The Caspian Language of Sahmirzad.

Author:Borjian, Habib

    The township of Sahmirzad (locally Samerza) sprawls along the southern slopes of the Alborz range, 15 miles north of Semnan, at 35.8[degrees] north latitude, 53.3[degrees] east longitude, and 2,000 meters above sea level. Two parallel mountain ridges separate Sahmirzad, which is in the Semnan district, from Dodanga and Savadkuh districts of the Mazandaran (Mazandaran) province. Downslope from Sahmirzad lie Sangesar and Semnan, each with its distinct Iranian language, forming a Sprachbund with the nearby Sorxa'i, Lasgerdi, and Aftari, all of which are crowded into the district of Semnan. The permanent population of the township of Sahmirzad, recorded as 7,273 individuals in the 2006 and 8,882 in 2011 censuses, swells significantly in summers upon the return of the residents who work elsewhere. The villages located in the valleys on the north of Sahmirzad, including Deh Sufian, Aseran, Jasm, and Garm Casma, speak varieties of Sahmirzadi. See Figure 1. (1) The local residents estimate the total number of speakers of Sahmirzadi to be around 5,000 individuals.

    1.1. Documentation

    Documentation of Sahmirzadi was begun in the 1880s by Valentin Zukovskij (henceforth Zuk.), who elicited from Persian short texts and anecdotes and a list of words and verb paradigms (Zukovskij 1922: 5-8 and glossary). This was followed by compilation of basic morphology along with two short texts by Arthur Christensen (henceforth Chris.; 1935: 142-78), (2) brief notes by Georg Morgenstierne (henceforth Morgen.; 1960: 108-9), a comparative lexicon with other languages of Semnan by Manucehr Sotuda (henceforth Sotu.; 1963), an elicited text by Iran Kalbasi (henceforth Kalb.; 2009: 531-33), and a typological study by Mohammad Dabir-Moqaddam (henceforth Dabir.; 2014: vol. II, 1034-90). The unpublished materials consulted in this study include a list of words and sentences from the 1960s collected by Husang Purkarim (henceforth Pur.), kindly shared by a colleague of his who preferred to stay anonymous; audio recordings in 2003 of a 227-item basic vocabulary, shared generously by Alexander Kolbitsch; and the present author's unpublished documentation (since 2009), which is left unmarked in the citations that follow.

    In this study I have chosen to draw upon all these credible sources rather than eliminating past knowledge in favor of my own field notes. In view of the incessant weakening of the local idioms under the pressure of Persian, I believe that no reasonable documentation should be ignored with the pretext that the methodologies used in the past are incompatible with current practice. Multiplicity of sources surely adds to the richness of language description even if the job becomes more complicated.

    The materials at hand, being from multiple sources, show considerable variance in transcription; for instance, the gloss 'eye' is elicited by Zukovskij as casm, by Christensen as cas, cas, casm, by Sotuda as cas; and 'wind' has the outputs Zuk. vo, Sotu. va, Chris, bad, and my field-notes [va ]. (3) In Christensen's documentation, it is evident that the speech of his chief informant is strongly persianized, marking phonology (barf, barg, bid, blst, which are v-initial words in other sources), morphology (verb personal endings, [section]4.4), and syntax (adpositions, [section]3.4); these discrepancies could partly be due to infected idiolects of seasonal workers who winter in Mazandaran and elsewhere. Note that the phoneme /a/ ([section]2.2.2) is rendered as variously as Zuk. , (4) Chris., Morgen. , Sotu., Kalb. , Pur., Dabir. , while /a/ ([section]2.2.7) is rendered in the latter two sources as . The original symbols are retained in the citations below. I decided to stay loyal to transcription symbols from the original sources, although it may introduce an additional level of complexity.


    2.1. Consonants. The consonantal inventory of Sahmirzadi is/pbtdkgcjfvszszx [gamma]hmnrly/.

    2.1.1. The phonemic value of z [3] is to be verified; its presence in my notes is limited to kazdom [[phrase omitted]] 'scorpion' and hizda [[phrase omitted]] 'eighteen'. As in Mazandarani proper, the existence of z therefore appears to be limited to regressive assimilation of the phoneme /j/ [d3] before /d/.

    2.1.2. Dorsals. The articulations of /k/ and /g/ are not exclusively velar as is the case in Mazandarani; younger speakers tend to follow the Persian model of palatalizing before front vowels; thus, gal 'throat' is heard as both [[phrase omitted]] and [[phrase omitted]]. Likewise, the prevalent Persian way of switching between back velar/uvular fricative and stop ([[phrase omitted]] and [G] in most speakers) is also the norm in Sahmirzadi, contrary to Mazandarani, in which these sounds have collapsed to voiced velar fricative [[phrase omitted]] in all positions. This study employs both [gamma] and q, e.g., qossar [[phrase omitted]] 'pressure' and maryona [[phrase omitted]] 'egg'; the distributional relationship between these allophones can only be established with sufficient data.

    2.2. Vowels. This section proposes the Sahmirzadi vowel inventory as /a a [epsilon] e i u o/. These phonemes are arranged together with their phonetic realizations in Table 1. The choice to include the fronted allophones as basic for the phonemes /u/ and /o/ is made due to their higher occurrences ([section]2.2.8), although this choice leads to an asymmetrical phonemic inventory. The sections that follow present a detailed analysis of each vowel sound and its considerable variation among speakers. Variations are mostly due to different speakers, but they may also be the perception of different listeners to the same audiotapes.

    2.2.1. Vowel length. Although Middle Iranian length distinction has not reached Sahmirzadi systematically, /a u i e/ are perceivably articulated long when compared to other vowel sounds. Besides, as in Persian, slight lengthening precedes consonant clusters in coda, as in abr [[phrase omitted]] 'cloud', valg [[phrase omitted]] 'leaf. The lengthening in ras[epsilon]n [[phrase omitted]] 'rope' and asun [[phrase omitted]] 'last night' in my notes should be phonetic, due to pragmatic factors. In fact, vowel lengthening in initial syllables is common in many parts of Iran.

    2.2.2. /a/ is realized as [[alpha]], sometimes partially rounded to [[phrase omitted]], and is usually articulated half long: [[phrase omitted]] 'fire', [[phrase omitted]] 'river', [m[alpha]'r] 'mother', [[phrase omitted]] 'snake', [va'] 'wind', [[phrase omitted]] 'chick', [[phrase omitted]] 'sister'. It was heard as slightly raised in [[phrase omitted]] 'tree', [[phrase omitted]] 'horn', [[phrase omitted]] 'moon', [[phrase omitted]] 'navel', among other words. In [[phrase omitted]] 'hunger', /a/ is noticeably short, compared with [[phrase omitted]] 'hungry'. It is therefore hard to judge whether the position in stressed vs. unstressed syllables imposes a condition in this phonetic variation. More data is needed to establish free vs. conditional variation for this phoneme.

    2.2.3. /i/ is half long or long, especially in final position: [di'] 'smoke', [ta'li:] 'rock', [[phrase omitted]] 'mouse', [[phrase omitted]] 'hill', [[phrase omitted]] 'ant', [tim] ~ [ti'm] 'seed', [[phrase omitted]] 'root'. As is evident in these examples, historical *i and *u have merged into /i/; for example, mis

    2.2.4. /e/ is half long and is principally derived from Middle West Iranian *e, as in jer [[phrase omitted]] 'below', reg [re:g] 'pebble', tej 'sharp', mes 'ewe', xes 'kin', aspe [[phrase omitted]] 'white' (

    There are a few cases in which /e/ has other origins: per [per] ~ [per] 'father' (probably, as other varieties of Caspian imply, with the protoform *[phrase omitted], from Mid. West Ir. *pidar), ses [[phrase omitted]] 'six', and the "inverse-ezafa" marker ([section]3.2), which is used to form possessive pronouns (me, te, e, etc.; Table 2).

    There are instances where [e] overlaps with [I], as in aspez [as'pez] ~ [[phrase omitted]] ~ [[phrase omitted]] iouse' (

    2.2.5. /[epsilon]/. The space between the phonemes /e/ and /a/ is filled with a range of open-mid vowels that includes [[phrase omitted]]. These sounds at times overlap with /a/, less likely with /e/, and therefore could reasonably be placed in the domain of either of these principal neighboring phonemes. On the other hand, the high frequency of the range [[phrase omitted]] affords a distinct phoneme, designated in this study as /[epsilon]/, which by and large agrees with the symbol a used by Dabir-Moqaddam. I could not identify any minimal pairs in the data between either /[epsilon]/and /e/ or /[epsilon]/ and /a/. However, in support of a phonemic status for /[epsilon]/ it should be added that, notwithstanding its frequent allophonic intersection with /a/, the Sahmirzadi speakers commonly perceive it as Persian kasra far more than fatha. The following examples from the data are meant to show the allophonic range of the sound: y[epsilon]k [y[epsilon]k] ~ [[phrase omitted]] 'one', S[epsilon]n [[phrase omitted]] ~ [[phrase omitted]] 'sand', l[epsilon]ng [[phrase omitted]] ~ [[phrase omitted]] ~ [[phrase omitted]] 'leg', v[epsilon]ni [[phrase omitted]] ~ [[phrase omitted]] 'nose', ask[epsilon]m [[phrase omitted]] ~ [[phrase omitted]] 'belly', at[epsilon]s [[phrase omitted]] ~ [[phrase omitted]] 'fire', ru[gamma][epsilon]n [[phrase omitted]] ~ [[phrase omitted]] 'ghee', b[epsilon]rar [[phrase omitted]] ~ [[phrase omitted]] 'brother', s[epsilon]tara [[phrase omitted]] ~ [[phrase omitted]] ~ [[phrase omitted]] 'star', ass[epsilon][gamma]on [[phrase omitted]] ~ [[phrase omitted]] 'bone', [epsilon]sm [[phrase omitted]] ~ [[phrase omitted]] 'name'.

    2.2.6. In final position a front-mid vowel is heard in the following words of obscure etymon: zaye 'grandchild' ( zadak), xal[epsilon] [[phrase omitted]] 'many' (cf. Tajik xale), lav[epsilon] [[phrase omitted]] 'pot', saz[epsilon] [s[alpha]'z[epsilon]] 'broom'. Nevertheless, the most...

To continue reading