SIL Electronic Working Papers 1997-002, June 1997
Copyright © 1997 Wolf Seiler and Summer Institute of Linguistics, Inc.
All rights reserved.

This paper first appeared in the Proceedings of the Eighth Inuit Studies Conference, Inuit Studies, Inuit and Circumpolar Study Group, Université Laval, Quebec, Canada, 1992.

Valence and Affix Ordering in Inupiatun

Wolf Seiler


0. Introduction
1. The notion of valence
2. Valence changers
3. Classes of verbal postbases
4. Conclusion
Key to abbreviations


Derivation by affixation is central to the description of Eskimo grammar. With hundreds of possible derivational affixes, defining constraints on affix ordering poses a challenge.

The hypothesis of this paper is that the notion of valence is central to understanding the order of affixes in Inupiatun verbs. I present evidence that there exist three distinct sets of affixes for verbal derivations: one set expands a verb base by adding components of meaning, another changes the inherent valence of a (potentially expanded) base and creates a new "compound base," and the third adds sentential level modification to the fully derived verb base. These classes relate to affix order in that verb expanders can precede valence changers (and thus fall within their semantic scope), while sentential modifiers must always follow them.

O. Introduction

Inupiatun is spoken as a first language by approximately 4,000 people in eleven villages in the region around the Kotzebue Sound of Alaska. Inupiatun is part of the larger Northern Eskimo language, which in Northern Canada is known as Inuktitun and in Greenland as Kalaallisun. All three designations name a continuum of interconnecting dialects generally referred to as Inuit. Inuit, along with the Yup'ik (Southern Eskimo) and Aleut languages, belongs to the Eskimo-Aleut family (Krauss 1973).

Descriptions of the Inupiatun language (as spoken in this area) are generally not extensive and have begun to appear only recently. They are, in chronological order, a "teach yourself" grammar (Webster 1968), three dictionaries (Webster and Zibell 1970, MacLean 1980, NBMDC 1979), an article on the Modalis case (Seiler 1978), a morphological description (MacLean 1979, MacLean 1986), and a description of the phonology of Inupiatun (Kaplan 1981). The data for this paper originate from my own language study in the area.

Inupiatun is a highly agglutinative language, characterized by a rich derivational morphology. All verbal and nominal forms in the language are built on an initial base (from an open class of the lexicon) and end with an obligatory inflectional suffix (from a closed class). Between the base and the inflection may occur a string of derivational affixes, which have traditionally been called "postbases" in studies of Alaskan Eskimo. The postbases are a nearly closed class, and there are several hundred of them. Though in theory the strings could be rather long, and forms containing five or more postbases have been attested, those with four or less postbases are more natural.

This paper concerns itself with the internal order of the "nearly closed" class of postbases. The basic claim is that the notion of valence is essential to describing ordering constraints among the postbases.

1. The notion of valence

The term "valence" was introduced into linguistics by the French structuralist L. Tesniere in 1959. Like valence in chemistry, from which the metaphor is borrowed, it has to do with what a partial construction would need to add to become complete. Comrie (1981:51) defines valence as "the number and kind of noun phrase arguments that a particular predicate (usually, a verb) can take."

Valence has three aspects--syntactic, semantic, and pragmatic (Ruzicka 1978). The syntactic aspect in Inupiatun is seen, in particular, in verbal and nominal inflection. Every verb in Inupiatun is inflected as either intransitive or transitive. On nominals, case inflections serve to crossindex the noun phrases in a sentence to the arguments of the verb. For example, with a verb root like aatchuq- 'give,' the ergative case is used to mark the agent (giver), the absolutive case to mark the patient (object given), and the terminalis case to mark the benefactee (receiver). The semantic aspect of valence has to do with semantic case roles (Cook 1989). Though two verbal expressions may have the same syntactic form in terms of inflections, they may differ in valence in that the semantic roles of the nominal, as agent, patient, benefactive, instrument, etc., may differ. The pragmatic aspect of valence has to do with the focus or topicality of the noun phrase arguments, such as when the same configuration of semantic case roles can be expressed by means of different surface inflections to signal differences in focus, as in a passive construction.

2. Valence changers

Of the many derivational affixes, there are some that change the valence of the base to which they are added. In this analysis, a denominalizer changes an avalent noun base into a monovalent (intransitive) verb base or, in the case of -(g)i- 'EQ.p' into a bivalent (transitive) one. Conversely, a nominalizer changes a monovalent or bivalent base to an avalent noun. Though denominalizers and nominalizers, as valence changers, seem to follow the pattern argued for in this paper, we limit our focus in what follows to postbases which are added to verb bases and produce a modified verb base.

The vast majority of the verb-to-verb postbases only add to the meaning of the base without changing the valence. For example, sentence (1) contains five such postbases:

(1) Putu-m savik paqi-tqi-qaagh-u-lgit-chaghi-gaa
    Putu-ERG knife find-REP-first-DES-again-HAB/t-3s.3s.IND
    'Putu again, as always, wants to be the first to find the knife.'

However, there are some verb-to-verb postbases that change the valence of the verb base. These include:

     -tit/pkaq- accusative (CS)'
     -(t)qu- 'obligative (OBL)'
     -kau- 'passive, action realized (PSV.r)'
     -naq- 'passive, action unrealized (PSV.u)'
     -/(u)t(i)- 'dative shift (DS)'

Examples (2)-(4) below show changes in syntactic and semantic valence. For instance in (2) and (3) a causative postbase changes the verb base by adding a participant who is the agent of causing. In (4) the postbase adds a participant who is the agent of obliging.

(2) a. Putu iglaq-tuq
       Putu.ABS laugh-3s.IND
       'Putu is laughing.'
    b. Miiyu-um Putu iglaq-tit-kaa
       Miiyu-ERG Putu.ABS laugh-CS-3s.3s.IND
       'Miiyuk caused Putu to laugh.'
(3) a. Putu iglau-ruq
       Putu.ABS travel-3s.IND
       'Putu is traveling.'
    b. Miiyu-um Putu iglau-pkagh-aa
       Miiyu-ERG Putu.ABS travel-CS-3s.3s.IND
       'Miiyuk caused Putu to travel.'
(4) a. Putu-m atiklhuk killaiyagh-aa
       Putu-ERG cover.ABS sew-3s.3s.IND
       'Putu sewed a parka cover.'
    b. Miiyu-um Putu killaiya-qu-gaa atiklhung-mik
       Miiyu-ERG Putu.ABS sew-OBL-3s.3s.IND cover-MOD
       'Miiyuk wants Putu to sew a parka cover.'

The next two examples, including "dative shift" and passivization, demonstrate a change in the pragmatic aspect of valences:

(5) a. Miiyu-um atiklhuk killaiyagh-aa Putu-mun
       Miiyu-ERG cover.ABS sew-3s.3s.IND Putu-TERM
       'Miiyuk sowed a parka cover for Putu.'
    b. Miiyu-um Putu killaiya-uti-gaa atiklhung-mik
       Miiyu-ERG Putu.ABS sew-DS-3s.3s.IND cover-MOD
       'Miiyuk sewed Putu a parka cover.'
(6) a. Putu-m atiklhuk killaiyagh-aa
       Putu-ERG cover.ABS sew-3s.3s.IND
       'Putu sewed a parka cover .'
    b. Atiklhuk killaiya-kau-ruq (Putu-miny)
       Atiklhuk.ABS sew-PSV.r-3s.IND (Putu-ABL)
       'A parka cover has been sewn (by Putu).'
    c. Atiklhuk killaiyagh-naq-tuq
       cover.ABS sew-PSV.u-3s.IND
       'A parka cover could be sewn.'

3. Classes of verbal postbases

At the simplest level of description, the Inupiatun verb can be described as consisting of a base, optionally followed by any number of derivational postbases, followed by an obligatory inflectional suffix:

               V --> Vbase (Postbase)* Inflection

This is, in fact, as far as many descriptions of Eskimo morphology have gone. To my knowledge, there have been no effective attempts to explain ordering constraints among the many hundreds of these postbases until the work of Fortescue (1980). He made a significant contribution in observing that postbases which function as sentential modifiers always come at the end; postbases which modify or extend a base always precede the sentential modifiers. My addition to his work is to separate base-expanding postbases into two groups, namely, those which change the valence of the base versus those which add components of meaning without changing valence.

My central hypothesis is that there are three major classes of postbases. First, there are the valence changers of which I showed examples above. Second, there are the postbases which can never precede a valence changer. Third, there is the class which can sometimes precede and sometimes follow valence changers. (There appear to be no postbases that can never follow a valence changer.) The distinction between the second and third class is not only a formal one; it also correlates with the functional distinction between sentential modification and verb expansion.

Verb expanders include Modality modifiers which relate to manner or degree, and Auxiliary modifiers which relate to effort, desire, or potential on the part of the agent. (The latter category form a compound predicate but do not change valence since the agents of the base predicate and the added predicate are coreferential.) Sentential modifiers include Aspect modifiers ("Tense" for Fortescue) which relate to frequency and time of actions, and Evaluative and Epistemic modifiers which relate to evaluation and factual knowledge of truth value, as well as Conjunctional modifiers and Negatives.

The fact that verb expanders sometimes come before and sometimes come after the valence changer reflects a difference in scoping and thus in meaning.

The following examples of verb stem expanding modifiers confirm this claim:

(7) a. ilyisima-ruq
       'he knows'
    b. ilyisima-qqaaq-tuq
       'he knows first' (before others)
    c. ilyisima-qqaa-tla-ruq
       'he can know first'
    d. ilyisima-qqaa-tla-niaq-tuq
       'he is going to be able to know first'
    e. ilyisima-qqaa-tla-nia-nghit-chuq
       'he is not going to be able to know first'

The examples in (7) show that it is possible to add to a base a number of expanding affixes without changing the valence of the base. In (7) we added one postbase, respectively, from the Mod, Aux, Asp, and Neg sets. This is the prescribed order of affixation under normal circumstances.

The set of valence changing affixes operates on a different basis than the expanding affixes. Rather than fitting into fixed position slots, they appear to vary with respect to affix order. This is because they cause embedding which shifts the preceding material to a lower level of the derivation. Thus it is possible to treat an expanded verb base with the addition of a Valence Changer as a new "compound base." As a new base it can accept more expanding affixes.

Examples (8) and (9) show this process. First, (8) illustrates a basic verb with two expanders:

(8) a. paqit-kaa
       'he found it'
    b. paqit-qaagh-aa
       'he found it first'
    c. paqit-qaa-tla-gaa
       'he can find it first'

(9) shows the four possible ways of incorporating a valence changing causative postbase with the same two expanders:

(9) a. paqit-qaa-tla-pkagh-aa
       'X causes Y to be able to find it first'
    b. paqit-tla-pka-qqaagh-aa
       'X first causes Y to be able to find it'
    c. paqit-qaaq-ti-tla-gaa
       'X can cause Y to find it first'
    d. paqit-tit-qaa-tla-gaa
       'X can first cause Y to find it'

In this example there are two underlying predicates ("find" and "cause"). Each of the four possible orderings means something different, depending on which of the predicates the expanding affixes attach to.

The data therefore demonstrate that valence changers operate on the verb base (or the expanded verb base) and, essentially, create a new base, which, in turn. can accept more base-expanding affixes. In practical experience, however, there is a limit to how much information a speaker can convey within a single verb structure. I have found that four, or perhaps five, expanding affixes are possible; more than that creates "information overload."

4. Conclusion

In this paper, I have shown that there exist three distinctly different sets of postbases for verbal derivations: one set expands a verb base by adding components of meaning, another changes the inherent valence of a (potentially expanded) base and creates a new "compound base," and the third adds sentential level modification to the fully derived verb base. The significance of these classes for explaining affix ordering is that only a verb base expander can precede a valence changer, and when it does it falls within its semantic scope.

Previous interpretations of derivational affixes of Inuit have grouped valence changers with other base expanders (Fortescue 1980). It is my claim that valence changers must be distinguished as a separate class, since they are the pivotal class forunderstanding the internal structure of Inuit verbs.

To summarize, I propose the following model for interpreting the verb structure of Inuit (terms in parentheses indicate optionality; an asterisk represents any number of repetitions):

  V              --> Vexpanded_base (Sentential mod.)* Inflection

  Vexpanded_base --> base (Expanders)*

  Vbase          --> Vroot
                     Vexpanded_base Vvalence_changer

Key to abbreviations

3s         third singular
ABS        absolutive
ABL        ablative
CS         causation (causative)
DS         dative shift
ERG        ergative
IND        indicative
INT        interrogative
MOD        modalis (case)
OBL        obligative
PSV.r      passive, action realized
PSV.u      passive, action unrealized
TERM       terminalis (case)


Cook, Walter A., S.J. 1989. Case Grammar Theory. Washington, DC: Georgetown University Press.

Comrie, Bernard. 1981. Language Universals and Linguistic Typology. Chicago: University of Chicago Press.

Fortescue, M.D. 1980. Affix ordering in West Greenlandic Derivational Processes. IJAL 46:259-78.

Kaplan, Lawrence D. 1981. Phonological Issues in North Alaskan Inupiaq. In Alaska Native Language Center Research Papers, Vol. 6. Fairbanks: University of Alaska.

Krauss, Michael E. 1973. Eskimo-Aleut. In Current Trends in Linguistics, Vol. 10, general ed. Thomas R. Sebeok. The Hague: Mouton.

MacLean, Edna A. 1979. Beginning North Slope Inyupiaq Grammar. Alaska Native Language Center, Fairbanks: University of Alaska.

MacLean, Edna A. 1980. Abridged Inyupiaq and English Dictionary (Barrow Dialect). Alaska Native Language Center, Fairbanks: University of Alaska.

MacLean, Edna A. 1986. North Slope Inyupiaq Grammar, Second Year. Alaska Native Language Center. Fairbanks: University of Alaska.

NBMDC. 1979. Kobuk Inyupiat Junior Dictionary. Anchorage, AK: National Bilingual Materials Development Center.

Ruzicka, R. 1978. Three aspects of valence. In Valence, Semantic Case and Grammatical Relations. Vol. 1 of Studies in Language Companion Series. Amsterdam: John Benjamins.

Seiler, Wolf. 1978. The Modalis Case in Inyupiat. In Workpapers of the Summer Institute of Linguistics, University of North Dakota, Vol. 22. Grand Forks, ND.

Tesniere, L. 1959. Elements de syntax structural. Paris: Librairie C. Klincksieck.

Webster, Donald. 1968. Let's Learn Eskimo. Fairbanks, AK: Summer Institute of Linguistics.

Webster, Donald and Wilfried Zibell. 1970. Inyupiat Eskimo Dictionary. Fairbanks, AK: Summer Institute of Linguistics.

Date created: 9-Mar-1996
Last modified: 3-Jun-1997

[SILEWP 1997 Contents | SILEWP Home | SIL Home]