TOWARDS A TYPOLOGY OF POSSESSORS AND EXPERIENCERS IN NEO-ARAMAIC : NON-CANONICAL SUBJECTS AS RELICS OF A FORMER DATIVE CASE

5 The Neo-Aramaic dialects are modern vernacular forms of Aramaic, which has a documented history in the Middle East of over 3,000 years. Due to upheavals in the Middle East over the last one hundred years, thousands of speakers of Neo-Aramaic dialects have been forced to migrate from their homes or have perished in massacres. As a result, the dialects are now highly endangered. The dialects exhibit a remarkable diversity of structures. Moreover, the considerable depth of attestation of Aramaic from earlier periods provides evidence for the pathways of change. For these reasons the research of Neo-Aramaic is of importance for more general fields of linguistics, in particular language typology and historical linguistics. The papers in this volume represent the full range of research that is currently being carried out on Neo-Aramaic dialects. They advance the field in numerous ways. In order to allow linguists who are not specialists in Neo-Aramaic to benefit from the papers, the examples are fully glossed.


Introduction 1
Predicative possessors and impersonal experiencer constructions are encoded by the dative preposition l-across Semitic languages, in addition to Aramaic, Hebrew (e.g. Berman 1982) and Syrian Arabic (e.g. Cowell 1964;Al-Zahre and Boneh 2010, 250). Like most non-European languages, Semitic languages do not have a designated possession verb. Predicative possessors equivalent to English have are based on locational expressions of prepositional possessor (Stassen 2009), as illustrated for Hebrew in (1a-b) below. Predicative possessors are originally intransitive constructions where the existential element agrees or used to agree with the possessee� It is a common phenomenon, sometimes termed 'havedrift' (Stassen 2009), that predicative possession undergoes transitivisation by assimilation of its morphosyntax to that of more typical and frequent agent-patient verb constructions because of their matching semantic-pragmatic properties (Stassen 2009, 208-43). While the agent-like possessor is still prepositional, the possessee has grammaticalised to a full-fledged object in colloquial Israeli Hebrew. It can be marked differentially by the object marker et, for example: Similarly, the preposition l-expresses the experiencer in impersonal experiencer verb constructions, as illustrated for Israeli Hebrew in (1e-f). The adjective or verb denoting the mental state is non-referential ms., while the subject-like experiencer is introduced by l-� The same preposition can also mark so-called external possessors� The possessor is not dependent on the nominal possessee itself but is expressed as an affectee part of the verbal predicate, for example: external possessor g� avad le-Dan ha-tik� lost�3ms to-Dan the-file.ms 'The file got lost on Dan.' (Berman 1982, 41) h� ima raxaṣa le-Dan et ha-panim� mom washed�3fs to-Dan dom the-face 'Mom washed Dan's face (for him). ' (ibid. 47) Such prepositional arguments can also be optionally added to co-refer to the subject with various semantic nuances such as (1i) below. Such subject-coreferential datives are also known as ethic or ethical datives (dativus ethicus) in Semitic linguistics 2 � subject co-referential dative i� ha-yeladim histalku la-hem the-children ran�away�3pl to-them 'The kids (upped and) ran away. ' (ibid. 51) All of these constructions are, of course, semantically and formally related to the expression of the recipient of ditransitive constructions in denoting often highly animate, typically human affectees (e.g. Berman 1989, 49;Naess 2007, 185-208).
Such subject-like prepositional affectees have been argued to be a common trait of Northwest Semitic (e.g. Pat-El 2018). Both full nominals and pronouns are marked prepositionally in all of the constructions illustrated above. Most Semitic languages lost case inflection presumably through phonetic erosion and other forces of change such as increasing fixation of word order. Thus instead of case declensions Neo-Semitic languages use zeromarked nouns and independent pronouns as the default citation form. They developed differential marking strategies of definite nominals, including cross-referencing through pronominal affixes. 3 Typically, the predicative possessor and the experiencer of impersonal experiencer verb constructions are marked by the same preposition l-and its allomorphs in Late Antique Aramaic languages such as Syriac. A key difference in Syriac is the optional use of additional 'pronominal copies'. That is, prepositional person markers that cross-reference a co-nominal. In (2a) below, for example, the prepositional possessor (l-ḡaḇrå ḥaḏ) is referred back to by a prepositional person marker (l-eh). The same holds for the experiencer in (2b). (2) Classical Syriac a� l-ḡaḇrå ḥaḏ ʾiṯ-wa-w l-eh to-man�ms one.ms exst-was-3mpl to-him tren bnin two�m son.mpl 'A certain man had two sons (lit� Him were two sons).' (Luke 15:11,Curetonian) b. kery-aṯ l-hun l-ḡaḇre grieved-3fs to-them�m to-man�mp w-eṯ-beš-∅ l-hun ṭåḇ and-medp-be.bad-s.3ms to-them�m well 'The men were grieved and very angry (Them grieved it F and angered it M ).' (Genesis 34:7, Pšiṭta) The possessor is stripped of its prepositional marking and becomes a zero-marked noun or pronoun, when it undergoes topicalisation. Its grammatical function as possessor or experiencer has to be resumed by the prepositional person marker such as l-eh in the following examples� c� gaḇrå ḥaḏ ʾiṯ-wa-w l-eh man�ms one.ms exst-was-3mpl to-him tren bnin two�m son.mpl 'A certain man had (lit. Him were) two sons.' (Luke 15:11,Sinaiticus) d� malka… kery-aṯ l-eh saggi king�ms grieved-3fs to-him much 'The king (who judged Daniel) felt very sorry (lit� Him grieved it F ).' (Aphrahat XXI: 411.20) Such agreement markers emerge out of topicalisation constructions through increasing obligatorisation (e.g. Givón 1976;Lehmann 1988, 62;cf. Diem 2012;Mor and Pat-El 2016) and accordingly transitivisation (see above). That is, the clause-initial position without prepositional marking is favoured for discourse topics. This position grammaticalises for 'non-canonical' subjects on the model of the 'canonical' subject in other clauses (i.e. transitivitisation) where sentence-initial position of the subject has become the default position. The remaining cross-referencing prepositional pronoun becomes effectively an inflectional crossindex like verbal affixes.
Neo-Aramaic languages have a set of person markers generally known as the L-suffixes that historically go back to such dative person markers based on the preposition l-� In a similar fashion as (2c-d) above, these L-suffixes are used to express the predicative possessor and impersonal experiencer, for example in the dialects of Ṭur ʿAbdin, i.e. Ṭuroyo (3a-b), and Christian dialect of Urmia, i�e� C� Urmi, (4a-b): (3) Ṭuroyo (Kfaerze, SE Turkey; Ritter 1967Ritter -1971 a� ú-həmmāl-ano kət-way-le əštó-abne. This article is a comparative survey of the morphological properties of such possessors and experiencers in Neo-Aramaic, concentrating on North-Eastern Neo-Aramaic (NENA) and Central Neo-Aramaic (i.e. Ṭuroyo and Mlaḥsó). Some comparative remarks concerning Western Neo-Aramaic will also be made. The data are mostly from NENA and Ṭuroyo grammatical descriptions 4 and fieldwork I conducted personally in the diaspora 4 For ease of comparison and accessibility, the various styles of transcription have been made uniform as follows. The reduced centralised vowel ([ɪ] ~ [ə] (~ [ɯ])) sometimes represented as <i>, <ı>, <ɨ>, <ĭ>, or <ə> is represented by the single grapheme <ə>.The voiceless and voiced interdental fricatives are marked by <θ> and <ð>, respectively, (as against <ṯ>, respectively, <ḏ> in some sources), and the pharyngeal and glottal stop by <ʕ> and <ʾ> (against half rings <ʿ> and <ʾ> in some sources). Post-velar unaspirated /k̭ /, in for example C. Urmi (Khan 2016), corresponding with /q/ in other dialects, is represented by <q> for simplicity's sake. Moreover, I have taken the liberty to adapt Prym and Socin (1881) and Ritter's (1967Ritter's ( -1971 detailed transcription of Ṭuroyo to a phonological transcription that matches NENA more closely like that of Jastrow (1992). Emphasis and glossing are mine in examples, unless stated otherwise� or in collaboration with G. Khan and/or D. Molin in Iraq 5 and with D. Molin in Jerusalem. There are notable differences and resemblances across Neo-Aramaic dialects, some of which go back to pre-modern Aramaic.
As the term used for these person markers already suggests, the L-suffixes are no longer prepositional in nature but have become inflectional suffixes. While their use in these constructions is still reminiscent of a formerly dative case, synchronically, they are no longer prepositional but serve to cross-index arguments in the clause. Obligatorisation of such cross-indexing is a well-known feature of the 'canonical' subject relation (e.g. Keenan 1976;Onishi 2001) contrary to objects, the marking of which remains conditioned by discourse-referential properties (e.g. Haig 2018a). Do these L-suffixes express a 'non-canonical' subject? To what extent have these L-suffixes become obligatory? And to what extent do they still interact with prepositional arguments? As we shall see, dialects have different strategies and not all of them operate on the same level as (2c-d) above.
First, we shall briefly review verbal inflection and how the recipient is expressed in ditransitive constructions. These findings are compared with the morphosyntax of predicative possessors and (impersonal) experiencer verb constructions in both subgroups of Neo-Aramaic.

Role Reference Inversion
Verbal person marking in NENA and Ṭuroyo is considerably complex and cannot be treated in full detail here. 6 Historically, verbal inflection goes back to participial constructions that 5 Data collection in Iraq was made possible by GCRF funding. 6 Overviews of the morphosyntax in NENA and Ṭuroyo can be found in Khan (2010), Coghill (2016, 55-101), Waltisberg (2016) and Noorlander (2018b, forthcoming). combined with clitic person markers. Two sets of person markers are used. They will be referred to as the E-suffixes and L-suffixes, which are respectively diachronically enclitic pronouns and participial agreement (E-suffixes) and prepositional pronouns based on l (L-suffixes). These are attached to the following inflectional bases. The imperfective base is derived from the active participle and the perfective base is derived from a verbal adjective that expressed result states. I will refer to themas qaṭəl-(< *qāṭel-) and qṭil-respectively after the inflection of stem I strong verbs. The NENA qaṭəl-base corresponds to Ṭuroyo qoṭəl-, where *ā has shifted to /o/ in open syllables. A so-called neuter class of mainly intransitive verbs in Ṭuroyo follows the pattern C 1 aC 2 iC 3 in the perfective, such as damixo 'she slept' for dmx� Historically, this goes back to a verbal adjective with a geminate second consonant, e�g� *dammīḵ 'asleep', which should not be confused with NENA qaṭəl-� Transitive clauses show a type of role reference inversion 7 conditioned by these inflectional bases (Noorlander forthcoming). The roles that the E-suffixes and L-suffixes refer to are different depending whether they attach to the imperfective or perfective base. This can be seen, for instance, in the following examples from Amidya (NW Iraq). While the L-suffixes mark the object in the qaṭəl-base for the verb šmʾ 'hear', they mark the agent in the qṭil-base, and vice versa for the E-series� a� k-šamʾ-i baxta ind-hear IPFV -they woman 'They hear a woman�' 7 Or "agreement inversion" (Doron and Khan 2012). See also Polotsky (1979, 209;1991, 266;1994, 95), Hoberman (1989:96, 113), Mengozzi (2002b, 44-5), Noorlander (2018b, 119-23, 129, 408-10 Prominent objects are marked differentially via cross-indexing and/or prepositional marking. The definite object in (7) below, for instance, is marked consistently by the preposition (ʾəl)l-and triggers agreement throughout the constructional qaṭəl-/qṭil-split� In (7a), however, the L-suffix attached to qaṭəl-cross-indexes the object, whereas the E-suffix attached to qṭil-does so in in (7b). (7) J. Arbel (NE Iraq; Khan 1999, 288-90)  In addition, agent focus can be expressed optionally by means of the preposition (e)l-combined with the agreement through L-suffixes in Ṭuroyo. The prepositional marking of the object and the agent are both optional. Additional cross-indexing of a prominent object is also optional in Ṭuroyo. 8 Contrast (8a) with (8b) below. Type (8b) is peculiar to the dialect of Raite (Waltisberg 2016, 186f.). Both can also be lacking altogether, as illustrated in (8c). The L-suffix that expresses the agent, however, is obligatory, cf. (8d) and (8e) below. Hence optional ergative prepositional marking is always accompanied by an agent L-suffix as illustrated in (8c).

[V-A] [l→O]
g-ḥoze-∅ l-i-dăvăre fut-see IPFV -he dom-the-breach.m 'He will find the breach (in the wall Thus both the nominal and verbal marking of objects is conditioned by the discourse salience of the argument. The verbal agreement with the agent, however, is obligatory. The prepositional marking of the agent is optional only in the preterite in Ṭuroyo.

Semi-Clitic L-Suffixes and Ditransitive Verbs
The L-suffixes show lingering features of their enclitic origin (Doron and Khan 2012, 231). First of all, they allow tense morphemes like -wa-to intervene, e�g� Secondly, verbs generally only take one object affix. There are a number of dialects, however, that allow a verb to take more than one L-suffix, i.e. to stack L-suffixes. This occurs across the constructional split illustrated above. Thus, the first L-suffix always marks the (T)heme, i.e. the entity transferred to somebody, and the second marks the (R)recipient role in the qaṭəl-base inflection. Example (10) illustrates this where the first L-suffix -nay (i�e� maxzən-+ -lay → maxzən-nay) expresses the T and the second L-suffix -lux expresses the R� This is generally only allowed when the T is third person� 9 (10) C. Marga (SE Turkey) In a number of dialects, a second L-suffix is added to the perfective to express the R. Thus we find perfective forms in dialects like C� Marga such as (11) below where the first L-suffix -li (i�e� mər-+ -li → mər-ri) marks the A, but the second one, -lux, marks the R� Moreover, stacking of L-suffixes may occur even when the T is marked by the E-suffixes such as -a in (12) below.
(12) C. Marga (SE Turkey) betrothed PFV -her -they -him 'They betrothed her to him.' This is also attested for rural dialects in Ṭuroyo (cf. Ritter 1990, 75), for example: húw -i -le -lalle gave PFV -them -he -them 'He gave them to them.' (Miden, Ritter 1967-1971 Ṭuroyo, however, prefers an unmarked set of bound person markers 10 to express third person Ts 11 when both the T and R are bound pronouns, as exemplified in (13b) below.
10 These are identical to the third person forms of the copula that historically goes back to bound person markers, e�g� e�g� ú-dawšo basímo-yo 'The honey is nice'� 11 See Jastrow (1985, 137-38), Waltisberg (2016, 296), Noorlander (2018b, 341-45 In addition, a prepositional indirect object construction is available to all persons as well as all types of full nominals. Various dialect-dependent prepositions are used to mark the R independently of the verb. The respective preposition will vary significantly across as well as within dialects. Variants of the preposition (ʾəl)l-still occur, such as: ʾát-tarʿone mər-re l-ú-malko the-doorkeeper�mpl said PFV -they to-the-king�ms 'The doorkeepers said to the king� (Ritter 1967(Ritter -1971 The prepositional recipient NP can trigger additional agreement by L-suffixes on the verb, to illustrate:

b.
Gorgis k-omar-∅-re l-áb-baqore Gorgis prs-say IPFV -he-them to-the-cowherder�mpl 'Gorgis says to the cowherders.' (ibid. 115/164) Several NENA dialects, however, make use of other (novel) prepositions such as ṭ(l)a-, ta-, ba(q)-, qa-etc., for example: Prepositional marking of the R is preferred when the T is a first or second person pronominal affix attached to the verb and when the R is a full nominal. 12 Thus, full nominal recipients are generally prepositional� An extra L-suffix can express pronominal recipients in both NENA and Ṭuroyo for both the qaṭəl-and qṭil-based person marking� When the verb selects an additional L-suffix, it is confined to recipients found throughout the verbal system in Ṭuroyo and several NENA dialects. Third person themes can be marked through a different, unmarked set of bound person markers.
When such additional L-suffixes of the first and second person are added to qṭil-in Ṭuroyo, they also express the object of monotransitive verbs, e.g. grəš-le-li 'He pulled me'� One cannot say **grəš-li-le for 'I pulled him' (e.g. Noorlander 2018b, 340). Generally, NENA dialects do not add such object L-suffixes to qṭilforms. Jewish dialects in Iranian Azerbaijan, however, such as Urmi and Salamas and several Christian dialects in SE Turkey such as Bohtan (Ruma; Fox 2009), Haṣṣan (Jastrow 1997; Damsma forthcoming), Umṛa and Jənnet (Noorlander field notes) use the L-suffixes for objects throughout the qaṭəl-/qṭil-split, i�e� grəš-li-le 'I pulled him', cf� garš-ax-le 'We pull him'� 13 The preterite illustrated in (6) above is known as the šmīʿ l-or qṭīl l-construction in Aramaic studies. Historically, it goes back to the resultative participle and an agent-like argument marked by l-� It developed from a stative-resultative to a preterite via a perfect. Views diverge as to its exact interpretation. It has been connected with possessors, experiencers and subject co-referential datives� 14 It lies beyond the scope of this article to address this issue here. It should be noted, however, that, while a connection between these 'non-canonical' subject construction types and the šmiʿ l-constructions developing into the preterite seems plausible to me in itself, we shall see that there are important distinctions� Forms like grəš-li 'I pulled' consist of L-suffixes that are marked for tense-aspect. They serve as inflectional agent suffixes of the preterite based on qṭil-. This is a notable distinction from the use of L-suffixes to express affectees, since they are found across different inflections and not just the qṭil-based forms. This difference is observed above for the recipient role but also extensions thereof that are the relics of a formerly dative argument.

Beneficiaries
Apart from recipients of ditransitive verbs, L-suffixes can be added to any monotransitive verb to express an additional R-like affectee, as if it were an additional argument of the verb. The Ṭuroyo L-suffix -lən in (16), for example, expresses a beneficiary in a construction that is clearly derived from ditransitive constructions. The same holds for -li in (17) below to illustrate this for lišana deni dialects of NENA like J� Dohok: He opened the door for them.' (Ritter 1967(Ritter -1971/237) He opened the door for me.' The T-like argument can be pronominalised through the same unmarked set as in ditransitive constructions added to the L-suffix expressing the beneficiary in Ṭuroyo, e.g.
(18) Ṭuroyo (Midən, SE Turkey) '(From a ball of threads) she made me a pair of stockings.' (Jastrow 1992, 138.12) Indeed, both the A and the R-like affectee can be l-marked and cross-referenced by L-suffixes� 15 The l-marking of the A is pragmatically conditioned (agent/narrow focus), for example: Towards a typology of possessors and experiencers in Neo-Aramaic '… what the sultan has done to the emir�' (Ritter 1967(Ritter -1971, 36/87)

Subject Co-referential L-suffixes
An additional R-like argument expressed by the L-suffix can also denote an interested party, indirect affectee or benefactor that is co-referential with the subject. This is found across the verbal system for many telic dynamic verbs, including Indeed, co-referential L-suffixes have become special (stressed) inflectional endings in the high frequency motion verb ʾzl 'go' as well as the imperative forms of ʾθy 'come' in Ṭuroyo, replacing the original subject encoding. Because of this, the verb ʾzl has an irregular and unique inflection that is identical to the L-suffixes except for the 2pl� and 3pl�, which take special endings, as shown in (22) below.
Presumably the final /l/ of the original root ʾzl played a role, yielding special endings because of the complete assimilation with the preceding /z/. The 3s forms can be enhanced with -yo, which mimics its use in ditransitive constructions and creates a penultimate stress as in the first person -no in forms like k-əzz-i-no 'I'm going' and k-əzz-an-o 'We're going'. Subject co-referential L-suffixes can even be added instead, e.g. Ritter 1967Ritter -1971 Importantly, no such conjugations are attested for ʾzl in the closely related Central Neo-Aramaic dialect Mlaḥsó� The imperfective and imperative do not take L-suffixes, e.g. ∅-oz-ina 'Let's go' and iz-ewun 'Go PL !'. The L-suffixes function as subject markers for the preterite, e�g� preterite azi-le 'He went', against the perfect azi-∅ 'He has gone' (Jastrow 1994, 156). Only the pl� imperative of ʾsy 'come' in Mlaḥsó, e.g. toxun 'Come PL !' does seem to parallel Ṭuroyo toxu� Subject co-referential datives also occur in NENA dialects. This is, for instance, common in the imperative of motion verbs 17 , e�g� C� Urmi ta-lux 'Come MS !, si-lux 'Go MS !' (Khan 2016 II :151-52). It can also combine with other verbs and verbal forms expressing a beneficiary, e�g� šqul-lux xa-dana ʾərba 'Take a sheep for yourself' (ibid. 152), zon-i-lay mexulta 'They buy themselves food', zvun-nux xaql-i 'Buy MS (yourself MS ) my field!' (Polotsky 1996, 37, transcription modified).
The verb ʾzl is also highly irregular in Christian NENA dialects in SE Turkay and northern Iraq, especially on the Mosul plain� Both the qaṭəl-base and qṭil-base take L-suffixes as subject coding, as shown in (25) below, including after the 'past convertor' -wa, e�g� k-zá-wa-la 'She used to go'. Khan (2002, 120) assumes the base za-is a reduced form of the infinitive ʾəzála� Note also that the imperative of ʾθy 'come' has similarly irregular forms inflected with L-suffixes. (The imperative of ʾzl does not take L-suffixes in this dialect.) (24) C. Qaraqosh (NW Iraq; Khan 2002, 120, 153, 155, 122) imperfective In Western Neo-Aramaic, subject co-referential L-suffixes are readily found in the imperative, e.g. zubnu-llxun 'Buy MPL yourselves (sth.)!', and are common with the verbs of motion ʾty 'come' and zyl 'go', and with the change-of-state verbs qʿy 'sit' and ðmx 'sleep, fall asleep' (Arnold 1990b, 238, cf. Spitaler 1938: The imperative forms of ʾθy 'come' is thus regularly fused with L-suffixes in Western Neo-Aramaic (Arnold 1990b, 173) similarly to Ṭuroyo and NENA dialects on the Mosul Plain: Subject co-referential datives (or ethical datives) were already common with such intransitive verbs in pre-Modern Aramaic and can be considered an archaic feature in Neo-Aramaic, e.g. qum leḵ! 'Arise FS !' qåm-∅ l-eh 'He has risen' (see Fassberg 2018;cf. Joosten 1989). Fassberg (2018), following Ullendorff, argues the so-called ethical dative reflects the colloquial language. Several scholars claim the ethical dative influenced the emergence of intransitive verbal forms inflected with L-suffixes like qəm-li 'I rose' in NENA and Mlaḥsó (Mengozzi 2002b, 44;Halevy 2008;Fassberg 2018, 115). While this is conceivable, one should note that this dative endured as additional L-suffixes in the spoken varieties and did not disappear as a result (pace Fassberg 2018, 116). Moreover, where the original dative pronominal is conventionalized as inflectional morphemes of the verb, it is attested across the inflectional system, and thus not an inflectional property of qṭilas verbal form per se�

Trans-Zab Jewish Dialects: əll-series
A different strategy comparable with subject co-referential L-suffixes exists in the so-called Trans-Zab Jewish dialects of NENA (Mutzafi 2008b). Certain intransitive verbs can take bound person markers derived from the independent set based on the preposition ʾəll-, constituting a secondary LL-series. They are impersonal, dummy pronouns belonging to the 3ms. or 3fs. in intransitive predicates functioning like a middle voice marker (Mengozzi 2006). They are not co-referential with the subject but seem to express the telic endpoint, for example:

Morphosyntax of Possessors in Neo-Aramaic
Possession can be expressed in various ways in Neo-Aramaic languages (Noorlander 2018b, 154-58). 18 The focus here will be on the possessor marking strategies that are related to the original dative preposition l-. I should note briefly, however, that possession can be expressed adnominally by means of nominal suffixes, e.g. 'This garden belonged to the king', lit� 'was the king's' (Midyat, Ritter 1967-1971).

Possessor Marked by L-suffixes Only
Predicative possession is based in existential clauses introduced by the dialectal reflexes of the existential marker *ʾiθ-'there is/are'. This uninflectable particle is negated by the negator la (in NENA and Ṭuroyo) in a form going back to *la-yθ-'there is/are not', and for past tense by the suffix -wa, e�g� *ʾiθ-wa 'there was/were' (in NENA and Ṭuroyo), similarly to verbs. The preverbal TAM-marker k-typical for the indicative-durative present is always combined with it in Ṭuroyo, e.g. k-ito 'There is'. Together with L-suffixes they express predicative possession, e�g� kət-li 'I have'� In Western Neo-Aramaic, the existential particle is reduced to ī-or ū-before L-suffixes, e.g. ī-le 'He has' (Arnold 1990a, 185). The negator is čū and the past particle is wa preceding the predicate, e�g� čū-le 'He has not', wa ī-le 'He had'. The L-suffix in Neo-Aramaic marks the possessor which is reminiscent of their use as markers of the recipient (i.e. 'T belongs to R'). The co-referential nominal, however, is usually not prepositional. Thus, (30a) below presents a simple existential predicate in Ṭuroyo. (30b) illustrates the additional L-suffix expressing the R-like possessor without a co-nominal referent. In (30b), the possessor NP ú-malk-ano 'this king' is zero-marked but the L-suffix cross-references it, indexing its role as the possessor. The unmarked set of independent pronouns is similarly used to express the possessor, as illustrated in (30c). Possessors are generally also expressed by an L-suffix in Western Neo-Aramaic such as ī-le 'He has' and ī-l 'I have' in the following examples� The possessor co-nominal itself is zero-marked� When L-suffixes are attached to the verb hwy, the construction can semantically entail a process, i.e. 'become', rather than a state, i.e. 'be'. The verb can be used to convey 'be born'. The L-suffix denotes an R-like affectee, i.e. the one who begot the child, for example: Pronominal objects are otherwise not marked through L-suffixes on the qṭil-based preterite verb in such dialects. Forms like **grəš-la-li for intended 'She pulled me' do not occur. The secondary L-suffix is clearly reminiscent of the stacking of L-suffixes in ditransitive constructions in dialects like C. Marga and lišana deni Jewish dialects, cf. (37) below. This indicates how the L-suffix is considered an R-like argument in the system and expressed by an L-suffix regardless of the inflectional base, cf. (37a) below taken from the Jewish dialect of Dohok. The verb hwy can also be inflected for person and combine with the L-suffix not to convey a strict sense of belonging but a broader sense of relation, i.e. 'X is/becomes Y with respect to somebody'. The construction parallels ditransitive verbs. Only in this sense can the pronominal possessee be expressed in the same way as the theme in ditransitives such as -yo , for example in (36b): The expression of ability through predicative possessors is also recorded in Trans-Zab Jewish varieties (e�g� Khan 2004, 311, 364), ma ʾit-wa-lu hol-i-wa 'What could they do?', ʾana kwe-li 'I will be able'.

3�2�2� External Possessors
Sporadically, L-suffixes can express a possessor-like affectee of verbal predicates akin to example (1g) from Hebrew� At least one such instance where the secondary L-suffix marks an external possessor is attested in Mlaḥsó: (39) Mlaḥsó (Lice, SE Turkey) ṭafloki mís -le -li a�child died PFV -he -me 'One child of mine died (on me).' (Jastrow 1994, 124.121) Such external possessors are also attested in NENA dialects where the second L-suffix marks the R in qṭil-� The possessor is added as an R-like affectee in both the qaṭəl-and qṭil-based inflection such as the construction in J. Dohok given in (40). Its usage in J. Dohok does not seem to have a clear distribution. Coghill (2019, 368) notes that apart from pyš 'remain', it is confined to telic intransitives in C. Telkepe (NW Iraq), such as myθ 'die', ʔθy 'come' and bry 'happen'� An L-suffix denoting an R-like argument can be added to intransitive verbs in Ṭuroyo. It can be combined with the verbs fyš 'remain', qyθ 'hit, touch, meet', ʾθy 'come' and mṭy 'arrive'� Since these motion verbs denote movement towards an endpoint, these constructions typically convey a sense of reception, e.g.

Trans-Zab Jewish Dialects: əll-and -la-l-series
Occasionally, one also finds prepositional external possessors in NENA attached to the verbal base. An LL-series of person markers based on the preposition (ʾəl)l-is used to express the external possessor as illustrated for J. Arbel below.
(43) J. Arbel (NE Iraq; Khan 1999, 292) yāle rāba míl-lu-llaw children very.much died PFV -they-her 'Many of her children died�' Western Iranian dialects such as J. Saqqiz and J� Sanandaj use the morpheme -la-20 as base for the L-suffixes to express predicative possession together with the verb 'become' (Khan 2009, 88-90, 301-02). This la-is possibly a relic of a former impersonal L-suffix -la 'it F ', i�e� xír-la-li 'It F became to me' → 'I have'. Full possessor NPs are zero-marked and can occupy preverbal position as illustrated in (44) below. The verbal base xir is invariable like the existential marker and does not agree with, for instance, indefinite plural nouns such as puḷe 'money' in J� Sanandaj ʾaná hămešá puḷé xír-la-li 'I have always had money' (Khan 2009, 302). This lal-series, therefore, serves as a special set of person markers, identifying their role as the most salient affectee� (44) 20 Khan (2009, 89) notes "the element la-is likely to be a fossilised form of a 3fs. copula form *ila". It also possible it is an L-suffix used impersonally.

Possessor Marked by l-Only
The independent possessor argument is generally zero-marked in NENA� Alternative expressions do exist where the possessor is prepositional in some varieties of NENA such as J. Sulemaniyya combined with a 3ms. copula in (33b) below.
The house belongs to him.' (Khan 2004a, 336, 362) Similarly, sporadically, a predicative possessor can be expressed independently by means of the preposition (e)l-in Ṭuroyo, e.g.

i-dŭkθo kul-a el-ŭx -yo
the-place�fs all-her to-you.ms -it�is 'The whole place belongs to you FS (Midən, Ritter 1967-1971 Unlike the rest of Neo-Aramaic, however, the predicative possessor is always independent in Mlaḥsó� The possessor is expressed as an independent dative (pro)noun such as eli 'to me' in (47). The possessee controls the agreement of the verb hwy 'be'. Jastrow (1994)  Note that, in these cases, the possessor is marked only by a preposition just like the examples from Hebrew in (1).

Possessor Marked by l-and L-suffixes
The possessor can be optionally marked through the preposition l-in addition to the L-suffix in Ṭuroyo. 21 This includes predicative possessors such as (49a) and R-like affectees such as (49b-c). The optionality of the prepositional marking of the possessor alongside the L-suffix is reminiscent of the morphosyntax of agents in the Ṭuroyo qṭil-based preterite (cf. Diem 2012). This strategy to combine the preposition l-and L-suffixes does not occur in NENA.

Transitivisation of Possessive Constructions
Predicative possessive constructions have undergone transitivisation in NENA and Ṭuroyo in that the L-suffixes are obligatory person markers like verbal inflection. Apart from the L-suffixes, the construction remains impersonal� The possessee does not control agreement and does not trigger differential object marking. Generally speaking, even when a possessee could still be contextualised through anaphora such as where English would use a pronominal object for 'to have', it will tend to remain implicit in Neo-Aramaic� Forms like ʾət-li or kət-li could also mean 'I have it F /it M ' or 'I have them'. This raises the question of how transitivised the predicative possessive construction are in being compatible with pronominal objects like transitive clauses in general� There are indeed cases where the transitivisation seems to be more advanced and pronominal objects are overtly expressed� This, for instance, applies when the possessee is first or second person. First and second person pronominal objects differ across dialects. The possessee can be expressed as a pronominal object either through the unmarked set of independent pronouns, e�g� This parallels the use of independent personal pronouns in transitive clauses to express focal objects, for example: b.

'We saw YOU FS '
If available, the possessee can also be expressed through a dedicated set of prepositional pronominal objects, e�g� In the latter, the marking of the possessee clearly patterns like that of objects of qṭil-based preterite verbal forms where the agent is expressed by the L-suffixes, e�g� b.
ʾaxnan qatux xze-lan we you saw PFV -us 'We saw you MS ' A few NENA dialects in SE Turkey such as Artun (Hertevin), Umṛa and Jənnet mark the object on the transitive qṭil-based perfective by means of additional L-suffixes, e�g� grəš-le-la 'He pulled it F '� The marking of the possessee is the same as the object in the predicative possessor construction, e.g. ʾət-le-la 'He has it F '. It has taken over the full agent and object marking morphology of the perfective (see the examples below). When object L-suffixes like -la 'it F ' are added to grəš-lax 'You FS pulled', first and second person agents are marked by a special set one could call the L-E-series yielding grəš-lət-ta 'You FS pulled it'. 23 The same transitive verbal coding occurs in the predicative possessor construction, e.g. ʾət-lət-ta 'You FS have it F '� Moreover, these transitive constructions are used when full nominal possessees trigger differential marking. Thus the indefinite possessee in (52a) functions like an indefinite object in (52c)but the definite possessee in (52b) triggers cross-indexing like a definite object in (52d).
(52) C. Artun (Hertevin, SE Turkey;Jastrow 1988, 67, 160.541-542) a� ana lət -li hay It should be noted, however, that this is not acceptable in the majority of dialects� Speakers of J� Dohok, for example, do not readily accept pronominalisation of the possessee in predicative possessor constructions. They disfavour expressions like **ʾətli ʾahat 'I have you FS ' and circumvent this by choosing constructions involving independent possessive pronouns akin to English 'You FS are mine'�

Verboid bas-'enough'
A related verboid construction in NENA based on the particle bas-'enough' is generally inflected with suffixes going back to possessor-like L-suffixes that have assimilated to the preceding /s/. The possessee-like complement of the quantifier bas, i�e� that which is possessed in a satisfactory amount such as xaye 'life' below, is prepositional (m-), e.g.

bass-i m-xáye
enough-me from-life�pl 'I have had enough of my life' (Khan 2008a(Khan , 1241.
The original L-suffix can still be observed in the past equivalent, e�g� bas-wa-li 'I had had enough'. Depending on the dialect, the copula can also be added to this to express the referent of the quantifier bass-, e�g� 'That F is enough for you.' (Khan 2016 I , 585) The same particle is fully inflectable for L-suffixes in Ṭuroyo, as illustrated below. Unlike (53) above, the possessee is not prepositional but zero-marked or expressed by a copula: The structure is at least superficially similar to ditransitives in that the T-like person markers are identical with the copula�

Morphosyntax of Experiencers in Neo-Aramaic
There are numerous ways in which experiencers are encoded in Neo-Aramaic. One should note that some of the constructions discussed in what follows also have equivalent expressions in other dialects involving a different structure. In impersonal experiencer constructions, for instance, experiencers can also be expressed adnominally through agreeing possessive suffixes, e.g.

brat-í qard-ev-ya
daughter-my cold�fs-her-it�is�f 'My daughter feels cold.' (lit. Her coldness is) (Israeli 1998, 170) This is an areal phenomenon found across languages in West Asia, including the Neo-Aramaic speaking area. 24 It regularly features in neighbouring Iranian varieties where the experiencer is marked in the so-called 'oblique' case or through pronominal clitics that also denote the possessor and the agent in the past (Haig 2018b(Haig , 132-33, 2018c There are cases where the experiencer is expressed as the object. For example, the verb ʿjb 'please, like', borrowed from Arabic, takes object suffixes in Western Neo-Aramaic just like the corresponding verb in Arabic, e.g. (60) Western Neo-Aramaic (Maʿlula, NW Syria) ana aʿžb-īš-n I pleased-you.fs-me 'I like you FS .' (Arnold 1991, 140.42) 24 See Khan (2016 II , 355-59) It will become clear, however, that experiencers are construed as R-like affectees similarly to possessors in the previous discussion.

4�1�1� Transitive Verbs
Apart from beneficiaries and predicative possessors, L-suffixes can denote experiencers. In several (Christian) NENA dialects (and Ṭuroyo), verbs like I bsm, II/III ʿjb and I hny (variants include nny and nhy) are impersonal experiencer predicates conveying more or less the equivalent to English 'like', 'please' or 'enjoy', as illustrated for Ṭuroyo and C. Barwar below.
When the experiencer verb is impersonal apart from the L-suffix, it takes non-referential 3ms� or 3fs� morphology� Unlike NENA, the verb bsm is stative 'was pleasant/nice' or inchoative 'became pleasant/nice' in Ṭuroyo. 25 It takes the C 1 aC 2 iC 3 -pattern in the perfective typical for non-referential 3ms� morphology, for example: was�pleasant PFV -it�m-him the-sitting�fs of-before 'He (lit� Him) enjoyed (once again) sitting idly like earlier times.' (Midən, Ritter 1967-1971 The stimulus can be pronominalised like themes in a ditransitive construction, such as -yo in the following example: It would seem that there are also constructions where -yo is effectively non-referential. This is at least the case in fixed expressions of the following kind:  (Ritter 1967(Ritter -1971 The stimulus, however, can still control agreement and be referential, as is the case with the stimulus of lzm 'need' (56d-e) below.
for-what pvb-need IPFV -they -me 'What do I need them (i.e. gold pieces) for? ' (Midən;Ritter 1967Ritter -1971 e� ú-yawmo d-lŭzm -at -lan itŭx the-day�ms rel need IPFV -you.s -us come�imp 'Come the day we need you S !' (Midyat, letter, Ritter 1990, 207) In NENA, the qṭil-based form of the experiencer predicate inflects for two L-suffixes such as (64b) and (65b) below. The first represents the impersonal coding, which is expressed by the E-suffix in the qaṭəl-based forms in (64a)  The only example known to me where Western Neo-Aramaic has similarly grammaticalised an experiencer L-suffix is the verb 'want' in the dialect of Jubbʿadin. The L-suffix attaches to an uninflected form be-, e�g� bē-le (< *bʿē l-eh 'Him wanted'), the originally 3ms. form of the resultative participle *bʿē of bʿy 'want' (Arnold 1990a, 192). bēle (like batte in the other Western dialects) developed under influence of the corresponding construction bədd-o 'He wants' < 'In his wish' in local Arabic varieties� The experiencer nominal is zero-marked and controls the agreement expressed by the L-suffix: Verbal experiencer predicates can comprise an NP denoting the mental state or process somehow reaching the experiencer expressed through the L-suffix as illustrated in (71). Note that in (71a) and (71b) the verb does not agree with the NP and is essentially impersonal. The key person marker being the L-suffix. aθi-∅-li ḥəmto qwiθo came PFV -it�m-me fever�fs heavy.fs 'I caught a heavy fever.' (ʿIwardo; Ritter 1967Ritter -1971 The experiencer can be added to intransitive verbs denoting physiological states such to 'be cold' in various dialects in SE Turkey. Thus the expression 'I am cold' corresponds with: There are several more intransitive verbs in Ṭuroyo that can express an experiencer in this way, notably kyw 'get ill', e�g� kayu-li 'I got ill', and nyḥ 'get well', nayəḥ-li 'I got well'. The verbs ḥrw 'be concerned' (lit. 'get destroyed') and ʿyq 'get distressed' combine with an additional prepositional stimulus. The verbal form is impersonal, for example:  (1998, records numerous examples where verbs in the qaṭəl-based inflection regularly combine with experiencers expressed in this way including verbs denoting pleasure such as bsm 'please' as illustrated below but also verbs denoting pain mry 'hurt', capability kšy 'find difficult' and merit such as mṭy 'deserve' (lit. reach) and špr 'befit'. The construction combines with a prepositional stimulus or a clausal complement.

4�2�1� Experiencer Marked by a Preposition only
So far we have observed that the experiencer NP is zero-marked like the 'canonical' subject and only expressed through L-suffixes on the verb. Nevertheless, prepositional marking of experiencer predicates does occur in several NENA dialects, reflecting an oblique status. It is common for physiological states� The independent ʾəllseries is part of fixed expressions for the sensations of heat and cold which themselves feature as nouns in this construction, for example in C� Marga:  Ritter 1967Ritter -1971

Transitivisation of Experiencer Verb Constructions
The verb ʿjb 'please, like', borrowed from Arabic, is a stem III causative verb in Ṭuroyo and is ambivalent as to its orientation. The verb of liking can be directed at the R-like affectee expressed by the L-suffix, for example:  Ritter 1967Ritter -1971 At the same time, the verb can also have undergone complete transitivization� Its coding is not distinct from primary transitive verbs. The experiencer is expressed like an agent, for example: b.
ú-greʿuno d-həzy-o-le maʿajb-o-le the-youngling.ms rel-see IPFV -she-him III:pleased PFV -she-him 'The young man that she sees (and she) likes' (Miden,ibid. 75/199) This also occurs in NENA dialects. At least in the preterite, the verb that is otherwise typically impersonal can also be used with 'canonical' transitive verbal coding, for instance in C. Urmi and C. Artun

Verboids
A few experiencer verbs have a distinct verbal base in the imperfective, comparable to the verb ʾzl in some NENA dialects (see Subsection 2.2), e.g. zəl-wa-li 'I had gone' (perfective) and k-za-wa-li 'I used to go' (imperfective). The verb zdy ~ zdʾ 'fear, be afraid' has a regular qṭil-based preterite construction, e.g. J.

Conclusions
Both the possessor and experiencer nominal or independent pronoun are generally clause-initial, zero-marked and obligatorily cross-referenced by the L-suffix in both NENA, Ṭuroyo and Western Neo-Aramaic. They are arguably 'non-canonical' subjects. Only sporadically do we find purely prepositional arguments. L-suffixes can be added to monotransitive and intransitive verbs to express an R-like affectee in similar fashion to ditransitive verbs. While the optional subject co-referential L-suffixes marking that can mark an affected subject like the middle voice or express dynamic telicity seem to be generally a common Aramaic phenomenon, they undoubtedly conventionalized to verbal inflectional morphemes in certain Neo-Aramaic languages, particularly the motion verbs *ʾzl 'go' and *ʾty 'come'� Impersonal experiencer constructions tend to diverge across dialects. It is common to find that verbs of liking take 'noncanonical' subject marking besides physiological states of 'cold' and 'heat'� Dialects can prefer distinct strategies for these physical sensations. In SE Turkey, for example, the experiencer of 'cold' is expressed by L-suffixes attached to a verbal predicate, while that of 'heat' by a preposition as a complement of a nominal predicate� The Neo-Aramaic languages have developed 'non-canonical' subject marking that exhibits similar structures as the agent in the perfective past in NENA and Ṭuroyo (e.g. grəš-li 'I pulled'). The 'non-canonical' subject, for instance, can be marked by both the preposition l-and L-suffixes in Ṭuroyo only. This closely parallels the optional ergative marking in the preterite� An important difference from agent L-suffixes in the preterite is that the L-suffixes that mark the 'non-canonical' subject are found across the inflectional system, just like other R-like affectees. Exceptions where the 'non-canonical' subject marking is confined to the imperfective are the verbs 'fear', which has a verboid base ṣad-, and the verb 'go', which has a base za-, in NENA dialects� These correspond with the 'canonical' verbal inflection in the preterite (e�g� zəl-lan 'We went' : za-lan 'Let's go!').
One important difference, however, is that the L-suffixes of the preterite are dependent on the inflectional base qṭil-and have an additional TAM function. This does not apply to the other uses of the L-suffixes that were subsumed under 'noncanonical' subjects in the previous discussion that can still be more R-like. The 'non-canonical' subject marking, therefore, is role-based. It is the construal as an R-like indirect affectee that makes it favour coding distinct from the 'canonical' subject. By contrast, the agent marking through L-suffixes in the preterite is not only role-based but also TAM-based. That is, the originally dative agent is dependent on the inflectional base (qṭil-) and hence, generally, perfective past aspect� Occasionally, however, the 'non-canonical' subject undergoes full transitivisation and takes over 'canonical' transitive coding� Sometimes it is only the transitive morphosyntax peculiar to the qṭil-based preterite that is taken over, identifying the L-suffixes that mark the possessor or experiencer with those that mark the agent� Nevertheless, it is also clear that in many cases where the L-suffix is used as an extension of an R-like affectee the suffix can maintain characteristics of a ditransitive construction. These subject-like affectees are still treated like recipients, presumably as relics of their formerly dative prepositional marking� This is evident in the stacking of L-suffixes to the qṭil-base in NENA where the first L-suffix is impersonal and the second L-suffix denotes the R-like affectee. Impersonal experiencers thus resemble the predicative possessor construction based on the invariable existential marker (cf. Polotsky 1979, 209-10), yet, since they are verbal, they select the regular verbal affixes, even L-suffixes expressing the impersonal agent in the qṭil-based forms (e�g� ʿjəb-le-le ʾalaha 'It M pleased God'). Pronominalisation of the stimulus can be expressed by the unmarked set of bound person markers (also serving as the copula) like -yo in Ṭuroyo, which are confined to third person themes in ditransitive clauses.
The topicalisation and hence zero-marking of the NP became increasingly obligatory and original independent prepositional pronouns have undergone complete verbalisation in most cases� L-suffixes, while originally prepositional and independent of the verb, exhibit a tendency to convert into verbal person markers and sustain referential continuity with the most topical argument in sometimes otherwise largely impersonal predicates� Both more conservative and more innovative patterns are found in Neo-Aramaic� Dialects also have the option to withstand the proclivity to convert a topicalised affectee into a 'non-canonical' subject. A dialect may still prefer to retain prepositional marking as a viable alternative besides verbal person marking or it may prefer an oblique status throughout for such arguments. In the end, each dialect 'can do its own thing' and a uniform category of 'subject' is not always readily identifiable.