Linguistic Chimeras. The Case of Kazakhstan

Chapter 1

What a chimera is

In the sixth book of the Iliad there is a monster that is a lion in front, a goat in the middle and a serpent behind. The Greeks called it χίμαιρα, “young she-goat”. The myth gave the word a meaning that outlived antiquity: a creature assembled from parts that never live together in nature.

In 1907 the German botanist Hans Winkler brought the word into biology. He called a chimera a plant obtained by grafting, whose tissues belong to two different organisms and never merge into one. The plant lives, grows, even flowers, yet each of its cells remains a cell of its own parent. That is the precise sense of a biological chimera, and it fits language remarkably well.

Take the word Төралқа (Toralqa, the presidium) in the Russian text of the constitutional law on the People’s Council. It lives inside the sentence but remains a Kazakh word: «заседания Төралқа созываются», «решение Төралқа оформляется протоколом» [4], Ф-12. The Russian sentence requires the genitive case here, and the word refuses it. The cell has not fused with the organism.

The report behind this longread calls a chimera any construction that meets three conditions at once:

1. it joins elements of two language systems — graphic, grammatical or word-forming;
2. the receiving language cannot absorb the foreign element by its own rules, because absorption is forbidden or not provided for by an instruction;
3. as a result the receiving system stops working properly: a word no longer declines, a sentence loses its links, a record cannot be found by search, a term becomes unclear to native speakers.

The first two conditions without the third give an ordinary quotation: the name of the society “Қазақ тілі” in quotation marks bothers nobody. The third without the second gives an ordinary mistake, which a proofreader will fix. A chimera appears only where all three meet.

Absorb or freeze

In 1950 the Norwegian-American linguist Einar Haugen proposed seeing two movements in borrowing: importing a foreign model and substituting one’s own means for it. The Russian word аул (village) shows substitution in its purest form. The Kazakh ауыл entered Russian, the vowel cluster unfamiliar to a Russian ear was shortened, and the word received every Russian case: аула, аулу, в ауле. Kumis, baiga and saxaul were absorbed the same way. Substitution is what keeps borrowing healthy: a language does not merely take something foreign, it digests it.

Test yourself. Below are five examples, from fully absorbed to failed. Which of them is a chimera?

The last two cards show why the subject cuts both ways. In Russian text an instruction prevents a Kazakh form from being absorbed. In Kazakh text an instruction introduces a form that speakers never asked for and the language cannot take in, because the place is already occupied. The mechanism is the same: the norm comes not from living usage but from an order, and in both cases the reader loses.

Chapter 2

A letter as a flag

A Russian reader who meets теңге in a Russian text will read it as [tenge]. Russian has no [ŋ], the sound the letter ң stands for, and the letter will not teach it to him. A Kazakh reader will read the word correctly without the letter, because he knows the word. So what is the letter for?

The answer lies in the idea of the grapheme. The word comes from Greek γράφω, “I write”, on the model of “phoneme”; Jan Baudouin de Courtenay was the first to treat a sound and its written sign as units of different levels. A grapheme is the smallest unit of writing that distinguishes meaning. The Kazakh Cyrillic alphabet approved in 1940 has 42 letters: all 33 Russian ones and 9 of its own. Each of the nine marks a sound that Russian lacks or that does not distinguish meaning in Russian.

Rendering foreign names with one’s own alphabet is called practical transcription. The standard Russian reference is the book by R. S. Gilyarevsky and B. A. Starostin, Иностранные имена и названия в русском тексте (“Foreign names in Russian text”). Its principle is simple: in running Russian text names are written in Russian letters; Latin spelling is kept only for trademarks, and a name that mixes two alphabets should go in quotation marks. Hence тенге, Конаев, Жетысу.

This practice faces a serious objection, and it deserves a hearing. Soviet rendering of Kazakh names into Russian was careless. In 2023 Z. K. Temirgazina, O. A. Andryushchenko and M. Valova examined a 1988 dictionary of Kazakh names and found that only 18.12% of names were rendered one-to-one [41], Ф-45. This is where the case for Kazakh letters in Russian text comes from: for a long time Russian rendering was not neutral. What follows from it, however, is the need for an accurate and uniform transcription, not the abandonment of transcription altogether.

The chronicle of one word

The most visible graphic decision of 2026 concerned the name of the national currency. Here is the year, from the referendum to yesterday’s headlines.

Three conclusions follow. First, the obligation differs by addressee: for state bodies it rests on the Justice Ministry’s explanation, for banks on the regulator’s demand, for textbooks on the update cycle, and for the media on nobody. Second, grammar has not suffered: Russian тенге never declined, just like евро or кенгуру. The sentence does not break; typing, search and consistency do. Third, practice departs from the instruction at the very top: the Russian text of the head of state’s speech, published almost four months after the Committee’s explanation, follows Russian spelling.

The numeric currency code 398 and the letter code KZT did not change. The decision affected only the Cyrillic name.

Two lines of this chronicle are worth rereading. For almost thirty years the 1997 decree on National Currency Day carried «тенге» in its Russian title; on 1 July 2026 that title was rewritten — not the text of the decree, only the name of the money in it [75], Ф-74. The state is consistent: it applies the new rule to old acts as well. Four months later the official Russian text of the head of state’s speech, discussing public spending, writes «6,5 трлн тенге» [6], Ф-24. Neither line is a mistake. The first shows how an instruction works; the second shows how the Russian language works where the instruction does not reach. The chimera lives between the two.

A control group

To see whether the letter is dictated by necessity, we need a control group, and there is one. The Mongolian alphabet has the same letters ү and ө as Kazakh, yet Russian texts write the Mongolian president’s name as Ухнаагийн Хурэлсух, without Mongolian letters. Transcription is possible; Kazakh letters in Russian text are a choice, not a necessity.

Nor is this choice made for every name. Within one and the same document some names are rendered in Russian and others are not:

A graphic chimera has two costs. The first is technical: in the standard Kazakh keyboard layout for Windows the nine special letters sit on the number row, and an operator typing Russian text either switches layouts twice per word or substitutes a similar-looking Latin character. Chapter 9 shows what this does to databases. The second cost is normative: if some names are rendered and others are not, neither an editor nor a translator nor a machine can infer a rule, and each case has to be memorised.

Chapter 3

A phrase that will not decline

The Russian text of the constitutional law on the People’s Council contains a phrase that occurs 18 times: 11 times as the full group «Қазақстан Халық Кеңесінің Төралқасы» (Presidium of the People’s Council of Kazakhstan) and 7 times as the short form «Төралқа» [4], Ф-11. Almost every time it stands in a case it cannot express.

The count can be repeated: the text is taken from an official source, without the table of contents, and each word is searched as a separate word form method — Appendix Г of the report. The first edition of the report gave 19: the table of contents had slipped into the count by mistake.

To see what is going on, we need a word rarely heard outside Turkic studies. Arab grammarians used iḍāfa, “annexation”, for a pair of nouns in which the second defines the first: “the teacher’s house”. Persian grammar named the same thing ezafe. Turkologists carried the term over to Turkic noun phrases with possessive markers. Kazakh grammar treats it as one of five kinds of syntactic link, матасу, literally “interlocking”.

The key difference between the languages lies in their structure. Russian is inflectional: words are linked by their endings. A word that takes no endings drops out of the link. Try it yourself.

Three kinds of breakage

A frozen apposition. Decree No. 887 of 3 May 2022 created three regions, called in Russian «область Абай», «область Жетісу» and «область Ұлытау» [9], Ф-33. Precision matters here. Russian itself allows “generic word plus name”, as in провинция Хэбэй or штат Айова. The form область Абай also has its own justification: the region contains an Abai District, and the adjective Абайская would be confused with it. What makes the construction a chimera is something else: a Kazakh letter that the Russian alphabet lacks, and a refusal to decline where Russian norm requires it — «в городе Қонаев» instead of в Конаеве.

A foreign plural. The street known in Kazakh as «Керей, Жәнібек хандар көшесі» is the street of the khans Kerey and Zhanibek, founders of the Kazakh Khanate. Хандар is “khan” with the Kazakh plural ending. Russian texts moved the generic word “street” to the front but left khandar as it was, and the 2026 press uses three spellings: «Керей, Жанибек хандар», «Керей Жанибек хандар», «Керей и Жанибек хандар» [33], [34], Ф-37. To a Russian reader “Khandar” looks just like a surname. We do not know the official Russian form in the naming commission’s decision М-08.

A two-suffix izafet in a law. The group «Қазақстан Халық Кеңесінің Төралқасы» carries two Kazakh endings: ‑нің for the genitive and ‑сы for possession. The drafters themselves felt the awkwardness: the short form Төралқа drops ‑сы, thereby already adapting the Kazakh word to the Russian sentence, but it stops half-way and gives it no Russian endings. Russian equivalents exist in the official texts: Article 70 of the Constitution gives “People’s Council of Kazakhstan”, Article 6 of the law gives “(Presidium)” [1], [4]. The translation exists, but it serves as a gloss, not as the name.

How the neighbours do it

A body’s national name can be kept without breaking the Russian sentence. Both the neighbours’ practice and Kazakhstan’s own past practice show this.

For three decades the Russian text of Kazakhstan’s laws kept the national names of bodies — аким, маслихат, Мажилис — writing them in Russian letters and declining them. The Kurultai in the new Constitution continues that tradition. The Khalyq Kenesi breaks it.

Why does this matter more than it seems? Under Article 9 of the Constitution the Russian text of a law in Kazakhstan is official Ф-04. Courts, lawyers, translators and machine translation systems use it. A sentence in which the subject cannot be told from the object by its form forces every reader to translate it mentally into Kazakh to understand its syntax. That defeats the purpose of a parallel official text: to be clear to someone who reads only that text.

Chapter 4

One rule, four languages

If chimeras were produced by a legal requirement, they would appear in every language in which the state publishes its texts. If they come from a choice concerning only Russian text, other languages will not have them. That can be checked.

Three languages of different structure serve for comparison. English is analytic: words are linked by order and prepositions. Spanish is moderately inflectional, with number agreement and strict written stress rules. Turkish is agglutinative and related to Kazakh, with the same izafet.

Four conclusions follow from the table. Kazakh letters pass into no foreign text. English has chimeras too — a preserved khandar, five different spellings of Toralqasy in the translation of one law, Cyrillic і inside a Latin word Ф-13, Ф-38, Ф-40. But the harm a chimera does is proportional to how inflected a language is: “a decision of Toralqa” is grammatically flawless, while «решение Төралқа» reads as a malfunction. And the most closely related language translates most fully: Turkish is built like Kazakh, yet it forms its own phrase — Halk Konseyi’nin, Kurultay’ın.

The three languages show three temperaments of absorption. English takes the name as it is but declines the Kazakh letters: Kurultai, the tenge, the city of Konaev. Spanish subjects the name to its own rules completely: it adds a stress mark by its own spelling rules — Kurultái — and gives the money a Spanish plural — millones de tenges [49]–[51], Ф-62. Turkish, Kazakh’s closest relative, translates the name outright — Kazakistan Halk Konseyi — and attaches its own genitive ending with an apostrophe [52], [53], Ф-63. None of the three breaks its grammar for the sake of a foreign script, and none is asked to.

Who the rule is aimed at

The same body, on the same day, names the same thing differently. On 1 July 2026 the National Bank announced in English: “Tenge Now Has a Constitutional Status as the National Currency of Kazakhstan” [80], Ф-78. In Russian the same bank moves its acts and press releases to «теңге». The State Revenue Committee writes “the city of Konaev” in English and «город Қонаев» in Russian [31], Ф-35.

The means to carry the Kazakh form into English exist. In the draft Latin alphabet put out for public discussion in spring 2021, ң is written Ñ and қ is written Q [81], Ф-79. One may object that the alphabet has not been approved and it is too early to use it; the Language Policy Committee says exactly that [47], Ф-60. Yet even without new characters English text could write Qonaev or Qazaqstan with ordinary Latin letters — and does not. Kazakh Cyrillic, by contrast, is approved, and its letters are carried into Russian text.

In English the state once proposed a change. In November 2016 the then Senate chairman Kassym-Jomart Tokayev wrote that in the English version QAZAQSTAN reflects the essence of the state more accurately than KAZAKHSTAN. Mazhilis deputy Olga Shishigina objected that the English spelling is already the country’s brand [82], Ф-80. The official English text still writes Kazakhstan, while Qazaqstan spreads on its own — in brand names and on young people’s social media. The state proposed; the language chooses. In English the mechanism works without an order.

Why Russian in particular

There is no official explanation for this selectivity, and attributing motives to it would be unfair. Two explanations are possible, and both deserve naming without choosing between them.

The legal one. Under Article 9 of the Constitution the Russian text is official, while English translations of laws are published as unofficial. It is natural to fix a state symbol where the text has legal force. But the argument cuts both ways: precisely because the Russian text is official, its own norm — and Article 21 of the Law on Languages requires spelling rules to be observed [7], Ф-31 — deserves protection.

The sociolinguistic one. In Fishman’s terms, a decision about the corpus is taken for the sake of status, and status is marked where a boundary runs. English text faces outward, and being understood matters more to the state. Russian text addresses the country’s own citizens, and there the boundary is marked with a letter.

Whichever explanation is right, the fact remains: a norm described as a requirement of law is applied to one of two languages, and to the very one whose grammar suffers most from it.

Chapter 5

Exonyms both ways

In 1944 the Kazakh regional newspaper «Көкшетау правдасы» began to appear in Kokchetav. From 1956 to 1963 the Russian paper of the same region was called «Кокчетавская правда», and later became «Степной маяк» [78], [79], Ф-77. One city, two forms of its name, each following the rules of its own language.

The name of a place in a language foreign to it is called an exonym; the name in its inhabitants’ language is an endonym, from Greek ἔνδον, “within”, and ὄνυμα, “name”. Kazakh uses its own exonyms freely: Мәскеу for Moscow, Орынбор for Orenburg, Омбы for Omsk, Сарытау for Saratov, Түмен for Tyumen, Қорған for Kurgan, and Ресей for Russia itself. Nobody demands that Kazakh text write “Moskva” or “Orenburg”, and rightly so. Russian did the same in the other direction: Alma-Ata, Kokchetav, Aktyubinsk are a foreign name absorbed by the language’s own rules.

The most curious part of the story is that the form of the name shifted with no decree at all. Gulbarshyn Salyq, chair of the Akmola Region Association of Local Historians, examined the regional archive. In the “place of issue” field, documents until December 1935 read «город Кокчетав», from December 1935 to July 1944 «город Кокшетау», and then Kokchetav again. She found no document on a renaming [78], Ф-76. In those years the form “Kokshetau” appeared in police circulars, on district executive committee letterheads and in a report of the Central Statistical Administration of the USSR State Planning Committee.

The timeline shows the main point. Model 1 is the Russian exonym: Alma-Ata, Kokchetav, Aktyubinsk, fully declined, with adjectives. Model 2 was a state decision just like model 3: the Russian form was changed from above. But it was changed with the means of Russian — the 1997 decree is even titled “On renaming… and changing their transcription” [74], Ф-72 — and Russian speech accepted the forms: Almaty, Kokshetau and Aktobe are ordinary words in Kazakhstan’s Russian texts today, declined and turned into adjectives. Model 3 changes the principle itself: Russian text is no longer given a transcription; a foreign script is carried into it.

Other voices have proposed a different path. Seitkazy Matayev, chairman of the Union of Journalists of Kazakhstan, proposed returning the name Alma-Ata to the city in the Russian version only, keeping Almaty in Kazakh [71], Ф-75. At the time Forbes.kz learned from the National Archive that the Supreme Council never discussed renaming Alma-Ata separately: the new name went straight into the text of the 1993 Constitution [71], Ф-70. The poet Olzhas Suleimenov has also argued more than once for bringing back the name Alma-Ata. In a 2021 interview with Вечерний Алматы he said the name suits the city better “in both sound and meaning”. He recalled that in twentieth-century Central Asia only three capitals had open, sonorous names: Samarkand, Ashgabat and Alma-Ata [84], Ф-81. The position has opponents: the publicist Dos Koshim saw a political rather than a linguistic meaning in the appeal [85]. Which form is better for Almaty is for society to decide; the example only shows that the two-name model has long been discussed.

The story has a legal tail. Article 19 of the Law on Languages protects “traditional, historically established” Kazakh names [7], Ф-30. But Alma-Ata and Kokchetav are also historically established names of the same places, only in Russian rendering. The law does not say how to tell a traditional Kazakh name from its traditional Russian rendering М-13.

Chapter 6

A name you cannot cancel

In the summer of 2026 a woman in Kazakhstan tried to register her daughter’s birth online with a surname ending in ‑eva and a patronymic ending in ‑қызы (“daughter of”). The system rejected the combination. She asked the Ministry of Justice whether the law forbids it.

On 12 August the ministry replied: there is no legal ban on mixed spelling, it is merely advisable to respect Kazakh grammar, and the information system’s refusal is a technical limitation, not a legal prohibition [14], Ф-42. A personal name is the one chimera that cannot simply be abolished: it belongs to a person. So the question is not which form is right, but that the state has no model that would let every lawful form be read and stored unambiguously.

The word “anthroponymy” comes from Greek ἄνθρωπος, “human”, and ὄνυμα, “name”. Until the twentieth century Kazakhs had no surnames in the European sense: a person was called by name and, for precision, by father, grandfather or clan. The passport campaign of the 1930s required a permanent surname, formed on the Russian model: Kunanbayev, Baitursynov, Satpayev. A presidential decree of 2 April 1996, No. 2923, allowed people to drop ‑ov, ‑ev, ‑ovich and to use Kazakh forms [13], Ф-41. No single model emerged: as late as 2011 Sherubai Құрманбайұлы, then chairman of the Languages Committee, counted 17 to 18 variants of how surnames were written [16], Ф-44.

Son of Kydyrbek, son of Algatbek

Now take a real person. The akim (head) of the capital’s Esil district appears on official websites in four ways [11], [12], [38], Ф-39:

To a Kazakh reader, ‑ұлы in the surname and in the patronymic are two different steps of a genealogy; to a Russian reader it is a tautology; to a database it is four different strings. The same imbalance produces paradoxes. A wife who takes her husband’s surname Muratuly becomes, on paper, “son of Murat”; if she changes the ending to ‑қызы, she becomes “daughter of her father-in-law”, to whom by custom she is a daughter-in-law. The suffix ‑келіні (“daughter-in-law of”), which would resolve the contradiction, is not allowed by the ministry, together with ‑тегі, ‑ұрпағы and others [15], Ф-43. How many such records exist in the registers we do not know М-10.

Honesty requires a caveat: the anthroponymic chimera does the least harm to Russian grammar. Russian has long had a class of indeclinable surnames — Georgian ones in ‑dze, Ukrainian ones in ‑enko — and surnames in ‑uly join it easily. What suffers is the unambiguity of records, database search and the meaning of Kazakh suffixes. The remedy lies in the law on names and in how registers are built, not in the rules of Russian text.

Three foreign models show what one can choose from. Iceland lives by patronymics instead of surnames — Jónsson, “son of Jón”, Jónsdóttir, “daughter of Jón” — and the model works because it is applied consistently: names do not change on marriage. Turkey made everyone take a surname by law in 1934: uniformity came quickly, at the price of a choice made for the person. Mongolia keeps the clan name in a separate field of the document next to the patronymic, without writing it into the surname as a suffix. In all three cases there is one model for everyone. In Kazakhstan there are seventeen or eighteen, and each is lawful.

Hence what can be done without touching anyone’s surname. The ministry itself admits that if an electronic system rejects a lawful variant, that is a technical limitation, and parents can register through the portal, a public service centre or the civil registry office [14]. So the rule should become an exhaustive list of permitted models, and systems should accept every one of them. People should be found by their individual identification number, not by spelling, and surname, first name and patronymic should be stored in separate fields. Then “son of Kydyrbek” and “Kydyrbek-uly” will stop being two different people for the machine while remaining one family history for humans.

Chapter 7

The pendulum

Until 1998 a museum in Kazakh was музей. In 1998 it became мұражай. In 2014 it became музей again [40], Ф-52. Sherubai Құрманбайұлы warned that if terms are changed every five to ten years, neither specialists nor translators can settle on any of them.

The pendulum has not come to rest even in the commission’s own decisions. On 23 October 2024 the Republican Terminology Commission, at the request of the Atyrau regional administration, approved petrochemical terms including «газфазалық термиялық үдеріс», “gas-phase thermal process” [42], Ф-53. The word үдеріс, formally returned to «процесс», lives on in approved terms.

The Terminology Commission is advisory; a norm becomes binding when it enters a document that people must follow. The building code SP RK 3.02-121-2012 on catering establishments contains «себезгі бөлмелері» (shower rooms) and «бұрқақ су», and SP RK 3.02-120-2012 contains «субұрқақ», although the commission approved «фонтан» [20], Ф-54. A building code is updated once a decade, a textbook once in five years, and a word lives on in them whatever was decided about it.

Eight words

On 10 November 2026, at the 10th Kurultai of the “Қазақ тілі” society, the President named 27 words that have come into wide use and eight coined words whose necessity, he said, seems doubtful [6], Ф-26. Both lists appear word for word in the official text of the speech. Turn the cards: not all eight are flawed in the same way.

Twenty-seven that took root

The grouping of these words by field was done for the report and can be disputed in individual cases; their etymologies still need checking against an academic dictionary М-12. But the pattern is plain: among the 27 there is not a single household object, and all eight rejected words are exactly that.

Four patterns follow. Abstract versus everyday. A speaker first hears legal and diplomatic concepts in official speech, where the state sets the form; a household object he has known since childhood by the name heard at home. Timing. Ғаламтор appeared just as the internet was entering the country’s life; күйсандық and жеделсаты came to objects that had other names for decades. Precision versus metaphor. The President counted «жасанды сана» (“artificial consciousness”) among the words that took root, but the law on artificial intelligence is titled «Жасанды интеллект туралы», because сана means human consciousness and the law regulates systems [43], Ф-55; both forms live on, each in its own field Р-08. Origin does not matter. At least eight of the 27 words have Arabic or Persian roots — ғарыш, жаһан, сынып, пайыз and others. Success depends on how well the root is absorbed, not on where it came from.

The dossier also yields a simple practical conclusion. The President’s list groups the words by one feature: speakers perceive them as artificial. Their linguistic defects differ. Себезгі and желкөз are well formed and could remain in technical language as synonyms; жеделсаты and үнқағаз rest on false images and should be removed from the norm outright. The useful question to ask of a new word is not “is it ours or foreign” but “is it clear, is it accurate, is the place free”.

Chapter 8

Statistics at odds with themselves

Public debate often repeats that the share of international terms approved by the Terminology Commission fell 3.3 times after 1991, from 41% to 12.3%. A check finds two incompatible figures for one and the same period.

The claim of a “3.3-fold contraction” rests on source B alone. A possible reason for the gap is different classification: a phrase like «авиациялық авария» (aviation accident) can be counted either as mixed or as keeping its international form. Arithmetic adds three more observations.

The overall total does add up: 2,748 terms of the first period plus 15,963 of the second give 18,711 — exactly the number source B gives as approved by 2017. Note too that the 2,748 terms were approved not in 1971–1991, as is often written, but in 1971–1981; another 79 in 1990, and none at all in 1991 [19], Ф-48 Р-02 closed. As of November 2026 the termincom.kz database holds about 391,000 terms.

The honest conclusion of this chapter is modest. Three incompatible counting methods in official sources give no right either to claim a “catastrophic displacement” or to deny it. Until the method is published, figures on shares cannot be trusted — neither one set nor the other. Hence one of the report’s recommendations: a single method and a recount of the data for 1971–2026.

Chapter 9

The digital cost

When a text is read first of all by a machine, a chimera stops being a matter of taste and becomes lost data. See for yourself: below are six records of the name of one and the same avenue in Astana, as they appear in documents and listings.

None of the three standard methods finds the record with Latin i letters, whatever the query. Why does even “diacritics-insensitive search” not help? The answer lies in how Unicode works.

Unicode is the international standard that assigns a number to every character of the world’s scripts. The name was proposed in 1988 by Xerox engineer Joe Becker: “uni” for unique, universal, uniform. The standard distinguishes composite letters — a base letter plus a diacritic — from atomic ones. Composite letters have a canonical decomposition, NFD, and nearly all search engines that promise diacritics-insensitive search decompose the letter and drop the mark. Try any letter: the check runs right in your browser.

For the standard, Қ is not “K with a tail” but a separate letter with no link to K. Kazakh ә, ң, і and Russian а, н, и are different letters, and there is nothing to fold. That is why теңге and тенге are two different words to an ordinary search.

Characters from different alphabets that look identical are called homoglyphs, from Greek ὁμός, “same”, and γλυφή, “carving”. The Unicode Consortium keeps a list of them in its security standard, because fraudsters use them. The substitution runs both ways. On the English page of the egov.kz portal an address is written “Tauelsіzdіk Ave.”: inside the Latin word there are two Cyrillic і letters [38], Ф-40. Nobody has measured how common such records are in state registers М-07.

Where homoglyphs come from is easy to see. In the standard Kazakh keyboard layout the nine special letters sit on the number row. An operator who types Russian all day must switch layouts twice to write one word with і — or press the Latin i, which looks the same on screen. The result is a record a person reads without effort and a machine will never find. The error comes not from carelessness but from the design of the tool: a rule that demands Kazakh letters in Russian text multiplies the occasions for such a substitution.

The key is not all-powerful. For the records of the Esil district akim it reduces «Кыдырбек-улы Дарменияр Алгатбекулы» and «Қыдырбек-ұлы Дәрменияр Алғатбекұлы» to one string, but «Дарменияр Кыдырбекулы» stays different: another word order, no patronymic. That is no longer a task for search but for data structure: surname, first name and patronymic belong in separate fields, and a person should be linked to a record by an individual identification number.

The same problem affects machines that learn language. In December 2024 the Institute of Smart Systems and Artificial Intelligence at Nazarbayev University released KazLLM, an open language model trained on about 150 billion tokens in four languages [44], [45], Ф-56. Its Kazakh corpus was prepared by more than 140 scientists and staff from 26 institutes and universities [66], Ф-66. To the model every spelling variant is a separate sequence of characters. If “fountain” appears in the corpus as four different words and one name in Kazakh, Latin and mixed letters, the model links them less well. How many such duplicates the national corpus contains has not been measured publicly М-02.

Chapter 10

Lessons from abroad

The history of language building knows both the successes and the failures of purism. They share one rule: a new word wins when it fills an empty place and loses when it pushes out an absorbed one.

Go through the seven stories and note three things that recur everywhere. The first is timing: tölva and ordinateur arrived before the English word, mot-dièse and тихогром after it. The second is the accuracy of the image: “number prophetess” describes a computer better than “chest of kuis” describes a piano. The third is the limit of power: a state controls the form in its own texts but cannot order another language what to call a country or a city.

Chapter 11

Who should do what

Kazakh script in Russian text and new Kazakh words both have serious defenders. Before turning to measures, their arguments deserve a hearing in their strongest form.

The report’s recommendations rest on one principle: separate three things that are now mixed up — the official name in the state language, its rendering in a text in another language, and its technical storage.

Chapter 12

Fix the phrase yourself

Back to the line where it all began. Assemble the Russian title so that both halves of the sentence live by one grammar.

If you chose «Конаев», you did exactly what Russian did with Kazakh names in 1993–1999: you gave the name a new Russian transcription and kept all its Russian cases. The city’s Kazakh name remains Қонаев — in Kazakh text, on Kazakh signs, in Kazakh speech. Two names for one place, each following the rules of its own language, like the two Kokchetav newspapers eighty years ago.

None of this requires choosing between languages. All it takes is a rule that respects the inner laws of each. A language left with its grammar accepts a new name easily — the history of Almaty and Kokshetau shows that. A language deprived of its grammar answers with a chimera.

The same holds for Kazakh. A word imposed in place of an absorbed one is rejected just as the Russian sentence rejects a motionless Төралқа. A word that arrives in time and fits accurately is absorbed so well that twenty years later nobody remembers it was ever coined — just as nobody remembers that ғаламтор was once new. In both cases a chimera is the sign of a decision taken for the language rather than with it.