Stammbaum or Continuum? The Subgrouping of Modern Aramaic Dialects Reconsidered Ronald Kim

1. INTRODUCTION

    The terrible and still largely unknown series of persecutions suffered by speakers of Modern Aramaic (MA) outside of Syria and their emigration en masse from their homelands mean that most MA dialects today are either extinct or highly endangered. Under these circumstances, all researchers in MA have been encouraged to concentrate on the urgent task of fieldwork, so that we might be able to record as many texts and write as complete grammatical descriptions of as many dialects as possible (see Krotkoff 1985: 133).
Yet although many varieties still remain to be documented, we now have a far fuller picture of MA dialectal diversity than was available twenty or even ten years ago. Jastrow’s monographs on Hertevin and Mlahsô (1988, 1994a) were followed by preliminary reports of previously unknown Christian MA dialects of southeastern Turkey (1994b, 1997b) and Sinha’s grammar of Bespan and the neighboring Cudi dialects (2000). A number of studies have begun to fill in the blanks in our knowledge of the pre-1915 dialects of the Mountain Nestorians (Bergnestorianer) of the Hakkâri district, including Talay’s descriptions of the dialects spoken by their descendants living along the Khabur River in northeastern Syria (1999, 2008), and Fox’s research on Jilu (1997) and Bohtan (2002). Hoberman 1989 and more recently Khan 1999a, 2002a, 2004, and Mutzafi 2004b have given us detailed treatments of [Graphic Character Omitted]Amadiya (Jewish), Arbel (Jewish), Qaraqosh (Syrian Christian), Sulemaniyya and Halabja (Jewish), and Koy Sanjaq (Jewish) in Iraq, respectively; while Sabar’s 2002 dictionary provides a wealth of invaluable data for the Jewish dialects of Zakho and neighboring villages in northwestern Iraq.
As for the MA of Iran, Younansardaroud (2001) provides a precise description of a dialect very close to the Christian speech of Urmia, the basis of the Modern Syriac literary language. The previously all-but-unknown dialects of Iranian Kurdistan are now becoming accessible, thanks to the work of Hopkins (1989, 2002) and others on the Jewish dialects, and of Panoussi (1990) and Heinrichs (2002) on Senaya, the language of the Christians of Sanandaj (Aramaic Sena). Further monographic descriptions are due to appear shortly, e.g., from Geoffrey Khan’s students at Cambridge University, who have conducted fieldwork on other Iraqi MA varieties.
The information available on MA dialectology, then, has increased so markedly over the past generation, and especially the last decade, that one need not apologize for incorporating new findings into a comparative and historical study. Indeed, comparative investigations can play an essential role in determining which grammatical features or patterns need to be the focus of further studies or of fieldwork on unrecorded dialects.(FN1) The time therefore seems ripe for a reevaluation of earlier views on the historical development and dialectal diversification of present-day Aramaic.

2. THE SUBGROUPING OF MA DIALECTS

2.1. PREVIOUS TREATMENTS
The oldest and most fundamental classification of MA distinguishes between two groups: Modern West Aramaic (MWA), spoken in the three villages of Ma[Graphic Character Omitted]lula, Bax[Graphic Character Omitted]a (today officially Sarxa), and Gubb[Graphic Character Omitted]adin (MWA Guppa[Graphic Character Omitted]d) in the Qalamun valley of the Antilebanon mountains, north of Damascus; and Eastern MA, comprising all the other dialects of Syria, Turkey, Iraq, and Iran. Such a bifurcation makes sense geographically, as the Qalamun is close to Palestine, home to the three major Western Middle Aramaic languages of the 1st millennium A.D.: Jewish Palestinian, Christian Palestinian, and Samaritan. With respect to the isoglosses distinguishing Western and Eastern Middle Aramaic dialects of the 1st millennium A.D., MWA clearly sides with the western dialects, whereas the rest of MA has participated in the innovations of the eastern group.(FN2)
1. 3sg. m. y-in the subjunctive (continuing the old imperfect), vs. eastern n- ˜ l-: Ma[Graphic Character Omitted]lula batte yidmux ‘he will sleep’ vs. Syriac nedmok.(FN3)
2. Masculine plural -o (in variation with older -oya in Gubb[Graphic Character Omitted]adin) ˜ -o(y)-before pronominal enclitics < MidAr *-ayya (status emphaticus), vs. eastern -e:(FN4) Ma[Graphic Character Omitted]lula tar[Graphic Character Omitted]a ‘adoor,’ pl. tar[Graphic Character Omitted]-o ‘doors,’ tar[Graphic Character Omitted]-oy-e ‘his doors’ vs. Syriac kalb-a ‘dog,’ pl. kalb-e ‘dogs’ (with b from the sg., like Turoyo kalb-e, C. Urmia koelb-i ‘dogs’; cf. J. Zakho kalw-e < *kalb-e).(FN5)
3. Infixed -nn- between imperfect verbs and accusative pronominal suffixes, vs. eastern -0-: Ma[Graphic Character Omitted]lula subj. yit[Graphic Character Omitted]un ‘that he carry,’ ytu[Graphic Character Omitted]n-enn-ax ‘that he carry you (m.)’ vs. Syriac net[Graphic Character Omitted]an, net[Graphic Character Omitted]n-ak.
4. Retention of the semantic contrast between status absolutus and status emphaticus (e.g., *malk ‘king,’ *malk-a ‘the king’), vs. loss of the latter’s determinative function in the east (cf. Syriac malk-a ‘(the) king’). MWA has generalized the status emphaticus in nouns, but the contrast survives in adjectives, e.g., Ma[Graphic Character Omitted]lula psona iz[Graphic Character Omitted]ur ‘(a) small boy’ vs. psona z[Graphic Character Omitted]ora ‘the small boy’; the status absolutus is used after numerals as a count plural, e.g., Ma[Graphic Character Omitted]lula tlota tara[Graphic Character Omitted] (tar[Graphic Character Omitted]i) ‘three doors’ < m. pl. *-in, etlat warkan ‘three pieces of paper’ < f. pl. *-an.(FN6)
In addition, the old West Semitic perfect and imperfect (i.e., suffixed and prefixed conjugations) are retained to this day in MWA, but have been lost everywhere else.(FN7)
This two-way split, which goes back to the earliest days of MA studies, remains the principal framework for more recent treatments of MA dialectology: For example, Jastrow’s survey article (1997a) begins with a description of MWA, then proceeds to examine the eastern dialects together as a group. Nevertheless, scholars have long been aware of the enormous diversity within Eastern MA, which includes such mutually unintelligible languages as Turoyo, standard Modern Syriac based on the Christian dialect of Urmia, the Jewish dialects of Iranian Kurdistan, and Modern Mandaic. Nöldeke (1881: 675) first emphasized the separation between Turoyo and the MA dialects east of the Tigris, based on features such as the following:
1. MidAr *a is unconditionally raised and rounded to o (phonemically/o/) in Turoyo and Mlahsô, but not in NENA.(FN8)
2. Old geminate consonants are simplified in Turoyo and Mlahsô (i.e., MidAr *VCC > *VC [V:C]), but retained in NENA. Cf. MidAr *gilla ‘grass,’ *[Graphic Character Omitted]ezza ‘goat’ < Turoyo gelo, [Graphic Character Omitted]ezo, but Hertevin gella, [Graphic Character Omitted]ezza, C. Urmia gilloe, [Graphic Character Omitted]zza.(FN9)
3. MidAr *x (i.e., *k, the spirantized postvocalic allophone of *k) and *h are preserved in Mlahsô and Turoyo, but merged as h in Hertevin and x elsewhere in NENA (Jastrow 1988: 6, 1997a: 349).
4. NENA, including the dialect of Hertevin, has undergone certain innovations in the pronominal system not shared by Turoyo (Hoberman 1988: 572):
a. The spread of initial a to the 3rd person independent pronouns: cf. Hertevin sg. m. ahu, f. ahi, pl. ahni, J. Zakho awa, aya, ani, C. Urmia aw, oej, oenij vs. Turoyo (Midyat) huwe, hiya, hinne, (Mid[schwa]n) hiye, hiya, h[schwa]nn[schwa]k.
b. The addition of final -t to the 3pl. independent pronoun in NENA (see above), but not in Turoyo (Midyat) hinne.
c. The 1pl. verbal suffix in NENA is -ax (Hertevin -ah), compared with Syriac m. -innan, f. -annan (< nominal pl. -in, an + enclitic -nan) and Turoyo -ina.
5. The clitic copula in Turoyo and Mlahsô for the most part continues the MidAr unstressed subject pronouns, e.g., 1sg. -no, 1pl. -na < MidAr *-na, *-nan (cf. Syr. -na, -nan). In contrast, NENA has created a new copula from a syntagm containing *it ‘there is,’ e.g., Hertevin 3sg. m. -ile, 1sg. m. -iwn, C. Urmia -ijli, -ijvin, J. Azer. -He, -Hen. See Nöldeke 1868: 200-206; Fox 1990: 74-75, Jastrow 1997a: 372-73.
6. The MidAr passive stems (Etp[sup[schwa]][Graphic Character Omitted]el, Etpa[Graphic Character Omitted][Graphic Character Omitted]al, and Ettaf[Graphic Character Omitted]al) survive in Turoyo and Mlahsô (albeit with various modifications and remodelings; Jastrow 1996), but have been lost in the NENA dialects, which express passive voice by means of a periphrastic construction with [Graphic Character Omitted]ys ‘become’ and the past participle (historically the MidAr passive participle in status emphaticus, m. sg. *C[sub1]C[sub2]iC[sub3]-a-; see § 3.1, no. 6).
7. The imperative of G-stem verbs (P[sup[schwa]][Graphic Character Omitted]al) is C[sub1]C[sub2]aC[sub3]-in Turoyo and Mlahsô,(FN10) but C[sub1]C[sub2]uC[sub3]-in NENA: cf. Mlahsô gras, pl. grásun, Turoyo gras, pl. grásu vs. Hertevin gros, pl. grúsen, Mangesh gros, pl. grúsu, C. Urmia grus, pl. grusun, Kerend grus ˜ gúrus, pl. grúsmn ˜ gúrusmun ‘pull!’ (Jastrow 1997a: 365).(FN11) The two groups have generalized the vocalism of stative resp. non-stative verbs in Middle Aramaic; cf. Syr. dmak ‘sleep!’ vs. ktob ‘write!'(FN12)
Consequently, he proposed a division of MA dialects into three subgroups: 1) Western, or MWA; 2) Central, i.e., Turoyo;(FN13) and 3) Eastern. This classification has been followed by most other scholars of Modern Aramaic, including Socin (1882: v), Duval (1896: 125-26),(FN14) and more recently Tsereteli (1977) and Jastrow (1985a: xx).
Macuch’s discovery of a modern descendant of Mandaic (see especially Macuch 1993) adds a fourth subgroup to Nöldeke’s scheme. The resulting four-way classification was adopted by Hoberman (1988: 557-58, 1989: 4, 1997: 313), who proposed renaming the third subgroup Northeast Neo-Aramaic (NENA) in order to distinguish it from Turoyo and Modern Mandaic, and also from older East Aramaic dialects such as Jewish Babylonian. This label has since been adopted by scholars working on MA, including Fox (1994, 1997), Jastrow (1997a: 347ff., 2002: 366ff.), Murre-van den Berg (1999: 3-5), Khan (1999a: 2ff., 2002a: 8ff., 2004: 3ff., 2007), Mutzafi (2004b: 9ff), and others.
On the other hand, and despite the significant isoglosses separating Turoyo and NENA (see below), Hoberman (1988: 558, 572, 1989: 4-6, 1990: 80) has also emphasized the similarities and shared innovations between the two which set them apart from MWA as well as Modern Mandaic.

Along with the well-known replacement of the OA [Old Aramaic] perfect and imperfect conjugations by participial constructions, these shared innovations in the pronominal system make it clear that Turoyo and NENA together form a single branch of the Aramaic family, separate from documented varieties of OA on the one hand and from modern Mandaic and Ma’lula on the other. The shared innovations of NENA and Turoyo are the following:

    A. The spread of initial a to the first person plural independent pronoun, which thus begins with ax, ahA-: PNENA *axnan, *axni, Turoyo dhna.
B. A shift in the vowels of the Nominal suffixes, second person singular, from masc. -ax, fem. -ex to masc. -ox, fem, -ax, Turoyo masc. -ux, fem. -ax (Jastrow 1985a: 35).
C. The gender differentiation of the Verbal suffix, second person singular, as masc. -[Graphic Character Omitted]t, fem. -at in consonant-final verbs. (In OA this differentiation existed only for vowel-final verbs.)
D. The third person plural Nominal suffix *-ay(h[Graphic Character Omitted]n); that is, the ay which appears in OA with plural nouns attaches to singular nouns as well (Hoberman 1988: 572).
Following his lead, Jastrow (1990: 90) set up a three-way division of the MA speech area into Western, Central (i.e., Turoyo and NENA), and Mandaic. More recently, he has fluctuated between these classifications, e.g., distinguishing the four groups MWA, Turoyo/Mlahsô, NENA, and Mandaic in his “state of the art” survey (2002), but also referring to “das zum Ostaramäischen gehörige, doch recht eigenständige Sprachgebiet des Turoyo” together with “das Neuostaramäische im engeren Sinne,” i.e., NENA (2005: 138).

2.2. A STAMMBAUM MODEL FOR ARAMAIC?
One may ask, however, whether the synchronic contrast between MWA and the rest of present-day Aramaic, or between Turoyo and the trans-Tigris dialects, is sufficiently great to justify these hypotheses. The subgroupings proposed above implicitly assume a Stammbaum model for the evolution of (pre-)Modern Aramaic, in which geographically defined dialect areas separate from and lose all contact with each other; in other words, “once having diverged, [the various sub-groups] did not significantly re-converge” (Boyarin 1981: 613). For instance, Hoberman and Jastrow’s classification may be represented by the diagram below:(FN15)
According to this interpretation, the ancestor of MWA separated from the remainder of the Aramaic dialects at some point (the traditionally assumed division into Western and Eastern Middle Aramaic of the 1st millennium a.D.); then Mandaic diverged from “Central” Aramaic, and finally Turoyo/Mlahsô split off from the main body of NENA.(FN16) Each of these postulated dialectal groupings, e.g., Western Middle Aramaic or NENA, “ha[s] a clear and separate identity as the consequence of exclusively shared common innovations… And these innovations in turn reflect a period of exclusively shared common prehistory during which the dialects were in contact only with each other, so that innovations spread only through these dialects” (Hock 1991: 448).
This assumption is certainly inaccurate, as there is no evidence that Aramaic ever underwent a clean split into two or more separate dialects which men developed independently of each other. On the contrary, the considerable evidence for extensive human contact (migration, trade, communication) across the contiguous Aramaic-speaking regions of Palestine, Syria, and Mesopotamia, from pre-Achaemenian times down to the first Islamic centuries, would lead one to expect a complex network of intersecting isoglosses. Other examples include the Romance-speaking territory from Portugal up to Belgium and down to southern Italy, the South Slavic dialects of southeastern Europe, and the Arabic dialects of the modern Middle East. As Hock (1991: 450) puts it,

The linguistic relationship between the neighboring dialects of the same language very commonly cannot be stated in terms of tree diagrams. This is a consequence of the fact that these speech varieties remain mutually intelligible, stay in close contact, and therefore continue to interact with each other on a day-to-day basis, with shifting realignments as political and social circumstances change. It is therefore unrealistic to expect clear, “tree-diagram” splits in such dialect continua. Clear splits seem to result only when originally closely affiliated dialects become separated, through migration or radical political and social realignment, such that they cease to be mutually intelligible and become different languages.

    By the twentieth century, of course, the totality of MA dialects were neither intelligible nor in close contact; not only had MWA and Mandaic been isolated for many centuries, but geographical and political barriers to communication and the social divide between Jews and Christians (and among the latter, among Syrian Orthodox, Nestorians, and Chaldeans) greatly reduced the opportunities for contact and incentives for accommodation to other speech forms–the driving force behind diffusion of linguistic change. It is important to remember, however, that this state of affairs was preceded by well over a millennium of geographically uninterrupted development, from the Persian Empire to the early Islamic period. We shall see below that the distribution of several important features in MA strongly argues for their local diffusion from one dialect to another in the past, and across the boundaries of the generally accepted subgroups.

2.3. MA DIALECTOLOGY RECONSIDERED
In reconsidering the historical evolution of MA dialectal diversity, we must address two fundamental questions. The first concerns the relationship between Turoyo (along with Mlahsô) and the trans-Tigris or NENA dialects. It is important to keep in mind that the seemingly sharp division between Turoyo and NENA is due in large part to a lack of reliable data for the MA dialects spoken in the highlands of southeastern Turkey until 1915. Until Jastrow’s discoveries, there was a considerable gap between the Tur ‘Abdin and the westernmost NENA dialects for which reliable linguistic information was available; the nearest was probably that of the Jews of Zakho.(FN17) Now that far more data has become known, both on the surviving dialects of southeastern Turkey and on the Jewish and Christian dialects of northern Iraq, is the linguistic boundary between Turoyo and NENA as clear-cut, or as significant, as traditionally believed? In this connection, Polotsky’s (1964: 108-9) cautionary remarks on the classification and subgrouping of Semitic apply equally well to MA, in particular to the Turoyo-NENA divide:

[We encounter] the familiar experience that once a dialectological framework has been set up, it may prove too rigid for fresh discoveries. Newcomers will almost inevitably be looked upon as intruders. There is the temptation to leave the existing framework intact and to accommodate additions in such a way as to cause minimal disturbance. It is thus likely that new entities which have every claim to a central or even independent place are assigned a marginal and subordinate position. The only way to obviate this danger is not to regard with too much reverence even such well-established and time-honored sub-groups as “Canaanite.”

    In fact, as Takashina 1990: 116ff. points out, certain features of Hertevin are shared with Turoyo and others with the NENA dialects; aside from lexical items, note the preterite (m)C[sub1]aC[sub2]eC[sub3], maC[sub1]C[sub2]eC[sub3] and infinitive (m)C[sub1]aC[sub2]oC[sub3]a, maC[sub1]C[sub2]oC[sub3]a of derived stems, the former with stem vowel a like Turoyo (vs. NENA u), the latter like NENA (m)C[sub1]aC[sub2]oC[sub3]e, maC[sub1]C[sub2]oC[sub3]e (see §3.1, no. 5 and n. 37). Based on this “twofold nature” of Hertevin, he assigns it to an intermediate position in the “continuum” of Aramaic dialects going back to pre-Islamic times.
The second issue has to do with the linguistic relationship between Turoyo/Mlahsô and the only surviving MA dialects to the west, i.e., Modern West Aramaic. Since me late nineteenth century, Turoyo has generally–and with justification–been treated as a modern variety of “East Aramaic.” Along with me MA varieties to the east, i.e., NENA and Modern Mandaic, it was considered to descend from a dialect closely related to the three great Eastern Middle Aramaic literary languages of the first millennium a.D., namely Syriac, Jewish Babylonian, and Classical Mandaic. However, Boyarin 1981 has argued mat Syriac, despite sharing major isoglosses with Jewish Babylonian and Mandaic (see §2.1 above), also underwent innovations common to one or more of the Western Middle Aramaic languages, e.g., 3sg. masculine possessives of plural nouns in *-auhi, or the spread of prefixed m-in the infinitives of the derived conjugations. Consequently, Syriac is better viewed as me sole literary representative of “Central Middle Aramaic,” occupying an intermediate location within me range of Aramaic dialects spoken in the first millennium a.D.(FN18)
On purely geographical grounds, of course, Turoyo may likewise be considered “Central Modern Aramaic.” However, given the existence of isoglosses grouping Syriac with the Western Middle Aramaic literary languages, we may go one step further and ask the corresponding question for their approximate modern descendants. Do Turoyo (plus Mlahsô) and MWA, now separated by a solid band of Arabic-speaking territory, share any significant innovations vis-à-vis older stages of Aramaic and/or the MA dialects to the east? Such shared features would hearken back to an earlier period when me Qalamun valley and northern Mesopotamia (including the Tur ‘Abdin) formed part of a dialect continuum spanning (historical) Syria, within which innovations beginning in one place could–and did–spread to neighboring areas.
Finally, the existence of distinct coterritorial Christian and Jewish varieties is a well-known characteristic of MA dialectology (see, e.g., Hopkins 1999: 321-22), but the interaction of the geographical and religious dimensions has not yet been sufficiently explored. Now that descriptions are available for far more dialects, including both Christian and Jewish varieties of several localities (Aradhin, Koy Sanjaq, and Sena, in addition to Zakho, Urmia, and Salamas), we should compare the geographical distribution of linguistic features among Christian and Jewish MA dialects, and seek to determine whether any post-Middle Aramaic innovations which began in one religious community have spread to the other.
These questions will lead us to reexamine the classification schemes listed in § 2.1, taking into account the dialect studies published in recent years. Based on principles of historical dialectology, dialect contact, and diffusion, I propose that the Modern Aramaic speech area should be viewed as a geographically discontinuous dialect continuum, stretching from Syria through southeastern Turkey and northern Iraq to northwestern Iran.(FN19) The distribution of a number of post-Middle Aramaic phonological, morphological, and lexical innovations (§3) suggests that Turoyo/Mlahsô and NENA together form a single chain of dialects, within which the easternmost NENA dialects, Qaraqosh, on the southwestern edge of Iraqi Kurdistan; and Turoyo and particularly Mlahsô in the west, occupy peripheral positions. In addition, Turoyo and Mlahsô share at least two significant innovations with MWA (§4), suggesting that the former occupy an intermediate place between MWA and NENA, just as Syriac does between the Western and Eastern Middle Aramaic languages.

3. EVIDENCE FOR A (PRE-)MA DIALECT CONTINUUM
It is a well-established principle of historical linguistics that only innovations are of diagnostic value for the subgrouping of related languages. This principle, stated by Hetzron 1976 within the context of Semitic, goes back at least to the 1880s and the Neogrammarian Carl Osthoff. Since inheritances (archaisms) might happen to be preserved in any particular descendant, two or more languages belong to the same subgroup if and only if they share a significant number of innovative features with respect to the proto-language (Hock 1991: 579; cf. Harrison 2003: 233).(FN20) Whether phonological, morphological, or lexical, these features must also be sufficiently idiosyncratic that the probability of their independent occurrence in multiple speech communities is not significant.(FN21) If these conditions are met, we may with reasonable confidence postulate a period of common evolution of the languages in question, following the breakup of the larger family to which they belong.
It follows that most of the arguments usually adduced for the subgrouping of MA, or for divisions within the NENA dialects, are of little or no diagnostic value. For instance, many of the features separating the NENA dialects from Turoyo (and Mlahsô) involve comparatively trivial phonological changes (e.g., *a > o and simplification of geminates in Turoyo/Mlahsô) or mergers (e.g., of MidAr *k and *h in NENA). Furthermore, loss of morphological categories is one of the most easily repeatable linguistic changes, and therefore among the least diagnostic for subgrouping (cf. Hetzron 1976: 96-97 with refs.). Thus little weight can be attached to the disappearance of the old prefixed and suffixed conjugations, or of the distinction between status emphaticus and status absolutus in the adjective, in Turoyo and NENA vs. their retention in MWA; here MWA has simply preserved an old formation and/or distinction. Phonetically natural and crosslinguistically common sound changes are likewise unsuited for subgrouping, e.g., the treatment of MidAr *t and *d, which Maclean 1895: xii-xv took as the basis for his initial classification of the NENA dialects.
On the other hand, idiosyncratic (morphological or other) innovations limited only to small sets of dialects tell us little about the larger patterns, the “big picture” of MA dialect differentiation. For instance, the fate of the MidAr Pa[Graphic Character Omitted][Graphic Character Omitted]el and Af[Graphic Character Omitted]el conjugations varies greatly from one NENA dialect to the next (Jastrow 1997a: 360-62; Mutzafi 2004a: 259-60; Kapeliuk 2005: 356-57), without clear indication of larger dialectal groupings. Similarly the evolution of vowel harmony in the Christian and Jewish varieties of Iranian Azerbaijan is a significant innovation shared only by those dialects, but offers little clue as to their geographical or historical relation to the rest of MA. The same applies to other locally restricted changes, e.g., the various modifications of endings and paradigms in standard Modern Syriac, many confined to the Christian dialects of Urmia and its immediate vicinity (Polotsky 1961: 23ff.; cf. Younansardaroud 2001: 73ff.); the characteristic features of the Jewish dialects of Arbel and the surrounding plain (the “Trans-Zab” group; Khan 1999a: 8-9, 1999b: 222-24; Mutzafi 2004b: 9-10); or the complex remodeling of the preterite in the Jewish dialects of eastern Iraqi and Iranian Kurdistan (see §3.1, no. 6).
In fact, probably the only useful criteria among those listed in §2.1 are the development! in the personal pronouns (both those common to Turoyo/Mlahsô and NENA, and NENA-only innovations) and the vocalism of G-stem imperatives. However, a number of isoglosses of equal or greater significance may be drawn across the MA speech area, based on precisely the sort of idiosyncratic innovations stipulated above.

3.1. PERIPHERAL ARCHAISMS IN “CENTRAL” MA
In the following cases, an innovation has taken place and diffused throughout a geographically central subset of Mlahsô/Turoyo and NENA, leaving unaffected peripheral areas to the west, east, and/or south which preserve the earlier state of affairs.
1. Placement of stress (Jastrow 1997a: 353). In general, word stress falls on the final syllable of nominal and verbal stems in Mlahsô (Jastrow 1985b: 267, 1994a: 14, 26) and the eastern Jewish NENA dialects, i.e., those of Arbel, Rustaqa/Ruwanduz, Sulemaniyya anc Halabja, and Koy Sanjaq in eastern Iraqi Kurdistan (Khan 1999a: 70-74, 2004: 61ff. Mutzafi 2004b: 55f.),(FN22) as well as Iranian Azerbaijan (Garbell 1965: 34-35) and Iraniar Kurdistan (Hopkins 2002: 282 and passim; Jastrow 1997a: 353). In the remaining MA dialects, stress is penultimate; this includes Turoyo (Jastrow 1985a: 26-27), Hertevin (Jastrow 1988: 17-18), Bohtan (Fox 2002: 167), Hassane and the other Cudi dialects (Sinha 2000: 61ff.), Baz (Mutzafi 2000: 300-301), Jilu (Fox 1997: 11-12), Tkhuma (Jacobi 1973: 30ff.). C. Zakho,(FN23) J. Zakho (Sabar 2002: 36-37), C. Aradhin (Krotkoff 1982: 16), J. Aradhin (Mutzafi 2002b), C. Nerwa (Talay 2001: passim), J. [Graphic Character Omitted]Amadiya (Hoberman 1989: 151), J. Barzani (Mutzafi 2002a: 52-53), the Christian dialects of Alqosh, Kultepe, and other villages of the Mosul plain (Sachau 1895: 64),(FN24) Qaraqosh (Khan 2002a: 67-68), C. Koy Sanjaq (Mutzafi 2004a, 2004b: 10-11), the Christian dialects of Iranian Azerbaijan (Hetzron 1969: 114; Younansardaroud 2001: 65ff.), and Senaya.(FN25)
2. In certain MA dialects, the 3sg. pronominal suffixes attached to verbs to denote a direct or indirect object differ from those found with nouns and prepositions. Whereas the former continue MidAr m. *-eh, f. *-ah, the latter appear to reflect preforms m. *-ew, f. *-aw < *-ehu, *-ahu, whose exact origins remain unclear.(FN26) Other dialects have leveled the distinction in favor of the postverbal allomorphs, e.g., Turoyo abr-e ‘his brother,’ min-a ‘from her’ like k-obé-no-le ‘I give him,’ ko-hozé-la ‘he sees her.’ The following data is taken from Hoberman 1988: 563 and Jastrow 1997a: 355-56, augmented by more recent descriptions.
We see that the distinction between nominal and verbal suffix is preserved in the peripheral dialects, namely Mlahsô, Hertevin, and the other dialects of the Bergnestorianer in the west and the Jewish dialects of eastern Iraq and Iran, plus C. Urmia, in the east. In contrast, the centrally located dialects have generalized the verbal suffixes. Cf. Hertevin bet-ew, J. Azer-baijani bel-ew vs. Turoyo bayt-e, J. [Graphic Character Omitted]Amadiya be:theta-e ‘his house,’ the latter having the same ending as Turoyo gr[schwa]s-le, J. [Graphic Character Omitted]Amadiya gr[Graphic Character Omitted]s-le ‘he pulled.'(FN30)
3. The 1sg. present ending was originally just the enclitic personal pronoun *-na (< independent *ana), attached to the present active participle. Cf. the Syriac present paradigms for the strong root [root]dmk ‘sleep’ and the III-y root [root]hzy ‘see’:

1sg     m.      damekna   hazena      ˜ hazen

             f.      damkana ˜ damkan       hazyana ˜ hazyan

2sg.         m.      damkat                               hazet

             f.      damkat                               hazyat

3sg.         m.      damek                                haze

             f.      damka                             hazya

    In most NENA dialects, the 1sg. m. and f. endings have been remodeled on the basis of the 2sg., for which all varieties of (eastern) MA share forms descended from m. *-et, f. *-at. The 2sg. m. and f. endings were originally identical for sound verbs (e.g., Syr. 2sg. m./f. damkat < *damek-att, *dameka-(a)tt), but came to be differentiated by generalizing the endings proper to Hl-y roots (Hoberman 1988: 565-68, 571-72; Khan 1999a: 91). The alteration of the 1sg. endings in most NENA dialects has resulted in a uniform present (or rather, subjunctive) inflection, in which all the sg. forms are disyllabic. Cf. the subjunctive sg. forms of [root]pth ‘open’ in Hertevin, the Jewish dialect of Arbel, and Christian Urmia:

                Hertevin        J. Arbel        C. Urmia

1sg.         m.      páthen      palxén         poetxin

             f.      páthan      palxán         poetxoen

2sg.         m.      páthet      palxét         poetxoet

             f.      páthat      palxát         poetxoet

3sg.         m.      páteh       palíx        poetix

             f.      pátha       palxá          patxoe

    However, the MidAr ending *-na is at least partially preserved in three sets of dialects, located on the western, eastern, and southern edges of the Central/Northeastern MA area.
1. In Mlahsô domémo, hozéno and Turoyo m. domáxno, hozéno, f. d[schwa]mxóno, h[schwa]zyóno, where 1sg. -no has the synchronic status of a clitic and does not affect the position of stress (R. Kim forthcoming a).
2. The Jewish dialects of Iranian Kurdistan likewise preserve the original ending, although the 1sg. f. also exhibits alternants in -an, apocopated from *-a-na: cf. Kerend pres. 1sg. m. garísna, f. garsán(a) ‘I pull,’ intransitive pret. 1sg. m. qímna, f. qíman(a) ‘I stood.'(FN31) In J. Arbel, palxéna and palxána ‘I open’ “exist as rare alternatives” to usual 1sg. m. palxén, f. palxán (Khan 1999a: 90); similarly for J. Koy Sanjaq, e.g., baxén(a), baxyán(a) ‘I cry’ or perfect rxís-en(a), rxís-an(a) ‘has walked’ (Mutzafi 2004b: 11, 82).(FN32) The Baz dialect described by Mutzafi (2000: 307) has a single form for 1sg. m. and f. in -[schwa]n M~ -[schwa]na, e.g., y-káns[schwa]n(a) ‘I sweep.’
3. The Jewish dialects of Sulemaniyya/Halabja and Rustaqa/Ruwanduz have gone one step further, eliminating the longer ending in the 1sg. f.: hence palxán, but m. palíxna (Khan 2002b: 400, 2004: 83-84). In Qaraqosh as well, the 1sg. f. is in -an, but the 1sg. m. optionally maintains the old ending, as in m. pátx[schwa]n ˜ pat%[schwa]xna, f. pátxan (Khan 2002a: 88-89). To the west, the Cudi dialects and C. Nerwa have -ena beside -[schwa]n in the 1sg. m. (Sinha 2000: 101; Talay 2001: 16-17);(FN33) and Jilu has 1sg. m. i-pát[schwa]xna (f. i-patxan) beside imperfect i-pátx[schwa]n-wa ‘I opened’ (Fox 1997: 35).
The variation in the latter two groups of dialects is also observed in the seventeenth-century texts from northern Iraq edited by Mengozzi (2002: 30): alongside usual 1sg. m. -en, Israel of Alqosh uses the ending -ena, like J. Arbel palxéna ˜ palxén.
4. The preterite of intransitive verbs is formed by adding the subject pronominal endings to the passive participle in status absolutus (MidAr *C[sub1]C[sub2]iC[sub3[sup-]]) in the Jewish dialects of eastern Iraqi Kurdistan (Rustaqa, Koy Sanjaq, Sulemaniyya/Halabja; Khan 2002b: 403-5, 2004: 86-87; Mutzafi 2004b: 82) and Iranian Kurdistan (Hopkins 1989: 426ff.). Similarly in Turoyo, many intransitive preterites, including the most frequent, add the subject endings to an originally déverbal adjective in status absolutus (MidAr *C[sub1]aC[sub2]C[sub2]iC[sub3[sup-]]), e.g., Turoyo 1sg. m. dam%[schwa]x-no, f. damixó-no ‘I slept.'(FN34)
In the remaining dialects, the preterite of intransitive verbs had adopted the pattern of transitive verbs: the original pass. ptcp. *C[sub1]C[sub2]iC[sub3[sup-]] is followed by l- + object pronominal suffixes in Hertevin, Hassane, the Christian and Jewish dialects of all but easternmost Iraqi Kurdistan, Christian Urmia, Jewish Azerbaijani, and Senaya.(FN35) See Jastrow 1997a: 366-67; cf. the forms for ‘I (m., f.) stood’ and ‘I pulled’ in six MA dialects, arranged from west to east.

        Turoyo                       Hertevin                J. Zakho                C. Urmia

intrans.     [qay%[schwa]m-no, qayimó-no]       qem-li               qim-li             qim-lij

trans.       gr%[schwa]s-li                   gres-li           gris-li         gris-lij

        J. Iran. Azer.          Kerend(FN36)

intrans.     qím-li               qím-na, qíma-n(a)]

trans.       grís-li           grís-li

    5. A second or actual/continuous present is found in the MA dialects of the Hakkâri Nestorians, most of Iraq, and Iranian Azerbaijan, including the Cudi dialects (Sinha 2000: 106-7, 131), Baz (Mutzafi 2000: 309ff.), Jilu (Fox 1997: 22ff.), the Tiyari dialects (Talay 1999: 172-73), Tkhuma (Jacobi 1973: 141-44), Iraqi Koine (cf. Fox 1990: 74-75), C. Zakho, J. Zakho (Sabar 2002: 48), C. Aradhin (Krotkoff 1982: 33), C. Nerwa (Talay 2001: 20), J. [Graphic Character Omitted]Amadiya (Hoberman 1989: 79-82), the Christian dialects of the Mosul plain (Sachau 1895: 50-51; Rhétoré 1912: 86), J. Koy Sanjaq (Mutzafi 2004b: 82-84, 112-13), J. Sulemaniyya (Khan 2004: 100-101, 2005: 366-67), C. Urmia (Polotsky 1961: 22-23, 1991; Hetzron 1969: 114ff.; Murre-van den Berg 1999: 201ff.; Younansardaroud 2001: 73ff., 2002: 844-48), and J. Iranian Azerbaijan (Garbell 1965: 63-65). This formation consists of the innovative NENA copula, described above on p. 3, and (b- +) the (post-)MidAr infinitive (*C[sub1]C[sub2]aC[sub3]a for the P[sup[schwa]][Graphic Character Omitted]al conjugation; Pa[Graphic Character Omitted][Graphic Character Omitted]el *mC[sub1]aC[sub2]oC[sub3]e, Af[Graphic Character Omitted]el *maC[sub1]C[sub2]oC[sub3]e);(FN37) see Jastrow 1997a: 361ff., esp. 365-66 and Goldenberg 2000: 82-84, the latter comparing English be on X-ing > be a-X-ing –> be Xing.
On the other hand, this formation is absent in the western dialects of Mlahsô, Turoyo, Hertevin, and Bohtan (Fox 2002: 174); in the Christian dialect of Qaraqosh; and at the eastern end of the MA-speaking area, the Jewish dialects of Iranian Kurdistan (Hopkins 1989: 417 n. 10), as well as Senaya (Heinrichs 2002: 256-60). Some of these have retained the MidAr present active participle in both general and actual usage, e.g., Turoyo k-oxál-no, Hertevin axlen, Qaraqosh k-([Graphic Character Omitted])áxl[schwa]n ˜ k-([Graphic Character Omitted])ax%[schwa]lna ‘I (m.) am (in the process of) eating’ as well as ‘I (m.) eat (in general, all the time, every day).'(FN38)
Other MA dialects have evolved different means for expressing the actual present, but formal discrepancies among them make it clear that they are independent local innovations. Thus Qaraqosh and Senaya both make use of the general (unmarked) present plus the copula, but whereas the Qaraqosh continuous present takes the form 3sg. m. k-il[schwa] k-sat[schwa], f. k-ila k-satya ‘he, she is drinking’ (Khan 2002a: 18, 331ff.), Senaya cliticizes the copula to the present, e.g., 3sg. m. paseh-ile, 1sg. f. páshan-yan ‘he, I (f.) am opening’ (Panoussi 1990: 118; Heinrichs 2002: 260).(FN39)
On the other hand, a formation la-palix ‘he is opening’ occurs in the Jewish dialects of Arbel and Rustaqa/Ruwanduz, and similarly la-k-pát[schwa]x in the Christian dialect of Koy Sanjaq (Mutzafi 2004a: 255-56). Here the prefix la- seems to be some sort of presentative particle, perhaps a fossilized form of the copula (Khan 1999a: 111-14, 271-74, 1999b: 219, 2000: 323ff., 2002b: 402; Heinrichs 2002: 261-62; see no. 6 below).(FN40) A comparable construction [Graphic Character Omitted]ale k-sate, [Graphic Character Omitted]ale y-sate ‘he is drinking’ is found in the Jewish dialects of Barzani, Shahe, and B[schwa]jil (Mutzafi 2002a: 42, 59).
6. A perfect based on the MidAr passive participle in status emphaticus ([Graphic Character Omitted][sup[schwa]][Graphic Character Omitted] sg. m. *C[sub1]C[sub2]iC[sub3]-a, f. *C[sub1]C[sub2]iC[sub3]-ta, p1. m. *C[sub1]C[sub2]iC[sub3]-in) occurs in Bohtan (Fox 2002: 172ff.), the Cudi dialects (Jastrow 1997b: 279-80; Sinha 2000: 106), Baz (Mutzafi 2000: 309ff.), Jilu (Fox 1997: 22ff.), Tkhuma (Jacobi 1973: 141-44), C. Zakho, J. Zakho, C. Aradhin (Krotkoff 1982: 34-35), C. Nerwa (Talay 2001: 19-20), J. [Graphic Character Omitted]Amadiya (Hoberman 1989: 82-90), the Christian dialects of the Mosul plain (Rhétoré 1912: 102-4), Qaraqosh (Khan 2002a: 340ff.), C. Koy Sanjaq (Mutzafi 2004a: 254-57), J. Koy Sanjaq (Mutzafi 2004b: 82-84, 105-9), C. Urmia (Polotsky 1961: 22-23, 1991; Hetzron 1969: 114ff.; Murre-van den Berg 1999: 201ff., Younansardaroud 2001: 73ff., 2002: 844-48), and J. Azerbaijan (Garbell 1965: 68-70; see below). This perfect is construed with the same innovative copula found in the actual present (see no. 5), e.g., C. Urmia ptijxevin ‘I (m.) have opened’ (<ptijxoe + -ijvin), like biptoexevin ‘I (m.) am opening’ (< bi-ptoexoe + -ijvin). The perfect in the Jewish dialects of extreme eastern Iraqi Kurdistan (Sulemaniyya, Halabja) and Iranian Kurdistan has the same origin, but the originally voice-neutral formation has been remodeled so as to differentiate transitive and intransitive; for details, see Hopkins 2002 and Khan 2004: 85-90, 97ff., 313-18, 2005:362-66.
No such perfect occurs in Mlahsô and Turoyo; Hertevin; the Jewish dialects of Arbel and Rustaqa in eastern Iraqi Kurdistan; or Senaya. Turoyo simply adds ko- to the preterite, e.g., Turoyo h[schwa]zyó-li ‘I saw her,’ ko-h[schwa]zyó-li ‘I have seen her’; this is surely the same ko- which forms the present from the subjunctive stem, e.g., ko-hozé-no-la ‘I see her.'(FN41) Similarly, the Jewish dialects of Arbel and Rustaqa make use of the same particle la- as in the actual present (see no. 5), e.g., Arbel la-plix-le ‘he has opened,’ Rustaqa la pil ‘he has fallen (and is lying on the ground)’ (Khan 1999a: 274-75, 2000: 328-29, 331, 2002b: 403-4).(FN42) Senaya prefixes gi- to the preterite, as in gi-pseh-le ‘he has opened’ (Heinrichs 2002: 261-63).
In two other dialects, we find verbal systems which reflect intermediate stages in the evolution of the perfect. Jastrow (1988: 58-59; cf. Goldenberg 1993: 302-3) observes that the perfect in Hertevin has not yet been fully elaborated: the construction “passive participle in status absolutus + enclitic subject personal pronoun” denotes the perfect of intransitive verbs but, as in Turoyo and Jewish Iranian Kurdistan, the perfect passive of transitive verbs. Cf. Hertevin intransitive [Graphic Character Omitted]iz-en ‘I have gone,’ transitive passive qtil-en, -an ‘I (m., f.) have been killed’ vs. preterite zi-li ‘Iwent,’ qtel-li ‘I killed.’ This dialect thus lacks a grammaticalized perfect for transitive active verbs, which would correspond to C. Urmia xizyevin (< xizyoe ijvin) ‘I (m.) have seen’ or Turoyo ko-ft[schwa]h-le, J. Arbel la-plix-le ‘he has opened.'(FN43)
The Jewish Azerbaijani dialects also appear to form a sort of transition zone for the spread of the status emphaticus perfect. These dialects, like most others (no. 4), form the preterite of both intransitive and transitive verbs with the passive participle in status absolutus and/- + object pronominal suffixes, e.g., qím-li ‘I stood,’ grís-li ‘I killed.’ But whereas the perfect of intransitive verbs consists of the passive participle in status absolutus plus pronominal subject suffixes, transitive verbs make use of the same status emphaticus + copula as C. Urmia and the other Christian dialects of Azerbaijan.
See Garbell 1965: 68-70 for further examples, and also Heinrichs 2002: 261.
Finally, the Jewish dialects of Barzani and vicinity have a morphologically suppletive perfect: only the negative is formed from the (negative) copula and the preterite participle in status emphaticus, while the positive consists of [Graphic Character Omitted]ale plus the preterite, similar to actual present [Graphic Character Omitted]ale k-sate, [Graphic Character Omitted]ale y-sate ‘he is drinking’ Thus we find [Graphic Character Omitted]ale qt[schwa]le ‘he has killed’ vs. lewe qtila ‘he has not killed,’ and [Graphic Character Omitted]ale s[schwa]dri-le ˜ [Graphic Character Omitted]ale sd[schwa]re [Graphic Character Omitted][schwa]lu ˜ [Graphic Character Omitted]ale sd%[schwa]relu ‘he has sent them’ vs. lewe sdira-lu ‘he has not sent them’ (Mutzafi 2002a: 42, 60, 65-66). These dialects were apparently in the process of replacing the status emphaticus perfect with the new “particle + preterite” construction; see below, pp. 521-22.
7. In most MA dialects, the preterite stem can agree only with a third-person patient, in which case it is inflected for gender and number: cf. Turoyo gr%[schwa]s-le, grisó-le, grisí-le ‘he pulled (him), her, them,’ J. Zakho sqil-li, sqila-li, sqila-li ‘I took (him), her, them.’ This continues the situation in Middle Aramaic, where the passive participle *C[sub1]C[sub2]iC[sub3]- in status absolutus agreed with its referent for number and gender, just like other adjectives. To express a first- or second-person patient, these dialects must add I- + clitic pronoun after the subject: hence Turoyo grisó-le ‘he pulled her,’ but gr%[schwa]s-le-li ‘he pulled me,’ gr%[schwa]-le-lalxu ‘he pulled you (pl.).'(FN46)
On the other hand, several MA dialects of Iraq and Iranian Azerbaijan allow the preterite stem to be marked for objects of any person, thereby creating (partial) parallelism of agentand patient-marking between the preterite and present: cf. J. [Graphic Character Omitted]Amadiya k-patx-an-nux ‘I (f.) open you (m.),’ ptix-an-nux ‘you (m.) opened me (f.)’ (Hoberman 1989: 36), and see Mutzafi 2002a: 65, 2002b: 481-82 for J. Barzani and J. Aradhin, and Hetzron 1969: 117ff.; Polotsky 1979 for C. Urmia. The evidence of the Nerwa texts (e.g., mulp-ax-lu ‘they taught us,’ lit-an-nox ‘you (m.) cursed me (f.)’; Goldenberg 1992: 119ff.) and religious poetry of northern Iraq (e.g., mxols-itu-le ‘he saved you (pl.),’ muxw-en-ne ‘he showed me’; Mengozzi 2002: 33ff., 2005a: 249-52) shows that such “intraconjugational” object affixes existed in Iraqi Aramaic of the seventeenth century, but had been lost by the nineteenth century in the Mosul plain dialect of the Sachau manuscripts. This intraconjugational marking of first- and secondperson objects is best understood as a post-MidAr innovation of the Iraqi dialects, which never spread to outlying areas such as Tur ‘Abdin or Iranian Kurdistan.(FN47)
For the most part, these innovations are found within a common core of dialects, namely those of trans-Tigris Turkish Kurdistan; western and central Iraqi Kurdistan, excepting the south(west)ernmost Christian dialect of Qaraqosh; and to a lesser extent, Iranian Azerbaijan, especially the Christian varieties of Urmia, Salamas, etc. In contrast, the areas on the periphery — Mlahsô and Turoyo to the west, Qaraqosh to the southwest, and eastern Iraqi Kurdistan and Iranian Kurdistan to the southeast — have preserved more conservative varieties of MA, lacking such typical features of the northwestern Jewish dialects (e.g., Zakho) or the Christian literary standard of Urmia as pres. lsg. m. -in, f. -an paralleling 2sg. m. -it, f. -at, uniform preterite inflection for intransitive and transitive verbs, or actual/continuous presents and perfects built respectively on the post-MidAr infinitive and the passive participle in status emphaticus.(FN48) For details, see Figures 1 and 2.
These isoglosses thus present the familiar pattern of focus vs. relic areas, with innovations concentrated in the former and peripheral archaisms in the latter (Hock 1991: 440-41; Chambers and Trudgill 1998: 94, 167-68). Moreover, we find that within the areas having an actual/continuous present and perfect, new formations consisting of invariable particle + present resp. preterite (la-palix ‘he is opening,’ la-plix-le ‘he has opened’) were in the process of spreading among the dialects of northeastern Iraq: they occur inter alia in Jewish Arbel, Rustaqa/Ruwanduz, and the vicinity of [Graphic Character Omitted]Aqra (Dobe; Barzan, Shahe, Bijil), and in Christian Koy Sanjaq (Mutzafi 2004a: 260-62); but the new perfect remains confined to the positive in the Barzani dialects. The resulting “donut-hole” picture may be observed in Figure 2.
The above observations strongly suggest that Mlah.s6/Turoyo and NENA cannot be neatly divided into discrete subgroups, but rather constitute a single dialect continuum. The existence of Randarchaismen shared by Mlahsô and Turoyo with eastern NENA dialects does not presuppose any historical connection between these two areas (pace Hopkins 1989: 417 n. 10, 429-30), but merely reflects their position at opposite ends of the main “central MA”-speaking region.

3.2. THE POSITION OF MLAHSÔ
Further support for the hypothesis that Mlahsô, Turoyo, and NENA form one dialect continuum is provided by at least two innovations common to Turoyo and all of NENA, but absent in Mlahsô.
1. Turoyo and the NENA dialects have shortened vowels in closed syllables, leading to the characteristic alternations that play such a prominent role in MA morphology, e.g., Turoyo pres. ko-gór[schwa]s ‘he pulls’ ˜ ko-gúrs-o ‘she pulls,’ pret. gr%[schwa]s-li ‘I pulled (him)’ ˜ gris-ó-li ‘I pulled her’; C. Urmia pres.k-poetix [ae:] ‘he opens’ ˜ k-poets-oen [ae] ‘I (m.) open,’ perf. ptijxeli ‘he has opened’ ˜ ptixteloe ‘she has opened.’ By contrast, vowels in Mlahsô are of roughly the same length and quality in both open and closed syllables, e.g., in doméx ‘he sleeps,’ domxína ‘we sleep,’ dmíx-li ‘I slept’ (Jastrow 1985b: 267, 1994a: 20-23).
2. In a number of cases, Turoyo and NENA have replaced an inherited word with a derivative or a borrowing (usually from Kurdish), while Mlahsô preserves the original Aramaic form. Examples include

Akahsô avó ‘father’ < Old Syriac (OSyr.) aba vs. Turoyo bábo, J. Zakho baba, Nerwa texts baba, C. Urmia babae, J. Azer. bába <– Kurd., Turk. baba (and/or Lallwort?);

    Mlahsô dozó ‘paternal uncle’ < OSyr. dada vs. Turoyo [Graphic Character Omitted]ámmo, Tiyari [Graphic Character Omitted]ama, J. Zakho [Graphic Character Omitted]amoya, J. Koy Sanjaq [Graphic Character Omitted]amona, J. Azer. amona (*ama preserved in ámi ‘my uncle!’; Garbell 1965: 296) <– Arabic [Graphic Character Omitted]amm;(FN49) and

Mlahsô renyó ‘trouble’ < OSyr. renya ‘thought, anxiety, trouble’ vs. Turoyo, J. Zakho, J. Azer. dard ‘pain, trouble, sickness’ <– Kurd., Turk. derd.

    Other lexical archaisms in Mlahsô include ahó ‘brother,’ lyav ‘in, into,’ mzitó ‘town,’ l-mun ‘why’ (‘for what’) < OSyr. aha, lgab, mditta, l-mun; cf. Turoyo ahúno, láwy[schwa]l (with suffixation and metathesis), waláye (<– Arabic), qay (<– Kurdish). For discussion and other examples, see Jastrow 1985b: 269, 1994a: 15-16 and Goldenberg 1998: 65.
These isoglosses suggest that Mlahsô occupies a peripheral, outlying position with respect to both Turoyo and NENA (see Fig. 2). Despite sharing most characteristic features with Mla’sô (see § 3.1, nos. 3-6; Jastrow 1985b: 266-67, 1994a: 13-16), Turoyo here sides with all the NENA dialects against Mlahsô. Along with the distribution of innovations in § 3.1, this suggests that the division between Mlahsô/Turoyo and NENA is not as significant as generally believed, and that both traditional groupings in fact belong to a single “central” or Central/Northeastern MA dialect continuum. If this interpretation is correct, shortening of vowels in closed syllables and the replacement of, e.g., *aba, *dada by *baba, *[Graphic Character Omitted]amma must have begun somewhere in the middle of the MA-speaking territory — perhaps also in northern Iraq, as suggested for the changes in § 3.1 — and eventually spread to all dialects except for that of two villages in the far northwest, near Diyarbakir.(FN50)
3.3. INNOVATIONS SHARED BY MWA AND TUROYO/MLAHSÔ
Having examined the MA dialects to the east of the Tigris, let us now turn west to reconsider the relationship between Turoyo/Mlahsô and MWA. A number of scholars have remarked that the two languages share only superficial phonetic details; thus Hoberman (1989: 4-6) justifiably doubts the value of “a few shared sound changes (a > o, p > f) and… certain morphological archaisms.” Yet although the backing and rounding of stressed a > o is phonetically natural and attested in dozens of languages worldwide, the relative chronology of Turoyo sound changes suggests that the shift of a > o there was a very early change — as one might infer from the presence already in West Syriac — and so may have been part of the same phenomenon as the MWA change.(FN51) Furthermore, Turoyo/Mlahsô and MWA share at least two other developments of a more idiosyncratic nature:
1. Introduction of a prothetic vowel *e- before word-initial clusters of obstruent + sonorant, e.g., Ma[Graphic Character Omitted]lula ebra, Mlahsô ebro, Turoyo abro < MidAr *br-a ‘son’ or Ma[Graphic Character Omitted]lula ema, Turoyo [schwa]mo <MidAr *m-a ‘name.’ In contrast, most NENA dialects resyllabify the initial cluster by geminating the sonorant, as in simma ‘name’ or dimma < MidAr *dm-a ‘blood.’
Interestingly, Hertevin sometimes follows the western pattern and sometimes that of the NENA varieties to the east (e.g., ebra like Ma[Graphic Character Omitted]lula ebra, Turoyo abro, but demma like NENA dimma; Jastrow 1990: 91-92). This suggests that the dialect of Hertevin belonged to the transition zone for this particular feature, and that the correspondence between MWA and Mlahsô/Turoyo does indeed reflect the diffusion of a single sound change, rather than independent innovations.
2. Use of the deverbal adjectives in *C[sub1]aC[sub2]C[sub2]iC[sub3]- as a verbal base for the perfect of intransitive verbs, e.g., *dammik- ‘sleeping, having slept,’ [Graphic Character Omitted]*nappiq- ‘coming out, having come out.’ The perfect of most verbs in MWA goes back to the old passive participle *CCiC-, but the majority of intransitive verbs have a perfect that continues *CaCCiC- (Arnold 1990: 76): cf. Ma[Graphic Character Omitted]lula it[Graphic Character Omitted]en ‘has carried,’ f. t[Graphic Character Omitted]ina vs. marrek ‘has gone by,’ f. marrika. Discrepancies within the three MWA-speaking villages indicate that there must have been variation in the choice of perfect formant at an earlier stage of pre-MWA, possibly going back to late Western Middle Aramaic: for instance, Arnold (1990: 76 n. 4) notes that many speakers of Bax[Graphic Character Omitted]a say dammex, f. dammixa for Ma[Graphic Character Omitted]lula and Gubb[Graphic Character Omitted]adin idmex, f dmixa ‘s/has slept, is asleep.’ Such variation must have characterized earlier stages of Turoyo as well; in the contemporary language, a majority of the most common intransitive verbs, and a handful of transitives, form their preterite on a stem C[sub1]aC[sub2]iC[sub3]-, e.g., 3sg. f. damíx-o ‘she slept,’ samí[Graphic Character Omitted]-o ‘she heard,’ gawír-o ‘she married,’ while others have adopted the pattern of transitive verbs, e.g., s[Graphic Character Omitted]%[schwa]l-la ‘she coughed’ (Jastrow 1985a: 71ff.; Mengozzi 2005a: 244-45; see § 3.1, no. 4).(FN53)
These shared innovations, although hardly impressive evidence of linguistic diffusion, need not be dismissed as mere coincidence or parallel evolution. They date from a time when MWA and Mlahsô/Turoyo were still connected by a chain of Aramaic dialects — i.e., before most of the population of Syria shifted to Arabic, hence no later than c. A.D. 1000. Within this former dialect continuum, the speech of the Tur ‘Abdin would naturally have more closely resembled that of nearby villages across the Tigris than the dialects of the Anti-Lebanon, but this is no reason to conclude that “Turoyo is closely connected with the other Neo-Syriac dialects, rather than with the Ma’lula group” (Blau 1968: 605 n. 1; emphasis mine). Only after the geographically intervening Aramaic varieties had all disappeared did the Anti-Lebanon dialects become completely isolated from developments taking place in northern Mesopotamia and points farther east. The consequences of this finding for the traditional Stammbaum model of Aramaic dialectology will be considered in § 4.

4. CONCLUSIONS
We have examined a number of important post-Middle Aramaic innovations which characterize a subset of the MA dialects, and seen that their distribution calls into question the standard view of clearly defined subgroups. With these results in mind, let us return to the questions in § 2.
As foreshadowed in § 2.3, the Tigris River has lost much of its significance as a dialect boundary separating Turoyo from the NENA dialects. To the extent that Turoyo stands out, it is as a relatively conservative Central/Northeastern dialect which has escaped many of the innovations that originated farther to the east. Whereas the isoglosses that run between Turoyo/Mlahsô and NENA are well known (e.g., *a > o in Turoyo/Mlahsô, merger of MidAr *x, *h and *gamma, *[Graphic Character Omitted] in NENA, or several innovations in personal pronouns in NENA; see § 2.1), other features group Hertevin with Turoyo/Mlahsô (§ 3.1, nos. 3 and 5), and still others separate Mlahsô from Turoyo (§ 3.2, nos. 1 and 2). The Randarchaismen of § 3.1, in which Turoyo o patterns with peripheral dialects, in particular those of Qaraqosh and eastern Iraqi and Iranian Kurdistan, can be understood only if Turoyo belongs to the same dialect continuum as NENA. We may infer from their present distribution that the innovations studied here — shift of primary stress from final to penultimate syllables; leveling of the 3sg. m. and f. pronominal suffixes after nouns/prepositions and verbs; remodeling of the present I sg. on the 2sg.; replacement of enclitic subject pronouns by l- + pronominal suffixes in the preterite of intransitive verbs; creation of an actual/continuous present from the infinitive, and of a perfect from the status emphaticus of the MidAr passive participle — originated somewhere in the center of the MA speech area, perhaps in northwestern Iraq, then spread east and west to varying degrees.
Similarly, the isoglosses shared by MWA and Turoyo/Mlasô suggest that the latter occupy an intermediate position between MWA and the NENA dialects (§ 3.3). Although not nearly as numerous as those connecting Turoyo/Mlahsô with the dialects east of the Tigris, these shared innovations imply the continued existence of an Aramaic dialect continuum across Syria during late antiquity and the early Islamic period, before the shift to Arabic.
The isoglosses in § 3.1 also indicate that the Christian and Jewish dialects of northwestern Iraq have participated together in almost all major post-MidAr innovations, including the creation of an actual present and perfect. Many scholars have previously noted that the communal dialects are less far apart in this region but diverge greatly to the east, in Iran. Since it is improbable that identical formations would have arisen separately in the two communities by chance, the most likely explanation is that individual changes arose within one community and were adopted by the other. This in turn presupposes a degree of interaction and mutual linguistic accommodation between Christians and Jews which did not exist farther east, where certain innovations spread only to the Christian dialects (e.g., penultimate stress, intransitive preterites formed like transitive preterites), while others were confined to the Jewish varieties (e.g., the la- actual present and perfect in Arbil and surroundings, or the remodeling of preterite and perfect in Suleimaniyya/Halabja and Iranian Kurdistan).(FN54)
The above findings, then, have interesting implications for the prehistory of the surviving MA dialects. First, MWA need not be considered one half of Modem Aramaic, parallel to the ancient Anatolian branch of Indo-European, which most scholars now consider to have been the first to separate from the Proto-Indo-European speech community.(FN55) The few innovations which characterize all the dialects to the east, such as loss of the old Aramaic prefixed conjugation and of a separate feminine plural in adjectives and pronouns, do not support the notion of a monolithic ancestral Proto-Eastem-Aramaic or a deep division between those dialects and MWA.(FN56) There is no denying, of course, that Turoyo shares significantly more isoglosses with NENA than with MWA, yet the very existence of post-MidAr features shared by MWA and its nearest MA neighbor vitiates the widely held assumption that MWA has been diverging from the rest of MA for “well over two thousand years” (Hoberman 1997: 313). The relative linguistic isolation of MWA is due to relatively “recent” historical circumstances, i.e., the Islamic conquest and subsequent linguistic Arabization of Syria, which disrupted a previously continuous chain of Aramaic dialects stretching from Palestine up through Syria to northern Mesopotamia.(FN57)
Another significant result is that Mlahô constitutes an outlier, not only with respect to Turoyo, but to the rest of Central/Northeastern MA as well. This is not necessarily inconsistent with the founding legend of the village of Mlahsô (Jastrow 1994a: 79), according to which the village was established by two monks from the Tur ‘Abdin several centuries earlier. In that case, Mlahsô must have split off from Turoyo at a time when the latter had not yet undergone certain characteristic innovations, i.e., when it still had word-final stress, unshortened vowels in closed syllables, and various MidAr lexical items now preserved only in Mlahsô (e.g., avó ‘father,’ renyó ‘trouble’). Like the loss of the prefixed and suffixed conjugations or the pronominal innovations on p. 508, vowel shortening in closed syllables and adoption of certain loanwords (§ 3.2) must have begun at one or more points within the Aramaic-speaking area and spread by diffusion to all surviving dialects except for Mlasô — ust as the innovations in § 3.1, which probably began at a later date, were adopted by a subset of contiguous dialects.
We must conclude that Turoyo/Mlahsô and the mass of NENA dialects are not descended from a single post-Middle Aramaic ancestor, and that the Stammbaum model does not adequately represent the linguistic history of Modem Aramaic, or for that matter of earlier stages of Aramaic (Boyarin 1981: 640-45; cf. Huehnergard 1995). I do not mean thereby to deny the linguistic distance between MWA and the rest of present-day Aramaic, or the isoglosses separating Turoyo (and Mlahsô) from the NENA dialects. The former is obviously connected with the disappearance of all intervening dialects as a result of language shift; there is no reason to think that the spoken Aramaic dialects of Syria in the late first millennium A.D. would have exhibited any sharp breaks from one region to the next. As for the uroyo-NENA divide, the bundle of isoglosses along the upper Tigris is surely due less to the meager physical barrier afforded by this river than to its importance as a border between the Eastern Roman and Persian empires in late antiquity, and likewise as a religious fault line between the Syrian Orthodoxy of the Tur ‘Abdin and the Nestorian Church of the (ex-)Sassanian lands.(FN58) In any case, historical linguists now generally agree that there is no necessary contradiction between “tree” and “wave” models (cf. Hock 1991: 450-52 for Indo-European), but the branching nodes in a Stammbaum cannot in general be interpreted literally as “clean breaks” into two or more speech communities, and certainly not for Aramaic over the past two thousand years.
In sum, the multitude of Modem Aramaic dialects, with all their considerable phonological, morphosyntactic, and lexical diversity, do not lend themselves to clear-cut geographical division into Western, Central, and Northeastern subgroups, but constitute the surviving pieces of a dialect continuum whose geographic continuity was interrupted-in some cases, relatively recently — by language shift. This does not mean that we must dispense entirely with labels such as NENA, as long as we keep in mind that they are valid only as convenient “geographical expressions” — to appropriate Prince Metternich’s notorious description of Italy — and not as historically justified linguistic subgroups. By viewing the Aqlamun valley dialects, Mlahsô, Turoyo, and the many Christian and Jewish NENA varieties as occupying western, central, and (north)eastern portions of a dialect continuum — one which became fragmented through population movements, interrupted communications, and especially language shift — we obtain a more precise understanding of the historical development of these individual dialects, as well as of medieval and modem Aramaic as a whole.
ADDED MATERIAL
RONALD KIM WROKLOW UNIVERSITY
A preliminary version of this paper was presented at the 28th North American Conference on Afroasiatic Linguistics in March 2000. I wish to thank the participants on that occasion, especially Samuel Ethan Fox, Simon Hopkins, Geoffrey Khan, and David Testen, for their helpful comments. The rapid appearance of descriptive grammars of Modern Aramaic dialects since then has necessitated extensive revisions, but I am pleased to learn that my conclusions have been confirmed and even strengthened by recent fieldwork discoveries. Thanks also to Wolfhart Heinrichs, Otto Jastrow, and Geoffrey Khan for reading part or all of earlier drafts, and to colleagues and students at Swarthmore College for inviting me to share my thoughts on Modern Aramaic dialectology in an informal lecture in February 2007.

                                m.                      f.

Mlahsô                           -av                     -a                              (Jastrow 1994a: 29-30)

Turoyo                            -e                      -a

Hertevin                             -ew                     -o (< *-aw)                     (Jastrow 1988: 22-23, n. 2)

Bohtan                               -ew                     -aw                             (Fox 2002: 169)

Bespan                         -u (< *-ew)             -aw                             (Sinha 2000: 70-71)

Baz                                  -[schwa]f                   -aw                             (Mutzafi 2000: 303-4)

Jilu                                 -e                      -o                              (Fox 1997: 44)

Tkhuma                               -e                      -a ˜ -o(FN27)                 (Jacobi 1973: 204-10, 218-20)

C. Zakho                             -eh ˜ -e(h)        -ah ˜ -a(h)(FN28)          (Sabar 1995: 34-35, 37)

J. Zakho                             -e                      -a

C. Aradhin                           -e                      -a(h)                           (Krotkoff 1982: 20)

J. Aradhin                           -e                      -a                              (Mutzafi 2002b: 480)

C. Nerwa                             -e                      -a                              (Talay 2001: 10)

J. [Graphic Character Omitted]Amadiya                        -e                      -a                              (Hoberman 1989)

J. Barzani                           -e                      -a                              (Mutzafi 2002a: 54)

C. Mosul plain                       -eh ˜ -eh          -ah ˜ -ah

Qaraqosh                             -[schwa]h                -ah                          Khan 2002a: 76ff.)(FN29)

J. Arbel                             -eu                     -aw                             (Khan 1999a: 8, 82)

C. Koy Sanjaq                        -ew                     -aw                             (cf. Mutzafi 2004a: 257)

J. Koy Sanjaq                        -ew                     -aw                             (Mutzafi 2004b: 60)

J. Rustaqa                           -eu                     -aw                             (Khan 2002b: 398)

J. Sulemaniyya/Halabja            -eu, -ew                -aw                             (Khan 2004: 74)

C. Urmia                             -u (< *-ew)             -o (< *-aw)

J. Azerbaijani                       -ew                     -aw                             (Garbell 1965: 59)

Senaya                               -e                      -e(!)                           (Panoussi 1990: 111)

J. Iranian Kurdistan (Kerend)        -ef                     -af

                preterite               perfect

intransitive         qím-li               m. qím-en, f. qím-an

transitive           grís-li           m. grisélen(FN44) f. gritélan (< grisá-ilen, gristá-ilan) ˜

                                             m. gris-én, f. grist-án(FN45)

MidAr                                   Ma[Graphic Character Omitted]lula    Mlahsô   Turoyo   Hertevin    J. Zakho    C. Urmia

*bra ‘son’                            ebra            ebro         abro        ebra        [brona]    [brunoe](FN52)

*bnayya ˜ *bne ‘sons’            bno                       abne

*dma ‘blood’                          edma                         admo        demma       dimma      dimma,

*sma ‘name’                        esma                      [schwa]smo   semma    simma   simmoe

*snayya ˜ *ne ‘yers’       isno      [snaye]   [schwa]sne   senne               sinni

*tmal ‘yesterday’                                     esmol        atm[schwa]l     etmal       timmal     timmoel

FIG. 1. Distribution of features 1-3 in § 3.1, with approximate isoglosses for Christian (C) and Jewish (J) dialects: 1) final vs. penultimate stress; 2) 3sg. pronominal suffixes with nouns and prepositions < *-ew, *-aw vs. *-eh, *-ah; 3) lsg. present ending -na vs. -n (<– MidAr *-na). (This and the following diagram are adapted from Mutzafi 2004b: 13.)
FIG. 2. Distribution of features 4-6 in § 3.1, with approximate isoglosses for Christian (C) and Jewish (J) dialects: 4) intransitive preterite (subject endings vs./- + object suffixes); 5) presence vs. absence of an actual/continuous present based on the post-MidAr infinitive; 6) presence vs. absence of a perfect based on the MidAr passive participle in status emphaticus.

FOOTNOTES
1. See the comments by Hoberman (1990: 79-81) and Jastrow (1990: 89). Especially poignant is Jastrow’s lament (1994a: 5ff.) that had he realized the enormous significance of the Mlahsô dialect earlier on, he might have taken greater effort to locate and record more material from its surviving speakers.
2. Except for no. 1, where the relevant forms (prefixed conjugation) have all disappeared. The features are taken from Boyarin 1979: 614-15.
3. The change in Syriac may be assigned to approximately A.D. 200, since y-is attested in inscriptions and magic bowls from before that date (Drijvers and Healey 1999: 29-30).
4. Contracted from *-ayya, which survives only in monosyllabic stems such as bnayya ‘sons,’ snayya ‘years,’ pl. tant, mayya ‘water.’
5. C. Urmia forms are cited in a modified version of the Soviet novyj alfavit, with oe for [schwa], [Graphic Character Omitted] for the soft jer, and s for ¸s.
6. Also in the (indeclinable) Turoyo elative, which continues the status absolutus of the m. sg. in predicative use: safiro ‘beautiful,’ hawri saf[schwa]r mina-yo ‘my friend is more beautiful than her.’
7. With the sole exception of the perfect of [Graphic Character Omitted]hwy, which survives as Turoyo enclitic (Midyat) -wo, (Mid[schwa]n) -wa ‘s/he was’ < *hwa, *hway- (cf. Syriac 3sg. m. hwa, 2sg. m./f. hwayt), and as an invariant past-tense clitic -wa in the NENA dialects. Modern Mandaic has preserved the perfect but, like Turoyo/Mla[Graphic Character Omitted]sô and NENA, eliminated the imperfect in favor of the MidAr periphrasis based on the present active participle in status absolutus (Macuch 1993: 68ff.; Jastrow 1997a: 360ff., 370).
8. Certain varieties of NENA also appear to have raised and rounded stressed MidAr *a to *o, e.g., the Bohtan dialect spoken in Georgia (Fox 2002: 167). In these dialects, however, o < *a in open syllables regularly alternates with short a in closed syllables, e.g., pótax ‘he opens’ vs. pátxa ‘she opens,’ revealing that this is a relatively recent change (cf. Jastrow 1997a: 352). See also below, §3.3 and n. 51.
9. Jastrow (1985a: xxi) describes this change as “Degeminierung mit Ersatzdehnung des vorhergehenden Vokals,” but phonemic vowel length was eliminated in all non-MWA dialects through shortening of inherited long vowels in closed syllables and lengthening of short vowels in open syllables; subsequently it was reintroduced through loanwords (from Kurdish, Arabic, and Turkish), reanalysis of later phonological developments, etc.
10. Except for II-y verbs, e.g., Turoyo (Midyat) su[schwa]m, pi. súmu ‘do!’ to [Graphic Character Omitted]sym.
11. And Modern Mandaic, e.g., m. getól, f. g[supe]/[subu]túl, although Macuch (1993: 70) also gives forms with stem vowel a, e.g., obád, obód vs. f. obúd. Exceptions to this pattern are the result of local innovations, e.g., J. Rustaqa and Sulemaniyya qátil, pl. qátilmun ˜ qátlun ‘kill!’ <– present stem qatíl with retracted stress, after the pattern of derived stems (e.g., mázdir, pl. mázdirmun ˜ mázdirun ‘send!’ beside present mazdír; Khan 2002b: 405-6, 2004: 93).
12. This leveling was facilitated by the loss of the old imperfect in the eastern Aramaic dialects, which left the imperative stem synchronically isolated.
13. And the now extinct dialect of the villages of Mlahsô and [Graphic Character Omitted]Ansa, discovered by Jastrow in the late 1960s. See Jastrow 1985 and 1994a.
14. For a complete list, see Tsereteli 1977: 245-47.
15. I follow Huehnergard (1995: 266ff.) in assuming that the dialectal diversity of early Aramaic (beginning of the 1st millennium B.c.) was to a large extent leveled by the spread of Imperial Aramaic, and that this latter is the approximate ancestor of the subsequent dialects, standing in much the same relationship to Middle and Modern Aramaic as, say, Hellenistic Koine to (almost all) medieval and modern Greek dialects.
16. Hoberman’s diagram (1989: 5) actually has Turoyo, NENA, and Modern Mandaic branching off from different points within East Aramaic, with the first two much closer to each other than to Mandaic. Although this model comes closer to the notion of a dialect continuum (see below, §2.3), it still treats the NENA dialects as having a single common intermediate ancestor separate from both Turoyo and Mandaic, i.e., a “Proto-NENA,” the personal pronouns and suffixes of which he attempts to reconstruct in Hoberman 1988.
17. Nakano 1973 gives a collection of texts recorded from a speaker of the Jewish dialect of Gzira (Cizre), just east of the Tur ‘Abdin, but provides no grammatical analysis.
18. This interpretation is followed by Kaufman 1997: 117-18, who distinguishes three Middle Aramaic varieties (Western, Syrian, and Eastern) rather than the traditional Western vs. Eastern dichotomy.
19. On the dialectal position of Modern Mandaic, see n. 56.
20. As Warren Cowgill put it in his survey of the Indo-European family, “Two or more languages, related to one another, are placed together in a subgroup when they have a significant number of innovations in common which are not present in any other related language (or when they can be proven to have arisen independently elsewhere)” (Cowgill and Mayrhofer 1986: 16; my translation).
21. See Harrison 2003: 233-38, although his confessed “despair” is not entirely justified. After all, many phonetically “natural” and/or crosslinguistically common sound changes are much more likely to occur in one direction than the other, e.g., debuccalization, rhotacism of *z > *r, palatalization of velars before front vocalics, etc. Similarly, the distribution of lexical features is often not probative for subgrouping, but unidirectional semantic shifts can be: e.g., Proto-Indo-European *wiH-ro-and *yeb[suph]-preserve their original meanings of ‘young’ resp. ‘enter’ in Tocharian, but have shifted to ‘man’ resp. ‘f*ck’ in most other Indo-European languages. Nevertheless, morphological innovations do usually play a crucial role in subgrouping arguments. Note for instance that of the 37 non-lexical characters in Ringe et al.’s (2002) computational investigation of the best-fitting cladistic tree for Indo-European, fully 15 are morphological.
22. With variable retraction to the preceding syllable, especially in certain categories such as vocatives; for examples, see Khan 2002b.
23. Data for this dialect are based on Hoberman 1993 and Sabar 1995.
24. I.e., the so-called Fellihi dialects. Data for these are drawn from the classic descriptions of Sachau 1895 and Rhétoré 1912, along with Sabar 1978.
25. I will argue elsewhere (R. Kim forthcoming a) that within the latter group, Turoyo, Hertevin, and C. Aradhin have underlying final stress which is retracted one syllable by a (near-)categorical constraint against word-final surface stress, whereas other dialects (e.g., Jilu) have gone one step further and reanalyzed underlying stress placement on the penultimate.
26. The 3sg. m. must go back to *-ayhu, the allomorph after plural nouns and prepositions which originally ended in *-ay (cf. Imperial Aram, -auhi, Babylonian Aram. -e(h)); this presumably served as the basis for the creation of f. *-ahu. On the complicated history of these variants, see Boyarin 1981: 615-18 and (for MA) Polotsky 1961: 16-17.
27. The variant -o is “äusserst selten gebraucht” (Jacobi 1973: 204 n. 1).
28. The variants m./f. -u(h), f. -aw in the speech of Rev. Harboli (Sabar 1995: 34-35) may reflect interference from his native MA dialect of Harbol, in the Bohtan area (cf. Bohtan -ew, -aw, Besp[schwa]n -u, -aw).
29. In this and some neighboring Christian dialects (e.g., Alqosh -eh, -ah, C. Zakho -eh, -ah; cf. Bart[schwa]lla -eh, -ah), final *-u in *-ehu, *-ahu was apparently apocopated, and the now word-final *h was strengthened to a pharyngeal (Khan 2002a: 17). The dialect of Hassane has -ah for both m. and f., e.g., betah ‘his, her house’ (Jastrow 1997b: 277-78).
30. Admittedly, the somewhat irregular distribution of the two groups might make one wonder whether the leveling of the two sets of suffixes in Turoyo might have been independent of the same change in the northwest Iraqi dialects. The Tiyari dialects now spoken along the Khabur River in Syria present a mixed picture: four tribal varieties have leveled m. -e, f. -a, while the rest retain separate nominal and verbal suffixes (Talay 1999: 171).
31. Forms from Jastrow 1997a: 363; Hopkins 1989: 427.
32. These dialects have substituted C[sub1]aC[sub2]C[sub3]-ena (with the ending -ena from III-y verbs, cf. Syr. hazena) for the older 1sg. m. C[sub1]aC[sub2]iC[sub3]-na, no doubt under the influence of 1sg. f. C[sub1]aC[sub2]C[sub3]a-na, which could be reanalyzed as C[sub1]aC[sub2]C[sub3]-ana.
33. The Cudi dialects have actually extended the pattern of 1sg. m. -[schwa]n ˜ -ena to 2sg. m. -[schwa]t ˜ -etena and 1pl. -ux ˜ -uxena, whereas C. Nerwa has created 3sg. m. y-dámxena beside y-dam[schwa]x ‘he sleeps.’ Interestingly, the J. Koy Sanjaq long variants in e.g., 2sg. m. baxét([schwa]n), f. baxyát([schwa]n) ‘you cry’ and 1pl. baxéx([schwa]n) ‘we cry’ could be from earlier *-etena, *-atena, *-ixena/*-exena, comparable to the long forms in Cudi.
34. These are the “neutrische Verben” of Jastrow (1985a: 71ff.). For more details, and the situation in Mlahsô, see §3.3, no. 2 and n. 53.
35. On the intransitive perfects Hertevin [Graphic Character Omitted]iz-en ‘I (m.) have gone’ and J. Azerbaijani qim-en ‘I (m.) have stood,’ see no. 6 below.
36. Cf. also J. Sakkiz (Iranian Kurdistan) intransitive (y)tíw-na ‘I (m.) sat down’ vs. transitive xzé-li ‘I saw’ (Goldenberg 1998: 68).
37. Attested in Jewish Babylonian Aramaic and Mandaic, e.g., Mand. mpaqade ‘order(ing),’ matsote ‘hear(ing)’ (Nöldeke 1875: xxvii, 142-44, 233-34). The Jewish Azerbaijani dialects have eliminated the distinction between P[sup[schwa]][Graphic Character Omitted]al and Pa[Graphic Character Omitted][Graphic Character Omitted]el/Af[Graphic Character Omitted]el and collapsed them into a single binyan for all verb roots, with continuous present C[sub1]aC[sub2]oC[sub3[sup-]] originally proper to the Pa[Graphic Character Omitted][Graphic Character Omitted]el/Af[Graphic Character Omitted]el conjugation (Garbell 1965: 62ff.; Hoberman 1991: 60-61); similarly for J. Sulemaniyya and Halabja, with garosé ‘pulling,’ 3sg. m. garosá-y ‘he is pulling’ like mardoxé ‘boiling,’ mardoxá-y ‘he is boiling’ (Khan 2004: 80-82, 2005: 360-61).
38. This is still largely true in dialects such as J. Sulemaniyya and Halabja, where the actual/continuous present with active participle + copula (e.g., ‘o kwása-y tèx’ ‘he is coming down’; Khan 2004: 318-19) is “not very extensively attested in the text corpus.” Khan infers that this formation is “losing ground to the indicative qatil form, with which it overlaps in function,” but it seems more likely that the grammaticalization of the actual present was just getting underway in these dialects, and that the inherited *C[sub1]aC[sub2]iC[sub3[sup-]] type continued to be used for most present functions. (Cf. Qaraqosh, where the construction copula + active participle is restricted to the verbs ‘go’ and ‘come,’ e.g., k-it[schwa] ‘azola’ ‘he is going’ [Khan 2002a: 349].)
39. Cf. also the copula + finite verb constructions in Besp[schwa]n and J. [Graphic Character Omitted]Amadiya, e.g., Besp[schwa]n 3sg. m. hole pat[schwa]x ‘he is opening’ (Hoberman 1989: 45; Sinha 2000: 131), beside the usual infinitive + copula.
40. Khan 2002a: 18 describes la-palix as a “typologically more advanced form of the progressive construction that is used in the Qaraqosh dialect, since this particle appears to be a fossilized form of the copula that has come to be used in an invariable form throughout the paradigm.” The Jewish dialect of Dobe has na- for Arbel la- (Khan 1999b: 219, 2000: 323).
41. Goldenberg 1992: 131-33 examines a sample of such forms and suggests that their common value is rather one of “facts realized or presumed inferentially on the part of the persons involved,” i.e., completed action inferred but not personally attested. But as it is often difficult to draw a clear line between perfect and narrative value (cf. the Turkish suffix -mis-, which Goldenberg cites in this connection), I retain the label “perfect” for the Turoyo formation for the time being. On the perfect in Mlahsô and its prehistory, see n. 53.
42. This particle recurs in Christian Koy Sanjaq, but there it is prefixed to the past participle (and in the first and second persons, the copula) to form a Stative perfect. Cf. 3sg. m. la-skíra, f. la-sk%[schwa]rta ‘is lost,’ 1sg. m. la-skíray[schwa]n, f. la-sk%[schwa]rtayan vs. dynamic perfect skírele, sk%[schwa]rtela ‘has lost; has gotten lost,’ 1sg. Skíray[schwa]n, sk%[schwa]rtayan (Mutzafi 2004a: 255-56, 260-62).
43. In contrast, the Cudi dialects immediately to the east do have transitive perfects, although only the full (tonic) copula is attested, e.g., hole mlita ‘he has filled’ (Sinha 2000: 106).
44. 44. In the dialects of southern Iranian Azerbaijan, m. gríslen.
45. Contracted from the longer forms, and reanalyzed as consisting of past participle + subject suffixes (I sg. m. -en, f. -an).
46. 46. In a few dialects, the preterite has now become completely resistant to pronominal object marking, so that one must have recourse to preterite prefix (qam-, kim-, etc.) + present: cf. C. Nerwa sq[schwa]l-li ‘I took’ vs. q[schwa]m saql[schwa]nna ‘I(m.) took her,’ C. Koy Sanjaq pt[schwa]x-le ‘he opened’ vs. qa-pat[schwa]x-le ‘he opened it (m.),’ or Senaya pseh-le vs. qam-pseh-le ‘id.’ (Talay 2001: 18; Mutzali 2004a: 256; Heinrichs 2002: 242-43).
47. 47. Less clear is the prehistory of the pattern first documented in Hertevin, whereby the subject marker takes the endings of the present when followed by an (extraconjugational) object marker, e.g., ‘lzé-li ‘I saw,’ gwer-roh ‘you (m.) married’ (< *-loh) but hzé-len-nah ‘I (m.) saw you (f.)’ (< *-len-lah), gwer-ret-ta ‘you (m.) married her’ (< *-let-la) (Jastrow 1988: 61-62). The subsequent discovery of examples in early MA manuscripts from northern Iraq, e.g., maxwaw li zuza de-hvel-len-nawxon ‘show me the money that I gave you!’ (seventeenth century), dex d-mer-ren-nux ‘as I told you’ (nineteenth century; Pennacchietti 1991, 1994: 272-75; Goldenberg 1993: 300-302, 2000: 84-86; Mengozzi 2002: 34-35, 45-47), leads Mengozzi to the conclusion that this is “an archaic feature preserved in marginal areas, rather than an innovation of the Hertevin dialect.” More precisely, forms such as *hzelin-lax, *gwir-lit-la arose as variants of original *ze-li-lax, *gwir-lux-la sometime before the seventeenth century in dialects of southeastern Turkey and western Iraqi Kurdistan, but were later eliminated everywhere except in Hertevin.
48. 48. Heinrichs 2002 makes the same observation for the distribution of the (general) present with prefix kV-(258-59, 260) and the perfect based on the passive participle in status emphaticus, i.e., feature no. 6 here (261). In a similar vein, Khan 2002a: 17 ascribes certain morphological peculiarities of the Qaraqosh dialect to “a greater tendency to archaism… rather than to the sharing of innovations with other dialects. These archaisms may have been preserved in the Qaraqosh dialect on account of its location on the periphery of the NENA area.” Especially striking is the preservation of the original form of derived stem imperatives and infinitives in Qaraqosh, whereas Mlahsô, Turoyo, and the rest of NENA have generalized prefixal m- from the present and preterite (as already in Syriac; Khan 2002a: 11-13, 86, 2007: 10-13).
49. Mutzafi 2004b: 212 derives the latter forms from an unattested masculine *[Graphic Character Omitted]am(m)a corresponding to Syr. [Graphic Character Omitted]amta ‘paternal aunt,’ but this will not account for the geminate of Turoyo [Graphic Character Omitted]ámmo.
50. And perhaps other pre-1915 villages to the north and west of the Tur ‘Abdin, including those whose MA speakers had shifted to Anatolian Arabic in comparatively recent times (e.g., the Christian villagers of Quturbul, who reportedly spoke a MA dialect until c. 1850; Heinrichs 2002: 246 n. 28).
51. Specifically, raising and rounding of *a > *o in Turoyo and Mlahsô must have preceded *VCC > *V:C (§ 2.1) as well as the Turoyo shortening of vowels in closed syllables, which in turn preceded *iw, *uw > u (e.g., subjunctive 3sg. kotu ‘that he writes’ < *kotiw). For details, see R. Kim forthcoming b.
52. The inherited word for ‘son’ has been extended to brona (C. Urmia bruna) everywhere east of Hertevin.
53. The discrepancy in preterite and perfect formation between Turoyo and Mlahsô (Jastrow 1994a: 14-15) is due to later innovations in both languages. The generalization of *CaCCiC- > *CaCiC- as the perfect of intransitives led to an asymmetry in the verbal system, as no corresponding perfect existed for transitive verbs. Mlahsô resolved this imbalance by extending the intransitive perfect *CaCiC- to transitives, whereas Turoyo lost the aspectual distinction between preterite and perfect. Subsequently, Turoyo reintroduced an aspectual contrast by prefixing ko- to the preterite to form a new perfect. Cf. the paradigms for ‘sleep’ and ‘pull’ in Mlahsô, Turoyo, and the latest stage ancestral to both:

                        preterite       perfect

pre-M/T   intrans.        (*dmíx-le)   *damíx

             trans.          *gris-le     –

Mlahsô   intrans.        dmíx-le      damíx

             trans.          grís-le   garís

Turoyo    intrans.        dám[schwa]x        [ko-dám[schwa]x]

             trans.          gr%[schwa]s-le   [ko-gr%[schwa]s-le]

54. Heinrichs 1990: xii, 2002: 245 suggests that the main opposition within NENA is not one of Christian vs. Jewish, but rather of Christian and Northwestern Jewish vs. the remaining Jewish dialects. Although this is largely true for the features examined in § 3.1, the distribution of nos. 2 and 3 and the variation at the western edge of nos. 4-6 are best explained by the geographical diffusion of innovations originating in a focal area in northwestern Iraq.
55. Edgar H. Sturtevant first suggested this interpretation in the 1930s (in rather extreme form as the “Indo-Hittite hypothesis”); it was largely abandoned in following decades, but since the 1970s has emerged as the consensus view among Indo-Europeanists. See Cowgill 1974 and more recently Ringe et al. 1998; Jasanoff 2003: passim.
56. The survival of the MidAr perfect in Modem Mandaic suggests that the prefixed conjugation was lost before the suffixed conjugation in spoken Aramaic of the eastern regions. Although I have not investigated it in detail, the relation between Modem Mandaic (MM) and the rest of MA also supports the historical interpretation of MA as a dialect continuum. In many cases, MM, like Mlahsô/Turoyo (and MWA), preserves MidAr forms which have been lost or remodeled in NENA. Thus the MM 3sg. pronoun is m. huy, f. hid (cf. Turoyo [Midyat] huwe, hiya), without the initial a- of the NENA dialects; and the Ipl. verbal suffix ends in -en (cf. Turoyo -ina, Syr. m. -innan), not -ax as in NENA. For other features, MM sides with NENA, at least partially: e.g., the 3pl. pronoun can end in -i(hanni, beside hánnex), like NENA ani. Note also that Mandaic retains pres. lsg. -na (m. qa-gatelna, f. qa-galana ‘I kill’) like Mlahsô/Turoyo, Qaraqosh, and the eastern Jewish dialects, confirming this as a Randarchaismus. (All data are from Macuch 1993.)
57. And down into central Mesopotamia, where Jewish Babylonian Aramaic was spoken through the first millennium A.D., and the homeland of the Mandaeans in Khuzistan; perhaps also east into present-day Kurdistan and Iranian Azerbaijan, although the early migratory history of these MA speech communities remains largely obscure (cf. Hopkins 1999: 326 on the Aramaic-speaking Jews of northwestern Iran).
58. On isogloss bundles and their frequent correlation with cultural traits, see Chambers and Trudgill 1998: 94-96, 100-102. Interestingly, the cascade model, in which innovations begin in larger population centers and diffuse to smaller communities, and later to rural areas (Chambers and Trudgill 1998: 166ff.), is of limited usefulness for MA dialectology: most of the Turoyo/NENA innovations discussed in § 3 appear to have spread in a largely horizontal manner, among adjacent dialects. However, the existence of so many changes common to the dialects of northwestern Iraq and Iranian Azerbaijan suggests that cascade-style diffusion from the Mosul plain to the Urmia area may have played a role. That, for example, the C. Urmia dialect shares penultimate stress and a fully grammaticalized perfect (formed from the status emphaticus of the MidAr passive participle) with nearly all the Christian and Jewish varieties of Iraq, whereas the Jewish dialects of Iranian Azerbaijan retain the older features, implies that such contacts between the two regions were more significant and/or longer-lasting within the Christian communities than the Jewish ones.

REFERENCES
Arnold, Werner. 1990. Das Neuwestaramäische. V. Grammatik. Wiesbaden: Harrassowitz.
Arnold, Werner, and Hartmut Bobzin, eds. 2002. “Sprich doch mit deinen Knechten aramäisch, wir verstehen es!” 60 Beiträge zur Semitistik: Festschrift für Otto Jastrow zum 60. Geburtstag. Wiesbaden: Harrassowitz.
Blau, Joshua. 1968. Review of Hellmut Ritter, Turoyo, die Volkssprache der syrischen Christen des Tur ‘Abdin. A: Texte, Bd. I and Otto Jastrow, Laut- und Formenlehre des neuaramäischen Dialekts von Midin im Tur ‘Abdin. BSOAS 31: 605-10.
Boyarin, Daniel. 1981. An Inquiry into the Formation of the Middle Aramaic Dialects. In Bono Homini Donum: Essays in Historical Linguistics in Memory of J. Alexander Kerns, ed. Yoël L. Arbeitman and Allan R. Bomhard. Amsterdam: Benjamins. Vol. 2, pp. 613-49.
Chambers, J. K., and Peter Trudgill. 1998. Dialectology. 2nd ed. Cambridge: Cambridge Univ. Press.
Cowgill, Warren. 1974. More Evidence for Indo-Hittite: the Tense-Aspect Systems. In Proceedings of the Eleventh International Congress of Linguists, ed. Luigi Heilmann. Bologna: Il Mulino. Vol. 2, pp. 557-70.
Cowgill, Warren, and Manfred Mayrhofer. 1986. Indogermanische Grammatik, vol. I. Heidelberg: Carl Winter.
Drijvers, Han J. W, and John F. Healey. 1999. The Old Syriac Inscriptions of Edessa and Osrhoene: Texts, Translations, and Commentary. Leiden: Brill.
Duval, Rubens, 1896. Notice sur les dialectes néo-araméens. Mémoires de la Société de Linguistique de Paris 9: 125-35.
Fox, Samuel Ethan. 1990. Cliticization in Neo-Aramaic. In Studies in Neo-Aramaic, ed. Wolfhart Heinrichs. Atlanta: Scholars Press. Pp. 69-78.
Fox, Samuel Ethan. 1994. The Relationships of the Eastern Neo-Aramaic Dialects. JAOS 114: 154-62.
Fox, Samuel Ethan. 1997. The Neo-Aramaic Dialect of Jilu. Wiesbaden: Harrassowitz.
Fox, Samuel Ethan. 2002. A Neo-Aramaic Dialect of Bohtan. In Arnold and Bobzin 2002. Pp. 165-80.
Garbell, Irene. 1965. The Jewish Neo-Aramaic Dialect of Persian Azerbaijan: Linguistic Analysis and Folkloristic Texts. The Hague: Mouton.
Goldenberg, Gideon. 1992. Aramaic Perfects. Israel Oriental Studies 12: 113-37.
Goldenberg, Gideon. 1993. Review of Jastrow 1988. JSS 38: 295-308.
Goldenberg, Gideon. 1998. Review of Jastrow 1994a. JSS 43: 63-70.
Goldenberg, Gideon. 2000. Early Neo-Aramaic and Present-day Dialectal Diversity. 755 45: 69-89.
Harrison, S. P. 2003. On the Limits of the Comparative Method. In The Handbook of Historical Linguistics, ed. Brian D. Joseph and Richard P. Janda. Oxford: Blackwell. Pp. 213-43.
Heinrichs, Wolfhart, ed. 1990. Studies in Neo-Aramaic. Atlanta: Scholars Press.
Heinrichs, Wolfhart, ed. 2002. Peculiarities of the Verbal System of Senaya within the Framework of North Eastern Neo-Aramaic (NENA). In Arnold and Bobzin 2002. Pp. 237-68.
Hetzron, Robert. 1969. The Morphology of the Verb in Modern Syriac (Christian Colloquial of Urmi). JAOS 89: 112-27.
Hetzron, Robert. 1976. Two Principles of Genetic Reconstruction. Lingua 38: 89-108.
Hetzron, Robert, ed. 1997. The Semitic Languages. London: Routledge.
Hoberman, Robert D. 1988. The History of the Modern Aramaic Pronouns and Pronominal Suffixes. JAOS 108: 557-75.
Hoberman, Robert D. 1989. The Syntax and Semantics of Verb Morphology in Modern Aramaic: A Jewish Dialect of Iraqi Kurdistan. New Haven: American Oriental Society.
Hoberman, Robert D. 1990. Reconstructing Pre-modern Aramaic Morphology: the Independent Pronouns. In Heinrichs 1990. Pp. 79-88.
Hoberman, Robert D. 1991. Formal Properties of the Conjugations in Modern Aramaic. In Yearbook of Morphology 1991, ed. Geert Booij and Jaap van Marie. Dordrecht: Kluwer. Pp. 49-64.
Hoberman, Robert D. 1993. Chaldean Aramaic of Zakho. In Semítica: Serta Philologica Constantino Tsereteli Dicata, ed. Riccardo Contini, Fabrizio A. Pennacchietti, and Mauro Tosco. Turin: Zamorani. Pp. 115-26.
Hoberman, Robert D. 1997. Modern Aramaic Phonology. In Phonologies of Asia and Africa, ed. Alan S. Kaye. Winona Lake, Ind.: Eisenbrauns. Pp. 313-35.
Hock, Hans. 1991. Principles of Historical Linguistics. 2nd ed. Berlin: Mouton de Gruyter.
Hopkins, Simon. 1989. Neo-Aramaic Dialects and the Formation of the Preterite. 755 34: 413-32.
Hopkins, Simon. 1999. The Neo-Aramaic Dialects of Iran. In Irano-Judaica IV. Studies Relating to Jewish Contacts with Persian Culture Throughout the Ages, ed. Shaul Shaked and Ammon Netzer. Jerusalem: Ben-Zvi Institute for the Study of Jewish Communities in the East. Pp. 311-27.
Hopkins, Simon. 2002. Preterite and Perfect in the Jewish Neo-Aramaic of Kerend (Southern Iranian Kurdistan). In Arnold and Bobzin 2002. Pp. 281-98.
Huehnergard, John 1995. What is Aramaic? Aram 7: 261-82.
Jacobi, Heidi. 1973. Grammatik des thumischen Neuaramäisch (Nordostsyrien). Wiesbaden: Franz Steiner.
Jasanoff, Jay H. 2003. Hittite and the Indo-European Verb. Oxford: Oxford Univ. Press.
Jastrow, Otto. 1971. Ein neuaramäischer Dialekt aus dem Vilayet Siirt (Ostanatolien). ZDMG 121: 215-22.
Jastrow, Otto. 1985a. Laut- und Formenlehre des neuaramäischen Dialekts von Midin im Tur ‘Abdin. 3., ergänzte Auflage. Wiesbaden: Harrassowitz.
Jastrow, Otto. 1985b. Mlahs%o: an Unknown Neo-Aramaic Language of Turkey. 755 30: 265-70.
Jastrow, Otto. 1988. Der neuaramäische Dialekt von Hertevin (Provinz Siirt). Wiesbaden: Harrassowitz.
Jastrow, Otto. 1990. Personal and Demonstrative Pronouns in Central Neo-Aramaic: A Comparative and Diachronic Discussion based on Turoyo and the Eastern Neo-Aramaic Dialect of Hertevin. In Heinrichs 1990. Pp. 89-103.
Jastrow, Otto. 1994a. Der neuaramäische Dialekt von Mlahsô. Wiesbaden: Harrassowitz.
Jastrow, Otto. 1994b. Neuentdeckte aramäische Dialekte der Türkei. In XXV. Deutscher Orientalistentag vom 8. bis. 13.4.1991 in München: Vorträge im Auftrag der Deutschen Morgenländischen Gesellschaft, ed. Cornelia Wunsch. Stuttgart: Franz Steiner. Pp. 69-74.
Jastrow, Otto. 1996. Passive Formation in Turoyo and Mlahsôo. Israel Oriental Studies 16: 49-57.
Jastrow, Otto. 1997a. The Neo-Aramaic Languages. In Hetzron 1997. Pp. 334-77.
Jastrow, Otto. 1997b. Zum neuaramäischen Dialekt von Hassane (Provinz ¸Sirnak). In Humanism, Culture, and Language in the Near East: Studies in Honor of Georg Krotkoff, ed. Asma Afsaruddin and A. H. Mathias Zahniser. Winona Lake, Ind.: Eisenbrauns. Pp. 275-81.
Jastrow, Otto. 2002. Neo-Aramaic Dialectology: the State of the Art. In Semitic Linguistics: The State of the Art at the Turn of the Twenty-First Century, ed. Shlomo Izre’el. Winona Lake, Ind.: Eisenbrauns. Pp. 365-77.
Jastrow, Otto. 2005. Der bestimmte Artikel im Aramäischen: ein Blick auf 3000 Jahre Sprachgeschichte. In Studia Semítica et Semitohamitica: Festschrift für Rainer Voigt anläbetalich seines 60. Geburtstages am 17. Januar 2004, ed. Bogdan Burtea, Josef Tropper, and Helen Younansardaroud. Münster: Ugarit-Verlag. Pp. 137-50.
Kapeliuk, Olga. 2005. From H. J. Polotsky’s Nachlass on the Verb in Urmi. In Mengozzi 2005b. Pp. 349-58.
Kaufman, Stephen A. 1997. Aramaic. In Hetzron 1997. Pp. 114-30.
Kim, Ronald I. Forthcoming a. Clitics, Stress, and Verbal Inflection in Modern Aramaic Dialects.
Kim, Ronald I. Forthcoming b. Towards a Historical Phonology of Modern Aramaic: the Relative Chronology of Turoyo Sound Changes. To appear in Proceedings of the XIII Incontro Italiano di Linguistica Afroasiatica. Udine: Forum Editrice Universitaria Udinese.
Khan, Geoffrey. 1999a. A Grammar of Neo-Aramaic: The Dialect of the Jews of Arbel. Leiden: Brill.
Khan, Geoffrey. 1999b. The Neo-Aramaic Dialect Spoken by Jews from the region of Arbel (Iraqi Kurdistan). BSOAS 62: 213-25.
Khan, Geoffrey. 2000. The Verbal System of the Jewish Neo-Aramaic Dialect of Arbel. JAOS 120: 321-32.
Khan, Geoffrey. 2002a. The Neo-Aramaic Dialect of Qaraqosh. Leiden: Brill.
Khan, Geoffrey. 2002b. The Neo-Aramaic Dialect of the Jews of Rustaqa. In Arnold and Bobzin 2002. Pp. 395-410.
Khan, Geoffrey. 2004. The Jewish Neo-Aramaic Dialect of Sulemaniyya and Halabja. Leiden: Brill.
Khan, Geoffrey. 2005. The Verbal System of the Jewish Neo-Aramaic Dialect of Sulemaniyya. In Mengozzi 2005b. Pp. 359-70.
Khan, Geoffrey. 2007. The North-eastern Neo-Aramaic Dialects. JSS 52: 1-20.
Krotkoff, Georg. 1982. A Neo-Aramaic Dialect of Kurdistan: Texts, Grammar, and Vocabulary. New Haven: American Oriental Society.
Krotkoff, Georg. 1985. Studies in Neo-Aramaic Lexicology. In Biblical and Related Studies Presented to Samuel Iwry, ed. Ann Kort and Scott Morschauser. Winona Lake, Ind.: Eisenbrauns. Pp. 123-34.
Lipi%nski, Edward. 1997. Semitic Languages: Outline of a Comparative Grammar. Leuven: Peeters.
Maclean, Arthur John. 1895. Grammar of the Dialects of Vernacular Syriac as Spoken by the Eastern Syrians of Kurdistan, North-West Persia, and the Plain of Mosul with Notices of the Vernacular of the Jews of Azerbaijan and of Zakhu near Mosul. Cambridge: Cambridge Univ. Press.
Macuch, Rudolf. 1993. Neumandäische Texte im Dialekt von Ahwaz. Unter Mitwirkung von Guido Dankwarth. Wiesbaden: Harrassowitz.
Mengozzi, Alessandra. 2002. Israel of Alqosh and Joseph of Telkepe: A Story in a Truthful Language; Religious Poems in Vernacular Syriac (North Iraq, 17th Century): Volume 2: Introduction and Translation. Louvain: Peeters.
Mengozzi, Alessandra. 2005a. Neo-Aramaic and the So-called ‘Decay of Ergativity’ in Kurdish. In Proceedings of the 10th Meeting of Hamito-Semitic (Afroasiatic) Linguistics (Florence, 18-20 April 2001), ed. Pelio Fronzaroli and Paolo Marrassini. Florence: Dipartimento di Linguistica, Università di Firenze. Pp. 239-56.
Mengozzi, Alessandra, ed. 2005b. Studi Afroasiatici: XI Incontro Italiano di Linguistica Camitosemitica. Milan: Franco Angeli.
Murre-van den Berg, H. L. 1999. From a Spoken to a Written Language: The Introduction and Development of Literary Urmia Aramaic in the Nineteenth Century. Leiden: Nederlands Instituut voor het Nabije Oosten.
Mutzafi, Hezy. 2000. The Neo-Aramaic Dialect of Maha Khtaya d-Baz. Phonology, Morphology and Texts. JSS 45: 293-322.
Mutzafi, Hezy. 2002a. Barzani Jewish Neo-Aramaic and its Dialects. Mediterranean Language Review 14: 41-70.
Mutzafi, Hezy. 2002b. On the Jewish Neo-Aramaic Dialect of Aradhin and its Dialectal Affinities. In Arnold and Bobzin 2002. Pp. 479-88.
Mutzafi, Hezy. 2004a. Features of the Verbal System in the Christian Neo-Aramaic Dialect of Koy Sanjaq and Their Areal Parallels. JAOS 124: 249-64.
Mutzafi, Hezy. 2004b. The Jewish Neo-Aramaic Dialect of Koy Sanjaq (Iraqi Kurdistan). Wiesbaden: Harrassowitz.
Nakano, Aki’o. 1973. Conversational Texts in Eastern Neo-Aramaic (Gzira Dialect). Tokyo: Institute for the Study of Languages and Cultures of Asia and Africa.
Nöldeke, Theodor. 1868. Grammatik der neusyrischen Sprache am Urmia-See und in Kurdistan. Leipzig: Weigel. (Rpt. Hildesheim: Olms, 1974.)
Nöldeke, Theodor. 1875. Mandäische Grammatik. Halle: Buchhandlung des Waisenhauses. (Rpt. Darmstadt: Wissenschaftliche Buchgesellschaft, 1964.)
Nöldeke, Theodor. 1881. Review of Eugen Prym and Albert Socin, Der neu-aramaeische Dialekt des Tur ‘Abdin. 1. Teil: Die Texte. 2. Teil: Uebersetzung (Syrische Sagen und Maerchen aus dem Volksmunde gesammelt und uebersetzt). ZDMG 35: 218-35.
Panoussi, Estiphan. 1990. On the Senaya Dialect. In Heinrichs 1990. Pp. 107-29.
Pennacchietti, Fabrizio. 1991. Gli allomorfi della flessione preteritale del dialetto neoaramaico orientale di Hertevin (Turchia) in prospettiva storica. In Semitic Studies in Honor of Wolf Leslau on the Occasion of His Eighty-Fifth Birthday, November 14th, 1991, ed. Alan S. Kaye. Wiesbaden: Harrassowitz. Pp. 1197-1202.
Pennacchietti, Fabrizio. 1994. II preterito neoaramaico con pronome oggetto. ZDMG 144: 259-83.
Polotsky, H. J. 1961. Studies in Modern Syriac. JSS 6: 1-32.
Polotsky, H. J. 1964. Semitics. In The World History of the Jewish People, volume One: At the Dawn of Civilization: A Background of Biblical History, ed. B. Netanyahu and E. A. Speiser. New Brunswick, N.J.: Rutgers Univ. Press. Pp. 99-111.
Polotsky, H. J. 1979. Verbs with Two Objects in Modern Syriac (Urmi). Israel Oriental Studies 9: 204-27.
Polotsky, H. J. 1991. Modern Syriac Conjugation. JSS 36: 263-77.
Rhétoré, J. 1912. Grammaire de la langue Soureth ou chaldéen vulgaire, selon le dialecte de la plaine de Mossoul et des pays adjacents. Mosul: Imprimerie des Pères Dominicains.
Ringe, Donald, Tandy Warnow, and Ann Taylor. 2002. Indo-European and Computational Cladistics. Transactions of the Philological Society 100: 59-129.
Sabar, Yona. 1978. From Tel-Kepe (‘A Pile of Stones’) in Iraqi Kurdistan to Providence, Rhode Island: the Story of a Chaldean Immigrant to the United States of America in 1927. JAOS 98: 410-15.
Sabar, Yona. 1995. The Christian Neo-Aramaic Dialects of Zakho and Dihok: Two Text Samples. JAOS 115:33-51.
Sabar, Yona. 2002. A Jewish Neo-Aramaic Dictionary: Dialects of Amidya, Dihok, Nerwa and Zakho, Northwestern Iraq. Wiesbaden: Harrassowitz.
Sachau, Eduard. 1895. Skizze des Fellichi-Dialekts von Mosul. Berlin: Verlag der Königlichen Akademie der Wissenschaften in Commission bei Georg Reimer.
Sims-Williams, Nicholas. 1989. Eastern Middle Iranian. In Compendium Linguarum Iranicarum, ed. Rüdiger Schmitt. Wiesbaden: Reichert. Pp. 165-72.
Sinha, Jasmin. 2000. Der neuostaramäische Dialekt von Bespsn (Provinz Mardin, Südosttürkei): eine grammatische Darstellung. Wiesbaden: Harrassowitz.
Socin, Albert. 1882. Die neu-aramäischen Dialekte von Urmia bis Mosul. Tübingen: Laupp.
Takashina, Yoshiyuki. 1990. Some Remarks on Modern Aramaic of Hertevin. Journal of Asian and African Studies 40: 85-132.
Talay, Shabo. 1999. Überblick über die neuaramäischen Dialekte der Assyrer am Khabur (Nordostsyrien). Mediterranean Language Review 11: 156-75.
Talay, Shabo. 2001. Grammatikalische Anmerkungen und Texte zum neuaramäischen Dialekt von Nerwa (Nordirak). Mediterranean Language Review 13: 1-37.
Talay, Shabo. 2008. Die neuaramäischen Dialekte der Assyrer am Khabur in Nordostsyrien. Wiesbaden: Harrassowitz.
Tsereteli, Konstantin. 1977. Zur Frage der Klassifikation der neuaramäischen Dialekte. ZDMG 127: 244-53.
Younansardaroud, Helen. 2001. Der neuostaramäische Dialekt von Särdä:r’id. Wiesbaden: Harrassowitz.
Younansardaroud, Helen. 2002. Der neuostaramäische Dialekt von Särdä:rid. In Arnold and Bobzin 2002. Pp. 841-51.