Which raga is this?

Two steps, in this order

Identifying a raga is not one problem. It is a wide problem followed by a narrow one, and people fail at it mostly by trying to solve both at once.

The wide problem is which notes are in play. That is mechanical, and it reduces hundreds of ragas to a handful. The narrow problem is which of these few, and that is decided by phrases rather than by the note set. Ragas that share a scale are common; ragas that share a scale and a phrase vocabulary are rare.

So: get the notes first, then stop looking at notes and start listening to movement.

Step one: find the tonic

Everything downstream depends on this and nothing downstream can survive getting it wrong. If you place Sa a note off, you will transcribe a real melody into a raga that does not exist, and the shortlist will be confidently, uselessly wrong.

1. 01Listen for the drone. If there is a tanpura, the tonic is being handed to you — the most prominent sustained pitch is Sa.
2. 02With no drone, listen for where phrases come to rest. Ragas resolve; the note a line settles onto at the end of a phrase is almost always Sa, occasionally Pa.
3. 03Hum the resting note and hold it against the recording. If it sits inside the music rather than fighting it, you have Sa. If it beats against sustained notes, try a fifth up or a fourth down.

Step two: write the notes relative to Sa

Write swaras, not pitches. C D E tells you nothing portable; S R2 G3 tells you everything, because raga identity is a pattern of intervals from the tonic and is entirely independent of what the tonic happens to be.

Keep the ascending and descending phrases in separate lists as you go. The difference between them is not a detail — for a large share of ragas it is the fingerprint.

At this point a note set of five to seven swaras will usually match somewhere between three and ten named ragas. That is a good outcome, not a failure. Feed the set into a reverse lookup and read the candidates.

Step three: separate the look-alikes

Now the listening starts. Four kinds of evidence do nearly all the work, roughly in order of how quickly they resolve an ambiguity.

1. Asymmetry between ascent and descent

Many janya ragas drop a note going up and restore it coming down. Bilahari and Shankarabharanam contain the same seven swaras, but Bilahari skips M1 and N3 on the way up and uses everything on the way down. Hearing that omission once settles it.

2. Vakra motion

Some ragas do not climb straight. A vakra phrase doubles back — P M1 R2 G2 R2 S rather than a clean descent. Vakra shapes are strongly characteristic and are almost impossible to confuse once you know to listen for them.

3. The jiva swara

The note the raga keeps returning to and sits on. Two ragas with an identical scale often have different centres of gravity, and the ear picks this up quickly even when it cannot name what it noticed.

4. Gamaka on a specific note

Often the deciding evidence. One raga holds a note plain where another oscillates it heavily; the same G2 can be a steady pitch in one raga and a wide kampita in another. When two candidates survive everything else, this is usually what remains.

Pairs that catch people, and what decides them

A working reference for the ambiguities you will actually meet. The right-hand column is the single piece of evidence to listen for.

What software can and cannot tell you

It is worth being precise about this, because the gap between the two is where wrong answers come from.

• ·Scale detection is reliable. Given a clean monophonic recording and a stable tonic, pitch tracking will recover the note set accurately. This is a solved problem.
• ·Tonic detection is mostly reliable. With a drone present it is nearly certain. Without one, and on a short excerpt, it can land a fifth out — and then everything after it is wrong.
• ·Raga naming is a ranked guess. The evidence that separates same-scale ragas is phrase-level and gamaka-level. Systems that claim a confident raga name from four seconds of audio are reporting a scale match and calling it a raga.

The honest use of a tool here is as the wide step: let it collapse hundreds of possibilities to a shortlist in a second, then do the narrow step yourself with your ears. Treat a single confident raga name from a machine as a hypothesis to check, not a result.

Getting faster at it

• ·Learn the melakarta arithmetic. Knowing that mela n and n + 36 differ only in madhyama, and that neighbours within a chakra differ only in Dha and Ni, turns most parent-scale ambiguities into a single yes-or-no listen.
• ·Learn the big janya families. Kharaharapriya, Shankarabharanam and Harikambhoji parent a large share of the performed repertoire. Knowing their children covers an enormous amount of ground.
• ·Collect deciding notes, not raga lists. The useful memory is not “Bilahari is a janya of 29” but “Bilahari skips Ma going up”. One is a fact; the other is a test you can run while listening.
• ·Sing your guess back. Singing the candidate over the recording exposes a wrong answer within a phrase or two, far faster than reasoning about it does.

Tools for this

Common questions

Can you identify a raga from its notes alone?

Only to a shortlist. Many ragas share an identical note set and differ in how the notes are approached and ordered — Mohanam and Bhoopali share a scale, and Shree and Madhyamavati overlap heavily. The note set is step one of two.

What is a jiva swara?

The note a raga keeps returning to and lingers on — its centre of gravity. Two ragas with the same scale often have different jiva swaras, and hearing which note the phrases settle on is frequently what separates them.

What is the difference between arohana and avarohana?

Arohana is the ascending sequence of notes, avarohana the descending one. In many janya ragas they differ: a note may be skipped on the way up and included on the way down, and that asymmetry is a fingerprint.

Do I need to know Sa before identifying a raga?

You need a Sa, not the Sa. Raga identity is a pattern of intervals, so as long as you pick one note as the tonic and stay consistent, the pattern is readable. Choosing the wrong tonic is the single most common cause of a wrong answer.

Can an app identify a raga automatically?

An app can identify a scale reliably and suggest ranked raga candidates. Naming a raga with confidence usually needs phrase-level evidence — the characteristic gamakas and turns — which is harder to detect and easy to get wrong on a short sample.

How do I tell two similar ragas apart?

Look for the one note that behaves differently. Usually it is a note that is skipped in one direction, or is held plain in one raga and oscillated in the other. Find that note and the ambiguity generally resolves.