Research Notes

Date: 2026-09-10. Follows raag-accuracy-audit.md (round 1, applied in PR #3). Status: applied on 2026-09-10 (Bhakti added, §2 corrections, §3 ras table, §4 Tier-A additions, §5–6 fields and rendering, §8 season addendum).

Summary

• What: every one of the 87 raags re-checked against at least two independent sources, plus a source-by-source trust evaluation, a two-tier emotion (ras) proposal with a confidence level for every raag, a ranked list of 35 raags the app lacks, a map of variant families, and a data-schema proposal to hold all of it.
• How: 5 index sources for coverage, 7 content sources for facts (The Raga Guide book text, tanarang, Parrikar, Wikipedia, ragajunglism, chandrakantha, two peer-reviewed listener studies). A fact counts as corroborated only when two sources of different lineage agree.
• Headline findings: 11 further field corrections (one thaat, nine vadi/samvadi, one time). The current ras field disagrees with the corroborated sources on 34 of 87 raags; the biggest are Malkauns, Shree, Marwa, Bhairav, Hameer, Shankara, Kalingada and Lalit. Seasons: Shree has a sourced winter association that round 1 removed. 10 raags absent from the app appear in four or more indexes.
• The hard part: no authoritative per-raag rasa table exists anywhere. The proposal below is built from the mood language of the sources, weighted by source tier, and every row carries its evidence and a confidence grade.

Notation: (k) komal, (t) tivra, ' upper octave. Source tags: RG Raga Guide, T tanarang, P Parrikar, W Wikipedia, RJ ragajunglism, CK chandrakantha, B99 Balkwill & Thompson 1999, M15 Mathur et al. 2015.

1. Source evaluation

Corroboration rule used in this report. A fact is corroborated when sources from at least two different lineages agree: RG or a tier-1 book, tanarang (Gwalior pedagogy), Parrikar (Ramrang lineage), or a Wikipedia page that cites tier 1. ragajunglism and swarganga are treated as restatements, not votes.

How well the sources agree with each other. On thaat and scale they agree almost everywhere; the disagreements are the known edge cases (Bihag: Bilawal in RG and Wikipedia, Kalyan in tanarang; Jog: Khamaj-based in Parrikar and Wikipedia, Kafi in tanarang; Chandrakauns and Saraswati: no consensus thaat). On time they agree to within one prahar for most raags; the outliers are listed in §2. On vadi/samvadi they disagree often, which the RG itself explains: the vadi concept is "a theoretical construct" and many raags have several strong notes. On mood they agree far more than expected, once the vocabulary is normalised, which is what makes §3 possible.

2. Corrections found in round 2

These are on top of round 1. Each line gives the change and the sources that agree.

Thaat

• Deshkar → Bilawal. Bhatkhande places it in Bilawal; T, P and W agree. The app has Kalyan (same five notes as Bhoopali, but Deshkar's Dha-centred, uttarang treatment is what puts it with Bilawal). Fix.
• Chandrakauns → move to the "Carnatic / no thaat" bucket and rename that bucket "Other / no thaat". T says "not defined; some say Bhairavi or Kafi". RG says it follows Malkauns rules. Keeping it under Kafi is defensible only by habit.
• Saraswati: T files it under Kalyan; P and the scale (tivra Ma with komal Ni) say it fits no thaat. Keep the current bucket, add T's view to the narrative.
• Bihag: RG and W say Bilawal (current); T says Kalyan "because of increased tivra Ma in modern practice". Keep Bilawal; add the note.
• Jog: P and W say Khamaj-based (via Tilang); T says Kafi. Keep Khamaj; add the note.

Vadi / samvadi (app → recommended, sources)

Confirmed as already correct after round 1 (two or more sources): Yaman Ga/Ni; Kedar Ma/Sa; Hameer Dha/Ga; Kamod Pa/Re; Nand Sa/Pa; Bhoopali Ga/Dha; Deshkar Dha/Ga; Hindol Dha/Ga; Alhaiya Dha/Ga; Bihag Ga/Ni; Durga Ma/Sa; Shankara Ga/Ni; Gaur Sarang Ga/Dha; Shuddha Sarang Re/Pa; Gaud Malhar Ma/Sa; Des Re/Pa; Jhinjhoti Ga/Ni; Rageshri Ga/Ni; Tilang Ga/Ni; Gorakh Kalyan Ma/Sa; Bhairav Dha/Re; Ramkali Pa/Sa; Vibhas Dha/Re; Lalit Ma/Sa; Bhatiyar Ma/Sa; Marwa Re/Dha; Puriya Ga/Ni; Sohini Dha/Ga; Shree Re/Pa; Purvi Ga/Ni; Basant Sa/Pa; Bhairavi Ma/Sa; Malkauns Ma/Sa; Bilaskhani Dha/Ga; Bhupali Todi Dha/Ga; Bageshri Ma/Sa; Bhimpalasi Ma/Sa; Patdeep Pa/Sa; Vrindavani Re/Pa; Madhyamad Re/Pa; Bahar Ma/Sa; Adana Sa/Pa; Jaunpuri Dha/Ga; Darbari Re/Pa; Asavari Dha/Ga; Todi Dha/Ga; Multani Pa/Sa; Gurjari Dha/Re; Madhuvanti Pa/Sa.

Still split, keep as is and note in the UI: Hansadhwani (app Ga/Ni; T Sa/Pa; W Re/Pa), Komal Rishabh Asavari (app Dha/Ga; T Dha/Re; W Dha/Ga), Miyan ki Malhar (app Ma/Sa; T Pa/Sa; W Ma/Sa), Ahir Bhairav (app Ma/Sa; T Ma/Sa; W Dha/Re).

Time of day

Season

• Shree: RG says "traditionally this raga should be performed during the winter, in the early evening", and that the name also refers to the early-winter harvest season. Wikipedia says winter too (weak source, but independent). Round 1 removed Winter; restore it with the RG citation.
• Kaushik Dhwani / Bhinna Shadja (a candidate addition): Parrikar cites the shastric association with Hemant (early winter).
• Kafi: RG says hori-dhamar in Kafi belongs to Holi; keep Spring as an association.
• Tilak Kamod: T alone says "particularly sung in rainy season". One source; do not add.
• Des: none of RG, T or P attach it to the monsoon. Its monsoon link comes from thumri and film usage. Keep it, but as an association, not a rule (see schema, §6).

3. Two-tier emotion (ras) proposal

The problem. The tradition never fixed one rasa per raag. Deepak Raja's position (that rasa depends on the performer and the composition) is shared by RG, which reports musicians' descriptions rather than assigning rasas, and by both listener studies, which found the same raag reads "calm" in alaap and "happy" in gat. The sources do, however, use a stable vocabulary: sweet, playful, light, romantic, longing/viraha, devotional/bhakti, serene, solemn/grave/gambhir, heroic/vigorous, austere, anxious. Mapped onto the app's nine ras:

• Shringar: romantic, sweet, love, viraha (love-in-separation), charm
• Karuna: pathos, plaintive, melancholy, longing where sorrow dominates
• Shanta: serene, peaceful, calm, meditative, solemn, gambhir; and Bhakti (devotional), which the sources use constantly but which is not one of Bharata's nine
• Veer: heroic, vigorous, majestic, bold, energetic, king-like
• Hasya: playful, joyful, lively, sprightly, buoyant, festive
• Adbhuta: wonder, mystery, haunting, the monsoon-storm sublime
• Raudra, Bhayanaka, Bibhatsa: used by the sources only for Adana (B99 "anger") and in historical notes on Bhairav and Bhinna Shadja

Recommendation on Bhakti. Add it as a tenth ras tag. Tanarang names "Bhakti ras" for Ahir Bhairav, Vibhas, Jogiya, Lalit, Todi, Multani, Bhupal Todi, Gunkali, Bahar and Bairagi; Wikipedia for Bhoopali and Bilaskhani; RG for Bhairav and Lalit. Folding it into Shanta loses the most common single mood word in the corpus. The table below uses Shanta(B) to mark rows where the source word was "devotional".

Method. Primary = the ras named by the most sources, weighted RG and Parrikar above tanarang above Wikipedia; the empirical studies break ties. Secondary = the next most-cited. Confidence: High = three or more sources agree on the primary; Med = two; Low = one source, or sources diverge. "App" is the current ras field. Rows where the proposal changes the primary are marked ▲.

Count: primary changes on 34 raags; High confidence on 33 rows, Med on 26, Low on 28.

Patterns worth knowing. The app leans on Karuna and Raudra where the sources lean on Shanta/Bhakti and Veer. Raudra is not supported for any raag except as a historical note. Bhayanaka is not supported anywhere. Adbhuta survives only for the Malhars and Chandrakauns/Bhatiyar. The "grave, majestic" raags (Darbari, Shree, Marwa, Malkauns, Bhairav) are the rows most likely to draw disagreement, because "gambhir" sits between Shanta and Veer; the table follows the sources' own words.

4. Raags the app lacks, ranked by corroboration

Coverage matrix: RG (74), tanarang index (~120), chandrakantha (42), Parrikar essays (~100 raags discussed), Wikipedia performance-time list (~80). Number in brackets = indexes that carry the raag.

Tier A: in the Raga Guide and two or more other indexes. Add these.

Tier B: two indexes, or one strong source. Add if you want depth in a family.

Kalavati (Khamaj; Pa/Sa; S G P D n; T "simple, melodious"), Kaunsi Kanada (Asavari; Ma/Sa; late night; Malkauns + Kanada; T, P), Malgunji (Kafi; Ma/Sa; late night; Khamaj + Bageshri; T, P, W), Gunkali (Bhairav; Dha/Re; S r m P d; T "Bhakti and Karuna"; P), Ramdasi Malhar (Kafi; Ma/Sa; rains; T, P), Khambavati (Khamaj; Ga/Dha; near Jhinjhoti; T, P, RG mention), Bihagda (Bilawal; Ga/Ni; Bihag + Khamaj; T, P), Jogkauns (Bhairavi per T; Gunidas 1940s; T, P), Madhukauns (Kafi; Pa/Sa; S g M(t) P n; T, P), Gauri (Bhairav-ang; Re/Pa; sunset; T, P), Narayani (Khamaj; Sa/Pa; T, P), Yamani Bilawal (Bilawal; Sa/Pa; both Ma; T, P), Devgandhar (Asavari; Dha/Ga; both Ga; T, P), Sorath (Khamaj; Des's progenitor; P, RG mention), Jait (Marwa; Pa/Sa; S r G P D; T, P), Hemant (Bilawal; Ma/Sa; T, P mention), Manj Khamaj (RG only; Ma as secondary tonic), Bibhas of Marwa thaat (RG, P; S r G P D with shuddha Dha), Vachaspati (Carnatic; T), Suha/Sughrai (Kafi; Dha varjit; T, P), Barwa (Kafi; Agra; P), Gara (Khamaj; P), Nat (Bilawal; P; rarely standalone), Khat (Asavari; sankeerna; P).

Do not add: anything appearing in only one index and without a tanarang or RG page.

5. Variant families

Parrikar's essays and the RG give a consistent picture of which raags are treated as one family. This is the basis for a family field and a variants list.

The RG's "same tone material, different raag" pairs are worth surfacing in the UI as "easily confused with": Bhoopali/Deshkar/Shuddha Kalyan; Kedar/Kamod/Hameer/Chhayanat/Gaud Sarang; Marwa/Puriya/Sohini; Purvi/Basant/Paraj/Lalita Gauri; Darbari/Adana/Kaunsi Kanada; Bhairavi/Bilaskhani Todi/Komal Rishabh Asavari; Bhimpalasi/Patdeep/Dhani; Des/Tilak Kamod/Sorath; Malkauns/Chandrakauns; Miyan ki Malhar/Bahar; Bhairav/Kalingada/Gauri/Ramkali.

6. Proposed data schema

Current fields: name, alt, thaat, time, season, ras[], char and swaras, vadi, samvadi, arohana, avarohana, aalap, narrative. Proposed additions, all optional so existing rendering keeps working:

family:        "Kanada"                    // §5
jati:          "Audav-Sampurna vakra"      // from T and RG
pakad:         "Sa Re Ma Pa, Ni(k) Dha Pa, Ma Ga Re"   // raag-vachak phrase (T, P)
ras:           ["Veer", "Shanta"]          // keep: [primary, secondary]
rasConfidence: "High" | "Med" | "Low"
rasEvidence:   "RG 'raga of kings'; P 'monumental gravitas'; W 'grave'"
timeNote:      "RG 12–3; tanarang 9–12"    // only when sources split
season:        "Any"                       // rule, not mood
associations:  ["Monsoon (thumri, film)"]  // informal links, shown differently
variants:      ["Adana", "Nayaki Kanada", "Kaunsi Kanada", "Shahana"]
confusedWith:  ["Adana", "Kaunsi Kanada"]
history:       "Named Darbari by Tansen per Faqirullah (1666); ragamala: victorious king"
sources:       [{tag:"RG", p:48}, {tag:"T", url:"..."}, {tag:"P", url:"..."}]
corroboration: 3                           // number of independent lineages agreeing on thaat+scale

Also: add Bhakti to rasClass and the filter chips; add a family filter; show rasConfidence as a small badge; render sources as footnote links in the detail panel; render associations in a lighter style than season.

7. What to apply next, in order

1. Field fixes from §2: Deshkar thaat; the 11 vadi/samvadi rows; Hindol and Dhani time; Shree season Winter; thaat notes for Bihag, Jog, Chandrakauns, Saraswati.
2. Decide on Bhakti (§3). Then apply the ras table, starting with the High-confidence ▲ rows (Yaman, Hameer, Bhoopali, Alhaiya Bilawal, Shankara, Bhairav, Kalingada, Malkauns, Bhimpalasi, Dhani, Bahar, Shree, Marwa, Puriya, Lalit) since those are the ones most likely to be noticed as wrong today.
3. Add the 12 Tier-A raags (§4) with full detail blocks; the sources above have everything needed except aalap lines, which I would write from the pakad and the RG melodic outlines.
4. Add family, variants, confusedWith (§5) for the existing 87.
5. Add sources and rasEvidence per raag from the notes file, and the schema changes in the renderer.

Working notes with every extracted fact and source are in the scratch file notes.md from this session and can be moved into the repo on request.

8. Addendum: season, re-examined (2026-09-10)

Round 1 stripped Summer, Autumn and Winter from the app on the grounds that no source supported them. That was right about the app's old values and wrong about the tradition: there is a documented six-season scheme, and stripping the field lost it. This addendum restores what is sourced and records the rest as association rather than rule.

The six ritu. Vasant (spring), Grishma (summer), Varsha (monsoon), Sharad (autumn), Hemant (early winter), Shishir (deep winter).

The treatise scheme. The raga-ragini systems were built on six patriarchal raags: the Raga Guide reproduces Damodara's (c.1625) version of the Hanuman scheme: Bhairav, Kaushik (Malkauns), Hindol, Dipak, Shri, Megh; and those six were mapped to the six seasons. The Archives and Research Centre for Ethnomusicology gives the most widely followed mapping, from the Sangita Ratnakara (13th c.) and Sangita Darpana (c.1625):

A competing scheme in the Manasollasa (Someshwara, 1131), reported by Deepak Raja, gives spring to Vasanta, summer to Bhairava, monsoon to Megh, autumn to Panchama, early winter to Natanarayana, deep winter to Shree. The two schemes agree only on Megh's monsoon and Shree's winter, which is why those two are treated as rules and the rest as history.

What survives in practice. Deepak Raja: "the enthusiasm of the performing tradition has been limited largely to Vasanta (spring) and Varsha (rainy season)", because both seasons signify nature renewing itself. Shailaja Khanna in The Tribune reports the autumn, pre-winter and summer traditions as having "passed into oblivion". The Raga Guide is consistent with both: of 74 raags it names a season in the time field for only Bahar and Basant (spring), Gaud Malhar, Megh, Miyan ki Malhar and Sur Malhar (rains), plus Shree (winter) and a note that Kafi's hori belongs to Holi.

Changes applied. The ritu framework itself stays in this report rather than becoming a page in the app; the per-raag detail panels carry the associations and the season notes.

• Vrindavani Sarang → Summer (rule). Corroborated by two lineages: tanarang, "the intensity of noon time and the summer season to which the melody is particularly oriented", and Wikipedia's season field citing Yagnik's Shastriya Raag Darshan. The Summer filter chip is restored.
• Associations added (shown in the detail panel, not filterable as a rule): Hindol, Megh, Bhairav, Shree and Malkauns each carry their ritu-scheme slot with the treatises named; Bhairav additionally carries the Manasollasa's summer and a note that it is sung in any season today; Tilak Kamod carries monsoon on tanarang's single mention; Kafi carries the Holi link; Bahar carries the spring jod-raags; the four Malhars carry Varsha.
• seasonNote added to Bhairav, Malkauns, Hindol, Shree, Megh and Vrindavani Sarang explaining rule versus history.
• New sources: ARCE (vmis.in) for the ritu scheme; Shailaja Khanna, The Tribune (tier 3) for current practice; Deepak Raja's 2013 seasonal essay folded into his catalogue entry.

Not changed, and why. Bhimpalasi, Patdeep, Multani, Madhuvanti and the other Sarangs stay at Any: their "summer" was a gloss on the afternoon hour, and no source gives them a season. Bhupal Todi's "spring" appears in one Wikipedia infobox against the Raga Guide's silence, so it stays out. Pilu's monsoon appears only in an unsourced infobox field. Des keeps monsoon as an association, from thumri and film rather than the treatises.