Chapter 273 is one paragraph. Inside it are roughly seventy name-tokens for Śiva liṅgas — sixty-six of them distinct — each with a place attached or a fruit promised, and almost none with a story.1
"Tilabhandeshvara is in Dashashvamedha. Sangamesha is said to be at the confluence of Ganga with the ocean… Batukeshvara is on the banks of Gandaki… Gopeshvara, Rangeshvara, Vameshvara, Nagesha, Kamesha, Vimaleshvara, Vyaseshvara, Sukesha, Bhandeshvara, Humkaresha, Surochana, Bhuteshvara and Sangamesha are also said to be famous as the destroyers of great sins."1
That last sentence lists thirteen liṅgas in a row with a single shared predicate.
Two chapters earlier, twelve names in a short verse acquired twenty chapters of story and a promise that reciting them on waking clears your sins. Here, sixty-six names get one paragraph and no promise beyond a general one.
⚑ Both are lists of holy places produced by the same compiler within three chapters of each other. One is the most famous list in Śaiva practice and the other is essentially unread. The difference is not the quality of the sites, and this page is about what the difference actually is.
Chapter 273 closes by declaring a structure: "I have spoken about the Shiva lingams that are in the eastern direction, the ones that are ordinary and the rest. O best among sages! I will now speak about the Shiva lingams that are in the southern direction."1
A four-direction survey, and the volume delivers all four. It delivers them in four completely different formats.
⚑ The frame is symmetrical and the contents are wildly not. One direction gets a gazetteer, one gets a novella, one gets a single shrine at length, one gets a paragraph. The four-direction structure is a container inherited from somewhere, and the compiler filled it with whatever material he had for each quarter — which is a more informative thing to find than a tidy survey would have been.
The eastern list can be checked mechanically, and it does not survive the check.
Counting distinct names against total mentions in that single paragraph: six names appear twice — Saṅgameśa, Bhūteśvara, Nāgeśa, Vimaleśvara, Siddheśvara and Rāmeśvara.1
Some of those are plainly different sites carrying one name: Saṅgameśa means lord-of-the-confluence and there are many confluences; the paragraph itself gives one "at the confluence of Ganga with the ocean" and mentions another unlocated. But the text never says so. It lists them as separate items in a single enumeration, twice each, with no note that the repetition is intentional and no differentiating detail for four of the six.
⚑ And the collisions run outward as well as inward. The eastern catalogue contains Vaidyanātha, Nāgeśa, Rāmeśvara and Bhīmeśvara — four of the twelve jyotirliṅgas — listed among "the ones that are ordinary and the rest." It also contains Kardameśa, which chapter 272 had just assigned as the upaliṅga of Oṃkāra.5
So a name can be a jyotirliṅga in one chapter, an upaliṅga in the next, and an ordinary eastern liṅga in the one after that, with nothing marking the difference.
That is not necessarily error — Indian sacred geography is full of genuine homonyms, and the volume elsewhere has documented name-collisions (a second Dārukavana; two figures called Sudharma). But it means the gazetteer cannot be read as a register. A reader cannot tell, from the text alone, whether the Vaidyanātha in the eastern list is the jyotirliṅga in Jharkhand or a different shrine that shares its name, and the compiler gives no way to find out.
The clearest single case sits at the seam between the western and northern sections.
Chapter 281 places Gokarṇa unambiguously: "Gokarna is an excellent tirtha on the shores of the western ocean." It is the western section's whole subject, and its liṅga is Mahābala.3
Chapter 282, answering a request for liṅgas "in the northern direction", opens: "There is another kshetra known as Gokarna and it destroys great sins… There is Shiva's excellent lingam there, known as Chandrabhala."4
⚑ The word another is doing a great deal of work, and it may be doing it honestly — the text may be signalling a genuinely different place of the same name, which would make this a name-collision rather than a contradiction.
But note what happens either way. If they are two places, the volume has just filed two same-named tīrthas in two directions three chapters apart with the single word another as the entire disambiguation. If they are one place, the directional frame has broken. Neither reading gives you a usable geography, and the text supplies nothing to choose between them.
The catalogue is not uniformly storyless. Four or five entries carry a clause of narrative — a compressed remainder of an episode — and where they point is the interesting part.
"Krittivasheshvara is simultaneously in the form of an aged man and a young boy." "The one known as Japyeshvara is at the place where Dadhichi fought." "So is Andhakesha, the place where, in ancient times, Shankara killed the daitya Andhaka." The eastern list opens with Avimuktaka as Kāśī's principal liṅga.1
Every one of those points back into material this volume has already narrated at length. Kṛttivāsa is the flayed hide of Gajāsura. Dadhīci's fight and the taking of his bones is chapter 253. Andhaka is chapters 212 to 219. Avimukta is Kāśī, which the volume works over repeatedly.
⚑ The catalogue's story-traces are all internal. Not one of the sixty-six carries a clause pointing to an episode the volume has not told somewhere else — which suggests the compiler was not surveying a countryside of shrines but indexing his own contents, attaching a location to each episode he had already delivered and then padding the list to length with names.
The clause on Kṛttivāseśvara is the exception worth noting, because it is the only genuinely new image in the paragraph — a liṅga "simultaneously in the form of an aged man and a young boy." It is unexplained, and Debroy's note supplies only its position: "near the Kashi Vishvanatha temple."6 One vivid detail, stranded in a list, with nothing attached and nowhere to go.
Read the eastern catalogue for its rewards rather than its names and a second structure appears.
Almost every entry carries one: "When men see Siddhanatheshvara, they obtain siddhi" · Brahmeśvara "bestows the four objectives of human existence" · Bharadvājeśvara "enhances the energy of the brahman" · Śūlaṭaṅkeśvara "directly grants whatever is desired" · Candreśvara "bestows a radiance like that of the moon" · Someśvara "wards off all adversities."1
⚑ Several are puns: the moon-lord grants moon-like radiance, the siddhi-lord grants siddhi, the jñāna-bestower is Nandeśvara. The reward is generated from the name.
That is a composition method, and it is visible because it is repeated. Where a site has a story, the reward comes from the story — Mahābala clears brahma-hatyā because a cow and a king were cleared there. Where a site has only a name, the reward is derived from the name itself, which is the cheapest possible way to give sixty-six entries a fruit each.
Which is worth knowing before citing any single entry in this catalogue as evidence of anything about that site's actual cult.
Set the two lists side by side and the difference is not prestige.
The twelve are short, ordered, versified, and each eventually attached to an episode. A person can hold them, say them, and tell one.
The sixty-six are long, unordered, unversified, and attached to nothing but their own names. Thirteen of them share a single predicate in one sentence. There is no way to hold the list, and no reason to — the reward for the whole class is generic, so remembering any particular member buys nothing.
⚑ The compiler had the same material available for both: names, places, fruits. He gave one list a structure that could be carried and the other a structure that could only be consulted — and only one of them was ever carried.
The chapter's own closing phrase says as much without meaning to: "the ones that are ordinary and the rest." The gazetteer is explicitly the residual category, and residual categories are where information goes to be preserved rather than used.
You have inherited a reference document — a component library, a controls register, a list of every known customer segment, an internal wiki's index — and you need to know whether anybody actually uses it.
Count the distinct entries against the total mentions. Duplicates inside a single list mean it was assembled from more than one source and never reconciled. Six repeats in one paragraph, here, and none of them acknowledged.
Check whether the entries collide with a more prominent list elsewhere. ⚑ A name appearing in the flagship list and in the residual catalogue, with nothing distinguishing them, is the signature of a document nobody has read end to end. Four of the twelve jyotirliṅgas are sitting in the eastern gazetteer among "the ones that are ordinary."
Look at where the entries' descriptions came from. If most of them are derived from the entry's own name — the caching layer handles caching, the escalation policy handles escalations — the descriptions were generated to fill the column, not written from knowledge. That tells you the register is a shape rather than a record.
And ask whether the list has a form anybody can carry. Short, ordered and versified gets recited. Long, unordered and generic gets consulted, which means it gets consulted almost never and corrected never at all. If you need a reference to actually shape behaviour, the twelve-item version with a story attached to each will beat the exhaustive one every time — and the exhaustive one will quietly rot while remaining technically correct.
Evidence. Ch 273 read in full and its names counted directly — roughly 72 name-tokens, 66 distinct, six repeated;1 the directional structure traced across ch 273 to 282 by locating every occurrence of "southern/western/northern direction" in the body;1234 ch 272's upaliṅga list checked against ch 273's names for collisions;5 the two Gokarṇa passages read in place.34
Tension 1 — the repeats may be genuine homonyms. Saṅgameśa at two different confluences is entirely plausible. The finding is not that the text erred; it is that the text supplies no way to tell, which makes the list unusable as a register whichever is true.
Tension 2 — the two Gokarṇas.34 The word another may signal a genuine second site or may be papering over a directional inconsistency. Preserved, not resolved.
Tension 3 — four jyotirliṅgas appear in the residual catalogue.15 Either the volume holds them to be different shrines of the same name, or the two lists were compiled from different sources and never cross-checked. Nothing in the text chooses.
Open question. The northern section promises to speak "briefly" and delivers a handful of sites. Is there any point in the corpus where the four directions receive comparable treatment — or is the asymmetry here simply what the available material looked like? Not determinable from the range read.
Against The Twelve Jyotirliṅgas: the necessary pairing, and the more interesting half of it. That page shows a list built to be carried; this one shows the same compiler, three chapters later, building one that could only be consulted. The material was the same and the outcome was not.
Against The Tīrtha Launders Anything: Mahābala at Gokarṇa is the western section's whole content, and it is the site that clears a premeditated brahmin-killing. The gazetteer's one narrated direction is the one that produced this build's sharpest doctrinal finding — story-bearing sites are where the arguments live.
Against The Five Faces and the Eight Forms: a fourth instance of enumeration failing under its own frame. There a schema announced eight and explained seven; here a catalogue announces distinct entries and repeats six.
Against The Four Doublets and the Four Disclosures: the same non-disclosure habit at list scale. A name recurring inside one paragraph is the smallest possible unflagged repetition, and it is unflagged for the same reason the large ones are.
The Steadicam Voice: Show, Don't Explain — that page holds the craft principle that a reader retains what they were walked through and forgets what they were told about.
These four chapters are an unusually clean natural test of it, because the same author, on the same subject, in the same fortnight of composition, does both.
The eastern direction is explained: sixty-six liṅgas, each with a location and a fruit, thirteen of them sharing one predicate in a single sentence. The western direction is shown: one site, three chapters, a cow who costs out a murder and a king who cannot shake the sin he ate. Nobody has ever needed to be told which of those two the tradition kept.
⚑ And the mechanism the pairing exposes is sharper than "stories are more memorable." Look at where the content came from. In the shown material, the site's power is derived from what happened there — Gokarṇa clears brahma-hatyā because the text just watched it clear brahma-hatyā twice. In the explained material, the site's power is derived from its own name: the moon-lord grants moon-like radiance, the siddhi-lord grants siddhi. Explaining did not merely produce weaker prose; with nothing to walk the reader through, it left the author nothing to say, so he generated the promises from the labels.
Which turns the craft principle into a diagnostic that works before you have written anything: if the only thing you can say about an item is a restatement of what it is called, you do not have material, and no amount of listing will produce any. The instruction to show is not a stylistic preference — it is a check on whether you know anything.
Sharpest implication. Three chapters after issuing the twelve jyotirliṅgas as a memorisable verse, the same compiler produces sixty-six liṅgas in one paragraph, six of them named twice inside it, four of them jyotirliṅgas demoted to "the ones that are ordinary and the rest", and one of them the upaliṅga he had assigned to Oṃkāra a chapter earlier. He announces a four-direction survey and delivers it in four incompatible formats — a gazetteer east, eight chapters of narrative south, a single shrine given three chapters west, and "briefly" north — with Gokarṇa appearing in two of the four quarters, disambiguated by the word another and nothing else. And the promises in the catalogue are generated from the names: the moon-lord grants moon-like radiance, the siddhi-lord grants siddhi. Where the volume had a story it derived the site's power from the story; where it had only a name it derived the power from the name. One of these two lists is the most famous thing in Śaiva geography and the other is unread, and the difference was decided at the level of form, not of holiness.
Generative questions.