Sie sind auf Seite 1von 16

ON THE FORM AND MEANING OF DOUBLE NOUN INCORPORATION

1



ALESSIO MURO
Universit di Padova


Noun Incorporation Constructions (NICs) commonly involve one noun root (N) and
one verb root (V), but in some languages the V may be compounded with two Ns,
which almost always have the semantics of a theme and a non-theme (a problematic
fact for Bakers (1988) theory). Moreover, if the non-theme N specifies instrument
or manner, the linear order in which the two Ns appear may vary, one N taking
scope over the other. This fact too can be a problem for Bakers theory, as well as
for any theory making use of a fixed Thematic Hierarchy of semantic roles (TH). By
reviewing the most salient features of NICs in general, a theoretical apparatus is
proposed which dispenses with the notion of head movement in favor of XP
movement, yet preserves the idea of a TH by means of a phase-based model of word
formation.

Keywords: Noun incorporation, phases, XP movement.


1. Introduction

A Noun Incorporation Construction (NIC) is a verbal form built on a complex base,
which results from the compounding of a verbal root (V) and a nominal one (N), as
can be seen in the example below from Chukchi (Polinskaja & Nedjalkov
1987:240):

(1) a. otloge motqomot kawkawok kilinin (Chukchi)
otlog-e m mm motqo otqo otqo otqo.[R] kawkaw-ok kili kili kili kili-nin
father-ERG butter.ABS bread-LOC spread.on-3SG.A>3SG.P

b. otlog-on kawkaw-ok m mm motqo otqo otqo otqo-rkele rkele rkele rkele-ge
father-ABS bread-LOC butter-spread.on-3SG.S
Father spread butter on the bread.

In (1a), the word motqomot butter appears as a free-standing form, whereas in
(1b) it appears compounded with the V root (thereby losing its functional
morphology). The most influential proposal to explain contrasts like (1a) vs. (1b)
above is probably Bakers (1988, 1996, Baker et al. 2004) head movement account,

1
I would like to thank the Scientific Committe of IMM15 for their valuable comments on an earlier
version of this paper. For helpful suggestions I also thank Mark C. Baker, Paola Beninc, Guglielmo
Cinque, Federico Damonte, Alberto M. Mioni and Cecilia Poletto. The usual disclaimers apply.



according to which the head of an NP (base-generated as a complement to the V
root, as in (2a)) moves to adjoin to the V head (2b):

(2) a. [
S
[
NP
otloge ] ... [
VP
[
NP

N
motqomot ]
V
kili- ]]
b. [
S
[
NP
otlogon ] ... [
VP
[
NP
t
i
]
N
motqo
i
-
V
rkele ]]

This analysis makes two important predictions: first of all, only themes (direct
objects and unaccusative subjects) should be incorporated; moreover, noun
incorporation (NI) is expected to target only the head of the theme NP (and never
an XP).
In this paper, I will concentrate on a phenomenon that has received very
little attention, i.e. the incorporation of two N roots into one single V root (double
NI), as in the following Classical Nahuatl example (Launey 1978:168):

(3) ni-mitz-m mm m - -- -tn tn tn tn- -- -nmiqui nmiqui nmiqui nmiqui (Cl. Nahuatl)
1SG.A-2SG.P-hand-lip-meet
I kiss your hands, lit. I meet your hands with my lips.

According to Baker (1988), double NI should not exist at all, since a V is supposed
to assign one and only one structural case: if a second incorporated noun (IN) were
added to the core NIC, at the very least a violation of the principles of case theory
would arise.
2

In the next section, I will clarify what types of constructions qualify as
double NI. In order to provide a theoretical account of the phenomenon, I will
introduce in section 3 three alternations to be found in NICs cross-linguistically.
My theoretical proposal is exposed in section 4 and applied to the data in section 5,
where two subtypes of double NI will be distinguished on the basis of a
morphosyntactic dimension (M-phase structure), which is conditioned by the
semantics of -roles (only indirect arguments can trigger M-phase mismatches).

2. What counts as double NI?

Not every verbal word containing a V root and two N roots qualifies as a double
NIC. In this section, I will illustrate the methodology I followed in collecting the
data for this work, first of all by showing three types of data which superficially
look similar to double NI, but actually instantiate (or interact with) different
phenomena.



2
A double NIC where the outer IN should designate the possessor of the first one is ruled out by
both case theory and the Empty Category Principle.


2.1. Abnormal speech types

The following example from Koyukon Athapaskan is reported as a case of double
NI (Axelrod 1990:183):

(4) %netlaatlooketsedenletso (Koyukon)
ne-tlaa tlaa tlaa tlaa- -- -tloo tloo tloo tloo - -- -ke-tse-de-ne--le-ts ts ts tso oo o
there.and.back-head-rope-something-HUM.S-TH-TH-M/ASP-STAT-
move(FLEXIBLE.IN.S)
Well go around with a rope tied to our necks.

In the original source (Attla 1983:88-89), the speaker is a bear, and in the
following sentence, the storyteller translates (4) into more normal Koyukon, which
is very likely to mean that the word is grammatically odd: in the absence of more
evidence, this example alone cannot prove that the language has double NI. Rather,
a form like (4) seems to fit into what Sapir (1915) terms abnormal types of speech:
among the phenomena he mentions, there are morphological processes as well. To
qualify as a NIC in my sense, a word must be fully interpretable.

2.2 Incorporation of N-N compounds

Double NI is attested in the Chukotko-Kamchatkan language family, as we shall
see (18, 20); but not all cases of multiple incorporated N roots are amenable to such
an explanation. In the following Koryak example, the INs are not semantically
independent of each other; rather, they form a compound N-N root (M.Kurebito
2001:45):

(5) t-o-[ [[ [l ll loq oq oq oql ll la aa a - -- -kemet kemet kemet kemet' '' 'a aa a] ]] ]- -- -n nn ncocmav cocmav cocmav cocmav-o-k (Koryak)
1SG.S-EV-[winter-clothes]-prepare-EV-1SG.S.PST
I prepared winter clothes.

I consider two INs an instance of double NI only if they are semantically
independent of each other.

2.3 Double NICs where one IN is an applied object

NI can interact with other morphological processes, like applicative formation. This
is interesting, since in Bakers (1988) theory, a difference is assumed between
cyclic and acyclic combinations of grammatical function changing processes.
Applicative formation, in particular, should apply only after NI, and there should
be no cases of incorporation of an applied object. As I showed in Muro (2009), this
prediction too is not borne out, as can be seen in the following example from



Wichita. In this word, the outer IN is the applied object of the portative applicative
-ri (Rood 1976:266):

(6) kiya--a:-ki-riwa:c riwa:c riwa:c riwa:c-aras aras aras aras-[r]a-ri-kita kita kita kita- -- -a a a a- -- -hi:rik-s (Wichita)
QUOT-3SG.A-CIS-large-meat-COLL.P-PORT-top-come-ITER-IPFV
He brought the large (quantity of) meat to the top in several trips.

Here, we have an instance of NI followed by applicative formation, but, contrary to
Bakers prediction, theme NI occurs a second time after the applicative is merged.
A different case is attested in Takelma: here the applied object is the inner IN,
which is introduced by an instrumental applicative (Sapir 1922:68):

(7) gwen gwen gwen gwen- -- -waya waya waya waya- -- -sg sg sg sg
u uu u
t tt t-hi (Takelma)
neck-knife-cut-INS
He cut their necks off with his knife.

After the incorporation of the applied object, another instance of NI takes place
(lexical affixation, in Muros (2006) terms, since the root is a bound body part
prefix). Examples like (7) should be distinguished from double NI with M-phase
mismatch (see 5.1).
The last two examples are highly instructive and relevant to the topic of this
paper, but since they involve an additional complication (i.e. the presence of an
applicative formation process), they have to be left to future work.
For the above illustrated reasons, in selecting the data I tried to choose
examples that meet the requirements of full interpretability, semantic independence
of the INs, and absence of intervening grammatical function changing processes
between the two instances of NI. Moreover, I excluded examples where the inner
IN is a term referring to the body or a body part, since these Ns are particularly
prone to grammaticalization, and it is often difficult to tell whether, in a particular
construction, one has to do with a true IN or a functional morpheme. The examples
I was able to find all include one IN performing the role of theme and one with
non-theme semantics.
3


3. More about NI: the alternations

3.1 The directionality alternation

In NICs, INs can either precede their incorporating V hosts (preverbal NI, as in (8),
from Mohawk) or follow them (postverbal NI, as in (9), from Mapudungun)
(examples from Baker et al. 2004:139):

3
Rice (2000:106) reports an example involving two adjunct-like INs, for which, however, I have
not yet been able to provide an analysis.


(8) wa?-ke-nakt nakt nakt nakt-a-hninu hninu hninu hninu-? (Mohawk)
FACT-1SG.A>3N.P-bed-EV-buy-PUNC
I bought the/a bed.

(9) i chao kintu kintu kintu kintu-waka waka waka waka-le-y (Mapudungun)
1SG.POSS father seek-cow-PROG-IND-3SG.S
My father is looking for the cows.

According to Baker et al. (2004), this alternation can be analyzed in terms of left or
right adjunction of the moved X to its host, dispensing with the antisymmetric
view of head movement (cf. Kayne 1994, Baker 1996). This has undesirable
results, since the two types show different properties (see Muro 2006, 2009):
preverbal INs can undergo several kinds of morphological processes such as
reduplication, plural marking, compounding with adjectival modifiers (both to their
left and right sides) and, in the case of nonadjacent NI (which will be introduced in
3.2), they can even take possessive morphology (see the discussion in Muro
2009:145ff.); in any case, they almost never undergo any kind of phonological
erosion or simplification. The picture is very different with Ns incorporated in
postverbal position, which almost never take any kind of modification and quite
often undergo phonological simplification. As in Muro (2009), I assume preverbal
NI to instantiate the basic order, whereas the postverbal order can be derived by
movement (see further, 4.2).

3.2 The adjacency alternation

NICs also differ with respect to whether the IN occurs immediately adjacent to the
V root (adjacent NI, as in (10), from Sapir 1911:260), or else other morphemes
intervene (nonadjacent NI, as in (11), from Axelrod 1990:185):

(10) ni-naca naca naca naca- -- -qua qua qua qua (Cl. Nahuatl)
1SG.S-meat-eat
I eat meat.

(11) kkosotseeyhyeetaayh (Koyukon)
kko-sotseeyh sotseeyh sotseeyh sotseeyh-ye-ee--taayh taayh taayh taayh
around-happiness-3SG.P-ASP-CAUS-lie(NPL.AN.S)
S/he is very happy, lit. happiness carries him/her around.

This alternation only occurs with preverbal NICs (postverbal ones always have
adjacent members: see Muro 2009). This should be all the more reason to deal with
each of the two types in different ways (see below, 4.2).




3.3 The transitivity alternation

The last alternation I am going to illustrate does not directly concern the core of the
NIC (i.e. the N-V or V-N compound), but rather its interactions with object
agreement: in some languages (again only with preverbal NICs), theme NI does not
affect the transitivity of the verbal complex, which agrees with the IN itself
(valence-neutral NI, as in (12), from the Tanoan language Southern Tiwa: Allen et
al. 1984:294-5); in other languages, NICs with incorporated themes are intransitive
(antipassive NI, as in (13), from Classical Nahuatl: Launey 1978:166) (see also (1)
above):

(12) a. seuan seuan seuan seuan-ide ti ti ti ti-m mm mu uu u-ban (S. Tiwa)
man-AN.SG 1SG.A>3AN.SG.P-see-PST

b. ti ti ti ti-seuan seuan seuan seuan-m mm mu uu u-ban
1SG.A>3AN.SG.P-man-see-PST
I saw the/a man.

(13) a. ni-c-nmaca nmaca nmaca nmaca tlaxcal tlaxcal tlaxcal tlaxcal-li (Cl. Nahuatl)
1SG.A-3SG.P-sell tortilla-NMLZ

b. ni-tlaxcal tlaxcal tlaxcal tlaxcal- -- -nmaca nmaca nmaca nmaca
1SG.S-tortilla-sell
I sell tortillas.

The valence-neutral character of NI is used by Baker (1988, 1996, Baker et al.
2004) as a diagnostic for the syntactic nature of NI, as opposed to a lexical nature.
Antipassive NI may also be syntactic, but valence-neutral NI is more clearly so,
since the visibility of the IN to a more clearly sentence-level process such as
agreement is directly testable on the basis of overt morphology, which proves that
NICs of this type are radically different from lexical compounding.
4


4. Theoretical framework

4.1 Preliminary assumptions

In this work, I will pursue the approach outlined in Muro (2009:ch.2): the basic
lines of Antisymmetry (X-bar theory, right-branching only, leftward movement)
are maintained; unlike Baker et al. (2004), I will propose an account of the

4
A theoretical account of the Transitivity Alternation is not possible in this work for space reason.
For a discussion, see Muro (2009: ch. 2).


observed facts in terms of XP movement (Koopman & Szabolcsi 2000, Cinque
2007, 2010), rather than head movement.
This will be made possible by dispensing with the traditional view that the
theme NP is the complement of the V root (V): rather, the latter is argued to
instantiate an XP starting the derivation of the verbal complex (cf. Hale & Keyser
2002), and the theme argument is assumed to vary in its internal structure (it can be
an XP, which will give rise to preverbal NI, or an X, resulting in postverbal NI).
5

I follow Baker (1988) in the assumption that the sequence of projections
instantiating argument structure is projected in a fixed order (in compliance with
the Uniformity of Theta Assignment Hypothesis, UTAH). In particular, I assume
the following hierarchy of thematic projections, building on recent work by
Damonte (2004:90 [with my additions]):

(14) The Thematic Hierarchy (TH)(relevant portion):
Ag(ent)P > ... > Goal/Source/ReasonP >
Ins(trumental)/Man(ner)/Loc(ative)P > ... > Dat(ive)P > Th(eme)P > V

According to Chomsky (1999), the syntactic derivation is sent off in clusters of
structure to LF and PF for interpretation; the derivation must converge at vP and
CP, his two syntactic phases. Building on previous work (mainly Pylkknen 2008),
I propose a specifically morphological dimension of phases (morphological phases
or M-phases, notated between vertical bars). An M-phase should be understood as
a segment of syntactic structure which is sent to Spell Out as a unit, thus defining a
morphological boundary; the morphological cohesion between any two elements
should be determined in an incremental fashion (the activation of a higher phase
implies the activation of all the lower phases; the deeper any two elements are
embedded inside M-phases, the more tightly they will be bound together
morphologically, and the more impenetrable their phase will be to higher functional
processes). The lowest M-phase is the root of a lexeme (with its category-checking
projection P as its edge, cf. Wiltschko 2009). Since languages vary a lot in
defining their word boundaries (i.e. there is considerable variation among
languages in terms of degree of synthesis), higher M-phases may or may not be
activated. All of the languages exemplified in this paper display polysynthetic
morphologies, which means that in these languages, all or nearly all M-phases are
active. For our present purposes, I assume three M-phases: ||, ||, and |I| (where ||
is conceived as corresponding to Chomskys vP, and |I| to the edge of the field of
Agr projections).
A key feature of M-phases is that they can be reanalyzed as lower phases: I
argue that this is due to a morphosyntactic operation I term Relabel: thus, a || M-
phase can be relabeled and enter the derivation as a || phase, as in (15):

5
This will of course imply a redefinition of the concept of government, which can no longer play a
key role in the theory I propose; for the time being, this problem must be left to future work.



(15) Relabel || ... || |||| ... ||

The double vertical bars indicate that Relabel implies additional morphological
cohesion among the constituents of the phase. Under Relabel, the -role of any
argument previously merged within || is neutralized, and the exponent of that
argument is reinterpreted as a simple modifier of the host V root.
6
This need not be
a diachronic process, and the modifying element can retain its referential
properties, only changing its syntactic status. With respect to double NICs, this
proposal allows to explain the exceptions to the TH (what I call M-phase
mismatches, see 4.1) without having to assume total lexicalization of the
constructions in question: this will be shown to come in useful (see the Chukchi
data in (18) below).

4.2 The core proposal

Given the premises set forth above, NI appears not to be the result of the movement
of the IN; in fact, it seems to be more convenient to think that the latter does not
move at all. The process thus turns out to be a special configuration in the context
of V movement (|V|,|IN|=|P|):

(16) Preverbal (theme) NI

a.* ThP b. ThP
2 2
|IN| ThP |IN| ThP
5 2 5 2
Th |V| Th |V|
2 2
P P
5 5



As can be seen, in (16a), the IN is an XP blocking V movement to [Spec, ThP]. For
the derivation to succeed, the [IN-V] complex must move as a whole, resulting in
preverbal NI (16b). Postverbal NI results from the presence of an X-sized IN,
which allows for the movement of the V root to [Spec, ThP]:



6
In this respect I depart from the original explanation proposed in Muro (2009: ch.2), where I
hypothesized that an adjunct IN could be merged directly as a modifier to the V root in a ModP
projection inside ||.


(17) Postverbal (theme) NI

ThP
2
|V| ThP
5 2
Th |V|
IN 2
P
5


XP movement can also be useful to explain the adjacency alternation introduced in
3.2: the fact that the IN in (11) appears higher than agreement could mean that it
has moved to a projection above the edge of the |I| phase. This kind of long
movement is widely assumed to be a property of XPs, not of Xs.
With these premises in mind, let us now proceed to a survey of the available
data about double NI, trying to derive the attested variation by means of the
proposed theoretical apparatus.

5. Languages with double NI

5.1 Languages with M-phase mismatch

In some of the languages displaying double NICs, the innermost IN is not the
theme: (3) above, from Classical Nahuatl, instantiates such an example. A more
detailed picture can be gained from the Western Chukchi data below (T. Kurebito
1998:107-8):

(18) a. kuke kuke kuke kuke-o miml miml miml miml-e t-o-nilu nilu nilu nilu-gen (W. Chukchi)
pot-ABS.SG water-INS 1SG.A-EV-wash-3SG.P

b. t-o-kuk kuk kuk kuk- -- -iml iml iml iml-o-nilu nilu nilu nilu-gek
1SG.S-EV-pot-water-wash-1SG.S

c. kuke kuke kuke kuke-o t-iml iml iml iml-o-nilu nilu nilu nilu-gen
pot-ABS.SG 1SG.A-water-wash-3SG.P

d. *t-o-kuke kuke kuke kuke- -- -nilu nilu nilu nilu-gek miml miml miml miml-e
1SG.S-EV-pot-wash-1SG.S water-INS
I washed out the pot with water.




In (18a), the V bears transitive agreement, and both Ns appear as independent
words; (18b) shows that this dialect allows double NI. Of the two arguments, the
one appearing closer to the V root in (18b) performs an instrumental role;
furthermore (18c-d) illustrate that it is not possible to incorporate the direct object
N at the expense of the incorporation of the instrumental N. It can be observed that
the agreement on the V in (18c) is still transitive, notwithstanding the incorporation
of the instrumental N (Chukchi has antipassive NI). The agreement markers in
(18b), on the other hand, show intransitive agreement, as expected from the fact
that the direct object N has been incorporated.
I take this as evidence that the lowest -role visible to agreement is still
Theme, notwithstanding the fact that the theme IN comes outside of the
[instrumental.IN-V.root] complex. But then, given the TH and the facts from
agreement, how can the unexpected linear order be explained? I propose that the
two INs in (18b) are merged under an M-phase mismatch, with the theme IN
merged after the phase containing the [instrumental.IN-V.root] complex has been
relabeled. The derivation of the relevant portion of (18b) is shown in (19):

(19) M-phase mismatch:
|
V
|, Merge [
Ins/Man/LocP
|
N
|
Ins/Man/Loc
] ||[
Ins/Man/LocP
|
N
|
Ins/Man/Loc
]|
V
|;

||[
Ins/Man/LocP
|
N1
|
Ins/Man/Loc
]|
V
|, Relabel || ||
V
||[
ModP
|
N1
|
Mod
]|
V
|;
||
V
||[
ModP
|
N1
|
Mod
]|
V
|, Merge [
ThP
|
N2
|
Th
]
||[
ThP
|
N2
|
Th
]||
V
||[
ModP
|
N1
|
Mod
]|
V
|.

Note that one cannot simply assume that the [instrumental.IN-V.root] complex is
lexicalized, since the double NIC (18b) is fully transparent and is even described as
a synonym to the fully analytical version (18a), with both Ns left unincorporated.
The range of -roles associated with double NICs with M-phase mismatches
seems to coincide with the range of roles attributed by Damonte (2004) to the
thematic projection Ins/Man/LocP. Below I provide another example from Western
Chukchi, where the inner IN shows the semantics of a manner adjunct (T. Kurebito
1998:111):

(20) a. kuke kuke kuke kuke-o j jj jom om om om-nen k kk keli eli eli eli-mil (W. Chukchi)
pot-ABS.SG put.on-3SG.A>3SG.P cap-like

b. koka koka koka koka- -- -k kk kale ale ale ale- -- -j jj jop op op op-ge
pot-cap-put.on-3SG.S
He put on the pot like a cap.

The reason why an M-phase mismatch should be triggered by this projection only
remains to be worked out. It can be observed, however, that even if every double
NIC with an M-phase mismatch I was able to identify has an Ins/Man/LocP


component as the inner IN, the reverse is not true, i.e. not all double NICs
involving an Ins/Man/LocP must necessarily show an M-phase mismatch (see (21)
below).
On a theoretical level, it should also be noted that double NICs with M-
phase mismatches are not necessarily problematic for Bakers theory: it could be
said, in principle, that only the outer IN is an instance of syntactic NI, the inner one
being the result of a lexical derivation. The best evidence to evaluate Bakers
theory on the basis of double NICs should therefore not involve an M-phase
mismatch. Let us now survey the available data that appear to meet this
requirement.

5.2. Double NICs with no M-phase mismatch

A double NIC, as I just said, does not necessarily imply an M-phase mismatch:
there are cases where both INs must be assumed to be projected within the same
phase (in full compliance with the TH). The following Southern Paiute example
(Sapir 1921:16) illustrates this (the language has M-phase mismatches in other
constructions):

(21) [ [[ [[wii [wii [wii [wii] ]] ]- -- -[to [to [to [to- -- -kuchum kuchum kuchum kuchum- -- -punku punku punku punku] ]] ]- -- -rgani rgani rgani rgani] ]] ]-yugwi-va-nt-m() (S. Paiute)
[[knife]-[black-buffalo-pet]-cut.up]-sit(PL.S)-FUT-PART-AN.PL
they who are going to sit and cut up with a knife a black cow (or bull)

Even if the above construction is a participial form, we must assume the IN to be
adjacent to the V root, since the core of the double NIC (the complex [knife]-
[black-bovine]-[cut.up]) is in turn incorporated into the V sit: if one or both of
the INs lay within a different phase from ||, scope ambiguity could arise, since the
INs could in principle be arguments of either V root. The derivation of the relevant
portion of (21) must therefore be as follows:

(22) |
V
|, Merge [
ThP
|
N1
|
Th
] ||[
ThP
|
N1
|
Th
]|
V
|;
||[
ThP
|
N1
|
Th
]|
V
|, Merge [
Ins/Man/LocP
|
N2
|
Ins/Man/Loc
]
||[
Ins/Man/LocP
|
N2
|
Ins/Man/Loc
][
ThP
|
N1
|
Th
]|
V
|.

Note that in (21) the outer IN again performs an instrumental function; the data I
am going to illustrate next, instead, involve direct and indirect objects of
prototypical ditransitive Vs (i.e. give).
The first example is from Lakota (De Reuse 1994:215):








(23) s(g) s(g) s(g) s(g)- -- -mni mni mni mni- -- -ku ku ku ku (Lakota)
dog
7
-water-give
to water horses

The author shows evidence that the segment s(g)-mni dog-water is not to be
interpreted as a compound; unfortunately, this example does not show the
interactions with agreement. In this respect, more instructive evidence comes from
Taos Northern Tiwa (Harrington 1910:28,35):

(24) a. sn-na (Taos)
man-AN.SG

n
-k u uu u- -- -p p p p- -- -w ww w -hu
3SG.AN.A>3.P>3SG.AN.D-horse-water-give-IPFV
The man is giving the horse some water.

b. sn-na
man-AN.SG

n
-k k k ku uu u- -- -p p p p- -- -h h h h -hu
3SG.AN.A>3.P>3SG.AN.D-horse-water-take.away.from-IPFV
The man is taking the water away from the horse.

In these examples, agreement is overt; moreover, the agreement prefix is from the
ditransitive set, i.e. it bears the features of both the INs (NI is valence-neutral in the
Kiowa-Tanoan languages). This proves that syntax can access the features of both
the INs, and that the two cannot form a compound.
In my opinion, the examples in (24) are important evidence: given the facts
from agreement, which prove the syntactic visibility of the INs, the data show
clearly that Bakers prediction regarding the impossibility of double NI following
from the uniqueness of structural case assignment by the V root is not borne out.
The derivation for the core portions of (24a,b) should be as in (22), with DatP
replacing Ins/Man/LocP.
So far, I have only discussed examples from languages with preverbal NI;
however, double NI is also available in languages with postverbal constructions, as
in the following Sora data (Anderson & Harrison 2008:355-6):

(25) a. ji ji ji ji- -- -lo: lo: lo: lo:- -- -si: si: si: si:-t-am (Sora)
stick-earth-hand-NPST-2.P
Mud will stick to your hand.


7
In Lakota, a horse is termed ska wakha (dog be.holy), holy dog. In NICs, only the head N
appears compounded as s(g)-; the constructions are nevertheless understood as referring to horses.


b. en aj-ja: ja: ja: ja:- -- -dar dar dar dar- -- -si: si: si: si:-am
1SG.A NEG-receive-cooked.rice-hand-2.P
I wont receive rice from your hand.

c. jo jo jo jo- -- -me me me me- -- -bo:b bo:b bo:b bo:b-dem-te-n-ai
smear-oil-head-REFL-NPST-INTR-1.S
I will anoint myself with oil.

These examples show the theme IN as the inner member of the NIC: they are thus
compatible with a derivation along the lines of (17) above, with no TH violation, as
shown in (26) (XP=Ins/Man/LocP for (25a,c), Goal/Source/ReasonP for (25b):

(27) |
V
|, Merge [
ThP Th
N1] ||[
ThP

Th
N1]|
V
|;
||[
ThP

Th
N1]|
V
|, Move |
V
| ||[
ThP
|
V
|
Th
N1];
||[
ThP
|
V
|
Th
N1], Merge [
XP X
N2] ||[
XP X
N2][
ThP
|
V
|
Th
N1];
||[
XP X
N2][
ThP
|
V
|
Th
N1], Move [
ThP
|
V
|
Th
N1]
||[
XP
[
ThP
|
V
|
Th
N1]
X
N2].

At the end of the derivation, the surface order of the components is the mirror
image of that found in the preverbal double cases (21, 23, 24).

6. Conclusions

In this paper, I tried to show that the phenomenology of double NICs, analyzed
from the viewpoint of the linearization of their components, involves several
factors, to my knowledge never considered together in previous accounts of NI.
First of all, it should be said that the alternations in directionality (3.1) and
adjacency (3.2) (and very likely the transitivity alternation as well, see Muro
(2009:77-87)) seem to suggest an internal structure for preverbal INs, which appear
to be more complex than heads. The same should be said about V roots, since their
movement properties are more easily explained in terms of XP movement.
The complex nature of V roots also emerges from the varying degrees of
morphological cohesion that the latter can exhibit with respect to theme and non-
theme INs in double NICs. In some cases, the surface order appears as non-theme
IN > theme IN > V root; in other cases, we have the order theme IN > non-theme
IN > V root (in violation of the TH). The observed TH violations could result from
M-phase mismatches, due to a morphosyntactic operation (Relabel) targeting the
edges of M-phases. From the point of view of the semantics of -roles, a (one-way)
correlation can be observed between M-phase mismatches and the presence of INs
with values fitting within the range encompassed by the projection Ins/Man/LocP
on the TH. Dative INs do not seem to trigger M-phase mismatches.



When the TH is respected, I assume no M-phase mismatch has taken place.
The existence of double NICs of this type is a problem for Bakers theory, which
predicts that only theme arguments should undergo NI. By considering the
interactions of NI with agreement as illustrated in 3.3 (valence-neutral NI should be
syntactic), I tried to show that the available data (most clearly (24), from Northern
Tiwa) cannot be explained away as lexical idiosyncrasies; rather, they require a
radical revision of the head movement account for NI. I hope the solution proposed
here will provide a viable alternative.


ABBREVIATIONS

1,2,3=1,2,3 person; A=agent-like argument of transitive verb; ABS=absolutive;
AN=animate; ASP=aspect; CAUS=causative; CIS =cislocative; COLL=collective; D=recipient-
like argument of ditransitive verb; ERG=ergative; EV=epenthetic vowel; FACT=factual
mood; FUT=future; HUM=human; IN=inanimate; IND=indicative; INS=instrumental;
INTR=intransitive; IPFV=imperfective; ITER=iterative; LOC=locative; M=mood; N=neuter;
NEG=negative; NMLZ=nominalizer; NPL=nonplural; NPST=nonpast; P=patient-like
argument of transitive verb; PART=participle; PL=plural; PORT=portative; POSS=possessive;
PROG=progressive; PST=past; PUNC=punctual; QUOT=quotative; R=reduplication;
REFL=reflexive; S=single argument of intransitive verb; SG=singular; STAT=stative;
TH=thematic prefix.

REFERENCES

Allen, Barbara J., Donna B. Gardiner & Donald G. Frantz. 1984. Noun
Incorporation in Southern Tiwa. International Journal of American
Linguistics 50.3, 308-309.
Anderson, Gregory D.S. & K. David Harrison. 2008. Sora. The Munda
Languages ed. by Gregory D.S. Anderson, 299-380. London & New York:
Routledge.
Attla, Catherine. 1983. Sitsiy Yugh Noholnik Tsin: As My Grandfather Told It.
Fairbanks, Alaska: Alaska Native Language Center.
Axelrod, Melissa. 1990. Noun Incorporation in Koyukon Athapaskan.
International Journal of American Linguistics 56.2:179-95.
Baker, Mark C. 1988. Incorporation: a Theory of Grammatical Function
Changing. Chicago: The University of Chicago Press.
1996. The polysynthesis parameter. Oxford: Oxford University Press.
Baker, Mark C., Roberto Aranovich & Luca A. Golluscio. 2004. Two Types of
Syntactic Noun Incorporation: Noun Incorporation in Mapudungun and its
Typological Implications. Language 81.1:138-76.
Chomsky, Noam. 1999. Derivation by Phase. MIT Occasional Papers in
Linguistics 18. Cambridge, Mass.: Department of Linguistics, MIT.


Cinque, Guglielmo. 2007. The Fundamental Left-Right Asymmetry of Natural
Languages. University of Venice Working Papers in Linguistics 17:77-107.
2010. The Syntax of Adjectives: A Comparative Study. Cambridge, Mass.: MIT
Press.
Damonte, Federico. 2004. The Thematic Field: The Syntax of Argument Structure
Enhancing Morphology. Ph.D. diss., Padua University.
Hale, Kenneth & Samuel J. Keyser. 2002. Prolegomenon to a Theory of Argument
Structure. Cambridge, Mass.: MIT Press.
Harrington, John P. 1910. An Introductory Paper on the Tiwa Language, Dialect
of Taos, New Mexico. American Anthropologist 12.1:11-48.
Kayne, Richard S. 1994. The Antisymmetry of Syntax. Cambridge, Mass.: MIT
Press.
Koopman, Hilda & Anna Szabolcsi. 2000. Verbal Complexes. Cambridge, Mass.:
MIT Press.
Kurebito, Megumi. 2001. Noun Incorporation in Koryak. Languages of the North
Pacific Rim 6, ed. by Osahito Miyaoka & Fubito Endo, 29-57. Osaka:
Faculty of Informatics, Osaka Gakuin University.
Kurebito, Tokusu. 1998. A Report on Noun Incorporation in Chukchi. Languages
of the North Pacific Rim 4, ed. by Osahito Miyaoka & Minoru Oshima, 97-
113. Kyoto: Graduate School of Letters, Kyoto University.
Launey, Michel. 1978. Introduction la langue et la littrature aztques. Tome 1:
grammaire. Paris: LHarmattan.
Muro, Alessio. 2006. Lexical Affixation in Salish and Wakashan and its
Relevance for a Theory of Polysynthesis. Padua Working Papers in
Linguistics 2:1-28.
2009. Noun Incorporation: A New Theoretical Perspective. Ph.D. diss., Padua
University.

(paduaresearch.cab.unipd.it/2113/1/Muro_NI.pdf)
Polinskaja, Marija & Vladimir P. Nedjalkov. 1987. Contrasting the Absolutive in
Chukchee. Lingua 71:239-69.
Pylkknen, Liina. 2008. Introducing Arguments. Cambridge, Mass.: MIT Press.
de Reuse, Willem J. 1994. Noun Incorporation in Lakota (Siouan). International
Journal of American Linguistics 60.3:199-260.
Rice, Keren. 2000. Morpheme Order and Semantic Scope: Word Formation in the
Athapaskan Verb. Cambridge: Cambridge University Press.
Rood, David S. 1976. Wichita Grammar. New York & London: Garland.
Sapir, Edward. 1911. The Problem of Noun Incorporation in American
Languages. American Anthropologist 13:250-282.
1915. Abnormal Types of Speech in Nootka. Canadian Department of Mines,
Geological Survey Memoir 62, Anthropological Series 5:1-22.
1921. Language. An Introduction to the Study of Speech. New York: Harcourt,
Brace & World.




1922. The Takelma Language of Southwestern Oregon. Handbook of
American Indian Languages, Part 2. Bureau of American Ethnology,
Bulletin 40, ed. by Franz Boas, 1-296. Washington, D.C.: Smithsonian
Institution.
Wiltschko, Martina. 2009. Root incorporation: Evidence from Lexical Suffixes
in Halkomelem Salish. Lingua 119:199-223.

Das könnte Ihnen auch gefallen