2.sg.nom them.pl.dat lord.m.sg.voc
ee
OJoa
aogo
AtAd
ACa
d
at
a
a
s. a
Ac
mvrQaCx
x
s. ar m
c
a
power.n.sg.acc and
e
(Yasna Haptangh
aiti, 29.10)
Hittite:
a.
n Mursilin
nu ka
kuennir
nu e
sar ieir
nu
and prt Mursilis.acc they.killed and blood shed.3.pl and
Hantilis
Hantilis
nahsariyatati
feared.3.sg.m
And they killed Mursilis and they shed blood and Hantilis
was afraid.
(2BoTU. 23.1.33-35)
b.
an
su.kur.ra.me
s lu
.mesis.guskin ya humandan
charioteers
grooms.golden and all
Charioteers and all the golden grooms.
(StBoT. 24.ii.60-61)
c.
[ud]u.a.lum
za uru-az parnanzass a
a
kass
this.nom and ptc city.nom house.nom and ram
` -ru
du
become.3sg.imp
and let (both) this city and house become the ram
(KUB
41.8 iv 30.)
Old Church Slavonic also boasts a pair of coordinators: an initial i and
12
Mitrovi
c
vetu
tvori
knezi
su
su
su
counsel.m.sg.acc took.3.sg.aor with princes.pl.inst
svoimi
i
s
Moravl
eny
poss.refl.inst and with the Moravians.pl.inst
carju
ku
posla
Mixailu
sent.3.sg.aor to tsar.sg.dat M.sg.dat
the Prince of Moravia took counsel with his Moravian princes
2
ze gljo
...
Azu
vamu
I
but tell.1.sg.pres you.dat
(Codex Marianus, Mat. 5:28)
Similarly, Old Irish possessed a complex ocus, which occupied the first
position, and a simplex ch (< kwe, cf. IIr. ca, Lat. que, etc.).
(27) Old Irish:
a. bo
Conchubur
ocus maithi
was.3.sg.aor C.m.nom.sg and the nobles.pl.nom
N
i nEmuin
Ulad
Ulstermen.m.pl.gen in Emain Macha
Conchobar and the nobles of the Ulstermen were in Emain
(Compert Con Culainn, 1.1)
Macha.
b. ba ch ri
Temrach
cop and king Tara.gen
And he was king of Tara.
(Laws, 4.179)
Among the old Germanic languages, only Gothic boasts a double set
of coordinators differing in the linear placement: an initial jah and an
13
enclitic uh.
(28) Gothic:
a. ak
ana lukarnastavin j ah liuteiv
ak
ana lukarnastain
jah liutei
neither on candle.dat.sg
and light.ind.3.sg
allaim
vaim
in vamma
garda
allaim
aim
in amma
garda.
all.dat.pl it.dat.pl in that.m.dat.sg house.m.dat.sg
Neither do men light a candle, and put it under a bushel.
(Codex Argenteus, Mat. 5:15)
b. (galaiv
in praitauria
aftra
(galai
in praitauria
aftra
came.pret.3.sg in judgement hall.acc.sh again
peilatus j ah) woida
iesu qav
uh
Peilatus jah) wopida
Iesu qa
uh
P.nom
and called.pret.3.sg J.acc said.pret.3.sg and
imma
imma
him.m.dat.sg
(Then) Pilate entered into the judgment hall again, and called
Jesus, and said unto him.
While Gothic still shows the dual type of coordination (28), there is no
such evidence for other early Germanic languages. The only early Runic
inscription we have is the one in (29), where a medial conjunction andi is
employed.
(29)
Runic Germanic:
a. 1aigil1 andi1 aIlrun1
.andi .alrun.
.aigil
Aigil.pn and Alrun.pn
Aigil and Alrun.
(Looijenga 2003: 253Pforzen I (Bavaria), mid 6th c., silver
belt buckle)
14
Mitrovi
c
w
dharme
ca arthe
ca k
ame
ca
dharma/law.loc and commerce.loc and pleasure.loc and
moks. e
ca bharata r. s. abha yad
iha asti
tad
liberation.loc & Bharata giant which here is.3.sg that
anyatra yad
na iha asti
na tat kvacit
elsewhere which not here is.3.sg not that anywhere
Giant among Bharatas whatever is here on Law, and on commerce, and on pleasure, and on liberation is found elsewhere,
but what is not here is nowhere else. (Mah
abh
arata, 1.56.34)
b.
s ca cetathah
ajinvas
u
va
yav ndra
n
am
. suta
. v
strength-bestowing
Vayu Indra and rush.2.dl rich
Vayu and Indra, rich in spoil, rush (hither).
a
Homeric Greek:
a.
te eonta
te
os
ede
ta
ta
which were (=know.plup) the and exist.part the and
essomena pro
te eonta
exist.fut before and exist.part
That were, and that were to be, and that had been before.
(Homer, Iliad A: 70)
b.
eia te pteroenta
aspidas eukuklous lais
and feathered
pelt
shields round
The round shields and fluttering targets.
(Homer, Iliad M: 426)
(32) Classical Latin:
15
a. iam
tum tendit que fovet que
already then pursue and favour and
Already then, she both pursued it and (also) favoured it. (Vir.,
Aen., 1.18)
b. vam sam
utem que
life safety
and
the life and safety
a.
t
elo
d
so
i
o
staago i
e pace mog
ze s
boite
fear but refl rather which may and soul and body
pogubiti
destroy
16
Mitrovi
c
But rather fear that which is able to destroy both soul and body.
(CM, Mat. 10:28)
b.
eko gol
b
od
ete
ze m
odri
eko zmij
e
i
c
eli
obe
be
but wise as serpents and harmless as doves
Rather be wise as serpents, and harmless as doves.
(CM, Mat. 10:16)
Note that the focal additive meaning related to polysyndeticity has
been retained in some of the contemporary varieties of Slavonic. The
following are parallel examples from Mathew in SerBo-Croatian:
(34)
Synchronic SerBo-Croatian:
a. Bojte se vi
se onoga koji
mo
ze i
du
su i
tijelo
fear refl more that which may and soul and body
pogubiti
destroy
But rather fear that which is able to destroy both soul and body.
(Mat. 10:28)
b. budite dakle
mudri kao zmije
i
bezazleni kao
be
therefore wise as serpents and harmless as
golubovi
doves
Rather be wise as serpents, and harmless as doves.
(Mat.
10:16)
In this subsection weve shown that IE indeed freely allowed reduplication of the coordinator. Having explored the possible semantic side2
In Mitrovi
c (2014: ch.
6), the historical and cross-linguistic dimensions of
non/reduplicative coordination is investigated.
17
While peninitial coordinators are monomorphemic, the initial coordinators are not. Initially placed coordinators are bimorphemic and as
such are decomposable synchronically or diachronically into two coordinators, each underlying a morpheme. Greek kai, for instance, derives
(< h2 (e)u + te); Gothic coordinators jah and jau result from yo + kee and
of four coordinators for PIE. One series is orthotone and another enclitic
3
as shown in Tab. 1 .
orthotone
w
w
k
o / k i
h2ew
y
o
t
o
enclitic
-k e
-h2 u
-yo
-te
18
Mitrovi
c
dependent / composed
w
k e
k e
te
h2 u
Gr. kai
te
OIr. to-ch
Hit. tak-ku
h2 u
Skt. u ca
Lat. at-que
IIr. u-ta
Gr. au-te
Lat. au-t
yo
Goth. ja-h
Goth. j-au
nu
OIr. na-ch
OIr. na-de
independent
nu
[+]
[]
IIr. ca
Lat. que
OIr. ch
Goth. uh
Gaul. cue
Ven. ke
Celtib. ku
Gr. te, de
Alb. dhe
Skt. tu
Sl. ze
Sl. to
Sl. i-no
IIr. u
Gr. au
CLuw. ha
Sl. i
Hit. ya
TochA. yo
Myc. jo
Hit. nu
OIr. no
Sl. n
u
yo
nate morphemes are distributed between J and the head of its comple0
ment, , as per Tab. 2. This idea is summarised in (35) with the three
types of coordinate construction; Classical Latin (at)que is taken as an ex-
The notation [] in Tab. 2 refers to whether a particle is a Wackernagel element, requiring second-position ([+]), or not ([]). The theory and details behind the notations
are addressed below.
19
que
.
que
.
que
.
at.
que
.
skrit u-t
a are spread between and J , also lends itself to a diachronic
analysis of the development of linear placement of coordinators in synchronic IE, which is uniformly head-initial. The analysis put forth here
also makes an empirical prediction for IE. Our having assigned the lower
-headed coordination structure a category status, we predict the independence of P. According to (8), the syntax of coordination is broken
0
down into categories of two kinds. While the higher J is taken to join
coordinate arguments, its substructural P is thus, mutatis mutandis, predicted to be an independent phrasal category. By virtue of being junc0
tional, J establishes a two-place relation between coordinands (a formal default of coordination). P, on the other hand, does not establish a
two-place coordinate relation, which leads us to the possibility there are
mono-argumental and morphosyntactically coordination-like constructions headed by in IE. Given the generalisation on monomorphemic
0
20
Mitrovi
c
in IE mono-argumental constructions headed by monomorphemic particles like Latin que, Sanskrit ca or OCS i. This is in fact what we find in
all IE branches. Independent Ps are of three types: polarity constructions (I didnt see anyone), free-choice constructions (You may have
any/whichever one) and focus constructions (Even he came in). In the
0
former two, Ps contain a and a wh-element. The following examples show a consistent spread of Ps, marked with brackets, across the
board of old IE languages.
Moving westward, we start with Indo-Iranian. Both R
. gvedic and postVedic Sanskrit show the non-coordinate use of the coordinating particle
ca, where it forms a free-choice expression of the wh-ever-type (36a,36b),
or a negative polarity item (36c). When not combined with a wh-host,
the particle forms an additive expression with focus semantics, akin to
the function ofalso/even in English, as shown in (36d).
(36) Vedic & Classical Sanskrit:
a.
dhi
tda
m
ca]
a
vi
s
vam
modate
y
a
t
[k
m
pra
pr
thivy
a
m
.
.
.
world exults which [what ] world.f.acc upon
this
dhu
yady- abhyupetam
[kva
ca]
s
a
dhu
asa
.
if
promised to be accepted where honest dishonest
va kr. tam
may
a
.
or done.pst.part 1.sg.instr
If you accept whatever I may do, whether honest or dishonest.
(Bh
agavatapur
an. a, 8.9.12)
c.
m
ay
a?
ca] tititarti
[ka
s
na yasya
neg whom.gen [who.m.sg ] able to overcome illusions.pl
No one [=not anyone] can overcome that (=the Supreme Personality of Godheads) illusory energy.
(Bh
agavatapur
an. a, 8.5.30)
21
d.
[cintayam
ca] na pasyami bhavatam
. s. prati
thinking.pres.part neg see.1.sg you
unto
vaikr. tam
offence.acc
Even after much thinking, I fail to see the injury I did unto
(Mah
abh
arata, 2.20.1)
you.
(Gell. 4,2,3)
22
Mitrovi
c
Gothic:
a. visxad uh
gaggis
[ishvad uh] (. . .) gaggis.
[where ]
go.2.sg.pres.act.ind
wherever you go
(Mat. 8:19)
b. j ah xaz
uh saei
hauseiv
waruda
jah [hvaz
uh] saei
hausei
waurda
and who.m.sg and pro.m.sg hear.3.sg.ind words.acc.pl
meina
meina
mine
And every one that heareth these sayings of mine
(Mat. 7:26)
In Old Church Slavonic, there were two kinds of particles: i and ze,
both of which were conjunctive; ze was adversative in nature and historically related to Greek (translating as but). In non-coordinate uses,
i was additive-focal (cf. Sanskrit ex. 36d), while ze combined with whhosts to form a negative polarity item or a free-choice expression. The
former additive and the latter free-choice functions are shown in (39a)
and (39b), respectively.
(39)
posu
[i togo]
ku
sent.3.pl.aor [ him.m.sg.acc] to then.pl.dat
He sent also him to them.
(Mar. 12:6)
b.
ei
izdrece
etvoj
o
su kl
with oath.f.sg.ins promised.3.pl.aor her.f.sg.dat
sprositu
ste vu
ze] a
[ego
dati
give.inf [what ] if ask.sg.pres
With an oath he promised to give her whatsoever she would
ask.
(Mat, 14:7)
23
We also find the additive use of the coordinator pe (< k e?) in Tocharian:
(40)
Tocharian:
h
emintuyo
rt
a. [n
ypic olyiyam
sa
.
[jewels.pl.inst full ship.f.sg.loc caravan.m.sg.obl
Jambudvipac
pe]
Jambudvipa.m.sg.allt and/]
mura
s. ,
t
ya
s. pa
having been made.supp.abs.m.sg.abl seven
ukac
kom
kn
wram
. sa
.
day.m.pl.perlt neck.sg.allt water.sg.loc
With a caravan to Jambudvipa also having been made in a
ship filled with jewels [. . . ]
taka, 5 )
(tA, Pun
. yavanta-Ja
w
While Classical Armenian did not possess the enclitic k e-type coordinator, we can ascertain its loss in the pre-Classical period, since the
w
remnant k e still shows in fossilised non-coordinate form, with the semantics aligned with the remainder of other IE languages.
(41)
Classical Armenian:
a.
ete [ok] . . .
if who-
If anyone [strike (thee) upon thy right cheek . . . ]
(Mat.,
[erbek] . . .
[time.loc ]
At any time/ever.
24
Mitrovi
c
(42)
Hittite:
a.
ul
nu-wa
[kuit ki] sakti
and-quot neg [who ] know.2.sg.pres
You know nothing (=not anything)
(KUB XXIV.8.I.36)
b.
nu dumu.me
s-U [kui
s
s-a] kuwatta
utn
e
paizzi
J sons.his
who- = somewhere country.loc went
Each of his sons went somewhere to a country.
(KBo.
3.I.1.1718)
c.
nu [kuitt-a]
arhayan kinaizz[i
J what- = seperately sifts
She sifts everything seperately.
(KUB XXIV.11.III.18)
Old Irish ch, itself a reflex of PIE k e, aside from the coordinate function, also creates free-choice (43a) and universal quantificational (43b,43c)
expressions.
(43)
Old Irish:
a. [ce
ch] taibre
[what ] give.2.subj
what[so]ever thou mays give.
(Zu ir. Hss. 1.20.15; Thurneysen 2003: 289)
b. [ce
ch] orr
[what ] slay.3.m.subj
whichever he may slay.
(Anecd. ii.63.14.h; Thurneysen 2003: 289)
c. a
25
* in kdo
J who
anyone/whoever
(45)
i (t)ko
who
anyone/whoever
26
Mitrovi
c
The second position effect has its traditional aetiology in what is known
as Wackernagels Law. Wackernagel (1892) is credited to have dubbed the
one generalisation that applies to the syntax of PIE, namely that some
elements consistently occupy the second position in a given string of
words, or, in modern terminology, in a given constituent. Suffice it to
say that the 1892 generalisation is far beyond explanatory: it is solely
a descriptive observation pertaining to word count. An explanation is,
however, feasible in a theory of syntax which, for instance, attributes all
configurational (word order related) differences to differences in movement. There have essentially been two theoretically different approaches
to the explanatory account of Wackernagels Law. Although both theories see the cause of the second position effect in movement, one confines this movement to narrow syntax while another places the movement in the post-syntactic module where prosody is king.
The purpose of this section is not to categorically suggest a confinement space wherein the W(ackernagel)-movement takes place, but to
suggest an over-arching factor of the distribution of the second position
effects that the IE coordination data suggests. This factor, as it were,
is the phasal architecture, to which not only the syntactic derivation is
subject but also the phonological and prosodic processes that follow it.
A Wackernagel element like our (Lat. -que, Hom. -te, Lat. que, Goth.
Lat. -uh, Skt. Lat. -ca, etc.) has a requirement which demands be pre5
min
The clitic host is necessarily (of the size of) a head; we do not come across entire categories preceding enclitics.
27
head, X transferred to the two interfaces for semantic and phonological processing (interpretation and externalisation respectively). A phase
therefore not only partitions narrow syntactic derivation into logical building blocks but also delimits post-syntactic operations and synchronises
them with narrow syntax. In this direction, Samuels (2009: 242) takes
as a starting point the conceptual argument laid out in the foundational
work by Marvin (2003: 74): If we think of levels in the lexicon as levels of
syntactic attachment of affixes, we can actually say that Lexical Phonology suggests that phonological rules are limited by syntactic domains,
possibly phases. Samuels thus proposes a Phonological Derivation by
Phase (PDbP), which relies on a cycle that is not proprietary to phonology. (Samuels, 2009: 243) Combining Samuelss theory with a recognition of post-syntactic movement, we should predict the domain or scope
of such operations based on the narrow syntactic derivation. Assume
in (46) is a Wackernagel-type coordinator specified with [ ], which represents the requirement for peninitial placement. Lets assume it takes
a phasal complement X P, which has ZP as its specifier and YP as its complement.
(46)
.
P
.
[]
.
.0
.Z
X.P
.
ZP
0
X.
.
.
YP
. ...
.
a. -checkable terminals narrow syntactically:
28
Mitrovi
c
janayan ma
nave
si ca ti
a
apa
ks. a
m
for.men created.mid.3.sg.m earth (J) water
For men he created the earth and water.
On the other hand, a structure like the one in (48) could only be an instance of post-syntactic movement since the target of movement is syntactically inaccessible and incorporable (head-immovable) as the set of
0
-accessible terminals is in fact empty (null C ) and does not contain the
wh-terminal, which originates within the specifier of the k
artv
a-headed
CP. Assuming phonology doesnt have to readsyntactic boundaries,
since it just applies to each chunk as it is received (Samuels, 2009:
250), the syntactically inaccesible wh-temrinal y
a is made available to
ni
rtv
ca ti ka
a
ya
i
kr. ta
made.prt. (J) which.rel to.be.made.fut.part
29
so
hanti
raks. a
slay.pres.3.sg demons.acc.pl
a
The present verb hanti seems to sit in T with the object, the demons,
lower in the structure, presumably in its V-complementing in situ position. Assuming the category of (49) is that of CP, we see that CP edge is
0
dination, the [ ] feature on would not be deleted. Given our assumptions, the derivation would crash due to this. Structure in (50) sketches
this scenario, where there are no syntactically or post-syntactically ac0
material below .
30
Mitrovi
c
(50)
.
JP.
.. ...
J.
.
.
.
[]
.
.
P
C.P
.
.
.
empty edge
C.
0
empty X
.
TP
.
.
inaccessible
otherwise silent J receives phonological realisation for -checking reasons. The full internal coordination structure of (49) is given in (51). The
0
last resort mechanism qua phonological realisation of J may be analogised to expletive subjects in a language like English. Just as there is no
subject (in the vP) eligible to raise to [Spec, TP] in sentences like it is raining,, an expletive subject is realised as last resort. Equally, when there
0
(51)
so
raks. a
a hanti
u -t
J slay.pres.3.sg demons.acc.pl
31
a
The proposed analysis is also an explanation of an empirical generalisation that has not only been extensively shown to hold not only in
R
. gvedic (Klein 1985a,1985b) and Old Persian (Klein, 1988) but across the
vast array of ancient IE languages (Klein 1992, Agbayani and Golston 2010).
(52)
categorial generalisation:
Peninitial coordinators do not head clausal coordinations.
Since clauses (CPs) are inherently phasal (Chomsky 2001, et seq.), they
provide the selecting head with far less search space, or in the case of
(51), an empty set of possible incorporees. In non-CP coordinands, [ ]
0
may be checked by virtue of access to terminals in s complements interior. The derivation of non-clausal coordination is therefore strictly
0
/scalar semantics (36)(43). The third J -cycle obtains a sysntactic structure for coordination. Diachronically, the change occurs in the collaps0
ing of the second and third cycles, whereby and J feature in a single
cycle and thereby inherently yielding bimorphemic coordinators, mor0
under J , as instantiated in (53b). The interdependence of the J- complex is empirically and technically analogous to proposals by Chomsky
0
(2008) and Richards (2007), among others, who claim that T is lexically
defective, bearing no -features of its own, and instead inherits its 0
32
Mitrovi
c
(53)
a. III.
. JP.
II.
.Pint
.
I.
. XP
.
.. ...
.
[ ]
.
. .
0
.
.ZP
J.
.
.
.
YP
.0
.X
. ...
b. II.
I.
. XP
.
. JP.
.. ...
.
J.
0
.
.
.
.Pint
.
0
X.
.
YP
.
.
. ...
.
.*
.
[.]
.
This view predicts that the loss of enclitic monomorphemic coordinators, and the inverse rise of the inherently initial bimorphemic coordinators, entails the loss of independent P, which features in focal additive, polar and scalar construction as in (36)(43). This is in fact confirmed.
7
0
1. Clausal coordination type generalises to all categories as comes preinstalled with a hosting morpheme.
The only exception to this diachronic interlock between changes in word order and se0
mantics, would be a case where would not carry [ ] and thus would not get buried
0
under J in time. The Slavonic branch is such an exception, which has lexically syncre0
0
tised the entries for J and as i but the semantics of the coordinate/non-coordinate
constructions clearly shows that two forms of i existed in OCS, which is preserved in
most branches of synchronic Slavonic. See ? for details.
ge
II
a
st
P
JP
st
fir
e
ag
st
33
p
res
se
on
JP
0
en
em
ov
-m
r
st
fir
ns
po
es
t
sor
t re
las
34
Mitrovi
c
J , which acts as checker. We have thus derived the two empirical generalisations on IE coordination.
(54)
a.
b.
35
Gonda (1954: 182) continues to note that the relation between the copulative [coordinate] ( A) and the epic [non-coordinate] ( B) has
never been correctly formulated. It is hard to envisage a correct formulation without the the precise tools that theoretical models make available
and with which we have proposed a rather detailed formulation of this
very relation.
36
Mitrovi
c
references
Agbayani, B. and Golston, C. (2010), Second-position is first-position:
Wackernagels law and the role of clausal conjunction, Indogermanische Forschungen: Zeitschrift f
ur Indogermanistik und allgemeine Sprachwissenschaft
115, 121.
Bach, E. (1964), An Introduction to Transformational Grammars, New York:
Holt, Rinehart & Winston.
Beekes, R. (2010), Etymological Dictionary of Greek, Leiden Indo-European Etymological Dictionary Series, Leiden: E.J. Brill.
Bloomfield, L. (1933), Language, New York: Holt, Rinehart & Winston.
37
tsverlag.
Heidelberg: Carl Winter Universita
Klein, J. S. (1985b), Toward a Discourse Grammar of the Rigveda. Part 2., Vol. II,
tsverlag.
Heidelberg: Carl Winter Universita
Klein, J. S. (1988), Coordinate Conjunction in Old Persian, Journal of the
American Oriental Society 108(3), 387417.
Klein, J. S. (1992), Some Indo-European Systems of Conjunction:
Rigveda, Old Persian, Homer, Harvard Studies in Classical Philology 94, 1
51.
Klein, J. S. (1997), Indefinite pronouns, polarity and related phenomena
in Classical Armenian: A study based on the Old Armenian gospels,
Transactions of the Philological Society 95(2), 189245.
Looijenga, T. (2003), Texts & Contexts of the Oldest Runic Inscriptions, Vol. 4 of
The Northern World, E.J. Brill, Leiden/Boston.
38
Mitrovi
c
Marvin, T. (2003), Topics in the Stress and Syntax of Words, PhD thesis,
Massachusetts Institute of Technology.
Meillet, A. (1908), Introduction `
a letude comparative des langues indoeuropeennes,
Paris: Hachette.
Mitrovi
c, M. (2014), Decomposing connectives, PhD thesis, University
of Cambridge.
Muadz, H. (1991), Coordinate structure: a planar representation, PhD
thesis, University of Arizona.
Myler, N. (2012), A note on den dikkens (2006) arguments for a j(unction)
head. Ms. NYU.
Richards, M. (2007), On feature inheritance: an argument from the
phase impenetrability condition, Linguistic Inquiry 38, 563572.
Richards, N. (2014), Uttering Theory. Unpublished Monograph. MIT.
Rizzi, L. (1997), The fine structure of the left periphery, in L. Haegman,
ed., Elements of Grammar, Dordrecht: Kluwer, pp. 281337.
Samuels, B. D. (2009), The Structure of Phonological Theory, PhD thesis,
Harvard University.
Slade, B. M. (2011), Formal and philological inquiries into the nature of interrogatives, indefinites, disjunction, and focus in Sinhala
and other languages, PhD thesis, University of Illinois at UrbanaChampaign.
Szabolcsi, A. (2013), What do quantifier particles do? Ms. NYU.
Thurneysen, R. (2003), A Grammar of Old Irish, Dublin: Dublin Institute for
Advanced Studies.
39