What if every Indian language merged into one? The maths of it

A short equation from Cornell says two languages cannot coexist for ever. Fit it to India's census and a single mother tongue takes 90 to 500 years, unless history, or bilingualism, gets in the way.

A signboard at Imphal airport written in Meitei, Hindi and English
Photo: Haoreima, CC BY-SA 4.0, via Wikimedia Commons

If every Indian language merged into one, India would lose most of its linguistic diversity: its Greenberg diversity index, about 0.78 on Census 2011 language shares, would fall towards zero. Mathematically, a merger is not absurd: the Abrams-Strogatz model of language competition, fitted to the rise of Hindi from 37.0% to 43.6% of mother tongues between 1971 and 2011, gives 90 to 500 years for one language to reach 99%. In practice it would not happen, because that model assumes no bilingualism, and India runs on it: 26% of Indians were bilingual in 2011, and similar languages can coexist stably when people speak both.

Key takeaways
  • 121 languages and 19,569 raw mother-tongue names were recorded in Census 2011; 22 are in the Eighth Schedule.
  • 43.63% of Indians gave Hindi as their mother tongue in 2011, and 57.1% spoke it as a first, second or third language.
  • About 0.78: India's Greenberg diversity index at the language level, the chance two random Indians have different mother tongues (my calculation).
  • a = 1.31: the exponent Abrams and Strogatz found across 42 regions; with it, two languages cannot coexist unless something holds them apart.
  • About 70 deaths: the official toll of the 1965 anti-Hindi agitation, which ended plans for Hindi as the sole official language.

India is often called a country of a thousand languages, and the census agrees in its own careful way. This post asks the opposite question with real tools from linguistics and physics: how diverse India is, how fast one language could absorb the rest, what the merged language would sound like, and what stops it. For the raw rankings, see the most spoken languages in India.

Has India ever tried to make one language win?

Yes, and it backed down in 1965. The Constitution set Hindi as the official language of the Union, with English allowed alongside it for 15 years. The deadline was 26 January 1965.

The story

As 26 January 1965 approached, Madras State erupted. On the eve, police took DMK leader C. N. Annadurai and about 3,000 party members into preventive custody. On the day, about 50,000 students marched from Napier Park towards Fort St. George. Clashes spread across the state for weeks, with police firings, burning of Hindi name boards and several self-immolations; the official death toll was about 70, and unofficial estimates ran far higher. On 11 February, Prime Minister Lal Bahadur Shastri promised on radio that English would continue. The Official Languages (Amendment) Act passed in December 1967, making Hindi and English bilingualism effectively indefinite.

Wikipedia: Anti-Hindi agitations of Tamil Nadu (from contemporary reports)
The memorial hall in Tamil Nadu honouring those who died in the language agitations
A memorial in Tamil Nadu to the "language war" martyrs. A forced merger has already been tried once; this is what it cost.Photo: Surya Prakash.S.A., CC BY-SA 3.0, via Wikimedia Commons

How diverse are India's languages, in one number?

About 0.78 on the Greenberg index, meaning that two Indians picked at random have a 78% chance of different mother tongues, counting at the level of the census's 121 languages.

The worked derivation: the Greenberg index

The linguist Joseph Greenberg's diversity index is the probability that two randomly chosen people speak different first languages:

\[ D = 1 - \sum_{i} p_i^{\,2} \]
  1. Square each share (Census 2011 shares, via Wikipedia). Hindi: \( 0.4363^2 = 0.190 \). Bengali: \( 0.0803^2 = 0.0064 \). Marathi: \( 0.0686^2 = 0.0047 \). The next twelve languages above 0.5% add about 0.0158.
  2. Sum. The 15 largest languages give \( \sum p_i^2 \approx 0.217 \). The remaining 4.43% (100% less the 15 largest languages' 95.57%, Census 2011 table), split among many small languages, adds almost nothing.
  3. Subtract from one.
    \[ D \approx 1 - 0.217 = 0.78 \]
  4. Sensitivity. Hindi alone contributes 0.19 of the 0.22. Counting Hindi's many census mother tongues (Bhojpuri, Rajasthani and others) separately would push D higher; counting Hindi and Urdu as one Hindustani would push it down to about 0.75.
Mother tongue share of India's population, Census 2011
See the data as a table
Share of population (%)
Hindi43.63
Bengali8.03
Marathi6.86
Telugu6.7
Tamil5.7
Gujarati4.58
Urdu4.19
Kannada3.61
Odia3.1
Malayalam2.88
Punjabi2.74
Assamese1.26
Maithili1.12
Santali0.61
Kashmiri0.56

Source: Census of India 2011, Paper 1 of 2018 (as tabulated in Wikipedia, Languages of India)

Use this to check what you just read. Set Hindi's share and the number of other languages, splitting the rest equally, and watch the index fall as one language grows.

Diversity if one language grows

How fast could one language take over?

Somewhere between 90 and 500 years, if you take a famous physics model at face value. The model is the Abrams-Strogatz equation, the best-known mathematical theory of language competition.

Modelling the dynamics of language deathDaniel M. Abrams and Steven H. Strogatz, Cornell University · Nature 424, p. 900 · 2003

The authors treat two languages as competing for speakers, with the attraction of a language rising with its number of speakers and its "status", the social and economic opportunity it offers. They fitted the model to historical counts from 42 regions in Peru, Scotland, Wales, Bolivia, Ireland and Alsace-Lorraine, covering Welsh, Scottish Gaelic and Quechua, and found the exponent \( a \approx 1.31 \pm 0.25 \) almost constant across cultures. With \( a > 1 \), only the all-X and all-Y states are stable: one language eventually wins. The limit, which the authors state: the model assumes a well-mixed population of monolinguals, and real bilingual societies exist, as Quebec French shows.

Read the paper (free)

The worked derivation: fitting the model to India

  1. The equation. Let \( x \) be the share speaking language X (here, Hindi), \( s \) its status between 0 and 1, and \( c \) a rate constant:
    \[ \frac{dx}{dt} = c\left[(1-x)\,x^{a}\,s \;-\; x\,(1-x)^{a}\,(1-s)\right], \qquad a = 1.31 \]
  2. The data. Hindi's census share: 36.99% (1971), 38.74% (1981), 39.29% (1991), 41.03% (2001), 43.63% (2011).
  3. The fit. I integrated the equation numerically and searched for the \( c \) that best fits those five points for several assumed statuses. Each fit matches the data to within about 0.4 percentage points.
  4. Project forward from 2011. The time for \( x \) to reach 90% and 99%:
Assumed status sFitted c (per year)Years to 90%Years to 99%
0.550.2496691
0.600.059170260
0.700.024248419
0.800.015285508

Sensitivity and honesty. Five census points cannot separate status from speed, which is why the range is so wide. Worse, much of Hindi's rise is probably faster population growth in Hindi-speaking states, not people switching language, and the model counts both as "conversion". So read this table as an upper bound on how fast a merger could happen, not a forecast.

One language takes over: three possible futures0%50%100%19712011210022712519s = 0.55: ~91 yrs = 0.60: ~260 yrs = 0.80: ~508 yrcensus points
The census record (dots) and three Abrams-Strogatz projections to 99%. The model cannot tell which future is right from 40 years of data.

Why hasn't one language already won?

Because Indians are bilingual, and Indian languages are similar to each other. The Abrams-Strogatz model has no bilinguals in it. A follow-up paper added them.

Interlinguistic similarity and language death dynamicsJorge Mira and Ángel Paredes, University of Santiago de Compostela · Europhysics Letters · 2005

Mira and Paredes extended the Abrams-Strogatz model with a third group of bilingual speakers and a similarity parameter \( k \) between 0 (no overlap) and 1 (identical languages), which sets how easily people become bilingual. Fitting it to Galician and Castilian Spanish in north-west Spain, two closely related Romance languages, they found \( a = 1.50 \), Galician status \( s = 0.26 \), \( c = 0.1 \) and \( k = 0.80 \), and showed that a stable bilingual state can exist when languages are similar enough, which the original model rules out. The limit: one pair of languages, fitted with few parameters.

Read the paper (free)

India fits that picture well:

  • Bilingualism is common: 31.49 crore people, 26% of the population, reported a second language in 2011.
  • A link language exists without a merger: Hindi reached 57.1% as a first, second or third language, while English serves as a second link.
  • Languages already converge in structure: linguists have described South Asia since the 1950s as a linguistic area or Sprachbund, where unrelated families such as Indo-Aryan and Dravidian share features like retroflex consonants, echo words and verb-final word order (Murray Emeneau, 1956).
  • Status is protected by states: after linguistic reorganisation, each major language has a state government raising its status, which is exactly the "control on s" that Abrams and Strogatz say can stabilise coexistence.
A railway ticket counter sign at Bhubaneswar station written in Odia, Hindi and English
Odia, Hindi and English at Bhubaneswar station. The three-language sign is India's working answer to the merger question.Photo: କଳିଙ୍ଗପୁତ୍ର, CC BY-SA 4.0, via Wikimedia Commons

What would the merged language sound like?

Something like Hindustani with a heavy load of borrowed words. Languages that merge by contact usually keep the grammar of the majority and soak up vocabulary from the rest; Hindustani itself grew that way, from a Khari Boli base with Persian, Arabic and later English words.

How long until the old languages are unrecognisable?

Glottochronology, the much-criticised but still useful dating method, gives a rough clock. It assumes a language keeps a fixed share \( r \) of its core vocabulary per thousand years. Robert Lees measured \( r = 0.805 \) on the 200-word list; Swadesh's 100-word list gives \( r = 0.86 \).

  1. Shared core words after time t (in millennia) between two languages that split apart:
    \[ c = r^{2t} \quad\Longleftrightarrow\quad t = \frac{\log c}{2\log r} \]
  2. Example: for two daughter dialects of the merged language to keep only half their core vocabulary in common, with \( r = 0.86 \) (Swadesh's value, Wikipedia: Glottochronology; the sum is calculated in this post):
    \[ t = \frac{\log 0.5}{2\log 0.86} = \frac{-0.301}{2(-0.0655)} \approx 2.3\ \text{millennia} \]
  3. Sensitivity. With \( r = 0.805 \), it is about 1.6 millennia. Critics note real rates vary widely: Bergsland and Vogt found about 4% per millennium for Icelandic but as much as 20% for Literary Norwegian, so treat this as order of magnitude only.

The point for our question: a single merged language would not stay single. Spread over a subcontinent, it would split into new regional dialects within a couple of thousand years, just as Sanskrit and the Prakrits once did.

Merge, then split againHindiBengaliTamilTeluguMarathi+ 116 moreone languagenorth dialecteast dialectsouth dialectabout 1.6 to 2.3 millennia to share only halftheir core words (glottochronology)90 to 500 years (Abrams-Strogatz fit)
The full counterfactual timeline. Merging is the fast half; splitting again is the slow half that always follows.

What would India lose?

Literatures, scripts and a lot of knowledge stored in words. The 22 scheduled languages each carry centuries of written work, and many of the 121 census languages have no other home.

MeasureIndia in 2011Merged India
Languages counted by the census1211
Greenberg diversity indexabout 0.78about 0
Language families by populationIndo-Aryan 78.07%, Dravidian 19.64%, Austroasiatic 1.11%, Sino-Tibetan 1.01%one
Bilingual people31.49 crore (26%)none needed
A merged language would be one of the fastest-changing things in India: give it two thousand years and it would be many languages again.The working of this post

So, could every Indian language merge into one?

  1. Diversity: India starts at a Greenberg index of about 0.78.
  2. Dynamics: the Abrams-Strogatz fit gives 90 to 500 years to 99% for one language, as an upper bound on speed.
  3. Brakes: 26% bilingualism, similar languages (Mira and Paredes) and state support for status all stabilise coexistence.
  4. History: the one serious push, in 1965, ended in about 70 official deaths and a law protecting bilingualism.
  5. Aftermath: even a merged language would split into new dialects within about 2,000 years (1.6 to 2.3 millennia, calculated in this post from the Lees and Swadesh retention rates).
Verdict: possible with changes

A single shared language is already partly real: Hindi reaches 57.1% of Indians as a first, second or third language, and English links the rest. A single mother tongue for everyone is mathematically possible only if India drops bilingualism and lets one language's status run unchecked for 90 to 500 years, and 1965 showed the political cost of trying. The version that works is the one India runs now: many mother tongues, plus one or two link languages, which the Mira-Paredes model says can stay stable when languages are close and people speak more than one.

What if next?

Also on this blog: why India counts in lakh and crore. More questions live on the Kyun page.

Questions people ask

How many languages does India have?

Census 2011 grouped 19,569 raw mother-tongue names into 1,369 rationalised mother tongues and then into 121 languages, of which 22 are scheduled in the Constitution.

Is Hindi India's national language?

No. The Constitution names Hindi in Devanagari as the official language of the Union, with English continuing as an official language under the Official Languages Act; India has no "national language" in law.

Why did Tamil Nadu oppose Hindi in 1965?

Because Hindi was due to become the sole official language on 26 January 1965. Weeks of protests and about 70 official deaths led to Shastri's assurance on English and the 1967 amendment securing bilingualism.

Are Hindi and Urdu the same language?

In everyday speech they are close varieties of Hindustani with a shared grammar; they differ in script and in formal vocabulary drawn from Sanskrit or Persian and Arabic. The census counts them separately, at 43.63% and 4.19%.

Can two languages survive side by side?

Yes, when many people speak both and the languages are similar. Mira and Paredes showed this formally for Galician and Spanish, and India's 26% bilingualism is a real-world example.

Sources

  1. Abrams and Strogatz, Nature: Modelling the dynamics of language death (2003) (source for the model and a = 1.31)
  2. Mira and Paredes, Europhysics Letters: Interlinguistic similarity and language death dynamics (2005) (source for k = 0.80 and stable bilingualism)
  3. Wikipedia: Languages of India (Census 2011 tables) (source for shares, 121 languages and family shares)
  4. Office of the Registrar General: Census of India 2011, Paper 1 of 2018, Language (source for 1971-2011 Hindi shares)
  5. Wikipedia: Multilingualism in India (source for 31.49 crore bilinguals and Hindi's 57.1%)
  6. Wikipedia: Anti-Hindi agitations of Tamil Nadu (source for the story and the 1967 Act)
  7. Wikipedia: Glottochronology (source for r = 0.805, 0.86 and the criticisms)
  8. Wikipedia: Sprachbund (source for South Asia as a linguistic area)
  9. Jolad and Agarwal, The India Forum: what the census obscures (2022)
  10. Business Today: Census 2027 to be conducted in two phases (2026)
Indian languagesCensus 2011Abrams-Strogatz modelGreenberg diversity indexHindustaniglottochronologyanti-Hindi agitation 1965linguistic area
Sapta

Sapta (Saptarshi Sarkar) is a mechanical engineer turned marketer in Bengaluru and a solo touring rider. He writes the seven Saptaverse blogs. Work with him or read more of Whyblog.

Keep reading in Whyblog

From the other blogs