Pirka Itak

SOAS Working Papers in Linguistics Vol.
18 (2016): 3-28
Morphological alignment in Saru Ainu:

A direct-inverse analysis
Elia Dal Corso

615195@soas.ac.uk
Abstract
The Saru dialect of the Ainu language displays a full system of person agreement affixes, that
constitutes a fundamental part of the verb‘s morphological structure. Although this agreement
system can be said to be one of the most studied features of Saru Ainu, there is not yet, at the
best of my knowledge, a unitary approach to account for its peculiarities. Among these
peculiarities, the uneven formal variation of affixes to indicate subject and object referents
depending on grammatical person is perhaps the most striking one. The analysis in this article
stems from previous research on this topic, and attempts to propose that Saru Ainu‘s
agreement system represents a case of direct-inverse alignment. This direct-inverse approach
can account for the apparent discrepancies in the formal realization of affixes and it gives a
smoother picture of morphological marking of referents.
Keywords: morphology; alignment; Ainu; person; grammatical function; transitivity;

inverseness
1. Introduction
1.1. The Ainu language
Ainu is an endangered indigenous language of Japan, spoken in its northernmost island of
Hokkaidō. Other varieties of Ainu were also spoken throughout Sakhalin and the Kuril
Islands in today‘s Russia, though today the Ainu language survives just in its Japanese
variety. Hokkaidō Ainu is believed to have less than 20 native speakers remaining (Vovin,
1993; Bradley, 2007). The other two varieties of the language (Sakhalin and Kuril Ainu) are
extinct today. Hokkaidō Ainu can be further subdivided into smaller dialects according to
differences in lexicon, morphology and, more rarely, syntax (Bugaeva, 2004:7). Ainu has no
acknowledged genetic relation with any of the neighboring languages of Japan or continental
Russia. Although scholars have proposed different categorizations for the language 1, it is still
difficult to obtain a reliable proof for these theories.
This article focuses on the Saru dialect of Ainu (henceforth SA). This is included among the
Southern Hokkaidō dialects and it was originally spoken along the Saru river, from which it
takes its name. A corpus of recorded and transcribed materials2 in this dialect is used as the
main resource for this study. These materials were collected through elicitation from native
speakers by Tamura Suzuko in the 1950s and 1960s during fieldwork in Hokkaidō.
1.2. Aims and structure of the paper

With this paper I propose a unitary description of morphological alignment in SA. The
speculation on Ainu alignment has not gone far in the past literature, so the kind of alignment
displayed by the language is still an unsettled matter. The main observations on alignment
have been made by Bugaeva (2006). She considers Ainu‘s morphological alignment looking
at the realization and distribution of verbal personal affixes. Although this contribution is
1
See for example proposals by Street, 1962; Shafer, 1965; Vovin, 1993; Greenberg, 2000-2002.
2
Tamura, Suzuko. Ainu go onsei shiryō. online resource –http://dspace.wul.waseda.ac.jp/dspace
4 Elia Dal Corso
important as it provides a preliminary understanding of the nature of Ainu alignment, the

author does not linger on this subject, with many issues left unsolved. Following from
Bugaeva‘s observations, I then carry the analysis forward with the aim of enhancing the
description of SA‘s morphological alignment.
I should point out that this paper is not concerned with syntactic alignment. This kind of
alignment is reportedly sensitive to particular syntactic operations, such as control, reflexive
binding or argument gapping in coordination among others (Dixon, 1979; Manning, 1996).
Following Dixon and Manning, I assume that there are different levels of structure within one
language. These different levels may be sensitive to different kinds of alignment and,
therefore, are better treated separately. One of these levels (also assumed in Dixon‘s and
Manning‘s framework) is the morphological level. On this level of structure, alignment is
sensitive to the formal realization of verbal arguments. When looking at SA from this
perspective, a number of discrepancies arise in the formal realization and use of verb personal
affixes. This appears to depend mostly on grammatical person, seemingly suggesting
different kinds of alignment for each one of them (see §2.2. and §2.3). In addressing these
discrepancies, I pay particular attention to the linear order of verb arguments, thematic roles,
and referencing of speech act participants. The interaction of these features in connection to
morphological alignment eventually explains the apparently unexpected formal realization of
arguments witnessed in SA. Moreover, the outcome of this approach is a proposal for the
existence of a morphologically direct-inverse alignment in this dialect (see §4). This direct-
inverse approach has the ultimate value of bringing all grammatical persons together, and it
gives a smoother picture of alignment. To the best of my knowledge, this has never been
proposed before for SA. The analysis contained in this paper is thus useful to deepen our
knowledge about the Ainu language altogether and, more specifically, it is a valuable
contribution to the speculation on Ainu‘s morphological alignment. Moreover, from a wider
perspective, this work may enhance our understanding of inverseness as it is found cross-
linguistically. The case of SA, in fact, deviates in many aspects with regard to the alleged
prototypical characteristics of inverseness. This new contribution could improve the
typological profile of inverseness, eventually refining our approach to other inverse systems
around the world.
The paper is organized as follows. In §2. I present SA‘s personal affixes after I introduce the
verb classification I assume throughout the analysis. In this section I also present and discuss
briefly Bugaeva‘s observations on alignment, setting the background information necessary
to develop my own analysis. I dedicate §3. to highlighting faults and weakness of previous
proposals and to introducing new tools to analyze alignment. In §4. I discuss the direct-
inverse approach to SA‘s alignment, while in §5. I summarize my findings and underline
some issues that remain unclear.
2. Agreement and alignment

2.1. Verb classes
Before I discuss how personal agreement is marked in SA, it is necessary to give a brief
introduction of the verb classes that are distinguished within the language. In descriptive
grammars, Ainu verbs are organized into separate classes according to their valency. The
basic two-way distinction that separates intransitives and transitives is expanded in Ainu to
also include ditransitives and complete 3 verbs (Tamura, 2000: 41-42). This is not a
3
Following Tamura, I here use the term ‗complete verb‘ based on the translation of 完全動詞 kanzen dōshi,
which is the accepted term used in Japanese literature on Ainu.
Morphological alignment in Saru Ainu: A direct-inverse analysis 5
prerogative of SA, rather it appears to be the same also for other dialects or varieties, for
which the same terminology is used (Murasaki, 19764; Refsing, 1986; Tamura, 2010). Verb
valency interacts with the category of grammatical person, as there may be different
realizations of the same personal affix depending on whether it marks the S, A or O argument
of the verb. Since here I am concerned specifically with the formal realization of personal
affixes, throughout my argumentation I assume that verb valency is fundamental for the
description of morphological alignment.
Complete verbs (as their name suggests) do not subcategorize for any argument (‗verb‘ < - >).
Complete verbs, among the other types of verb, are few in number and describe states or
conditions, especially related to time and the weather.
(1) Orano, sir-kunne kor…

and condition-be.dark when
‗And, when the night came…‘ (Tamura, 1985: 6)
(2) Me-an.
cold-be.PC
‗It is cold.‘ (Tamura, 2000: 41)
Intransitive verbs subcategorize for just one argument – the subject (‗verb‘ <SUBJ>5).
(3) A-ekasi […] Ø-soyne.

4SUBJ-father 3SG.SUBJ-go.out
‗My father […] went out.‘ (Tamura, 1985: 48)
Transitive verbs subcategorize for a total of two arguments – the subject and the primary
object (‗verb‘ <SUBJ, OBJ>).
(4) Ene Ø-iki hi a-Ø-nukar.

like.this 3SG.SUBJ-do NMLZ 4SUBJ-3SG.OBJ-see
‗I saw that he did like this.‘ (Tamura, 1985: 48)
Ditransitive verbs, finally, subcategorize for a total of three arguments – the subject, the
primary object, and the secondary object (‗verb‘ <SUBJ, OBJ, OBJ2>). If the two OBJs are
overtly expressed with an NP, they are told apart by linear order, with OBJ being closer to the
verb than OBJ2. As we see in (5) this is not mirrored in the position of personal affixes on the
verb, where OBJ2 overrides OBJ, appearing next to the predicate. This is discussed in §2.4.
below.
4
Murasaki (1976) also uses the labelling V0, V1, V2 to classify verbs, thus indicating the number of arguments
for which a certain verb subcategorizes.
5
The notations SUBJ, OBJ, OBJ2 are intended, here and throughout the argumentation, as devices to represent the
arguments of a given verb – i.e. they represent the verb‘s argument structure. Similarly to what is assumed
within the LFG framework (Bresnan et al., 2016: 326), I consider argument structure to be an interface between
thematic role and grammatical function of a verb‘s predicators (see §2.4). Differently from this approach,
however, I here intend grammatical functions as being indicators of the functional relationship of predicators not
at the syntactic level, but rather exclusively at the morphological level. This fits with the scope of the analysis,
i.e. morphological alignment, and with the assumption that syntactic and morphological alignment should be
treated separately (as pointed out in §1.2).
6 Elia Dal Corso
(5) Caca Ø-wen sisam icen Ø-Ø-kore6.

old.man 3SG.SUBJ-be.bad japanese money 3SG. SUBJ-3SG.OBJ2-give
‗The old man gave money to the poor Japanese.‘ (Tamura, 2000: 42)
Throughout my argumentation I focus specifically on transitive and ditransitive verbs,

considering intransitives only briefly. Complete verbs, on the other hand, are not relevant for
the analysis to come, so they will not be discussed further.
2.2. Personal affixes in SA

2.2.1. Personal affixes as agreement markers
Personal affixes are possibly one of the most described aspects in Ainology. Exhaustive
descriptions of the personal affix systems are available for both the Hokkaidō and Sakhalin
varieties of Ainu7. In discussing SA personal affixes in this paper, I refer to Tamura (1970,
1972, 2000).
Personal affixes are portmanteaus that indicate the functional features of a referent, which is a
participant in the event expressed by the verb. The referent‘s features they encode are
grammatical person (first, second, third or fourth), number (singular or plural), and
grammatical function (SUBJ, OBJ or OBJ2). They are dependent parts of speech that form a
unitary morphophonological word with the host verb 8. Evidence of this comes from the cases
of vowel elision observed at morpheme boundaries and from stress9 shift, that may affect the
verb when a personal affix is added (Tamura, 1970: 580-587). Personal affixes are
obligatorily expressed on each verb in a sentence, even though the referent they indicate may
be clearly understandable from the context. Example (6) shows two coordinated verbs with
the same second person singular subject referent which appears marked via a personal affix
just on the first coordinated verb – this sentence is unacceptable.
(6) * Eani su e-Ø-suke wa Ø-Ø-e.

you broth 2SG. SUBJ-3SG.OBJ-cook and 2SG. SUBJ-3SG.OBJ-eat
‗You make broth and eat it.‘
SA also employs what in the literature are called ‗personal pronouns‘ (Tamura, 1970: 578;
2000: 47). Personal pronouns are independent words and they precede the verb, accordingly
to Ainu‘s canonical word order (Tamura, 2000: 25-35). Personal pronouns may never be a
substitute for their relative personal affix. That is to say, the personal affix must appear on the
verb even though a personal pronoun is overtly present. Examples (7) and (8) show an
unacceptable sentence where the personal pronoun kani ‗I‘ is used as a substitute for the affix
ku- and the corresponding grammatical sentence.
(7) * Kani Ø-arpa.

I 1SG. SUBJ-GO.PC
‗I go.‘
6
Why the OBJ is not included among verb personal affixes in the glossing is also discussed in §2.4.
7
For an analysis of personal affixes in Sakhalin Ainu see Hattori (1961) and Murasaki (1976), or for other
Hokkaidō dialects see Refsing (1986), Bugaeva (2004), Tamura (2010), Bugaeva (2012), Takahashi (2015).
8
An exception to this seems to be the fourth person suffix -an. According to Tamura (1970: 587-589) this is the
only personal affix that can be said to stand on a borderline between a suffix and an independent word. The
reason of this may be an ongoing process of grammaticalization.
9
I use here the term ‗stress‘ following again Tamura‘s (2000) terminology. However, many other scholars have
proposed that Ainu has in fact pitch accent (see for example Vovin, 1993).
(8) Kani k-arpa.

I 1SG. SUBJ-GO.PC
‗I go.‘ (Tamura, 1972: 24)
Ainu verbs rarely encode grammatical categories (such as tense, aspect, number or person).
These categories are expressed through the use of separate morphology or syntax (Tamura,
2000). It follows that, in order to relate the verb to its NP arguments, we expect some kind of
obligatory agreement to match categories like person and number between V and NPs. In
light of the behavior of personal pronouns in interaction with personal affixes in SA, it
appears that the latter have the function of signaling agreement. Considering this and bearing
in mind also their morphophonological characteristics, I believe it is advisable to redefine
personal affixes. Following Oku (2008), I assume that personal affixes are better recognized
as agreement markers. Personal pronouns, on the other hand, are simply used as emphatic
devices for pragmatic purposes10. Relabelling personal affixes as ‗agreement markers‘ gives a
clearer idea of their function. On the other hand, it does not make a substantial difference
whether I use the term ‗personal pronouns‘ or another one to describe words like kani in (8),
at least as far as this analysis is concerned. I presented personal pronouns here first to
underline the properties of agreement markers, and to provide an all-round summary on
referent marking on SA verbs, but they will not be discussed further.
2.2.2. Formal realization of agreement markers

One characteristic of agreement markers is the possibility to have different formal
realizations, according to the grammatical function covered by their referent. Different formal
realizations of agreement markers has been documented for all Ainu dialects, and in
particular by Tamura (1970, 1972, 2000) for SA. The outcomes of previous research on this
topic formed the basis for speculation about Ainu morphological alignment (Bugaeva, 2006).
The difference that shows formally in the expression of agreement markers involves
grammatical functions, in that one form is available to mark the referent when it is SUBJ,
while a separate one is available to mark the same referent when it is either OBJ or OBJ2.
Given this distinction, the terminology used to refer to the two different forms of a same
agreement marker is ‗nominative‘ for the one marking SUBJ, and ‗accusative‘ for the one
marking OBJs (Tamura, 2000: 58)11. For the time being, I accept this terminology for the
10
It is worth mentioning how the emphatic use of personal pronouns seems to be possible only when their
referent has the grammatical function SUBJ. In Tamura (2000: 62-71) all the examples provided by the author
feature a personal pronoun with a SUBJ referent, expressed via agreement marker on the verb. The same is also
true for the cases of personal affixation found in the texts collected by Tamura. The only exception to this
behavior is witnessed with the second persons. In this case the personal pronoun may have a referent with the
grammatical function OBJ.
Ecioka ka eci-tak-pa.
you even 1SG.SUBJ>2PL.OBJ-INVITE-PL
‗I invite you too.‘ (Tamura, 2000: 66)
Here the prefix eci- is a portmanteau of a first person acting on a secon person object (see §4.2. below). Having
an OBJ reiterated via a personal pronoun for any other grammatical person is not possible. Moreover, it appears
to be unacceptable to reiterate any grammatical person with the grammatical function SUBJ if the OBJ is a second
person referent. Although this discrepancy seems unexpected, it is correctly predicted by the direct-inverse
approach presented in §4.
11
The same terminology is used in Tamura (1972), where the Japanese terms shukaku and mokutekikaku
(translating ‗nominative‘ and ‗accusative‘ respectively) are present. Tamura (1970: 578) uses the terms
‗nominative‘ and ‗objective‘ following Hattori (1961).
8 Elia Dal Corso
different forms of agreement markers. However, this soon proves to be inadequate for SA, as
the following examples suggest.
First person singular is marked with the marker ku-12 (‗nominative‘) when the first person
referent is SUBJ of the verb (S or A argument), while it is marked with en- (‗accusative‘)
when the referent is an OBJ (O argument).
(9) a. Kani k-arpa.

I 1SG. SUBJ-GO.PC
‗I go.‘ (Tamura, 1972: 24)
b. Ponkurmat eun ku-Ø-ye ka Ø-Ø-ki.

little.girl towards 1SG. SUBJ-3SG.OBJ-say even SLV-V.OBJ-do
‗I even said it to the little girl.‘ (Tamura, 1984: 12)
c. A-en-ipe-re ka somo Ø-Ø-ki.

4 SUBJ-1SG.OBJ-eat-CAUS even not SLV-V.OBJ-do
‗People did not even feed me.‘ (Tamura, 1972: 18)
In the same way, first person plural is marked differently for ‗nominative‘ and ‗accusative‘ –
the former being marked with the prefix ci- (A argument) or with the suffix -as (S argument),
and the latter with un- (O argument). This set of agreement markers strictly expresses a first
person plural exclusive – that is, ‗we‘ includes the speaker and some other third participant,
but not the listener.
(10) a. Cis-as kor arki-as.

cry-1PL.SUBJ.EXCL while go.PL-1PL. SUBJ.EXCL
‗We went while crying.‘ (Tamura, 1972: 18)
b. Unu ka ona ka ci-Ø-sak

mother even father even 1PL.SUBJ.EXCL-3PL.OBJ-not.have
‗We do not have a mother nor a father.‘ (Tamura, 1972: 18)
c. Ø-un-toykokikkik
3PL.SUBJ-1PL.OBJ.EXCL-beat.violently
‗They beat us up.‘ (Tamura, 1972: 18)
Second person singular is always marked with the prefix e- independently from the
grammatical function covered by its referent (S, A and O arguments) – the ‗nominative‘ and
‗accusative‘ affixes thus appear to have the same morphological realization. This same
behavior shows for the second person plural. In this case too we have the same morphological
realization (eci-) regardless of the grammatical function of the referent.
(11) a. E-eraman ruwe?

2SG.SUBJ-understand EV.DIR
‗Did you understand?‘ (Tamura, 1972: 27)
12
Both first person singular ku- and first person plural ci- may appear respectively as the allomorphs k- and c-
after vowel elision. This happens when ku- is followed by a verb starting with any vowel but i and when ci- is
followed by a verb starting with any vowel (Tamura, 1970: 581-584).
b. A-e-ko-nu yakun e-Ø-ye […]

4SUBJ-2SG.OBJ-APPL-hear if 2SG.SUBJ-3SG.OBJ-say
‗If people ask you about it, you say it […]‘ (Tamura, 1972: 28)
(12) a. Eci-iki
2PL.SUBJ-do
‗You do.‘ (Tamura, 1985: 64)
b. Eci-Ø-nu ya?
2PL.SUBJ-3SG.OBJ-hear FIN
‗Do you hear it?‘ (Tamura, 1972: 28)
c. A-eci-tak kusu […]

4.SUBJ-2PL.OBJ-invite because
‗Since they invite you […]‘ (Tamura, 1985: 64)
Third person singular and third person plural are marked via the zero prefix Ø-. This is
consistent for S, A and O arguments equally. As an example take sentences (3), (11b) and
(10c) above, repeated here as (13).
(13) a. A-ekasi […] Ø-soyne.

4SUBJ-father 3SG.SUBJ-go.out
‗My father […] went out.‘ (Tamura, 1985: 48)
b. A-e-ko-nu yakun e-Ø-ye […]

4SUBJ-2SG.OBJ-APPL-hear if 2SG.SUBJ-3SG.OBJ-say
‗If people ask you about it, you say it […]‘ (Tamura, 1972: 28)
c. Ø-un-toykokikkik.
3PL.SUBJ-1PL.OBJ.EXCL-beat.violently
‗They beat us up.‘ (Tamura, 1972: 18)
Also in this case the ‗nominative‘ and the ‗accusative‘ affixes are joined by the same
morphological realization. In this respect third person behaves like the second person.
2.2.2.1. The fourth person

SA‘s fourth person 13 has the characteristic of not being univocally linked to one fixed
participant in natural discourse (i.e. ‗I‘, ‗you‘, or some other external third person). In this
respect it is thus different from other grammatical persons treated in §2.2.2. Due to its
polysemous usage, it deserves to be treated separately.
Like first person plural, fourth person can be marked in three different ways. The ‗nominative‘
affixes are a- (A argument) and -an (S argument), and the ‗accusative‘ affix is i- (O
argument).
13
I use the term ‗fourth person‘ following the trend found in recent works on Ainu (for instance Bugaeva, 2004;
Tamura, 2010). This is an umbrella definition, in substitution to Tamura‘s (2000) term ‗indefinite person‘, to
group all meanings borne by this grammatical person, that are not limited to referencing to an indefinite agent.
Fourth person is not intended as an obviative person like in some approaches to inverse languages (see §4.2).
10 Elia Dal Corso
(14) a. Nisapno ar-siknak-an.

quickly completely-be.blind-4SUBJ
‗I became completely blind quickly.‘ (Tamura, 1985: 2)
b. Hunak un ka a-i-y-ani.
where to even 4SUBJ-4OBJ-0-carry
‗They carried me to somewhere.‘ (Tamura, 1985: 4)
Fourth person is used to express the inclusive first person plural – that is, a ‗we‘ that includes
both the speaker and the listener. With this use, it compensates for the restriction on first
person plural that solely indicates an exclusive first person, as seen in (10) above.
(15) Hetak, paye-an wa ipe-an ro.

INT go.PL-4SUBJ and eat-4SUBJ FIN
‗Come on, let us go and eat.‘ (Tamura, 1972: 19)
It can also mark an honorific second person. According to Tamura (1972) it was customary
for Ainu women to substitute second person with fourth person when speaking to men. This
switch was otherwise a general way to show respect towards the listener. In (16) a woman is
worried about her husband.
(16) A-sik-ihi a-Ø-arka-re hawe?

4SUBJ-eye-POSS 4SUBJ-3SG.OBJ-hurt-CAUS EV.DIR
‗Did you hurt your eye?‘ (Tamura, 1972: 22)
Fourth person is also used to mark an indefinite agent. This happens when the action is
performed by an agent that is either not important or unknown – i.e. when the agent bears a
low pragmatic and/or semantic content (Tamura, 2000: 71).
(17) Nupuri a-Ø-nukar.

mountain 4SUBJ-3SG.OBJ-see
‗Someone sees the mountain.‘ / ‗The mountain is visible.‘ (Tamura, 2000: 71)
Finally, fourth person is used to mark first person within direct quotation. In this case it has a
logophoric function (Bugaeva, 2006).
(18) “Arpa-an kusu ne na” sekor Ø-hawean.

go.PC-4SUBJ FUT FIN COMP 3SG.SUBJ-say
‗He said: ―I will go‖.‘ (Tamura, 2000: 66)
A logophoric function is recognized also in the use of fourth person to mark a first person
narrator in traditional folktales, where the narrator identifies itself with the character acting in
the tale (see also (14) above).
(19) Sine po Ø-Ø-kor unarpe a-ne

one child 3SG.SUBJ-3SG.OBJ-have middle.age.woman 4SUBJ-COP
hine an-an.
and be.PC-4SUBJ
‗I was a middle-aged woman with one child.‘ (Tamura, 1985: 2)
Starting from this information, I now continue by looking at how formal realization of
agreement influences our perception of morphological alignment.
2.3. Different alignments

I used examples in §2.2. to present the different realizations of agreement markers on SA
verbs according to the grammatical function and argument function (S, A or O) or their
referent. This is better represented in the following table.
Table 1: Formal realization of personal agreement

SUBJ OBJ
S argument A argument O argument
1SG ku- en-
1PL -as ci- un-
2SG e-
2PL eci-
3 Ø-
4 -an a- i-
Up to this stage of the analysis, I assumed a division of agreement markers into ‗nominatives‘
and ‗accusatives‘ following Tamura (2000). However, a careful look at their realization
suggests that these labels may be faulty. Why are there cases like e- where we have no formal
change among different argument functions? Why are there cases, like the fourth person,
where a different formal realization is available for S and A arguments, that both mark the
same grammatical function (i.e. SUBJ)? These questions challenge Tamura‘s terminology.
Moreover, why do we witness such an apparently random distribution in the number of
available affixes throughout different grammatical persons?
Bugaeva (2006: 185-186) makes sense of this discordant behavior of agreement markers by
proposing that Ainu simultaneously displays three different kinds of morphological alignment.
Based on first person singular markers, she states that Ainu has a nominative-accusative
alignment, that formally groups together S and A and treats O differently. In the same way,
fourth person and first person plural on one hand, and second persons and third person on the
other are her basis for arguing that a tripartite and a direct alignment are present too. Three
different alignments seem to coexist within the same language – each one is linked to one or
more specific grammatical person.
Table 2: Three different alignments

Nominative-
Tripartite Direct
accusative
S S S
A O A O A O
Given what we see at the morphological level, Bugaeva‘s deduction seems indeed sensible.
However, it is not clear what the productivity would be for such an organization of agreement
on the same structural level. Moreover, a closer look at SA‘s agreement system shows some
characteristics of agreement markers that may prompt a revision of Bugaeva‘s model.
12 Elia Dal Corso
2.4. Constraints on argument marking: grammatical-function-oriented approach

By looking at agreement affixation on ditransitives, we can notice an important constraint of
verbs in regards to argument marking. That is, in SA it is not acceptable to have more than
two agreement markers, each one indicating a different referent, affixed on the same verb. It
follows that only a maximum of two arguments included in the subcategorization may be
expressed via agreement marking on a given verb. This can be recognized as a structural
constraint that applies on the level of verb morphology. While it does not affect intransitives
and transitives for obvious reasons, ditransitives must employ some strategy to avoid the
ungrammaticality shown below. (20) exemplifies an unacceptable verb form where
ungrammaticality is triggered exactly by the simultaneous presence of the prefixes a-, i- and
e-, that refer to the three subcategorized arguments of the ditransitive ekoyayirayke ‗thank
for‘.
(20) * A-i-e-e-ko-yayirayke.
4 SUBJ-4OBJ-2SG.OBJ2-APPL-APPL-thank.for
‗I thank you for something.‘
If one of the arguments is implied or expressed via a separate NP, and so the verb form hosts
just two agreement markers, then this verb form is grammatical.
(21) Tapne a-i-siknu-re wa a-e-e-ko-yayirayke.

like.this 4SUBJ-4OBJ-live-CAUS and 4SUBJ-2SG.OBJ2-APPL-APPL-thank.for
‗You saved my life like this and I thank you (for that).‘ (Tamura, 1984: 12)
The implication of this constraint is clear – ditransitives have one subcategorized argument
that must be excluded from the morphological verb structure. How this argument gets
expressed alternatively, if not with an agreement marker, is not an issue here. On the other
hand, I am concerned with which argument is the excluded argument, and what parameters
give a certain argument the priority to appear affixed on the verb. In order to understand this,
I firstly assume the internal verb structure below14.
Table 3: Assumed internal structure of the verb

agreement slot 1 agreement slot 2 verb
The assumption here is that, for any given SA verb, there are up to two available slots for
agreement markers. Whether these slots are actually occupied or not is decided by verb
valency. So, even though two slots are available, they are empty for complete verbs or
partially occupied for intransitives. This is because there are no arguments in the
subcategorization of these verb classes that can occupy a given slot in the first place. It
follows that the only cases where both slots are taken are with transitives and ditransitives.
These two verb classes are the perfect environment to speculate whether each slot allows only
arguments with specific parameters or the assignation is arbitrary.
From the observations in §2.2. we have the first evidence to say that the agreement marker
referring to the referent with the grammatical function SUBJ can never be omitted from the
14
This structure is intended to represent the internal structure of all SA verbs independently from their class. It
does not mirror the actual linear distribution of agreement markers that, on intransitives, may also be suffixed
and then follow the verb. Both the slots are here theoretically assumed to precede the verb for the sake of
convenience.
verb structure15. For instance, having the agreement marker substituted by a personal pronoun
results in an ungrammatical sentence (see (7) above). This shows a formal obligatoriness.
Moreover, the SUBJ referent cannot be the one subcategorized argument excluded from
agreement markers within verb structure. This is shown in (22) where a first person SUBJ does
not occupy slot 1 nor slot 2. This alternatively shows a structural obligatoriness.
(22) * Kani i-e-e-ko-yayirayke.

I 4OBJ-2SG.OBJ2-APPL-APPL-thank
‗I thank you for something.‘
This limitation in the affixation of agreement markers shows that one of the two slots must be
reserved for the SUBJ referent. Then we may wonder whether SUBJ can appear in slot 1 as well
as in slot 2 when an OBJ referent is also present. Examples (23) and (24) give the answer to
this question.
(23) A-en-ipe-re ka somo Ø-Ø-ki.

4SUBJ-1SG.OBJ-eat-CAUS even not SLV-V.OBJ-do
‗People did not even feed me.‘ (Tamura, 1972: 18)
(24) * En-a-ipe-re ka somo Ø-Ø-ki.

1SG.OBJ-4SUBJ-eat-CAUS even not SLV-V.OBJ-do
‗People did not even feed me.‘
It appears that it is not possible to reverse the order of two agreement markers, not even when
the grammatical function of their referents is unmistakably understandable from their
‗nominative‘ or ‗accusative‘ formal realization. We see then that not only must SUBJ always
be expressed with an agreement marker, but also that this agreement marker needs to occupy
slot 1 within the verb structure. The univocality by which slot 1 can be only taken by the
agreement marker referring to SUBJ also suggests that each slot may be linked to a particular
grammatical function.
When dealing with OBJs, I rely on thematic roles to distinguish OBJ from OBJ2. The following
examples sustain the proposal by which each slot in the verb structure indeed allows just
agreement markers with referents that have one particular grammatical function. Slot 2 in fact
allows exclusively agreement markers with an OBJ referent. Moreover, it appears that verb
slots are also selective of a specific thematic role of referents. Slot 1 (reserved for SUBJ
referent) allows in fact an agreement marker whose referent has also the thematic role
agent/experiencer (depending on verb semantics). On the other hand, the thematic role of OBJ
referent allowed by slot 2 varies according to verb class. With transitives OBJ has the thematic
role patient/theme and with ditransitives it has the thematic role goal/beneficiary.
(25) Rur takup a-i-kore.

broth only 4SUBJ-4.OBJ2-give
‗They give me nothing but broth.‘ / ‗They give nothing but broth to me.‘
(Tamura, 1984: 2)
15
This can be said to be true also for cases of portmanteau (see Footnote 10) where the SUBJ referent is included
with the OBJ within one sole agreement marker.
14 Elia Dal Corso
(26) ?Po icen a-Ø-kore.

child money 4SUBJ-3SG.OBJ-give
‗Someone gives the money to the child.‘
Example (25) shows the ditransitive verb kore ‗give‘ where slot 1 and slot 2 indicate
respectively SUBJ and OBJ2 referents. In this case the excluded argument is OBJ, expressed
here as the separate NP rur ‗broth‘. In (26) the verb is the same, but now the excluded
argument is meant to be OBJ2 (the beneficiary), while OBJ (the theme) occupies slot 2. This
sentence is not ungrammatical, but the meaning with which it is intended is different from its
actual one. (26) in fact can only have the reading ‗Someone gives the child to the money‘ – it
is the one OBJ with the thematic role of goal/beneficiary which has the priority to appear
among agreement markers. Goal/beneficiary cannot be expressed via an oblique to leave the
slot free for OBJ patient/theme either, as this would go against the subcategorization of a
ditransitive that wants both OBJs expressed as core arguments. At the same time, there is also
structural evidence that the one OBJ with thematic role goal/beneficiary must be the secondary
object. With transitives slot 2 cannot refer to anything but an OBJ referent with a thematic role
patient/theme – this OBJ never has a thematic role goal/beneficiary. In (28) one of the
arguments (tanpe ‗this thing‘) is expressed with an oblique, so as not to violate the
subcategorization constraint of the transitive ye ‗say‘. The goal/beneficiary ponkurmat ‗little
girl‘ is here treated as the only (primary) OBJ and still the sentence is unacceptable.
(27) Ponkurmat eun ku-Ø-ye ka Ø-Ø-ki.

little.girl towards 1SG.SUBJ-3SG.OBJ-say even SLV-V.OBJ-do
‗I even said it to the little girl.‘ (Tamura, 1984: 12)
(28) * Tanpe ani ponkurmat ku-Ø-ye ka Ø-Ø-ki.

this.thing with little.girl 1SG.SUBJ-3SG.OBJ2-say even SLV-V.OBJ-do
‗I even said this thing to the little girl.‘
As shown here, the thematic role of the referent cannot be changed as it is univocally linked
to a particular slot according to verb class. This ultimately suggests that thematic role is in
fact one of the parameters shaping slots within the verb structure. The necessity of
distinguishing two kinds of OBJs follows from this. In fact, to argue for a unitary grammatical
function OBJ whose referent may have a goal/beneficiary or patient/theme thematic role
would mean the admission of acceptability for sentences like (26) and (28), that are incorrect
or subject to misinterpretation. The main implication of this, however, is that there is a one-
to-one correspondence of grammatical functions and thematic roles when it comes to verb
slots. Precisely, the grammatical function OBJ subsumes the thematic role patient/theme and
the grammatical function OBJ2 subsumes the thematic role goal/beneficiary. The fact that
grammatical function OBJ2 and thematic role goal/beneficiary override grammatical function
OBJ and thematic role patient/theme is due to the theoretical assumption of a thematic
hierarchy as in Grimshaw (1990)16. From this, I follow Dalrymple (1990: 169-170) in arguing
that, with regards to grammatical functions, among SUBJ, OBJ and OBJ2 all are equally ranked,
and that it is the thematic hierarchy of each grammatical function that determines their
relative superiority. Given the univocal correspondence of grammatical functions and
thematic roles discussed here, it follows that the only grammatical function hierarchy I can
assume ranks SUBJ> OBJ2> OBJ.
16
Grimshaw‘s hierarchy, where goal outranks theme, in fact organizes thematic roles as agent > experiencer >
goal/location/source > theme.
Table 4: Assumed grammatical function and thematic role hierarchies

agent/ goal/ patient/
TR > >
exp. benef. theme
GF SUBJ > OBJ2 > OBJ
These hierarchies formalize the accessibility of referents that can potentially be expressed via
agreement markers on the verb. The two referents whose grammatical function is included in
the subcategorization of the verb and that also bear the highest thematic role on the hierarchy
are eligible to appear affixed as agreement markers. These agreement markers may have
different morphological realizations according to the functional features of their referent, as
presented in §2.2. Discussing grammatical functions of referents was necessary to introduce
thematic roles and their correspondence to them. Thematic roles of referents are in turn
fundamental to develop the direct-inverse proposal for SA‘s alignment presented in §4.
My assumption here is that SA must have a way to morphologically mark which is the most
important referent involved in an event. Considering the data available and the observations
at hand, I could define this ‗importance‘ by means of grammatical functions and thematic
roles. I could assume that a language usually tends to give prominence to the grammatical
function SUBJ (the function around which the sentence revolves), or to the thematic role of
agent/experiencer (the one who performs/perceives the event). This concept of ‗markedness‘
would be based on fixed hierarchies (in my case Grimshaw‘s hierarchy), that define which
grammatical function and thematic role outranks the others. From this I should expect SA to
mark a grammatical function like SUBJ more overtly than or differently from OBJs, in order to
have a systematic way to understand the dynamics of interaction among their referents within
a sentence. At the morphological level, this would be obtained exactly by means of separate
formal realizations of same-referent agreement markers, like the ones presented above for
first person singular (ku-/en-). This eventually supports Tamura‘s ‗nominative‘-‗accusative‘
terminology. Nonetheless, cases like the second person singular marker e- or the third person
zero marker challenge this view. In these cases two grammatical functions and two thematic
roles that should be clearly distinguished in fact appear formally as the same. Moreover, the
fact that just one of the available persons (i.e. the third) is never overtly marked is not in line
with this approach either. To clarify this, I may need to abandon Tamura‘s terminology.
However, this could also mean to negate the importance that grammatical functions and
thematic roles are argued to have in deciding the formal markedness of referents in the first
place. This grammatical-function-oriented approach then proves as misleading as the
‗nominative‘-‗accusative‘ one. Nevertheless, this is not the case. In fact, one more parameter
besides the already established grammatical functions and thematic roles is necessary to grasp
the logic of SA morphological alignment. This eventually allows a revision of Tamura‘s
terminology without affecting the grammatical-function-oriented approach. This parameter is
linked to the pragmatic primacy of participants (i.e. the referents of agreement markers)
within the event. Its characteristics and implications on alignment are presented in §3, where
I also introduce an alternative terminology to describe SA alignment itself.
3. Re-defining markedness
3.1. (Di)transitive personal affixes
The best environment to speculate upon participants‘ interaction within the event described
by the verb is again with transitives or ditransitives. Table 5 below17 schematizes how the
17
Adapted from Tamura (2000: 59).
16 Elia Dal Corso
pragmatic interaction between two participants is reflected by agreement markers on

morphology. Participants are defined by person, number (except third and fourth persons) and
grammatical function.
Table 5: Transitive agreement markers of SA

OBJ
1SG 1PL 2SG 2PL 3 4
1SG -18 - eci- eci- ku-Ø- ku-i-
1PL - - eci- eci- ci-Ø- ci-i-
2SG en- un- - - e-Ø- e-i-
SUBJ
2PL eci-en- eci-un- - - eci-Ø- eci-i-

3 Ø-en- Ø-un- Ø-e- Ø-eci- Ø-Ø- Ø-i-
4 a-en- a-un- a-e- a-eci- a-Ø- a-i-
From this table we can see that in most cases both participants in the event can be separately
recognized from a clearly retrievable agreement marker included in the verb structure slots.
This agreement marker figures among the ones available for those particular persons.
Moreover, we see that in all these cases the participant with the grammatical function SUBJ is
expressed via the ‗nominative‘ affix and precedes the OBJ, occupying thus slot 1. This is
exactly what was predicted by the analysis above. Six of the boxes in the table (highlighted in
grey), on the other hand, show an anomalous case of agreement affixation. These are the only
cases where two separate affixes are not straightforwardly recognizable, and where there is
no ground to speculate whether slot 1 or slot 2 is the one occupied by the affix in the structure.
These are portmanteau affixes that bear not only functional information about person, number
and grammatical function, but also pragmatic information about participant interaction (e.g.
un- expresses a second person singular acting on a first person plural). They are in fact
special cases and, as such, they are treated separately in the following sections (see §3.2. and
§4.2.2). For the time being, it is important to notice that the participants these portmanteaus
refer to are always first and second persons.
3.2. Speech act participants – the saliency hierarchy

Speech act participant (SAP) is a term to indicate first and second person referents, separating
them from third person referents, which in turn are defined non-SAPs. The concept of SAP is
based on a pragmatic approach to the category of person, that groups first and second persons
together as referring to the two central participants in a given conversational context (i.e.
‗I‘/‗we‘ and ‗you‘). In this sense person becomes a category of natural discourse more than a
grammatical category (Klaiman, 1992: 236). By including this pragmatic-based paramenter
of SAPs to the framework, I imply that if a referent is recognized as a SAP then it must be
morphologically marked differently from non-SAPs. This assumed different marking entails
some kind of primacy of referents based on pragmatics. I call this primacy ‗saliency‘ in
substitution for the more loose term ‗importance‘ mentioned above.
This kind of referent saliency felicitously explains SA‘s zero marked third person (non-SAP),
in that formal unmarkedness distinguishes it from overtly marked first and second persons
(SAPs). At the same time, however, this referent saliency fails to account for cases of
agreement marker formalization that are not predicted by the SAP/non-SAP disambiguation.
It is not clear, for instance, why first and second persons (both SAPs) may or may not have
18
Hyphens indicate interaction of one or more participant(s), either with itself or with each other, that are
expressed morphologically via specialized reflexive or reciprocal affixes.
different agreement marker forms depending on the grammatical function of their referent (i.e.
ku-/en- vs. e-). This falls out of the scope of the SAP approach, as further subdivisions
internal to the SAP/non-SAP division are not necessarily implied.
Moreover, in the case of SA, there is one more person to include in the frame – fourth person.
I introduced fourth person in §2.2.2.1. highlighting its polysemous use and paying attention to
the fact it does not have a referent in the real world which corresponds to a specialized
grammatical person. Rather, its possible referents are already formalized in SA as other
grammatical persons (e.g. logophoric fourth person is in fact a first person and honorific
fourth person is a second person). Nevertheless, a separate morphological realization for
these referents is present. I argue that the reason for this is again found in pragmatics. Fourth
person seems to have the function of relating the SAP and non-SAP domains by turning SAPs
into non-SAPs and vice versa. Logophoric and honorific fourth person takes a participant and
distances it from the SAP dimension (either for stylistic or politness reasons), making it more
like an external participant to the event (i.e. non-SAP). Conversely, inclusive and unknown-
agent fourth person takes a non-SAP participant and brings it closer to the SAP dimension19.
In this sense fourth person has a deictic function. All these special pragmatic relations need to
be expressed in morphology, and this is what the fourth person is for. Considering its
behavior, it appears that fourth person stands on the borderline between referencing to SAPs
and non-SAPs. Trying to fit SA‘s fourth person either among SAPs or non-SAPs presents a
theoretical issue. When we try to associate fourth person, as a unitary person category of SA
with its polysemous value, to SAPs or non-SAPs, it inevitably violates the constraints that
define either one of these domains. This is exactly because fourth person simultaneously
encodes different entities whose pragmatic centrality to a given conversational context is not
the same (e.g. central first-second persons vs. non-central unknown agent). It is then
troublesome, but also beyond the scope of the present analysis, to try and speculate whether
SA‘s fourth person should qualify as a SAP or non-SAP20.
What is relevant here instead is that, although the SAP/non-SAP distiction proves to be a
feature to which SA morphological marking is indeed sensitive (see §4), the issues outlined
above show that it cannot be taken as the one feature that decides SA‘s saliency. It follows
that a more thorough definition and a better organization of saliency is necessary in order to
analyze the cross-personal morphological difference in agreement markers.
To shape a saliency hierarchy for SA I focus on two characteristics of agreement markers:

formal markedness and overtness. Formal markedness refers to the actual shape of agreement
markers possible for a given grammatical person, and assumes that a more marked version of
agreement marker may be available simultaneously with a less marked version for that same
person. To define which is the more marked version of a given agreement marker I look at
intransitive verbs. Agreement markers affixed on intransitives are argued to refer to the S
argument of the verb and were presented in §2.2. They are summarized here for convenience.
19
I assume inclusive fourth person to have this function even though the participant argued to be brought closer
to the SAP dimension (i.e. a second person) is a SAP itself. This is because the two prototypical SAPs ‗I‘ and
‗you‘ are here fused together as one single participant. The lack of an ontological separation between the two
SAPs is the ground to assume a stronger closeness in saliency. This emulates the case where an actual non-SAP
becomes more similar to a SAP (i.e. the unknown-agent fourth person).
20
This would possibly require the assumption that person categories can be ―split‖ between the two domains of
SAP and non-SAP – something that, at the best of my knowledge, has never been discussed before.
18 Elia Dal Corso
Table 6: Intransitive agreement markers of SA

1SG 1PL 2SG 2PL 3 4
SUBJ ku- -as e- eci- Ø- -an
Leaving aside ‗nominative‘-‗accusative‘ markers featured on transitives, I assume that if a

given person is marked by a certain form of agreement marker on an intransitive, then that
particular form of agreement marker must be the more marked one that is available for that
person. This follows from the fact that, since intransitives have just one single argument
within their subcategorization, that argument is necessarily the more marked one
morphologically. Agreement markers included in Table 6 all represent the more marked
version of agreement marker available for each person, and they are thus assigned the feature
[+MAR]. Subsequently, the alternative agreement marker form that may be present for certain
persons is the less marked one, and it is assigned the feature [–MAR] agreement markers of
SA can be then reorganized as follows.
Table 7: Formal markedness of agreement markers

1SG 1PL 2SG 2PL 3 4
[+MAR] ku- ci-/-as e- eci- Ø- a-/-an
[–MAR] en- -un Ø-21 i-
I treat [+MAR] agreement markers for first person plural and fourth person as functionally the
same, even though they appear formally different. Tamura (1970) and Shibatani (1990) argue
that the two different forms of agreement marker displayed by first person plural and fourth
person in SA most likely have developed from a common origin 22 . This historical
development sets the ground to propose that ci-/-as and a-/-an respectively are allomorphs of
the same affix. Which is the underlying form of this affix is not an issue here; rather I now
have evidence to treat formally different agreement markers as variants of the same [+MAR].
Secondly I look at overtness. I intend overtness in terms of whether a certain person can be
straightforwardly retrievable from agreement marker realization. This is independently from
the fact that this person may be expressed via a [+MAR] or [-MAR] agreement marker. For
instance, a first person plural can be said to be overtly retrievable on the transitive affix eci-
un-, even though its relative agreement marker appears in the [-MAR] form un-. From this
agreement marker, in fact, the involvement of a first person plural referent is unmistakably
clear.
Bearing these two parameters in mind I turn back to transitive affixes. By looking at
agreement marker realization in this environment I expect to draw two saliency indexes for
each person. These indexes are respectively based on formal markedness and overtness. As
21
Why the zero affix is considered both the [+MAR] and [-MAR] variant of third person agreement marker is
explained in §4.2.
22
Tamura (1970: 589) argues that affixes a- and -an may have the same origin as *an, which in turns has
originated from the existence verb an ‗exist‘. She bases this assertion on the fact that SA‘s a-, which is in
complementary distribution with -an, corresponds to the prefix an- in many other dialects, among which the
Rayciska dialect of Sakhalin Ainu (Murasaki, 1976). This is further supported by observations on the prosodic
characteristics borne by these variants, as both an- and a- do not affect accent position on the verb. On the other
hand, Shibatani (1990: 29) notices that in the Ishikari dialect first person plural intransitive suffix -as is also
used on transitives in place of the prefix ci- featured in SA and other Southern Hokkaidō dialects. The author
does not provide any further insight about this fact, and indeed more information would be needed to speculate
an actual common historical development for these two affixes. However, here I use Shibatani‘s preliminary
observation to assume that, seemingly to a-/-an, also ci-/-as are allomorphs of a same affix.
an example, consider the transitive affix a-en-. Fourth person agreement marker is here
overtly retrievable and in the variant [+MAR] – it responds positively to both formal
markedness and overtness. First person singular en-, on the other hand, is overtly retrievable
but in the variant [-MAR]– it responds positively just to overtness. Proceeding like this for all
transitive affixes, I count when they respond positively to the parameters I set. The final
outcome of this process eventually tells how persons are ranked in respect to each other and
consequently how the saliency hierarchy is organized. The only assertion possible since this
stage is that third person must be at the bottom of the hierarchy. Since I am dealing with
morphological marking, it follows that a formally unmarked person must be the least salient
among all. This again fits with the SAP approach presented above, by which non-SAPs are
considered less salient. For this reason third person is not included in the counting.
As it regards portmanteaus featured in Table 5, it has already been mentioned that they are
treated as special cases of agreement marker affixation. This is due to the fact that they
encode pragmatic information regarding two separate referents within one single formal
realization. For the purpose of the analysis at hand, they are always assumed to respond
positively to the parameter of overtness. That is to say, a portmanteau like 1SG>2PL eci-
overtly expresses both first and second person, although two morphologically separate
agreement markers are not present. On the other hand, portmanteaus respond to the parameter
of formal markedness like all other agreement markers, depending on whether the [+MAR]
variant is available for a given person. For instance, again 1SG>2PL eci- responds positively
for formal markedness just for second person, as eci- is in fact a second person [+MAR].
Contrarily, the portmanteau 2SG>1SG en- does not respond positively for any of the two
persons it encodes. This is because en- is a [-MAR].
Table 8 presents the results of counting. The higher the total, the higher the relative person is
ranked on the saliency hierarchy.
Table 8: Saliency in SA
1st person 2nd person 4th person
markedness 4 14 6
overtness 16 16 12
total 20 30 18
Adding the previously excluded third person, the saliency hierarchy to account for
morphological markedness of agreement markers appears as follows.
2nd>1st>4th>3rd
It is worth mentioning how, at this stage, I did not mention semantics as a decisive
component in the definition of SA‘s saliency. This is in contrast with the cross-linguistic
tendency that discusses saliency as tightly linked to semantic-based features such as animacy
(Klaiman, 1993). Nevertheless, also in the case of SA, semantics is indeed relevant to
saliency, under the shape of thematic role of participants. However, rather than being
something that is used to define saliency, semantics works parallel to it eventually deciding
proximality (see §4).
3.3. Conflicting hierarchies

Now that the saliency hierarchy is formalized, I can apply it together with grammatical
functions and thematic roles to speculate SA morphological alignment. In §2.4. I proposed
20 Elia Dal Corso
that agreement markers are linked to a particular slot within the verbal structure depending on
grammatical function and thematic role of their referents. The agreement marker whose
referent grammatical function and thematic role outrank the ones of another referent,
according to the relative hierarchy, appears in slot 1. Following from this I would assume that
this is the case also for saliency – if a referent corresponds to a person ranked higher on the
saliency hierarchy respectfully to another, then the former should appear in slot 1. It is clear
that this presumes all three hierarchies discussed until now to work parallel to each other. Slot
1 should eventually host the morphologically more marked agreement marker between the
two affixed on the verb. This is a felicitous assumption in cases like the transitive affix ku-i-.
Table 9: Verb slot values for ku-i-

slot 1 slot 2
verb
ku- i-
GF SUBJ OBJ
TR agent patient23
saliency 1st 4th
[+MAR] [-MAR]
Here slot 1 hosts an agreement marker whose referent has all values (SUBJ, agent, 1st) that
outrank, on the relative hierarchies, all values borne by the referent expressed with the
agreement marker in slot 2. As expected, agreement marker in slot 1 is marked as [+MAR] and
agreement marker in slot 2 as [-MAR]. However, there are other cases where the same
assumption is indeed infelicitous. This is for instance the case of a-en-.
Table 10: Verb slot values for a-en-

slot 1 slot 2
verb
a- en-
GF SUBJ OBJ
TR agent patient
saliency 4th 1st
[+MAR] [-MAR]
In this case fourth person appears lower than first person on the saliency hierarchy and yet it
figures in slot 1 within the verb structure. The sentence is grammatical despite agreement
markers linearly appear in such an order that is opposite to the one dictated by saliency.
Moreover, it is impossible to invert the agreement markers (i.e. en-a-) in order to fit with the
saliency hierarchy, since this gives an unacceptable result as discussed in §2.4. This behavior
seems to suggest that the validity of the saliency hierarchy holds unless the thematic
hierarchy is not corrupted. That is to say, if a first person referent (even though highly ranked
for saliency) has the thematic role OBJ/ OBJ2, its relative agreement marker cannot appear in
slot 1 because the features it bears are linked to a fixed position within verb structure. One
problem follows naturally. If it is true that thematic role have priority over saliency, why do I
need a saliency hierarchy in the first place? Alternatively, if saliency is indeed relevant, how
should I apply it without conflict with other hierarchies?
23
For the sake of convenience, here and in the following table agent and patient are taken as representative of
the possible thematic role of the referents.
4. The direct-inverse approach

I propose that the working of SA‘s agreement markers is recognized as a case of direct-
inverse alignment. SA deviates in some aspects from what is said to be the prototype of an
inverse language. Nevertheless, these aspects are in line with less common behaviors featured
in a number of inverse systems. This approach is effective to account for the seeming
discrepancies in the application of hierarchies, and it also explains the peculiarity of
portmanteaus.
4.1. Assumptions
Klaiman (1993: 227) defines inverseness as the property of a language to have transitive,
non-reflexive predication specially marked in case a first or second person referent
corresponds to a non-subject role. In his overview on inverse languages he underlines four
main generalities of this language type – sensitivity to argument ontological ranking,
systematicity, directionality and transitivity (Klaiman, 1993: 235). Inverse languages do not
display one common single behavior, rather they can relate differently to what is implied by
the four generalities above. These in fact simply present a prototype, the conformity to which
is strictly language dependent.
Sensitivity to argument ontological ranking is based on ontological saliency. An argument‘s

ontological saliency depends on its referent‘s centrality in relation either to the physical
universe or to the conversational context (Klaiman, 1993: 235). Klaiman says that the
hierarchy defining the ranking of referents is usually described as a natural hierarchy like in
Table 11. This in turn is recognized as an animacy hierarchy, where statuses refer more to
categories of natural discourse than to categories of grammatical person.
Table 11: Animacy hierarchy (Klaiman, 1993: 236)
1st 2nd 3rd proper human animate inanim.

person person person noun noun noun noun
Under systematicity Klaiman highlights the property by which inverseness involves just
transitive predicates or predicates of a certain lexical domain. While transitivity appears to be
a fundamental requirement to have the direct-inverse alternation, the lexical domain of
inverseness is highly language specific. Nonetheless, the scope of inverseness is clearly
defined so that just specific predicates may feature the alternation – in this sense inverseness
is systematic (Klaiman, 1993: 237). Directionality refers to the direction in which referent
interaction expressed by the verb proceeds. Inverseness assumes a distinction between verbs
denoting events in which referent interaction evolves following the arrow in Table 11, and
events in which the interaction evolves in the opposite direction (Klaiman 1993: 239-240).
This distinction is supposed to be grammaticalized. In many languages grammaticalization of
inverseness is obtained via a specialized morpheme. This may be in complementary
distribution either with another specialized morpheme or with a zero morpheme to distinguish
direct and inverse constructions. Finally, transitivity is a formal characteristic that
distinguishes direct-inverse alternation from active-passive or ergative-antipassive
alternations (Klaiman, 1993: 242).
While I do not include transitivity in my analysis of SA, I indeed acknowledge the other three
generalities drawn by Klaiman to define inverseness in this language. In §3.2. I discussed a
saliency hierarchy for SA based on the centrality of referents in respect to the conversational
context. The concept of saliency applied above fits with the one of ontological saliency
22 Elia Dal Corso
assumed in Klaiman‘s framework. There are, however, some remarkable differences. In SA‘s
ontological hierarchy second person outranks first person, where in the hierarchy assumed in
Table 11 the opposite is true. Nevertheless, there are instances of languages (Plains Cree and
other Algonquian languages) that indeed differ from this generic hierarchy by ranking first
and second person like SA (Klaiman, 1993: 236, 239). Another difference (also pointed out in
§3.2.) is that SA seems not to differentiate (at least not morphologically) among all the
statuses, included in the hierarchy in Table 11, that are below third person. I argue this
signifies that in SA animacy is not a substantial feature of saliency. Rather than semantics,
pragmatic relations between referents in discourse play a central role in shaping this concept24.
These discrepancies with the generic model reported by Klaiman do not undermine the
possibility of relating SA‘s saliency to it 25. Inverseness is clearly systematic in SA. Although
there is no particular lexical domain over which inverseness has its scope, it indeed applies
just to transitive predicates. This restriction shows systematicity. SA deviates from the
prototype of inverseness also regarding directionality. This is because grammaticalization of
inverseness does not happen via a specialized morpheme to indicate how the interaction
between participants evolves. The coding of directionality happens in SA through alternation
of formal realizations of agreement markers on the verb (see §4.3). This is also the case of a
small number of inverse languages like Mapudungun, where alternation in shape of person-
encoding affixes compensates for the lack of overt inverse morphology (Klaiman, 1993: 240).
Looking at formalization of agreement markers from the perspective of directionality is a
relevant issue for the analysis at hand, as it solves the alledged incompatibility among
different hierarchies that shows in some cases.
4.2. The proximate-obviative distinction

Inverseness may encompass the proximate-obviative distinction. In non-local domain (i.e.
when non-SAP referents are involved) languages may differentiate between more and less
salient third person referents (Jacques and Antonov, 2014: 303). Saliency of non-SAPs is
normally determined on animacy of referents, as outlined in Table 11. This distinction is
encoded by expanding the system of person categories including a fourth person 26, which
represent third person referents of lesser centrality (Klaiman, 1993: 247). Fourth person is
otherwise called obviative (OBV) as opposed to a more central third person that is defined
proximate (PROX). These labels help illustrate distal relations of referents based on their
ontological status, which is guided by semantics and pragmatics (Jacques and Antonov, 2014:
303). Proximality is usually marked through morphology or syntax, alternatively in different
shapes depending on directionality.
In §4.1. I noticed how SA does not operate any distinction among third person referents on
the morphological level, and how this subsequently affects the definition of saliency. This
happens because in SA semantics has no influence in shaping the concept of saliency. Since
the proximate-obviative dichotomy usually entails a semantic-based distinction in animacy,
SA should not have the requirements necessary to feature it. Nevertheless, the concept of
proximality is applicable in SA too. I propose that in this language proximality is defined by
saliency and indeed semantics, that, under the shape of thematic roles, applies once saliency
24
In §3.2. SA‘s saliency hierarchy was in fact developed based on morphological realization of referents,
assuming exclusively the relevance of pragmatics but no influence from semantics. In this sense the concept of
―saliency‖ in SA differs from its general cross-linguistic definition (as we find it in Klaiman), that subsumes the
relevance of semantic-based features like, for instance, animacy.
25
In fact, Klaiman defines the hierarchy in Table 11 a mere approximation of how ontological saliency is
determined in languages (Klaiman, 1992: 236).
26
This must not be confused with the language-specific definition of ‗fourth person‘ given above for SA.
has been defined. The major difference with Klaiman‘s model is thus that semantics
influences proximality not through saliency, but rather working parallel to it. Moreover,
thematic roles have a wider scope in SA than animacy does in other inverse systems. In fact,
we can hardly speculate on animacy of first and second person referents, as they will rarely
be inanimate. On the contrary, differences in thematic roles are equally valid for all referents,
so that even most salient second persons may be either agents or patients. It follows that in
SA any referent is possibly treated as PROX or OBV according to two factors – saliency and
thematic primacy.
Proximality is outlined in SA via mathematical operation. Persons from second to fourth,

according to their saliency, are assigned a value going from 2 to 0 (with external third persons
being assigned a negative value -1) and a value of either 2 or 0 depending on the thematic
role of their referent27. By applying these values to transitive affixes I can define proximality
for each instance of referent interaction they express. The process highlights cases when the
OBV corresponds to the OBJ referent and others when it corresponds to the SUBJ referent. In
the former case slot 1 hosts the PROX referent agreement marker while in the latter it hosts the
OBV referent agreement marker. As an example, consider the two instances below.
Table 12: Deriving proximality for affixes ci-i- and Ø-eci-

direction of persons saliency thematic total proximality
interaction index index
1PL>4 1st 1 2 3 PROX
4th 0 0 0 OBV
3>2PL 3rd -1 2 1 OBV
2nd 2 0 2 PROX
Table 13 gives an overview of transitive affixes, separating cases where PROX precedes OBV
within the verb structure (white) and cases where the opposite happens (light grey). The
latters also include portmanteaus (see §4.2.1).
Table 13: Proximality of transitive affixes

OBJ
1SG 1PL 2SG 2PL 3 4
1SG - - eci- eci- ku-Ø- ku-i-
1PL - - eci- eci- ci-Ø- ci-i-
2SG en- un- - - e-Ø- e-i-
SUBJ
2PL eci-en- eci-un- - - eci-Ø- eci-i-

3 Ø-en- Ø-un- Ø-e- Ø-eci- Ø-Ø- Ø-i-
4 a-en- a-un- a-e- a-eci- a-Ø- a-i-
From Table 13 we understand that the PROX referent is marked by what was previously
recognized as the [+MAR] variant of agreement marker. Conversely, the OBV referent is
marked by the [-MAR] variant. Thanks to this approach the discrepancy discussed in §3.3. is
cancelled in that PROX, that indicates the more central referent, is systematically more marked
27
Agent/experiencer is assigned the value 2 while goal/beneficiary and patient/theme are assigned the value 0.
There is no higher value assigned to OBJ2 even though it is ranked higher on the thematic hierarchy because it
can never appear as agreement marker within verb structure together with OBJ (see §2.4). For this reason it
suffices to have separate values for SUBJ and OBJs, which nevertheless appears simultaneously as agreement
markers. Furthermore, the 2-0 distinction is here necessary to maintain the same scale for both factors (i.e.
saliency and thematic primacy).
24 Elia Dal Corso
in morphology28. Boxes highlighted in dark grey show special cases where the derivation
gives the same proximality for both referents.
Table 14: Deriving proximality for affixes Ø-un and a-e-

direction of saliency thematic
persons total proximality
interaction index index
3rd -1 2 1 OBV
3>1PL
1st 1 0 1 OBV
4th 0 2 2 PROX
4>2SG
2nd 2 0 2 PROX
While having multiple OBVs on the same verb is attested in other inverse systems, many
languages consider impossible for two PROXs to coexist (Jacques and Antonov, 2014: 307).
Recognizing both referents as PROX in the case of a-e- and a-eci- follows from the fact that
both these agreement markers never indicate an OBV referent in any other case within the
affix paradigm. In this respect SA deviates from most instances of inverseness.
4.2.1. Portmanteaus
Despite their peculiar behavior, SA‘s portmanteaus respond to inverseness like other
transitive affixes. According to Klaiman (1993: 246), some inverse systems encode local
interaction among SAPs with special 1st>2nd and 2nd>1st forms. This is a strategy for
languages to highlight SAP-only interactions and differentiate them from other instances of
interaction. These special forms discussed by Klaiman are found in SA as first-second person
portmanteaus. They are special in that they not only encode functional features of referents
but also pragmatic information about their interaction (§3.1). Moreover, they are transitive
affixes where just one of the referents appears overtly expressed. In all cases this is the OBJ
referent, that thus appears in the agreement marker variant appropriate for this grammatical
function. The SUBJ referent is in turn implied by the affix, and so formally unmarked. As it
regards morphological alignment, I assume that the only referent whose agreement marker is
formally retrievable from portmanteaus‘ realization is indeed the PROX. In these instances
being the PROX does not entail being marked by a [+MAR] agreement marker, like the case of
2SG>1PL un- exemplifies. Rather, proximality simply involves overt markedness, as opposed
to formal unmarkedness of OBVs. Portmanteaus‘ definition of proximality clearly overcomes
saliency and thematic criteria, and thus cannot be predicted by the derivation process applied
above. In light of this, I argue portmanteaus to imply that the OBJ referent (realized as PROX)
actually structurally follows the SUBJ (i.e. OBV) on the verb. This latter, even though encoded
in the portmanteau, is projected as a zero agreement marker on slot 1.
28
It is necessary to assume that third person zero marker indicates both PROX and OBV and so that Ø- is in fact
the [+MAR] and [-MAR] variant of third person agreement marker (see §3.2). This clearly results from the
derivation. It could be speculated that third person PROX was indeed realized differently from OBV at some
stage of the language (i.e. [+MAR] was different from [-MAR]) – overt marking for third person alternative to Ø-
is reported for instance in Sakhalin Ainu (Murasaki, 1976). This distinction might have gone lost later in the
history of the language, leaving both variants of agreement marker as formally the same. The same cannot be
said for second person. Even though, seemingly to third person, only one realization of agreement marker is
available, this corresponds in fact solely to [+MAR]. The derivation shows in fact that second person is never
OBV.
Table 15: Verb structure with portmanteau eci-

slot 1 slot 2
verb
Ø- eci-
GF SUBJ OBJ
TR agent patient
proximality OBV PROX
This way I can look at portmanteaus, in their own peculiarity, as structurally the same as
other transitive affixes.
4.3. Direct and inverse constructions

In §4.1. I introduced how inverseness encompasses directionality. This means that inverse
languages distinguish verbs according to how referent interaction evolves, considering also
the ontological saliency of these referents. Many languages operate this distinction via an
overt, specialized morphosyntactic device, that usually manifests itself in two different forms.
One of these forms systematically appears when, on a transitive verb, the logical subject
outranks in terms of saliency the logical object. Conversely, the other form appears when the
logical subject is outranked in terms of saliency by the logical object (Klaiman, 1993: 240).
The two forms of this morphosyntactic device are then mutually exclusive and discern clearly
direct from inverse constructions. Example (29) (borrowed from Klaiman, 1993) shows two
Plains Cree sentences where direct and inverse morphemes -āw and -ik are indicative of
directionality. Given that in this language first person outranks third person as it regards
ontological saliency, we know that in (29a) first person ‗I‘ must be the logical subject since
the direct morpheme -āw is used. On the other hand the opposite must be true in (29b), where
the inverse morpheme -ik appears to signal that it is the logical object to bear a higher
saliency.
(29) a. Ni-wāpam-āw (nīýa) atim.

1-see-DIRECT (I) dog
‗I see the dog.‘ (Klaiman, 1993: 245)
b. Ni-wāpam-ik (nīýa) atim.

1-see-INVERSE (I) dog
‗The dog sees me.‘ (Klaiman, 1993: 245)
It was argued above that SA does not resort to any overt morphosyntactic device as a way to
specify directionality. In fact, no added morphology or syntax is featured on verbs hosting
transitive affixes, despite we witness clear alternations in referent proximality (see Table 13).
Even more importantly, directionality in SA takes a slightly different scope than it does in the
inverse systems reported by Klaiman.
Firstly, in Klaiman‘s acceptation, ‗logical subject‘ and ‗logical object‘ are recognized
respectively as the ‗doer‘ and the ‗patient‘ of a transitive predicate. Consecutively, these roles
are easily relatable to the prototypical thematic roles of agent and patient. It is inferable then
that Klaiman assumes logical subject to be assigned the highest thematic role available in the
argument structure of the transitive verb, while the logical object takes the available lower
thematic role, as in Bresnan (2001). It follows from this that I can summarize his overview on
directionality saying that this feature is indeed sensitive to thematic roles and ontological
saliency of referents. As for SA, to argue that thematic roles decide directionality means to
automatically acknowledge the inclusion of grammatical functions into the derivation too.
26 Elia Dal Corso
This is due to the one-to-one correspondence of thematic roles and grammatical functions
within transitive verb structure, as discussed in §2.4. The involvement of grammatical
functions in directionality is just indirect, but it eases the formalization of this feature of
inverseness for SA. In fact, I can assert that direct constructions of SA entail that SUBJ> OBJ,
while inverse constructions entail that OBJ>SUBJ. This avoids resorting to labels and
definitions not included in my framework (i.e. ‗logical subject/object‘).
Secondly, in Klaiman‘s model thematic roles apply together with saliency. This indeed
happens in SA too, but one implication follows from this. It has been proved that thematic
roles and saliency are the factors that also define proximality in SA (§4.2). The logical
consequence of this is that, if directionality is sensitive to these same factors, then it must be
sensitive to proximality. That is, for the case at hand, I can better restate my previous
assertion saying that direct constructions entail that PROX>OBV, while inverse constructions
entail that OBV>PROX.
To compensate for the lack of a specialized morphosyntactic device SA resorts to

morphology that is already available. To express directionality this language uses caselike
forms of agreement markers (i.e. [+MAR]/[–MAR]) and organizes transitive verb structure into
slots connected to fixed functional-semantic features (§2.4). Caselike forms indicate referent
proximality, thus clarifying which referent outranks the other. On the other hand, position
within verb structure unarguably illustrate whether this referent is the SUBJ or the OBJ. The
direct and inverse constructions discussed by Klaiman where logical subject outranks logical
object and viceversa are ultimately mirrored in SA respectively as linear PROX-OBV and OBV-
PROX alternation of transitive verb affixes.
Fixed features linked to verb slots are also useful to not misinterpret grammatical functions in
cases of even proximality (see Table 13). In these instances it is impossible to recognize if the
construction displays the direct or the inverse morphology, so it should be virtually
impossible to understand the kind of referent interaction. The same is attested in other
languages (Klaiman, 1993: 246), where a possible even ontological saliency constitutes a
problem for the correct interpretation of the predicate. To overcome this obstacle, these
languages resort to different strategies, among which the use of a fourth person obviative. In
SA this even-proximality problem is overcome in principle and no further strategy is needed.
No misinterpretation is in fact possible because position of agreement markers subsumes all
functional-semantic features necessary for the right comprehension of the verb form.
5. Conclusions
In this article I discussed morphological alignment in SA. The analysis proposes that SA‘s
morphological alignment follows a direct-inverse organization. Crucial for the argumentation
of this proposal were the interaction of grammatical functions, thematic roles and pragmatic
primacy or referents, and the assumption of a fixed internal structure for transitive verbs.
Evidence for the sensibleness of this assumption comes from (un)grammatical instances of
personal agreement use, that show systematic correspondences of verb structure slots to
functional-semantic features of referents. SA deviates in some aspects from prototypical
inverse systems (Klaiman, 1993). In particular, it organizes referent saliency on pragmatic
relations, ranking second persons higher that first persons. It does not set a lexical domain for
inverseness to apply on verbs, rather the domain is restricted to predicate transitivity and it
also redifines the scope of proximality, extending it outside of the non-SAP domain.
This analysis constitutes an attempt to formalize SA‘s morphological alignment. The direct-
inverse approach to alignment is effective in that it provides a unitary organization of
morphological referent marking for this language, that was previously dealt with recurring to
different (and sometimes contrasting) approaches. Inverseness of morphological alignment
may be also useful to solve issues regarding other categories, that are still unexplained or
overlooked. One such case is the restriction in the use of emphatic second person personal
pronouns (see Footnote 10). Nonetheless, some other issues remain unsolved. For instance, it
is not clear why local predication 2PL>1SG/1PL is not expressed via a portmanteau, as it is the
case for other instances of 1st-2nd person interaction. Moreover, having PROX and OBV third
person formalized in the same way (i.e. Ø-) does not quite fit with the behavior of all other
persons. These, in fact, either have separate forms for PROX/OBV or, if not, they are never OBV
(e.g. second persons). While historical reasons may be at the root of these behaviors, further
speculation is needed to properly explain them. This is left for future research on this topic.
From a wider perspective, this analysis may also have an impact on our typological
understanding of inverseness. I showed above that SA‘s inverse system distinguishes itself,
among other things, for how saliency is defined and how the scope of proximality extends
over all kinds of participants. Such behaviors are said to deviate from the alleged prototype of
inverseness, which in turn shapes the typological profile of this category. The ―exceptional‖
case represented by SA may be the cue that these behaviors that are said to fall outside of the
prototype could indeed be cross-linguistically more common than what previously thought. In
light of such new accounts, we would need to refine our approach to inverse systems
acknowledging the relevance of linguistic variation, which would eventually result in an
improvement of the typological description of inverseness.
Abbreviations
0: epenthetic vowel or glide; 1: first person; 2: second person; 3: third person; 4: fourth
person; AM: agreement marker; APPL: applicative; CAUS: causative; COMP: complementizer;
EV.DIR: direct evidential; EXCL: exclusive; FIN: final particle; FUT: future; GF: grammatical
function; INT: interjection; +MAR: more marked variant of an AM; –MAR: less marked variant
of an AM; NMLZ: nominalizer; OBJ: primary object; OBJ2: secondary object; OBV: obviative; PC:
paucal; POSS: possessive; PL: plural; PROX: proximate; SG: singular; SLV: subject of a light
verb construction; SUBJ: subject; v. OBJ: nominalized verb in a light verb construction
References
Bradley, D. 2007. Languages of Mainland South-East Asia. In Miyaoka, O.; Sakiyama, O.;
Krauss M. E. (eds.), The vanishing languages of the Pacific Rim, 301-336. Oxford: Oxford
University Press.
Bugaeva, Anna. 2004. Grammar and folklore texts of the Chitose dialect of Ainu (idiolect of
Ito Oda). ELPR Publication Series A2-045. Suita: Osaka Gakuin University.
Bugaeva, Anna. 2006. Applicatives in Ainu. In Chiba daigaku yūrashia gengo bunka ronshū
9, 185-196.
Bugaeva, Anna. 2012. Southern Hokkaidō Ainu . In Tranter Nicolas (ed.), The languages of
Japan and Korea, 461-499. London/New York: Routledge.
Bresnan, Joan. 2001. Lexical-functional syntax. Oxford, UK: Blackwell Publishers.
Bresnan, Joan & Asudeh, Ash & Toivonen, Ida & Wechsler, Stephen. 2016. Lexical-
functional syntax – second edition. Wiley Blackwell.
Dalrymple, Mary E. 1990. Syntactic Constraints on Anaphoric Binding. Unpublished PhD
Thesis, Stanford, California: Stanford University.
Dixon, R. M. W. 1979. Ergativity. In Language 55:59-138.
28 Elia Dal Corso
Greenberg, Joseph H. 2000–2002. Indo-European and Its Closest Relatives. Stanford:

Stanford University Press.
Grimshaw, J. 1990. Argument Structure. Cambridge: MIT Press.
Hattori, Shirō. 1961. Ainu go Karafuto hōgen no ‗ninshō setsuji‘ ni tsuite (アイヌ語カラフ
ト方言の「人称接辞」について) [About ‗personal affixes‘ of the Karafuto dialect of
Ainu]. In Gengo Kenkyū 39, 1-20.
Jacques, Guillaime; Antonov, Anton. 2014. Direct/inverse systems. In Language and
Linguistics Compass 8/7, 301-318.
Klaiman, M. H. 1992. Inverse languages. In Lingua 88, 227-261. Elsevier Science Publishers.
Manning, Christopher D. 1996. Ergativity: Arguement structure and grammatical relations.
Stanford, California: CSLI Publications
Murasaki, Kyōko. 1976. Karafuto Ainu go: bunpō hen ( カラフトアイヌ語 : 文法編 )
[Karafuto Ainu – grammar]. Tokyo: Kokushokan Kōkai.
Oku, Satoshi. 2008. Tasōgōteki gengo toshite no ainu go: fuhen bunpō riron no shiten kara no
shiron (多総合的言語としてのアイヌ語:普遍文法理論の視点からの試論) [Ainu as a
polysynthetic language: essay from the point of view of the theory of universal grammar].
In Izutsu Katsunobu (ed.), Ainu gogaku to gendai no gengo riron (アイヌ語学と現代の
言語理論) [Ainu studies and the contemporary linguistic theory], 161-200.
Refsing, Kirsten. 1986. The Ainu language: The morphology and syntax of the Shizunai
dialect. Aarhus: Aarhus University Press.
Shafer, R. 1965. Studies in Austroasian II. In Studia Orientalia 30 (5).
Shibatani, Masayoshi. 1990. The languages of Japan, 3-86. Cambridge/New York:
Cambridge University Press.
Street, John C. 1962. Review of N. Poppe ‗Vergleichende Grammatik der altaischen
Sprachen‘, Teil I (1960). In Language 38:92–98.
Takahashi, Yasushige. 2015. Ainu go Tokachi hōgen ni okeru 1 ninshō to hōkatsu ninshō ni
tsuite (アイヌ語十勝方言における１人称と包括人称について) [About first person
and inclusive person present in the Tokachi dialect of Ainu]. In Hoppō gengo kenkyū 5,
205-213.
Tamura, Masashi. 2010. Ainu go Shiranuka hō gen no bunpō kijutsu (アイヌ語白糠方言の
文法記述) [A grammatical description of the Shiranuka dialect of the Ainu language].
Chiba: Chiba University. (Doctoral dissertation.)
Tamura, Suzuko. 1970. Personal affixes in the Saru dialect of Ainu. In Studies in general and
oriental linguistics : presented to Shirō Hattori on the occasion of his sixtieth birthday,
577-611. Tokyo: TEC Company.
Tamura, Suzuko. 1972. Ainu go Saru hōgen no ninshō setsuji no shurui (アイヌ語沙流方言
の人称接辞の種類) [The types of personal affixes of the Saru dialect of Ainu]. In Gengo
Kenkyū 61, 17-39.
Tamura, Suzuko. 1984. Ainu go onsei shiryō 1 (アイヌ語音声資料１) [Audio materials of
the Ainu language 1]. (http://dspace.wul.waseda.ac.jp/dspace/handle/2065) (Accessed
2016-09-26).
Tamura, Suzuko. 1985. Ainu go onsei shiryō 2 (アイヌ語音声資料２) [Audio materials of
the Ainu language 2]. (http://dspace.wul.waseda.ac.jp/dspace/handle/2065) (Accessed
2016-09-26).
Tamura, Suzuko. 2000. The Ainu language. Tokyo: Sanseido Co.
Vovin, Alexander. 1993. A reconstruction of Proto-Ainu. Leiden: E. J. Brill.

Pirka Itak

Uploaded by

Copyright:

Available Formats

Pirka Itak

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Pirka Itak

Uploaded by

Copyright:

Available Formats

SOAS Working Papers in Linguistics Vol.

Morphological alignment in Saru Ainu:

Elia Dal Corso

Keywords: morphology; alignment; Ainu; person; grammatical function; transitivity;

1.2. Aims and structure of the paper

important as it provides a preliminary understanding of the nature of Ainu alignment, the

2. Agreement and alignment

(1) Orano, sir-kunne kor…

(3) A-ekasi […] Ø-soyne.

(4) Ene Ø-iki hi a-Ø-nukar.

(5) Caca Ø-wen sisam icen Ø-Ø-kore6.

Throughout my argumentation I focus specifically on transitive and ditransitive verbs,

2.2. Personal affixes in SA

(6) * Eani su e-Ø-suke wa Ø-Ø-e.

(7) * Kani Ø-arpa.

(8) Kani k-arpa.

2.2.2. Formal realization of agreement markers

(9) a. Kani k-arpa.

b. Ponkurmat eun ku-Ø-ye ka Ø-Ø-ki.

c. A-en-ipe-re ka somo Ø-Ø-ki.

(10) a. Cis-as kor arki-as.

b. Unu ka ona ka ci-Ø-sak

(11) a. E-eraman ruwe?

b. A-e-ko-nu yakun e-Ø-ye […]

c. A-eci-tak kusu […]

(13) a. A-ekasi […] Ø-soyne.

b. A-e-ko-nu yakun e-Ø-ye […]

2.2.2.1. The fourth person

(14) a. Nisapno ar-siknak-an.

(15) Hetak, paye-an wa ipe-an ro.

(16) A-sik-ihi a-Ø-arka-re hawe?

(17) Nupuri a-Ø-nukar.

(18) “Arpa-an kusu ne na” sekor Ø-hawean.

(19) Sine po Ø-Ø-kor unarpe a-ne

2.3. Different alignments

Table 1: Formal realization of personal agreement

Table 2: Three different alignments

2.4. Constraints on argument marking: grammatical-function-oriented approach

(21) Tapne a-i-siknu-re wa a-e-e-ko-yayirayke.

Table 3: Assumed internal structure of the verb

(22) * Kani i-e-e-ko-yayirayke.

(23) A-en-ipe-re ka somo Ø-Ø-ki.

(24) * En-a-ipe-re ka somo Ø-Ø-ki.

(25) Rur takup a-i-kore.

(26) ?Po icen a-Ø-kore.

(27) Ponkurmat eun ku-Ø-ye ka Ø-Ø-ki.

(28) * Tanpe ani ponkurmat ku-Ø-ye ka Ø-Ø-ki.

Table 4: Assumed grammatical function and thematic role hierarchies

pragmatic interaction between two participants is reflected by agreement markers on

Table 5: Transitive agreement markers of SA

2PL eci-en- eci-un- - - eci-Ø- eci-i-

3.2. Speech act participants – the saliency hierarchy

To shape a saliency hierarchy for SA I focus on two characteristics of agreement markers:

Table 6: Intransitive agreement markers of SA

Leaving aside ‗nominative‘-‗accusative‘ markers featured on transitives, I assume that if a

Table 7: Formal markedness of agreement markers

3.3. Conflicting hierarchies

Table 9: Verb slot values for ku-i-

Table 10: Verb slot values for a-en-