How close is today’s Bible to its earliest copies?

Codex Sinaiticus was written in Greek in the middle of the fourth century and contains the earliest complete copy of the Christian New Testament. It also contains two early Christian works not found in modern Bibles, the Epistle of Barnabas and the Shepherd of Hermas, appended after Revelation.

The book itself is now in four places: London, Leipzig, St Petersburg and St Catherine’s Monastery at Sinai. It was photographed and transcribed in full and put online, which is why anyone can now check a verse against a fourth-century page.

Codex Vaticanus, catalogued as Vat.gr.1209 and dated by the Vatican Library’s own record to the fourth century, has also been digitised page by page. The Sinaiticus project calls it the only other nearly complete Christian Bible of a similarly early date.

Two substantial witnesses from the same century make comparison possible. They are not wholly independent: textual critics class both as Alexandrian, one line of transmission among several, so their agreement weighs less than agreement across lines would. Where they differ, the difference is the evidence.

Until 1947, the oldest Hebrew manuscripts of most of the Old Testament were medieval. Then a shepherd found scrolls in a cave near Qumran, and over the following nine years ten more caves nearby were found. Together they held a few near-complete scrolls and tens of thousands of fragments, representing more than nine hundred texts copied between the third century BC and the first century AD. About 230 of the manuscripts are biblical, with at least part of every book of the Hebrew Bible except Esther.

That is roughly a thousand years older than anything previously available. It was, in effect, an unplanned experiment on a millennium of copying, and many of the biblical scrolls turned out to be close to the medieval text. Not all of them were.

In Codex Sinaiticus, Mark 16 runs to verse 8 and stops. Printed Bibles still carry the longer ending, and the note attached to it can be informative in itself: the World English Bible prints verses 9 to 20, reports that the critical text footnotes their omission by a few manuscripts, and then tells the reader its own translators regard the passage as reliable. That is a note taking a side, not only flagging a fact.

A handful of familiar passages are in this position. They are known and they are flagged, and the flagging is older than printing: some manuscripts that contain Mark’s longer ending carry a scribe’s note against it, or mark it with asterisks or obeli, the signs copyists used for a spurious addition.

The number of differences across all known manuscripts is very large, perhaps a third of a million for the New Testament alone, and most of them are minutiae that make no difference to translation or exegesis. Both of those things are true at once, and quoting only one of them is how this subject gets misrepresented, in either direction.

Where scholars argue over particular readings, the set is much smaller: passages where a change looks deliberate, and where the wording bears on doctrine. How many such passages there are, and how much weight they carry, is a live discussion rather than a settled count. Some go further and doubt there was ever a single original to recover: David Parker argues that the text stayed fluid for centuries.

Editors weigh how old and how good the manuscripts behind a reading are, and how widely it is spread across separate lines of transmission; whether it suits the author’s style and context; and which reading best explains how the others arose. Rules of thumb such as preferring the shorter or the harder reading are a last resort, and applied mechanically they seldom convince. The considerations sometimes point in different directions, and then judgement decides.

This is why printed Greek New Testaments come with a critical apparatus of variant readings, and why the United Bible Societies edition grades its committee’s confidence in each reading from A to D. The edition is an argument, and it shows its working.

Between composition and the earliest substantial copies there is a gap filled almost entirely by fragments. What happened to the text in that period, before copying became institutional, is reconstructed rather than observed.

It is also the period in which the most consequential changes would have been easiest to make and hardest to detect, which is precisely why it cannot be quietly assumed to have been uneventful.

Many Christians hold that the text has been providentially preserved, and read the manuscript record as evidence of care rather than as a merely lucky outcome. The Westminster Confession says the Hebrew and Greek were "kept pure in all ages" by God’s "singular care and providence". Churches do not all name the same text as authoritative: the Council of Trent made the Vulgate the authentic Latin edition for public use, which Pius XII later said was a juridical judgement rather than primarily a critical one.

Textual criticism can describe how the copies relate to each other. It has nothing to say about whether anything was superintending the copyists, which is a different kind of claim about the same events.

Churches do not all print the same Old Testament, and the Greek and Hebrew traditions differ in more than the book list: some texts are longer in one and shorter in the other. The Qumran copies caught that variation while it was still live. Many of them are close to the Masoretic text, known before from manuscripts more than a thousand years later, which the scrolls project calls remarkable given the gap, and some differ from it in language, in spelling, and in what has been added or dropped, in a way the same project reads as scribes feeling free to modify what they copied. The biblical manuscripts from the later refuge caves do not vary in that way: they match, which is why the Hebrew text is thought to have settled by the second century.

So a question about closeness to the earliest copies has to be asked about a particular passage, in a particular tradition. Asked in general, it has no single answer.