Textual transmission
What the collections share
Where does Rigvedic material reappear in the Samaveda?
Eight different relationships between two verses get called 'similar' in ordinary speech, and they are not degrees of one measure. One of them records a direction — this verse carries that one's wording — and it was established for a single pair of collections. Everything else on this plate is context for that fact.
Every pair, against every kind of connection
Six corpus pairs against eight relationship classes. All 48 cells are shown, including the ones that are not numbers, and each of those prints a mark that says why.
| Pair | Exact parallel | Near parallel | Directed reuse | Variant | Shared vocabulary | Parallel, internal | Semantic resemblance | Semantic assertion |
|---|---|---|---|---|---|---|---|---|
| AV-RV | 551 | 752 | 311 | 22 | 1,176 | — | ∅ | — |
| RV-SV | 89 | 1,180 | 1,684 | 415 | 421 | — | ∅ | — |
| RV-YV | 24 | 473 | 0 | 165 | 291 | — | ∅ | — |
| AV-SV | 9 | 326 | 0 | 132 | 84 | — | ∅ | — |
| AV-YV | 7 | 162 | 0 | 41 | 136 | — | ∅ | — |
| SV-YV | 70 | 156 | 0 | 13 | 33 | — | ∅ | — |
- 0 Measured, and zero.
- — This class joins verses inside one collection, so it cannot reach a pair.
- ∅ The layer does not exist anywhere in this graph.
- 26 of 48 cells carry a measured count.
A cell is read as: this many edges of this class join these two collections. Directed classes are counted once per edge regardless of which end is in which collection.
Rigveda to Samaveda, verse by verse
Direction was established for 2 corpus pairs — AV-RV and RV-SV. 1,995 directed reuse edges exist across them; below are 14 Rigveda-to-Samaveda ones, as the two verses they connect.
1,995directed reuse edgesAcross 2 corpus pairs (AV-RV and RV-SV). The other 4 have no directed reuse, which means nobody established one.
- Rigveda, the sourcereused bySamaveda, the reuse
- RV 7.27.3match level not recordedSV ARANYA 1.2
- RV 9.97.58match level not recordedSV ARANYA 1.5
- RV 9.61.12sandhi insensitiveSV ARANYA 1.7
- RV 9.61.11match level not recordedSV ARANYA 1.8
- RV 8.93.13sandhi insensitiveSV ARANYA 2.1
- RV 9.83.3match level not recordedSV ARANYA 2.2
- RV 1.7.2sandhi insensitiveSV ARANYA 2.3
- RV 1.7.4sandhi insensitiveSV ARANYA 2.4
- RV 10.181.1match level not recordedSV ARANYA 2.5
- RV 2.41.2match level not recordedSV ARANYA 2.6
- RV 8.89.5match level not recordedSV ARANYA 2.7
- RV 1.32.1match level not recordedSV ARANYA 3.11
- RV 1.91.18sandhi insensitiveSV ARANYA 3.2
- RV 1.91.22match level not recordedSV ARANYA 3.3
The order of these rows is the order the endpoint returned, which is the Samavedic citation order. It carries no ranking: the first row is not the strongest example of anything.
What the figures show
26 of the 48 cells carry a measured count. Directed reuse reaches 2 corpus pairs, while the class with the most edges — near parallel at 3,049 — reaches 6 of 6. 1 of the 8 classes are not built at all and are shown anyway, because a row that vanished would be a row nobody could ask about.
One way to read it
That the Samaveda reuses Rigvedic verses is not a discovery of this graph — it is the oldest thing anyone says about the Samaveda. What the matrix adds is the shape of the claim: direction is recorded for AV-RV and RV-SV and not for the other 4, so a reader comparing pairs is comparing what was measured rather than what happened.
How to read this plate
- What is measured
- Relationship edges between passages in different collections, counted by the kind of relationship and by the pair of collections joined.
- What one mark is
- In the matrix, one cell is one corpus pair against one relationship class, and its weight is the number of edges. In the flow view, one line is one Samavedic verse and the Rigvedic verse it reuses.
- What is in scope
- All six corpus pairs against all eight relationship classes: 48 cells, all 48 returned.
- What is excluded
- Generic similarity. There is no single similarity score here, because the eight classes are not degrees of one measure.
- What this does not show
- That an empty cell means the two collections share nothing. Most of the 48 cells are not a measurement — some were never built, some hold a relationship class that cannot enter a corpus pair at all, and some were measured and came back zero. Each cell prints which it is, and those are facts about the graph rather than about the texts.
Reads Parallel layer, directed reuse edges, variant layer, formula mediation, shared-vocabulary layer