The same name or word can appear in several different Roman-letter spellings.

北京 may be written Beijing, Peking, or Pei-ching. The Japanese word し may become shi or si. 부산 appears officially as Busan, while older sources commonly use Pusan. The Russian name Чайковский may appear as Čajkovskij, Chaĭkovskiĭ, Chaykovskiy, or the established English form Tchaikovsky.

These differences do not necessarily mean that some writers are careless or that only one spelling is correct.

Romanized spellings differ because languages, governments, scholars, libraries, mapmakers, passport authorities, and software systems do not all need the same result. One system may preserve the original written characters. Another may help English-speaking readers approximate pronunciation. A third may avoid diacritics so that names can be typed on basic keyboards.

Romanization is therefore not one universal operation. It is a family of methods shaped by language, history, national policy, technical constraints, and intended use.

Quick Answer

Romanized spellings differ because different systems prioritize different kinds of information.

A romanization may be designed to:

• Preserve the source spelling
• Represent pronunciation
• Support reverse conversion
• Help readers of a particular language
• Standardize geographic names
• Catalog books
• Record passport identities
• Work without diacritics
• Preserve historical usage
• Match local or national policy

Unicode’s transliteration guidance states that the same text can be transliterated in many different ways. It also notes that Latin letters do not have universal sound values: a letter chosen to help an English reader may suggest another sound to a French, German, Spanish, or Indonesian reader.

The correct question is usually not:

Which spelling is universally correct?

It is:

Which system produced this spelling, what information does it preserve, and what purpose does it serve?

The Main Reasons Romanized Spellings Differ

A spelling should therefore be evaluated within its system rather than compared only by appearance.

The Latin Alphabet Does Not Have Universal Sound Values

A major reason romanization differs is that Latin letters are already used differently across languages.

The letter j may suggest:

• The sound in English judge
• A sound similar to English y in some European languages
• A sound similar to French j
• A velar or glottal sound in Spanish conventions

The letter c may suggest:

• The sound in English cat
• The sound in English city
• A ts-like sound
• A ch-like sound in another orthographic tradition

Unicode gives Japanese romanization as an example. Hepburn uses j with a broadly English-oriented value, but its vowels do not always follow ordinary English pronunciation. Kunrei-style Japanese uses forms such as ti, which must be read according to Japanese romanization rules rather than as an English syllable.

This creates an unavoidable design problem.

A spelling that is helpful for English speakers may be less intuitive for:

• French speakers
• German speakers
• Spanish speakers
• Portuguese speakers
• Indonesian speakers
• Speakers of other Latin-script languages

There is no completely language-neutral pronunciation spelling based on ordinary Latin letters.

Script and Language Are Not the Same Thing

Romanization must identify the source language, not merely the source script.

The Cyrillic script is used for Russian, Ukrainian, Bulgarian, Serbian, Macedonian, Mongolian, and many other languages. Arabic-derived scripts are used for Arabic, Persian, Urdu, Pashto, Kurdish, Sindhi, and additional languages. Devanagari is used for Sanskrit, Hindi, Marathi, Nepali, and other languages.

The same visible letter may therefore receive different Latin equivalents according to the language.

For example, Cyrillic Г is commonly represented as:

• g in Russian
• h in Ukrainian under the current national geographic convention
• Another language-specific value in other Cyrillic alphabets

ISO 9 takes a broad script-oriented approach by covering Cyrillic characters used in Slavic and non-Slavic languages, while language-specific systems can incorporate the orthography and pronunciation of one language more directly.

Similarly, Han characters cannot be assigned one universal Roman reading.

香港 can be represented according to Mandarin pronunciation, but the internationally established form Hong Kong reflects Cantonese naming history. Japanese kanji must be interpreted through Japanese readings rather than automatically converted with Mandarin Pinyin.

A converter that detects “Cyrillic,” “Arabic script,” or “Chinese characters” but does not identify the language may return a technically consistent yet linguistically inappropriate form.

Transliteration and Transcription Preserve Different Information

Romanized spellings also differ because some systems follow written characters while others follow pronunciation.

Transliteration

Transliteration primarily represents the source writing system.

A strict transliteration may attempt to:

• Assign a distinct Latin form to every source character
• Preserve spelling even when pronunciation changes
• Retain silent or historical letters
• Support reverse conversion
• Avoid merging separate source characters

ISO 9 is an example of a system built around systematic Cyrillic character transliteration. ALA-LC guidelines similarly state that library romanization tables should generally be transliteration schemes rather than pronunciation guides and should support machine transliteration and reversibility where practical.

Transcription

Transcription represents spoken pronunciation.

A pronunciation-oriented system may:

• Change a consonant according to its position
• Add vowels absent from the writing
• Reflect assimilation
• Ignore distinctions no longer heard
• Depend on a standard dialect

South Korea’s Revised Romanization is explicitly based on standard Korean pronunciation and avoids non-letter symbols where possible. As a result, the same Hangul consonant may receive different Latin representations in different phonetic positions.

Hybrid systems

Many widely used systems combine both principles.

They may preserve most written distinctions while applying selected pronunciation rules. This makes them more practical for public use but less reversible than a strict character transliteration.

The words transliteration, transcription, and romanization are also used somewhat differently by different institutions. The behavior of the system matters more than its label.

Different Systems Are Built for Different Purposes

Romanization systems cannot be understood without considering why they were created.

Scholarly systems

A scholarly system may preserve:

• Vowel length
• Tone
• Retroflex consonants
• Historical spelling
• Morphological distinctions
• Separate source characters

It may use forms such as:

• ā
• č
• ḥ
• ṛ
• ṣ
• š
• ṭ
• ū
• ž

These forms are precise but require readers to understand the notation.

Library systems

ALA-LC Romanization Tables are designed for bibliographic records, authority control, searching, citations, sorting, shelving, and record exchange. The Library of Congress maintains separate current tables for numerous languages and scripts because cataloging requires stable and reproducible distinctions.

A library form may look unfamiliar to general readers because cataloging consistency is more important than immediate pronunciation.

Geographic-name systems

UNGEGN promotes romanization systems for geographic names that are scientifically sound, nationally implemented, and sufficiently reversible for their intended purpose. Its approach begins with national standardization rather than imposing one foreign spelling system on every country.

Map and gazetteer systems may therefore differ from library or academic transliteration.

Passport systems

Passports must operate across legal, administrative, and machine-readable environments.

ICAO Doc 9303 requires a Latin transcription or transliteration when mandatory personal data is written in a non-Latin script. The machine-readable zone uses a restricted character repertoire, so diacritics, punctuation, or other features may be transformed or omitted.

A passport form is therefore not necessarily a full scholarly romanization or the spelling a person would choose for publication.

Public-facing systems

Road signs, tourist information, schools, and news media often prefer spellings that are:

• Easy to recognize
• Easy to type
• Free from specialist symbols
• Consistent with established public usage

A practical public system may intentionally sacrifice reversibility for readability.

National Policies Produce Different Spellings

Governments may adopt their own romanization standards for:

• Road signs
• Place names
• Passports
• Education
• Public records
• Maps
• Official publications

These systems can change over time.

South Korea introduced Revised Romanization to reflect standard Korean pronunciation using ordinary Roman letters where possible. This produced modern place-name forms such as Busan, Daegu, and Jeju, replacing older McCune–Reischauer-style forms such as Pusan, Taegu, and Cheju in many official contexts.

Japan issued new national Roman-letter guidance on December 22, 2025, replacing its earlier government notice. The current main table uses familiar Hepburn-like forms such as shi, chi, tsu, fu, and ji, although the national rules are not identical to every convention historically described as Hepburn.

National changes do not instantly erase older forms. Previous spellings may remain in:

• Historical documents
• Company names
• University names
• Family names
• International organizations
• Maps
• Search indexes
• Established English vocabulary

This is why multiple generations of romanization can coexist for decades.

Historical Systems Leave Long-Lasting Traces

Romanization often reflects the period in which a name entered international use.

Before Hanyu Pinyin became dominant, Chinese names were represented through systems such as Wade–Giles, postal spellings, and regional pronunciation traditions.

This produced familiar contrasts such as:

ISO 7098 now describes the romanization principles of Modern Chinese Putonghua and applies to bibliographies, catalogs, indexes, and toponymic lists. ALA-LC Chinese romanization is also Pinyin-based but applies library-specific rules and omits tone marks.

Older spellings survive because they may be attached to:

• Famous historical figures
• Published works
• Institutional names
• Philosophical vocabulary
• Archival records
• Diaspora family identities

A historical form should not necessarily be “corrected” when quoting or cataloging a source from the period in which it was standard.

Romanization Can Be Filtered Through Another Language

Some Roman spellings did not develop directly from the present-day source language.

A name may have passed through:

• English
• French
• Dutch
• Portuguese
• Spanish
• Russian
• Ottoman Turkish
• A regional trade language
• A colonial administration

For example, North African place and personal names may have French-influenced spellings, while names in Southeast Asian communities may preserve Dutch, Portuguese, English, or local Latin-script conventions.

This process is sometimes called indirect romanization.

The path may be:

Local language → another regional or colonial language → Latin spelling

rather than:

Local language → modern international romanization standard

Indirect forms may look irregular under a modern table but remain historically or officially established.

Dialects and Language Varieties Produce Different Forms

A writing system may represent several spoken varieties.

Arabic names may be pronounced differently in:

• Modern Standard Arabic
• Egyptian Arabic
• Levantine Arabic
• Gulf Arabic
• Maghrebi Arabic
• Iraqi Arabic

The Arabic letter ج, for example, does not have one identical pronunciation throughout the Arabic-speaking world. Short vowels, the definite article, and consonants such as ق also vary regionally.

Chinese characters likewise receive different readings in:

• Mandarin
• Cantonese
• Hokkien
• Teochew
• Hakka
• Other Sinitic languages

A Cantonese-derived surname should not automatically be rewritten in Mandarin Pinyin merely because both are represented by the same Han character.

A romanization tool must therefore distinguish:

• The source script
• The source language
• The language variety
• The chosen standard
• The person’s or place’s established usage

Diacritics Create Full and Simplified Forms

A complete romanization system may require diacritics, while public or technical forms remove them.

Examples include:

The simplified form is easier to type, but information may be lost.

Diacritics can distinguish:

• Long and short vowels
• Tones
• Different source consonants
• Retroflex and dental sounds
• Emphatic consonants
• Separate Cyrillic letters
• Nasalization

ALA-LC and ISO systems often retain such marks because differentiation and possible reversibility matter. Public signage and web addresses frequently omit them for convenience.

The best practice is not to choose between marked and unmarked forms permanently. Store the full form and generate simplified versions for search or technical use.

ASCII and Keyboard Limitations Change Spellings

Older computer systems, typewriters, telegraph networks, passports, URLs, and basic keyboards could not always represent extended Latin letters.

A marked letter might therefore be replaced by:

• A plain letter
• A digraph
• A doubled vowel
• An apostrophe
• A number
• A capital-letter convention

Examples include:

These substitutions improve compatibility but can make different source forms identical.

A URL such as /tokyo/ may be entirely appropriate even when the article displays Tōkyō. The URL is a technical identifier, not the complete linguistic representation.

Word Division and Hyphenation Differ

Romanization does not only decide which letters to use. It also decides how to divide words.

Scripts may not use spaces in the same way as English.

A system may need to determine:

• Whether a compound is one word
• Whether particles are attached
• Whether name syllables are joined
• Whether an administrative suffix is hyphenated
• Whether an apostrophe marks a syllable boundary

Chinese Pinyin generally joins syllables into words, but specialized library rules can divide them differently. Korean given names may appear joined, hyphenated, or spaced. Older Wade–Giles names commonly hyphenated syllables, while Pinyin usually joins them.

Examples include:

Two systems can use similar character mappings but produce visibly different output because their word-division policies differ.

Personal Names Follow Identity, Not Only Linguistic Rules

Names frequently resist standardization because people and families already use established Latin spellings.

Examples include Korean surnames:

South Korea’s official romanization framework allows personal and company names to continue using established spellings.

Similar variation occurs elsewhere:

• محمد → Muḥammad, Muhammad, Mohammed, Mohamed
• 大野 → Ōno, Oono, Ohno, Ono
• علي → ʿAlī, Ali, Aly
• Чайковский → Čajkovskij, Chaĭkovskiĭ, Chaykovskiy, Tchaikovsky

For identity use, the practical priority is usually:

1. Official document spelling
2. Person’s preferred spelling
3. Established professional form
4. Systematic romanization

A generated form can be shown as a linguistic alternative, but it should not be presented as a correction of someone’s identity.

Geographic Names May Be Romanizations or Exonyms

A foreign-language name is not always a romanization.

Consider Москва:

• Moskva is a direct Roman representation of the Russian name.
• Moscow is the conventional English name.

Similarly:

• القاهرة → Al-Qāhirah through romanization
• القاهرة → Cairo in conventional English

UNGEGN promotes nationally standardized geographic names and romanization systems, while recognizing that international naming practice also involves endonyms, exonyms, historical names, and language-specific conventions.

This distinction explains why maps in different languages may use different labels even when everyone agrees on the identity of the place.

An English atlas may use:

• Moscow
• Athens
• Cairo
• Vienna

A local-name or linguistic database may use:

• Moskva
• Athína
• Al-Qāhirah
• Wien

Neither approach is automatically wrong. They answer different editorial questions.

Examples Across Major Languages

Chinese: Beijing, Peking and Pei-ching

For 北京:

• Beijing is Hanyu Pinyin.
• Pei-ching is associated with Wade–Giles.
• Peking is a historical conventional or postal-era form.

Pinyin represents Modern Standard Mandarin and is reflected in ISO 7098. Older forms remain important in historical sources and established names.

Japanese: Shi and Si

The kana し can appear as:

• shi in Hepburn-style systems and Japan’s current principal guidance
• si in Kunrei-shiki and Nihon-shiki

Hepburn provides a pronunciation cue familiar to many English readers. Kunrei and Nihon emphasize the regular structure of the Japanese kana rows.

Unicode uses this type of contrast to show that a transliteration may require users to learn a special convention rather than read the form according to their ordinary spelling habits.

Korean: Busan, Pusan and Pwusan

For 부산:

• Busan follows Revised Romanization.
• Pusan follows McCune–Reischauer.
• Pwusan follows Yale linguistic romanization.

Revised Romanization is designed for modern South Korean public use. McCune–Reischauer remains important in historical scholarship. Yale is designed to reveal linguistic structure rather than provide a simple public pronunciation spelling. South Korea’s official system is based on standard pronunciation and favors ordinary Roman letters.

Russian: Čajkovskij, Chaĭkovskiĭ and Chaykovskiy

For Чайковский:

• Čajkovskij reflects ISO 9-style transliteration.
• Chaĭkovskiĭ reflects ALA-LC library practice.
• Chaykovskiy reflects a practical English-oriented system.
• Tchaikovsky is the established English form of the composer’s name.

ISO 9 covers Cyrillic alphabets across Slavic and non-Slavic languages, while ALA-LC provides separate language-specific tables for bibliographic use.

Arabic: Muhammad, Mohammed and Muḥammad

For محمد:

• Muḥammad preserves the distinct Arabic consonant represented by ḥ.
• Muhammad removes the diacritic.
• Mohammed, Mohammad, and Mohamed reflect different language traditions, pronunciations, and established identities.

Arabic writing also normally omits many short vowels, so complete pronunciation-oriented romanization may require lexical or grammatical knowledge. Passport authorities may apply state-specific transliteration rules within ICAO’s broader document framework.

Indic Scripts: Kṛṣṇa and Krishna

For कृष्ण:

• Kṛṣṇa is a full scholarly transliteration.
• Krishna is a simplified conventional form.

The diacritics distinguish vocalic ṛ, retroflex ṣ, and retroflex ṇ. ISO 15919 provides a broader framework for Devanagari and related Indic scripts, while simplified public spellings commonly omit the distinctions.

Politics and Cultural Identity Can Influence Romanization

Romanized names can carry political and cultural significance.

A change may signal:

• Preference for the local national language
• Rejection of an earlier colonial spelling
• Independence from another language’s naming tradition
• Restoration of an Indigenous name
• A new official language policy
• Alignment with international standards

The difference between Kyiv and Kiev, for example, reflects source-language choice as well as spelling. One is based on Ukrainian; the other is historically associated with a Russian-influenced international form.

Geographic-name standardization is therefore not purely mechanical. UNGEGN’s work connects national naming with language, culture, history, local implementation, and geographic administration.

A neutral reference should explain the source and status of each form rather than treating romanization software as an authority on political legitimacy.

How to Decide Which Spelling to Use

Begin with the use case.

Ask five questions

Before choosing a spelling, ask:

1. Which language is being represented?
2. Which romanization system applies?
3. Is the purpose pronunciation, spelling preservation, identity, cataloging, or mapping?
4. Is there an official or established Latin form?
5. Which distinctions are lost if the spelling is simplified?

These questions resolve most apparent conflicts.

Recommended Editorial Policy

A publication can adopt the following policy.

Contemporary personal names

Use the spelling the person uses in reliable official or professional sources.

Geographic names

Use the current officially standardized form unless the publication has a documented conventional-name policy.

Historical subjects

Preserve historically meaningful spellings and provide modern equivalents where useful.

Technical terms

Use the romanization standard expected by the discipline.

First mention

Where helpful, show the original script:

Tōkyō (東京)

or:

Muhammad (محمد)

Alternative systems

Mention a variant when it helps readers connect older and newer sources:

Mao Zedong, written Mao Tse-tung in Wade–Giles

Diacritics

Preserve them in the full reference form. Provide simplified search aliases separately.

Recommended Database Structure

A multilingual database should not rely on one field called romanized_name.

Use separate fields such as:

This structure allows several spellings to coexist without implying that they have equal authority.

Best Practices for Romanization.org

Romanization.org should explain variation rather than hide it.

A conversion result should ideally display:

Original script

The unchanged source text.

Language and script

For example:

• Mandarin Chinese — Han characters
• Japanese — kanji and kana
• Korean — Hangul
• Ukrainian — Cyrillic
• Arabic — Arabic script
• Hindi — Devanagari

Primary system

The exact standard and version.

Alternative systems

A side-by-side comparison.

Official or preferred form

Where known, this should be separated from generated results.

Simplification status

Labels such as:

• Fully marked
• Diacritics removed
• ASCII fallback
• Pronunciation-oriented
• Reversible
• Partially reversible
• Not reversible

Historical notes

Explain whether the spelling is:

• Current
• Historical
• Conventional
• Colonial-era
• Library-specific
• Geographic
• Passport-derived

Information-loss warning

State which distinctions disappeared during simplification.

Common Misconceptions

One word has one correct romanization

Many words have several correct forms under different standards.

The easiest spelling is the most accurate

A simple spelling may omit tones, vowel length, source-character distinctions, or grammatical information.

A technical transliteration is always the official name

Passports, personal preferences, governments, and institutions may use established alternatives.

Romanization works the same for every language using a script

The language must be identified before applying a table.

A newer system makes every older spelling wrong

Older forms remain valid in historical, institutional, and identity contexts.

Diacritics are decorative

They may distinguish different source letters or sounds.

English spelling conventions are universal

Readers of different Latin-script languages interpret letters differently.

A map name is always a direct romanization

It may be an exonym, historical name, translation, or conventional form.

Romanization can always be reversed

Simplification, pronunciation rules, omitted vowels, and merged letters often make reverse conversion impossible.

Frequently Asked Questions

Why does one name have several English spellings?

The spellings may represent different romanization systems, languages, dialects, historical periods, or identity preferences.

Which romanization system is the most accurate?

Accuracy depends on the goal. A strict transliteration may be most accurate about spelling, while a transcription may be more accurate about pronunciation.

Why is Beijing also called Peking?

Beijing is the modern Pinyin form. Peking is an older conventional spelling that entered international use through earlier naming traditions.

Why is Busan also written Pusan?

Busan follows South Korea’s Revised Romanization. Pusan follows the older McCune–Reischauer tradition.

Why is Japanese し written shi or si?

Hepburn-style systems use shi to assist many international readers. Kunrei-shiki and Nihon-shiki use si to preserve the regular structure of the kana table.

Why is Muhammad spelled Mohammed or Mohamed?

Arabic vowel representation, regional pronunciation, European-language traditions, passport rules, and personal preference all contribute to the variants.

Why is Krishna written Kṛṣṇa in scholarly works?

The marked form preserves distinct Sanskrit sounds and letters. The common English spelling removes the diacritics.

Why is Moscow not written Moskva in English?

Moskva is a direct Roman representation of Москва. Moscow is the established English exonym.

Should personal names be converted to the newest standard?

Not when the person has an official or preferred spelling. Generated forms should be treated as alternatives.

Should websites use diacritics?

Use them in the authoritative display form when the standard requires them. Create separate unmarked forms for search and URLs.

Is an ASCII spelling a romanization system?

It may be a defined simplified system, but often it is merely a technical fallback derived from a fuller form.

Can software automatically choose the correct spelling?

Software can generate possibilities, but official names, personal preference, historical context, dialect, and ambiguous readings often require authoritative data or human review.

Final Checklist

Before selecting a Roman spelling, confirm:

• The source language is known.
• The source script is known.
• The intended use is defined.
• The romanization system is named.
• The standard version is recorded.
• Official or preferred forms have been checked.
• Historical usage has been considered.
• Dialect and regional pronunciation have been considered.
• Diacritics are preserved where required.
• Simplified forms are labeled as lossy.
• Word division and hyphenation follow the chosen system.
• Exonyms are distinguished from direct romanizations.
• Original-script text remains stored.
• Search aliases do not replace the primary form.
• Automated output has been reviewed where ambiguity exists.

Conclusion

Romanized spellings differ because romanization is never only about replacing one set of letters with another.

Every system makes choices about:

• Language
• Pronunciation
• Orthography
• Audience
• Reversibility
• Diacritics
• National policy
• Historical continuity
• Identity
• Technical compatibility

A form such as Beijing, Peking, or Pei-ching represents a different historical or systematic route from the same Chinese characters.

Forms such as Busan, Pusan, and Pwusan reflect different goals: modern public communication, older pronunciation-oriented scholarship, and linguistic analysis.

Forms such as Kṛṣṇa and Krishna differ because one preserves scholarly distinctions while the other prioritizes ordinary typing and public familiarity.

The existence of several spellings does not mean standards have failed. It means the spellings were designed for different users and purposes.

A trustworthy romanization reference should therefore preserve the original script and explain:

1. Which language is represented
2. Which system produced the Latin form
3. What audience the system assumes
4. Whether the output follows spelling or pronunciation
5. Which information is preserved or removed
6. Whether an official, historical, or preferred form takes priority

Romanization.org can make these relationships visible. Instead of presenting one spelling as universally correct, it can show why the alternatives exist—and help users select the form appropriate to their map, passport, catalog, publication, database, or language-learning context.