Where the water enters

The four witnesses

The legend of the four who entered the pardes -- Ben Azzai, Ben Zoma, Aher, and Rabbi Akiva -- survives in four layers, each translated here under the same controlled vocabulary and the same merkavah-ascent frame, so the differences are textual and not artifacts of translation. Read across the row for "water, water" and the story tells itself.

1. The warning is a Babylonian addition

The marble / "water, water" warning -- the single most-quoted sentence of the whole tradition -- is absent from the two earliest witnesses. The Tosefta (tannaitic) and the Yerushalmi (Palestinian) both tell the story of the four who entered, name the fate of each, and adduce their proof-texts -- but neither has Akiva warn anyone about marble or water. The warning appears in the Babylonian Talmud. Two independent pre-Babylonian witnesses lacking it is the kind of evidence that lets a reader see the warning as a development rather than as the kernel of the story. (Each translation confirms the absence mechanically: the Tosefta and Yerushalmi pericopes contain pardes and the four sages but neither mayim mayim nor avnei shayish.)

2. The water was cosmological before it was optical

Where the earlier witnesses do speak of water, it is not the shimmer of marble but the waters of creation. Both the Tosefta and the Yerushalmi attach to the four-entered legend the episode of Ben Zoma gazing on maaseh bereshit and declaring that there is "but a handsbreadth" between the upper and the lower waters of Genesis 1 -- argued, in the Yerushalmi, by a gezerah shavah on the word "hover" (the spirit of God hovering over the waters; the eagle hovering over its young, "touching and not touching"). This cosmological water -- the upper and lower waters of the firmament -- is precisely the substrate David Halperin reads behind the later optical warning. The Bavli's "do not say water, water" takes a word already loaded with the cosmology of creation and turns it toward an optical test at the palace gate. The Targum renders the same controlled term, mayim, as "waters" in the early witnesses and "water" in the Bavli, and preserves the register-shift as a flagged ambiguity rather than smoothing it.

3. The Yerushalmi swaps the fates

The sharpest single divergence is in the Palestinian telling. In the Bavli and the Tosefta, Ben Azzai gazed and died (and receives Psalm 116:15, "Precious in the eyes of the Lord is the death of his devoted ones"), while Ben Zoma gazed and was stricken (and receives Proverbs 25:16, the warning against a surfeit of honey). The Yerushalmi reverses it: Ben Azzai is the one stricken, Ben Zoma the one who dies. The proof-texts travel with the fate -- death keeps the precious-death verse, the wound keeps the honey-surfeit verse -- not with the named sage. It is a clean illustration of how a tradition can hold its exegetical furniture fixed while rotating the figures through it.

4. The Hekhalot recasts the danger as a question

In the Hekhalot literature proper -- Heikhalot Rabbati 26 -- the water danger returns in a third, distinct form. The gate-guardians of the sixth palace "act as though casting upon him thousands of waves of water, yet there is not a single drop there," and the descender who fails is the one who questions the illusion -- "these waters, what is their nature?" -- run down with stoning, rebuked as a descendant of the Calf-worshippers, and pinned with iron bars. Where the Bavli's danger is mis-naming the marble ("do not say 'water, water'"), the Hekhalot's is questioning a guardian-cast illusion. (The Munich manuscript Scholem made famous combines the two -- the marble plates and the question.) So the danger is not merely added in Babylonia; once present, it is realized in more than one shape. This is the witness the cluster had flagged as source-blocked -- and it turned out to be in the same public-domain Wertheimer recension as Heikhalot Rabbati 1, 19, and 24. The earlier "no source" verdict was an artifact of searching for the Bavli's words (marble, the doubled water, water) rather than the Hekhalot's (waves of water, "those who kissed the Calf").

One safe passage, three idioms

A smaller detail rewards the side-by-side reading: how each witness describes the one who came through unharmed. The Tosefta has Akiva "ascended in peace and descended in peace"; the Yerushalmi, "entered in peace and went out in peace"; the Bavli, simply "departed in peace." The ascent-and-descent language of the earlier witnesses is the verbal seed of the later Heikhalot paradox of the yored merkavah -- the one who "descends" to the chariot by ascending through the palaces.

On method

This is what the synopsis is for, and why no one of these passages, translated alone, would be much of a contribution: the bare Bavli warning is in every handbook. The contribution is the cross-witness comparison under a single controlled vocabulary -- the same rendering rules applied to all four layers so the differences are the text's, not the translator's. All four witnesses are translated from public-domain sources (Tosefta: Machon-Mamre; Yerushalmi: Venice 1523; Bavli: Vilna; Heikhalot Rabbati: Wertheimer) and were extracted verbatim, not retyped. Each translation carries its own apparatus, drift audit, registry check, and verified-citation manifest; the citations are shared across the cluster, so later witnesses inherit the verification of the first. The fourth witness had been flagged as source-blocked -- Scholem's famous version is from a manuscript, and the critical printings are copyrighted -- but the Hekhalot form of the danger proved to be in the same public-domain Wertheimer recension all along; the arc is complete from clean sources. The fuller Munich-manuscript recension (which adds the marble pavement) remains a copyrighted comparand, characterized in the apparatus rather than reproduced.

Read the witnesses: Tosefta · Yerushalmi · Bavli · Heikhalot Rabbati.