The body of chapter 20 says almost nothing about shudras. Two clauses:
A person who serves brahmanas, kshatriyas and vaishyas is referred to as shudra. A shudra who tills the land is vrishala, the others are dasyus.1
Three words, two grades, done.
Everything else this volume knows about the category lives in a single endnote, and the endnote is where the sorting actually happens:
Eight shudra sub-classes are mentioned — vyadha (hunter), vyalagrahi (snake-catcher), vagatita (one with whom one does not speak), chandala, and others. A chandala is the offspring of a brahmana mother and a shudra father.2
Eight sub-classes. And one of them is not an occupation at all.
Take the named items in turn, because they are not the same kind of thing.
Vyadha — a hunter. An occupation, and a polluting one in this system's terms: it involves killing.
Vyalagrahi — a snake-catcher. Also an occupation, also polluting, and specifically dangerous.
Chandala — not an occupation. A parentage, defined below.
Vagatita — one with whom one does not speak.
That fourth term is doing something the others are not.
A hunter is defined by what he does. A chandala is defined by who his parents were. A vagatita is defined by what other people withhold from him.
There is no act, no trade, no birth in the term. The category consists entirely of the behaviour of everyone else. You are a vagatita because you are not spoken to.
That deserves its own look, because it is the most unusual item in any of this volume's social taxonomies.
Every other status term in these chapters names a property of the person. Vidvan names his learning. Vrishala names his tilling. Rajan names his ruling. Even the insult chandala brahmana names his jealousy.3
Vagatita names an absence in the community.
Two consequences.
It is self-sustaining and needs no institution. A rule requiring enforcement can be relaxed by whoever enforces it. Silence has no enforcer. Every individual withholding of speech is a private act that nobody authorized and nobody can revoke, and the aggregate is a status.
And it cannot be appealed, because there is nobody to appeal to. A man declared impure by a priest can seek out that priest. A man nobody speaks to has, by definition, no interlocutor. The category's mechanism and its inescapability are the same fact.
Debroy translates the compound and does not comment further, so the vault should be careful: this is a gloss on a term, not a passage of description. What the endnote establishes is that the category existed as a named thing in the classification the compiler was working from — which is enough. Somebody had found it necessary to have a word for the man nobody talks to, and to file it among the shudra sub-classes.
The chandala definition is the item with the largest consequences, and its structure is worth pulling apart.
A chandala is the offspring of a brahmana mother and a shudra father.4
Note which way round it runs.
The mother is of the highest varna. The father is of the lowest. That is the specific configuration — and endnote 450, on a different passage, gives the general term for it: a pratiloma marriage, "where the mother's varna is superior to that of the father's."5
Pratiloma means against the grain, against the direction of the hair.
So the system distinguishes two directions of crossing. A high-varna father with a low-varna mother produces one class of offspring; the reverse produces another, and the reverse is the one that generates the most excluded category the tradition has.
Three things follow, and the third is the one that matters.
The child has done nothing. This is a status fixed at conception by two other people's act. Every other item in the eight-fold list at least involves the person — a hunter hunts, a snake-catcher catches. The chandala is defined entirely by an event that preceded him.
The variable is the direction of the crossing, not the fact of it. Mixing per se is not what produces a chandala. Mixing upward on the mother's side is.
And what that reveals is whose status is being protected. A brahmana man's child by a shudra woman is a manageable case. A brahmana woman's child by a shudra man is the worst category available. The asymmetry only makes sense if what the rule is guarding is access to brahmana women — and the child's permanent exclusion is the enforcement mechanism, applied to the one party who could not have consented to anything.
You are eleven and you have worked out that something is wrong with your family.
Not from anything anyone said. From the shape of ordinary days — where people stand when they talk to your father, which houses your mother does not enter, the particular way the man at the well looks at a point slightly beside your head.
Your grandmother was a brahmana. You know this the way you know the age of the well.
And the rule, when somebody finally says it out loud, is not about you at all. It is about her and a man who died before you were born.
There is nothing to do about it. There is no act you could perform, no count you could accumulate, no conduct that would touch it — and this is the same volume that offers a shudra brahmana-status for two and a half million japas, and tells a brahmana his good conduct makes him a vidvan and his jealousy makes him a chandala.6
Those doors are open. This one is not, because it was closed before you existed.
The taxonomy that regrades the elite by behaviour fixes the bottom by descent, and both facts are in the same chapter, and nobody in the text puts them side by side.
Worth returning to what the body actually says, because it is thinner than the endnote and its thinness is informative.
A shudra who tills the land is vrishala, the others are dasyus.7
Two grades. And the lower one — dasyu — is not a caste term at all. It is the oldest hostile word in the Vedic vocabulary, used in the Rig Veda for the enemy, the outsider, the non-Aryan.
So the body text's shudra taxonomy is: those who farm, and the enemy.
Compare what the same passage does one paragraph up. Brahmanas get seven grades. Kshatriyas get two. Vaishyas get two. Shudras get two, and the second one is a word for a foreign people.8
The resolution is not uniform, and the direction of the falloff is perfectly consistent: the finer the grading, the higher the varna. A system describes most precisely the people it is written by and for.
Which makes the endnote's eight-fold list interesting from a different angle. Somebody, somewhere in the transmission, did have a fine-grained shudra taxonomy — hunter, snake-catcher, the unspoken-to, the chandala, and four more. It just is not in the body of this Purana. It survives in the annotation layer, which is where this volume keeps turning out to store the things its body assumes and never says.9
Now the finding that neither list produces alone, and which was flagged during the Cluster A endnote pass as the strongest structural result of that lens.10
The volume has a second social-deviance taxonomy, also in the endnotes, also unstated in the body. It grades transgressive women:
Svairini, kamini, pumshchali — graded not by frequency but by whether the partner crosses varna lines.11
Put the two beside each other.
The women's taxonomy runs downward from a woman's behaviour: what she did, and specifically whether she did it across a varna boundary.
The shudra taxonomy runs upward from a child's parentage: who his parents were, and specifically whether they crossed a varna boundary.
They index on the same variable. Not frequency. Not consent. Not marriage. Sexual contact across varna lines, viewed from two ends of the same act.
One taxonomy classifies the woman. The other classifies the child. Between them they cover the complete consequence set of a single event, and neither list mentions the man.
That is the sharpest thing in either endnote, and it is invisible from inside either one. The body of the Purana contains neither list. Both survive only because a translator supplied the definitions the text assumes — which means this entire architecture of consequence was, for the text's own audience, too obvious to write down.
The structural close, and it is worth stating plainly rather than gently.
The chandala definition is a rule that assigns a permanent, hereditary, unappealable exclusion to a person on the basis of an act committed by two other people before he was born, and it does so asymmetrically, penalizing only the direction of crossing that involves a woman of the highest varna.
It is the most severe status in the system and it is the one least connected to anything its holder did.
And it sits in a volume that is elsewhere remarkably generous about status. The same book offers liberation to a murderer who overheard a recitation while dying of fever, purification to anyone who wears ash "voluntarily or involuntarily," and a route from shudra to brahmana by recitation count.12 Its dominant mechanism is a grace that does not check credentials.
None of it reaches here.
Which is worth recording precisely, because the standard defence of such passages is that the tradition's soteriology outruns its sociology — that liberation is open even where society is not. In this volume the soteriology is remarkably open. The sociology sits underneath it, in the endnote layer, unargued and unmentioned, and the text never once brings the two into contact.
The generosity and the exclusion do not conflict in the text because they never meet in it.
Strongest support is the convergence of the two endnote taxonomies — women graded by crossing, shudra sub-classes defined by crossing — on a single variable that neither list names.
Tension one. The body gives shudras two grades and brahmanas seven, and the lower shudra grade is dasyu, a word for a hostile foreign people.
Tension two. Vagatita is a status constituted entirely by other people's withholding, with no property of the person in it, and therefore no possible appeal.
Tension three. The chandala definition is asymmetric — pratiloma only — which locates the rule's real object as access to brahmana women rather than mixing as such.
Tension four. The volume offers extraordinarily open soteriology (grace without consent, liberation by proximity, varna-change by japa count) and this fixed hereditary exclusion, and never places them in the same passage.
Open question. What are the remaining four of the eight sub-classes? Debroy names four and writes "and others," which means the full list was available to him and he did not reproduce it. Whether the omitted four are occupational like the first two or structural like the last two would settle what kind of taxonomy this is, and Volume 1 does not supply them.
The vault's Śaiva material has almost no caste sociology, and this page is where the gap becomes a problem rather than an absence.
Nirvāṇa-Ṣaṭkam: I Have No Caste, No Bond holds Śaṅkara's hymn as Śiva's self-description — no caste, no parent, no bond. It is the tradition's most-quoted refusal of exactly this apparatus.
And it is a statement about Śiva, not about a chandala.
That distinction is the whole difficulty, and the vault should hold it rather than letting the hymn stand as the tradition's position on caste. A god who has no caste is not a social claim. The social claim is in the endnote, and it says a child's status is fixed by his grandparents' varna.
Brahmana: Not Caste but Knower of Brahman is the vault's philological reform of the parallel term. It applies to the top of the system. Nobody has offered a corresponding reform of chandala, and there is no obvious philological route to one — the word's definition here is a parentage, not a quality.
The Chandala's Wood Block: Caste Architecture (history). The historical record shows the category enforced through physical infrastructure — a required signalling object, prohibited routes, avoidance built into the plan of settlements.
The endnote supplies what that architecture was built around: a definition by parentage. The block, the route, the distance all follow from a rule about who someone's mother was.
What the pair produces: an exclusion defined by descent requires infrastructure precisely because it is invisible — a hunter can be identified by his bow and a snake-catcher by his basket, but a chandala looks like anyone, so the system has to manufacture a marker. The wood block is not an intensifier of an obvious status; it is a prosthetic for a status with no natural signal. That reframes the whole apparatus: the more arbitrary a category, the more physical enforcement it needs, and the elaborateness of caste marking is evidence of how little the categories corresponded to anything perceptible.
Identity Scripts: How Shame Becomes Self-Concept (psychology). The developmental material shows that a status conferred before a child can act on it is absorbed as a fact about the self rather than as a judgement about circumstances — and that this is what makes early-assigned shame structurally different from later-acquired guilt.
Vagatita and the chandala parentage are both pre-agentive assignments. One is constituted by an absence of speech, the other by an event before birth. Neither offers the person anything to have done.
The insight neither gives alone: a status assigned before agency cannot be experienced as unjust from inside, because injustice requires a comparison to what you deserved, and no claim about desert was ever available — which is why such categories produce not grievance but self-concept, and why they are more stable than any enforced rule. The psychology pages document this as a clinical mechanism in individuals. Here it is being used as social architecture, and it identifies why the pratiloma rule is placed where it is: the enforcement is aimed at adults, and the cost is loaded onto a child, who is the only party structurally incapable of experiencing it as a penalty.
Sharpest implication. The body of chapter 20 gives shudras two grades where brahmanas get seven, and the lower one is dasyu — the Rig Veda's word for the enemy. The real taxonomy survives only in an endnote: eight sub-classes including vagatita, "one with whom one does not speak" — a status constituted entirely by other people's withholding, with no property of the person in it and therefore no possible appeal — and chandala, defined as the child of a brahmana mother and a shudra father. That is pratiloma, against the grain, and its asymmetry names its real object: the rule is guarding access to brahmana women, and it loads the permanent cost onto the one party who could not have consented. Read against the volume's other endnote taxonomy — transgressive women graded by whether the partner crossed varna lines — both systems index on the same single variable, from opposite ends of one act, and neither mentions the man.
Generative questions.