The Language of Yoga

Pāṇini and the Aṣṭādhyāyī

In the 4th century BCE, a grammarian from northwest India codified the entire Sanskrit language in approximately 4,000 sūtras. The Aṣṭādhyāyī — "The Eight Chapters" — does not describe Sanskrit after the fact. It generates it. It is a productive system: inputs (roots, suffixes, conditions) plus transformation rules applied in a specific order produce grammatically correct forms. No exceptions.

The structure is remarkable. Pāṇini defines metarules that determine application order: a specific rule overrides a general one, a later rule overrides an earlier one. It is a precedence system identical to what modern parsers use to resolve ambiguities.

He also invented compact notation — the pratyāhāras, which function as macros. ac means "all vowels." hal means "all consonants." It is programming with macros, 24 centuries before programming existed.

In 1948, linguist Paul Kiparsky demonstrated that Pāṇini's rules are formally equivalent to Noam Chomsky's generative grammars — published 2,300 years later. The BNF notation that John Backus created in 1959 to define programming languages is conceptually identical to Pāṇini's metalanguage. Backus did not know Pāṇini.

There is something profound in this: a 4th-century BCE pandit and a 20th-century computer scientist independently arrived at the same architecture — productive rules, metarules, compact notation — because both were answering the same question: how do you generate a complete formal system from a finite set of rules?

Devanāgarī

Devanāgarī (देवनागरी) is the script in which Sanskrit is written — along with Hindi, Marathi, and Nepali. Its name means "the script of the city of the gods" (deva + nagara).

What distinguishes it from nearly all other writing systems is its near-perfect correspondence between grapheme and phoneme. Each character represents exactly one sound. There is no irregular orthography, no silent letters, no homographs with different pronunciations. If you know the rules, you can correctly pronounce any word written in Devanāgarī. If you know the rules, you can correctly transcribe any word you hear.

The system is organized with phonetic logic: vowels first (a ā i ī u ū ṛ e ai o au), then consonants grouped by place of articulation — gutturals (ka kha ga gha ṅa), palatals (ca cha ja jha ña), retroflexes (ṭa ṭha ḍa ḍha ṇa), dentals (ta tha da dha na), labials (pa pha ba bha ma), and finally semivowels, sibilants, and aspirate. Each group of five follows the same pattern: voiceless, voiceless aspirated, voiced, voiced aspirated, nasal. It is not arbitrary — it is a matrix of phonetic features.

Every consonant carries an inherent a vowel. द is read da, not d. To remove that vowel, a virama ( ्) is added below: द् is d without a vowel. It is an economical system — the syllable da needs a single character, while d alone needs only a diacritical mark of elimination.

IAST — International Alphabet of Sanskrit Transliteration

If Devanāgarī is so precise, why not use it throughout the site? Because it is designed for those who already read Devanāgarī. For everyone else, it is a barrier. IAST — International Alphabet of Sanskrit Transliteration — is the bridge: an unambiguous encoding of Sanskrit in Latin characters that allows correct reading and pronunciation without learning a new alphabet.

The convention was formalized at the Congress of Orientalists in Geneva in 1894, based on the system developed by British philologist William Dwight Whitney and adopted by the Calcutta Sanskrit series. Every Sanskrit phoneme has exactly one IAST representation. No ambiguity is possible.

The difference between ś, ṣ, and s is not decorative — they are three distinct phonemes with three distinct places of articulation. Writing "shiva" instead of "śiva" or "ṣiva" loses phonetic information. IAST preserves it. Writing "Sankhya" instead of "Sāṃkhya" loses vowel length, nasality, and place of articulation. IAST keeps them all.

That is why on hatha.es you will find Aṣṭāvakra Gītā and not "Ashtavakra Gita." The former tells you exactly how to pronounce it. The latter is an approximation that loses the nuances.

Sandhi — Combination as function

Sandhi (सन्धि, "junction") is the system of phonetic transformations that occur when sounds meet at the boundary between words or between components of a compound. In Sanskrit, sandhi is not optional — it is mandatory and completely deterministic.

It works exactly like a mathematical function: same input, same output, always. No exceptions, no dialects that alter it, no speakers who apply it halfway. It is a phonetic law, not a convention.

This has a direct consequence for anyone reading Sanskrit texts: words transform on contact. Yoga Sūtra 1.2 is written yogaś citta vṛtti nirodhaḥ with sandhi, but the underlying forms are yogaḥ citta vṛtti nirodhaḥ. The final ḥ of yogaḥ becomes ś before the c. Without understanding sandhi, the word yogaś looks like a different word entirely.

On hatha.es we keep sandhi forms in direct quotations — because that is how they appear in the original texts — but break them down in notes and the glossary, where what matters is the isolated form of each term.

Sanskrit on hatha.es

This entire site breathes IAST. Every Sanskrit term you encounter — in the texts, in the glossary, on the practice pages — uses the IAST convention. It is not an aesthetic decision. It is a precision decision.

Sanskrit is the technical language of yoga. Its terms are not interchangeable with their translations. Dhyāna is not exactly "meditation." Samādhi is not exactly "concentration." Prāṇa is not exactly "breath." Each Sanskrit term carries a technical meaning that is lost when we substitute an approximate translation. Keeping IAST is keeping precision.

Pronunciation matters in practice. A mispronounced mantra is not the same thing. A bīja (seed sound) with the accent on the wrong syllable changes meaning. IAST gives you the information you need to pronounce correctly — if you know how to read the diacritics.

If you have never read IAST, start with the most frequent diacritics: the line above a vowel indicates length (ā is long, a is short), the dot below retroflexes the consonant (ṭ is pronounced with the tongue curled back), the tilde over ñ is the same as in Spanish. With those three, you can already read 80% of the IAST that appears on this site.