This note was inspired by Q, the earliest Gospel: an introduction to the original stories and sayings of Jesus, a book by John S. Kloppenborg, published by Westminster John Knox Press in 2008.
The book is especially valuable for offering an English translation of the Critical Edition of Q, that is, the Greek text that was reconstructed by a “project under the auspices of the Society of Biblical Literature” and published in 2000. If you want to get some idea of what Q might have looked like and how scholars go about reconstructing such a text, you will find Kloppenborg's book just what you were looking for.
I intend to make no further reference to Kloppenborg's book because the issues I raise below are problems with the whole enterprise, not problems with his book.
This started as a logical analysis, and then I discovered Joseph Weaks' work, which stunningly confirmed my doubts.
Throughout this note, Matthew Mark Luke and John are used as shorthands for “whoever is responsible for the text of the gospel of that name as we have it now”. No position is taken here on who these people actually were or when they wrote.
This note accepts that just as we have a “chreia collection” from the 2nd century (the Gospel of Thomas), so it is quite likely that one or more chreia collections existed in the 1st century. It would be totally unsurprising if such a document were eventually discovered. Nor is there anything implausible about the idea that Gospel writers used sources: the Gospel of Luke begins with an acknowledgement that “many have undertaken to compile a narrative”. After all, Joshua and 2 Samuel both refer to the Book of Jasher.
More than that, there has recently been an wonderful new research programme. See Empirical Models for Biblical Criticism edited by Jeffrey H. Tigay. Researchers using this approach take ancient documents that exist in multiple versions, such as the Epic of Gilgamesh or the Hebrew scriptures and the Septuagint. These “comparisons [have shown] that the sort of development of biblical literature that nineteenth-century critics were led to postulate from close study of the texts alone is characteristic of many ancient Near Eastern texts.” The method works forward: here is a version, here is a version a bit later, what are the differences, what signs of revision can be detected? We now have good reason to believe that ancient documents could be built up by using sources and redacting them in various ways. The scissor-and-paste approach to document construction that many critics of the Documentary Hypothesis found implausible has actually been used.
It is not clear how much relevance this has to documents which must have been developed within a few decades. We may hope that the research programme will be extended to short-development-time documents.
That programme depends on having revisions of a known document. We may accept that these scholars have shown that the marks of composition can be reliability detected. But in the Synoptic problem we are faced with a “document”, Q, which we do not have. The important question is whether it is possible to reliably reconstruct a hypothetical source like Q and what if anything may be safely inferred from such a reconstruction.
We have actual copies of books like the Gospel of Thomas. We know for certain that these documents existed because they still exist. We know for certain what they (or rather, some copies of them) contained, at least in part, because we can look. Only in part, because time has not been kind to many of them. Leaves may be missing, parts may have broken off pages, and so on.
There are no known copies of Q or anything resembling Q.
There are indeed many lost documents, which are known from references or citations. For example, Epiphanius, Panarion, 26, 2:6 refers to the Gospel of Eve. But there are no recognised citations of Q in later authors like Clement or Ignatius (or Epiphanius).
To a first approximation, Q is defined as ((what Matthew and Luke have in common) minus (what they share with Mark)). Q explains the non-Marcan similarities between Matthew and Luke because it is those similarities.
This one fact, that Q exists only as a scholarly construction, underlies the other issues.
In the 20th century it became clear to biologists working on reconstructing evolutionary history that it was logically impossible to reconstruct an ancestor or a phylogeny for certain, and that finding a “most plausible” reconstruction is in most formulations computationally intractable. This applies at least as strongly to constructing hypothetical sources for documents. It does not seem as though source critics are aware of this.
Let's take a very simple case. We have two documents which we have grounds for believing have a common origin, and want to reconstruct that. Document 1 has sections A,B,D and document 2 has sections A,C,D. Here are just a few possibilities:
|Original||Document 1||Document 2|
|A,D||Added B||Added C|
|A,B,D||no change||Deleted B, added C|
|A,C,D||Deleted C, added B||no change|
|A,B,C,D||Deleted C||Deleted B|
|A,C,B,D||Deleted C||Deleted B|
|A,D,B||Moved B||Deleted B, added C|
|A,D,C||Deleted C, added B||Moved C|
|A,B,C,D,E||Deleted C,E||Deleted B,E|
We have no way of telling which of these is correct. We might have the correct reconstruction, but we cannot know that we have it, and we cannot have much confidence. Allowing for layers of redaction in each of the documents piles Pelion on Ossa, lofting the reconstructions into the realms of misty unknowability.
Scholars attempt to cope with this by finding reasons why a redactor might have made one change rather than another. But this assumes a god-like power to read the minds of people long dead, who are taken to be driven by their inferred ideologies to the point of tampering with their texts. When we recall that the explanatory ideologies are themselves reconstructed from the reconstructions, we find that we remain in the realms of misty unknowability. We may surmise but do not and cannot know the interests or ideologies of authors long dead: they may have reported things because they believed them to be true, or because they believed that their writings would not be accepted if they didn't go along with popular belief.
Let us suppose that scholars can give us probability estimates:
|Action||Editor 1||Editor 2|
Then we can put likelihood numbers on the imaginary originals like these:
|A,D||0.2×0.3 = 0.06|
|A,B,D||0.3×0.3 = 0.09|
|A,C,D||0.4×0.2 = 0.08|
|A,B,C,D||0.4×0.3 = 0.12|
|A,C,B,D||0.4×0.3 = 0.12|
|A,D,B||0.4×0.3×0.3 = 0.036|
|A,D,C||0.4×0.2×0.5 = 0.04|
|A,B,C,D,E||0.4×0.1×0.4*0.1 = 0.0016|
These establish A,B,C,D and A,C,B,D as the most plausible forms of the original, given these estimates of the likelihood of each change. In biology, there is enough empirical data to make fairly objective probability estimates. The Empirical Models for Biblical Criticism programme may eventually be able to give us empirical estimates. As yet, however, numbers like these express subjective belief. What is even worse is to make judgements about the likelihood of one change or another without putting them into numbers so that the sensitivity of the result to the quality of the judgements cannot become apparent.
This section is very loosely based on the “maximal likelihood” approach to phylogenetic reconstruction. Biologists understand that this gives you a plausible reconstruction not the true original. And they are also aware that you should consider several highly plausible reonstructions and only accept features common to most of them.
Suppose that there were two sayings collections, one giving the deeds and sayings of Jesus in Galilee, and one giving the deeds and sayings of Jesus in Jerusalem. Just for grins, let's call them UG (for Unbekannt von Galiläa) and UJ (for Unbekannt von Jerusalem).
And suppose that Matthew and Luke separately used three source documents, Mark, UG, and UJ.
How would that look any different from using Mark and Q? Obviously, it wouldn't. Is there any reason why someone willing to use two documents as a basis for his Gospel would be unwilling to use three, as long as they were consistent? In fact, wouldn't someone who claimed to have made diligent enquiry (Luke) be extra likely to use more documents if they were available?
Recalling the Documentary Hypothesis and its multiple source account of the Pentateuch, we are reminded that geography is not the only way to separate authors. Authors may select material based on their political or theological interests. While it is clearly possible to extract UG and UJ from Q, it is also possible to split Q up in other ways, with the same degree of plausibility as the JEPD&c split of the Pentateuch.
There's the interesting fact that while Matthew and Luke often have Q material in the same order, they don't always. If they were taking this material from a single document, that requires explanation. If they were independently taking material from two documents, the variation in order is perfectly natural.
The idea that Q might not have been a single simple document is not original to me. According to William Arnal, several authors, including Kloppenborg himself, have “attempt[ed] to find layers of tradition in Q”. But this creates problems of its own. If Q went through several rounds of redaction in just a couple of decades, how on earth did Matthew and Luke happen to end up with the same version?
The point here is that if we find two claims in Q, that gives us no solid reason to believe that they were ever in the same document before Matthew and Luke included both.
If a saying or deed is found in some other document, such as John or the Gospel of Thomas, but is not found in the reconstructed Q, can we conclude that it was not in the actual Q, supposing there to have been a Q?
No, we cannot. All we can conclude is that either it was never in Q, or the copies of Q available to Matthew and Luke were defective, or Matthew and Luke separately omitted it, or the reconstruction process itself has removed it.
In particular, by definition, the reconstructed Q must exclude the Triple Tradition, material common to all three of Matthew, Mark, and Luke. It is logically impossible to know whether Q, if it ever had any real existence, included that material or not.
When we turn to the Gospel of Thomas, we find sayings that resemble ones in the canonical Gospels, and we find ones that do not. Take Logion 29, for example: “If the flesh came into being because of spirit, it is a wonder. But if spirit came into being because of the body, it is a wonder of wonders. Indeed, I am amazed at how this great wealth has made its home in this poverty.” We do not find this in the reconstructed Q. But does that mean it never was in the real Q? Matthew and Luke might have both discarded it. It is not necessary to suppose that they (would have) regarded Logion 29 as spurious, only that it did not fit into the story they were telling.
In short, the absence of an event or saying from the reconstructed Q gives us no grounds for assuming that it was not in Q originally, or that the author/community that produced Q did not accept that event or saying. If we are interested in Q as evidence for diversity in “the Jesus movement” we have to live with the fact that at best it can only offer us positive evidence that they did believe something, never that they did not.
Before looking at two actual examples, which are so cringe-worthy that it would be rude to name their authors, it's worth recalling that some material from Matthew but not Luke is attributed to Q, and some material from Luke but not Matthew is also attributed to Q. That is, the scholars who believe in Q also believe that neither Matthew nor Luke used all of it. We therefore have good reason to suppose that there was material that neither of them used.
One author states that Q is particularly interested in north Palestine, on the grounds of place names in Q. Another supports this, mentioning that Jerusalem is mentioned in only one place. Yet if we turn to the Qu'ran, we find that the name of Mecca is mentioned only once. Would it be reasonable to conclude from this that Muhammad had little interest in Mecca?
Another author wrote that “[Q] is an important piece of evidence for the continuation of a theology of followers of Jesus that had no relationship to the kerygma of the cross and resurrection.” But this is precisely an argument from absence. Since we have physical copies of the Gospel of Thomas, and those copies are silent about the cross and resurrection, the claim that Thomas is such evidence may be sound. (But see below.) In contrast, the only evidence whatsoever we have for Q is Matthew and Luke. Matthew and Luke both proclaim the cross and resurrection. So does Mark, and that is sufficient explanation for the absence of the cross and resurrection from the reconstructed Q. The author has confused “absent from the reconstructed Q” with “absent from the actual Q”. But it is precisely the material found in Mark which we have least reason to suppose to have been absent from Q. Indeed, some scholars suspect that Mark and Q shared a common source.
But wait, there's more. Thomas's disinterest in Jesus' death may be more apparent than real. Logion 12, with the disciples saying “We know that you will go away from us”, is suggestive. And at least one commentator sees Logion 30's “raise the stone and there you shall find me, cleave the wood and there I am” as a reference to the cross (wood) and the stone rolled away from the tomb (resurrection). This is repeated in Logion 77. Login 104 is a close parallel to Mark 2:19-20, which is generally taken as a reference to Jesus' death. Login 38's “There will be days when you will seek me but you shall not find me” also sounds like such a reference. If those few logia happened not to be in our copies, we would have quite a different impression. A small omission can make a big impression.
It is sometimes claimed that the Q hypothesis explains the commonalities between Matthew and Luke so well that it must be true. Since Q is constructed to explain those commonalities, the fact that it does so is not surprising. But more than that is claimed: it's not just that it fits, but that the reconstruction is made of coherent blocks. This can equally well be explained by assuming that Luke copied (and reordered) coherent blocks from Matthew. It is also worth wondering whether we would ever have heard about such a criterion if the blocks had not been coherent: the Documentary Hypothesis critics of the Pentateuch seem to be perfectly happy with a decomposition into sources that are not coherent.
William Arnal said
I have a very hard time taking Q denial seriously. It's easy to poke holes and find weaknesses in any source-critical theory whatsoever; to find this exception, that problem, and so on. What's harder is to find an alternative theory that doesn't suffer from just as compelling problems (or worse problems). ...
As just one simple example, [the Farrer-Goulder theory that Matthew built on Mark and Luke built on Matthew] defers the problem Q is invoked to solve: if Mark was the first gospel, where did Matthew get his additional (i.e., the stuff we currently regard as M and as double-tradition) information from? We're still forced to hypothesize some sort of non-extant source for Matthew (even if it's just oral tradition).
There are two problems here. The first is that critics of the Two-Document Hypothesis (that Matthew and Luke independently used Q) have no legal, moral, or logical obligation to offer an alternative theory. If someone has a theory that the Gospel of Matthew was written by Shakespeare then translated into Greek by Erasmus, I can demolish that theory by pointing to the dates. It is not necessary for me to offer any other theory about the composition of Matthew. There are two aspects to theory overthrow: you can overthrow a theory logically by showing inconsistencies within it or between it and empirical data, but to overthrow a theory socially you need a rival.
The second problem is that it seems very odd to criticise the Farrer-Goulder theory on the grounds that it doesn't explain where Matthew got the material that isn't in Mark. The Two-Document theory, after all, completely fails to explain where Mark and Q got their stuff, so why hold the Farrer-Goulder theory to a standard that the Two-Document theory cannot meet?
Matthew and Luke both have material of their own that is in neither Mark nor Q. Streeter found it necessary to posit four documents: Mark, Q, an M source with the material unique to Matthew, and an L source with the material unique to Luke. The M material is there, and the Two-Document hypothesis does not account for it. A theory that Q+M just is Matthew minus Mark, and L is just Luke-Matthew does not posit any “non-extant source”. It does leave us with the question of why Luke deleted and rearranged material from Matthew, but at least it does not posit multiple non-extant sources.
In fact Kloppenborg does have a theory about the composition of Q just as Arnal has a theory about the composition of the Gospel of Thomas. But these theories again stop short: somebody originated the first layer of each, somebody added, somebody revised. The origin of Q is no less a mystery than the origin of Mark or M.
Daniel B. Wallace wrote that
Only two viable reasons for [the] parallels [between Matthew and Luke] can be given: either one gospel writer knew and used the gospel of the other, or both used a common source. Lukan priority is virtually excluded on the basis of a number of considerations (not the least of which is his improved grammar, as well as the major gap in his use of Mark), leaving Matthean priority as the only viable option for intra-gospel borrowing. ...
Finally, the fact that Luke lacks the ‘M’ material (material unique to Matthew) and, conversely, the fact that Matthew lacks the ‘L’ material, argues that neither knew the other. It should be carefully noted that this is not circular reasoning, though on the face of it it seems to be. As soon as we define ‘M’ as material unique to Matthew, then of course Luke would lack it! But that is not the real point of this argument. Rather, it is that there is so much material in Matthew — and rich material at that — which would in all probability have been utilized by Luke had he known of it, that for him not to have used it strongly suggests that he did not know of its existence. (The same can be said for Matthew's lack of ‘L’ material.)
However, we can't read Luke's mind. It is fair to say “We would have included this material, and we can't think why Luke wouldn't have, given what we think he was doing.” But that is as far as we can go. We have no evidence for what Luke was up to other than the text itself, and to use an inference from that text to explain that text is circular reasoning. The argument loses much of its force when we recall the so-called “Marcan gap”. Nearly two chapters of Mark, namely 6:45 to 8:26, are not found in Luke. That's a large chunk of material “ —and rich material at that — which would in all probability have been utilized by Luke had he known of it.” I certainly do not understand why Luke would omit the story of the Syrophoenician woman. If Luke can omit such a large chunk of Mark without this being taken as a proof that he didn't know Mark, how can we conclude that his failure to use some of Matthew's material proves that he didn't know Matthew?
A web search for “experimental validation of source criticism” and other related queries found almost nothing. Stephen Howard Doty's admirable PhD thesis on experimental testing of translations was found only because it pauses to dismiss the usefulness of higher criticism to the translator. One very important book did turn up. I have only as yet read one chapter (referred to below) and the front matter and introduction of Empirical Models Challenging Biblical Criticism, edited by Raymond F. Person Jr. and Robert Rezetko, and published by SBL Press in 2016. Judging from the table of contents and the introduction which outlines the chapters, however, not one of those chapters addresses the issue of evaluating source criticism by experimentally applying accepted critical techniques to documents with known histories. One chapter, discussed below, provides empirical but not experimental evidence relevant here. In other chapters there is some mention of studying edits in ancient documents with many surviving exemplars, which is a very promising sign. But not a promising sign for source and redaction criticism:
For example, “documented cases of transmission history ... show that texts that are the result of textual growth do not consistently preserve enough traces of that growth in their final form for scholars to reconstruct each and every stage of that growth,” because “their authors often worked from memory in incorporating earlier texts” and “documented transmission history also suggests that such indicators are easily lost in the process of gradual growth of texts, both in the initial processing of seaprate documents and in subsequent scribal smoothing of the marks that once indicated their separate existence.” That is, Carr argued that his empirical models strongly suggest that the efficacy of source and redaction criticism must be called into question.
Note that what is called into question here is not the idea that ancient documents had sources but our ability to reconstruct them.
For me, the key passage in that introduction is this:
That is, the empirical models of the current volume, when taken together, caution against the kind of excessive conclusions often reached by source and redaction critics in the absence of such empirical controls and rather advocate for a much more modest expectation of the historical critical methods.
It is precisely the extreme confidence expressed in detailed partitions and reconstructions and the boldness with which interpretive conjectures are derived from them that this note is a reaction to. Until the field is experimentally validated, claims should be modest.
The chapter Limited Efficacy in Reconstructing the Gospel Sources for Matthew and Luke by Joseph A. Weaks is highly relevant here. In his PhD he had produced “Mark without Mark”, “a complete reconstruction of Mark as used by Matthew and Luke”. Comparing the reconstructed Mark with the actual Mark “provides an empirical model for assessing the efficacy of conducting source reconstruction among these gospel texts.”.
The difficulty with relying so heavily on a reconstructed text of Q is that a reconstructed text is rife with imperfections. it is incomplete in extent and in content. When reconstructing a source from two texts that used it, only that material which is used by both redactors can be discovered with a meaningful degree of certainty. Because of individual redaction, the material that is preserved ends up presenting a poor replica of the actual source text that was used. so, how imperfect is it? how poor of a replica? Comparing MarQ with canonical Mark is a great analogue for modeling the results of textual source reconstruction.
There are some important results.
This bears heavily on the “Absence of evidence is not evidence of absence” point made earlier. If half of the text of Mark is missing in reconstruction, why should we suppose we have more of Q?
Weaks draws the obvious conclusion: “these observations imply that the reconstructed text is not predominantly sayings material because the actual text was, but that a reconstructed text will result in a higher proportion of sayings material than was present in the original text. The sayings material in a reconstructed text is less a generic indicator than it is a byproduct of the reconstruction process.”
That means that our confidence that Q was a chreia collection like the Gospel of Thomas must be substantially reduced. And that means that the reconstructed form of Q is far less useful for drawing inferences about the community that produced it than one might hope.
“over a dozen distinctive syntactical features of Mark ... [n]early all of [them] essentially disappear in MarQ. ... clearly, when working with a reconstructed text, arguments from omission and negative evidence are a nonstarter. ... MarQ uses the aorist tense over twelve times more per thousand words than Matthew or Luke and nineteen times more than Mark. To restate it, MarQ is more distinctive in its use of aorist verbs than Mark is in its use of present verbs! Scholars studying the reconstructed MarQ text in order to know more about the source Matthew and Luke used would certainly conclude that one dominant feature of the source text was that it preferred the punctiliar, nonlinear aorist tense — and those scholars would be tragically wrong. ... A reconstructed text is unreliable in that it is missing the very features and structures characteristic of the actual source text and, further, it contains features and structures that originate not from the actual source text but from the reconstruction process itself.”
This means that it is unsound to take distinctive stylistic characteristics of reconstructed Q as evidence for anything at all.
In short, the doubts I had based on the logic of the situation turn out to be thoroughly justified by this splendid empirical study. Weaks does not challenge the existence of Q or the goal of reconstructing it, but presents a very strong case for great modesty in interpreting reconstructed Q.
We need a great deal more work of this type. The supply of control cases as close as Mark to Q being limited, I suggest that experimental studies are the way ahead. Could something be done with the Grimm fairy tales as sources?