a syntactic approach to noun incorporation in cree

12
A Syntactic Approach to Noun Incorporation in Cree DEAN MELLOW McGill University 1.0 Introduction In order to suggest an account of Algonquian medial morphemes, this pa- per outlines a syntactic analysis of complex verbs in Plains Cree, as it is spoken in Northern Alberta. 1 The difficulties in accounting for the facts of polysynthetic word forms are considerable. This paper extends a research program that was outlined in Wolfart (1971:511): Since the verb stems of Algonquian languages usually consist of more tha one morpheme, two questions have long been of interest to Algonquianists: (1) How should these morphemes be classified, and (2) What are the rela- tions among the resulting classes . . . While the problem has been under consideration since at least the times of DuPonceau, it is not much closer to a satisfactory solution today than it was in Kroeber's day. Like Wolfart (1971), this paper will be concerned specifically with one sitional class of morphemes, the medial. Since medials often alternate with nouns that occur outside of a verb stem, they have been said to be the result of noun incorporation. The alternation is illustrated below ( A G R = inflectional affix which indicates agreement for the grammatical categories of person, number, order, mode, tense, etc.): (1) nooc -i -h -iiw wacaskwa hunt 0 TA AGR muskrat 'He hunts the muskrat.' ^his research was funded by FCAR grant 88 EQ 3630, by a Sir James Lougheed Award of Distinction from the Alberta Heritage Scholarship Fund and by a Max Bell Fellowship in Canadian and Northern Studies; I most gratefully ac- knowledge this support. In addition, I would especially like to thank Mark Baker and Yvonne Carifelle for, respectively, theoretical and empirical assistance All unacknowledged citations of Cree data are from Yvonne. 250

Upload: others

Post on 22-Apr-2022

2 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: A Syntactic Approach to Noun Incorporation in Cree

A Syntactic Approach to

Noun Incorporation in Cree

DEAN MELLOW

McGill University

1.0 Introduction

In order to suggest an account of Algonquian medial morphemes, this pa­

per outlines a syntactic analysis of complex verbs in Plains Cree, as it is

spoken in Northern Alberta.1 The difficulties in accounting for the facts of

polysynthetic word forms are considerable. This paper extends a research

program that was outlined in Wolfart (1971:511):

Since the verb stems of Algonquian languages usually consist of more than one morpheme, two questions have long been of interest to Algonquianists: (1) How should these morphemes be classified, and (2) What are the rela­tions among the resulting classes . . . While the problem has been under consideration since at least the times of DuPonceau, it is not much closer to a satisfactory solution today than it was in Kroeber's day.

Like Wolfart (1971), this paper will be concerned specifically with one po­

sitional class of morphemes, the medial. Since medials often alternate with

nouns that occur outside of a verb stem, they have been said to be the

result of noun incorporation. The alternation is illustrated below ( A G R =

inflectional affix which indicates agreement for the grammatical categories

of person, number, order, mode, tense, etc.):

(1) nooc -i -h -iiw wacaskwa hunt 0 TA A G R muskrat

'He hunts the muskrat.'

^his research was funded by FCAR grant 88 EQ 3630, by a Sir James Lougheed Award of Distinction from the Alberta Heritage Scholarship Fund and by a Max Bell Fellowship in Canadian and Northern Studies; I most gratefully ac­knowledge this support. In addition, I would especially like to thank Mark Baker and Yvonne Carifelle for, respectively, theoretical and empirical assistance All unacknowledged citations of Cree data are from Yvonne.

250

Page 2: A Syntactic Approach to Noun Incorporation in Cree

N O U N INCORPORATION IN C R E E 251

(2) noot -acaskw -ii -w hunt muskrat AI A G R

'He hunts the muskrat.'

Thus, it might be said that the noun wacaskwa 'muskrat' of (1) is incor­porated into the verb stem of (2) as a medial. This paper will adopt a

specific theory of grammar, Government-Binding (Chomsky 1981; Baker

1988), and will build upon Wolfart (1971) by suggesting that a syntactic movement analysis may explain certain facts of word formation, including the nature of the medials. Thus, the research questions have been nar­

rowed to: (i) Are all medials the result of syntactic incorporation? and (ii)

Which nouns may be incorporated? This account answers these questions

by suggesting that only deverbal medials are incorporated and that only

structural objects may be incorporated.

2.0 Theoretical Assumptions

This analysis makes several assumptions about the nature of grammatical

systems. First, it assumes that different aspects of word formation are the

result of different components of the grammar. Derivation and compound­

ing are said to result from principles of morphology. Inflectional affixes are

the result of interpretational (spell-out) rules which occur at PF, i.e., after syntactic movement. Complex predicates, which include noun incorpora­

tion, are the result of syntactic movement. A second significant assump­

tion is that Cree and all languages have a configurational phrase structure.

While it is beyond the scope of this paper to justify this assumption, the

primary motivation lies with learnability considerations such as those noted

in Williams (1984) or Mellow (1989). In addition, the asymmetries in terms of the types of nouns which may incorporate may also be considered to be

an argument for hierarchical structure in Cree. The details of the syntactic movement constitute the third assump­

tion. These details are specified by the Government-Binding framework as described in Baker (1988). Briefly, the G B theory that is assumed here

suggests that the variable position of certain elements within a sentence or complex predicate can be characterized as the result of movement from a

base-generated or underlying position to another position. In addition to

the movement of phrases (XPs), this analysis also assumes that heads (X°s,

categories such as nouns and verbs) can move and are then adjoined to other heads. The base-generated positions are defined by certain components of

the grammar such as X-bar theory, theta-theory and lexical properties. Constraints on movement are proposed to account for the restricted range of grammatical sentences. This account of noun incorporation will make

crucial reference to the following constraint:

Page 3: A Syntactic Approach to Noun Incorporation in Cree

252 D E A N M E L L O W

(3) Head Movement Constraint: /Travis An X° may only move into the Y° which properly governs it. lira 1984:131)

This constraint restricts movement such that a head may only move into the next head above it in the phrase structure tree. In sum, the analysis in this paper assumes that the best characterization of a sentence or a complex

predicate is a derivation which can be represented graphically as a hierar­chical structure whose terminal elements may move, with ungrammatical

sentences eliminated by constraints. Given this theoretical framework, we will now investigate its value for answering the research questions outlined

above. W e will address the question of internal relations first, as aspects of

it will bear upon the question of classification.

3.0 The Relation Between Incorporated Medials and the Verb

An interesting fact of noun incorporation is that not all of the nouns in a

sentence may be incorporated into a verb stem; only a limited range of re­lations exist within the verb stem. A number of linguists have investigated

the nature of the relations that exist within a verb stem. Wolfart (1971) identified two types of internal relations: action-goal, and action-local com­

plement. He does not list any examples of agents or more oblique semantic roles such as the benefactive. A syntactic analysis makes a different claim

about which nouns may incorporate. The distinction is based on structural

rather than semantic criteria. In addition, the syntactic analysis goes be­yond description and provides a principled explanation of the absence of

agents and benefactives. W e noted above that the head movement involved

in noun incorporation is restricted by the Head Movement Constraint. The basic implication is that heads cannot move down to lower heads nor can

they move up such that they skip over the first head above them. Thus, a

subject noun cannot adjoin to the verb below it, and adjuncts, such as the

benefactive object of a preposition, cannot be adjoined to a verb because a preposition is the first head above them. As a result, this theory predicts that only structural objects can incorporate. This structural distinction is reflected in the semantics because objects are canonically patients, sub­jects are canonically agents, and benefactives are the object of prepositions. The presence of the action-goal relation noted by Wolfart and the absence of agentive and benefactive relations is predicted by the syntactic analy­sis. There are two apparent exceptions to the predictions of the movement anaylsis: subjects of intransitives and locatives can be incorporated.

3.1 Subjects of intransitives

The incorporated subject of an intransitive is illustrated below:

Page 4: A Syntactic Approach to Noun Incorporation in Cree

N O U N I N C O R P O R A T I O N IN C R E E 253

(4) siipjhk -was _kwaa _w

is blue sky/clouds II A G R

'The sky is blue.'

Syntactic theory accounts for this type of incorporation because it assumes that many intransitive verbs are unaccusative. The claim is that the sole ar­gument of this type of intransitive verbs is base-generated as an object and is then moved into subject position. This also explains why unaccusative subjects are patients and not agents. Since these verbs have similar mean­ings in different languages, we can illustrate this with the English verb 'break':

(5) a. Bill broke the vase. b. [e] broke the vase. c. The vase broke.

Example (5a) is the transitive form of the verb where the object 'the vase' plays a patient role. Example (5b) is the base-generated D-structure of

the intransitive in which 'the vase' occupies the object position. Example

(5c) is the surface form of the unaccusative intransitive. If this analysis

is correct, then it explains why the subjects of intransitives incorporate in Cree.

This syntactic theory predicts that the other class of intransitives,

unergatives, whose subjects begin in subject position, should not be able

to incorporate. The Cree data supports this prediction:

(6) iskwiiw pimipah -taa -w woman runs AI A G R

'The woman runs.'

(7) *pimipah -iskwiiw -taa -w run woman AI A G R

'The woman runs.'

An unergative verb such as 'run' does not allow its subject, an agent, to incorporate. This pattern seems to hold for other unergative verbs in Cree such as 'cry' or 'gallop'. Thus, not only can the syntactic approach account

for this apparent exception, but it also provides an insightful explanation of the asymmetry between different types of intransitive verbs.

3.2 Locatives

A second problem arises because medials can indicate a locative relation and the semantic role of location is often associated with the structural position "object of a preposition". A n example is given below:

Page 5: A Syntactic Approach to Noun Incorporation in Cree

254 DEAN MELLOW

(8) niim -iskote -en -ew hold fire TA AGR

'He holds him over the fire by hand.' (Wolfart 1971:517)

However, while the gloss indicates 'over the fire', an equally accepted trans­lation might be 'somewhere around the fire'. It is not clear that a preposi­

tion is involved. In addition, some direct objects do have a locative meaning

as we see in the following English examples:

(9) George Washington crossed the Delaware.

(10) Robert Frost reached a fork in the road.

A third explanation of the locative relation refers to the argument which will follow in section 4.0, i.e., not all medials are incorporated. This is important because the great majority of the locatives are body-part medials

and some of the body-part medials are not considered to be incorporated. In

sum, the structural correlate of the semantic notion of location is not clearly object of the preposition, nor is it clear that all or even most of medials which indicate location are the result of noun incorporation. The internal relations of the non-incorporated medials seem to belong in the domain of morphological, not syntactic, explanation. Thus, locative relations are not necessarily problematic for this approach. Overall, the syntactic analysis

provides an insightful explanation of the internal relatious of a verb stem by defining which nouns may be incorporated.

4.0 The Classification of Medial Morphemes: Are all medials incorporated?

Wolfart (1973:63, 66) has divided Plains Cree medials into two types: 1) derived or deverbal, which are paralleled by independent stems from which

they are said to be derived; and 2) simple, which show no internal structure, consist of only one morpheme, and are not paralleled by independent stems. Simple medials may be further divided into two classes: "body-part" and

"classificatory". W e can further divide the body-part medials into those which may occur in dependent nouns (i.e., nouns which must be possessed) and those which do not. The classificatory medials are only attested as medials; they do not occur as noun stems. Given this information, we can tentatively classify the medials into four classes: (a) deverbal, can occur as an independent noun stem; (b) simple, body-part, can occur as a dependent noun stem (deverbal); (c) simple, body-part, cannot occur as a noun stem-and (d) simple, classificatory, cannot occur as a noun stem. I suggested above, based on their locative meaning, that some body-part medials may be classified as non-incorporated. The complete classification of medials suggested in this paper is as follows: (a) and (b) are the result of syntactic noun incorporation; (c) and (d) are not. Thus, we will refer to media]0

Page 6: A Syntactic Approach to Noun Incorporation in Cree

N O U N I N C O R P O R A T I O N IN C R E E 255

(a, b) as "incorporated" and (c, d) as "non-incorporated". The criteria for justifying this classification will be the interaction of medials with four facts

of Cree: nouns external to the verb stem, the presence of external specifiers

and modifiers, secondary derivation, and types of co-occurring finals.

4.1 Nouns external to the verb stem

Perhaps the earliest and best known criterion for determining if a given medial is the result of incorporation is that suggested by Sapir (1911:251): "verbal affixes that refer to nouns . . . are not instances of noun incor­poration if they are etymologically unrelated to the independent nouns or

noun stems with which they seem logically connected." As I interpret this criterion, an incorporated medial can occur outside of a verb stem, as part

of either an independent or dependent noun; it is not relevant whether the noun must be possessed. This criterion is quite compatible with the

G B theory adopted. However, external occurrence is not required by G B

theory, since some nominals could be marked as being obligatory for incor­poration, just as some morphemes in many languages are marked as being

bound. Four examples of medials are listed below, corresponding to categories

(a)-(d) above:

(11a) noot -acaskw -ii -w hunt muskrat AI A G R

'He hunts the muskrat.'

(lib) saki -stikwaan -een -eew seize head T A A G R

'He seizes his head by hand.' (Wolfart 1973:67)

(lie) saki -nisk -iin -iiw seize arm/hand T A A G R

'He seizes him by the arm.'

(lid) pakam -aaskw -ii -w hit wood AI A G R

'He hits wood.'

In these examples, (a) and (b) may be considered incorporated because they are paralleled by the forms wacaskwa 'muskrat' and nistikwaan 'my head'.

However, in (c) and (d), nisk and aaskw never occur externally as noun

roots. Instead, the external nouns are mispitoon 'an arm' and mihti 'wood'. Thus, this first criterion supports the classification of deverbal body-part

medials with the deverbal medials; these medials are incorporated and the

others are not.

Page 7: A Syntactic Approach to Noun Incorporation in Cree

256 D E A N M E L L O W

4.2 The presence of external modifiers

The second type of fact which supports the classification is noted by Baker (1988:93-96): "Noun Incorporation can create discontinuous dependencies ... the incorporated noun root can often be modified or specified by a nonadjacent word or phrase that remains morphologically outside the verb complex." This so-called "stranding" of determiners and numerals is illus­

trated in the following Cree sentence:

(12) kiit -askisin -iin -iiw oohi niiso remove shoes T A A G R these two

'He removes these two shoes.'

In (12), the medial 'shoes' is interpreted with the demonstrative and the

numeral, i.e., 'these two shoes', even though they do not form the unit oohi

niiso-maskisma. The presence of these external modifiers can be explained by a syntactic movement analysis, since the noun 'shoes' actually heads the

noun phrase at D-structure and a co-indexed empty noun phrase, a trace,

is said to remain in order to obey the Projection Principle and represent

the continuing structural relationship between the moved noun and the

transitive verb which originally required the presence of the noun in that

position. Interestingly, while the stranded elements are permitted with the dever­

bal medials, they do not appear to be permitted with classes (c) and (d),

the non-derived body-part medials and the classificatory medials. These facts are illustrated below:

(13) *saki -nisk -iin -iiw oohi niiso seize arm/hand TA A G R these two

'He seizes these two arms.'

(14) *pakam -askw -ii -w oohi hit wood AI A G R that

'He hits that wood.'

In (13) and (14), the medials cannot be modified by stranded elements. It is important to note that (13) can be a grammatical sentence when it has the interpretation 'he seizes these two by arm'. Thus, while the modifying elements may appear, they may not modify this medial. In addition the transitivity of the final may play a role in (14); it is ungrammatical with the transitive final -nn 'by hand': *pakamaskwimitw. Since (12) and (13) both

have a transitive final, and (14) does not, it may be the intransitive final with the stranded element which makes (14) ungrammatical However as we will discuss in section 4.4, the transitive value of the finals is n t clear. Overall, these facts suggest that classifying medials as incorDorat^ or non-incorporated is valid.

Page 8: A Syntactic Approach to Noun Incorporation in Cree

N O U N I N C O R P O R A T I O N IN C R E E 257

4.3 Secondary derivation

Cree word formation is very productive and it is possible to have more than one root, medial, and/or final. W h e n secondary derivations yield more

than one medial, G B theory again provides a criteria for distinguishing between classes of medials. The syntactic movement analysis suggests that

only one noun should be incorporated. Baker (1988:121) notes that double

incorporation would create ambiguity: "when two nouns are incorporated, the information as to which one is associated with which thematic role

begins to be lost." Such an analysis presents several other problems as

well. First, the Head Movement Constraint would require that both NPs

were properly governed by the verb. This would require two NPs dominated by the same VP. This is a controversial claim even for the double object construction in English, e.g., "I [gave [Leslie] [a Valentine]]". Second, we

have a number of technical dificulties such as: which noun moves first, what

structural unit does the second attach to, etc. The third problem is that

finals may appear between the medials. While this will be discussed further

in the next section, it appears that finals are sensitive to S-structure, i.e.,

they vary according to the lexical content of the complements of the verb after movement has occurred. For example, when a transitive verb has its

object incorporated, it often takes an intransitive final. If this is true, finals

cannot be assigned until at least S-structure, and perhaps not until P F as I a m proposing for inflectional affixes.2 It the outermost medial moves in

the syntax and the internal final is assigned at PF, then it is not clear how finals would be spelled out appropriately if at all. In sum, the syntactic

approach strongly implies that only one medial can be incorporated per verb and that this incorporated medial should be the outermost medial.

The innermost medial(s) would be part of the stem to which a noun was

adjoined. Once again, when we examine the Cree data, it becomes desirable to

classify the medials into the two groups suggested above. Several examples

of verbs with two medials are listed below:

(15) kisii -ipeek -in -ayoowinis -ee -w agitate liquid T A clothes AI A G R

'He washes his own clothes by hand.' (Wolfart 1973:65)

(16) saki -nisk -ii -piton -iin -iw siezes arm/hand 0 arm T A A G R

'He (3) seizes him (3' by his (3' arm.'

2However, as pointed out by Kenneth Miner (personal communication), as­signing finals at S-structure is also problematic because many finals have lexical content or are lexically associated with various stems, and therefore should be assigned at D-structure.

Page 9: A Syntactic Approach to Noun Incorporation in Cree

258 DEAN MELLOW

(17) *noot -acaskw -ii -apisk -isi -w hunt muskrat 0 metal AI A G R

'He hunts muskrat with something metal (i.e,. a trap).'

(18) *noot -acaskw -ii -nisk -iin -iw hunt muskrat 0 arm/hand T A A G R

'He hunts muskrat by hand.'

(19) *noot -acaskw -ii -piton -iin -iw hunt muskrat 0 arm T A A G R

'He hunts muskrat by arm (i.e., with a trap on its arm).'

In the grammatical examples, (15) and (16), the inner medials are the non-incorporated classificatory and body-part types. The outer medial in

(15) can appear as an independent noun and the outer medial in (16) corre­sponds to the dependent noun mispiton 'an arm'. Syntactic theory correctly

predicts that these two sentences are grammatical. Alternatively, the un­grammatical examples in (17), (18) and (19) have a deverbal medial in the inner position. The syntactic theory correctly predicts that these should be

ungrammatical. Thus, the forms produced in secondary derivation support

this classification of the medials.

4.4 The nature of the finals

The positional class of morpheme which follows a medial is called a final. Finals reflect the transitivity of the verb, the animacy of the subject or

object, and may also have concrete meaning. Wolfart (1973: 39) notes that

"the syntactic and semantic properties implied by the label "transitivity" are not always shared by the entire class. The morphological basis of the present classification therefore needs to be emphasized." However, while

it is not clear which facts are encoded by the choice of a given final, there are some facts of their occurrence which are suggestive for the classification of medials. The transitive nature of Cree finals is significant because of the following criteria for noun incorporation reviewed in Miner (1982-36)" "Verbs containing incorporated objects become grammatically intransitive (along with whatever follows syntactically from this)." It is not entirely clear how to translate such an observation into G B theory. After all incorporating verb is still transitive in that there is a patient for the t' and there is also a structural object position which is occupied by a category. Thus, as noted above, the finals may not be sensitive t^tf/ existence of a structural object, but to the absence of lexical cont°t -G

that position. m

Page 10: A Syntactic Approach to Noun Incorporation in Cree

N O U N I N C O R P O R A T I O N IN C R E E 259

While it is not clear what the finals refer to, the facts of Cree again suggest an asymmetry: verbs with incorporated medials have only intran­sitive finals, and the other medials m a y take either type of final. Wolfart

(1971:516) notes that in "verbs exhibiting the action-goal relation ... all

verbs of this type are intransitive; further derivatives are, however, freely made." W e see an example of a typical deverbal intransitive in (20) and transitive derivative of it in (21):

(20) keet -askisin -ee -w remove shoes AI A G R

'He takes his (own) shoes off.' (Wolfart 1971:516)

(21) kiit -askisin -iin -iiw remove shoes T A A G R

'He removes the shoes by hand.'

The only transitive exceptions to the action-goal relation which are listed by Wolfart are those with classificatory medials. A n example is illustrated below:

(22) kaasiy -aapisk -ah -am wipe metal TI A G R

'He wipes it as metal.' (Wolfart 1971:517)

Thus, for one of Wolfart's two types of internal relations in complex verbs, action-goal, the nature of the finals supports the proposed classification.

However, in Wolfart's other type of stem internal relation, the action-

local complement relation, the verbs typically take transitive finals. While the great majority of this relation are body part medials, we do not find a

clear distinction because many of these body-part medials are deverbal and can appear in dependent nouns. For examples of these transitive body-part

medials, we repeat (lib) and (lie):

(lib) saki -stikwaan -een -eew seize head T A A G R

'He seizes his head by hand.' (Wolfart 1973:67)

(lie) saki -nisk -iin -iiw seize arm/hand T A A G R

'He seizes him by the arm.'

W e see that (lib), with its deverbal medial, has a transitive final. It is pos­sible that (lib) is a further derivative in the way that (21) is a transitive

derivative of the intransitive verb in (20). While (lie) follows the predic­

tion that non-incorporated medials may be found in transitive verbs, the

Page 11: A Syntactic Approach to Noun Incorporation in Cree

260 DEAN MELLOW

situation is confused because the body-part is not the object of the verb. In general, the correspondence between the two classes of body-part medials and their finals is not clear. Thus, the data suggest that the deverbal class (a) and the classificatory class (d) follow the predictions for Miner's cri­terion. However, the body-part medials are not clearly distinguished into

two classes. Before this argument can be used convincingly, the nature of Algonquian finals must be investigated further to determine what they

are sensitive to and therefore what this implies for the G B theory assumed

here. Overall, these four facts of Cree suggest that determining which nouns

may be incorporated provides an insightful and significant way to classify

the medial morphemes. While the facts indicate that deverbal class (a) is

incorporated and the classificatory class (d) is not incorporated, the dis­

tinction is not as clear for the body-part medials. In general, dividing the semantic unity of the body-part class seems justifed by the facts of inde­

pendence and secondary derivation.

5.0 Conclusion

We have seen that a syntactic analysis of certain Algonquian medials pro­vides insightful answers to two questions about the nature of complex words. In addition, many intriguing questions are raised by this analysis: W h a t

rules of morphology account for the position and characteristics of those

medial affixes which are not accounted for by noun incorporation? Are some of these complex stems lexicalized? What crucial properties deter­mine the form of the finals and what component of the grammar can best account for them? Can the syntactic approach provide insight into other aspects of the order and structure of complex predicates? The analysis

presented here seems to be a valuable step towards accounting for the facts of Algonquian word forms.

REFERENCES

Baker, Mark

1988 Incorporation: A Theory of Grammatical Function Changing. Chicago-University of Chicago Press.

Chomsky, Noam

1981 Lectures on Government and Binding. Dordrecht: Foris.

Lees, Jim

1979 A Mini-Grammar of Cree-Montagnais. Linguistique amermdienne I. Montreal Working Papers in Linguistics 12:109-147.

Page 12: A Syntactic Approach to Noun Incorporation in Cree

N O U N INCORPORATION IN C R E E 261

Mellow, Dean

1987 Cree: A Brief Grammatical Sketch/Field Notes. Ms.

1988 A Modular Approach to Morphology: A Syntactic Account of Noun Incorporation in Cree. Ms.

1989 On Triggers: A Parameter-Setting Approach to the Acquisition of Cree, A Free Word Order Language. McGill Working Papers in Linguistics 5:2. [To appear].

Miner, Kenneth

1981 Algonquian Medials as Incorporated Nominals. Annual Meeting of the American Anthropological Association. Handout.

1982 A Note on Noun Incorporation. Algonquian and Iroquian Linguistics 7:36-37.

1986 Another Verb-Initial Characteristic of Algonquian. Algonquian and Iroquoian Linguistics 11:1.

Sapir, Edward

1911 The Problem of Noun Incorporation in American Languages. Amer­ican Anthropologist 13:250-282.

Starks, Donna

1987 Word Ordering: More Than Ordering Subjects, Objects and Verbs. Pp. 215-231 in Native American Language and Grammatical Typol­ogy. Paul Kroeber and Robert Moore, eds. Bloomington: Indiana University Linguistics Club.

Travis, Lisa

1984 Parameters and Effects of Word Order Variation. Ph.D. thesis, MIT.

Williams, Edwin

1984 Grammatical Relations. Linguistic Inquiry 15:639-673.

Wolfart, H. Christoph

1971 Plains Cree Internal Syntax and the Problem of Noun Incorporation. International Congress of Americanists, Proceedings 38:511-518.

1973 Plains Cree: A Grammatical Study. Transactions of the American Philosophical Society 63, Part 5.

Wolfart, H. Christoph, and Janet F. Carroll

1981 Afeet Cree; A Guide to the Language. Edmonton: University of Alberta Press.