Chapter 5 Below the Tigris: The Neo-Aramaic Dialects of Ṭur ʿAbdin and Mlaḥsó

In: Ergativity and Other Alignment Types in Neo-Aramaic
Author:
Paul M. Noorlander
Search for other papers by Paul M. Noorlander in
Current site
Google Scholar
PubMed
Close
Open Access

The Neo-Aramaic dialects of Ṭur ʿAbdin (‘Ṭuroyo’) and Mlaḥsó constitute a separate subgroup in Southeast Turkey called Central Neo-Aramaic. In terms of alignment, dialects of Ṭur ʿAbdin are typologically similar to the Southeastern Trans-Zab Jewish varieties of NENA. The now extinct dialect of Mlaḥsó, in turn, is similar to Christian NENA dialects in SE Turkey such as Borb-Ruma (Bohtan) as well as Jewish dialects of Iranian Azerbaijan, such as J. Urmi. Ṭur ʿAbdin Neo-Aramaic dialects are much less diverse than their Northeastern Neo-Aramaic kin, but there are some notable differences among them. We will contrast them with the Trans-Zab Jewish dialects of NENA and conclude with a comparison of Mlaḥsó with Ṭur ʿAbdin and NENA dialects in general.

A major difference between Central and Northeastern Neo-Aramaic is found in the verbal stems and derivations, since Central Neo-Aramaic is characterized by an extensive system of verbal derivations. Each stem derivation (IIV) has its own mediopassive pendant (IMIVM), e.g. stem IM fṣoḥ-o ‘She is happy’. In addition, stem I verbs also include a special ‘perfective’ pattern CaCiC, i.e. qaṭil-, e.g. damix-o ‘She slept’, which will be represented by its historical origin *qaṭṭil- for *CaCCiC, e.g. damixo < *dammiḵå, to avoid confusing with the NENA qaṭəl-base, which corresponds to Central qoṭəl-. The Neo-Aramaic dialects of Ṭur ʿAbdin and Mlaḥsó differ greatly in the usage of these bases.

Hemmauer and Waltisberg (2006) and, recently in more detail, Waltisberg (2016) argue that the preterit in Ṭuroyo is essentially tripartite. The distinction in verbal stems between intransitive and transitive clauses plays a key role in their argumentation. A more nuanced view will be offered here: ergative alignment is indeed manifested in Neo-Aramaic dialects of Ṭur ʿAbdin, at least in terms of pro-indexes and, to some extent, also prepositional marking. The latter is more distinctly ergative than what is found in NENA. Recently, Coghill (2016, 84–90) and Khan (2017, 894–895) also briefly treated alignment in Ṭuroyo and Mlaḥsó in comparison with NENA, and their observations are comparable to mine.

5.1 Morphosyntactic Traits of Central Neo-Aramaic

5.1.1 Stems Disengaged: *məqṭol- vs. *qoṭəl-

Central Neo-Aramaic is noteworthy in comparison to NENA for having mediopassive stem derivations. The system is represented for the dialects of Ṭur ʿAbdin in Table 32 above.

Table 32

The Ṭuroyo stem derivations

Active

Mediopassive

ipfv

pfv

ipfv

Ia:

qoəl-

qṭil-

qṭil-

-qṭol-

Ib:

doməx-

damix-

II:

m-zabən-

m-zabən-

m-zabən-

mi-zabən-

III:

m-a-dməx-

m-a-dməx-

m-t-a-dməx-

mi-t-a-dməx-

IV:

m-farqəʿ-

m-farqəʿ-

m-farqəʿ-

mi-farqəʿ-

Notes: dmx ‘sleep’, zbn ‘sell’, frqʿ ‘burst’. Stems in shaded cells take L-suffixes to express a. Data based on Jastrow (1985).

‘Imperfective’ (ipfv) bases corresponding to qoṭəl- are given on the left and right and ‘perfective’ (pfv) bases corresponding to qṭil- in the middle of the table. This arrangement serves to show the convergence between the two voice systems in the perfective past. The active and mediopassive are differentiated only by inflectional base in the ‘imperfective’, qoṭəl- vs. məqṭol-. The inflectional bases for the ‘perfective’ are generally the same for both active and mediopassive with the following exceptions:

  • verbs belonging to what is called class ‘Ib’ of stem I, which distinctively has active CaCiC- and only possibly CCiC- in the mediopassive;

  • verbs having a mediopassive of stem III with a typical -t-infix (mtaCCaC-).

Stem I verbs may be divided into two distinct classes: (Ia) takes CCiC- and (Ib) follows CaCiC-, which are, respectively, qṭil- and *qaṭṭil-,1 but the ‘imperfective’ base of both of these is CoCəC, i.e. qoṭəl-. Otherwise, what applies to stem Ia verbs generally also applies to derivational stems. The shaded area indicates forms that take agent (or subject) indexes of the L-set. The rest takes subject (and/or agent) indexes of the E-set.

Overall, voice is marked differently in the verbal morphology of the ‘perfective’ and ‘imperfective’. The ‘imperfective’ anticausative pendants consist of distinct mediopassive stem derivations. As we will see, the ‘perfective’, by contrast, shows valency alternations similar to what is observed for Southeastern Trans-Zab Jewish dialects of NENA.

Mlaḥsó distinguishes approximately the same stem derivations as Ṭuroyo. The stem derivations are represented in Table 33 above. The shaded area indicates, where the L-suffixes are employed as subject and agent indexes. Interestingly, we find more or less the distribution opposite of Ṭuroyo (see Jastrow 1996).

Table 33

The Mlaḥsó stem derivations

Active

Mediopassive

prs

pret

prs

perf

ipfv

pfv

ipfv

I:

qail-

qoel-

qṭil-

me-qṭel-

me-qṭel-

II:

zaben-

zaben-

m-zaben-

m-zaben-

III:

m-a-dmex-

m-a-dmex-

m-t-a-šoġ-

m-t-a-šoġ-

IV:

qarveʿ-

qarveʿ-

Notes: zbn ‘sell’, dmx ‘sleep’, šyġ ‘wash’, qrvʿ ‘chase away’. Stems in gray shade take L-suffixes. Stem IIIM is only attested for weak verbs. Source: Data from Jastrow (1994, 33–34).

As Table 33 illustrates, mediopassive stem derivations, such as meqṭel- ‘be killed’ and mtašoġ- ‘be washed’, correspond to the ‘imperfective’ (ipfv) in both the preterit and present. This is unlike Ṭuroyo, where, apart from stem III, the mediopassive merges with the active in the ‘perfective’, e.g. qṭil- for the preterit of both qoṭəl- ‘kill’ and məqṭəl- ‘be killed’, which will be further discussed in § 5.3.2.

5.1.2 Stems Entangled: Phonological Reduction

Vowel reduction leads to slight differences in the inflection of the ‘imperfective’ base qoṭəl-. First of all, as a rule, ə is lost before a CV-sequence and turns to a before a closed syllable, so that ˚doməx- ‘sleep’ with -no of the 1ms. becomes ˚domax-no ‘IM sleep’. Furthermore, rural dialects, such as Miden, have long i [i:] and o [o:] in verbal forms, these are shortened and neutralized to ə [ɪ] or ŭ [u] in urban dialects in and around Midyat in an unstressed open syllable directly before the stressed syllable. Compare the following verbal forms:2

(1)

‘IM sleep’

‘IF went to sleep’

˚domax-no

damix-ono (rural)

˚dŭmax-no

daməx-ono (urban)

Miden, in turn, has almost completely merged the short vowel ŭ with ə. The differences in vowel reduction lead to the following paradigms in comparison to Mlaḥsó:

(2)

Miden

Midyat

Mlaḥsó

1ms.

‘IM go to sleep’

domax

-no

dŭmax

-no

domex

-no

1fs.

‘IF go to sleep’

dəmx

-ono

dŭmx

-an

domx

-ono

3ms.

‘He goes to sleep’

doməx

-∅

doməx

-∅

doméx

-∅

Consonant clusters with ə can be readjusted in the Midyat dialect, whereby ‘perfective’ nšəq-o-le ‘He kissed her’ alternates with nəšq-o-le against Miden iq-o-le (Ritter 1990, 63).

Phonological phenomena such as the ə-deletion rule and role reference inversion can yield ambiguous forms, whereby the ‘perfective’ and ‘imperfective’ bases are identical (Jastrow 1985, 144–145). While ə becomes a before suffixes with an initial consonant, it is normally deleted in an open syllable. Since the subjunctive is the unmarked ‘imperfective’ form, this leads to ambiguity for stem II and IV verbs, for example II ḥlq ‘throw’ in (3). Similarly, a transitive form like mḥalq-i-le (stem II) can be either subjunctive or preterit. This resembles the situation in the NENA dialect C. Artun (Hertevin, SE Turkey; Jastrow 1988, 38) where the ‘perfective’ and ‘imperfective’ bases are identical for such verbal derivations.

subjunctive

preterit

(3)

mḥalaq-no

‘that I throw’

‘I was thrown’

mḥalq-i

‘that they throw’

‘They were thrown’

mḥalq-i-le

‘that they throw him’

‘He threw them’

Moreover, the difference between the two inflectional bases is neutralized for final-/y/ verbs belonging to stem Ia in rural dialects like Miden, which merge ŭ with ə. This may be illustrated by a comparison to NENA:

Ṭuroyo (Miden)

NENA (C. Artun)

(4)

subjunctive

∅-əzy-o-li (< *ḥŭzy- < *ḥozy-)

∅-ḥazy-a-li

preterit

əzy-o-li

ḥəzy-a-li

The ambiguity does not apply when the verb does not take both agent and object indexes, but only subject indexes. In that case, the choice of affixes distinguishes subjunctive from preterit, for example in the intransitive verb hlx ‘walk’ belonging to stem II:

(5)

subjunctive

∅-mhalax-no

‘that IM walk’

preterit

mhalax-li

‘I walked’

5.1.3 Unmarked and Prepositional Pronouns

Table 34 below provides an overview of the unmarked and dative independent pronouns in Central Neo-Aramaic dialect. The Midyat prepositional pronouns are based on the unmarked independent ones rather than pronominal suffixes, as elsewhere in Neo-Aramaic, i.e. l- ‘to’ + ŭno ‘I’, l- ‘to’ + huwe ‘he’, in analogy to demonstrative pronouns, e.g. l-ano from l- ‘to’ + hano ‘this’, l-ani from l- ‘to’ + hani ‘these’. In the second person, we also find the forms l-ŭxat for the masculine singular and l-ŭxatu for the plural (Ritter 1990, 3), which appear to be contaminations of expected l-ox and l-oxu, and the independent pronouns hat and hatu. In Mlaḥso, the first person plural is eləna throughout, and the 3ms. dative is different from Miden, inflected with the suffix -av and distinct from the L-suffix -le.

Table 34

Independent pronouns in Central Neo-Aramaic

Miden

Midyat

Mlaḥsó

1sg.

ono

el-i

ŭno

l-ŭno

onó

l-i, el-í

1pl.

aḥna

el-an

aḥna

l-aḥna

el-əna

el-əna

2ms.

hat

el-ŭx

hat

l-ox

hat

el-óx

2fs.

hat

el-ax

hat

l-ex

hat

el-éx

2pl.

hatu

al-xu

hatu

l-oxu

hatun

el-ekun

3ms.

hiye

el-e

huwe

l-uwe

hiye

el-áv

3fs.

hiya

el-a

hiya

l-iya

hiya

el-á

3pl.

hənnək

al-le

hənne

l-ənne

hiyen

el-én

Data based on Jastrow (1992, 1994) and Ritter (1990)

5.1.4 Differential Object Marking and Word Order

Generally speaking, object nps follow the verb in Ṭuroyo, but precede the verb in Mlaḥsó, compare:

(6) Ṭuroyo (Miden, Ritter 1967–1971, 71/51)

[v]

[p]

qṭəl-le

tloṯ-

arbʿó-

sowe

killPFV-a:3pl

three

four

old:pl

‘They killed three, four elderly people.’

(7) Mlaḥsó (Jastrow 1994, 126.123)

[p]

[v]

ḥamšó

predé

qṭi-len

five

people:pl

killPFV-a:3pl

‘They killed five people.’

The Neo-Aramaic dialects of Mlaḥsó and Ṭur ʿAbdin may use differential prepositional marking of objects, although it is largely optional in the latter, as a result of which definite object nps generally remain unmarked. Contrast (8) with (9) below. The nominal marking can thus remain largely neutral in Ṭuroyo.

(8) Ṭuroyo (ʿIwardo)

a.

[v]

[p]

d-qŭṭl-ina

ád-dew-ani

sbjv-killIPFV-a:1pl

the-wolf-dem:pl

‘… so that we may kill these wolves.’ (Ritter 1967–1971, 91/24)

b.

[v]

[p]

qṭi-lan

í-kŭrf-ayḏ-an

killPFV-a:3pl

the-snake:fs-lk-our

‘We killed our snake.’ (ibid. 92/50)

(9) Mlaḥsó (Jastrow 1994, 148.18)

a.

[v]

[domp]

ḥoze-

l-a-roʿye

seeIPFV-a:3ms

dom-the-shepherd:pl

‘He sees the shepherds.’ (Jastrow 1994, 88.93)

b.

[domp]

[v]

l-a-gavre

qṭi-len

dom-the-man:pl

killPFV-a3pl

‘They killed the men.’ (Jastrow 1994, 77.1)

Ṭuroyo speakers from the village of Rayite as represented in texts 95–113 of Ritter (1967–1971) constitute a notable exception, which prepositionally mark definite object nps; both patients and themes.3 This holds for both qoṭəl- and qṭil-, for example:

(10) Ṭuroyo (Rayite)

a.

[v-a]

[domp]

g-ḥoze-

l-í-dăvăre

fut-seeIPFV-a:3ms

dom-the-breach:fs

‘He will find (lit. see) the breach (in the wall).’ (Ritter 1967–1971, 107/90)

b.

ḥze-li

l-ú-tadbir

diḏ-ŭx

seePFV-a:1sg

dom-the-measure:ms

lk-your:ms

‘I saw your measurements.’ (ibid. 104/44)

5.2 The Neo-Aramaic Dialects of Ṭur ʿAbdin

While comparable to South-Eastern Trans-Zab Jewish dialects of NENA, such as Sanandaj and Saqiz in western Iran (see Chapter 3, especially § 3.3.1.1.), the Neo-Aramaic dialects of Ṭur ʿAbdin are typologically more straightforward. The ergative and non-ergative alignment types are complementary in Ṭuroyo, each confined to the third or non-third person category. The neat combination of ergative verbal person marking and ergative prepositional marking is only found in this subgroup.

5.2.1 Patient-Related Factors

5.2.1.1 Monotransitive Person Marking: Ergative and Horziontal

Morphological ergative person marking is confined to third in the inflection of qṭil- in Ṭuroyo and alternates with horizontal person marking for the first/second persons.

As in the majority of NENA dialects, the E-set of person markers groups s and p for third person markers only, for example:

(1) Ergative alignment of pro-indexes (third person only)

a.

damix-o

(intransitive)

sleepPFV-s:3fs

She went to sleep.’

b.

ḥəzy-o-le

(transitive)

seePFV-p:3fs-a:3ms

‘He saw her.’ (lit. Him saw she)

Generally, since differential object indexing does not occur as frequently as in NENA, person markers only serve as cross-indexes for s and a. When the E-set marks p, it is strictly speaking a pro-index:4

(2) Miden

a.

ftəḥ-le

ʿayn-e

(no cross-indexing of definite p)

openPFV-a:3ms

eye-his

‘He opened his eyes.’ (Ritter 1967–1971, 81/18)

b.

ʿayne

d-ú-babo

ftiḥ-i

(cross-indexing of definite s)

eyes

of-the-father

openPFV-s:3pl

‘Father’s eyes opened.’ (ibid., 57/237)

c.

ṭəm-le

ʿayn-e

u

ftiḥ

-i

-le

(pro-index p)

closePFV-a:3ms

eye-his

and

openPFV

-p:3pl

-a:3ms

‘He closed his eyes and opened them (again).’ (73/400)

d.

ftiḥ-i

(pro-index s)

openPFV-s:3pl

They opened.’

Ergativity is thereby confined to pro-indexes in Ṭuroyo, as illustrated for the labile verb ftḥ ‘open’ in (2) above. The trigger potential for agreement is higher for s and a (a=sp): they always trigger cross-indexing.

Cross-indexing of p is possible, but rare: a form without object indexes like ftəḥ-le ‘He opened’ in (2a) is generally preferred at least in the Miden dialect (Jastrow 1985, 137). Nevertheless, differential cross-indexing of definite full nominal objects is occasionally also found,5 for example:

(3)

[a]

[v-p-a]

[p]

hăma

Aḷoho

sim-o-le

mujiza

but

God:ms

doPFV-p:3fs-a:3ms

miracle:fs

haṯe

(diff. indexing of p)

dem:fs

‘But God performed this miracle.’ (Miden, Talay 2004, 128.335)

First/second person markers, however, pattern horizontally (sa=p). The L-series groups both a and p, as exemplified and schematized below.

(4) Horizontal alignment for non-third person arguments

a.

damix-ono

(intransitive)

sleepPFV-s:1fs

‘IF went to sleep.’

b.

ḥzé-li-lax

(transitive)

seePFV-a:1sg-p:2fs

‘I saw youFS.’ (lit. Me saw your)

The object affix always follows the agent affix in stacking of L-suffixes. Since the order and role designation of the two L-suffixes is fixed, there is no ambiguity.

The two alignment types are complementary, both are restricted by a person category as third vs. first/second person. Table 35 above illustrates the distinct strategies in object marking conditioned by person.6

Table 35

Person-conditioned alignment in Ṭuroyo (Miden)

s = E-set

p = E-set

daməx-

He slept’

grəš--la

‘She

pulled

him

damix-o

She slept’

griš-o-la

‘She

her

[3rd]

damix-i

They slept’

griš-i-la

‘She

them

s = E-set

p = L-set

damix-ət

‘YouMS slept’

grə́š-li-lŭx

‘IF

pulled

youMS

[1st/2nd]

damix-at

‘YouFS slept’

grə́š-li-lax

‘IM

youFS

damix-utu

‘YouPL slept’

grə́š-lan-lalxu

‘We

youPL

daməx-no

‘IMS slept’

grə́š-lax-li

‘YouFS

meM

damix-ono

‘IFS slept’

grə́š-lŭx-li

‘YouMS

meF

damix-ina

‘We slept’

grə́š-xŭl-lan

‘Youpl

us

In actual transitive clauses, the coding of the agent is stable and does not vary depending on person, e.g. griš-o-lan ‘We pulled her’, grə́š-la-lan ‘She pulled us’ (Jastrow 1985, 38–139).

From a comparative perspective, horizontal verbal person marking is rare in the NENA subgroup,7 although stacking of L-suffixes does occur, particularly in SE Turkey (see § 4.4.3.). In the Trans-Zab Jewish bundle, the ʾəll-series is preferred for the first and second person, e.g. xze-li ʾəll-ax ~ xzé-li-llax ‘I saw youFS’, and in several Christian dialects of NENA, mainly in Turkey, the L-suffixes can be used instead, e.g. C. Jinnet ḥzí-le-laḥ ‘They saw youFS’.

NENA constructions conditioned by the person of p are somewhat different in distribution from Ṭuroyo. Third person markers are generally available in both alignment patterns, but the first and second only in the non-ergative pattern. Nevertheless, in Ṭuroyo, forms like grə́š-la-le cannot be used to denote ‘She pulled him’.

Some intransitive verbs are compatible with a-like coding in Ṭuroyo, such as nwəḥ-le ‘ItM barked’, depending on semantic and/or morphological factors. Conversely, some two-argument state verbs such as šmʿ ‘hear’ are incompatible with a-like coding and have transitive coding exactly like qoṭəl-, e.g. šamiʿ-o-le ‘She heard him’ (see § 5.2.3.).

Hemmauer and Waltisberg (2006) argue that the preterit is only superficially ergative and that a tripartite system points to an underlying accusative pattern similar to qoṭəl-. Recently, Waltisberg (2016, 20, 176) denied any manifestation of ergativity in Ṭuroyo and emphasizes that the alignment is essentially tripartite.

First of all, our approach does not differentiate between deep and superficial alignment and no alignment pattern is subsumed under another. It does differentiate agreement in terms of morphological marking and trigger potential, which Hemmauer and Waltisberg seem to conflate. They rightly show that agent and (especially) subject agreement are ultimately primary to the verbal system.

As expected for Aramaic, in terms of trigger potential, the indexing of full nps is indeed accusative in Ṭuroyo, in most varieties even similarly to Classical Arabic. When full nominals are considered, subject nps and agent nps each take morphologically distinct sets (mainly E-set vs. L-set), while object nps generally do not trigger overt indexing (∅) and, if they do, this is conditioned by definiteness. Since s and a are still distinguished morphologically, this is a tripartite type of verbal person marking (asp).

Nevertheless, ergative verbal person marking may still incidentally be observed for definite nps, where definite objects do trigger the same overt morphology as subjects. Such overt coding of p is taken as starting point for the basic characterization of an alignment type in my approach.8 And when we consider the person category and its morphological marking, the verbal person marking is unmistakably ergative for the third person and horizontal for the first and second person.

Coghill (2016, 85–87) reaches a similar conclusion. In her model, however, the fixed v-a-p order of the two L-suffixes leads her to characterize the first/second person as tripartite.9 Affix order is only considered in our approach when prefixes are contrasted with suffixes (see § 2.3.2.3.). In his critique of Coghill (2016), Waltisberg (2016, 20, 176) points to the important fact that the inflectional base of certain intransitive verbs (CaCiC- as in damix-o ‘She fell asleep’) differs from that of transitive verbs (CCiC- as in ftiḥ-o-la ‘She opened itF’) in the perfective past, arguing that one cannot consider this system therefore to be ergative. Nevertheless, there is no reason why the same stem would be required nor why a different stem would hinder the identification of the same set of person markers for s and p functions of arguments in one and the same clause type. Irrespective of the shape of the stem, it is the same E-set that expresses the properties of s and p; the inflectional base, though it correlates with transitivity, does not express the syntactic roles of arguments. It does confirm that verbal person marking alignment in Neo-Aramaic is primarily structurally dependent on the type of inflectional base (qṭil-), and not perfective aspect per se (see § 5.2.3.).

In essence, the observations for Ṭuroyo are rather similar to those for Southeastern Trans-Zab Jewish dialects of NENA. All else being equal, s and a always trigger indexing irrespective of person reference in both qṭil- and qoṭəl-. Object indexes come in two sets depending on person: the E-set for third person aligning ergatively with p and the L-set for the other persons aligning horizontally with a. Moreover, the two sets of object indexes (E-set vs. L-set) are complementary in Ṭuroyo, while in NENA third person object indexes generally occurs in both the E-set and an alternative strategy, of which there are several.

5.2.1.2 Ditransitive Person Marking

Unlike NENA, a second L-suffix cannot express third person patients, so that forms like **ftə́ḥ-la-le for ‘She opened itM (i.e. the door’) are disallowed. This restriction is germane to their function as indicators of the patient (Jastrow 1985, 137–138) while L-suffixes are favored to express recipients across the system.

When third person markers do feature in stacked L-suffixes, the second L-suffix expresses the pronominal recipient or beneficiary, for example:

(5)

[v-a

-r]

[t]

ftíḥ-ḥan

-ne

ú-tarʿo

openPFV-a:3pl

-r:3ms

the-door:ms

‘They opened the door for him.’ (Miden, Ritter 1967–1971: 73/371)

When attached to qṭil-, the second L-suffix always expresses r when it is third person. For non-third person markers, however, p and r are identical. Compare:

[v-a-p]

[v-a-p/r]

(6)

a.

grə́š-le-la

b.

grə́š-le-li

‘He pulled for her

‘He pulled (for) me

(lit. Him pulled her)

(lit. Him pulled me)

Ṭuroyo usually does not allow more than one object affix on the verb in ditransitive constructions. Two object suffixes rarely occur, but if they do, the E-suffix expresses t, the last L-suffix expresses r (see Ritter 1990, 75), for example:

(7)

[v

-t

-a

-r]

húw

-i

-le

-lalle

givePFV

-3pl

-3ms

-3pl

‘He gave them to them.’ (Miden, Ritter 1967–1971, 73/371)

It is much more common, however, for t to be marked by a special enclitic series (the same as the ‘copula’) when both t and r are pronominal. This is confined to third person reference: -yo for the singular and -ne for the plural, for example:

(8)

a.

[v

-a

-r

-t]

-li

-lalle

-yo

givePFV

-a:1sg

-r:3pl

-t:3ms

‘I gave them itM (the milk).’ (Miden, Ritter 1967–1971, 75/375)

b.

maḥát

-la

-lalle

-ne

putPFV

-a:3fs

-r:3pl

-t:3pl

‘She prepared them for them.’ (Miden, ibid. 115/110)

Only third person pronouns, therefore, exhibit distinct sets of dependent person markers for each grammatical function (p, t, r) while these are not distinguished for their first and second person counterparts.

The use of L-suffixes to mark recipients and similar affectees also occurs in NENA dialects, particularly in NW Iraq, such as Jewish ʿAmedia and Dohok (see also Noorlander 2021), but this is not dependent on person (see § 4.4.3.2.). Neutralization of all functions, including p and t, for all persons, occurs in NENA dialects nearby Ṭur ʿAbdin as well as Jewish varieties of NW Iran (see § 4.4.3.1.). In C. Artun (Hertevin; SE Turkey), however, the situation is exactly the reverse of Ṭuroyo: stacking of L-suffixes (ḥze-le-le) is confined to third person agents (see § 4.4.4.).

Another person-role constraint is found in the inflection of object indexes attached to the imperative in Ṭuroyo (Jastrow 1985, 140–143, 1992, 128–130). A special set, namely 3ms. -e, 3fs. -a and 3pl. -ene, marks third person ps, e.g. graš-e ‘Pull itM!’ (**graš-le). Themes are marked in this same way when r is a prepositional full nominal. To illustrate:

(9)

[v-t: pro]

[datr: fnp]

haw-e

l-Baṣuṣ

give:impv-t:3ms

r:dat-prn

‘Give itM to Baṣuṣ!’ (Miden, Ritter 1967–1971, 115/283)

This is exactly like the E-suffixes for qṭil-, for example:

(10)

[v-t: pro]

[datr: fnp]

hiw-o-le

l-Šalliṭa

givePFV-t:3fs-a:3ms

r:dat-prn

‘He gave itF to Šalliṭa.’ (Miden, Ritter 1967–1971, 86/27)

The L-suffixes always express r, such as -le in the following example, when the theme is a full nominal:

(11)

[v-r: pro]

[t: fnp]

haw-le

məlyun

give:impv-r:3ms

million

‘Give him a million!’ (Miden, Talay 2004, 114.266)

When both arguments are pronominal, the object suffix expresses t, while r is expressed independently as a prepositional argument (the el-series), for example:

[v-t]

[r]

(12)

-le-lan

el-e / al-xu

‘He gave us to him/to youPL.’

Thus, the L-suffixes express all objects for non-third person markers, synthesizing p, t and r. First/second person indexes therefore follow the object coding of qoṭəl- in the entire verbal system. This is a striking difference from NENA dialects, where the E-set may equally synthesize p, t and r. That is, forms like mir-a-li could be interpreted as ‘I told her (r)’ just like mir-ət-ti- (<*mir-ət-li) is interpreted as ‘I told youMS’ (see § 4.4.3.2.). By contrast, Ṭuroyo distinguishes between mir-o-li ‘I told itF (t)’ and mə́ḷ-ḷi-la (< mə́r-li-la) ‘I told youMS (r)’. In clusters of dependent pronouns, however, third person themes are distinguished from the patient by means of the enclitic set, e.g. -le-la-yo ‘He gave itM to her’ as opposed to grəš--le ‘He pulled itM’.10

In conclusion, r is marked in the same way for all persons throughout the verbal system, while it is third person pronouns that are marked differently depending on their syntactic role (p, t and/or r). It is furthermore noteworthy that only a and r are marked by the same L-set (a=r) in the third persons, while the L-set can be used to mark all functions for the first/second persons (p, t and r) just as it does for qoṭəl-.

5.2.2 Agent-Related Factors: Optional Flagging

Turning to independent pronouns and full nominals, speakers of rural and urban dialects in Ṭur ʿAbdin can choose to mark such arguments prepositionally in a function, and sometimes also p function.

5.2.2.1 Optional Prepositional Marking of Objects and Agents

Generally, a definite object np remains unmarked in Ṭuroyo dialects. Prepositional marking and cross-indexing are occasionally observed for qoṭəl- only, for example:

(13) Miden (Ritter 1967–1971, 81/49)

[v

-a

-p]

[domp]

k-ŭḏʿ

-i

-le

l-ú-zlām

ind-knowIPFV

-a:3pl

-p:3ms

dom-the-man:ms

‘They know the man.’

Across Ṭuroyo dialects, the same preposition can be used to mark the agent only in qṭil-, for example:11

(14) ʿIwardo (Ritter 1967–1971: 33/34.37, 55/25)

a.

[v-a]

[erga]

[p]

ḥze-le

l-ú-Ṭay-awo

ú-med-ano

seePFV-a:3ms

dat-the-Muslim-dem:ms

the-thing:ms-dem:ms

That Muslim saw this thing.’

The same holds for independent pronouns, including demonstratives, of the el-series, for example:

b.

lo

el-i

u

lo

l-ú-ḥawr-ayḏi

lə́-ḥze-lan

ú-mede

d-əmm-at

neg

dat-1sg

and

neg

dat-the-friend:ms-my

neg-seePFV-a:1pl

the-thing

subr-sayIPFV-a:2sg

‘Neither I nor my friend found the thing youSG speak of.’

One may contrast this with s, which does not get marked as such:

c.

[v-s]

[s]

aṯi-

ú-Malke

aʿm-a

(no overt marking of s)

comePFV-s:3ms

the-prn:m

with-3fs

‘Malke came with her.’

The subject np (Malke) of a basic intransitive verb like ʾṯy ‘come’ in (14b) is indexed, but not marked prepositionally. A similar np in a-function can be marked both prepositionally (l-) and verbally (L-suffixes), while p is zero-marked, as shown in (14a). This is a type of optional ergative nominal marking, especially when the agent is focal.12

Only intransitive verbs that take L-suffixes are compatible with such flagging (Waltisberg 2016, 176). For example, the subject of the stem III verb hlx ‘walk’:

(15)

l-Nari

malax-le

(flagging of sa)

dat- prn:ms

walkPFV-s:3ms

Nari walked.’ (Rayite; Ritter 1967–1971, 96/229)

The optional prepositional marking is always conjoined with indexing of a. The prepositional argument functions as a and is syntactically equivalent to the subject in, for instance, cross-clausal anaphoric deletion, e.g.

(16)

l-anii

hjəm

-me

aʿl-ayye,

u

(∅)i

falit-i

aʿl-

ayye

b-ax-xanejər

dat-dem:pl

attackPFV

-a:3pl

on-them

and

(s)

fallPFV-s:3pl

on-

3pl

with-the-dagger:pl

These attacked them and (∅) fell on them with daggers.’ (Iwardo; Ritter 1967–1971, 33/32)

There is no construction in Ṭuroyo equivalent to those in NENA dialects where the agent is prepositional but not overtly indexed, e.g.

xze-

l-naša

‘ItM was seen by somebody’ / ‘Somebody saw itM’.

A construction that would potentially parallel this is exemplified below. The verb ḥzy is intransitive denoting a spontaneous event (‘appear’) and the prepositional noun expresses a recipient-like argument rather than the agent.

(17) Flagging, but no indexing (Midyat; Ritter 1967–1971, 11/107)

[s]

[v-s]

[obl]

Malaxo

Gábriyel

b-ú-ḥŭlmo

ḥze-

l-Mor

Šəmʕon

angel:ms

prn

in-the-dream:ms

seePFV-s:3ms

dat-hon

prn

‘The angel Gabriel appeared to Lord Simon in the dream.’

The optional a-marking thus does not appear to occur in passives, where instead the preposition me(n)- is used. However, it is possible that in fronting a topical patient this type of ergative marking can be included in constructions that are seemingly equivalent to a passive, such as (18) below.

(18) Kfarze (Lahdo 2013, 210.14)

ú-mšiḥoy-ayḏox

ṣluw-we

l-áy-yəḏoye

the-anointed:ms-your:ms

crucifyPFV-a:3pl

dat-the-Jews

‘But your Christ was killed by the Jews.’

The ergative prepositional marking of nps may combine with the ergative indexing of nps, as illustrated in the following examples. The word order often seems to be p-v-a. The full nominal aḥḥeṭani ‘this wheat’ and demonstrative pronoun haṯe ‘this’ are indexed by the E-set, like s, and the agent np is marked differently both nominally and verbally.

(19) ʿIwardo (Ritter 1967–1971, 55/11, 46/25)

a.

[p]

[v-p-a]

[erga]

áḥ-ḥeṭ-ani

xil-i-le

l-ú-moro

the-wheat:pl-dem:pl

eatPFV-p:3pl-a:3ms

dat-the-master:ms

‘The owner ate this wheat.’

b.

haṯe

sim-o-le

l-ú-Qanda

dem:fs

doPFV-p:3fs-a:3ms

dat-the-prn

‘(It was) Qanda (who) did this.’

Ṭuroyo varieties such as the dialect of the village Rayite which employ differential prepositional marking of p, may also use this dative agent construction, as shown in (20a–b). The resulting prepositional marking alignment pattern is horizontal (sa=p).

(20) Rayite (Ritter 1967–1971, 107/85.116)

a.

[v-a]

[datp]

madʿal-le

l-ʿAli

aʿm-e

(p flagged)

takePFV-s:3ms

dom-prn:m

with-3ms

‘He (i.e. the son) took along Ali.’

b.

[data]

[v-a]

[datp]

l-ʿAli

grəš-le

l-ú-sayfo

(a and p flagged)

dat- prn:ms

pullPFV-a:3ms

dom-the-sword:ms

Ali drew the sword.’

Waltisberg (2016, 177–180) points out that salient and highly referential arguments are marked by l-, and Diem (2012, 45) that prepositional as tend to favor post-verbal position. Recently, Kuzin (2018) explored the correlation between word order as well as argument referential properties and the presence or absence of prepositional marking of a in a corpus study of Ritter’s (1971) texts. His results indicate that there is a weak correlation with word order, but not with animacy or definiteness per se. Thus, while word order cannot predict the occurrence of optional a-marking, the post-verbal (v-a) order seems to be favored for prepositional as, but preverbal (a-v) order for unmarked as.

Optional overt marking of unexpected agents is well-known in typology.13 In the Australian Aboriginal language, Warrwa, for example, ergative case-marking is optional and not predictable, but manifests itself by means of distinct coding depending on focus and the degree of agentivity (McGregor 2006). Zero-marking of a is what defocuses it, signaling an expected actor with little impact. Overt flagging of a is diffused across an ordinary ergative marker and a focal ergative marking. The former adds no significance to a, while the latter adds salience to a, highlighting it as being counter to expectation and having an exceptionally powerful impact on p.

Prepositional marking in Ṭuroyo seems to be parallel to this. It generally increases the agent focus, enhancing the sense of responsibility and unexpectedness, as illustrated in the following examples.

(21)

urban

rural

(Prym-Socin 1888, 133.9–10)

(Ritter 1967–1971, 59/41)

a.

xlo l-ŭno qṭi-li bab-ox

b.

lo el-i qṭi-li í-ḥŭrmayḏŭx

‘Do you think I killed yourMS

‘(It was) not I (who) killed

killed yourMF wife.’

yourMS father?’

It also typically occurs in the situation where two agents are contrasted (i.e. X on the hand, Y on the other hand), for example:

(22) l-uwe mamṭé-le-lan u-l-ano qṭi-∅-le xanejər

That one brought us (here) attacked them, but this one slayed him.’ (ibid. 33/32)

Independent pronominalization of a focal constituent is common for Neo-Aramaic in general. Depending on the dialect, this may be accompanied by additional person marking on the verb (see § 2.3.1.2.).

While the indexing is obligatory in Ṭuroyo, the prepositional marking is optional. The unmarked counterpart of full nominals and independent pronouns is also available, but it is not specific to the a role. The unmarked independent pronouns may also express focus and freely alternate with a prepositional counterpart. Compare, for example, el-ŭx and hat below.

(23) Pronominal a (ʿIwardo, Ritter 1967–1971, 48/60.48)

a.

[(erg→)a]

[v-a]

ma

lo

el-ŭx

məḷ-∅-ḷŭx?

qay

ġbin-at!

q

neg

dat-2ms

sayPFV-t:3ms-a:2ms

why

be.angry-s:2sg

‘But didn’t youMS yourself say so? Why! Are youSG angry?’

b.

ma

lo

hat

məḷ-∅-ḷŭx

ma

ġbin-at?

q

neg

youMS

sayPFV-t:3ms-a:2ms

q

be.angry-s:2sg

‘Did youMS not say so? Are you angry?’

Unmarked full nps may equally alternate with a prepositional pendant in a-function, compare l-babi and babi in the following examples:

(24) Full nominal a (Miden, Ritter 1967–1971, 73/106)

a.

[erga]

[v-a]

l-bab-i

lo-moláf-le-li

dat-father:ms-my

neg-teachPFV-a:3ms-r:1sg

My father did not teach me (to do it that way).’

b.

[v-a]

[a]

haṯe

ono

hawxa

moláf-le-li

bab-i

dem:fs

I

thus

teachPFV-a:3ms-r:1sg

father:ms-my

‘This (is) how my father taught me (to do it).’

It is difficult to pinpoint an exact semantic difference between the absence and presence of the prepositional marking of the agent. An increase in agentivity seems to be more readily implied by the use of the preposition l-, but this is not always apparent. Nevertheless, it would be mistaken to consider the differential flagging of a in the ‘perfective’ functionally equivalent to the differential flagging of p in the ‘imperfective’.

The distinct patterns in the interaction of indexing and flagging observed thus far are recapitulated in Table 36 below. p aligns with s ergatively mainly in terms of indexing. Flagging may target either a or p, as well as both a and p. The unmarked instances of both agent and object nps are most common, while prepositional marking of both is least common: either ergative or accusative flagging, then, appears to be favored. The combination of both indexing and flagging of salient objects in qṭil- does not appear to occur. This would require further study to be ruled out completely.

Ṭur ʿAbdin dialects, therefore, concur with the cross-linguistic tendency to avoid the combination of ergative person markers with accusative nominal marking (Dixon 1979, 92; 1994, 95; see § 3.3.2.). Moreover, even from a language-internal perspective, it is likely that there is an additional morphological factor for why this combination is avoided. The dative prepositional marking by means of the preposition (e)l- correlates with the L-suffixes in marking the same role. This can be observed not only in the differential marking of p in qoṭəl- in (13) above, but also in the following constructions.

5.2.2.2 On l- and L-suffixes Elsewhere: Agent-Recipient Parallels

Prepositional objects are typically marked by (e)l- independently of the verb or, if a dependent person marker, as an L-suffix attached to the verb. Certain verbs, such as qry ‘call (for)’ and ʾmr ‘say, tell’ always takes such a complement. Indexing and prepositional marking may also be combined:

Table 36

Indexing and prepositional marking of a and p

s

E-set

maṭy-o í-kalo

‘The bride arrived.’

L-set

mhalax-la í-kalo

‘The bride walked.’

l-np

mhalax-la l-í-kalo*

a

p

L-set

nšəq-le ú-ḥaṯno í-kalo

‘The groom kissed the bride.’

l-np

nšəq-le l-ú-ḥaṯno í-kalo

l-np

nšəq-le ú-ḥaṯno l-í-kalo

E-sfx

nšiq-o-le ú-ḥaṯno í-kalo

l-np

nšiq-o-le l-ú-ḥaṯno í-kalo

l-np

nšəq-le l-ú-ḥaṯno l-í-kalo

These sentences serve as hypothetical examples of the relevant pattern. *sA verbs only.

(25) Rayite (Ritter 1967–1971, 99/6, 96/207)

a.

qre-le l-ú-abro navoyo

‘He called for his middle son.’

b.

qré-le-le

‘He called for him.’

c.

qré-le-le l-ú-malko

‘He called for the king.’

Similarly, recipients marked by l- can trigger additional suffixes, such as the addressee of the verb ʾmr ‘say’:

[a]

[v-a-r]

[datr]

(26)

u-zlām

mə́ḷ-ḷe-le

l-u-zʿuro

‘The man said to the little one.’ (Miden, ibid. 76/65)

The coding of focalized agents as such is identical to the differential marking of recipient nps in qṭil-. Thus, a construction involving a prepositional full nominal recipient such as mər-ḷe l-np based on ʾmr ‘say’ is ambiguous to the role of the dative argument, it can either denote r ‘He said to np’ or anp said’, for example:

(27) ʿIwardo (Ritter 1967–1971, 35/35, 40)

r: mər-le l-ú-mŭstašārayḏe

‘He said to his counselor

a: mər-le l-ú-ʿmiro

The emir said’

The two are not mutually exclusive and can even co-occur, for example:

(28)

a.

[(dat→)a]

[v-a-r]

[datr]

ú-šŭlṭono

mə́ḷ-ḷe-le

l-ú-wazir-ayḏe

the-overlord:ms

sayPFV-a:3ms-r:3ms

dat-the-vizier-his

The sultan said to his vizier.’ (Anḥəl, ibid. 64/2)

b.

l-ú-šŭlṭono

mə́ḷ-ḷe-le

l-ú-waziro

dat-the-overlord:ms

sayPFV-a:3ms-r:3ms

dat-the-vizier

The sultan said to the vizier.’ (Anḥəl, ibid. 64/12)

The key difference is that the flagging of a is optional, while the r of a ditransitive verb like ʾmr ‘say’ is always marked prepositionally. Moreover, l-marked recipients are not necessarily additionally indexed by L-suffixes, while the l-marked agent is always marked as such.

There is a stronger parallel with the l-marked possessor in predicative possession based on the existential marker kət- or the suppletive verb hwy ‘be’. The possessum/possessee remains unmarked. In example (29), prepositional marking of the possessor is variable, but, here, the L-suffix always indexes the possessor. Thus, L-suffixes marking a cannot be omitted in Ṭuroyo, similarly to the L-suffixes expressing the possessor.

(29) Predicative possessor (ʿIwardo, Ritter 1967–1971, 58/3, 57/12)

a.

[pssr]

[exist-pssr]

[pssm]

ú-zlām-ano

kə́t-way-le

arbʿi

kaloṯe

the-man-dem:ms

exst-pst-3ms

forty

daughter-in-law:pl

This man had forty daughters-in-law.’

b.

[pssm]

[exst-pssr]

[datpssr]

ma

kət-le

l-ú-malk-ano

q

exst-3ms

dat-the-king-dem:ms

‘What does the king have?’

The combination of L-suffixes and l-marking is readily found elsewhere within the language (see Noorlander 2021), except for p in qṭil-. It is only in qṭil-, then, that differential prepositional marking of p by means of l- cannot be combined with L-suffixes. Presumably, the combination is morphosyntactically linked with the use of a morphologically or at least historically related similar set of dependent person markers. The preposition (e)l- links an—often focal a—in the perfective past with the same marking typical of the predicative possessor, recipients and other prepositional arguments.

5.2.3 Voice and Other Verb-Related Factors: *qṭil- vs. *qaṭṭil-

The verbal person marking is part of a larger system of stem derivations. When we examine the valency alternations in Neo-Aramaic dialects of Ṭur ʿAbdin, there are close parallels with NENA varieties, especially the Trans-Zab Jewish varieties, as well as major differences among them, primarily in verbal stems. Ṭuroyo makes a two-dimensional split in the inflection of intransitive verbs: one with respect to the type of subject indexes (E-set/L-set) and another with respect to the morphological class for stem I verbs (qṭil-/*qaṭṭil-).

5.2.3.1 Ergative and ‘Neuter’ Verbs

Virtually all transitive verbs of stem Ia can be ambivalent in a causative/inchoative alternation in Ṭuroyo (cf. Ritter 1990, 124). We can, however, only speak of lability (i.e. no change in basic morphology), for the ‘perfective’. The mediopassive generally expresses the inchoative of the equivalent causative. Consider, for example, the verb ftḥ ‘open’ in the following alternation. The inchoative marks the subject like an object, while the causative takes an agent index from the L-set.

(30) Labile alternation

a.

[s]

[v-s]

ʿayne

d-ú-babo

ftiḥ-i

(inchoative, no agent)

eye:pl

lk-the-father

openPFV-s:3pl

‘Father’s eyes opened.’ (lit. they opened) (Miden; Ritter 1967–1971, 81/18)

b.

[v-a]

[p]

ftəḥ-le

ʿayn-e

(causative, specified agent)

openPFV-a:3ms

eye-his

‘He opened his eyes.’ (lit. Him opened) (ibid. 57/237)

We can compare this to SE Trans-Zab Jewish varieties of NENA such as J. Sulemaniyya. The verbs pqy in NENA and frqʿ IV in Ṭuroyo pattern alike:

(31)

Ṭuroyo (Miden)

J. Sulemaniyya

(Jastrow 1985, 112)

(NE Iraq; Khan 2004a, 297)

tr.

mfarqaʿ-le

pqe-le

(a = L-set)

He burst (sth.)’

‘id.’

intr.

mfarqʿ-o

pəqy-a

(s = E-set)

ItF (was) burst’

‘id.’

A cause may be expressed overtly by the preposition me ‘from’, as illustrated in (32). me may also simply express the cause in other intransitive constructions, for example:

(32) Ṭuroyo (Qamishli, NE Syria)

a.

u-tarʿo

ftəḥ-∅

me

hawa

qwiṯo

the-door:ms

openPFV-s:3ms

from

wind:fs

strong:fs

‘The door opened because of (or: was opened by) a strong wind.’

b.

i-dawmo

qayiṯ-o

b-i-nuro

m-u-barqo

the-tree:fs

start.burnPFV-s:3ms

with-the-fire:fs

from-the-lightening:ms

‘The tree caught fire because of the lightning.’

Anticausatives are known to be compatible with causal phrases, but the implication is not as strong as that of the passive prototype.

What we have seen thus far is similar to NENA, but there are also noteworthy differences. First of all, the inchoative/causative alternation is not labile in the valency alternation in the ‘imperfective’. A distinct anticausative stem is used, i.e. məqṭol-, for the intransitive valence pattern, while transitive valence patterns are morphologically distinguished only by choice of argument coding in the ‘perfective’:

(33) Valency alternations: məqṭol- vs. qoṭəl-

perfective

imperfective

tr.

ftəḥ-la

˚fətḥ-o

(causative)

‘She opened (sth.)’

‘She opens (sth.)’

intr.

ftiḥ-o

˚məftoḥ-o

(inchoative)

‘ItF (was) opened’

‘ItF opens, is being opened’

The ‘imperfective’ therefore maintains a voice distinction at the level of inflectional base only, whereas the ‘perfective’ does so at the level of verbal person marking.

Some stem I verbs such as fṣḥ ‘be(come) glad’ are middle only (IM), e.g. fṣiḥ-∅ ‘He was/became glad’. They evince no labile alternation (e.g. **fṣəḥ-le ‘He gladdened’). This also parallels SE Trans-Zab Jewish varieties of NENA, although NENA has no corresponding separate mediopassive base in the ‘imperfective’. Compare the cognate verb pṣx in Jewish Sanandaj (W Iran; Khan 2009, 523):

(34) Emotive response middle in Ṭuroyo and NENA

Ṭuroyo

J. Sanandaj

pfv

fṣiḥ-∅

pṣix-∅

‘He rejoiced’

‘id.’

ipfv

˚məfṣəḥ-∅

(≠ qoṭəl-)

păṣəx-∅

(= qaṭəl-)

‘He rejoices’

‘id.’

When we consider object omission, Ṭuroyo does not show distinctions in the marking of the agent. A verb like šty ‘drink’ can freely occur without the object and the coding of the agent does not alter:

(35) Miden

a.

[v-a]

[p]

štalle

i-qaḥw-aṯṯe

drinkPFV:a:3pl

the-coffee:fs-dem:fs

‘They drank the coffee.’ (Ritter 1967–1971, 115/63)

b.

štalle

(∅)

maqraṭ-ṭe

drinkPFV:a:3pl

III:breakfastPFV-s:3pl

‘They drank (and) had breakfast.’ (73/113)

An antipassive, where a becomes s and p becomes oblique, is not found in Ṭuroyo. Nevertheless, there is a class of verbs that evinces some features of antipassive typology. Stem I verbs come in two subclasses depending on their pattern for the ‘perfective’: (Ia) qṭil- and (Ib) *qaṭṭil-. The verbs of (Ib) the *qaṭṭil-class are mainly intransitive and mostly do not occur in labile alternations. Jastrow (1985, 71) refers to them as “neutrische Verben” (‘neuter verbs’), i.e. belonging to neither the passive nor active voice. The E-set is used as subject indexes. The transitive valence pattern is derived, for example the verb tym ‘finish’ in the following alternation:

(36) Causative alternation

a.

[s]

[v-s]

í-măsăl-ayḏ-an

tayim-o

(inchoative, stem Ib)

the-story:fs-lk-our

finishPFV-s:3fs

‘Our thing is finished.’ (Ritter 1967–1971, 115/149)

b.

[v-a]

[p]

[a]

matəm-le

ú-šŭġl-ayḏ-e

ú-malko

(causative, stem III)

finishPFV-a:3ms

the-business:ms-lk-his

the-king:ms

‘The king finished his business.’ (77/21)

The causative counterparts mainly belong to either stem III or II as shown for a few verbs, given below.

inchoative (Ib)

causative

daməx-∅

‘sleep, fall asleep’

III madmax-le

‘put to sleep’

bašel-∅

‘cook’ (intr.)

II mbaše-le

‘cook’ (tr.)

barəm-

‘turn’ (intr.)

Ia brəm-le

‘turn’ (tr.)

mali-∅

‘be(come) full’

Ia mle-le

‘fill’

(rare)

Only rarely do verbs alternate between stem Ia and stem Ib, but it is possible, such as Ib mali-∅ ‘be(come) full’ (intr.) and Ia mle-le (tr.) ‘fill’ below. However, the ‘perfective’ of IM (məftoḥ- : ftiḥ-∅) merges with Ib in final-y verbs in some villages,14 e.g. Ritter (1990, 378)

imperfective

perfective

məmle

:

mle

məmle

:

mali

Stem Ib verbs generally express only one argument. They can be combined with a prepositional complement, e.g. krx ‘search’ in

karəx-

aʕla

‘He looked for her.’

Generally, ‘neuter’ verbs do not display a distinction in the coding of transitivity. Unlike in NENA, the verb ylf ‘learn’ shows no difference for the transitive and intransitive valence patterns:

(37) Intransitive and transitive CaCiC-‘perfective’

a.

yaləf-no

ṭowo

(intransitive)

learnPFV-1ms

good:ms

‘I learnt well.’ (Iwardo, Ritter 1967–1971, 37/11)

b.

yaləf-

ʿələm

(transitive)

learnPFV-3ms

science

‘He learnt science.’ (Midyat, ibid. 24/257)

Thus, a few stem Ib verbs can be morphosyntactically transitive, even though lexically speaking they are not transitive (see further below). These transitive ‘neuter’ verbs may take clausal complements, full nominal objects and object indexes from the L-set, which is indistinct from the transitive coding in the ‘imperfective’, for example:

(38) Miden

a.

í-naqla

d-i-qriṯo

šamiʿ-i

ú-xabr-ano

the-moment:fs

subr-the-village:fs

hearPFV-1pl

the-word:ms-dem:ms

‘When the people of the village heard the news.’ (Ritter 1967–1971, 71/16)

b.

čirok̭-áṯe-ze

ʿəsrí-kore

šamiʿ-ína-la

story-dem:fs-add

twenty-times

hearPFV-1pl-3fs

‘This story, too, we (already) heard itF twenty times.’ (115/14)

These transitive verbs typically express two-argument experiencer predicates, such as šaməʿ-∅ ‘hear’ and aḏəʿ-∅ ‘know’ (Jastrow 1985, 71; Furman and Loesov 2014). Semantically speaking, they are not primary transitive verbs, for the agent-like argument is not an actual instance of a in the same sense as primary verbs like qṭl ‘kill’ or twr ‘break’ but rather an experiencer. These two-argument experiencer states are not compatible with the a-like coding of transitive verbs, and one could compare this to the antipassive in some languages where ergative morphosyntax predominates. The Polynesian language Samoan, for instance, employs ergative alignment for primary transitive verbs. Some stative verbs, especially two-argument experiencer verbs, such as ‘love’, always occur in the antipassive, while action verbs never do (Comrie 1978, 373). Here in the Neo-Aramaic dialects of Ṭur ʿAbdin, we observe that two-argument states take identical transitive coding across the qoṭəl-/qṭil-split:

perfective

imperfective

yalif-o-le

‘She learned it.’

˚yəlf-o-le

‘She is learning it.’

yalif-o

‘She learned (quickly).’

˚yəlf-o

‘She is learning (quickly).’

In the unlikely case when roots occur in two classes, there is often a subtle semantic shift accompanying detransitivization, e.g. Ritter (1990, 51)

intr.

tr.

Ib

qaṭəʿ-

‘He crossed’

Ia

qṭəʿ-le

‘He cut through’

Ib

naṭər-

‘He waited’

Ia

nṭər-le

‘He guarded’

The fact that these verbs belong to the largely intransitive neuter class could be because they do not (as strongly) imply an effect on a patient-like argument. This is what makes them similar to the antipassive, a voice that generally serves to decrease the effect implicature (Cooreman 1994).

5.2.3.2 Personal and Impersonal Alternations

Contrasting with NENA, the agentless qṭil-form is also compatible with two-argument state verbs and even intransitive verbs (Ritter 1990, 124). Verbs denoting a state, such as ḥzy ‘see’ in (39) below, may occur in a labile alternation. The intransitive valence pattern has a spontaneous reading.

(39) Labile alternation for ḥzy ‘see’ (Midyat)

a.

[s]

[v-s]

[obl]

Malaxo

Gábriyel

b-ú-ḥŭlmo

ḥze-

l-Mor

Šəmʿon

angel:ms

prn

in-the-dream:ms

seePFV-s:3ms

dat-hon

prn

‘The angel Gabriel appeared to Lord Simon in a dream.’ (Ritter 1967–1971, 11/107)

b.

[v-a]

[p]

ḥze-li

b-ḥŭlm-i

ḥa

k-omər-

seePFV-a:1sg

in-dream:ms-my

one:ms

ind-sayIPFV-a:3ms

‘I saw in my dream one saying…’ (23/9)

Transitive verbs belonging to stem Ib that take a *qaṭṭil-base in the ‘perfective’ can have a mediopassive counterpart (IM), even though there is no corresponding form in stem Ib. The mediopassive (IM) iḏiʿ-∅ ‘be renowned’ is for example reported to exist for (Ib) aḏəʿ-∅ ‘know’ for the verb ʾdʿ ‘know’ (Jastrow 1985, 76; Ritter 1990, 727), but there is no (Ia) verb **iḏiʿ-le ‘know’.

The mediopassive may also be used to express an impersonal passive. A causal origin is more strongly implied for a verb, such as qṭl ‘kill’ in (40b) below, but the verb does not cross-index the object and takes the unmarked 3ms. form. Thus, the perfective is characterized by a type of impersonal labile alternation.

(40) Miden

a.

qṭəl-le

tloṯo

gawre

mən-aye

killPFV-a:3pl

three

man:mpl

from-3pl

‘They killed three men of them.’ (Ritter 1967–1971, 85/22)

lit. ‘Them killed three men of them.’

b.

qṭil

tloṯo

gawre

me-Midən

killPFV

three

man:mpl

from-Miden

‘Three men from Miden were killed.’ (85/12)

lit. ‘Three men from Miden killed.’

A major difference between NENA and Ṭuroyo is that even intransitive verbs may be impersonalized (Ritter 1990, 124 ff.). This is illustrated for dmx ‘sleep’ and rʿm ‘come together’ below. The verb dmx ‘sleep’ belongs to stem Ib (*qaṭṭil-) and the impersonalization involves a change in inflectional base and absence of indexing.

(41) Impersonalization in Ṭuroyo (Ritter 1990, 124–125, 127)

a.

daməx-∅

‘He fell asleep.’

(*qaṭṭil-, intransitive)

b.

dmix(-∅) larwal

‘People (lit. ItM) slept there.’15

(qṭil-, impersonal)

An ambitransitive verb, such as rʿm ‘come together’, however, is labile in both personal and impersonal contexts:

c.

rʿim-i

ám-maye

(qṭil-, inchoative)

gatherPFV-s:3pl

the-water:pl

‘The water (pl.) accumulated.’

d.

rʿim(-∅)

harke

šəšwone

(qṭil-, impersonal)

gatherPFV

here

ant:pl

‘(ItM) swarmed here (with) ants.’

Here, for (41d), a construction with subject indexing, e.g. rʿim-i harke šəšwone ‘Ants swarmed here’, would theoretically also have been available. What restrictions there are to this impersonalization in Ṭuroyo requires further investigation, but nothing like (41b) or (41d) is attested in NENA.

5.2.3.3 Lexical Transitivity and Intransitivity

Only those verbs that take a qṭil-form in the ‘perfective’ show a split in patient-like or agent-like subject indexes. The subject marking split parallels that of SE Trans-Zab Jewish varieties. Subjects are always coded in a patient-like fashion in the *qaṭṭil-class. Table 37 below illustrates the main semantic classes and respective coding that are compared with NENA below.

Although it is impossible to predict exactly on the basis of semantics what type of coding is preferred, there are notable tendencies.

Similarly to Jewish dialects like Sulemaniyya, it is noteworthy that, from a cross-linguistic perspective, the semantically most agent-like class of verbs denoting controlled activities (Croft 1998, 52–53) includes many verbs that do not take sA coding, such as raqəḏ-∅ ‘dance’ and šaġəl-∅ ‘work’ and čik-∅ ‘sneak in’.

Interestingly, the verb sḥy ‘swim; bathe, wash’ and other controlled activities do take sA coding in Ṭuroyo (sḥe-le), while the cognate verb sxy in Jewish Sulemaniyya does not (səxe-∅). A semantically similar verb is ḥayəf-∅ ‘wash (oneself)’ in Ṭuroyo, which does not take sA coding, e.g. ḥayif-i an-noše eba ‘The people washed with itF’ (Miden, Ritter 1967–1971, 78/213). Similarly to NENA, reflexives relating to dress and grooming, such as lwš ‘dress’, show agent-like coding and may also take an object, e.g. lwəš-še aj-julaṯṯe ‘They put on their clothes’ (Miden, Ritter 1967–1971, 76/33).

The agentless counterpart of transitive verbs, which receive patient-like subject coding generally belong to the mediopassive stem derivations. There are only few exceptions. An example is the verb xlṣ ‘save, escape’, which has a ‘perfective’ form xaləṣ-∅ ‘be saved’.16 Verbs expressing uncontrolled processes generally do not take sA coding, irrespective of morphological class: either a qṭil- or *qaṭṭil-base, and this situation corresponds to NENA, as given in (42) and (43) below. The verb yaqəd-∅ ‘burn’, for example, belongs to stem Ib and has a derived causative. Essentially, the qaṭṭil-base is used to decrease the effect implicature and centralize a state or entering into a state affecting the subject (see § 5.2.3.1.).

Table 37

Patient-like or agent-like marking of s in Ṭuroyo

Lexical class

s

qṭil-

*qaṭṭil-

state, (dis)position

E-set

ġbin-∅

‘be angry’

zayəʿ-∅

‘fear’

change of state, (dis)position

(sP)

ṯniḥ-∅

‘rest’

yaṯu-∅

‘sit’

uncontrolled process

ḥniq-∅

‘suffocate’

nafəl-∅

‘fall’

čik-∅

‘sneak in’

ʿabər-∅

‘enter’

controlled activity

sḥe-le

‘swim’

raqəḏ-∅

‘dance’

zmər-le

‘sing’

šaġəl-∅

‘work’

reflexive

lwəš-le

‘dress’

ḥayəf-

‘wash’

šləḥ-le

‘undress’

gawər-

‘marry’

sound emission

(sA)

nwəḥ-le

‘bark’

object omission

L-set

xi-le

‘eat’

yaləf-∅

‘learn’

Data based on Jastrow (1985); Ritter (1990); Noorlander’s field notes

(42) Derived causative (*qaṭṭil-class)

Ṭuroyo

J. Sulemaniyya (Khan 2004a)

‘burn’

intr.

yaqəḏ-∅

intr.

qil-∅ (~ yəliq-∅)

tr.

moqaḏ-le

tr.

mqəl-le

(43) Labile (qṭil-class)

a.

‘break’

intr.

twir-∅

intr.

twir-∅

tr.

twəḷ-ḷe

tr.

twər-re

b.

‘suffocate’

intr.

ḥniq-∅

intr.

ḥniq-∅

tr.

ḥnəq-le

tr.

ḥnəq-le

NENA and Ṭur ʿAbdin dialects diverge more strongly when it comes to the agent-like coding of subjects, as illustrated in (44) below. Verbs that denote a controlled event are treated differently, whereby šaġəl-∅ ‘work’ and gawər-∅ ‘marry’ do not take a-like coding in Ṭuroyo, but they do in NENA, whereas ṣhe-le ‘swim; bathe, wash’ takes a-like coding in Ṭuroyo, but it does not in NENA. Moreover, there is an exceptional group of transitive verbs belonging to subclass Ib (*qaṭṭil-) that mainly expresses mental states, where the experiencer subject is (indirectly) affected by some mental experience, including more controlled mental activities, such as yaləf-∅ ‘learn’ (instigating) and uncontrolled mental processes, such as ṭaʿi-∅ ‘forget’ (non-instigating) (Jastrow 1985, 72; Ritter 1990, 93; Furman and Loesov 2014). These correspond to sA forms in NENA, as shown by the comparison with Jewish Sanandaj below.

(44) Subject coding in Ṭuroyo and Jewish Sanandaj

Ṭuroyo

J. Sanandaj (Khan 2009)

a.

raqəḏ-∅

‘dance’

=

rqil-∅

b.

yaləf-

‘learn’

yləp-le17

c.

sḥe-le

‘swim’

səxe-∅ (also ‘wash’)

d.

šaġəl-∅

‘work’ (< Ar.)

ḥaštá wi-le (< Ir.)18

e.

gawər-∅

‘marry’

gəwr-e (< *gwər- + -le)

f.

aḏəʿ-∅

‘know’

ʾli-le

g.

šaməʿ-∅

‘hear’

šmi-le19

There are several verbs that have similar semantic characteristics as the (Ib) subclass taking a *qaṭṭil-base but belong to the (Ia) subclass taking a qṭil-base and transitive coding (Ritter 1990, 733), for example ḥzy ‘see’ and bʿy ‘want’:

qṭil-

*qaṭṭil-

(45)

ḥze-le

‘see’

vs.

šaməʿ-∅

‘hear’

bʿe-le

‘want’20

vs.

abəʿ-∅

‘want’ (roots bʿy vs. ∅bʿ)

This is consistent with the cross-linguistic tendency for ‘see’ to be the most salient of perception verbs (Viberg 1983) and receive transitive coding more likely than ‘hear’ (Haspelmath 2015).

Conversely, some middle-only verbs belonging to stem IM, e.g. ṯniḥ-∅ ‘rest’, are similar to class Ib (*qaṭṭil-) in terms of semantics (stative, experiencer), but occur in a derived causative alternation (Jastrow 1985, 77, 92), for example:

(46)

intr.

IM

fṣiḥ-∅

‘be(come) glad’

tr.

III

mafṣaḥ-le

‘gladden’

Moreover, there are intransitive verbs belonging to other stem derivations than stem I that receive agent-like subject coding, such as II hlx ‘walk’, e.g. mhalax-le (alongside Ib rahəṭ-∅ ‘run’) and III syw ‘become old’, e.g. masu-le.

Verbs from class (Ib) have also been attested in class (Ia), e.g. šməʿ-le ‘He heard’ instead of šaməʿ (Furman and Loesov 2014), thus showing an alternation between Ia, qṭil- + L-set and Ib, *qaṭṭil- + E-set. Ritter (1990, 15, 85, 619) offers examples of the following kind:

(47)

kafən-∅

‘er hungerte’

kfəl-le21

‘er bekam hunger’

fahəm-∅

‘er begriff’

fhəm-le

‘er verstand’

hawi-∅

‘geschehen’

hwe-le

‘entstehen’

The semantic difference between the two variants does not seem to be very obvious, but Ritter (1990, 85) hints at an aspectual distinction of punctuality, noting that forms like šməʿ-le and dməx-le are used “when one wants to emphasize the sudden occurrence of the event or its completed nature”.22 It seems to me that Ritter is referring to punctuality, which could be comparable to the role of punctuality in subject coding in, for instance, the Jewish dialect of Sulemaniyya (Khan 2004a, 301; see § 3.5.). There is no indication this is a productive process, however. Nevertheless, the fact that both transitive and intransitive verbs can occur in this alternation could suggest that we are not simply dealing with a dominant verbal form encroaching on others’ domain, as if speakers make errors and generalize stem (Ia) verbs, although it is conceivable that the anticipation of a transitive verbal form could trigger morphological adaptation to such a form, as in the following case, e.g. Ritter (1967–1971: 29/172)

qəm-la

í-k̭ačk̭e

nšəq-la,

bene

ʿayn-e

risePFV-a:3fs

the-girl:fs

kissPFV-a:3fs

between

eye:pl-hiss

‘The girl rose (and) kissed him on the forehead.’

This alternation could be comparable to an ‘antipassive’, which can be conditioned by aspect (e.g. Hopper and Thompson 1980). The form based on the intransitive construction, such as fahəm-∅, is durative, meaning ‘He knew, was able to perceive’, while the one based on the transitive, e.g. fhəm-le, is punctual, meaning ‘He realized’. It is possible that yaləf-∅ in (48a) below, for example, is used to focus on the learning process over time, while the agent-like form iləf-la in (48b) focuses on the moment of its completion and reaching a concrete effect (Ritter’s “completed nature”), even though both are perfective in terms of grammatical aspect (Ritter 1990, 656);23 this is also a distinction in the coding of transitivity.

(48) Punctuality vs. durativity (Midyat; Prym-Socin 1881, 157.25, 201.6)

a.

yaləf-∅

ú-kŭrrəko

qroyo,

msək-le

ás-saḥrat

b-í-qrayto

learnPFV-a:3ms

the-boy

read:inf

seizePFV-a:3ms

the-magical.power:pl

prp-the-reading

‘The boy learnt to read, (and) acquired magical powers through reading.’

b.

iləf-la

qroyo?

omər

iləf-la,

mayiṯ-o

learnPFV-a:3fs

read:inf

he.says

learnPFV-a:3fs

diePFV-s:3fs

Did she (i.e. the camelF) finally become able to read? He said: She did (and) died.’

It is possible that an additional semantic difference in dynamism plays a role, as observed for Jewish Sulemaniyya (see § 3.5.1 and § 3.5.2.). This is compared in (49a–b) below. A verb like tym ‘finish’ would focus on the cessation of an action and is more stative and endpoint-oriented than a verb like bdy ‘begin’, which implies a dynamic initiation.

(49) Dynamic vs. stative

Ṭuroyo

J. Sulemaniyya

(SE Turkey)

(NE Iraq; Khan 2004a)

a.

‘finish’

tr.

matəm-le

tr.

mtim-le

(stem III, L-set)

intr.

tayəm-∅

intr.

tim-∅

(stem Ib, stative, E-set)

b.

‘begin’

intr.

bde-le

intr.

bde-le

(stem Ia, dynamic, L-set)

However, one equally finds lexical alternatives that are not triggered by this semantic difference, such as xlṣ for ‘finish’ in examples like maxlaṣ-li ú-mŭklo ‘I finished eating’ (Ritter 1990, 221).

Four main lexical classes thus interact and overlap, as summarized in Table 38 below. Each may attract other verbs of similar semantics or derivational patterns.

Table 38

Ṭuroyo stem I subclasses in the ‘perfective’

qṭil-base

*qaṭṭil-base

tr.

ḥze-le (Ia)

‘see’

šaməʿ-∅ (Ib)

‘hear’

intr.

sḥe-le

‘swim’

raqəḏ-∅

‘dance’

intr.

fṣiḥ-∅ (IM)

‘be(come) glad’

saməq-

‘be(come) red’

The *qaṭṭil-form stands out system-internally. It is largely confined to basic single argument verbs that hardly occur in a labile alternations and to two-argument verbs denoting mental situations. In other respects, split subject-marking in Ṭuroyo shows strong similarities to that in NENA. sA coding (i.e. the L-set) becomes increasingly more likely under semantic conditions similar to those found in NENA (cf. Khan 2004a, 304–305), where an effect is more strongly implied and the event is punctual and dynamic. Nevertheless, lexicalization largely obscures these tendencies.

5.3 The Neo-Aramaic Dialect of Mlaḥsó

Mlaḥsó is rather distinct from the dialects of Ṭur ʿAbdin and similar to peripheral NENA dialects in SE Turkey. The neutral indexing and differential flagging of p are also comparable to Jewish dialects in Iranian Azerbaijan. Passive and anticausative voice phenomena in Mlaḥsó are different from all other dialects. Finally, the realis perfect is based on the *qaṭṭil-form regardless of lexical semantics and comparable to the situation in Christian Borb-Ruma.

5.3.1 Alignment of Person Marking

Mlaḥsó groups all grammatical functions under the L-set in the perfective past, treating s, a and p alike in morphological marking.24 The E-set is never used as object indexes in Mlaḥsó. This is similar to Christian NENA dialects in SE Turkey, particularly C. Borb-Ruma (Bohtan, SE Turkey; Fox 2009), but also to the Northwest Iranian Jewish dialects, such as Urmi (NW Iran; Khan 2008b). Example (1) offers a comparison for the verbs ‘take’ and ‘rise’ between Mlaḥsó and C. Borb-Ruma:

(1) Neutral alignment

Mlaḥsó

C. Borb-Ruma

(Jastrow 1994, 146.10)

(SE Turkey; Fox 2009)

a.

ṣíd-len-li

c.

ṣə́d-lan-ni

They took me.’

They took me.’

b.

qim-li

d.

qəm-li

I rose.’

I rose.’

An (e)l-series of independent object person markers is treated like full nominals and occurs in preverbal position (Jastrow 1994, 14). It may also alternate with the L-set as dependent person marker,25 which is comparable to the ʾəll-series in NENA.

e.

Mlaḥsó

l-i ṣid-len

‘They took me.’

The distinction between dependent and independent person markers is marginal in Mlaḥsó. The difference between the L-set and (e)l-series is most conspicuous in the 3ms. and 1pl. where the preposition takes the distinct suffixes -áv and -əna. Compare (2a) and (2b) below.

(2) Mlaḥsó (Jastrow 1994, 96.164,167)

a.

hiv-le

el-áv

mo

dahvé

(independent)

givePFV-a:3ms

r:dat-3ms

hundred

gold:pl

b.

hív-le-le

mo

dahvé

(dependent)

givePFV-a:3ms-r:3ms

hundred

gold:pl

‘He gave him one hundred pieces of gold.’

Pronominal objects are limited in general in Mlaḥsó. An object index is not obligatory and is frequently lacking when the referent is considered to be clear enough from the context. It is generally only expressed once and not continued by other constructions with the same referent (Jastrow 1994, 56).

Finally, agents are not marked prepositionally as in Ṭuroyo, except for the first person plural. The first person plural does not distinguish between dative and unmarked independent person markers. While other persons distinguish between unmarked and dative forms, such as the first person singular ono ‘I’, as opposed to (e)li ‘me’ and third masculine singular hiye ‘he’ as opposed to eláv ‘him’, the first person plural is eləna throughout and can also mark s or a even in the ‘imperfective’ (compare Ṭuroyo aḥna and elan) (Jastrow 1994, 28, 63). It is based on the dative preposition (e)l- and the first person plural ‘possessive’ suffix -əna. Thus unlike other independent person markers, the 1pl. eləna is completely neutral to its syntactic role, merging s, a, p, t and r (Jastrow 1994, 63),26 for example:

(3) First person plural pronoun in Mlaḥsó27

a.

eləna pišlan tamo

We stayed there.’

(s)

b.

eləna emirlan

We said.’

(a)

c.

eləna mapleṭlen

‘They helped us escape.’

(p)

d.

eləna mobele

‘He brought us there.’

(t)

e.

eləna hivlen

‘They gave to us.’

(r)

Generally speaking, then, Mlaḥsó prepositional marking is accusative (a=sp), but neutral for the first person plural (a=s=p). Indexing is morphologically neutral. Differential indexing and prepositional marking of arguments does not appear to be combined.

5.3.2 Neutralizing Subject Coding: Mediopassive with L-suffixes

Transitive and intransitive verbs inflect alike in the ‘perfective’ in Mlaḥsó. Mlaḥsó makes no distinction between the coding of s or a, for example:

(4)

dmix-lan

‘We slept.’

ḥze-lan

‘We saw.’

šmiʿ-lan

‘We heard.’

The E-set does not occur in combination with qṭil- under any conditions.

The L-set marks s in all intransitive constructions alike, including the passive. Only a few anticausatives remain in the active stem I that correspond to verbs belonging to stem Ib (*qaṭṭil-) in Ṭuroyo, for example ḥrv ‘destroy’, whose corresponding causative is stem III:

(5) The verb ‘destroy’ in Mlaḥsó and Ṭuroyo

Mlaḥsó28

Ṭuroyo

intr.

beyt-í ḥriv-le

intr.

bayt-i ḥaru-∅

(stem I)

‘My house got destroyed.’

‘id.’

tr.

maḥrev-le

tr.

maḥru-le

(stem III)

‘He destroyed (sth.).’

‘id.’

The s of a passive is similarly marked by the L-set. The -t-infix is the only morphological difference between the active and mediopassive of stem III verbs, such as ∅ḥt ‘put’:

(6)

tr.

III

maḥet-le

‘He put (sth.).’

intr.

IIIM

mtaḥet-le

‘He was put.’

Voice distinctions, therefore, are completely attuned to the type of stem in Mlaḥsó (Jastrow 1994, 41). In Ṭuroyo, by contrast, this is mainly dependent on verbal person marking. We can contrast this stem neutralization in Mlaḥsó to the voice distinctions in Ṭuroyo for the labile stem I verb ‘open’ and the transitive stem III verb ‘sell’ (cf. Jastrow 1996). The inflectional base is modified depending on TAM in Ṭuroyo, but on valency in Mlaḥsó.

(7) Stem neutralization in Mlaḥsó (Adapted from Jastrow 1994, 83.53–54, 88.99; 1996)

Mlaḥsó

Ṭuroyo

a.

tarʿó mepseḥ-∅

f.

ko-məftəḥ-∅ tarʿo

(present)

‘A door opens.’

‘id.’

b.

tarʿó mepseḥ-le

g.

ftiḥ-∅ tarʿo

(preterit)

‘A door opened.’

‘id.’

c.

tarʿó psiḥ-le

h.

ftəḥ-le tarʿo

(active, preterit)

‘He opened a door.’

‘id.’

d.

mzaben-no

i.

ko-mizaban-no

(passive, present)

‘I am sold.’

‘id.’

e.

mzaben-li

j.

mzaban-no

(passive, preterit)

‘I was sold.’

‘id.’

The examples in (7) show that the Mlaḥsó mediopassive makes no distinction between ‘perfective’ and ‘imperfective’ inflectional bases.29 The mediopassive base (e.g. IM mepseḥ-, IIIM mzaben-) is stable throughout, but the subject and agent coding is entirely tense-aspect-sensitive (e.g. E-set in the present vs. L-set in the preterit) irrespective of lexical semantics. The levelling of mediopassive stems in Mlaḥsó is presumably analogical to the active counterparts of stem II and IV verbs (Jastrow 1996, 57). These similarly merge the ‘imperfective’ and ‘perfective’ in Ṭuroyo active forms,30 for example:

Mlaḥsó

Ṭuroyo

k.

zaben-no

m.

ko-mzaban-no

(present)

‘I sell.’

‘id.’

l.

zaben-li

n.

mzabal-li (< mzaban-li)

(preterit)

‘I sold.’

‘id.’

In the end, the crucial difference from Ṭuroyo is the complete mixing of stems in Mlaḥsó by extending the ‘imperfective’ bases to the expression of the perfective past. The single L-set, otherwise associated with agent coding in Ṭuroyo and NENA, covers the entire voice spectrum ranging from causative to passive, functioning as the main TAM marker (preterit) against the E-set (present).

5.3.3 Special Perfect Forms Based on *qaṭṭil-

The E-suffixes never express objects in Mlaḥsó, as they do in Ṭuroyo and numerous NENA dialects (see § 2.3.3.). As person markers of both s and a, they are not only found in the ‘imperfective’ forms of all verbs, but also in the perfect, only attested for stem I. The perfect is formed with the *qaṭṭil-base, for example:

(8) Mlaḥsó (Jastrow 1994)

a.

dmix-le

‘He fell asleep.’

(preterit, L-set)

b.

damíx-

‘He has fallen asleep.’

(perfect, E-set)

c.

qim-le

‘He rose.’

(preterit, L-set)

d.

qaym-

(< *qayyim-)

‘He has risen.’

(perfect, E-set)

The *qaṭṭil-forms can also be used to express result states, e.g. kla rumo kali ‘Look there, a soldier is standing’ (Jastrow 1994, 142.36).

These perfect forms as such, however, are not restricted to intransitive and lowly transitive verbs in Mlaḥsó. All verbs, even transitives, which do not feature in the so-called *qaṭṭil-subclass in Ṭuroyo, such as ḥze-le ‘see’ against šaməʿ-∅ ‘hear’, can be conjugated in the same way in Mlaḥsó, e.g. šmiʿ-le ‘He heard’ against šamiʿ-∅ ‘He has heard’.

This situation is similar to our observations for C. Borb-Ruma (SE Turkey) in NENA (see § 4.4.3.2.), although NENA does not show a change in inflectional base. (9) below offers a comparison of the verbs ‘see’ and ‘give’.

(9) Transitive realis perfect in Mlaḥsó and C. Borb-Ruma

Mlaḥsó

C. Borb-Ruma

(Jastrow 1994)

(Fox 2009)

a.

ḥze-li

e.

ġze-li

(preterit, L-set)

‘I saw.’

‘id.’

b.

ḥazi-no

f.

ġz-ən

(perfect, E-set)

‘IM have seen.’

‘id.’

c.

hiv-le

g.

hu-li

(preterit, L-set)

‘He gave.’

‘id.’

d.

hayv-

h.

hu-

(perfect, E-set)

‘He has given.’

‘id.’

The difference between Mlaḥsó and C. Borb-Ruma mainly hinges on the two verbal bases for stem I verbs, *qaṭṭil- for the realis perfect against qṭil- for the preterit. In both dialects, the perfect and preterit are distinguished by a distinct set of verbal person markers. The perfect is transitive and readily combines with object nps in the same fashion as the ‘imperfective’, for example:

(10)

a.

[p]

[v-a]

ḥelm-ano

ḥazi-no

dream:m-dem:ms

see:perf-a:1sg

‘I saw that dream.’ (Jastrow 1994, 130.139)

b.

[a]

[v

-a

-p]

em-i

w

ov-i

naṭir

-a31

-li

mother:f-my

and

father:m-my

look.after:perf

-a:3pl

-p:1sg

‘My parents looked after me.’ (ibid. 94.157)

Verbal forms that otherwise denote the perfective past can also express the present perfect or a result state in Ṭuroyo just as in NENA, e.g. aḏiʿ-at-li? ‘Do youSG still know me?’, and ftiḥ-i ayn-a ‘Her eyes were open’ (Midyat, Prym-Socin 1881, 88.21). Nevertheless, it is possible to mark the realis perfect by means of the actualizing preverb ko-, which may also be enhanced by additional particles ga and kal, for example:

(11) Ṭuroyo (cf. Jastrow 1985, 153–154)

a.

(∅-)qṭi-le

‘He killed (him).’

(preterit)

b.

ko-qṭi-le

‘He has killed (him).’

(perfect)

c.

(∅-)qayəm-∅

‘He rose.’

(preterit)

d.

ko-qayəm

‘He has risen.’

(perfect)

e.

(∅-)šaməʿ-∅

‘He heard.’

(preterit)

f.

ko-šaməʿ

‘He has heard.’

(perfect)

This system, where the only morphological distinction between preterit and perfect is preverbal TAM-marking, has nevertheless parallels in NENA.32

By contrast, the choice between the L-set or E-set in subject and agent coding depends wholly on aspect in Mlaḥsó similarly to NENA dialects, such as C. Borb-Ruma (SE Turkey). The *qaṭṭil-form is less grammaticalized along the path from resultative to perfective past, while the qṭil-form with L-suffixes has fully grammaticalized. In other respects, the *qaṭṭil-form in Ṭuroyo is less grammaticalized. Only two-argument state verbs belonging to stem (Ib) may take objects, so that the qaṭṭil-form has not grammaticalized fully to also include highly transitive verbs. This also confirms an earlier grammaticalization of the qṭil-form with L-suffixes, as this form, by contrast, is compatible with all types of verbs, including the primary transitive ones and sometimes even stem (Ib) verbs, whereas the other way around does not apply.

5.4 The Primacy of Intransitive Coding

The mediopassive inflectional base, e.g. *məqṭəl-, is extended from the ‘imperfective’ to the expression of the preterit, i.e. perfective past, in Mlaḥsó. This morphological adaptation proceeds in the opposite direction in NENA found in the preterit, where the transitive coding is analogical to the ‘imperfective’.

First of all, as we saw in the previous section, the E- and L-series are tense-aspect-conditioned subject and agent markers in Mlaḥsó. Remarkably, in some respects, the Mlaḥsó verbal system mirrors the transitive qam-qaṭəl-construction found in NENA dialects (see § 4.4.5.). While several NENA dialects use a dedicated transitive construction based on the ‘imperfective’, Mlaḥsó uses a dedicated intransitive construction on the basis of an ‘imperfective’ base. It is only the set of person markers that expresses the TAM distinction:

intransitive

transitive

(1)

mepseḥ

-o

‘ItF opens.’

posḥ-o-le

‘She opens itM.’

mepseḥ

-la

‘ItF opened.’

psíḥ-la-le

‘She opened itM.’

The qaṭəl-base is extended from the present to the preterit in NENA, while the məqṭəl-base of the intransitive pendant is extended from the present to the preterit in Mlaḥsó. The direction of morphological adaptation is schematized in (2) below.

(2) Mlaḥsó

pret

prs

pfv-base

ipfv-base

tr.

psíḥ-la-le

posḥ-o-le

intr.

mepseḥ-la

mepseḥ-o

Ṭuroyo finds itself in the middle. Consider the following examples.

(3)

ko-ipfv-E-L

ko-

madamx-o-li

‘She lulls me to sleep.’

(∅-)pfv-E-L

(∅-)

madamx-o-li

‘I lulled her to sleep.’

(4)

ko-ipfv-E

ko-

madmax-no

‘IM lull to sleep.’

pfv+L

madmax-li

‘I lulled to sleep.’

To some extent, the adoption of the stem from the ‘imperfective’ into the preterit is also found among speakers of the Neo-Aramaic dialect of Midyat, for example the stem IM verb jġl ‘speak’, e.g. ko-məjġəl-∅ ‘He is speaking’, jġil-∅ ‘He spoke’ like ko-məfṣəḥ-∅ ‘He is glad’, fṣiḥ-∅ ‘He became glad’. məqṭəl-forms of the following kind can be found instead of expected qṭil-forms, and with the additional L-suffix to indicate the distinction in TAM:

(5) Ṭuroyo (Midyat; cf. Ritter 1967–1971, 11/297)

pret

prs

pfv-base

ipfv-base

unmodified

jġil-∅

˚məjġəl-

adapted

məjġe-le

məjġəl-

Preverbal TAM-marking (ko-) is significant to differentiate between forms that are morphologically identical, such as stem III verbs like madməx- ‘lull to sleep’. Preterit and actual present are differentiated by the prefix ko- only when third person coding from the E-set (e.g. 3fs. -o) immediately follows the verbal base. When argument coding other than third person immediately follows the verbal base, no such ambiguity would arise due to the person role constraint.

The E-set (-no) and L-set (-li) arguably signal a shift in tense-aspect comparable to Mlaḥsó, where ko- practically serves only to distinguish the realis present from the subjunctive. Mlaḥsó uses the x-preverb only with initial weak verbs. The distinct set of verbal person markers is sufficient to keep the tense-aspect apart.

The system in Mlaḥsó, therefore, is not only grounded in the levelling of inflectional bases by means of morphological identity and analogy (Jastrow 1996, 57), but it is also facilitated by the TAM marking function of the respective sets of suffixes.33

5.5 Summary from Stem to Stern

Central Neo-Aramaic has much in common with Northeastern Neo-Aramaic. With respect to alignment, Ṭuroyo and Mlaḥsó are especially similar to the Trans-Zab Jewish dialects of NENA. Ṭuroyo is similar to Jewish dialects of Iraqi and Iranian Kurdistan. Mlaḥsó is similar to Christian dialects in SE Turkey such as Borb-Ruma (Bohtan) as well as Jewish dialects of Iranian Azerbaijan. What sets them apart from these NENA varieties is the use of mediopassive stem derivations and a distinct ‘perfective’ base *qaṭṭil-, whose original resultative-stative intransitive semantics lingers on and is reflected in different ways.

Object-marking is person-restricted for qṭil- in Ṭuroyo as in the majority of NENA. The E-set is limited to the third person, grouping s and p ergatively, while first and second person are marked by the L-set, grouping a and p horizontally. The alignment of dependent person markers is phonologically non-distinct for Mlaḥsó, where the E-series is unavailable to mark objects:

(1)

Ṭuroyo

Mlaḥsó

p[−1,2]

p[+1,2]

p[−1,2]

p[+1,2]

tr.

ftiḥ-o-le

ftə́ḥ-le-li

psíḥ-le-la

psíḥ-le-li

intr.

ftiḥ-o

mepseḥ-la

Differential prepositional marking as well as a series of independent object person markers are based on the preposition (e)l-. Although nouns are normally unmarked in Ṭuroyo, differential prepositional marking does occur. Ṭuroyo is unique in using (e)l- also to mark optionally a together with indexing (the L-suffixes). This yields an ergative prepositional marking pattern alongside ergative indexing of full nps, e.g. haṯe xil-o-le l-u-kalwoThe dog ate this’. The optional prepositional marking of the agent parallels the possessor in predicative possessor constructions, e.g. abro kət-le l-u-malko ‘The king has a son’.34 The possible prepositional marking patterns are illustrated below for the phrases ‘The king opened the door’ and ‘The door opened’. Differential object marking and optional a-marking are not mutually exclusive. In at least the dialect of Rayite, they may be combined, manifesting horizontal alignment (like first and second person markers). Ergative indexing appears to be combined only with ergative prepositional marking and never horizontal prepositional marking.

(2)

Ṭuroyo

a.

(a=s=p) neutral

(as=p) ergative

tr.

u-malko ftəḥ-le u-tarʿo

l-u-malko ftəḥ-le u-tarʿo

intr.

u-tarʿo ftiḥ-∅

u-tarʿo ftiḥ-∅

b.

(a=sp) accusative

(sa=p) horizontal

tr.

u-malko ftəḥ-le l-u-tarʿo

l-u-malko ftəḥ-le l-u-tarʿo

intr.

u-tarʿo ftiḥ-∅

u-tarʿo ftiḥ-

The dialects of Ṭur ʿAbdin have various verbal classes, in which the compatibility with a-like coding primarily hinges on the verbal base and not the verbal semantics. Basic verbs known as ‘neuter verbs’ generally do not occur in labile alternations and have a special *qaṭṭil-base in the ‘perfective’ in Ṭuroyo, e.g. damix-o ‘She fell asleep’ as opposed to ftiḥ-o ‘ItF opened’, particularly verbs whose subject is an affectee registering a state or change of state. Semantically, single argument states, change-of-state verbs and uncontrolled processes are typically incompatible with sA marking, while verbs with a stronger implication of a dynamic effect, such as sound emission verbs, e.g. nwəḥ-le ‘He barked’, typically take the L-suffixes like a. Many situation types, however, such as controlled activities are variably categorized in Ṭuroyo, e.g. raqəḏ-∅ ‘dance’ vs. zmər-le ‘sing’. A few transitive verbs that generally express two-argument mental states and activities, such as šmʿ ‘hear’ and ylf ‘learn’ are incompatible with a-like coding, and take transitive person marking similarly to that of the ‘imperfective’, e.g. šamiʿ-o-li ‘She heard me’: ˚šəmʿ-o-li ‘She hears me’. Primary transitives never occur in the *qaṭṭil-form as such, but some verbs of class (Ib) occasionally occur in the inflection of class (Ia), whereby šməʿ-le ‘He heard’ and dməx-le ‘He slept’ have been attested alongside šaməʕ-∅ and daməx-∅. The difference between the two is not always semantically obvious, but it may reflect a relic of fluid-subject marking that once existed in Ṭur ʿAbdin, as it did in NENA.

Mlaḥsó, in turn, has a fully productive distinction between preterit and perfect depending on both inflectional base (qṭil- vs. *qaṭṭil-) and related agent and subject indexes (L-set vs. E-set). The qṭil-form combines with the L-set to express the preterit, e.g. dmix-le ‘He fell asleep’, šmiʿ-le ‘He heard’, qṭile ‘He killed’, but the *qaṭṭil-form combines with the E-set to express the perfect, e.g. damix-∅ ‘He has fallen asleep, is asleep’, šamiʿ-∅ ‘He has heard’, qaṭil-∅ ‘He has killed’. Preterit and perfect are distinguished by the TAM-preverb ko- in Ṭuroyo, compare:

(3)

Ṭuroyo

Mlaḥsó

preterit

perfect

preterit

perfect

tr.

ftəḥ-le

ko-ftəḥ-li

psiḥ-le

paṣiḥ-∅

intr.

daməx-∅

ko-daməx-∅

dmix-le

damix-∅

Finally, Central Neo-Aramaic shows a more complex verbal derivation system than NENA. Ṭuroyo and Mlaḥsó diverge significantly here as well: the set of person markers is essentially valency-conditioned in Ṭuroyo, but tense-aspect-conditioned in Mlaḥsó, so that for the verb ‘open’ we observe:

(4)

Ṭuroyo

Mlaḥsó

pret

prs

pret

prs

causative

ftəḥ-le

ko-fotəḥ-∅

psiḥ-le

poseḥ-∅

inchoative

ftiḥ-∅

ko-məftəḥ-∅

mepseḥ-le

mepseḥ-∅

In comparison to NENA (cf. Mengozzi 1998, 84), the qaṭəl-base has a wider functional distribution in NENA, because of the lack of special anticausative morphology in the ‘imperfective’, as contrasted in (5) below.

(5) Inchoative ‘open’ in Central Neo-Aramaic and NENA

Ṭuroyo

Mlaḥsó

J. Sanandaj

C. Jinnet

(SE Turkey)

(SE Turkey)

(W Iran)

(SE Turkey)

pfv

ftiḥ-∅

mepseḥ-le

plix-∅

ptəḥ-le

ipfv

˚məftəḥ-∅

mepseḥ-∅

păləx-∅

potəḥ-∅

Another important difference from NENA is that the agentless qṭil-form (cf. Gutman 2008) may be used to express the impersonal passive of both transitive and intransitive verbs in Ṭuroyo, e.g.

(6) Ṭuroyo

ftiḥ-i-le

at-tarʿe

‘He opened them—the doors.’

(causative)

ftiḥ-i

at-tarʿe

‘The doors opened.’

(inchoative)

ftiḥ

tarʿe

‘People opened doors.’

(impersonal)

1

q-ṭ-l, although as a lexical root meaning ‘kill’, is purely a dummy here to illustrate the consonantal template for sound verbs, but does not occur in this class. The gemination and asterisk indicate its historical origin to avoid confusion with NENA qaṭəl- that corresponds to Ṭuroyo qoṭəl-.

2

A resulting sequence əw contracts to u. Compare Midyat kṯuwole (for *kṯəwole) ‘He wrote itF’ and Miden kṯiwole ‘id.’.

3

See Waltisberg (2016, 186 ff.) for more examples. An example of the prepositional marking of themes: gd-obe-n-ŭx l-í-barṯayḏi ‘I will give youMS my daughter’ (Ritter 1967–1971, 107/84).

4

See § 2.3.2.2. for this functional distinction.

5

See Waltisberg (2016, 188–190) for more examples.

6

The 2pl. and 3pl. L-suffixes have idiosyncratic allomorphs (Jastrow 1985, 138) due to historical retentions, which are not discussed here.

7

Horizontal alignment features in Jewish Saqiz for the first and second person (see § 3.3.1.2). Possibly, the realis perfect in C. Artun (Hertevin) also shows horizontal alignment for the third person, e.g. hole wed-le-lehen ‘He has made them’, where a and p are grouped, against (hole) dmiḥ-∅ ‘He has slept’ (see § 4.3.1. and § 4.4.4.).

8

See § 3.2.2. for a discussion and compare Comrie (2005) and Malchukov (et al. 2010).

9

Similarly, Coghill (2016, 64–65) subsumes constructions like qṭə́l-la-le ‘She killed him’ in NENA under accusative alignment. See § 4.4.3. for a discussion of such forms in NENA.

10

See also Jastrow (1985, 137–138) and Waltisberg (2016, 296). This is a tripartite type of ditransitive verbal person marking (tpr) that correlates with ergative verbal person marking (as=p).

11

Cf. Ritter (1990, 65) and Diem (2012, 43–45).

12

Cf. Coghill (2016, 87–90).

13

See § 4.2.2.2. Cf. Givón (1985a), McGregor (2006, 2010), Fauconnier (2012), Verbeke (2013a).

14

This form, e.g. ḥazi-∅ ‘he was seen’, developed within Ṭuroyo by analogy with the fs. forms of the intransitive final-y verbs: baxyono ‘I (f.) wept’: ḥazyono ‘I (f.) was seen’, baxyo ‘she wept’: ḥazyo ‘she was seen’, baxi ‘he wept’: x; x = ḥazi ‘he was seen’.

15

Compare the German original (ibid.): “es wurde (auf dem Dache) geschlafen”.

16

A sense of ‘escape; become safe’ may also be in view (Ritter 1990, 219 ff.).

17

The intransitive form in J. Sanandaj yəlip-∅ conveys ‘learn’ in the sense of knowledge reception (less control) rather than acquisition (more control), i.e. being taught by somebody else.

18

ḥaštá ‘work’, wil- ‘do’ + -le.

19

It is possible that the intranstive coding in local Arabic cognates has influenced a few verbs belonging to subclass Ib: Arabic stative saməʿ-tu ‘I heard’ and mediopassives f-t-aham-∅ ‘He understood’ and aš-t-aġal-tu ‘I worked’ (Mardin, SE Turkey; Grigore 2007) correspond to Ṭuroyo šaməʿ-no, fahəm-∅ and šaġəl-no.

20

This root seems to be obsolete in Ṭuroyo today and is retained only in the impersonal expression kə-bʿe ‘must’ (Ritter 1990, 733).

21

kfəlle < *kfən-le.

22

German original (ibid.): “wenn man das plötzliche Eintreten des Geschehens, oder seinen abgeschlossenen Charakter hervorheben will”.

23

Ritter (1990, 656) hints at such a subtle aspectual difference by his comment to (48b) “die Lehre ist abgeschlossen”.

24

For a different view, see Coghill (2016, 90) who considers this “fully accusative alignment”, presumably because she identifies alignment on the basis of affix order rather than phonological form. See § 2.3.2.3. for why I do not consider that determinative here.

25

Jastrow (1994, 54–56), however, suggests that, since his Turkish informants (Diyarbakır) predominantly use independent person markers instead, the higher frequency of object L-suffixes in the speech of his Syrian informant (Qamishli) is due to interference from Ṭuroyo. Although her speech probably does contain hybrid forms of Ṭuroyo and Mlaḥsó (Jastrow 1994, 35), one could conversely argue that the prevalence of independent person markers in the speech of Jastrow’s other informants is due to an overall stronger interference of Kurmanji Kurdish in Turkey, where such person forms are independent. Since the two co-existing object marking strategies are common to all his informants, I will not treat one as more genuinely Mlaḥsó over the other.

26

It appears, however, that a biform exists for its object-marking function on the basis of ʿal- ‘on, upon’, e.g. ʿalena ṣədlen ‘They took us (captive)’ (Jastrow 1994, 104.2).

27

Examples from Jastrow (1994, 104.2, 132.149, 104.11, 124.116.121).

28

Examples from Jastrow (1994, 118.85, 158).

29

The distinction between ‘imperfective’ and ‘perfective’ is also levelled in the 1ms. conjugation of hollow verbs belonging to stem I, cp. sim-no (~ səm-no) ‘I make (sth.)’ and sim-li ‘I made (sth.)’ (Jastrow 1994, 36).

30

There may also be another connection. It is possible to inflect certain ‘perfective’ forms of a mediopassive by means of L-suffixes to express a recipient referent in Ṭuroyo, e.g. mtawməṛ-ṛe (< mtawmər- + -le) tə-medeHe (lit. him) was told nothing’ (Jastrow 1992, 85.15).

31

It should be noted that the 3pl. index of the Mlaḥsó perfect is distinctly -a instead of -i, which thus far defies explanation.

32

Compare the discussions in §§ 3.4.2, 3.4.3, 4.1.2.2. and 4.3.1.

33

Ironically when I asked (educated) Ṭuroyo speakers (from Qamishli) whether forms like **nšiq-at-li ‘I kissed youFS’ were possible, they replied with disapproval and told me I was confusing tenses.

34

On this parallelism, see Diem (2012) and Noorlander (2021; forthcoming).

  • Collapse
  • Expand