📄 Arabic Dialects of Central Asia: Internal Development Tendencies of Arabic Language Material

Summary

The paper looks at the two Arabic dialects still spoken in Uzbekistan: Bukhara Arabic (BAD) and Qashqadarya Arabic (QAD). Earlier scholarship (Tsereteli, Vinnikov, Akhvlediani, Jastrow) focused mostly on Bukhara, so the author sets out to document Qashqadarya as well. The material comes from his own fieldwork in the villages of Jeinau and Qamashi and in the city of Qarshi. He recorded informants of different ages and both sexes, including weaker speakers whose Arabic is heavily mixed with Uzbek and Tajik.

The dialects have been cut off from the Arab world for centuries. Their speakers live with three languages, Arabic plus Tajik and Uzbek. Within this setting, the paper asks which features arose from Arabic's own internal evolution and which came from contact. It surveys these areas:

• roots
• pronouns
• plurals
• derived verb stems
• verbal particles
• object marking
• periphrastic tenses

The author's conclusion is that most of what makes these dialects distinctive comes from the absence of Classical Arabic influence, although contact effects show up at several levels. The data matter for the historical morphology of Arabic and of Semitic languages in general.

Key points

• Sociolinguistic setting. BAD is spoken in the Ghijduvan and Wobkend areas, and QAD in villages north-west of Qarshi. Bukhara Arabs are more fluent in Tajik, while Qashqadarya Arabs lean more on Uzbek.
• Pronouns.
  • BAD demonstratives (had, hadi, hat, hal) are linked to Syrian/Aramaic forms. The demonstrative hal is flagged as relevant to how the Arabic definite article arose.
  • The QAD 1st person plural alternates between nana and a double plural, nanat.
• Plurals in QAD.
  • The default plural suffix is -at for both genders.
  • Double plurals also occur, such as farasinat "horses".
  • Some old broken plurals carry singular meaning. The author suggests patterns like CuCuC were originally singular.
• Verb stems and particles.
  • Derived stems II–X all exist except IX.
  • In QAD, long and short vowels are merging, which blurs stems II/III and V/VI.
  • The BAD imperfect particle mi is treated as an Arabic-internal development rather than a Persian loan, because Persian mi also occurs in the past tense.
• Verb morphology.
  • In BAD, transitive verbs always carry an object pronoun suffix, even when the object noun is present.
  • The verb kana has been worn down into a suffix -k.
  • New periphrastic tenses include:
    • a habitual built from a verbal noun plus "do"
    • a past tense of participle plus suffixes, attributed to Turkic influence
    • an immediate future with a t- prefix
    • a progressive built with the participle of nām

Interesting ideas

• I1. Telling borrowing from internal change by distribution: a feature borrowed from another language should match where it appears in the source language, and a mismatch suggests the feature developed internally. — For Mova (learn): when a witness parser keeps producing a deviant analysis, compare its distribution across constructions with the suspected source of bias to decide between a parser artefact and a genuine phenomenon.
• I2. Object doubling: the verb always carries an object affix, even when a full object noun is also present, so two surface items point to one argument. — For Mova (tree): UD dialect conversion should attach such doubled markers so that the affix and the noun resolve to a single object slot instead of producing two objects.
• I3. Form and meaning number can drift apart: in these dialects, forms that are morphologically plural can mean a single thing. — For Mova (tokens): store morphological Number separately from semantic count, as for English "news", "data" and "police", so that the downstream world model counts entities correctly.
• I4. Aspect built from auxiliaries: habitual and progressive meanings come from auxiliary constructions ("do" plus a verbal noun, "sleep" as a progressive marker) rather than from the main verb's own meaning. — For Mova (frames): recognise English periphrases such as "used to", "keep -ing" and "be going to" as aspect operators on a frame, not as separate events.
• I5. Recording speakers of different proficiency: the fieldwork deliberately included weak, code-mixing speakers alongside fluent ones. — For Mova (learn): build evaluation slices from non-native or code-mixed English, so that robustness is measured separately and not hidden by clean text.