Chapter 21: The Masoretic Text: How Faithfully Has the Torah Been Copied?, section 1 of 3 Overview of chapter 21 Book contents Previous Next

Which Text?

A code, of the kind described in the previous chapter, depends on the text in a way that ordinary reading does not. A reader who meets a misspelled word still understands the sentence. A code allows no such tolerance. It is a precise count: a fixed number of letters from the first letter of the hidden word to the second, the same number to the third, and so on. Suppose my text and yours differ by a single letter, one extra or one missing, anywhere between the first and last letters of a code I have found. Then every letter after that point is shifted by one, the intervals no longer match, and in your text the code does not exist. The reverse is also true: a variant letter can create a code in one text that is absent from all the others.

A difference outside the span of the code does no harm, since it shifts the whole code together. Its letters keep the same distances from one another. A difference inside the span destroys it. The longer the interval, the longer the span, and the greater the chance that some variant falls within it. Anyone who reports a code is therefore reporting it in a particular text, and the first question to ask is: which one?

This raises a question that most readers of the Torah never have reason to ask. Every copy of the Torah was written by hand until the age of printing, and copies written by hand inevitably contain mistakes. Proof of this is found every day. There are organizations that check Torah scrolls by computer, comparing every letter with the traditionally accepted text, and they regularly find errors. Almost all are minor and do not affect the meaning: a letter missing here, an extra letter there, one letter mistaken for another of similar shape. In Jewish law, however, these errors matter. A Torah scroll may not be used for the public reading unless it is “kosher,” and a kosher scroll is one that matches the accepted text letter for letter. A scroll is invalid if even one letter is missing.

But what is this accepted text, and why should it carry more weight than any other? Several very old copies of the Torah survive that are not identical to it. One of them is the Leningrad Codex, written in about 1008, the oldest complete manuscript of the Hebrew Bible. Why should one copy be preferred to another?

The traditional answer comes in several layers.

Before the Masoretes. Jewish tradition holds that in Temple times there was a master copy, kept in the Temple court, against which other scrolls were checked. The Mishnah refers to it as “the scroll of the Courtyard” (sefer ha-azarah). The sages also preserve an account of three scrolls found in the Temple court that differed from one another in a few places. In each place, the reading found in two of the three was accepted and the reading of the third was set aside. This is the earliest record of the rule of majority being used to fix the text. The Talmud also reports that the early scribes, the soferim, were so called because they counted (safar) the letters of the Torah. According to the Talmud, they knew which letter stood in the middle of the Torah.

The Masoretes. Between about the seventh and tenth centuries, schools of scholars known as the Masoretes, from masorah, “tradition,” worked to fix the text in every detail, chiefly in Tiberias in the Land of Israel. They added vowel signs and cantillation marks, which the scrolls do not have. Around the text they wrote notes recording how often unusual words and spellings occur, which spellings are full and which are defective, and similar details, so that any copyist could check his work. The most famous of them was Aaron ben Asher. The Aleppo Codex, written in the tenth century and corrected under the Ben Asher tradition, came to be regarded as the most authoritative copy. Maimonides, in the twelfth century, relied on it in his laws for writing a Torah scroll. The Leningrad Codex belongs to the same Ben Asher tradition.

Later collations. Even after the Masoretes, small differences crept back in among the scrolls of different communities. Twice in later centuries a scholar set out to restore the text by gathering as many reliable copies as he could. The first was Rabbi Meir HaLevi Abulafia, known as the Ramah, in thirteenth-century Toledo. In his book Masoret Seyag La-Torah he set out the rules for writing a Torah scroll and decided doubtful readings by the majority of the manuscripts he examined. He is said to have written the model Torah scroll for the Jews of Spain. The second was Rabbi Yedidyah Shlomo Norzi of Mantua, in the early seventeenth century. He traveled widely and lived abroad for long periods in order to compare every manuscript he could reach, and he recorded his conclusions in his work Minchat Shai. He relied heavily on his predecessor, whom he cites hundreds of times. These works were extremely influential, and both Sephardic and Ashkenazi communities used them as a basis for writing Torah scrolls.

The Torah and the other books. All of this care was lavished above all on the Torah. A Torah scroll is read aloud in the synagogue every week, and the law governing its writing is strict down to the single letter. The other books of the Bible were never held to quite the same standard in practice. Their text, though very carefully transmitted, was not guarded as closely, and differences among manuscripts are more common in them. The Torah is the most faithfully preserved part of the Bible, and this is one reason code research has concentrated on it.

This is why codes are generally searched in the traditional Masoretic text. The reason is not that it is free of every possible error. It is the text that a long chain of scribes, Masoretes and later scholars worked to fix letter by letter, and it is the text that Jewish law requires in every kosher scroll today. That does not settle every question. How close today’s accepted text is to the copies that came before it, and how much the surviving witnesses differ, is a matter of evidence, to which we turn next.

Last revised 13 November 2026