Theiling Online    Sitemap    Conlang Mailing List HQ   

Ishtalo grammar sketch

From:Tommaso R. Donnarumma <trd@...>
Date:Saturday, November 18, 2000, 14:58

Here's a quick and dirty grammar sketch of my brand new
language Ishtalo.  I've been cuddling bits and fragments of
this language (silly me!) since the day I watched to (or
should I say listened to) _Dances with wolves_ (the movie
with Kevin Kostner;  the Italian title was like this, but
Italian translations of movie titles are often far away
from the original).

I'm no expert of Indian languages, so I didn't attempt to
recreate one, although a few features of Ishtalo are inspired
by actual Indian languages (the inversive voice, for example;
or the interaction between quality predicates and argument
marking, a trait I took from Chickasaw, together with the
word _chokma,_ "good").  Perhaps someone among you will be
able to tell me if Ishtalo happens to resemble an Indian
language nonetheless...

Comments and questions are welcome.  If any of you has a
strong grasp on general syntax and/or language typology, I'd
especially like to hear your opinion about the syntax of the
core (marking and cross-referencing of verb arguments).  As I
understand the question, I broke a so-called "language
universal" here;  yet, the result doesn't look especially
alien to me...


There are 14 consonants and 4 vowels.  The consonants are:

           1   2   3   4   5
  Labial   p       f   m
  Apical   t   ts  s   n   l
  Dorsal   k   ch  sh      y
  LabVel   q               w

  1. V.less stop;  2. V.less affricate;  3. V.less fricative;
  4. Nasal;  5. Approximant

Among dorsals, /k/ is velar [k];  /ch/ and /sh/ are prepalatal
(I know no ASCII-IPA for them);  y is the palatal glide [j].

The vowels are:

      e   o

/e/ stands for the schwa [@], /o/ is mid back unrounded [V]
(I think this is the ASCII-IPA for this sound, correct?).

Stress is always word-final, except for a few clitics.
Words more than two syllables long have secondary stress on
the first syllable (proclitics aren't counted, and don't
take secondary stress).

When two vowels run together (for affixation or in adjacent
words), vowel deletion may occur, according to the following

1) if one of the involved vowels is /e/, it is always deleted;
2) if the vowels are identical, the first one is deleted (this
may cause the stress to shift);
3) in informal speech, deletion of the weaker vowel may
occur even if the vowels are different.


I assume that this name was chosen by 19th Century grammarians when
they first accounted for the language. :-)  "Articles" really are
bound pronouns (indeed, there are first and second person articles
too), although the so-called "direct articles" do act like articles
when attached to nouns and attributes.

Articles come in two variants, called direct and oblique.  The
direct forms are:

               sing.   plur.
  masculine    she     se
  feminine     ne      me
  neuter       ye      ye
  1st pers.    en      om
  2nd pers.    a       te
  imperson.    che     che

The oblique forms are:

               sing.   plur.
  masculine    wa      qan
  feminine     le      ela
  neuter       i       i
  1st pers.    mi      mi
  2nd pers.    wi      wi

There is no impersonal oblique article.


Nouns inflect for number and case.  Gender is also grammatical,
but it only shows up in concord (through the articles).

There are 4 numbers:

  singular    -0
  singulative -wes (-mes after k)
  plural      prefix reduplication
  partitive   prefix reduplication + -wan (-man after k)

Depending on whether the stem is C-initial or V-initial, prefix
reduplication consists of a copy of the intial CV or VC
respectively.  Several irregular words exist that employ
suppletive stems instead of reduplication.

Of numbers, the so-called singular is really unmarked, and it
is also found in instances where the English translational
equivalent has a plural.  Singulative, OTOH, always entails
that the referent is one in number.  Partitive is a kind of
plural employed when the referents are percieved as separate
parts of a whole.

There are three cases:

  nominative -0
  accusative -ta (optionally -0 for neuter nouns)
  oblique    -ti (-si after k)

With unmarked verbs, nominative is used for S and A.  It is
also used (mostly) for the possessor.  Accusative is used for
the direct object;  oblique for the indirect object and for
the object of adpositions.  More on this below...

When number and case suffixes co-occur, the former are attached
directly to the stem, and the latter to the former.

Most nouns also take a mandatory direct article (that
concords in gender and number with the noun) and an optional
oblique article (for possessor cross-referencing).

Concord in number is straightforward:  singular articles
go with singular and singulative nouns;  plural articles go
with plural and partitive nouns.

Gender mostly (but not exclusively) depends on animacy and
natural gender.  Besides, there are nouns which exsist in
different gender categories.  Take, for example, _*oka_
("young person") and _*ishtalo_ ("Ishtalo").  These naked
words never occur on their own;  instead, we have the
following gender-marked forms:

 [1] sh'oka, masculine:  "the boy"
     n'oka, feminine:  "the girl"
     sh'ishtalo, masculine:  "the Ishtalo man"
     n'ishtalo, feminine:  "the Ishtalo woman"
     y'ishtalo, neuter:  "the Ishtalo language"

There is a further complication.  As seen in the translations
of the examples above, NPs with gender-marked articles are
always seen as determinate in reference.  Indeterminate NPs
take the impersonal article instead:

 [2] ch'oka:  "a boy," or "a girl"
     ch'ishtalo:  "an Ishtalo"

Only a few nouns occur with no article, the most relevant
class of exceptions being that of personal names.

First and second person articles may occasionally replace the
proper article in emphatical contexts, such as:

 [3] nok empoksheta enlawal
     nok      en  =poksheta en  =lawal
     EMPHASIS 1D.S=shaman   1D.S=come
     "I, the shaman, came!"

As said before, the possessor is cross-referenced on the
noun by means of an oblique article, which is interposed
between the direct article and the noun stem.  If there is
an over NP too, it usually takes the nominative case:

 [4] shepoksheta yewatsitsa
     she   =poksheta ye    =wa    =tsitsa
     3D.M.S=shaman   3D.N.S=3O.M.S=spear
     "The shaman's staff" (lit. "the shaman his staff")


Little is to be said about adjectives _per se._  Adjectives
share the morphosyntactic behaviour of both nouns (see
above) and verbs (discussed below).

When inflected as nouns, adjectives can act as stand-alone
NPs or as adjuncts to other nouns or "nominalised"
adjectives.  When inflected as verbs, adjectives act as

For complex forms, the presence of nominal vs. verbal
affixes is enough to disambiguate an adjective, but in the
basic forms only the context can help:

 [5] nechokma
     ne    =chokma
     "She is good" or "The good one (fem.)"

 [6] nechokma n'oka
     ne    =chokma ne    =oka
     3D.F.S=good   3D.F.S=young.person
     "The girl is good" or "The good girl"

When used as attributes, adjectives exhibit concord in
gender (via the attached direct article), number and
case with the head noun;  optionally, possessor concord
(via an attached oblique article) is also found:

 [7] ye(wa)chochokma shepoksheta yew'atatakam
     ye    =wa    =cho.chokma she   =poksheta
     3D.N.S=3O.M.S=PR :good   3D.M.S=shaman
     ye    =wa    =at.atakam
     "The shaman's good remedies"


Verbs (and predicative adjectives) inflect for tense,
voice and argument agreement.  The structure of verb
chains is:

  direct article-oblique article-tense-stem-voice

Tenses are 7:

  present    mi- (mik- in front of vowels, nasals and s)
  near past  ta- (tak- in front of vowels, nasals and s)
  far past   koke-
  imperfect  won-
  negative imperfect  qen-
  future     ton-
  negative future  len-

Note, though, that no verb has seven marked tense forms.
Indeed, state verbs (and adjectives) bear no overt mark
for the imperfect, and action verbs bear no overt mark
for the far past.

As to the usage, present and far past signify an action
(punctual aspect), respectively in the current time and
in the past.  The imperfect and, to a lesser extent,
the the near past signify a process (imperfective and
progressive aspects), respectively ongoing or "just
finished."  The imperfect is also employed for general,
timeless statements and, at times, for the incohative
future ("to be going to").

The negative imperfect and negative future (two other
labels likely from 19th Century grammarians) signify
"never" and "nevermore/never again" respectively.  In
spite of their names, they have little correlation to
the time of the action.

There are 3 marked voices, which contrast with the
default, unmarked voice (traditionally, but improperly,
labelled "active").  They are:

  inversive         -ko (-no after k)
  passive/reflexive -tok
  antipassive       -qi (-eqi after k)

Cross-referencing of verb arguments (via direct and
oblique articles) is perhaps the most complicated
feature of the Ishtalo verbal system.  Several factors
control cross-referencing in Ishtalo:  for transitive
verbs, they are the tense and voice of the verb, as
well as the person of the A and P arguments;  for
intransitive verbs and a bunch of transitive ones as
well, semantics play a major role.

In the active voice, the basic model of cross-referencing
follows an ergative strategy.  The mandatory direct
article encodes S and P.  The oblique article, if present,
encodes the recipient and beneficiary functions.  The A
of transitive verbs is encoded by a mandatory oblique
article suffixed to the last element of the clause:

 [8] shelawal
     she   =lawal
     "He came"

 [9] newaksaswa
     ne    =waksas=wa
     3D.F.S=heal  =3O.M.S
     "He healed her"

[10] yel'oskatawa
     ye    =le    =oskata=wa
     3D.N.S=3O.F.S=give  =3O.M.S
     "He gave it to her"

There are complications.

With most verbs of perception (i.e, when A is an experiencer,
rather than an agent), there is no postfix clitic, but the
A is encoded as if it were a dative complement:

[11] yel'iqim
     ye    =le    =iqim
     "She knows it" (lit. "It is known to her")

Some stative verbs know of such valency alternations as those
illustrated in 12 and 13 below:

[12] nechokma
     ne    =chokma
     "She is good" (see also example 5)

[13] chelechokma
     che=le    =chokma
     "She feels well" (lit. "There's good for her")

Note the usage of the impersonal article in example 13 above
to fill in the direct article slot of the predicate.  The
direct article is mandatory, so that _*lechokma_ would be

One further possibility for predicates like _-chokma_ is
shown below:

[14] chechokmale
     IMP=good  =3O.F.S
     "She did well"

Another element of asymmetry is thrown in by first and second
person articles.  Whenever one of these articles appears in
the direct slot of a transitive verb, it must be understood as
encoding the A, not the P function.  If the first or second
person is P, then one must resort to the inverse voice.
Contrast example 9 above and examples 15 and 16 below:

[15] emwaksas
     en  =waksas
     "I healed (someone)"

[16] emwaksaskowa
     en  =waksas-ko =wa
     1D.S=heal  -INV=3O.M.S
     "He healed me"

Except when a first or second person is A or P, one cannot drop
the suffixed oblique article cross-referencing the A function
-- unless the verb is marked for the passive voice:

[17] newaksastok
     ne    =waksas-tok
     3D.F.S=heal  -PASS
     "She was healed"

When the verb is in the passive voice, the A function must be
left unspecified, with only one possible exception.  If the
suffixed oblique article is co-referential with the prefixed
direct article, the whole contruction is equivalent to a
reflexive in English:

[18] newaksastokle
     ne    =waksas-tok =le
     3D.F.S=heal  -PASS=3O.F.S
     "She healed herself"

If a first or second person is concerned, either the passive
or the inverse voice are allowed.  The western dialects have
developed a different strategy, in this case:  the oblique
pronoun is dropped, and the passive and inverse voice markers
co-occur, in the form -kotok (i.e., you have _emwaksaskotok_
in place of _emwaksastokmi_ for "I healed myself").  In a few
dialects, this compound form has been reanalyzed as a single,
general-purpose reflexive marker.  In these dialects, the
equivalent of example 17 above is _newaksaskotok_, and a
separate passive voice need be recognized.

Finally, the antipassive allows to promote the A function so
that it is encoded with the direct article for third persons
too.  The P function is not deleted, and it is
cross-referenced as if it where a dative argument to the
verb.  If a recipient or beneficiary is present, it cannot
be cross-referenced within the verb chain.  Below, the
antipassive equivalents of examples 9 and 10 are given:

[19] shelewaksasqi
     she   =le    =waksas-qi
     3D.M.S=3I.F.S=heal  -AP
     "He possibly healed her"

[20] sh'ioskataqi
     she   =i     =oskata-qi
     3D.M.S=3I.N.S=give  -AP
     "Perhaps he gave it"

As seen in the translations of the examples above, the
antipassive voice entails possibility, uncertainty or
indirect knowledge.  Antipassive forms are often accompanied
by adverbs that clarify the nature of the doubt.  Another
adverb that is often found together with the antipassive
voice is _yoktelo_ ("now, soon, at once"):  this is another
means to express inchohative future.

Antipassive agreement is also triggered by the future tense
(but not the negative future).  In the future tense, no
voice morpheme may appear on the verb:

[21] sheletomwaksas
     she   =le    =ton-waksas
     "He will heal her"

In the antipassive voice, first and second person articles
show no peculiar behaviour.


Whatever the complexities of verb agreement, the morphological
marking of NPs serving as verb arguments is straightforward.

In the default voice, argument marking always follows a
nominative-accusative strategy, whereby S and A are in the
nominative case and P is in the accusative case.  The only,
marginal exception is provided by neuter nouns:  if the
subject is animate and definite and the object is neuter,
then it is optionally possible to drop the accusative mark on
the object.  The recipient function is encoded in the oblique

The following examples extend examples 8 to 13 above (remember
that the suffixed oblique article for A cross-referencing
attaches to the last element of the clause, not the verb):

[22] shelawal shepoksheta
     she   =lawal she   =poksheta
     3D.M.S=come  3D.M.S=shaman
     "The shaman came"

[23] newaksas n'okata shepokshetawa
     ne    =waksas ne    =oka  -ta  she   =poksheta=wa
     3D.F.S=heal   3D.F.S=young-ACC 3D.M.S=shaman  =3O.M.S
     "The shaman healed the girl"

[24] yel'oskata y'atakan(ta) n'okati shepokshetawa
     ye    =le    =oskata ye    =atakam  -ta
     3D.N.S=3O.F.S=give   3D.N.S=medicine-ACC
     ne=    oka  -ti  she   =poksheta=wa
     3D.F.S=young-OBL 3D.M.S=shaman  =3O.M.S
     "The shaman gave the medicine to the girl"

[25] yel'iqim n'oka
     ye    =le    =iqim ne    =oka
     3D.N.S=3O.F.S=know 3D.F.S=young.person
     "The girl knows it"

[26] nechokma n'oka
     "The girl is good" (cf. example 6)

[27] chelechokma n'oka
     che=le    =chokma ne    =oka
     ID =3O.F.S=good   3D.F.S=young.person
     "The girl feels well"

In the passive voice, the marking is as expected:  i.e.,
the A function is dropped and the P function is encoded
in the nominative case.  The examples below parallel
number 17 and 18 above:

[28] newaksastok n'oka
     ne    =waksas-tok  ne    =oka
     3D.F.S=heal  -PASS 3D.F.S=young.person
     "The girl was healed"

[29] newaksastok n'okale
     ne    =waksas-tok ne     =oka         =le
     3D.F.S=heal  -PASS 3D.F.S=young.person=3O.F.S
     "The girl healed herself"

In the antipassive voice, the P takes the oblique rather
than the accusative case:

[30] shelewaksasqi n'okati shepoksheta
     she   =le    =waksas-qi ne    =oka         -ti
     3D.M.S=3I.F.S=heal  -AP 3D.F.S=young.person-OBL
     she   =poksheta
     "The shaman possibly healed the girl"


The main syntactic constraint regarding word order is for
the verb to be in clause-initial position.

For the rest, order of constituents is rather free.  In
the most unmarked construction, the direct object is
adjacent to the verb and the subject is at the end of
the clause, the indirect object occurring with equal
frequency before and after the direct object.

At phrasal level, the order is even freer.  Indeed, it is
not uncommon to find that modifiers and possessors are
displaced from their nominal head, with many constituents
intervening in between.


The main source of non-initial predicates is topicalization,
which involves fronting and mandatory morphological marking
(via the particle _non_).  The topic also loses any case
marking it might have:

[31] non n'oka newaksas shepokshetawa
     non   ne    =oka   ne    =waksas she   =poksheta=wa
     TOPIC 3D.F.S=young 3D.F.S=heal   3D.M.S=shaman  =3O.M.S
     "(As of) the girl, the shaman healed her."

There are constraints on what may be topicalised:  namely,
only the nominative and accusative functions, plus the
possessor of the nominative and a few adverbs of time and
place, may undergo topicalisation.


The default focus position is clause-final.  Another
common position for the focus is after the main verb. In
both positions, the focus is optionally marked with the
particles _ak_ and _nok._

In a few cases of very strong emphasis, when the two
common focus positions are not available (typically, when
the clause is only made of the verb and the focused NP),
it is allowable to move the focus in front of the verb.
In this position, morphological marking (usually via
_nok_) is mandatory.  See example 3.


That's all with it.  Many details are still missing -- for
example, modality and embedded clauses --, but I think I've
layed out the foundations of the language.  I'll keep
working on Ishtalo! :-)