guardians of the threshold, eight: four to the right of the lintel and four to the left
The chapter is a gate-guardian inventory, and it is the structural spine of the ascent narrative rather than an ornamental list: passage through each palace depends on satisfying the guardians posted there, and the practitioner who knows the names and the seals advances while the one who does not is destroyed. The right/left symmetry around the lintel (המשקוף) gives the gate an architectural specificity that the ascent chapters then assume. Lesses reads the adjurations and angelic encounters of this literature as a coherent ritual system in which power over angelic beings is obtained by specifiable means, which is the register in which this catalogue should be read: it is operative inventory, not description. Frame relation: angelic-seal-passage (merkavah-ascent). Lesses, Rebecca Macy 1998; Schäfer, Peter 1992
the third palace … the fifth
The chapter states eight guardians per gate and then transmits seven names for the third palace and nine for the fifth. The lists are printed as transmitted and not emended. This is the characteristic behaviour of nomina barbara in transmission: because the names carry no lexical sense, a scribe has nothing to correct against, and the divergence between witnesses is widest precisely in this material — which is why Schäfer’s Synopse prints witnesses in parallel rather than constructing a critical text. Treated as ambiguities[amb-hr17-guardian-count]. Frame relation: stam-anonymity (peshat-derash). Schäfer, Peter 1981
ANAFI’EL
Anafiel appears here as the eighth and last guardian of the fourth palace. Elsewhere in Heikhalot Rabbati he holds a far higher office, opening the door of the seventh palace to the descender and standing above the gate-guardians in the hierarchy. The two placements are not reconciled, and they should not be reconciled by an editor: the divergence is one of the clearer indications that the macroform aggregates microforms of different provenance, exactly as Schäfer’s macroform/microform analysis predicts. Frame relation: hidden (pardes). Schäfer, Peter 1981; Schäfer, Peter 1992
TOTROSIAI, YHWH God of Israel
The chapter opens by naming the enthroned one with a nomen barbarum bound to the Tetragrammaton and the standard divine title. This formula, and the seventy-odd angelic names that follow, are transliterated and never translated: they carry no lexical sense to render, they vary widely between witnesses, and their function is recitative rather than descriptive. Where a name contains a transparent element (ANAFI’EL from anaf, GEVURATI’EL from gevurah, SHEKHINI’EL from the shekhinah root), the etymology is left in apparatus and never substituted into the body, since a name rendered by its etymology has ceased to be the name the rite requires. Frame relation: theurgic-recitation (merkavah-ascent). Schäfer, Peter 1981; Swartz, Michael D. 1996; Janowitz, Naomi 1989
all the mighty ones: fierce, strong and rigid
The seventh-palace gate breaks the pattern of the preceding six. Where each earlier palace is given a fixed roster of eight named guardians, the seventh receives no names at all and instead a sustained description of massed armed figures — bows drawn, swords sharpened, lightning from the eyes, fire from the nostrils, torches of coals from the mouth. The withholding of names at the last gate is the chapter’s rhetorical climax and probably its point: at the six lower gates the practitioner passes by knowing what to say, and at the seventh the text declines to tell him. Frame relation: ascent (pardes). Schäfer, Peter 2009; Davila, James R. 2001
at the entrance of each and every palace
CROSS-TRADITION NOTE, classified (S) structural parallel, not (T) documented transmission. The architecture of this chapter — a graded series of heavens or palaces, each with angelic gatekeepers who must be satisfied before the traveller may proceed, terminating in an approach to the divine throne — is realised independently in the Islamic ascension literature, the isra’ and mi’raj narratives, whose developed forms are studied critically by Colby and in the cross-cultural volume edited by Gruber and Colby. The parallel is structural and is recorded as such: the developed mi’raj narratives postdate the period in which the Hekhalot microforms took shape, the two literatures were transmitted in different languages and institutional settings, and the scholarship cited establishes the comparison as a field of inquiry rather than a line of dependence in either direction. No claim of influence is made here, and the theological differences are not flattened by the resemblance — the Hekhalot ascent is undertaken by a practitioner who may be destroyed if unworthy and terminates in the kavod as a manifestation-aspect, which is not the framing of the Prophet’s singular ascension. Frame relation: exegetical-ezekiel (merkavah-ascent). Colby, Frederick S. 2008; Gruber, Christiane, and Frederick S. Colby (eds.) 2010
four to the right of the lintel and four to the left of the lintel
The guardians are positioned relative to the המשקוף, the lintel — the horizontal beam ABOVE a doorway. Guards standing to the right and left of a lintel is architecturally odd: one expects them flanking the doorposts (מזוזות) or the entrance itself. The reading is nonetheless what the witness gives, and it is not emended here. Two construals are available and neither is decisive: the phrase may be loose for ‘either side of the doorway’, or it may locate the guardians at the two ENDS of the lintel beam, which would make the gate’s guardian-array a horizontal span above the opening rather than a pair of flanking files. The distinction matters for anyone reconstructing the visionary architecture, and it is recorded here rather than silently resolved. Note also that this chapter distinguishes סף (threshold, §1) from פתח (gate), and both are rendered consistently throughout this edition. Frame relation: exegetical-ezekiel (merkavah-ascent). Schäfer, Peter 1992
ANAFI’EL (the other pole, supplied by ch. 23)
CROSS-CHAPTER NOTE added 2026-07-26, after ch. 23 was translated. This chapter places ANAFI’EL as the eighth and last guardian of the FOURTH palace. Chapter 23 supplies the other pole in full: there he closes the seventh-gate ascent roster, holds the signet of the seal of heaven and earth, receives prostration that even the Prince of the Presence does not, appears in the descent roster as well, and is beloved above all other guardians for one stated reason, that he opens the gate of the seventh palace. Chapter 23 also derives his name, from the branch (anaf) of the crown of crowns on his head, citing Habakkuk 3:3. Neither chapter is emended. The divergence stands as a fact about this recension, and on Schäfer’s macroform analysis it is the expected consequence of assembly from independent microforms; what the complete sequence adds is that both poles sit six chapters apart in ONE transmitted work. Frame relation: hidden (pardes). Schäfer, Peter 1981; Schäfer, Peter 2009
Editorial decisions
- cross-tradition edge classification. The single cross-tradition reference in this output is classified (S) STRUCTURAL PARALLEL, never (T) documented transmission, and its target is marked proposed because no Rosetta map for the graded-gatekeeper ascent yet exists. Red line #1 forbids asserting a cross-tradition resonance without cited comparative scholarship; the citation here is Colby 2008 and Gruber & Colby 2010, both historical-critical treatments of the mi’raj literature, deliberately not comparative-mysticism or perennialist works. No direction of influence is claimed. This is the FIRST cross-tradition edge authored from HR material, and it was authorable only because the comparative seed unit was written first; the HR ch 8 pilot correctly emitted zero, because nothing in that chapter’s martyrological material has an adjacency this project can currently source. (Hekhal Targum Project, 2026-07-25)
- training-memory citations. training-memory-citations-disclosure: every work cited in this apparatus (Schäfer 1981/1992/2009, Lesses 1998, Swartz 1996, Janowitz 1989, Davila 2001, Colby 2008, Gruber & Colby 2010) is real and was web-verified for author, title, publisher and year during the 2026-07-25 seed-authoring pass. NO page loci are given, deliberately: page references have not been checked against the printed volumes, and asserting one from memory is the riskiest act available to this pipeline. Citations remain training-memory pending editor verification; status stays machine-assisted. (Hekhal Targum Project, 2026-07-25)
- scripture pre-flight. Gate 2 parsed ZERO scriptural citations and therefore verified nothing. Hand-confirmed: this chapter contains no she-ne’emar, di-khtiv, ka-katuv or parenthetical verse reference — it is a name-inventory with no scriptural citation at all. Recorded so a green gate is not mistaken for verification. Second HR chapter in a row with this property, which suggests the scripture gate will provide little coverage across this corpus and that the nomina-barbara character-level check matters correspondingly more. (Hekhal Targum Project, 2026-07-25)
- name transliteration scheme. Angelic names are transliterated on a single consistent scheme across the edition and rendered in CAPITALS to mark them as operative names rather than words: consonantal skeleton with the gershayim-marked -aleph-lamed termination rendered -I’EL. Where the witness’s vocalisation is uncertain the transliteration follows the consonants and does not invent a vowel pattern. No name is translated by its etymology in the body, including the transparent ones (ANAFI’EL, GEVURATI’EL, SHEKHINI’EL, PAHADI’EL). Per the nomina-barbara discipline, this material requires character-level verification against the source at editor sign-off, because a plausible-looking regularised name is indistinguishable from a correct one to every automated check in this pipeline, including an independent blind re-translation check, which would regularise in the same direction. (Hekhal Targum Project, 2026-07-25)
- drift: ‘chamber’ scope-violation. ACCEPTED on the record, not fixed. The drift scanner reports ‘chamber’ as a forbidden rendering for heikhal twice; both hits render חדר (cheder) in the phrase חדר בתוך חדר, not היכל. The glossary’s intent is fully honoured in this text: heikhal is rendered ‘palace’ at all seven of its occurrences and nowhere as ‘chamber’. Changing a correct rendering of חדר in order to satisfy a matcher that cannot see which source word produced the English would be letting the instrument dictate the scholarship. ENGINE-ISSUES Issue #1 (drift is not term-aware) remains the root cause; this is its third recurrence on this corpus in one session. (Hekhal Targum Project, 2026-07-25)
- name correspondence verification. Ran a deterministic block-by-block name-correspondence check: 8 source blocks against 8 English blocks, 50 gershayim-marked names in the source against 50 transliterated names in the English, matching 1:1 in every block. This is the only automated defence available for nomina barbara, because a silently regularised name is invisible to schema, drift AND an independent blind re-translation check (which would regularise in the same direction and thereby confirm the error). It verifies COUNT and POSITION, never spelling: character-level verification of each of the 50 names against the witness remains an editor sign-off item. The check was generalised into backend/harvest.py::name_correspondence and wired into the yield gate the same session, so every future edition inherits it. (Hekhal Targum Project, 2026-07-25)
- consistency of פתח. The independent blind re-translation caught a genuine inconsistency: פתח was rendered ‘entrance’ in §1 and ‘gate’ in §§2-7 and §8, three Englishes for one source word, which the auditor read as a deliberate distinction and tried to preserve. Corrected to ‘gate’ throughout. The source’s real distinction, סף (threshold) in §1 against פתח (gate) elsewhere, IS preserved. A reader must be able to tell which of two nouns a text used. (Hekhal Targum Project, 2026-07-26)
Manuscript variants
- internal variant recorded by the text itself (ואית דאמרי): שיביא”ל for ושויא”ל, first palace — A name variant transmitted inside the source rather than reconstructed by collation, using the same Aramaic formula that marks the divine-name variant at HR ch 8. Both forms carried into the translation.
- Wertheimer, Batei Midrashot recension (digital transcription via Sefaria): third-palace list transmits seven names, fifth-palace list transmits nine, against the chapter’s own rubric of eight per gate — Not emended. Counted directly from the witness. The expected transmissional behaviour for names carrying no semantic constraint.