Saramsha

No. 010  ·  26 September 2026

The Weight of Syllables

Sanskrit verse is measured by weight, not by stress or rhyme. Every syllable is light ⏑ or heavy —, and a metre is one fixed sequence of the two, repeated in each quarter of the verse. Type or paste a verse — Devanagari or roman — and it is scanned and named.

Hover or tap a syllable to see why it is light or heavy.

About this piece

Chandaḥśāstra, in a browser

English verse counts stresses and looks for rhyme. Sanskrit does neither. It counts weight: each syllable is either laghu, light, or guru, heavy, and a metre — a chandas — is one exact sequence of light and heavy, followed in all four quarters (pādas) of a verse. There is nothing subjective about it. Given the syllables, the weights follow from three rules, and this page applies them.

A syllable is heavy if its vowel is long (ā ī ū ṝ e ai o au — e and o are always long in Sanskrit), or if it is closed by anusvāra ṃ or visarga ḥ, or if two or more consonants follow it before the next vowel. Otherwise it is light. That third rule is the one that surprises people: weight is not a property of the syllable alone but of what comes after it, and it reaches across word boundaries, because the verse was spoken as one unbroken stream. In tat savitur, the ta is heavy only because the s of the next word arrives to close it. Aspirates are single consonants — bh, kh, th are one sound each, one letter each in Devanagari — so bhava opens light. And the last syllable of a pāda is free: heavy or light, it counts as heavy, because the pause after it fills the time. Those syllables are marked with a hollow ring here.

Because the weights are mechanical, metres can be quoted mechanically. Sanskrit does it in triplets called gaṇas, eight of them, one for each possible pattern of three. Their names are held in a single mnemonic word — ya-mā-tā-rā-ja-bhā-na-sa-la-gā — which is a marvel of compression: read any three consecutive syllables of it and their own weights are the pattern of the gaṇa the first one names. yamātā is light-heavy-heavy, and that is gaṇa ya. mātārā is heavy-heavy-heavy, gaṇa ma. Slide the window along and you get all eight, with la (light) and gā (heavy) left over at the end for the odd syllables. A metre such as Mandākrāntā is then defined in seven syllables of verse: ma-bha-na-ta-ta-ga-ga. The panel on the right shows the window sliding.

The metres here are the samavṛtta, in which all four pādas are identical, which covers most of classical poetry. Two cases work differently. Anuṣṭubh, the śloka, is the ordinary metre of the epics and the Gītā, and it is not a fixed sequence at all: eight syllables a pāda, of which only the fifth, sixth and seventh are pinned — light, heavy, then heavy in the odd pādas and light in the even ones. That looseness is exactly why it can carry a hundred thousand verses of Mahābhārata without becoming unbearable. Upajāti is a licence rather than a metre: Indravajrā and Upendravajrā differ only in their first syllable, so a poet may mix them pāda by pāda, and Kālidāsa opens the Kumārasambhava doing precisely that.

Why should poets accept a cage this tight? Partly for the sound: Mandākrāntā's four heavy syllables and five light ones really do step slowly, which is why Kālidāsa gave the Meghadūta's homesick cloud 111 verses of it; Pañcacāmara's unbroken light-heavy alternation really does beat like a drum, which is what the Śiva Tāṇḍava Stotra is for. But partly for memory. These texts were carried by voice for centuries before they were written, and a metre is an error-detecting code: drop a syllable, or swap in a synonym of the wrong shape, and the line audibly breaks. Piṅgala's Chandaḥśāstra, which set all this down around the third century BCE, went further still — enumerating every possible pattern of light and heavy in a line led him to binary numerals, the recursion we call Fibonacci's, and the table of combinations later known as Pascal's triangle, the better part of two thousand years early.

Scanning runs entirely in your browser; nothing is uploaded. Devanagari and roman input both work — IAST diacritics, or plain ASCII with doubled vowels (kaashcit), since only vowel length and consonant counts matter. A śloka written as two long lines of sixteen is split into four pādas automatically. What this does not do: mātrā-counting metres such as Āryā, the half-regular ardhasamavṛtta metres, Vedic chandas, and any verse whose pādas are not on separate lines or marked by a daṇḍa. Where nothing matches exactly, the nearest metre is offered with its disagreeing syllables ringed — often that is a genuine variant reading rather than a mistake.