A Radical Construction Grammar analysis of noun phrase

structure

William Croft, University of New Mexico ([email protected])

1. Introduction: Why Radical Construction Grammar?

1.1. The remarkable persistence of American (Post-Bloomfieldian) structuralism in con-
temporary syntax

In the next chapter of Syntactic Structures, Chomsky discusses the ‘Limitations of phrase
structure descriptions’ (1957: Ch. 5)…In Chomsky’s argument, what he called the
‘customary’ method served in that sense as a straw man. But, unlike the more usual kind
of straw man, it was not abandoned when its inadequacies had been demonstrated. In-
stead it was incorporated into the real man that replaced it…By adopting this solution
[transformations], Chomsky effectively protected the constituency model of the Post-
Bloomfieldians from any fundamental criticism. There was no objection to it that could
not be met by deriving one constituency structure from a series of underlying constituen-
cy structures. Therefore its merits were not seriously debated. (Matthews 1993:149, 153)

Chomsky achieved this major change of direction [syntactic theory as a theory of lan-
guage acquisition] without apparently having to change anything else. A ‘grammar’ was
now, in one sense, something internal to a speaker. Let us call this a grammar in sense 1,
or ‘grammar1’. In another sense, a ‘grammar’ was, as it had previously been, a series of
statements constructed by a linguist. Let us call that a grammar in sense 2, or ‘grammar2’.
The main point, therefore, was that a grammar2 was now to be interpreted not just as a
characterization of a language—that is, of a set of potential utterances in the sense of
Bloomfield, Harris or Syntactic Structures—but as an account of what enables a speaker
to speak and understand it. A grammar2 is thus an account of, or description of, or theory
about, a grammar1. But at the same time it was assumed that grammars1 had essentially
the properties that grammars2 already had…All that happened was that our object in writ-
ing them had been redefined.
As for grammars, so for theories of grammar…The reification is thus complete on
both levels. A theory2 is now interpreted as a linguist’s account of the child’s theory1, a
grammar2 as an account of a resulting grammar1. But both the theory1 and the resulting
grammars1 are assumed to have the properties that theories2 and grammars2 had had in
Chomsky’s earlier work…Therefore the ways in practice in which theories2 and gram-
mars2 were developed did not change in the slightest. (Matthews 1993:211-12)

-1-
1.2. Syntactic argumentation and its problems
Syntactic argumentation: find tests (or arguments) that a syntactic unit (e.g. a category or a con-
stituent) exists.
Example: direct object category (vs. oblique prepositional phrase; Croft 2001:35)

(1) a. Jack kissed Janet.

b. Janet was kissed by Jack.

(2) a. The old man walked with a cane.

b. *The old man walked a cane.
c. *A cane was walked with by the old man.

Some problems in using distributional analysis for justifying syntactic units are recognized even
in introductory formal syntax textbooks:
It is important to note that when these tests are successful then what you have is an argu-
ment that a particular sequence [of words in a sentence] forms a constituent, rather than
proof that it does. It is even more important to notice that when one of these tests fails,
we cannot conclude that the sequence is not a constituent, we can only conclude that
there is no evidence that it is. (Adger 2003:66, emphasis original)

Obviously this description of the VP-ellipsis construction assumes that there is a VP con-
stituent in English. But that is exactly what we have been trying to prove; are we guilty of
circular reasoning here? Actually, we are not. What we are doing is appealing to a basic
methodological principle in linguistics which states that a simple analysis is normally
preferred over a more complex analysis…If we assume that there is a VP constituent, we
can give a very simple analysis of the data. If there were no such constituent, then it
would be much more difficult to specify precisely what the omitted elements in this con-
struction must be. (Kroeger 2004:36)

But their solutions do not really solve the problem (Croft 2001:34-36):
(3) a. Jack weighs 160 pounds.
b. *160 pounds is weighed by Jack.

(4) a. Claude Debussy lived in this house.

b. *Claude Debussy lived this house.
c. This house was lived in by Claude Debussy.

-2-
•Adger is actually wrong in describing syntactic practice. Failing a test is often taken as an argu-
ment against the identity of a syntactic unit (e.g., failing to form a Passive is normally taken as
evidence against the Direct Object status of the argument in the counterpart Active).
•When one test succeeds, and another fails (cf. Adger), which test do we use as an argument in
favor/against the identity of a syntactic unit? Passive or Prepositionless Postverbal NP?
•When distributional tests do not match, then is it really simpler to assume the global (cross-con-
structional) identity of a syntactic unit, let alone cross-linguistic identity (Kroeger)? You have to
add some ad hoc device to explain why 160 pounds doesn’t take a preposition, and this house
can passivize. And almost all of the time, distributional tests do not match if you examine the full
range of units that are supposed to fill the role in the construction used for the distributional test.

Two other serious problems with the traditional form of syntactic argumentation

•Most “tests” do not exist in all languages, e.g. VP Ellipsis, or Passive (Croft 2001:29-32). If you
use different tests, how do you know that it’s the same syntactic unit across the languages (Direct
Object, Adjective, etc.)?
•Finally, this approach assumes that one can identify constructions to begin with, namely the
constructions that are used in the tests/arguments

1.3 Radical Construction Grammar (Croft 2001, 2003)

•Constructions are a pairing of form and meaning, organized in a taxonomic network, like other
construction grammars (see Croft and Cruse 2004, part III).
•Constructions are the basic, primitive units of syntactic representation. Nonreductionist: the
primitive units are complex structures (compare Gestalt psychology).
•Syntactic units are defined only in terms of the constructions in which they occur.
•Generalizations across constructions can only be discovered (and justified) by cross-linguistic
comparison (as done in typological theory; Croft 2003). This never reveals the uniformity of syn-
tactic structure/patterning assumed from Bloomfield to contemporary theories and presupposed
in the notion of ‘tests’ for categories, constituency etc. Instead it reveals complex patterns of
variation of the encoding of function in grammatical form, constrained at most by implicational
universals or semantic maps, and explained symbolically (how function is encoded in form).
•How does one identify constructions? (This is also a problem for the structuralist approach; see
just above.) Identifying constructions is a categorization problem, and displays the same charac-
teristics as other categorization phenomena (prototypes and internal structure to the category, un-
predictable extensions, frequency effects, manipulable boundaries, etc.) However, since con-

-3-
structions are pairings of form and function, functional properties as well as formal ones play a
role in categorization.
•A further hypothesis of RCG is that dependency relations are ultimately semantic (functional).
The morphosyntax traditionally taken to encode syntactic dependency relations (case marking,
indexation) are actually symbolic relations linking syntactic elements to semantic components.

2. Constituency

Constituency: a formal grouping of elements. The grouping is assumed to be global, that is, oc-
curs across constructions. The grouping forms a category, such as Noun Phrase, that is usually
assumed to be universal, that is, recurs across languages.

“Tests” for constituency (mostly from McCawley 1998, but widely found elsewhere; see Croft
2001:185-97)

(a) Movement, and the danger of assuming identity of units across constructions
(5) a. [A guy [who I hadn’t seen since high school]]?? came in.
b. [A guy] came in [who I hadn’t seen since high school].

(6) A man entered and a woman left [who had met in Vienna].
(McCawley 1998:771)

(7) A man and a woman [who had met in Vienna] entered and left respectively.

(b) Pronominalization

(8) a. [The little girl] sat [on the couch].

b. [She] sat [here].

(9) a. the [blue flowers]N'

b. I [bought some bread]VP.

(10) Compré [el libro]NP ayer.

‘I bought the book yesterday.’

(11) [Lo]Prn compré ayer.

‘I bought it yesterday.’

-4-
(c) Ellipsis

(12) Lydia won’t [go to San Francisco]VP, but I will [Ø].

(13) a. [Los gatos]NP estaban durmiendo.

‘The cats were sleeping.’
b. [Ø] Estaban durmiendo.
‘(They) were sleeping.’

(14) Stripping: John plays poker on Fridays, but Ø not Ø bridge Ø.

(McCawley 1998:55; elided: John…plays…on Fridays)

(d) Coordination

(15) a. [The boys]NP and [the girls]NP should come to the front.
b. I [went outside]VP and [looked for my cat]VP.
c. I’m not going because [I’m tired]S and [my feet hurt]S.

(16) Gapping: Jenny gave [the books to Randy]?? and [the magazines to Bill]??.
Jerry ate [the steak in the kitchen]?? and [the ice cream in the living room]??.

(17) Right Node Raising: [Jenny makes]?? and [Randy sells]?? the prints.

…constituents are neither essential nor fundamental to linguistic structure, but emerge in
special circumstances. They reflect the basic psychological capacity for grouping, which
is guided by the gestalt principles of contiguity and similarity (Langacker 1997:9, empha-
sis original).

Formal grouping: essentially contiguity (Langacker 1997)

•Separability is based in part on meaning (conceptual grouping)

(18) a. I found [the Ø tieclasp]

b. I found [the gold tieclasp]
c. I found [the *yesterday tieclasp]
d. I found [the whatsitcalled tieclasp]

-5-
•Contiguity comes in degrees

(19) a. “I never [got really the full story]”—Carol T, 24.5.84

b. “and [the father in the meantime of this family]…” [speaker went on to describe the
father]—Melissa B, 31.5.98
c. “These things aren’t yet in [wide enough spread] use”—Terry W, 23.4.84

•Formal grouping in discourse

–Occurrence in a single intonation unit, which is also a probabilistic affair (Croft 1995, 2007)
–Unit of first-position self-repair (Fox & Jasperson 1995, Hopper 2000)

3. Noun phrase structure in Radical Construction Grammar

The issue: some linguists argue that in certain languages, words that are analyzed as modifiers in
a noun phrase in English (and other familiar languages), should be analyzed as NPs on their own,
in apposition to each other and to an NP containing the noun they appear to be modifying.
Miraña (Seifart 2005:153, 155)

(20) ó ɯhkɯ́-ʔi maːkı́ nı́ -mɯ-βá -kɛ kɯʔɾı́ -mɯ-kɛ

1SG take-PRED three-ANIM.PL-PL-ACC pintadillo-ANIM.PL-ACC
‘I caught three pintadillo (fish, sp.).’

(21) dó ʔhḼ̇ -ko dó ʔhḼ̇ -ko nɛː-mɛ má Ḽ̇ ʔtsɯ-kó
carguero-1D.POINTED carguero-1D.POINTED say-ANIM.PL medium-1D.POINTED
‘Carguero (tree, sp.), carguero they call (this), a medium-sized one’

Instead of expressing modification in noun phrases with a hierarchical internal structure

(e.g. head noun and adjective), it is a common strategy in Miraña to use a number of noun
phrases in apposition in a clause to provide additional information about a referent. The
coreferential reading of such noun phrases in apposition is achieved by noun class
marking (possibly in addition to number and case marking). (Seifart 2005:150)

Yimas (Foley 1991:182, 181)

(22) ama-na-kn wayk-k wa-n

1SG-POSS-V.SG buy-IRR go-IMP
‘Go buy my (betelnut).’

-6-
(23) patn ama-na-kn mpa-n
betelnut(V.SG) 1SG-POSS-V.SG one-V.SG
mpa-n patn ama-na-kn
mpa-n ama-na-kn patn
patn mpa-n ama-na-kn
ama-na-kn patn mpa-n
‘my one betelnut’

The latter construction is a scrambled paratactic construction, with the two nominals in
apposition to each other. (Foley 1991:182)

3.1. Arguments used to support the appositional NP analysis

Arguments listed below are used to argue for apposition in 22 languages, most of which are cited
in Rijkhoff (2002), who argues for appositional NP analyses himself in some cases where the au-
thor of the language documentation does not
•Discontinuity (D): the modifier is sometimes not contiguous to head noun (7 languages)
•Free Order (F): the modifier may either precede or follow the head noun (3 languages)
•Prosody (P): the modifier can occur in a separate prosodic unit, usually defined by pause (4
languages)
•Agreement (A): the modifier inflects for some or all of the same categories as the head noun,
such as number, gender and case (3 languages)
•Agreement Alternation (A2): the modifier does not agree in one construction, but does in a sec-
ond construction, often when the modifier follows the head noun (A2p; see Croft 1994; Croft and
Deligianni 2001; Rijkhoff 2002:275) (10 languages)
•Null Head (N): the modifier can stand alone, denoting ‘the one that is Modifier’, with the same
inflections as when it is modifying a head noun (5 languages; in one language, Georgian, the
modifier takes the nominal forms of the inflections instead of the modifying forms of the
inflections)
•Article (T): the modifier takes an article as well as the head noun, in one construction but not
another, again often when the modifier follows the head noun (5 languages)
•Classifier (C): a numeral takes a classifier (2 languages)
•Pronoun: a personal pronoun occurs with the modifier expression (3 languages). In all cases,
this represents a relationship between an argument index (predicate agreement) and a referring
expression
•Other (4 languages): some other, language-specific process that is used to support the claim, in-
cluding an assertion of apposition without any evidence provided

-7-
3.2. Methodological assumptions: Global Extension of Construction-Specific Analyses
Global Extension of Construction-Specific Analyses—an analysis supported by evidence in one
construction is assumed to hold for all other constructions.
•Discontinuity: if constituency is defined in terms of formal grouping, then discontinuity implies
two NPs. But then it is assumed that the same structural analysis is valid for NPs in which modi-
fiers and the head noun are contiguous.
A reason for not extending the analysis from discontinuous NP constructions to contiguous NP
constructions: discontinuous NP constructions are special in their information structure (=dis-
course function):
–Polish (Siewierska 1984): initial part = contrastive focus, final part = contrastive topic
–Gooniyandi (McGregor 1997): initial part = theme, final part = unmarked focus (if final
part in different intonation unit then it is an afterthought; I suspect this is possible for any
language with anaphoric head NPs)
–Wardaman (Merlan 1994): initial part = thematic, final part = information focus
–Babungo (Schaub 1985:123): final numeral = counterassertive
–Ngiti (Kutsch Lojenga 1994:355): initial numeral = “strong emphasis”

•Null Headed NPs: since modifier words in null headed NP constructions are NPs in themselves,
and they are identical in form when they occur contiguous to a head noun, then the NP analysis
is extended to the contiguous modifiers
A reason for not extending the analysis from headless NPs to modifiers contiguous to head
nouns: a null headed NP must have a pragmatically accessible (Ariel 1990) head, usually ex-
pressed in a preceding clause, while a headed NP has its head in the same clause. The Georgian
data referred to above suggest that the null headed clause is a different construction.

•The alternative: Treat each construction individually, with its own form and function. Cross-
constructional similarities in analysis are abstractions inductively derived by speakers (depend-
ing on type frequency, iconicity, etc.), not the basic building blocks of constructions.

3.3. Methodological assumptions: Isomorphism

Isomorphism
(a) difference of form always entails difference in “underlying” structure or function;
(b) identity of form always entails identity of “underlying” structure or function.
•Difference in form, in languages with two alternative modifier constructions, is taken to be a
difference in “function”, or more precisely, abstract structure: modification vs. apposition.

-8-
But abstract structure is dichotomous; only two alternatives are possible (cf. unergative vs. un-
accusative). What if a language has three, or more, different constructions?
–Miraña: Genitives are contiguous and always precede; Numerals can be discontinuous
but always precede; Relative Clauses can be discontinuous and may either precede or
follow
The structural dichotomy cannot account for the three-way difference in constructions.
•Identity of a modifier’s form in discontinuous NP, in null headed NP, and in a headed NP re-
quires identity of analysis
Some distributional differences between true apposition and “appositive” modifiers:
–True apposition (at least the paradigm case; Matthews 1981:224-25) involves two core-
ferring object words: my brother, the poet. “Appositive” modifiers involve an object
word and a property word (deictic word, numeral word, etc.) denoting a thing possessing
the property (or deixis, cardinality, etc.) that is coreferential with the object word referent

–True apposition may have independent grammatical specifications: French mon plat
préféré, une truite meunière ‘My favorite dish [MASC], pan-fried trout [FEM]’. “Apposi-
tive” modifiers always share their grammatical specifications with the nominal NP (in
traditional parlance, they agree)

–True apposition almost always occur in separate intonation units; “appositive” modifiers
overwhelmingly occur in the same intonation unit

Table 1. Putative modifiers and true appositive phrases in Wardaman (Croft 2007, Table 21)
Whole Pct Broken Pct Total
True appositive NPs 18 12.9% 121 87.1% 139
“Modifiers” 352 89.1% 43 10.9% 395
Total 370 69.3% 164 30.7% 534
Chi-square = 276.72, p < .001

•The alternative: The variationist approach. Alternative structures are different ways of saying
the same thing (Weinreich, Labov and Herzog 1968), e.g. alternative constructions used for mod-
ification. The same structure may also be used to say different things, e.g. an inflected modifier
used in different ways. The mapping between form and function is many-to-many: multiple
forms for the same function, and multiple functions for a given form. The many-to-many map-
pings need not be captured by positing different underlying abstract formal structures.

-9-
3.4. Methodological assumptions: Methodological Opportunism
Methodological Opportunism: Choose tests that support the assumed analysis; assume tests al-
ways coincide and therefore support the existence of a syntactic unit independent of the construc-
tions used to test for it (see §1.2 above).
•The alternative: construct a typology of the constructions used, identify the universals gov-
erning them, and offer a functional explanation for the universals (Croft 2003)
•NP structure: A sample of 45 languages assembled by Mark Donohue and myself, geographi-
cally and genetically diverse, but with an emphasis on languages of Australia and New Guinea.
•Nine semantic types of modifiers were investigated:
–deixis
–definiteness
–possession
–numerals
–quantifiers
–set-member (picking out a member of a contextually established set, e.g. next, other,
last, first, second etc.)
–properties
–actions
–locations

•For each semantic type in each language, the following information was encoded: the
morphosyntactic coding strategy (see Croft 2003, chapter 2), including word order, for headed
modifier constructions; what construction was used for anaphoric head modifiers (‘the/a Modifi-
er one’); and observations if any on discontinuity. Other examples of discontinuity in the linguis-
tic literature were also examined.
•Of a maximum possible 384 examples of attributive constructions, data on morphosyntactic
strategy was obtained for 325 constructions (21 languages lacked definiteness markers). Data on
anaphoric head constructions was obtained for 152 constructions, somewhat less than half the
total.
Typological Universals:
I. Discontinuity of modifier and head ⊃ agreement of modifier ⊃ null head in anaphoric
head construction
Only two exceptions: one set-member word in Supyire, and the deictic modifiers in Vani-
mo, require a “dummy” head and yet the modifier agrees with the head noun

II. Free modifier-head order ⊃ null head in anaphoric head construction.

In other words, freedom of word order is not part of the implicational hierarchy in I.
Specifically, discontinuity does not entail freedom of word order.

- 10 -
The various “tests” for head+dependent vs. appositional NPs do not coincide. Therefore there is
no evidence for a dichotomy between these two constituent structure analyses, across languages
or within them (in the case of a language with two different constructions, such as Yimas).

3.5. Explaining the implicational universals

•The head+dependent vs. appositive NPs analysis may lead one to think that the proper account
for the implicational hierarchy in I is in terms of degrees of formal integration. But if that were
the case, you would expect discontinuity ⊃ null head ⊃ agreement. Agreement would permit null
heads, and a discontinuous modifier would be a null-headed NP that is not contiguous to its head.
•But that is not what we find. Agreement ⊃ null head, not the other way around. In fact this is
exactly the same pattern we find for indexation of arguments on predicates: null arguments (i.e.
where the antecedent is in the preceding discourse) occur even if agreement is absent on the
predicate (Gilligan 1987). This tells us that the determining factor for this implicational hierarchy
is not degrees of formal integration but the whereabouts of the nominal referent that functions as
the head:
–Discontinuity: modifier and head in the same clause
–Agreement: modifier and head in the same clause or the head is accessible in the pre-estab-
lished context (since agreement implies null heads)
–Null head: the head is accessible in the pre-established context
•We can see that in fact the function of agreement (indexation) is not to “permit” a modifier to be
null headed in a clause—since a modifier can be null headed in a clause without agreement—but
to index the relationship between a modifier and its head. Then it is possible (in some languages)
to permit discontinuity of modifier and head—precisely because agreement indexes the relation
of the modifier to its head.
•This suggests that an expression like Spanish la moderna ‘the modern (one).FEM.SG’ is not a
property concept that has been converted to a noun, with a semantic shift to ‘a thing possessing
the property’, but in fact a modifier with null instantiation of the “head” referring expression
(Fillmore 1986; Croft 2001:275-80). A discontinuous modifier is probably just that, a modifier of
a noun which is not contiguous to it, although in some cases—such as the Miraña example 21
above, and its English translation—it is a null headed, prosodically separate independent NP.
•This analysis of the function of indexation fits with the fact that the occurrence of indexation is
a manifestation of the typologically less marked structure (the modifier exhibits greater behav-
ioral potential by virtue of indexing a referent). For example, on the whole, semantically proto-
typical adjectives are more likely to index their heads than less prototypical semantic modifier
types (Croft 1991:82), and indexation is a indicator of typologically less marked arguments (core
arguments, especially the nominative/absolutive argument; Croft 2004:147, 152).

- 11 -
•So, in all cases above, one can make the case that the modifiers are performing a modification
function. Modification is a functional relationship between two concepts, the denotation of the
modifier (property, deictic relation, cardinality, etc.) and the referent. So it is the underlying de-
pendency, not a constituency relation, that these implicational universals reflect.
•What we see here, then, is a semantic relationship (modifier-referent), that is reflected in formal
grouping most of the time, sometimes with a null instantiation when the referent is highly acces-
sible in the discourse; but sometimes is not reflected in a formal grouping. In the latter case, this
is often because a formal grouping would violate a discourse-functional (information-structural)
separation, as noted above for discontinuous NP constructions.

4. Conclusions

•“Absence” of a head is best analyzed as a null instantiation of a referent. As with null instantia-
tion of arguments, “agreement” is not necessary to allow null instantiation. Null instantiation is
simply a common way for languages to encode a highly accessible referent.
•“Modifiers” really are that: performing a modification propositional act function, though some-
times with null instantiated referents. This reflects a dependency relation, not a constituency
relation.
•“Agreement” does not exist to “allow” a modifier to stand as an NP on its own; a modifier
doesn’t need “agreement” to do that. But agreement helps to link a modifier to its head, which al-
lows for a discontinuous dependency (not constituency) between modifier and head.
•When a modifier is contiguous with its head, it represents a formal grouping that reflects the
conceptual grouping represented by the modification propositional act function, as well as the
valency relation between modifier and head (i.e. the modifier denotes a property possessed by
the referent of the head). So the appositional NP analysis of contiguous modifiers is inappropri-
ate, at least for prosodically unified examples.
•“Free” modifier-head order of modifiers does not fit into the implicational hierarchy because
presence of a head noun and contiguity to it are what defines the implicational hierarchy, and
“free” modifier-head order presupposes contiguity to a head noun.
•“Discontinuous noun phrases” probably have split information/discourse functions which moti-
vates their lack of formal grouping (compare Croft 2001:201). Since these are specialized con-
structions, with specialized functions, one cannot assume that they can be identified with the
Noun Phrase (dependent modifier + head noun) construction found in other, more common sen-
tence constructions.
!

- 12 -
References

Adger, David. 2003. Core syntax: a minimalist approach. Oxford: Oxford University Press.
Ariel, Mira. 1990. Accessing noun phrase antecedents. New York: Routledge.
Chomsky, Noam. 1957. Syntactic structures. The Hague: Mouton.
Croft, William. 1991. Syntactic categories and grammatical relations: The cognitive
organization of information. Chicago: University of Chicago Press.
Croft, William. 1994. Asymmetries in word order. Paper presented to the Word Order Group,
EUROTYP Plenary Conference, European Science Foundation, Le Bischenberg, France.
Croft, William. 1995. Intonation units and grammatical structure. Linguistics 33.839-882.
Croft, William. 2001. Radical Construction Grammar: syntactic theory in typological
perspective. Oxford: Oxford University Press.
Croft, William. 2003. Typology and universals, 2nd edition. Cambridge: Cambridge University
Press.
Croft, William. 2007. Intonation units and grammatical structure in Wardaman and in
crosslinguistic perspective. To appear in Australian Journal of Linguistics.
Croft, William and D. Alan Cruse. 2004. Cognitive linguistics. Cambridge: Cambridge
University Press.
Croft, William and Efrosini Deligianni. 2001. Asymmetries in noun phrase order. Paper
presented to the International Symposium on Deictic Systems and Quantification in
Languages Spoken in Europe and Northern Central Asia, Udmurt State University, Izhevsk,
Russia.
Fillmore, Charles J. 1986. Pragmatically-controlled zero anaphora. Proceedings of the Twelfth
Annual Meeting of the Berkeley Linguistics Society, ed. Vassiliki Nikiforidou et al., 95-107.
Berkeley: Berkeley Linguistics Society.
Foley, William. 1991. The Yimas language of New Guinea. Stanford: Stanford University Press.
Fox, Barbara A. & Robert Jasperson. 1995. A syntactic exploration of repair in English
conversation. Alternative linguistics: descriptive and theoretical modes, ed. Philip W. Davis,
77-134. Amsterdam: John Benjamins.
Gilligan, Gary Martin. 1987. A cross-linguistic approach to the pro-drop parameter. Ph.D.
dissertation, University of Southern California.
Hopper, Paul. 2000. Grammatical constructions and their discourse origins: prototype or family
resemblance? Paper presented at the 28th LAUD Symposium, Landau, Germany.
Kroeger, Paul R. 2004. Analyzing syntax: a lexical-functional approach. Cambridge: Cambridge
University Press.
Kutsch Lojenga, Constance. 1994. Ngiti: a Central Sudanic language of Zaire. Köln: Rüdiger
Köppe Verlag.
Langacker, Ronald W. 1997. Constituency, dependency, and conceptual grouping. Cognitive
Linguistics 8.1-32.
Matthews, Peter H. 1981. Syntax. Cambridge: Cambridge University Press.
Matthews, Peter H. 1993. Grammatical theory in the United States from Bloomfield to Chomsky.
Cambridge: Cambridge University Press.
Merlan, Francesca. 1994. A grammar of Wardaman. Berlin: Mouton de Gruyter.
McCawley, James D. 1998. The syntactic phenomena of English (2nd ed.). Chicago: University
of Chicago Press.
McGregor, William B. 1997. Functions of noun phrase discontinuity in Gooniyandi. Functions of
Language 4.83-114.

- 13 -
Rijkhoff, Jan. 2002. The noun phrase. Oxford: Oxford University Press.
Schaub, Willi. 1985. Babungo. (Croom Helm Descriptive Grammars.) Dover, New
Hampshire: Croom Helm.
Seifart, Frank. 2005. The structure and use of shape-based noun classes in Miraña (north west
Amazon). (MPI Series in Psycholinguistics, 32.) Ph.D. dissertation, Radboud Universiteit
Nijmegen, the Netherlands.
Siewierska, Anna. 1984. Phrasal discontinuity in Polish. Australian Journal of
Linguistics 4.57-71.
Weinreich, Uriel, William Labov, & Marvin I. Herzog. 1968. Empirical foundations for a theory
of language change. Directions for historical linguistics, ed. Winfrid P. Lehmann and Yakov
Malkiel, 95-195. Austin: University of Texas Press.

- 14 -
Languages and arguments (page references to language documentation [just number] or to
Rijkhoff 2002 [R + number], modifiers are Dem, Art, Num, Qnt, Adj, RC)
Nama (Hagman 1974): A2p-Dem, Gen, Num, Adj (46, R21)—could be nonrestrictive or after-
thought (46)
Beja (Hudson 1976): A2p/T-Dem, Gen, Num, Adj, RC (106) (R275 - R’s claim, not Hudson’s;
the difference is that the definiteness inflection is added)
Wellegga Oromo (Gragg 1976): postposed Gen with relativizer (R272 - R’s claim, not Gragg’s)
Oromo (Bender et al. 1976, cited by R): postnominal position-Adj (144) (R271)—possibly
nonrestrictive
Tswana (Cole 1955): no argument given-Dem (130, 435-36)
Georgian (Testelec 1998, cited by R): A2p-Adj (652, R22, R125)
Burmese (Wheatley 1987): C-Num (851, R162)
Nivkh (Comrie 1989): A2p/C-Num (269; R21fn23, R161)
Korean (Lee 1989): no genitive linker (117-18; R164) (Sohn 1989:353 analyzes no-linker con-
struction as quantifier float; but case is repeated, so more like discontinuity)
Samoan (Mosel & Hovdhaugen 1992): A2, T (292-Dem, 320-Num, 317-RC; R{pg})
Galela (van Baarda 1908, cited by R): C/T-Num (44; R163)
Nasioi (Rausch 1912, cited by R): N/A2-Dem, Adj (122, 131-32; R269)
Alamblak (Bruce 1984): pronoun (96; R32)—refers to we people type of apposition
Yimas (Foley 1991): D, F, N, A2-Gen, Adj, Dem (181-83)
Ngiyambaa (Donaldson 1980): pronoun=definiteness marker (128; R192)—possibly clausal
indexation
Mangarayi (Merlan 1982): D-Gen [Num, Adj illustrated] (49; R256fn4), placement of focus par-
ticle bayi (49)—but the latter may be relevant for information structure
Ngalakan (Merlan 1983): D, P, F, N, A-Dem, Gen, Num, Adj (80-83; no arguments explicitly
given)
Nunggubuyu (Heath 1984): P, N {more} (499-506; R200, 207)
Hixkaryana (Derbyshire 1979): D-Num (103-4; Derbyshire describes them as ‘adjuncts’)
Miraña (Seifart 2005): D, P, N, A-Dem, Num, Adj, RC, F-RCs only; language-specific tonal
process (151, 154); positioning of clitics (154); pronoun
Cubeo (Morse & Maxwell 1999): D, P, A (94-95)—but illustrated with classic (headed) apposi-
tional structures; pronoun (95)—but this is focus or maybe a left-dislocation construction
West Greenlandic (Fortescue 1984): (118-19; R326fn14)-appears to be true apposition, or else
term is used synonymously with ‘modification’?)

- 15 -