Spoken Tamil vs Written Tamil: Understanding Diglossia

Spoken Tamil and written Tamil are two distinct registers of the same language: a formal, literary style used in books, news and official writing, and a colloquial style used in everyday conversation, films and most homes. The difference goes well beyond accent, covering verb endings, pronouns, word shapes and vocabulary, which is why linguists call Tamil a classic case of diglossia. Almost every Tamil speaker switches between the two without thinking, but learners, and computers, find the gap surprising.

This article explains the spoken Tamil vs written Tamil divide with real examples, and why it matters for anyone learning Tamil or building Tamil language technology.

#What is diglossia?

Diglossia describes a situation where a community uses two varieties of one language for different purposes. One is the "high" variety, learned at school and used for writing and formal speech. The other is the "low" variety, learned at home and used for everything else. The linguist Charles Ferguson popularised the term in English in 1959, using Arabic, Greek, Swiss German and Haitian Creole as his main cases, and Tamil is now widely cited as another classic example. The Wikipedia article on diglossia gives a good overview.

"High" and "low" are technical labels, not judgements. Colloquial Tamil is not broken or lazy Tamil. It has its own consistent grammar, and it is the variety in which people actually live their lives.

#Centamil and kotuntamil: the traditional names

Tamil grammarians have long recognised this split. Formal literary Tamil is often called செந்தமிழ் (centamiḻ, "refined" or "beautiful Tamil"). The traditional counterpart term is கொடுந்தமிழ் (koṭuntamiḻ, literally "crooked Tamil"), used for the everyday spoken forms.

In modern usage you will more often hear:

• எழுத்துத் தமிழ் (eḻuttut tamiḻ, "written Tamil")
• பேச்சுத் தமிழ் (pēccut tamiḻ, "spoken Tamil")

There is also a widely used middle ground, sometimes called standard or common Tamil, heard in news broadcasts and public speeches. It follows written grammar but is softened for listening.

#Spoken Tamil vs written Tamil: real examples

The best way to see the difference is side by side. The spoken forms below are typical of Tamil Nadu speech; Jaffna and other dialects differ again.

Meaning	Written Tamil	Spoken Tamil
I am going	போகிறேன் (pōkiṟēṉ)	போறேன் (pōṟēṉ)
What are you doing?	என்ன செய்கிறாய்? (eṉṉa ceykiṟāy?)	என்ன பண்றே? (eṉṉa paṇṟē?)
it is there / there is	இருக்கிறது (irukkiṟatu)	இருக்கு (irukku)
they	அவர்கள் (avarkaḷ)	அவங்க (avaṅka)
don't want / no need	வேண்டாம் (vēṇṭām)	வேணாம் (vēṇām)
come (respectful)	வாருங்கள் (vāruṅkaḷ)	வாங்க (vāṅka)

A few patterns jump out:

• Shortened tense markers. The written present-tense marker -கிறு- (-kiṟu-) collapses to -ற- (-ṟ-) in speech.
• Plural and respect endings. Written -கள் (-kaḷ) becomes -ங்க (-ṅka).
• Consonant changes. Clusters such as -ண்ட- (-ṇṭ-) simplify, so வேண்டாம் becomes வேணாம்.
• Different verbs. Spoken Tamil often prefers பண்ணு (paṇṇu, "do, make") where writing uses செய் (cey).

None of this is random. Once you learn the patterns, you can usually predict the spoken form from the written one, and vice versa.

#Where each style is used

Most Tamil speakers move between registers many times a day.

Written Tamil appears in:

• Newspapers, magazines and textbooks
• Government notices and legal documents
• Classical and modern literature
• Formal speeches, news reading and religious discourse

Spoken Tamil appears in:

• Conversation at home, at work and with friends
• Film and television dialogue
• Phone calls, voice notes and most YouTube content
• Increasingly, social media posts and chat messages

Songs and fiction often mix the two. A novel may narrate in written Tamil while characters speak in colloquial Tamil, sometimes in a specific regional dialect. Online, many people now type spoken Tamil in Tamil script, or in Latin letters mixed with English, which we discuss in Tanglish and code-switching.

#Why the spoken vs written gap trips up learners

Heritage learners and adult students both run into the same problem. Classroom Tamil teaches the written standard, so students learn போகிறேன். Then they visit family and hear போறேன், or in Jaffna, போறன். Formal Tamil spoken aloud can sound stiff or comical in a casual setting, a little like answering "How are you?" with "I am in a satisfactory condition, thank you."

Practical advice for learners:

• Learn both, deliberately. Treat spoken and written Tamil as two linked skill sets.
• Pick a spoken dialect. Choose the one used by the people you talk to most. Our guide to Tamil dialects can help you understand the options.
• Learn the conversion patterns. A handful of rules (like -கிறு- to -ற-) unlock hundreds of words.
• Listen a lot. Films, podcasts and conversations with relatives are where spoken Tamil lives.

#Why diglossia makes Tamil hard for AI

For language technology, the spoken Tamil vs written Tamil gap is one of the biggest challenges. We touch on it in why AI struggles with Tamil, but it deserves a closer look.

Text data is mostly written Tamil. Wikipedia, news sites and books are written in the formal register. A model trained mostly on those sources tends to reply in stiff, formal Tamil, even when someone writes to it casually.

Speech data is mostly spoken Tamil. When someone talks to a voice assistant, they use colloquial Tamil with all its contractions and regional flavour. A speech system trained only on people reading formal sentences will struggle with natural speech.

Transcription is a choice. If someone says இருக்கு, should a transcript write இருக்கு or the formal இருக்கிறது? Both are defensible, but a dataset has to be consistent, and a model has to understand that they mean the same thing.

Dialects multiply the problem. There is one written standard but many spoken varieties. A model needs to connect all of them to the same meanings.

This is one reason Eelam Lab is building Nila (நிலா, nilā, "moon"), a Tamil large language model trained from scratch rather than fine-tuned from an English model. Nila is still in development. To handle both registers well, it needs high-quality written text and a large amount of genuine, everyday spoken Tamil from many regions.

#Frequently asked questions

#Is spoken Tamil grammatically incorrect?

No. Spoken Tamil follows its own consistent rules; it simply differs from the written standard. Linguists treat colloquial Tamil as a full, rule-governed variety, not as a corrupted form of written Tamil.

#Do Tamil speakers write the way they speak?

Traditionally, no. Formal writing uses the literary register. That is changing online, where many people type colloquial Tamil in chats, comments and social media posts, either in Tamil script or in Latin letters.

#Should I learn spoken or written Tamil first?

It depends on your goal. If you want to talk with family or friends, start with spoken Tamil in their dialect, and learn the script alongside it. If you want to read news or literature, you will need written Tamil, and spoken Tamil will still help you understand the patterns.

#Is diglossia unique to Tamil?

No. Arabic, Greek (historically), Swiss German and several South Asian languages show similar patterns. Tamil is often cited because the gap between its registers is especially wide and long-standing.

#Contribute everyday Tamil, the way you speak it

Most datasets have plenty of formal Tamil and very little of the Tamil people actually speak. You can help close that gap in a few minutes.

In the contribution studio, you can talk freely for up to five minutes in your natural spoken Tamil, read a written Tamil sentence aloud, or type and upload text you own. Both registers are useful, and recordings of relaxed, everyday speech are especially valuable. Every contribution needs your explicit consent and is reviewed by people before it is used in training.

Open the contribution studio and record a short story the way you would tell it at home.