📄 A research about the effect of Targums (Aramaics Old Testament) on religious literature of Hebrew-Persian

Summary

The paper looks at how Aramaic shaped the Persian religious writing of Iranian Jews. The authors describe two language shifts in this community. Under the Achaemenids, Hebrew gave way to Aramaic. After the Muslim conquest, Aramaic gave way to local Iranian varieties of Persian. Scripture followed both shifts: the Torah was rendered into Aramaic as the Targums, and later into Judeo-Persian, which is Persian written in Hebrew script.

The study identifies Aramaic vocabulary in Judeo-Persian Bible translations, commentaries (tafsirs) and poems. These words are checked against Targum glossaries and against Hebrew, Syriac and Akkadian dictionaries.

The main findings are:

• Targumic influence is limited to borrowed words.
• The borrowed words cluster in religious domains, especially law (fiqh) and theology, with a few general words.
• They generally have Hebrew counterparts, which suggests Hebrew still steered which Aramaic words entered Persian.

The work matters because it is presented as the first Iranian study of its kind. It also offers a model for tracing layered contact through an intermediate translation.

Key points

• The authors describe two shifts: Hebrew to Aramaic in the Achaemenid era, then Aramaic to Persian in the Islamic era. Scripture moved Hebrew → Targum (Aramaic) → Judeo-Persian.
• Aramaic elements appear only in the community's religious texts. The paper reports none in the secular poetry of Shahin Shirazi and Baba ibn Lotf, nor in personal letters and contracts.
• The authors explain this by arguing that the translators depended on the Targums rather than reading the Hebrew Torah on their own. They cite Nehemiah 13:24 as evidence that Hebrew knowledge had faded.
• Loterai (from lo-torah, "non-Torah") is treated separately as a hybrid language. Its pronouns, adjectives, many nouns and verb stems are Semitic. Its verbal endings, modal prefixes, suffix pronouns and syntax are Iranian. For this reason it is not counted as Persian literature.
• The sources used include:
  • Judeo-Persian texts: the Judeo-Persian Pentateuch (British Museum Ms. Or. 5446), the Vatican Judeo-Persian Pentateuch, and Gindin's Ezekiel tafsirs.
  • Dictionaries and glossaries: Cook's Glossary of Targum Onkelos, Jastrow's Targum dictionary, Klein's Hebrew etymological dictionary, and the Chicago Assyrian Dictionary.

Interesting ideas

• I1. Loanwords are traced through a pivot translation (Hebrew → Aramaic Targum → Persian) rather than through direct contact, so a word's path runs through the text translators actually used. — For Mova (back): when translating through the meta-world, Mova could record which intermediate representation contributed each output word, so that lexical choices can be explained by their path.
• I2. Borrowing is measured separately by register (religious vs. secular vs. documentary texts from the same community), and the borrowed layer appears in only one of them. — For Mova (tokens): lexicon entries could carry register or domain tags, so the tagger and lemmatizer treat domain-specific vocabulary as expected in some genres and suspicious in others.
• I3. In a hybrid language, the lexical layer (stems, nouns, pronouns) is analysed separately from the grammatical layer (endings, prefixes, syntax), since the two can come from different sources. — For Mova (tokens): morphology could analyse a foreign stem plus native affixes as two independent layers instead of forcing the whole word into one paradigm.
• I4. Secular texts from the same community act as a built-in contrast: their lack of Aramaic shows the effect belongs to the religious channel, not the community as a whole. — For Mova (learn): this is a natural negative control; a learned rule should be checked on matched texts where its trigger is absent, to confirm it is not a corpus-wide artifact.
• I5. Each borrowed word is checked for a counterpart in the related source language (Hebrew), to test whether the older language guided which loans were accepted. — For Mova (world): ontology entries could store cross-language cognate links, and the system could prefer candidate senses that have support from a related, already-grounded entry.