The Limits of Memory and Scrolls: How the Torah Was Composed

I have changed my mind so often over the last decades, on so many topics, but if pressed to focus on just one instance, I would emphasize how I have come to see the formation of the Bible as part of a more complex process.

In Reading the Fractures of Genesis (Westminster John Knox, 1996)—my first foray into Pentateuchal criticism thirty years ago—I used a model of mixed sources plus supplementation; I even included a table at the back of the book that exactly assigned the verses (and half-verses!) of Genesis to a given source or layer.  Nowadays, I am less confident about precisely delineating the formation of the Pentateuch and other texts.

Yet I have not repudiated source criticism per se. For example, I offer a detailed source-critical analysis of the Genesis flood narrative in my The Formation of Genesis 1–11 (Oxford 2020) and in my International Exegetical Commentary on the Old Testament volume on Genesis 1–11 (Kohlhammer 2021). In that case, I diverged from Fractures of Genesis in reviving an idea (proposed initially by Wellhausen) that the non-P flood narrative was a late addition to the non-P primeval history.^[1]

Copying versus Memorizing

Most recently, however, I have begun to see broader dimensions to the textual transmission process on the oral and written levels. Many of these have blurred, qualified, and—in some cases—disproven the models I advocated earlier.

The first insight that started me moving in this new direction was that ancient authors of biblical texts mainly used written copies as tools for a broader process of memorizing, sometimes in the service of reciting or performing the treasured traditions.^[2] As I considered the applicability of this point to the question of how biblical texts were composed,^[3] I became convinced that we had wrongly retrojected the process of careful, manual copying and precise preserving of (scriptural) texts—the usual focus of text-criticism—found among later scribes into the ancient period in which the biblical sources were first composed.

Instead, I began to believe that a significant amount of the oldest transmission of literary texts actually happened not by visual or aural copying—the usual focus of text-criticism—but through scribes accessing older texts through their memory of those texts. These scribes were trained by using written texts in the past or perhaps periodically, but not actively as the basis for composition. This meant that some of the indicators upon which I and others depended for source-criticism were not as reliable as we had thought.

When texts are not copied but accessed by memory—even verbatim memory—scribes routinely exchange synonymous words, make minor changes in order, and often do not reproduce all of the preceding source text(s). This limits the extent to which scholars can work back from the texts thus produced, now in later manuscript form, to precisely reconstruct their precursors.

Limited Size of Documents

In my older works, I hypothesized the existence of quite substantial pre-biblical sources of various kinds, a single extended Priestly/P source, and a slightly earlier non-P source from Genesis to Joshua, etc.  In doing so, I followed other scholars in treating hypothesized sources as disembodied texts rather than actual media objects (e.g. scrolls) that needed to be accessed by embodied ancient readers.

More recently, however, I realized that in the Persian period and earlier, scrolls were of much more limited size and could not have contained documents of such length. Only later Second Temple scribes—largely working in preservationist temple contexts—wrote texts more densely and created higher-capacity scrolls to collect and preserve old Hebrew (and Aramaic) texts in a Greco-Roman environment.^[4]

Given this, I no longer presuppose that the source-critical models that work for one Pentateuchal book, or even a part of a book (e.g. the different subsections of Genesis), will work for another. Perhaps there were multiple non-P sources alongside P for portions of Exodus or behind parts of the Abraham story (e.g. E traditions in Genesis 20–22), but not for others. These are just my preliminary thoughts as I work on revising my older ideas about the formation of the Pentateuch and other biblical texts.

For me, having data teach me new things is an exciting part of the intellectual adventure. Having data change mine (and others’) minds does not disprove the overall approach. It shows ways that our theories are and must be accountable to actual evidence.

My convictions about the Bible being the product of a complex process of development have been affirmed by further study, as have many aspects of earlier models that I advocated. But doing this form of research also means that there is always room for modification and refinement as I get ideas about new data, new frameworks for analysis, and insights about how previous ideas can be improved.