The Thirteen and the Hundred

At the end of the Muktika Upanishad, the god Rama, in answer to a question from Hanuman about how a person attains liberation, gives a number. There are, he says, one hundred and eight Upanishads, and he lists them, sorted under the four Vedas. The list is the source of the figure everyone repeats. It is also, by the standards of this corpus, a very late document, and the moment you look closely at it the tidy number begins to come apart in instructive ways.

This chapter is the bibliographic one. It asks the flat questions the great dialogues make it easy to forget. How many of these texts are there. Which ones actually carry the weight. And how a scattered body of forest teachings, composed across many centuries by people who left no names, came to be counted and ranked at all.

The number that will not hold still

Start with the honest answer. There is no fixed number of Upanishads, and there never has been.

The 108 of the Muktika is the famous figure, and the number is not accidental. One hundred and eight is treated as auspicious across a great many Indian traditions, in the beads of a rosary and the names of a deity, and a canon that lands on exactly 108 has the shape of completeness about it. But other collections count differently. Some manuscripts and editions gather more than two hundred texts under the name Upanishad. The reason is simple and important: the word was an open category for a very long time. Composing a new text and calling it an Upanishad, to lend it the authority of the form, went on into the medieval period and beyond. There are Upanishads devoted to particular deities, to yoga, to renunciation, to sectarian doctrine, that are younger than the great commentators who never mention them.

So the 108 is not an ancient census. It is a late catalogue, drawn up long after the classical period, that fixed an auspicious number onto a body of texts of wildly different ages. To know how many Upanishads there are, you first have to decide which question you are asking.

The principal few

The question that matters is narrower. Which Upanishads are old, and which ones did the tradition build on. Here the field shrinks dramatically, from a hundred and eight to a little over ten.

The standard English edition, Robert Hume’s of 1921, made the boundary famous in its very title: The Thirteen Principal Upanishads. Hume’s thirteen are the Brihadaranyaka and Chandogya, the two great prose giants; the Taittiriya, Aitareya, and Kaushitaki; the verse texts Kena, Katha, Isha, Mundaka, and Shvetashvatara; and the late short texts Prashna, Mandukya, and Maitri.

But thirteen is not the only count, and the older tradition tends to a smaller one. The commentator Shankara, in the eighth century, is held to have written commentaries on a core set usually given as ten: Isha, Kena, Katha, Prashna, Mundaka, Mandukya, Taittiriya, Aitareya, Chandogya, and Brihadaranyaka. That set of ten, the dashopanishad, became for many lineages the definition of what counted. Add the Shvetashvatara, on which a commentary is attributed to Shankara though its authenticity is doubted, and you reach eleven. Add the Kaushitaki and the Maitri, both old enough and weighty enough to belong with the principals, and you reach Hume’s thirteen.

So the swing is real and small. The principal Upanishads are ten if you follow Shankara’s secure commentaries, and thirteen if you follow the modern scholarly inclusion of the borderline texts. The Maitri in particular is the least stable member, late and composite, and even Müller noted that parts of it read as supplements. Where to draw the line is a matter of convention, and this treatise simply names the convention rather than pretending there is a fact of the matter.

There is one genuine dispute inside this that is worth flagging, because it shows how alive these questions still are. The commentary on the Mandukya attributed to Shankara has been challenged. The scholar Hajime Nakamura argued that its language, including terms borrowed from Buddhist philosophy, does not match Shankara’s secure work, and that the commentary is probably by a later hand. If he is right, the count of Shankara’s own Upanishad commentaries drops, and one of the most influential of all the texts comes to us framed by a voice we cannot quite name. The matter is not settled. It is the kind of thing the tradition’s confident lists tend to paper over and the historian is obliged to leave open.

How they hang on the Veda

Every principal Upanishad belongs to a Veda. It is not free-floating. It hangs off a specific branch of a specific Veda, embedded in or attached to that branch’s Brahmana or Aranyaka, which is why the first chapter of this volume insisted on the four-layer structure. The Upanishad is the last layer of a particular school’s particular Veda.

The map runs roughly so. To the Rigveda belong the Aitareya and the Kaushitaki, sitting in the Aranyakas of their schools. To the Sama Veda, the singers’ Veda, belong the Chandogya, attached to its Brahmana, and the Kena. To the White Yajurveda belongs the Brihadaranyaka, which is literally the closing portion of the Shatapatha Brahmana. To the Black Yajurveda belong the Taittiriya, the Katha, and the Shvetashvatara, along with the late Maitri. And to the Atharva Veda, the fourth and last to be recognized, belong the three short late texts, the Mundaka, the Prashna, and the Mandukya.

There is one striking exception, and it is worth its own sentence. The Isha Upanishad is not embedded in a Brahmana or Aranyaka at all. It forms the fortieth and final chapter of the Vajasaneyi Samhita, the hymn-collection itself, of the White Yajurveda. It is the only principal Upanishad that sits inside a Samhita, in the very oldest layer, the hymns. A text of pure knowledge lodged in the book of ritual verses. The generalization that the Upanishads are always the last layer holds almost everywhere. The Isha is the almost.

Early, middle, late

If the Upanishads cannot be dated by their authors, who are anonymous, they can be ordered relative to one another by their language and their form, and this is where modern scholarship has done its most careful work.

The classification goes back to Paul Deussen, who in his 1906 study sorted the principal Upanishads into stages, and it was refined by Patrick Olivelle in his 1998 edition, now the standard. The scheme is three-fold. First, the oldest layer, in archaic prose: the Brihadaranyaka and the Chandogya, the two giants, the deepest and the least systematic, broadly held to predate the Buddha. Then the early prose texts close behind them, the Taittiriya, Aitareya, and Kaushitaki. Second, a middle layer composed in verse: the Kena, Katha, Isha, Shvetashvatara, and Mundaka, more polished, more doctrinally pointed, later. And third, a late layer that returns to prose: the Prashna, the Mandukya, and the Maitri, the Mandukya probably the latest of the principal group, and the most tightly argued.

This is a relative order, not a set of dates, and the scholars who built it are the first to say so. Olivelle’s own caution is worth quoting, because it is the proper attitude toward the whole question. As he put it, “any dating of these documents that attempts a precision closer than a few centuries is as stable as a house of cards.” The next chapter and the next will keep returning to that humility, because it is the honest ground under everything that can be said about when these texts were made.

What can be said is the shape. A small group of genuinely ancient texts, the oldest in sprawling prose and the later ones tightening into verse and then into compact prose argument, each hanging off a branch of one of the four Vedas, with one stubborn exception lodged in the hymns themselves. Around that core, over many centuries, a widening cloud of younger texts borrowing the name, until a late hand counted out an auspicious hundred and eight and called the canon closed.

It was never quite closed. But the principal few were enough to carry everything that came after. How those texts actually taught, by what methods they drove their impossible claim into a listener, is the subject of the next chapter.