I’m still reading Nicholas Ostler’s The Last Lingua Franca: English Until the Return of Babel (previously), and I thought I’d pass on his discussion of various names for Persian-speakers (pp. 134-35):
In Central Asia, from the eighth century, Persian-speaking “Tajiks” began to supplant Sogdians as the dominant merchant community, bringing their language too right across the continent. One or two references note merchants from the west turning up in the Far East in the eighth century, but the steps of their general commercial progress are not documented. The merchants would have been reinforced in the eighth century by a large number of émigré Persians looking for new homes and livelihoods beyond the reach of the Muslims. Still, the variety of names that they were called tells a story of its own.
First, the word Tajik itself, originally a Sogdian term for an incoming Muslim, began to add overtones. For example, in the Sanskrit classic Somadeva’s Ocean of the Streams of Story (written 1063–81) a party of innocent Indian travelers to “the north” is waylaid and then traded as chattels by Tājika merchants. The word spread by association to their characteristic stock-in-trade, coming also to designate an excellent breed of horse. With mercantile literacy came a patina of culture: Tājaka was also applied to astronomical treatises translated from Arabic or Persian. So much for the Indian reputation; for the Turks, Persian literacy in itself was endlessly impressive. Their first recorded use of the word is in Mahmud al-Kashgari‘s Compendium of the Language of the Turks in 1072.
Tejikler bitigde bitimiş munı | Bitigde yok erse kim okkay anı?
The Tajiks in a book set down this. In a book if it were not, who would have mentioned it?
The Chinese came to refer to the whole Arab empire—which naturally impinged on them from its eastern end—as ta dʑeĭək, writing it, perhaps in the light of experience, as 大食 ‘great food’, though a thousand years later the word is now pronounced more like dàshí. [See this LH discussion of the latter term from 2006.]
The Turks were responsible for two other words for Tajiks used widely across Asia: Sart and Tat. Each grew out of reference to a contingent fact about Persian speakers into a general quasi-racial term.
The ultimate origin of Sart is doubtful: it seems to be either a dialectal pronunciation of Soǧd or a shortening of the Sanskrit sārthavāha or Sogdian sārtpāw ‘caravaneer’. Either way, it could be a nice irony, the Persians being identified with the Sogdian caravan merchants that they were replacing, and quite comparable with the origin of Tajik itself—an Arab from the Tayy tribe, which was best known to the Persians. Sart‘s first recorded use is in the Turkic text the Qutadǧu Bilig (1070), where it already refers not to merchants but to settled populations (specifically, locally in Kashgar), whom the Turks of that age might have expected to be Persian speakers, in contrast with the free-ranging Turks. By the fifteenth century, when Mīr ‘Ali Shēr Nawā’ was robustly comparing the value of Persian and Turkish as literary languages, he happily used Sart as a strict synonym for ‘Persian’, so he could speak explicitly of Sart tili ‘the Sart language’.
As for Tat, it always seems to have been a convenient term for Turks to refer to many significant peoples that were not Turks (much in the same way as the Wall root has been used round Europe for non-Germans, Welsh, Walloon, and Wallachian). Tat was known to have been used at different times to designate Crimean Goths, Greeks, and sedentary peoples generally, but its primary reference came to be the Persians within the Turkish domains. Hence unlike Tajik and Sart it had nothing to do with any reputation of Persians as merchants. (Tat is nowadays specialized to refer to special groups with Iranian languages in the west of the Caspian Sea.)
A nice tidbit from a footnote on page 129: to illustrate “coiner’s conservatism” (“when the idea of coined money is introduced for the first time, it often continues to use quite slavishly the model of its foreign source”), Ostler points out that a mancus struck by King Offa of Mercia in the late eighth century “combines the central legend OFFA REX with the Islamic šahāda ‘There is no God but Allah…’, in Arabic, round the rim, and an Arabic inscription on the reverse. It is all copied off a 774 dinar of Manṣūr in the contemporary Abbasid Caliphate.” Here‘s an image of the mancus in question.
Addendum (Oct. 2015). I regret to report that, according to Philip Shishkin, sart is now apparently “a derogatory term the Kyrgyz use for Uzbeks, the local n-word.”