Atkan Aleut “unclitic” pronouns

and definiteness
A multimodular analysis

Anthony C. Woodbury

Atkan Aleut has non-subject pronominals that are attracted to a position just
before the verb but do not fuse with it. This clitic like behavior, termed unclitic,
is modeled using a version of the automodular or autolexical analysis proposed
by Sadock (1991). The unclitic pattern is proposed as the explanation for a set
of apparent counterexamples in the puzzling word-order-and-‘definiteness’
paradigms first presented by Bergsland & Dirks (1981: 31–33) and commented on
by Fortescue (1987), Leer (1988), and Sadock (2009).

1. Atkan Aleut1

Atkan Aleut, indigenous in modern times to Atka, an island about mid-way along
Alaska’s Aleutian chain, is one of two subvarieties of Western Aleut, which, together
with its close relative Eastern Aleut, forms the otherwise remote Aleut branch of the
Eskimo-Aleut language family. Aleut, including especially Atkan, has been extensively
documented by Knut Bergsland and later Moses Dirks, himself an Atkan speaker,
beginning with Bergsland (1959), a philological presentation of old and modern texts,
and continuing through a series of articles, a major book of texts (Bergsland & Dirks
1990), an extensive dictionary (Bergsland & Dirks 1990), and two grammars: one is
disarmingly titled Atkan Aleut School Grammar (Bergsland & Dirks 1981, hereafter
“AASG”) but contains extensive, challenging paradigms and examples; and the other
is a reference grammar (Bergsland 1997, hereafter “AG”) which builds on AASJ and

. I wish to thank audiences at the University of Chicago (May, 2008) and the Australian
National University (February, 2009) for many perspicuous comments on oral versions of
this work. I acknowledge a deep debt of gratitude to Knut Bergsland and Moses Dirks
for their magnificent documentation and scholarship on Aleut; and to Jerry Sadock for
introducing me to the world of Aleut and Yupik-Inuit linguistics and for his guidance and
friendship over the years. Finally, thanks to two anonymous reviewers for their thoughtful and
challenging comments.
 Anthony C. Woodbury

incorporates all Aleut varieties and especially copious textual exemplification. At least
one other linguist, Anna Berge (Berge & Dirks 2008; Berge 2009) has recently joined
the primary documentary effort on Atkan Aleut; and a few more have made secondary
contributions through study of the documentary corpus, including notably Michael
Fortescue (1985), Jerrold Sadock (2009) and his student Jeff Leer (1987, 1988).
The present work is of the latter kind and is based on the documentation.2

2. Pronominal attraction toward the verb

Without describing it as such, Bergsland & Dirks (1981) and Bergsland (1997) pres-
ent several patterns in Atkan Aleut in which certain predicate pronominal elements
appear to be attracted from the ‘natural’ position they would have as corresponding
full NPs to a position immediately before the verb (I follow AG in matters of basic
morphosyntactic analysis):

(1) a. Anĝaĝina-s Piitra-x̂ tuku-x̂ hiti-ku-s.

people-ab.pl Peter-ab.sg leader-ab.sg make-pres-3.pl.s
Subject Object PredNom Verb
‘The people made Peter a leader.’ (AASG: ll)
b. Piitra-x̂ tuku-x̂ txin hiti-ku-x̂.
Peter-ab.sg leader-ab.sg self make-pres1-3.sg.s
Subject PredNom Object Verb
‘Peter made himself a leader.’ (AASG: 18)

. Segments are transcribed in the standard Aleut orthography described in Bergsland 1997,
where all symbols have IPA values except that a doubled vowel symbol indicates phonemic
vowel length, <g>is [>], <ĝ> is [t], <x̂> is [χ], and <ng> is [ŋ]. The following abbreviations are
used:

1,2,3 first, second, third person

ab absolutive case
abl ablative case
ns non-subject
past1 remote past tense, single-agreement allomorph
past2 remote past tense, double-agreement allomorph
pl plural
pres present and non-remote past tense (single and double agreement)
rl relative case
s subject
sg singular
 Atkan Aleut “unclitic” pronouns and definiteness 

(2) a. Piitra-x̂ angagina-s ngi-in unanax̂ a-na-x̂.

Peter-ab.sg people-(rl.)pl to-3.pl cook.ab.sg be-past1-3.sg.s
Subject Dative Pred Nom Verb
‘Peter was a cook for the people.’ (AASG: 34)
b. Piitra-m unana-x̂ ngi-in a-qa-ngis.
Peter-rl.sg cook-ab.sg to-3.pl be-past2-3pl.ns/3.sg.s
Subject PredNom Dative Verb
‘Peter was a cook for them.’ (AASG: 34)

Thus the pronouns txin ‘(him)self ’ in (1)(b) and ngiin ‘to/for him’ in (2)(b) occur
immediately preverbally, while their corresponding nonpronominal forms in (1)(a)
and (2)(a) each precedes a preverbal predicate nominal.
It is this pattern and some related patterns that will be the subject of this paper.
First, however, we take a short excursus into some of the basic workings of Aleut clause
syntax and inflection, since these are a crucial part of the data to be discussed.
Aleut is a rigidly SOV language, except that NPs can be pre- and postposed as
adjuncts, and the ordering of O and other nonsubject constituents is, as we have seen,
somewhat freer. The verb has two agreement constructions (see fuller discussion in
AASG, AG, Leer (1987) and Sadock (2009)) which I will call single-agreement
(as in (1)(a)–(b) and (2)(a)) and double-agreement (as in (2)(b)).
The single-agreement construction has three distinctive morphosyntactic fea-
tures: first, the subject shows up in a form with the absolutive case, marked with -x̂
(singular), -s (plural), or -x (dual) e.g. Piitra-x̂ in (1)(b); second, verbs in certain tenses
and moods show a single-agreement allomorph of the tense or mood marker, e.g. -na-
‘remote past; single-agreement allomorph’ in (2)(a); and third, following the tense
marker is a single (intransitive or transitive) subject agreement marker.
The double agreement construction likewise has three distinctive morphosyn-
tactic features: first, the subject shows up with the relative case (which is also
the case for possessors), e.g. Piitra-m in (2)(b); second, verbs in certain tenses and
moods show a double-agreement allomorph of the tense or mood marker, e.g. -qa-
‘remote past; double-agreement allomorph’ in (2)(b); and third, following the tense
marker is an agreement marker that takes into account person and number values
found in the (intransitive or transitive) subject as well as the number of an unbound
anaphoric nonsubject.3

. Although sensitive in principle to subject and non-subject number values, the actual for-
matives involved are few, so that whose number they mark is a matter of considerable com-
plexity: e.g. -ngis can indicate a third person plural subject with an anaphoric nonsubject of
any number, or a third person subject of any number with a plural anaphoric nonsubject: see
Sadock (2009) for a thorough account.
 Anthony C. Woodbury

What exactly is an unbound anaphoric nonsubject? In (2)(b), ngiin ‘to/for them’ is

an unbound anaphor because it has no binding antecedent; and it is a non-subject
(specifically, a dative case oblique). Indeed, it is always (and only) the presence of an
unbound, anaphoric nonsubject that triggers the double-agreement construction.
Thus, (2)(a) is minimally different from (2)(b) in that ngiin ‘for them’ is bound by
angaginas ‘people’ and hence only single-agreement occurs. In (1)(b), txin ‘(him)self ’
is a pronoun, but as a reflexive pronoun it is bound, in this case by the subject, and
hence the single-agreement construction is used. There are many further twists, turns,
and details and they have been analyzed in various ways, but understanding this basic
pattern will suffice for now, which is summarized in Table 1.

Table 1. Distinctive features of single- and double-agreement constructions4

Feature Single-Agreement Double-agreement4

Unbound anaphoric nonsubject Absent Present

Subject case Absolutive Relative
Remote past tense allomorph -na- -qa-
Agreement takes account of: Pers/# of subject Pers/# of subject and unbound
anaphoric nonsubject

We may use phrase structure notation to specify the morphological rule forming
inflected (indicative) verb words from verb stems as:
(3) V[Word] = V[Stem] [Tense] [[Pers, Num, +Subj]
([Num, −Subj, +UnboundAnaphor])]

(3) says that a verb word consists of a verb stem, a tense marker, and pronominal
marking for person and number of subject, and, if present, the number of an unbound
anaphoric nonsubject.

3. Multimodular “unclitic” analysis

We now return to the pronoun pattern in (1) and (2). This pattern is presented as a
set of word-order differences in AASG and in Bergsland (1997), but no more general

. The double-agreement endings also occur on the verb of certain subordinate clause con-
structions, where the nonsubject marker indexes not an unbound anaphoric nonsubject of
its own clause, but looks ahead to an NP of the next clause; furthermore, its own NP subject
remains in the absolutive. But this is quite distinct from the main clause single- and double-
agreement patterns described here.
 Atkan Aleut “unclitic” pronouns and definiteness 

account is given; nor do I know of any accounts of it in other literature on Aleut.

In this paper, I would like to suggest an analysis of it, and of some related data,
for which I believe Sadock’s (1991) multimodular approach to grammar provides a
perspicuous framework.
On the analysis I have in mind, the attraction of the light pronouns to the verb
is clitic-like; yet, they are not quite clitics in that they do not appear to fuse with the
verb phonologically.5 I will term them unclitics, and will furthermore argue that the
unclitics become incorporated in the inflectional apparatus of inflected verb words.
That is, the (virtual) inflected verb words so created would correspond to the bracketed
stretches in (1)(b′)and (2)(b′):

(1) b.′ Piitra-x̂ tuku-x̂ [txin hiti-ku-x̂.]VerbWord

Peter-ab.sg leader-ab.sg self make-pres-3.sg.s
‘Peter made himself a leader.’ (AASG: 18)

(2) b.′ Piitra-m unanax̂ [ngiin a-qa-ngis.]VerbWord

Peter-rl.sg cook.ab.sg for.3.pl be- past1-3.pl.ns/3.sg.s
‘Peter was a cook for them.’ (AASG: 34)

I now present a technical implementation of this basic idea. In Sadock’s (1991) frame-
work, expressions must meet criteria for well-formedness in a series of autonomous
modules: originally Syntax, Morphology, and Semantics. Built into the theory are
default principles of correspondence among the modules which are presumed univer-
sal; as well as a constrained set of possibilities for mismatching among components.
Lexical entries are the main locus for the expression of interface relationships, since
a lexical item (like any entry in a dictionary) will bear information for each module,
e.g. in syntax its part of speech and subcategorization frame, in semantics its meaning,
in morphology its role in word-structure. Over the years, various different reckonings
of the number and extent of the modules have been offered. Here, I will use a version
having the three above-mentioned modules, along with a fourth module, Morphopho-
nology (Chelliah 1988) dealing with the arrangement of morphological formatives,
and for which I will assume mismatching principles I proposed in Woodbury (1996).

. For example, none of the examples of word-internal or even word-external sandhi offered
in AG: 34–44 treat pronominal-plus-verb sequences; on the other hand in Bergsland’s (1959)
transcriptions from audiorecordings, “word forms that are spoken together without pause,
in one ‘stretch’ are united by a hyphen” (1959: 8); there, one finds pronominal-plus-verb
sequences both joined (tgìdig-ukurtá:guzaqalinas-art[akus] ‘they had begun to see each
other constantly’ and separate (tgìdig [Pause] qâranilakan ‘themselves they put into a rage,
and’ (1959: 58). Clearly the exact prosodic characteristics of these sequences are a matter
of interest.
 Anthony C. Woodbury

In this framework, I suggest the following expression of ‘unclitic’ pronouns:

–– In Morphohonology, the unclitics are independent phonological words.

–– In Morphology, unclitic pronouns are incorporated into the verbal complex as
inflectional pronominal prefixes marking nonsubjects.
–– In Syntax, unclitic pronouns are pronominal versions of non-subject NPs or PP.
–– In Semantics, the unclitics have the meanings contributed by their person,
number, and case features.

To take the example of unclitic ngiin ‘for them’ in (2)(b′), the lexical entry would appear
as follows:
(4) Lexical entry for the unclitic ngiin ‘for them’
Morphophonology: [Word/ngiin/]
(I.e., An independent phonological word)
Morphology: [PRO 3,PL,DAT]
(I.e., Third person plural dative pronominal element)
Syntax: PP[3,PL,DAT]
(I.e., Postpositional phrase, (with feature values))
Semantics: ‘for them, to them, etc.’

To complete the morphological account, a slight revision of the morphological rule (3)
for forming inflected verb words from verb stems is needed so that it accommodates
the unclitic non-subject pronoun:
(3′) V[Word] = ([Pers, Num, −Subj]) V[Stem] [Tense]
[[Pers, Num, +Subj] ([Num, −Subj, +UnboundAnaphor])]

The added portion, shown in italics, achieves this; the effect of this rule is the morpho-
logical bracketing shown in (1)(b′) and (2)(b′).
It now remains to say more about the syntax. In particular, how do we account
for the differences in word order between these unclitic forms, and the corresponding
forms where the unclitics take NP complements, in (1)(a) and (2)(a)?
Suppose we claim that in the syntax, there is a VP which may be expanded as
follows:6

. The notion VP or predicate is relatively uncontroversial in Aleut and is used explicitly,

e.g. by Leer (1987) and Sadock (2009). Arguments for positing a VP include (a) the fact that
the domain of unbound anaphoric nonsubject agreement includes the canonically post-subject
(putative) VP but excludes the subject, which is the domain of subject agreement; (b) the
maintenance of strict SOV word order within clauses, and the clear marking as adjuncts of
preposed objects (AG: 141–3) and postposed subjects (AG: 138–9); and (c) the existence of
several productive subject-creating and subject-suppressing valency-changing verbal affixes
 Atkan Aleut “unclitic” pronouns and definiteness 

(5) VP = (NP[DirectObj]) (PP[{Dat,Abl,…}]) (NP[PredNom]) V

This corresponds to the forms in (1)(a) and (2)(a). For example in (2)(a), ngiin
‘for…’ functions as a postposition that agrees with its complement and is represented
lexically as:
(6) Lexical entry for the postposition ngiin ‘for them’
Morphophonology: [Word/ngiin/]
(I.e., An independent phonological word)
Morphology: [P[Word] P[Stem,DAT] [3,PL]]
(I.e., Dative postposition, inflected for third person plural
agreement)
Syntax: P
(I.e., Postposition)
Semantics: ‘for [them], to [them], etc.’

Note that in contrast to the unclitic in (4), it is morphologically a free word (not an
incorporated pronoun), and syntactically a P (not a pro-form for a whole PP).
Nevertheless, the unclitic forms in (1)(b′) and (2)(b′) pose a problem, since for
them, the unclitic appears between the predicate nominal and the verb. In other words,
there is a mismatch between the syntax so formulated, and our morphological analysis
in (4) and (3′).
Following Sadock (1991: 61), by default the linear order in two dimensions must
match; but in case of a linearity mismatch between morphology and syntax, morphology
wins, as it were; and that is the case here.
In summary, a key discrepancy between morphology and syntax is handled by a
general principle. But a careful look at the lexical representation of the unclitic in (4)
reveals a more subtle morphology-syntax mismatch, for in syntax the unclitic ngiin
‘to/for them’ is an independent word, while in morphology it is a part of the verbal
word. Technically, this mismatch is accounted for by a second principle, the Incor-
poration Principle, formulated in Sadock (1991: 103, 105) and presented in a slightly
(but for our purposes crucially) revised form in Woodbury (1996: 334):
(7) Revised incorporation principle. If a lexeme combines with another
element in the morphology and with a phrase in the syntax, its morpho-syntactic
association will conform to the Strong Constructional Integrity Condition.

(AG: 115–118; 161–177), including a causative, a passive with clearly distinguishable personal
and impersonal (subjectless) variations, and an indefinite-subject construction (AG 178–185).
Even so, the current proposal does not hinge on labeling this domain as a single (VP)
constituent, as long as it is somehow distinguished from the subject domain.
 Anthony C. Woodbury

(8) Strong Constructional Integrity Condition. If the elements of a complex

morphological expression X+Y each have syntactic reality, then one of these
elements combines in the syntax with a phrase of which the other is the head.
In particular, unclitic ngiin ‘to/for them’ combines with aqangis ‘he was, with respect
to them’ in the morphology and with a phrase in the syntax – namely, the PP of which
it is the sole surface constituent; moreover this relationship conforms to (8) in that
ngiin and aqangis each have syntactic reality, and in syntax, the verb aqangis com-
bines with the PP of which ngiin is the head, and, indeed, the sole surface constituent.
In plain language, a morphological host like aqangis can morphologically incorpo-
rate a syntactically independent element like ngiin, to which it bears the syntactic
relationship of head.

4. A definiteness puzzle

Bergsland & Dirks (AASG: 31–33) present a paradigm of examples in which the
relative case – which is supposed to mark the subject whenever a unbound ana-
phoric nonsubject is referenced in the verb – fails to appear when the apparent subjects
of existential (and related) sentences are glossed as indefinite. This set of examples has
received considerable attention (AG: 151; Fortescue 1985: 121–2; Sadock 2009: 110–2)
and has been treated as a curious – but communicatively handy – exception to the
relative case rule:
(9) a. Qalgada-m kug-an a-ku-u.
food-rl.sg on-3.sg be-pres-3.sg.ns/3.sg.s
‘The food is on it.’
b. Qalgada-x̂ kug-an a-ku-u.
food-ab.sg on-3.sg be-pres-3.sg.ns/3.sg.s
‘There is food on it.’ (AASG: 32)
(10) a. Kuusxi-m sitx-an unguchi-ku-u.
cat-rl.sg under-3.sg sit-pres-3.sg.ns/3.sg.s
‘The cat is sitting under it.’
b. Kuusxi-x̂ sitx-an unguchi-ku-u.
cat.ab.sg under-3.sg sit-pres-3.sg.ns/3.sg.s
‘A cat is sitting under it.’ (AASG: 32)
(11) a. Tayagu-m nag-aan hiti-ku-u.
man-rl.sg in-abl.3.sg go-pres-3.sg.ns/3.sg.s
‘The man went out of it.’ (lit., ‘from its inside’)
b. Tayagu-x̂ nag-aan hiti-ku-u.
man.ab.sg in-abl.3.sg go-pres-3.sg.ns/3.sg.s
‘A man went out of it.’ (AASG: 32)
 Atkan Aleut “unclitic” pronouns and definiteness 

Thus in (9)(a) ‘food’ is glossed as definite and marked with relative case -m while
in (9)(b) it is glossed indefinite and marked only with absolutive case -x̂. And the
same state of affairs obtains in (10) and (11). What is going on here? Note that
because the ‘it’ of kugan ‘on it’ in (9)(a) and (b) is an unbound anaphoric non-subject,
it is (duly) referred to with the double-agreement verb ending -u. But according
to Table 1, it should also be accompanied by a relative case subject. In that respect,
the definite subject in (9)(a) follows the expected pattern whereas in (9)(b) the corres­
ponding indefinite element does not. The same is the case in the pairs in (10) and
(11). Thus either the correlation proposed in Table 1 must be qualified (as suggested
in the literature), or else our assumptions about definiteness or subjecthood must be
revisited or clarified.

5. A definite-subject/indefinite-predicate nominal analysis

Sadock (2009) has argued that the relative case may fail in the (b) sentences of (9)–(11)
as a disambiguating mechanism, but that the apparent subjects are still subjects; and
that this indeed characterizes the communicative functionality of the attested pattern.
But why? And why should the relative case mark the definite and the absolutive case
the indefinite, rather than the reverse?
Bergsland himself does not directly resolve this problem, but he gives a hint which
I would like to pursue, in combination with some consequences of the unclitic analysis
already proposed.
About (9)(a)–(b) Bergsland (AG: 151) writes:
This construction has the semantic character of an existential sentence and the
subjecthood of the term in the absolutive case is as problematic as that of the
corresponding term in the English translation.

And he gives a second hint by translating what he calls the “apparent subject” of (9)(b)
with a predicate nominal in an English expletive there construction.7 Bergsland seems
to be suggesting that the definite, relative case subject is more of a subject than the
indefinite, absolutive case entity. And if the indefinite, absolutive case entity were not
a subject, then the correlation of relative case subject and double agreement, expressed
in Table 1, would remain in place.

. English itself is less happy with expletive there in translations of (10)–(11)(b): ?There
is sitting under it a cat and ?There went out of it a man, although they capture the gist of
Bergsland’s strategy.
 Anthony C. Woodbury

In the terms so far developed, Bergsland’s hint might take the following form:
the definite subject is a true subject that is outside of VP and receives the relative case
in the normal way; whereas the indefinite “subject” is not a subject (as suggested by the
case marking), but a predicate nominal inside the VP (or, at least, outside the subject
domain) and is not assigned the relative case. Thus for (9)(a)–(b) we have the syntactic
analyses in (12)(a)–(b):
(12) a. Qalgada-m [VP kug-an a-ku-u.]
food.rl.sg on-3.sg be-pres-3.sg.ns/3.sg.s
‘The food is on it.’
b. [VP Qalgadax̂ kug-an a-ku-u.]
food.ab.sg on-3.sg be-pres-3.sg.ns/3.sg.s
‘There is food on it.’ (AASG: 32)

Subjectless sentences like those posited here may exist in Atkan Aleut in any case:
Bergsland argues for subjectless sentences for weather verbs (Bergsland AG: 129);8 and
he argues that the general verbal passive marker -lga-~-sxa- forms impersonal subjects
of both transitive and intransitive verbs by suppressing the initial or underived subject
argument while not advancing a nonsubject into its place (AG: 167–178).

6. Word order and definiteness

Concerning (9)–(11), there is an interesting corollary pattern when the postposition

has an overt postpositional object:

(13) a. Qalgada-x̂ stuulugi-m kug-an a-ku-x̂.

food-ab.sg table-rel.sg on-3.sg be-pres-3.sg.s
‘The food is on the table.’
b. Stuuluĝi-m kug-an qalgada-x a-ku-x̂.
table-rl.sg on-3.sg food-ab.sg be-pres-3.sg.s
‘There is food on the table.’ (AASG: 32)

. The test Bergsland uses to demonstrate subjectlessness is illustrated by chix̂ta-ku-x̂ ukux̂

ta-qa-ng ‘I saw that it was wet’ where -ng is a two place ‘I-it’ ending in which the “it” refers to
an unbound anaphoric nonsubject, namely some (tangible) thing that was wet, indicating that
the complement clause ‘be wet’ has an actual subject; vs. chix̂ta-ku-x̂ ukux̂ta-na-q ‘I saw that it
was raining,’ where -q is a one-place 1sg subject ending whose lack of a non-subject marker is
taken by Bergsland to follow from the object complement’s lack of any subject. Unfortunately,
I have not yet found comparable complementation evidence in order to test my claim of
subjectlessness for (9)–(11)(b).
 Atkan Aleut “unclitic” pronouns and definiteness 

(14) a. kuusxi-x̂ stuulugi-m stix-an unguchi-ku-x̂.

cat-ab.sg table-rl.sg under-3.sg sit-pres-3.sg.s
‘The cat is sitting under the table.’
b. Stuulugi-m sitx-an kuusxi-x̂ unguchi-ku-x̂.
table-rl.sg under-3.sg cat-ab.sg sit-pres-3.sg.s
‘A cat is sitting under the table.’ (AASG: 32)

(15) a. Tayagu-x̂ ula-m nag-aan hiti-ku-x̂.

man-ab.sg house-rl.sg in-abl.3.sg go-pres-3.sg.s
‘The man went out of the house.’
b. Ula-m nag-aan tayagu-x̂ hiti-ku-x̂.
house-rl.sg in-abl.3.sg man-ab.sg go-pres-3.sg.s
‘A man went out of the house.’ (AASG: 32)

The specification of the postpositional object means that there is no unbound ana-
phoric nonsubject in these examples, and hence – following Table 1 – the verb shows
single agreement and no subject – definite or indefinite – bears the relative case. Inter-
estingly, the putative indefinite subject in the (b) forms appears after the postpositional
phrase – as would fit a predicate nominative but not an (unexceptional) subject; while
the putative definite subject in the (a) forms appears before the postpositional phrase,
as would fit a canonical subject.
On our analysis, these facts would follow from the treatment of the definite NP’s
of the (a) forms as subjects, and the indefinite NP’s of the (b) forms as predicate nomi-
nals, shown below for (13):

(13) a.′ Qalgada-x̂ [VP stuulugi-m kug-an a-ku-x̂.]

food-ab.sg table-rel.sg on-3.sg be-pres-3.sg.s
‘The food is on the table.’
b.′ [VP Stuuluĝi-m kug-an qalgada-x̂ a-ku-x̂.]
table-rl.sg on-3.sg food-ab.sg be-pres-3.sg.s
‘There is food on the table.’ (AASG: 32)

This account is supported to the extent that the definite “subjects” of the (a)-form of
(13)–(15) pattern like subjects and the indefinite “subjects” of the (b) forms pattern
like non-subjects. At least with respect to word order, they carry out the canonical
clause-internal subject-first ordering; and on such an analysis require no new rule for
locating (true) subjects after PP’s.
But why, then, is there no corollary word-order argument for the (b) forms of
(9)–(11)? It is here that the unclitic analysis applies. In (9)–(11) but not (13)–(15),
the locational argument is a light pronoun (like the forms in (1)(b) and (2)(b)
already discussed). On the unclitic analysis, the locational pronoun is attracted to
the verb as an unclitic, forming a morphological verb word with it. In the (a) forms
 Anthony C. Woodbury

this simply reinforces the syntactic order; but in the (b) forms there is a mismatch
and morphology “wins” over syntax, following Sadock (1991: 61) as discussed above,
leading to its appearance to the right of the indefinite predicate nominal, just as
in (2)(b). Schematically:

(16) a. Qalgada-m [VP [Verb Word kug-an a-ku-u]]

food-rl.sg on-3.sg be-pres-3.sg.ns/3.sg.s
‘The food is on it.’
b. [vp Qalgada-x̂ [Verb Word kug-an a-ku-u]]
food-ab.sg on-3.sg be-pres-3.sg.ns/3.sg.s
‘There is food on it.’ (AASG: 32)

Thus the unclitic pattern independently predicts the lack of a clear word order dif-
ference among the (a) and (b) forms of (9)–(11). It also shows that the correct
generalization is not simply that the relative case marks definiteness and the absolu-
tive case indefiniteness, as in (9)–(11) but not (13)–(15); but that definiteness and
indefiniteness, regardless of case marking, are connected in these examples to word
order and subjecthood.

7. A pragmatic-syntax interface rule of definiteness linking?

In Autolexical terms, our generalization can be formulated provisionally as a matching

principle between syntax and pragmatics, as shown in (17):
(17) S: NPSubject NPNonsubject
| |
P: Definite Indefinite
But is this generally true about Aleut, or just true about this particular set of existential
predications? Bergsland (AG: 151) states, “an initial subject is mostly definite but may
also be indefinite (new information)”. As an example of an indefinite subject he gives
uuquchiingikuchax̂ ‘little fox’ in:
(18) (uuquchiigni(m) tanaa hnukuqaang) uuquchiingikucha-x̂
(fox’s its.dens when.I.reached) little.fox-ab.sg
hunaa-m ila-gaan iga-na-x̂
hole-rl.sg inside-abl.3.sg run-past1-3.sg.s
‘(when I reached the fox dens) a little fox ran out of its hole.’

The example may be telling: although ‘little fox’ is tagged as ‘new information’ in
Bergsland’s gloss, it is, in the terms of Prince’s (1981) taxonomy of given and new infor-
mation, inferrable from the presence of fox dens and not brand new information
 Atkan Aleut “unclitic” pronouns and definiteness 

out of the blue. Perhaps likewise the rock in (19) is inferable from the mountain
setting (AG: 154):9
(19) kiiguusix̂ hangakum qugana-x̂ hla-ga-n
mountain when.he.climbed.it rock-ab.sg son-3.sg-rl.sg
kug-an husa-qa-a.
on-3.sg fall-past2-3.sg.ns/3.sg.s
‘when he climbed the mountain a rock fell on his son.’10
It is way beyond the scope of this paper to solve all the intricacies of word order and
information structure in Aleut (see AG: 138–161 for a detailed compendium of textual
examples); and it may be then that a different calibration of information categories is
relevant in Aleut than what Bergsland seems to assume in his glosses and comments,
especially when going from the well-organized and well-studied paradigm discussed
in the literature to more challenging naturalistic texts. Indeed, the claim that subjects
are definite may need weakening, although I will continue to use the terms definite
and indefinite as stand-ins. The point is that by factoring out the confounding effects
of unclitic attraction, it is possible to see more clearly what may be going on.

8. Extending the pragmatics-syntax interface rule

Let us now consider the final portion of Bergsland and Dirks’ definiteness paradigm,
which concerns the relationship of definiteness and word order within VP’s. Bergsland
(AG: 158) notes, “the order of the object and the oblique term seems to depend on the
context or situation. Preceding the oblique term, the object is perhaps definite while
following it it may be indefinite.” Consider the following:
(20) a. Ivaana-x̂ kanfiixta-s yaasika-m nag-an agi-ku-x̂.
John-ab.sg candy-ab.pl box-rl.sg in-3.sg give-pres-3.sg.s
‘John put the candies in a box.’ (AASG: 32)
b. Ivaana-x̂ yaasika-m nag-an kanfiixta-s agi-ku-x̂.
John-ab.sg box-rl.sg in-3.sg candy-ab.pl give-pres-3.sg.s
‘John put candies in the box.’ (AASG: 32)
c. Ivaana-m yaasika-m nag-an agi-ku-ngis.
John-rl.sg box-rl.sg in-3.sg give-pres-3.pl.ns/3.sg.s
‘John put them in a/the box.’ (AASG: 32)

. As one reviewer notes, it would be worth seeing if substituting ‘little bird’ for ‘little fox’ –
brand new information – would favor a shift in NP-PP order.
. The non-subject marker in ‘fall’ refers to the father, possessor of ‘son’; and by our account,
‘rock’ would be in the relative case if it were actually the subject.
 Anthony C. Woodbury

In (20)(a), a definite full-NP direct object kanfiixtas ‘candies’ precedes an indefinite

postpositional phrase yaasikam nagan ‘in a box’; whereas in (20)(b) these two con-
stituents are reversed, and the interpretation of definiteness is the opposite, suggesting
that within the verb phrase, definite precedes indefinite. When the direct object is
pronominal (20)(c) – showing up only as the unbound anaphoric nonsubject “them”
referenced in the double-agreement on the verb – the postpositional phrase, as the
lone constituent left in the VP, can be interpreted as either definite or indefinite.
Consider likewise:
(21) a. Piitra-x̂ sa-x̂ Ivaana-m nga-an agi-ku-x̂.
Peter-ab.sg duck-ab.sg John-rl.sg to-3.sg give-pres-3.sg.s
‘Peter gave the duck to John.’ (AASG: 33)
b. Piitra-x̂ Ivaana-m nga-an sa-x aĝi-ku-x̂.
Peter-ab.sg John-rl.sg to-3.sg duck-ab.sg give-pres-3.sg.s
‘Peter gave John a duck.’ (AASG: 33)
(21)(a) orders a definite direct object before a definite postpostional phrase (with
proper name postpositional object); whereas (21)(b) reverses the order of these con-
stituents, leading to the interpretation of sax̂ ‘duck’, a common noun, as indefinite.
(See also (1)(a) for a comparable case.)
On the basis of this we may refine our statement in (17) as follows:
(22) Definiteness linking (Final version)
S: [S ... NPSubj [VP ... NP ... NP ... ]]
| | |
P: Definite Definite[Degree-i] Definite[Degree-j] (i ≥ j)

(22) says that subjects – at least in constructions like these – are definite; and that
within a VP, the degree of definiteness of an NP will always be greater than or equal to
the degree of definiteness of any NP to its right. (In the cases of existential sentences
like (13)–(15) and predicate nominal sentences like (2), it may be possible to make
even stronger claims about indefiniteness within the VP.)
As in (9)–(11), the ordering specified by (22) is overridden in the case of pro-
nominal, unclitic VP constituents. Thus the pronominal version of each of (21)(a)
and (b) is the same:
(23) Piitra-m sa-x nga-an aĝi-ku-u.
Peter-rl.sg duck-ab.sg to-3.sg give-pres-3.sg.ns/3.sg.s
‘Peter gave him the/a duck.’ (AASG: 33)
In both cases unclitic nga-an ‘to him’ is attracted to the position just before the verb
and after ‘duck’, even when – in the case of an indefinite reading of ‘duck’ per (21)(b) –
it means that an indefinite (‘duck’) precedes a definite (‘to him’). Unclitic attraction
in Morphology thus “wins” over syntax and pragmatics, neutralizing the definite/
indefinite distinction otherwise carried only by word order. The same neutralization
 Atkan Aleut “unclitic” pronouns and definiteness 

holds when an unclitic speech act participant pronominal occurs in a ditransitive

sentence of this type:

(24) Piitra-x̂ sa-x̂ ngu-s agi-ku-x̂.

Peter-ab.sg duck-ab.sg to-1.sg give-pres-3.sg.s
‘Peter gave me a/the duck.’ (AASG: 33)

Does even our revised formulation in (22) scale up beyond Bergsland and Dirks’
paradigm to full-blown discourse and to predicates beyond existentials and verbs of
putting and giving? Examples presented in AG: 149–161 make it clear that multiple
factors beyond definiteness may affect the ordering of direct objects and obliques,
the identification of which is, again, beyond our scope here. Still, the unclitic analysis
allows us to factor out clearly at least one of the influences on word order and may
allow us to see better what is left.

9. Conclusions

Aleut ‘unclitic’ pronominals are light NP or PP pronominals that are free in syntax,
separate words in Morphophonology, and part of verb inflection in Morphology.
Morphology overrides Syntax in determining their arrangement; it also overrides the
pragmatic cline (or clines) that influence order. What we have described as a series
of overrides has up to now seemed a fairly capricious set of word order anomalies to
be catalogued.
The unclitic phenomenon is certainly not exotic and is attested widely in Spanish
and other Romance languages (Perlmutter 1971). But by taking morphological attrac-
tion and morphophonological fusion as separate features, we also get insight into
both synchronic distinctions and implicit historical process according to which a free
pronoun is reanalyzed as part of inflectional morphology. For Aleut unclitics, the pro-
cess is incomplete in the sense that they do not fuse morphophonologically with their
hosts. Nevertheless, some of the same pronouns which function as unclitic objects
underwent an earlier round of attraction to postverbal position where they underwent
phonological fusion to yield simple agreement subjects in verb inflection. Compare
(AASG: 17, AG: 139):

(25) unclitic object pronoun Cognate subject inflection

Txichix achixa-ku-s. Tingin achixa-ku-x̂-txichix.
2.pl teach-pres-3.pl.s us.pl teach-pres(-sg)-2.pl.s
‘They are teaching you.’ ‘You are teaching us.’
Ting achixa-ku-s. Ting achixa-ku-q. (-q < *-x̂-ting)
1.sg teach-pres-3.pl.s 1.sg teach-pres(-sg)-1.sg.s
‘They are teaching me.’ ‘I am teaching me (= learning).’
 Anthony C. Woodbury

In the cognate subject inflections, the pronoun fused with a form of the verb that was
cognate with the present third person singular -x̂ form.
One interesting area for future exploration is to see to what extent Aleut, with its
“new” preverbal pronominal proclitics, takes on properties of languages with richer
agreement. This is a particularly interesting exercise because the foundation – especially
the system of unbound nonsubject agreement – is itself quite typologically unusual.
A second area of exploration is Aleut’s system for encoding definiteness, given
and new information, perhaps focus, and related notions; and its relationship to sub-
ject marking, VP position, and clause-external adjunct positions. Recognizing the
unclitic phenomenon allows one to factor out one strand of perplexity in Atkan word
order, but with respect to what is left, this study barely scratches the surface. There
will probably be much more to say once the argument structures of different verbs
and complex predicates are taken more fully into account. We have an extraordinary
documentary corpus from which to learn more, even as, sadly, this brilliant language
is losing so many of its speakers.

References

Berge, Anna. 2009. Tracking topics: A comparison of topic in Aleut and Greenlandic discourse.
In Variations on Polysynthesis: The Eskaleut Languages [Typological Studies in Language 86],
Marc-Antoine Mahieu & Nicole Tersis (eds), 185–200. Amsterdam: John Benjamins.
Berge, Anna & Moses L. Dirks. 2008. Niiĝuĝis Mataliin Tunux̂tazangis = How the Atkan Aleuts
Talk: A Conversational Grammar. Fairbanks, AK: Alaska Native Language Center, University
of Alaska Fairbanks.
Bergsland, Knut. 1959. Aleut Dialects of Atka and Attu [Transactions of the American Philosophical
Society n.s. 49(3)]. Philadelphia PA: APS.
Bergsland, Knut. 1997. Aleut Grammar. Fairbanks AK: University of Alaska, Alaska Native
Language Center.
Bergsland, Knut & Dirks, Moses L. 1981. Atkan Aleut School Grammar. Anchorage AK: National
Bilingual Materials Development Center.
Bergsland, Knut & Dirks, Moses (eds). 1990. Unangam ungiikangin kayux tunusangin/Unangam
uniikangis ama tunuzangis: Aleut Tales and Narratives Collected 1909–1910 by Waldemar
Jochelson. Fairbanks AK: Alaska Native Language Center.
Chelliah, Shobhana. 1988. Level ordered morphology and phonology in Manipuri. Ms, University
of Texas at Austin.
Fortescue, Michael. 1985. Anaphoric agreement in Aleut. In Predicates and Terms in Functional
Grammar, A. Machtelt Bolkenstein, Caspar de Groot & J. Lachlan Mackenzie (eds), 105–126.
Dordrecht: Foris.
Leer, Jeff. 1987. Aleut Anaphora. MA thesis, University of Chicago.
Leer, Jeff. 1988. The relative case in Aleut. In American Languages and Grammatical Typology,
Paul Kroeber & Robert Moore (eds). Bloomington IN: Indiana University Linguistics Club.
Perlmutter, David M. 1971. Deep and Surface Structure Constraints in Syntax. New York NY:
Holt, Rinehart and Winston.
 Atkan Aleut “unclitic” pronouns and definiteness 

Prince, Ellen. 1981. Toward a taxonomy of given-new information. In Radical Pragmatics,

Peter Cole (ed.), 223–256. New York NY: Academic Press.
Sadock, Jerrold M. 1991. Autolexical Syntax. Chicago IL: University of Chicago Press.
Sadock, Jerrold M. 2009. The efficacy of anaphoricity in Aleut. In Variations on Polysynthesis:
The Eskaleut Languages [Typological Studies in Language 86], Marc-Antoine Mahieu &
Nicole Tersis (eds), 97–114. Amsterdam: John Benjamins.
Woodbury, Anthony C. 1996. On restricting the role of morphology in autolexical syntax.
In Autolexical Theory: Ideas and Methods, Eric Schiller, Elisa Steinberg & Barbara Need (eds),
319–363. Berlin: Mouton de Gruyter.