The specifications for the sanctuary in Exodus read like a contract drafted by someone who expected it to be audited. They give the length of the Ark in cubits and half cubits, the weight of the gold and the silver in talents and shekels, the colour of the yarn, the number of screens, the weight of the half-shekel that every man counted out, and the names of the two men who were to do the work, and they name the book in which the accounts were to be kept. This essay treats that document not as a blueprint for a building that may or may not once have stood in the wilderness, but as a text: ancient specification writing with a voice, a transmission history, and a long afterlife in manuscripts, in law, in legend and in memory. What follows is a close reading of Exodus 25 to 27 and 30 to 40, followed by an account of what the rabbis, the masoretes, the Septuagint, the Samaritan Pentateuch, the Qumran scribes and the Ethiopian church made of it, none of whom were willing to let a box stay a box.
The reason to begin with the specification rather than with the object is that the specification is the larger and stranger thing. The Ark, as later tradition understood it, is a chest with a lid and two winged figures on top; it is portable, precious and famous. The specification is a whole apparatus: a tent, a court, hangings, clasps, sockets, an altar, a laver, a lampstand of beaten almond blossom, curtains, screens, a priestly wardrobe of coloured stones, an anointing oil with a fixed recipe and a price attached to copying it. It is roughly two thirds of the book of Exodus. And it is written in a register we recognise from builders rather than from theologians: imperative mood, repeated measurements, materials enumerated in a fixed order, and a repeated formula of compliance that closes each phase of the work like a stamp on a delivery note. To read Exodus 25 to 40 as a document is to take the specification seriously, and then to ask the question the specification itself provokes: if this is what the divine voice wanted, what did the tradition do with the wanting?
The voice that measured
Open the account at its beginning and the first thing the sanctuary costs is money. The LORD speaks to Moses and Moses is told to speak to the children of Israel and to take from them a contribution, a terumah, from every person whose heart is so moved (Exodus 25:1 to 2 in the Masoretic numbering, which is one verse ahead of most English editions for the rest of this passage). The list of what is to be accepted follows immediately: gold, silver and copper; the dyed yarns, tekhelet, argaman and tola’at shani; fine linen, shesh; goats’ hair, ez; ram skins dyed red and the hide of the tekhash, whatever marine animal that is; acacia wood, shittim; oil for the light; spices; and lapis lazuli and other stones for the ephod and the breastpiece (25:3 to 7, repeated almost word for word at 35:5 to 9, where the materials are restated as a shopping list after the collection has begun). A building specification that opens with a materials requisition and closes with a census is unusual in the ancient world, and the opening is the more surprising of the two: the first divine instruction about the sanctuary is a fundraising instruction.
After the money comes the measurement. The tone is consistent and it is the tone of the joiner. The Ark is two and a half cubits long, a cubit and a half wide, a cubit and a half high (25:10). The inner sanctuary is ten cubits long, ten wide and ten high, with a veil dividing off the Holy of Holies, which is nine cubits by nine (26:1 to 6). The curtains of the tent are thirty cubits long and the frames that carry them are ten cubits, with a cubit overlapped at each joint so that the seams do not open (26:7). The screen at the entrance is twenty cubits by five (26:15). The court is a hundred cubits on each side (27:18). The couch of the lampstand is a cubit and a half, with arms of a cubit, and the whole thing stands eighteen cubits high (25:32, 37:18). Nothing in these passages explains why. The text gives the dimensions with the flat competence of a man writing down what has to be built, and it gives them again when the building is finished, so that the reader can check the two lists against each other (36:32 to 37:16 against 25:10 to 27:21). That double entry is itself an argument about the status of the document: this is a specification that assumes the possibility of error and provides for its detection.
The word for the blueprint is tabnit, pattern, and the sentence that introduces it is worth quoting closely, since it describes the transmission of the whole of the material that follows: and I will show you the pattern of the mishkan, the tent, and the pattern of all its furnishings, and so you shall make it (25:9). There is a voice, a showing, a recipient and a maker. The pattern is shown to Moses in a scene the book does not narrate, at a moment the book does not date, on a mountain the book elsewhere names only in retrospect. From Moses the pattern passes to Bezalel; from Bezalel to a company of craftsmen; from the craftsmen back to Moses for inspection; from Moses out to the people, who are told that the work was done as the LORD commanded (36:21, 38:21, 40:14, 40:27). Every link in that chain is named, and the naming is not decoration. A description would have told us what the sanctuary looked like. A specification tells us who is responsible at each stage, and it makes responsibility checkable: the tablets are the witnesses that the work was done as commanded. The refrain is the formula of a clerk closing a file.
Read the materials list as a document rather than as a catalogue and it becomes a portrait of the society that could pay. Gold, silver and copper imply a metal economy with a surplus and a way of weighing it. The three dyed yarns imply dyers working with the expensive molluscs and cochineal of the eastern Mediterranean, and a supply chain that reaches beyond the camp. Shesh and ez imply weaving at two levels of fineness, the second of them coarse and strong, suitable for curtains and coverings. Ram skins and tekhash imply tanners, and a liturgy that requires skins. Shittim implies carpenters and a supply of acacia that has to be found in the desert and brought to the builders. Spices imply a perfumer and a world of long-distance trade, since myrrh and cassia do not grow in the Sinai. And the stones for the ephod imply a lapidary with access to a lapis lazuli that came from afar. The specification is therefore not only a building order. It is an inventory of skills, and the congregation that appears at the end of Exodus 39 turns out to have had all of them.
Then the purpose, stated once and never repeated: and let them make me a sanctuary that I may dwell among them (25:8). After that sentence the book never again says why the tent exists. Everything else is logistics. The tent is described with the vocabulary of a house, and its furnishings are its rooms. But the house is a tent, and the tent is pitched in the middle of a camp whose geometry is itself religious (Numbers 1 and 2 assign the tribes to fixed positions around the sanctuary, with the Levites camped around it and the remaining three encampments set out beyond them). This is a dwelling that travels. The form is the argument: a people who are not going anywhere permanent build a house that can be struck, and the specification proceeds on that assumption, giving every board, socket and peg a design that can be taken down and set up again. When the rabbis later call the sanctuary the mishkan and the sanctuary in Jerusalem the hevel, they are preserving a distinction the text itself makes, between the tent and the house, between the wandering and the settled.
One more thing is easy to miss and hard to leave out: the tent that Exodus 25 to 27 specifies was not the first tent. Earlier in the same book, at the moment when Moses is refused entry into the promised land and asks to see God, the LORD answers by building: a tent is pitched outside the camp, at a distance from the camp, and it is called the tent of meeting; whoever sought God would go out to it; when Moses entered, the pillar of cloud descended and stood at the entrance, and the LORD spoke with Moses face to face, as one person speaks to another (Exodus 33:7 to 11). Here the whole community stands at the entrance of its own tent and gazes after Moses as he goes in, and no one but Joshua enters behind him. The sanctuary account is thus, in part, the institutionalisation of an older and less permanent practice, and the prophets and psalmists pick up the vocabulary of this first tent when they speak of God as a host who leads, shelters and feasts. What Exodus 25 to 40 describes is a house with a budget, a site, a priestly staff and an architectural plan, built to replace a tent pitched outside the camp whenever the mood took it.
Against all of this, the specification is silent about a great deal. It gives no date and no place name for the encampment. It gives no price for labour, and no number of workmen; the accounts that follow record metal and cloth, not hands. It does not say how the pattern was shown to Moses, or whether anyone checked it against the copy that Bezalel made, though the repeated formula of compliance implies that somebody did. It does not mention a source for the acacia, or a smith for the gold, or a weaver for the hangings. Read as an archive, these silences are as informative as the measurements, and they are precisely the points on which modern scholarship divides, and divides again. Some read the wilderness account as a literary or compositional invention, a sanctuary retrojected into an earlier period, and the elaborate dimensioning as the product of a scribal school that loved to specify. Others defend the antiquity of the cultic furniture, pointing to the linguistic and material parallels with New Kingdom Egyptian chests and to the difficulty of the theory. Between those positions lies a middle course, on which the account is neither a construction report nor an invention but a tradition of instruction, written down in a form that could be checked, and checked, and written down again. The form of the document survives the question of its historicity. What we have is a specification, and specifications are preserved because people keep using them.
Two names and a company of workmen
Bezalel is named six times in the book of Exodus: in the call (31:2), in the second call (35:30), in the assignment of the work (36:1), in the summons (36:2), at the making of the Ark (37:1) and in the closing account of the whole enterprise (38:22). Oholiab is named five times: at the call (31:6), in the charge to teach (35:34), in the assignment (36:1), in the summons (36:2) and in the closing account (38:23). These are remarkably few occurrences for the two men who run the entire operation, and the figure is more interesting when it is set beside the alternative that the text keeps offering. For most of the narrative the workers are not two men but a category: anyone with chokhmat lev, skill of heart, a phrase the Septuagint renders with a phrase about understanding and the English versions variously make skilled, wise-hearted, skillful or expert. The Hebrew treats skill as a shared substance that can be poured into many vessels. The two named men are the exceptional cases, and the text marks them as exceptional by naming them.
The names are programmes, and their meanings have been argued over for centuries. Bezalel is usually read as in the shadow of God, and the midrashic tradition takes it further, glossing the name as the place where the heavenly pattern was shown, so that the man himself becomes a location of revelation. Oholiab is read as the tent of my father, though the etymology is contested and some have preferred the tent of my father’s ark; the reading is disputed because the point of the name is architectural, and whether it points to a dwelling, a shrine or a chest, it points to a building rather than to a skill. The genealogies are given twice and are not neutral. Bezalel is the son of Uri, the son of Hur, of the tribe of Judah; Hur is the man who, in the rabbinic account, was killed for refusing to take part in the golden calf, and Bezalel is thus the grandson of a martyr and, according to the midrashic genealogy, a great-nephew of Moses himself through Miriam (Exodus Rabbah 48:3 to 4). Oholiab is the son of Ahisamach, of the tribe of Dan. One of the two men is the grandson of a man who died for refusing idolatry; the other has no ancestry at all in the text. The specification does not comment on either fact.
The pairing of Judah and Dan deserves more attention than it usually receives. Dan is the tribe of the serpent, the tribe that bites and heels, and in the camp it marches last of all (Genesis 49:16 to 17; Numbers 2:29 to 31), camped at the rear of the procession. Judah is the lion and marches first, camped on the east. The sanctuary stands between the two, and its chief artisans are drawn from the two extremes of the marching order: the first tribe and the last. Read this way, the specification makes a quiet argument about who the sanctuary is for. It is not the property of the powerful end of the camp. The house that stands in the middle is built by the rear of the column as well as the front, and it is built for the rearmost tribe as much as for the royal one. That both men are needed, and that they cannot work apart, is also insisted on: the spirit is given to both, and the skill is distributed, with Oholiab granted the heart of Bezalel and the heart of every one who is wise-hearted (31:3 to 6).
The formulation of the commission is the hinge of the whole section, and it is easy to skim. I have filled him with the spirit of God, with skill in all manner of work, to work in gold, in silver, in copper, in cutting of stones, in engraving and in embroidery, and to devise works of art, to work in all manner of precious stone and all manner of timber, and to teach (31:3 to 5; 35:30 to 35, where the second call repeats the inventory almost verbatim). Two things are happening. First, the endowment is vocational: the spirit arrives as a competence, itemised by material. Second, and more odd, the commission includes teaching. Bezalel is to devise and to teach; the work is to be reproducible. A god who wanted one perfect object would not have thought to require an apprenticeship clause, and the fact that the specification includes one is part of why the book has been read as a constitution rather than as a report. It is a document that thinks about transmission.
The second body of workmen arrives in chapter 36, and the shift of register is sharp. Bezalel and Oholiab are named, but they are not the workforce. The gifts keep coming, morning after morning, so that the people bring more than is needed (36:3 to 5); Moses halts the collection with a proclamation through the camp that no man and no woman is to make further effort for the offering of the sanctuary (36:6), and the text’s verdict is that there was more than enough (36:7). Then the craftsmen are described arriving each from his own task (36:4), which is a working description of a workshop with specialisations, and the sentence that follows shifts from the named to the anonymous: all the skilled among those engaged in the work made the tent of ten strips of fine twisted linen, worked in blue, purple and crimson, into which they worked a design of cherubim (36:8). The chief artisans are therefore supervisors of a much larger body, and the specification, characteristically, does not tell us how large it was.
That is worth stating plainly, because the section of Exodus usually described as the census of the craftsmen contains no such census. The count that does occur is a count of contributions, in chapter 38. What chapter 36 does contain, and what is genuinely strange, is the way the women are counted: not by number, but by labour and by their mirrors. Men and women whose hearts moved them bring gold ornaments (35:22). All the skilled women spun with their own hands and brought what they had spun, in the dyed yarns and in fine linen (35:25), and all the women who excelled in that skill spun the goats’ hair (35:26). The bronze laver in the court is made from the mirrors of the women who served at the entrance of the tent of meeting (38:8). The mirrors are the one mass-produced item in the whole account that is not ordered by the specification, and they are the only object in the sanctuary whose maker is described by a category rather than a craft. In a text otherwise absorbed with cubits and shekels, the women’s spinning is the labour that is praised without being measured, and the bronze basin is what the community ends up with: a public object made out of the private faces of the women who did the hardest work of all.
The rabbis, reading the same material, at once begin to supply the human machinery that the specification leaves out. The Bavli takes the appointment of Bezalel as the proof text for a principle of consent: when God chose Bezalel, the narrative has God ask Moses first whether the choice is acceptable in his eyes, Moses replies that if it is acceptable in God’s eyes it is acceptable in his, and God then instructs him to put the question to Israel as well, and the people agree (Berakhot 55a, citing Exodus 31:1 to 2 and 35:30). A divine appointment is thus made to pass through a human consultation, and the consent of the community becomes a condition of legitimate leadership. Alongside that, the tradition wants the skill to be explicable, and it credits Bezalel with the letters of the alphabet with which heaven and earth were created, so that his endowment is described in the vocabulary of creation itself rather than of craft. The midrash goes further and has Moses marvel at the speed with which Bezalel grasped the pattern of the lampstand, which was so complicated that Moses himself could not take it in, and conclude that the man must have been in the shadow of God when the heavenly model was shown (Exodus Rabbah 34:1; Numbers Rabbah 12:3 and 15:10). And Sanhedrin 69b does the arithmetic that the specification itself refuses to do: Caleb was forty when he went to spy out the land, Bezalel was at least fourteen at that time, the spying took place a year after the tent was erected, and three generations, Hur, Uri and Bezalel, fit into twenty-four years with each parent begetting a child at the age of eight. Bezalel was therefore a teenager when he made the Ark.
So the rabbinic argument about who taught Bezalel his craft ends where the modern reader would expect it to end, but by a different route. The rabbis do not decide between a natural gift and a learned one. They hold both, and they are careful about which is which. The spirit is given, and the community consents; the man is a teenager, and a teenager has a lineage and a teacher and a curriculum; and the specification’s insistence on teaching is not decoration but the point at which the divine endowment becomes a tradition rather than a spectacle. That is the shape of the material as it enters the Masoretic tradition, and it is worth holding in mind, because the Ark, which is the most famous object in the specification, is the one item in it that Bezalel is explicitly said to have made with his own hands (37:1). The specifications are long, the materials are inventoried, the accounts are kept and the book of records is named, but the chest itself has a maker, and the maker has a name, and the name is used six times.
The materials and the ledger
Chapter 38 turns the specification into an account, and it does so with a title. These are the records of the tabernacle, the tabernacle of the pact, that were drawn up at Moses’ bidding, the work of the Levites under the direction of Ithamar son of Aaron the priest (38:21). The noun is pqud, a census, a count, an inventory; the noun for the book is sefer, a scroll or book. The specification therefore ends by naming its own archive: there was a book of the tent, in which the figures were entered. The book does not survive, and the account stands in its place, which is a neat summary of the whole genre. We are reading the transcript of a ledger whose original has been lost.
The figures themselves are startlingly precise, and they are precise in a particular way: they are the figures of a community that has recently counted itself. The census is taken over the males from twenty years upward, and the number recorded is 603,550 (38:26). Every counted man contributes a beqa, a half-shekel by the sanctuary weight (38:26), and the silver so gathered comes to a hundred talents and 1,775 shekels (38:25). The arithmetic is checkable, which is the point: at three thousand shekels to the talent, 603,550 half-shekels amount to 301,775 shekels, which is exactly a hundred talents and 1,775 shekels. The gold that went into the work is entered at twenty-nine talents and 730 shekels (38:24), the bronze of the offering at seventy talents and 2,400 shekels (38:29). The silver is then broken down into what it actually bought: a hundred sockets for the frames and the screen, one talent apiece, and 1,775 shekels of hooks, overlays and bands for the posts (38:27 to 28). The bronze is broken down into the sockets of the entrance, the altar and its grating and all its vessels, the sockets of the court and of its gate, and every peg of the tent and of the court (38:30 to 31). The ledger closes with the smallest objects in the building.
Economically, the picture is of a society paying for its own sanctuary with money it had, plus wealth it had taken. The gold and the bronze of Exodus 12:35 come out of Egypt in the wake of the death of the firstborn, and the silver is the people’s own money contributed as a poll tax of half a shekel. The text does not distinguish between the two sources of the metal, and the ark of the covenant is thus, among other things, a monument to the exit from Egypt, or at least a place where Egyptian wealth was put to a different use. The census money is also the money of registration. A people that has just been numbered, camped and marched has learned how to count itself, and the sanctuary’s budget is built on that skill.
It is worth pausing over how exactly the specification knows what it knows. It gives the sanctuary weight of the shekel and the talent, the length and width of every textile, the number of strips and how they are joined, the number of pillars, posts, sockets and pegs, the proportions of the anointing oil (five hundred shekels of myrrh, two hundred and fifty of cinnamon, two hundred and fifty of fragrant cane, five hundred of cassia, and a hin of olive oil, 30:23 to 24) and of the incense (stacte, onycha and galbanum in equal parts with pure frankincense, 30:34), the colours of the yarns, the number of the census, the names of the two chief artisans, and the names of the two sons of Aaron who blow the trumpets. It is silent about almost everything else. It does not give the number of workers or the wages they received. It does not give the yardage of the cloth, the length of the timber, the price of the dyes or the source of the acacia. It does not say how the pattern was shown to Moses, how long the work took, or what it looked like from outside. Silence about labour and precision about material is the signature of a text that is interested in the offering and in the equipment, and not much in the builders.
Two further features of the ledger deserve attention. The first is the surplus. When the artisans report that the people are bringing more than the work requires, Moses stops the collection by proclamation and declares that there was more than enough (36:5 to 7). The economy of the sanctuary is thus a closed economy in which the only risk recognised is excess. The second is secrecy, and it is a striking one: two of the documents in the specification are protected knowledge. No one is to compound anything like the anointing oil in the same proportions, and no one is to compound anything like the incense in the same proportions; anyone who does is to be cut off from his people (30:32 to 33, 30:37 to 38). A building contract that contains trade secrets for a perfume is a document written by people who understood what they were specifying. The specification’s silences, in other words, are not the silences of ignorance. They are the silences of a text that knows exactly what it wants and says so in the currency of shekels.
The box and its cover
Strip the specification back and the Ark is a piece of joinery. Make an ark of acacia wood, two and a half cubits long, a cubit and a half wide and a cubit and a half high; overlay it with pure gold, inside and out, and make upon it a gold molding round about; cast four gold rings for it, to be attached to its four feet, two rings on one of its side walls and two on the other; make poles of acacia wood and overlay them with gold, and insert them into the rings on the side walls for carrying the ark (25:10 to 15). On a short cubit of roughly 45 centimetres that gives a chest about 1.1 by 0.7 by 0.7 metres, or in the older reckoning 1.3 by 0.8 by 0.8 metres. The text says nothing about its capacity, but the instruction to overlay the interior as well as the exterior tells the joiner what he needs to know: the box is hollow, built of boards, and its seams have to be tight enough to take gold leaf on the inside.
Then one instruction, which is a design decision about handling and not about theology: the poles shall remain in the rings of the ark, they shall not be removed from it (25:27). The staves are permanent. When the Ark is lifted it is lifted on its own furniture, and it never has to be re-threaded or unpacked in the middle of a camp in the desert. Everything else in the specification can be dismantled: the curtains, the frames, the sockets, the pegs. The Ark cannot. Whatever else it is, it is designed for the moment when the camp moves, and designed so that the moment requires no improvisation. This is the detail that makes the chest a piece of transport engineering, and it is the detail that the later tradition, which wanted the Ark to be a throne, kept dropping.
The cover is the most consequential noun in the book. The Hebrew calls it kapporet, from the root k-p-r, to cover or to atone; the Septuagint renders it hilasterion, a word built on the verb for propitiating, and from that Greek came the Latin propitiatorium, which is why English readers for two centuries have called the object the mercy seat or the atonement seat rather than a lid. The tradition’s own gloss on the root is instructively severe: the cover is what covers the Ark, and in the ritual of Leviticus 16 the blood is placed over the kapporet, which covers the Ark, so the noun is doing its theological work long before any commentator explains it. But there is a second, more material sense: a kapporet in ordinary Hebrew is a lid, a cover, the thing you put on top. The Greek translator chose the word for atonement; the Hebrew word was already a lid. Both readings are in the object.
The instruction is exacting: make a cover of pure gold, two and a half cubits long and a cubit and a half wide; make two cherubim of gold, of hammered work, at the two ends of the cover; make one cherub at one end and the other at the other end, of one piece with the cover shall you make the cherubim at its two ends; the cherubim shall have their wings spread out above, shielding the cover with their wings, and they shall confront each other, the faces of the cherubim being turned toward the cover (25:17 to 20). One sheet of gold, struck from the same metal as the cover itself, worked into two figures that face one another and face the lid beneath them. The technique named is that of the chaser, not the engraver: the metal is hammered from the back into relief. Two thousand years later, when Solomon builds the permanent house, he makes two cherubim of olivewood, each ten cubits high, with wings of five cubits, set in the inner chamber and overlaid with gold (1 Kings 6:23 to 28). The specification is followed to the letter; only the timber changes, because the timber has to be found.
What the box is said to hold is specified with the same economy. You shall deposit in the Ark the testimony that I will give you (25:16); at the erection Moses takes the testimony and puts it inside (40:20); and the Chronicler, looking back from the finished Temple, states flatly that there was nothing inside the Ark but the two tablets that Moses placed there at Horeb (2 Chronicles 5:10). The rabbinic inventory is longer, and the rabbis derive it by argument rather than by text: when the Ark was hidden, the jar of manna and the flask of anointing oil were hidden with it, and Aaron’s rod with its almonds and blossoms, and the coffer that the Philistines sent as a gift to the God of Israel (Yoma 52b), and the Tosefta adds that these lay in the Holy of Holies with the Ark itself. R. Elazar then makes the inference formal by drawing a verbal analogy between the words for keeping, in the passage about the jar of manna, the anointing oil and Aaron’s rod (Exodus 16:33, 30:31, Numbers 17:25). Whatever the biblical text gives, the rabbinic imagination gives more, and gives it as a set.
The rest of the inner chamber is furniture in the same register: the table of showbread with its plates, cups, spoons and jars for the portions of bread that were to be set before the LORD every week (25:29 to 30); the lampstand of hammered gold, its shaft, its branches and its almond blossoms worked in one piece (25:31 to 40), eighteen cubits high, standing with its olive oil by the south side of the tent (26:35, 27:20); the altar of incense at the entrance of the inner chamber (30:3 to 8); the altar of burnt offering in the court, with its grating and its fire (27:1 to 8); the bronze laver and its stand (30:18 to 21). And the textiles that divide the space: the first veil, of blue, purple, crimson and fine linen, to separate the Holy of Holies from the rest (26:31 to 33), and the second veil at the entrance of the tent itself (26:36 to 37). Two curtains, two chambers, one voice, and the voice speaks from above the cover, between the two cherubim that are on top of the Ark (25:22). The specification has built a room whose entire geometry is arranged so that a sound can travel from above a lid to a man standing outside it, and has written that geometry down in cubits.
The forbidden image
The intellectual difficulty of the sanctuary account is not its measurements. It is that the book which forbids the making of images commands the making of two of them, in the same voice, a few chapters apart. You shall not make for yourself a graven image, or any likeness of any thing that is in the heavens above or that is in the earth below or that is in the water under the earth; you shall not bow down to them or worship them (Exodus 20:4 to 5). And then, in the sanctuary specification, make two cherubim of gold, of hammered work, at the two ends of the cover, with wings spread out and faces turned toward one another and toward the cover (25:18 to 20). The prohibition is a prohibition of images and of bowing to them; the instruction is an instruction to make two. The rabbis, the philosophers, the illuminators and the archaeologists have all felt this as the sharpest joint in the book, and no reading of the sanctuary that ignores it is worth much.
The standard rabbinic handling is a distinction between what an image represents and what it is for. The Decalogue forbids making an image of God and bowing to it; the sanctuary’s cherubim are neither a representation of God nor an object of worship, but the furniture of the throne on which God appears to sit, and their function is to overshadow and to shelter. The Hebrew word that governs their posture is decisive if one reads it closely: their faces are turned toward each other and toward the cover, not toward the worshipper who stands in the court. Whatever else they are, they are not looking at you. The tabernacle’s own design contains its own defence. The instruction to make them is also an instruction about where they are not to look.
The tradition’s other answer is functional. A throne needs a guard, and a throne is also spoken of as having a footstool; the Ark, in the later reading, is at times taken to be the footstool of the divine seat, and the cherubim are the guardians who stand over it and overshadow it, as the text says, with their wings. To this the tradition added a further, more radical claim that became central to the mystical reading of scripture: that the Ark is the image of the image, and that the cover, on which the blood is placed once a year, is the point at which the visible world is brought into relation with the invisible. In the Zohar and its successors the Ark and the cherubim become the centre of a cosmology in which the sanctuary is a model of the structure of the divine itself, and the two figures are read as the visible form of the formlessness. The direction of travel is consistent: away from depiction and towards function. What the images are called matters less than what they do, which is to hold a space open and to keep a threshold guarded.
The medieval philosophical readings continue the same logic, and they tend to ask the same question in the same form: not what do the cherubim look like, but what is the form they signify. Read in that register, the cherubim become a diagram of the spheres, of the intellect, of the mediation between what is above and what is below, and the Ark becomes the point where the mediation is performed. The Christian tradition, meanwhile, took the same furniture in an entirely different direction, and read the Ark as a type: the ark that carried the manna in the wilderness, with its lid, its cherubim and the blood upon the lid, was read as a type of the Virgin, the vessel of God, and the patristic writers from the third and fourth centuries, Gregory Thaumaturgus, Ambrose, Ephraem and Augustine among them, treat the Ark as the prototype of Mary. The Roman catechism of the twentieth century develops the same typology at length, taking the Ark as the prototype of the Ark of the New Covenant. The object that Judaism inherited as a chest of acacia and gold was, within a few centuries, the most quoted image in Christendom.
The history of the prohibition itself has to be handled with some care, and recent scholarship has made it more complicated than the older consensus. The traditional account held that Judaism was aniconic in a broad sense, a religion without images; the study of Jewish art and of the Second Temple material has made that untenable as a general description. Jews were not under a blanket ban on visual art, and the ancient ban is better understood as a prohibition of images made for worship, and of images of God, than as a prohibition on making pictures. The synagogue’s zodiacs, the painted motifs from the Qumran caves and the abundance of figurative art in late antique Jewish settings all tell against the strong version. What survives is not a rule but a running argument, and it is an argument the sanctuary specification is on the inside of: what may be depicted, for what purpose, and in whose presence.
The Qumran material is the sharpest test case, and it is instructive precisely because of what the movement is usually said to have believed. The community at Qumran is routinely described as the most rigorous of the Second Temple sects about images, and its writings are full of the language of angelic hosts and heavenly warfare rendered in vivid visual terms. Yet the longest and most elaborate manuscript from the caves, the Temple Scroll, is a second and expanded sanctuary description, and it describes the inner chamber, the veil, the mercy seat and the cherubim in detail. The community that is supposed to have banned images preserved and enlarged the one set of images that Judaism did have. The resolution is the same one the Greek translator found: the figures are described, specified and located, and in being described they become furniture. The tension is resolved by a theory of what images do, and it is resolved in writing rather than in carving, which is one reason the Qumran solution is so legible to us and the artistic one so difficult.
There is a further irony in the history. The medieval Sephardi illuminators, working at exactly the moment when the prohibition was being codified and argued in Jewish law, painted the sanctuary in exhaustive and gorgeous detail: the curtains with their figures, the courtyard with its hangings, the Ark with its cover and its two winged guardians, the lampstand with its branches. The Cervera Bible, copied and illuminated in Castile in 1299 and 1300, is the most famous survivor of that tradition, and its miniatures of the tabernacle show how completely the culture that forbids the graven image nevertheless wanted to see the Ark. The prohibition did not produce an image-free religion. It produced a religion in which the image was endlessly discussed, and in which the object at the centre of the discussion was this one: a box with a lid, two figures that do not look at you, and a voice that comes from between them.
The cloud and the word
What arrives at the sanctuary is not described as a presence. The biblical vocabulary is meteorological and royal: anan, cloud; amud ha-anan, the pillar of the cloud; and kabod, glory. The cloud comes to Sinai, rests on the mountain like a consuming fire and covers it for six days (Exodus 24:15 to 18). It comes down to the tent: the cloud covered the Tent of Meeting, and the glory of the LORD filled the Tabernacle, and Moses could not enter the Tent of Meeting, because the cloud had settled upon it and the glory of the LORD filled the Tabernacle (40:34 to 35). By day a cloud rested over the tent and by night fire appeared in that cloud, visible to the whole house of Israel throughout their journeys (40:38). When Solomon’s Temple is dedicated, the same thing happens: the priests could not continue their service because of the cloud, for the glory of the LORD filled the house of God (2 Chronicles 5:14). The word that appears in both places is the same word, and the sequence is the same sequence, which is one of the clearest indications in the Bible that the sanctuary builders understood their tent as the continuation of Sinai.
In Numbers the cloud becomes a logistics system, and the book is candid about how irritating it was. The cloud is over the tent; when it lifts, the people set out; at the spot where it settles, they make camp (Numbers 9:15 to 17). At the LORD’s command they break camp and at the LORD’s command they make camp, and they remain as long as the cloud stays over the Tabernacle (9:18). The trumpets are then given as the human instrument corresponding to the cloud, to summon the assembly and to set the divisions in motion (10:1 to 8), and the correspondence is exact: the people have learned to read a sky. And at the centre of the procession goes the Ark of the Covenant of the LORD, before them (10:33), so that the object which holds the tablets is also the visible token that the guide is present. When the cloud moves, the camp moves; when it waits, the camp waits, and the murmurings begin, since a people that has to be on the move at the speed of a cloud is a people that complains.
The tent of meeting of Exodus 33 is the earlier form of all this and the place where the vocabulary of presence is first made explicit. The tent is pitched outside the camp, the whole assembly stands at the entrance of each of its tents and watches Moses go in, the pillar of cloud descends and stands at the entrance, and the LORD speaks with Moses face to face, as one person speaks to another (33:7 to 11). Nothing is said about the tent being full of anything. What is said is that speech happens there, and that it happens with a directness the rest of the book rarely permits. The sanctuary specification then institutionalises that relationship, and the word the specification uses for the result is not the word we would have expected. The Bible says cloud and glory. The later tradition says presence, and it uses a noun that does not occur in the Bible at all.
The noun is shekhinah, from the root sh-k-n, to dwell or settle. It does not appear in the Hebrew Bible; the only form of it that does is the masculine proper name Shecaniah. It does not appear in the Dead Sea Scrolls. It appears in the Aramaic targums, where its function is to make a divine presence sayable in a translation that has to render a God who speaks and shows himself: in the Targum Onkelos, the words for seeing the face of God become the words for seeing the face of the shekhinah, and the phrase for the LORD going among the people is paraphrased in the same vocabulary (Exodus 33:20 and 34:9). Then it enters rabbinic literature and stays there, and the tannaitic and amoraic texts use it for the presence that abides with Israel, and, in a famous exchange, for the presence that abides everywhere: R. Ishmael and R. Sheshet, two rabbis of the third century, teach that the shekhinah abides with the people of Israel, and that it does not depart from them even while they are in exile (Bava Batra 25a). In the kabbalistic literature the word acquires a further and quite different weight. The shekhinah becomes the tenth sefirah, Malkhut, the feminine face of the divine, the sefirah that corresponds to the world; as Gershom Scholem observed, no other element of the kabbalah won such a degree of popular approval, and the association of the shekhinah with the feminine distinguishes the kabbalistic vocabulary from the earlier rabbinic one, in which shekhinah is a mode of presence rather than a mode of divinity.
It is worth being precise about what this history means. The Bible has no word for the presence of God as such, and no need of one: it has the cloud, the glory, the pillar, the kabod that fills the tent and the house. The word that the tradition invents is a word about dwelling, and it is invented, in all likelihood, in the period after the Temple, when the sanctuary was gone and the question What now? could no longer be answered by pointing at a building. A shekhinah can be lost, and then found again, and then lost again; it can descend into exile, and it can be present in a synagogue, in a house, in a marriage, in a room where somebody is learning. The rabbis needed a word for a presence that survives the loss of its furniture, and they made one. The Ark is the object that was lost, and the shekhinah is the name for what was lost with it, except that what was lost with it was not lost. This is the pivot on which the last section of this essay turns: the object and the vocabulary that replaced it were never really separate, and the history of the Ark after the wilderness is the history of a presence that had to survive without a box.
The text and its witnesses
The Masoretic text of these chapters is a much more punctured thing than a printed Bible suggests. It is divided into sections with open and closed boundaries, and the divisions in these chapters fall in telling places: the paragraph on the surplus of offerings closes a section (36:7), the craftsmen’s description ends with a closed section immediately before the accounts begin (38:23 to 38:24), and the Book of Records is named inside the section that describes the fittings (38:21). The Masorah, the body of annotation that the Tiberian masoretes wrote into the margins of their codices, records these divisions and much besides, including the number of occurrences of common words and the cross-references that let a copyist check his work. Two authorities stand behind the printed text of the sanctuary chapters. The Aleppo Codex, written in Tiberias in the tenth century, penned by Shlomo ben Buya’a and vocalized, corrected and annotated by Aharon ben Asher, is the earliest known manuscript of the complete Bible and, after the losses of 1947, the most authoritative witness to his tradition; Maimonides used it as his reference text. The Leningrad Codex, also of the Ben Asher school, is the manuscript from which most modern editions descend. Their readings are not identical, and the divergences are catalogued verse by verse in the modern documentation of the masorah, which prints for each note the masorah parva entry, the variant readings, and the sigla of the manuscripts in which each reading is attested. The documentation exists because the Masoretes themselves built it to exist: they left a machine for checking a copy, and the machine is still running. What this means for Exodus 25 to 40 is that the sanctuary account is not merely ancient and authoritative but granularly auditable, down to the pointing of a single word, and that a modern editor can say precisely where two great codices differ and how.
The Greek translation is where the text stops being a specification and becomes an interpretation. The Septuagint supplies a name the Hebrew does not: the Greek calls the chest the ark of testimony, kibotos marturion, where the Hebrew at 25:10 says only ark, and it keeps the name through the rest of the book, using the testimonies at 40:18 for what Moses puts inside. It renders the cover by the word for propitiation, hilasterion, which is the ancestor of the Latin propitiatorium and of the English mercy seat, and it does so consistently, so that the English reader of the Greek is already reading a theological gloss rather than a Hebrew noun. And then there is the wood. The Hebrew says shittim, the acacia; the Greek says incorruptible wood, and says it not once but repeatedly, of the Ark and its staves, of the table and its staves, of the frames, the bars, the posts, the altar of burnt offering and the altar of incense (25:10, 25:13, 25:23, 25:28, 26:7, 26:15, 27:1, 30:1, 30:5). A desert shrub has been turned into a material that cannot decay. The translator, working in Egypt for a community that had already begun to build with stone, has done something a scribe would not do: he has kept the function of the word and replaced its meaning, and he has done it with such consistency that the whole sanctuary becomes a structure of imperishable wood, as if the building itself had been promised durability. The specification’s silences are not the only kind of translation error, or the only kind of translation: the Greek is also where the specification acquires a theology of materials that the Hebrew never stated.
The Samaritan Pentateuch, the independent recension that the Samaritans have carried for more than two millennia, differs from the Masoretic text in roughly six thousand places, and in about one thousand nine hundred of those it agrees with the Septuagint rather than with the Hebrew. Its book of Exodus has a character of its own, and the description of it distinguishes two layers: an ancient harmonistic form in which commands were expanded, narrative sequences completed, parallel laws coordinated and material drawn in from Deuteronomy, and a later layer of scribal expansion. The Samaritan text of Exodus does not run to the same chapter and verse divisions as the Masoretic text, so the comparison has to be made by content rather than by number, and the sanctuary legislation, where the Pentateuch’s legal material is densest and where harmonising has the most to work with, is precisely where the question of what the two traditions each think a sanctuary is has to be asked. A fragmentary Exodus scroll from Qumran, conventionally called 4QpaleoExod, stands close to the Samaritan text and shares its major expansions, which is one of the reasons the Samaritan recension is now discussed as a live witness rather than as a curiosity.
The fragments of Exodus from the Judean desert give the sanctuary account its own small archive. Cave 4 alone yielded a series of Exodus manuscripts, lettered from b upwards, several of them tiny, and their contents are recorded in the catalogues and surveyed by Ziemier: one preserves Exodus 7 to 18 in a long series of small fragments, another preserves only Exodus 13, and one, 4QExod k, preserves Exodus 36:9 to 10, two verses from the weaving of the tent’s curtains in the middle of the building account. Benjamin Ziemier’s survey of the Qumran Exodus manuscripts against the Masoretic text, the Samaritan Pentateuch and the Old Greek concludes that the caves demonstrate that a genuine diversity of textual witnesses for Exodus existed long before the Masoretic text was fixed, and that the later dominance of the Masoretic form was a development rather than an accident. The longer scrolls are more consequential still. The Temple Scroll, designated 11Q19 and discovered, it is generally thought, in 1956 in Cave 11, consists of eighteen sheets of extraordinarily thin skin, arranged in three or four columns each, eight and a half metres long, the largest scroll ever found at Qumran, with two further copies of the same composition recovered. Most scholars assign the composition to the second half of the second century BCE, but there is no consensus about its date, its author, or its relation to the community at Qumran. Its content is the sanctuary again, at a scale that dwarfs Exodus 25: an ideal Temple described in the voice of God to Moses, with the inner chamber, the veil, the mercy seat and the cherubim set out in far greater detail than the Pentateuch gives them, and with the sacrificial and purity law expanded to match. The Masoretic text is the specification; the Temple Scroll is a second specification, made a few centuries later, in a different century’s religious weather, and it is the one that tells us what the sanctuary chapters were taken to mean once they had become a scriptural inheritance rather than a building programme.
One last piece of transmission history belongs here, because it is so nearly a coincidence. The earliest dated Hebrew printed book is generally reckoned to be an edition of Rashi’s commentary on the Pentateuch, printed by Abraham ben Garton at Reggio di Calabria in February 1475. The first printed Pentateuch followed at Bologna in January 1482, with Rashi’s commentary above and below the text and the Targum Onkelos in the outer margin, an enterprise financed by Joseph ben Abraham Caravita and printed by Abraham ben Hayyim de Tintori of Pesara and corrected by Joseph Hayyim of Strasbourg, on vellum, in an edition of which the copy once owned by the bibliographer B. J. De Rossi is in the Library of Congress. The age of the sanctuary commentary and the age of the printed Hebrew book therefore begin within a decade of one another, and the chapters about the Ark were among the first things in the Hebrew Bible to be mechanically reproduced. Meanwhile the great Sephardi codices were being copied by hand for exactly the readers who would want those commentaries: the Cervera Bible, copied and illuminated in Castile in 1299 and 1300, is a parchment codex of 451 folios with vocalization and masoretic notes, illuminated by Yosef Asarfati, and among the oldest and finest of the surviving Sephardi Bibles; it is kept in the Biblioteca Nacional de Portugal in Lisbon. The Ark was read in a book that had been made to be looked at, and it was read by people who had begun to print it.
The Ark after
The Ark outlived the tent by a long way, and it outlived the tent because the Ark was the one piece of furniture that could not be folded. It stands at Shiloh for years, in the tent of Eli’s house, and the young Samuel sleeps in the sanctuary itself (1 Samuel 1 to 3). When the elders of Israel ask for the Ark to go out with them against the Philistines, they are sending the furniture of the sanctuary into the field, and the result is the most famous defeat in the book’s history: thirty thousand of Israel’s dead, the loss of the two sons of Eli, and the Ark itself carried off to Philistia (1 Samuel 4). The name given to the child born to Eli’s daughter-in-law that day is the book’s comment on the whole event: Ichabod, the glory has departed from Israel (1 Samuel 4:22). The Ark is returned, but not before the Philistines have put a coffer of golden tumours and five golden mice beside it as a guilt-offering, and the Ark has sat in the house of Obed-edom for seven days before the men of Kiriath-jearim take it up (1 Samuel 5 to 7). Its first great journey is therefore a defeat and a humiliation, and the tradition is careful never to let it forget that.
David brings it to Jerusalem, dancing in a linen ephod while Michal watches from the window and is contemptuous (2 Samuel 6), and puts it in a tent he has pitched for it in the city; and when he proposes to build a permanent house for it, Nathan the prophet forbids it, and the Ark stays in a tent inside the walls of the city for the rest of David’s reign (2 Samuel 7). Solomon inherits the tent, and it is still standing at Gibeon, the greatest of the high places, when the LORD appears to him in a dream and offers him whatever he asks (1 Kings 3:4 to 5). So the tent of meeting outlived the wilderness by generations and outlasted the judges, and it was still the shrine in the year Solomon became king. What the Temple changed was the permanence and the scale; it did not change the theology of the furniture. When the Ark is carried into the Holy of Holies it goes under the wings of the cherubim that are themselves an enlargement of the two figures beaten into the cover (1 Kings 8:6), and the Chronicler, looking down the corridor of tradition, records that there was nothing inside the Ark but the two tablets (2 Chronicles 5:10). The cloud comes to the Temple as it came to the tent, and the glory fills the house, and the priests cannot continue their work because of it (2 Chronicles 5:14). Sinai has become Jerusalem without anyone noticing the substitution.
Then the Ark disappears from the story, and the tradition begins to work hard on the absence. The Second Temple had no Ark. The space where it had stood was still called the place of the Ark, and the ritual of the Day of Atonement went on performing itself around an empty place, the coals set as though between the staves (Yoma 52b). The same passage records what was done with the object: when the Ark was hidden, the jar of manna and the flask of anointing oil were hidden with it, and Aaron’s rod with its almonds and blossoms, and the coffer which the Philistines had sent as a gift to the God of Israel; and the question of who buried it is answered with Josiah, king of Judah, who saw from the words of Deuteronomy that the exile was coming and judged it better that the Ark should not be disgraced in exile, so he hid it (2 Chronicles 35:3). The Tosefta adds that all these things lay in the Holy of Holies. R. Elazar then converts the episode into a rule of inference by drawing a verbal analogy between the words for keeping in three passages, and the hidden objects are hidden together. What the rabbis construct, in other words, is not a disappearance but an archive: a sealed deposit of the sanctuary’s most precious contents, laid away by a king who could read the future in the legislation. How the Ark’s hiding stands related to the high priest’s breastpiece, with its four rows of stones engraved as seals with the names of the twelve tribes and its Urim and Thummim (Exodus 28:30; 39:10 to 14), is a further question on which the sources do not speak with one voice; what is certain is that the Ark and the breastpiece are the two containers of Israel’s names in the sanctuary, one lying in the Holy of Holies and one worn on the high priest’s chest, and that the tradition which hides the first is also the tradition that makes the second the oracle of a people without a king.
The most famous of the Ark’s afterlives is not a Jewish one. The Kebra Nagast, the fourteenth-century Ethiopic chronicle of the Solomonic dynasty, tells how Menelik I, son of Solomon and the Queen of Sheba, carried the Ark down to Ethiopia, and the dynasty that ruled from 1270 until 1974 drew its legitimacy from that story. In Axum the Ark is said to rest in the Chapel of the Tablet, built in the mid-twentieth century with funds from Empress Menen, because the divine heat of the tablets had cracked the stones of the earlier inner sanctum. The custody is governed by a rule as strict as anything in the sanctuary specification itself: a single guardian monk is appointed for life by his predecessor, and only the guardian may view the Ark; if the guardian dies without naming a successor the monks of the monastery elect one. The claim is treated with respect and with scepticism in roughly equal measure, and the details of its twentieth-century history are contested: Edward Ullendorff declared in 1992 that he had examined the object in 1941 while serving as a British Army officer, describing it as empty and as a construction of the Middle or late Middle Ages, and the refusal of any inspection since has kept the question open in a way that no amount of scholarship can close. The Ark of Axum, in other words, has been preserved with the same care as the ancient cult objects it claims to be, and it is the refusal of access, itself a kind of sanctuary rule, that makes it impossible to settle.
From Axum the object travels further still, and it travels as a sign. It becomes the ark of the New Covenant in Christian typology, from the third and fourth century fathers onwards, and Mary becomes the vessel that carried the God of Israel; it becomes the object of quests, of expeditions, of films and of a modern spirituality that treats the Ark as a treasure with a spiritual address. None of these afterlives is an error. Each is the same object seen from a different point in its long life as a sign of presence and of memory, and the fact that the sign survives so much better than the chest is not a failure of transmission but the whole point of the story. An object made by direct instruction, at enormous cost, in a wilderness, at the centre of a camp that could not stay still, became a sign of a people’s memory; and the memory had to be able to travel without the box, which is why the rabbis invented the word for the presence and the mystics gave it a body in the sefirot, and why the chest turned up in Ethiopia guarded by a monk for life, and why it still sits at the centre of every account anyone gives of what it means for a people to be addressed by a voice they cannot see.
References
- Tanakh: Exodus — Sefaria editors, Sefaria (Tanakh with the Jewish Publication Society 1985 translation), accessed 23 October 2026
- Babylonian Talmud, Yoma 52b — Sefaria editors, Sefaria (William Davidson edition), accessed 23 October 2026
- Babylonian Talmud, Berakhot 55a — Sefaria editors, Sefaria (William Davidson edition), accessed 23 October 2026
- Babylonian Talmud, Sanhedrin 69b — Sefaria editors, Sefaria (William Davidson edition), accessed 23 October 2026
- Septuagint (Brenton 1879): Exodus, with the Greek text — Lancelot C. L. Brenton, Wikisource, 1879 text, accessed 23 October 2026
- Bezalel — Jewish Encyclopedia editors, The Jewish Encyclopedia (Funk and Wagnalls), 1901-1906
- Ark of the Covenant — Grokipedia contributors, Grokipedia, accessed 23 October 2026
- Shekhinah — Grokipedia contributors, Grokipedia, accessed 23 October 2026
- Aniconism in Judaism — Grokipedia contributors, Grokipedia, accessed 23 October 2026
- Digital Dead Sea Scrolls: The Temple Scroll — Israel Museum, Jerusalem, Digital Dead Sea Scrolls project, accessed 23 October 2026
- The Qumran Exodus Manuscripts and the Versions that Became Canonical (MT, SP, OG) — Benjamin Ziemer, Hebrew Bible and Ancient Israel 13/4 (2024), 551-582, DOI 10.1628/hebai-2024-0029
- Samaritan Pentateuch — Grokipedia contributors, Grokipedia, accessed 23 October 2026
- Aleppo Codex — Grokipedia contributors, Grokipedia, accessed 23 October 2026
- Miqra according to the Masorah, with documentation: A2 Exodus — Ben Benckla, MAM project, benckla.github.io, accessed 23 October 2026
- Hebrew incunabula — Grokipedia contributors, Grokipedia, accessed 23 October 2026
- The Bologna Pentateuch: Judaic Treasures — Jewish Virtual Library, Jewish Virtual Library, accessed 23 October 2026
- The Cervera Bible (Hebrew, 1299-1300, illuminated) — Digital Bible Society, Internet Archive, accessed 23 October 2026
- Church of Our Lady Mary of Zion — Grokipedia contributors, Grokipedia, accessed 23 October 2026