language menu item language menu item language menu item language menu item

Amy Myshrall and Timonthy Brown checking their transcription on a Codex Sinaiticus leaf at Leipzig University Library.
Amy Myshrall and Timonthy Brown checking their transcription on a Codex Sinaiticus leaf at Leipzig University Library.

Download the full image.

Reconstructing Codex Sinaiticus


The discovery of pages of Codex Sinaiticus along with many other treasures at St Catherine’s Monastery, Sinai, in 1975 was an important event in the life of one of the world’s most remarkable manuscripts. As well as increasing the amount of biblical text for which the codex is available, the New Finds make it possible to reconstruct the contents of the first thirty-four quires in a way that was not previously possible. The following report will describe how it is possible to reconstruct the original content of the codex, down to the page where each book ended and began, and then consider some of the implications of this reconstruction.


The careful study of all the new fragments reveals that there are nineteen leaves wholly or partially extant, along with a few tiny fragments in which the text cannot be identified. They contain portions of Genesis (a leaf consisting of parts of eight columns containing 21.26-22.17 and 22.21-23.16), a whole leaf (Leviticus 20.27-22.30), a sequence of complete leaves from Numbers (16.7-26.2), a mutilated leaf which has parts of Deuteronomy 3.8-4.21 and a complete leaf with Deuteronomy 28.68-30.16; another mutilated leaf with Joshua 12.2-13.16, a tiny fragment with Judges 2.20 on one side and 4.6 on the other, and then a sequence of five complete leaves containing Judges 4.7-11.2; Finally in the Old Testament, there are scraps of a leaf which contains parts of 1 Chronicles 17.14-18.11. There are fragments of a leaf containing parts of Hermas, Similitudes 6.5.5-66.6, and a complete leaf with IX.14.4-18.5.

There are other fragments found in the nineteenth century and now in St Petersburg. The ones which concern us here consist of a part of a leaf now preserving Genesis 23.19-24.19 and 24.20-24.46; another partial leaf now with Numbers 5.26-6.18, 22-7.20.


The starting point is to find places where a quire number is preserved. The original numbers were written at the top left corner of the first page of each quire. These were probably added at the time of writing as a guide to the binder, since many were cut off when the folios were trimmed. A later series was added in the top right corner, which are further in from the corner, and therefore usually preserved. The hand which added these also put a binding mark on the middle of the right hand edge of the central opening of each quire.

The second step is to know how the manuscript was put together. Like other Greek manuscripts of its time, it follows Gregory’s Rule.[1] That is to say, it was made with matching pages of flesh and hair, with flesh on the outside of every quire of eight leaves. Thus the first page of a quire is always flesh, the second and third are hair, the fourth and fifth flesh, and so on. Whether a page is hair or flesh is easily told. Hair is darker, and absorbs the ink much better than the greasier flesh side, from which it tends to flake off. The first, third, fifth and seventh rectos of every quire are flesh sides, and the even numbered ones are hair. The only exception to this in the main blocks of text is Quire 91, which consists of the first, third and fourth sheets of a regular four-sheet quire, so that the openings Quire 91 Folio 1v/2r and Quire 91 Folio 5v/6r are not matching hair and matching flesh as they should be, but hair/flesh and flesh/hair respectively). The reason for this lies in some problem with a transition between blocks copied by Scribes A and B. The reconstruction assumes that no such problem arose in the lost early quires of the codex, and, since the calculation works very accurately, the conclusion justifies the assumption.

The third step is to calculate the space taken up by missing blocks of text. The reconstruction follows the principle that the number of missing letters between extant blocks of text divided by an average number of letters on a folio would give the number of folios missing between extant portions. In order to calculate these figures, an unaccented text of the CCAT Septuagint (see was downloaded from Elements such as punctuation and chapter and verse numbers were removed, and the blocks of text placed in a Word document. The Word Count tool was used to calculate the number of letters excluding spaces. This was then divided by the number of letters per line in the local extant portion of text (usually 14, sometimes 13.5), multiplied by 48 lines to the column and 8 columns to the leaf. Three elements which might affect the calculations appear to balance each other out. Two of them tend to make the printed text longer than the manuscript would have been. The first is abbreviations in the codex (notably the ?a? compendium, superline for final nu and sacred names reduced to the first and last letter (in fact there are not so many in this part of the Bible as in the New Testament). Textual variants were also ignored. Since scribes tend to omit text, the printed version will always be likely to be somewhat longer. On the other hand, there will have frequently been short lines where the scribe made a section break. The assumption that these elements cancel each other out was substantiated by the evidence.

There are three areas of enquiry:

  1. locating the New Finds and St Petersburg fragments of the Heptateuch and 1 Chronicles and reconstructing the structure of the first 34 folios
  2. locating the New Finds fragments of Hermas
  3. reconstructing the structure of Quires 50-56

1. Locating the New Finds and St Petersburg fragments of the Heptateuch and 1 Chronicles and reconstructing the structure of the first 34 folios

Fixed points

There are quire signatures on the folios containing Leviticus 20.27-22.4 (? = Quire 10) and Numbers 16.7-31 (IB = Quire 12). There is a binding mark on the page containing Judges 7.4-7.22, which must therefore be Folio 5r of its quire.

Genesis 21.26-22.7 and 23.19-24.46

The first extant text is a leaf containing Genesis 21.26-22.17. This has a flesh recto, and so must be an odd-numbered leaf. The amount of extant text between Genesis 1.1 and 21.26 is 52,449 characters. The density of the leaf is 5245 letters, and 52,4449 divided by 5245 = 9.99 folios, indicating that 10 folios are lost at the beginning. Therefore this leaf is Folio 3 of its quire.

The next extant leaf is in St Petersburg, and carries straight on in the text. Therefore it is Folio 4 of its quire.

If the manuscript began on Quire 1, then these two leaves would be in Quire 2. But we have no grounds for assuming this, and cannot yet allocate a quire number

Leviticus 20.27-22.30

The next extant leaf contains the quire signature 10. It is therefore Quire 10, Folio 1.

To retrace our steps: the lost block of text Genesis 24.46-Leviticus 20.27 contains 281,324 characters.[2] This divided by 5376 is 52.3 folios, which may be rounded down to 52.[3] 52 folios consists of six quires and four folios. Thus the folios in Genesis must belong to Quire 3.

Genesis 1.1 therefore began on Quire 2 Folio 1. We now know that the manuscript began with a quire of prefatory material. The number of folios in this quire cannot be known.

Numbers 5.26-7.20

Leviticus 22.30 to the end of the book contains 20,177 characters. This divided by 5376 gives 3.75 folios, namely Quire 10 Folio 2, 3, 4 ,5. If one postulates two blank columns at the end of Leviticus, we have 4 missing folios. [4]

Numbers therefore began on Quire 10 Folio 6r.

Numbers 1.1-5.26 contains 21,481 characters. This divided by 5376 gives 3.99 folios. [5] These were Quire 10, Folios 6, 7, 8 and Quire 11, Folio 1. So Numbers 5.26, which has a hair recto, begins on Quire 11 Folio 2.

Numbers 7.20-16.7 has 31,423 characters. This divided by 5376 gives 5.85 folios. A slightly lower density of 5237 characters per folio gives 6, completing Quire 11.

Numbers 16.7-20.28

Numbers 16.7 is on the numbered leaf Quire 12, Folio 1. The next two folios are continuous, namely Quire 12, Folios 2 and 3.

This longer sequence also gives the opportunity to stop and check the letter density. ?hese three folios contain 15,350 characters. This divided by 3 gives a figure of 5117 characters per folio. [6] This new figure provides convincing calculations for the next two quires. [7]

Numbers 23.22-26.2

The missing block Numbers 20.28-23.22 has 10,412 characters. This divided by 5117 is 2 folios. These were Quire 12, Folios 4-5. Numbers 23.22-26.2 is therefore Quire 12, Folio 6.

Deuteronomy 3.8-4.21

The missing block Numbers 26.2 to the end of the book is 36,077 characters. This divided by 5117 amounts to 7.05 folios. These seven folios were Quire 12, Folios 7 and 8, Quire 13, Folios 1, 2, 3, 4 and 5.

Deuteronomy therefore began on Quire 13, Folio 6r. Deuteronomy 1.1-3.8 contains 9,897 characters. This divided by 5117 gives 1.93 folios, i.e. 2 folios, namely Quire 13, Folios 6 and 7. Deuteronomy 3.8-4.21 is therefore Quire 13, Folio 8.

Deuteronomy 28.68-30.16

Deuteronomy 4.21-28.68 contains 74,611 characters, which divided by 5376 gives us 13.87 folios. A round 14 folios means that Deuteronomy 26.68-30.16 is Quire 15, Folio 7r.

Given the slight mismatch between the calculation and the round number, another check on the letter density in this leaf is advisable. Quire 15, Folio 7 contains 5360 characters, which comes out at 13.96 characters per line, so 14, and therefore the figure 5376, is accurate. It appears that in the previous leaves, the density crept up somewhat from 5117.

Joshua 12.2-14.3

Deuteronomy 30.16 to the end of the book is 13,823 characters. This divided by 5376 indicates 2.53 folios. Deuteronomy therefore ended on Quire 16, Folio 2r, Column 4. and Joshua began on Folio 2v. There are 34,043 characters in Joshua down to 12.2, which divided by 5376 gives us 6.33 quires. The figure one would expect to make up the gap before the next page is 6.5. The letter density on these leaves must therefore have been 5237.

Joshua 12.2 therefore begins on Quire 17, Folio 1. Unfortunately, the top corners of the leaf are missing so we have no quire signatures.

Judges 4.7-11.2

It is easiest to calculate first the main block of Judges.

Joshua 13.17 to the end of the book is 35,259 characters, which divided by 5376 gives us 6.55 folios. Judges 1.1-4.7 is 12,041 characters, which divided by 5376 gives us 2.24 folios. Joshua 13.17-Judges 4.7 taken together is 47,300 characters, amounting to 8.8 folios or 9 folios. So Judges 4.7 is on Quire 18, Folio 3r.

But there is also a tiny fragment of Judges 2.20 from the first column of a hair recto, starting at line 44. Judges 1.1-2.20 contains 7262 characters, or 1.35 folios. If Judges begins in a column 3, then 11 columns gives the right calculation (and the calculation of 8.8 folios for the end of Joshua allows space). Judges probably began on Quire 17, Folio 8v, Column 3.

Calculating backwards to confirm this

The first large block of text in the manuscript begins at Quire 34, Folio 8r, which starts in 1 Chronicles 9.27. This is in fact within a block of text which had been copied by mistake, as the ‘Three crosses’ note on Quire 35, Folio 4v makes clear. [8] The repeat section of 1 Chronicles must have begun on Quire 34, Folio 6. Since the text on Quire 34, Folio 8r after the 1 Chronicles intrusion is the middle of 2 Esdras 9.9, the text before the 1 Chronicles intrusion was 2 Esdras 1.1-9.9.

2 Esdras 1.1-9.9 has 23,326 characters. This divided by 5376 is 4.3 folios. 1 Esdras is 46495 characters, which divided by 5376 gives 8.65 folios. Thus 1 Esdras and 2 Esdras 1.1-9.9 together contain 69,821 characters, amounting to 12.99 folios. It is therefore reasonable to argue that 1 Esdras began on Quire 33 Folio 1r. The precise division point between the two books is somewhat uncertain, but might have been between the third and fourth columns on Quire 34, Folio 1r.

2 Chronicles contains 104,889 characters. This divided by 5376 amounts to 19.5 folios. If 2 Chronicles went on 20 folios, it began on Quire 30, Folio 5r.

The end of 1 Chronicles was therefore on Quire 30, Folio 4v.

1 Chronicles has 78,412 characters divided by 5376 (characters per folio at 14 characters per line), amounting to 14.6 folios which can be rounded down to 14.5 folios. This means that 1 Chronicles begins on Quire 28, Folio 6v.

The fragment that begins at 17.14 must be placed within these 14.5 folios.

The verso side of the fragment begins at Line 1, Column 1 at Ch.18.1. 18.1 to the end of 1 Chronicles is 30,501 characters. 30,501 characters divided by 5376 = 5.67 folios, rounded down to 5.5 folios, means that 1 Chronicles 18.1 begins Quire 29, Folio 7v. This means that the flesh recto of the 1 Chronicles fragment is Quire 29, Folio 7r.

From 1 Chronicles 1.1–18.1 is 47,911 characters, which divided by 5376 gives 8.9 folios, that is 9 folios. Counting backwards, this indicates that 1 Chronicles began on Quire 28, Folio 6v.

1-4 Kingdoms has 373,210 characters. This divided by 5376 gives 69.4 folios; 69.5 folios indicates that 1 Kingdoms 1.1 began on Quire 20, Folio 1r. Within these four books:

  • 1 Kingdoms has 95,893 characters, which divided by 5376 gives 17.4 folios. This indicates that 1 Kingdoms ended on Quire 22, Folio 1v.
  • 2 Kingdoms has 85,164 characters, which divided by 5376 gives 15.84 folios. This indicates that 2 Kingdoms began on Quire 22, Folio 2r ended on Quire 24, Folio 1v.
  • 3 Kingdoms has 100,432 characters, which divided by 5376 gives 18.68 folios. This indicates that 3 Kingdoms began on Quire 24, Folio 2r ended on Quire 26, Folio 4v.
  • 4 Kingdoms has 91,762 characters, which divided by 5376 gives 17.07 folios. . This indicates that 4 Kingdoms began on Quire 26, Folio 5r ended on Quire 28, Folio 5v.

Judges 11.2 -end of the book is 36612 characters, which divided by 5376 gives 6.8 folios, i.e. 7 folios.

Ruth contains 9,735 characters. This, added to Judges 11.2-end amounts to 36,612 + 9,735, which is 46,347 characters, or 8.62 folios. Ruth occupied just under 15 columns, so there will have been a blank column on Quire 19, Folio 8v.

Working back from Quire 20, Folio 1r as the beginning of 1 Kingdoms gives 9 folios, confirming that the last of the whole folios in Judges is Quire 18, Folio 7.

2. Locating the New Finds fragments of Hermas

There are 30,908 characters between Quire 95, Folio 1v and the last full folio. The last full folio (containing Similitudes 9.14.4ff) has 5191 characters, or 5.95 folios, so the next folio is Quire 95, Folio 8.

Quire 95, Folio 8v needs a part line and nine full lines supplying, which at 14 characters per line means that it ended at Sim. IX.XIX.1 ?? t??.

The remainder of the text (including the Latin!) has 17137 characters, which at 5376 amounts to 3.19 folios. Can one find an equivalence between the number of letters in the Greek and the Latin version? Taking a random example, the Latin version of Sim. IX.I-IV has 5880 characters against the Greek’s 5677. So we can treat the figure as approximately right.

The last folio of Hermas was therefore Quire 96, Folio 4v. Whether this was the last folio of the manuscript cannot be determined: it may have contained other works. [9]

If this was the last work in it, then the manuscript ended with a short quire of two sheets.

3. Reconstructing the structure of Quires 50-56

Isaiah and Jeremiah, written by Scribe B, together contain 277,553 characters. [10] They take up 6 quires and 7 folios, or 55 folios, giving a density of 5,046 characters to a folio. So Scribe B writes this section with a lower density than was produced elsewhere. Actually the density is slightly higher than that. If one takes away the last page and a column of Jeremiah, it comes out at 6 quires and 6 folios for 221 fewer characters, i.e. 277,332 for 54 folios, which is 5136 characters per leaf. This gives an average of 13.4 letters per line.

Our transcription of Joel has 7829 characters. It starts Quire 57, Folio 1r, Column 1 and ends Quire 57, Folio 2v, Column 1, and has 592 lines, = 13.22 characters per line, so we need to calculate throughout this block at 13 characters per line.

The rest of Lamentations (2.20-end) is 6636 characters. Baruch has 12,793 characters. The Epistle of Jeremiah has 6718 characters. Ezekiel has 142,844 characters. Daniel, including Bel, Susanna and the Prayer of Zechariah, has 62,050 characters. If these folios are calculated at 13 characters to the line, there are 624 characters to the column and 4992 to the leaf. It is worth calculating by the column, since some of these short books such as the Epistle of Jeremiah or the Minor prophets are likely to have been taken within blocks. These calculations produce the following results:

  • Rest of Lamentations
    • Lamentations 2.21 to end has 6,636 characters, which divided by 624 characters per column amounts to 10.63 columns = 1 folio and 3 columns
    • Lamentations ends Quire 50, Folio 2r, Column 3
  • Baruch
    • Baruch starts Quire 50, Folio 2r, Column 4
    • Baruch contains 12,793 characters, which divided by 624 amounts to 20.5 columns, or 2 folios and 5 columns
    • Baruch ends Quire 50, Folio 4v, Column 4
  • Epistle of Jeremiah
    • The Epistle of Jeremiah starts on Quire 50, Folio 5r, Column 1
    • The Epistle of Jeremiah contains 6,718 characters, which divided by 624 characters per column amounts to 10.76 columns, or 1 folio and 3 columns
    • The Epistle of Jeremiah ends Quire 50, Folio 6r, Column 3
  • Ezekiel
    • Ezekiel starts Quire 50, Folio 6r, Column 4
    • Ezekiel contains 142,844 characters, which divided by 4992 characters per folio is 28.615 folios, or 28 folios and 5 columns)
    • Ezekiel ends Quire 54, Folio 2v, Column 4
  • Daniel
    • Daniel starts Quire 54, Folio 3r, Column 1
    • Daniel contains 62,050 characters, which divided by 4,992 characters per folio is 12.43 folios (98 columns)
    • Daniel contained 12 folios and 4 columns
    • Daniel ends Quire 55, Folio 6v, Column 4
  • Hosea
    • Hosea starts Quire 55, Folio 7r, Column 1
    • Hosea contains 19,798 characters, which divided by 624 characters per column is 31.7 columns, or 4 folios
    • Hosea ends Quire 56, Folio 2v, Column 4
  • Amos
    • Amos starts Quire 56, Folio 3r, Column 1
    • Amos contains 16,166 characters, which divided by 624 gives 25.9 columns, or 3 folios and 2 columns
    • Amos ends Quire 56, Folio 6r, Column 2
  • Micah
    • Micah starts Quire 56, Folio 7r, Column 3
    • Micah contains 11,828 characters, which divided by 624 characters per column = 18.95 columns, or 2 folios and 3 columns
    • Micah ends Quire 56, Folio 8v, Column 4


Genesis begins21r
Genesis ends/Exodus begins56r
Exodus ends84r
Leviticus begins4v
Leviticus ends105v
Numbers begins6r
Numbers end135v
Deuteronomy begins6r
Deuteronomy ends162r
Joshua begins2v
Joshua ends, Judges begins178v
Judges end197v
1 Kingdoms begins201r
1 Kingdoms ends221v
2 Kingdoms begins2r
2 Kingdoms ends241v
3 Kingdoms begins2r
3 Kingdoms ends264v
4 Kingdoms begins5r
4 Kingdoms ends285v
1 Chronicles begins6r
1 Chronicles ends304v
2 Chronicles begins5r
2 Chronicles ends328v
1 Esdras begins331r
1 Esdras ends /2 Esdras begins341r
2 Esdras5v
1 Chronicles duplicate page6r
1 Chronicles duplicate page6v
1 Chronicles duplicate page7r
1 Chronicles duplicate page7v
Lamentations ends/Baruch begins502r
Baruch ends/Ep Jer begins 3r
Ep Jer ends5v
Ezekiel begins6r
Ezekiel ends542v
Daniel begins3r
Daniel ends557v
Hosea begins8r
Hosea ends, Amos begins563r
Amos ends, Micah begins7r
Micah ends8v
Hermas ends964v


[1] C.R. Gregory, Canon and Text of the New Testament, Edinburgh, 1907, 323-4.

[2] The rest of Genesis 94,649; Exodus 120,587; Leviticus 1.1-20.27 66,088.

[3] Quire 10 Folio 1 contains about 5,167 characters. This rate per page would add another three folios to the gap, which does not obtain a result matching with hair/flesh sequence of the Genesis fragments. A compression somewhat below 14 letters is more accurate.

[4] Except in very short books (the minor Catholic Epistles), the scribes preferred to begin a new book on a new folio, or even a new quire.

[5] I.e. about two lines out from a precise match.

[6] If one adjusts the line density to an average between 5376 and 5117 for the missing block Numbers 7.20-16.7, then the revised calculation of 31,423 divided by 5247 (the mean figure) gives 5.99 folios. It seems therefore that the scribe began to space the text slightly more as this block progressed.

[7] Convincing in that the calculations match the hair-flesh information, and in the end give a precise match through to the point where the London and Leipzig leaves begin.

[8] μεχρι του ϲημειου τω(ν) τριων ϲταυρων εϲτι(ν) το τελοϲ των επτα φυλλων τω(ν) περιϲϲων κ(αι) μη οντω(ν) του εϲδρα.

[9] One might even speculate that the manuscript lost a similar number of leaves at each end, so that around sixteen leaves are lost at the end, allowing room for one or more further texts.

[10] The project has produced some evidence that there was a scribe B1 and a scribe B2. That does not affect these calculations.