Warning: As this post analyzes a text-critical question through statistics, it is uber-geeky. I know numbers give some people the hives, so watch out.
Recently I’ve been reviewing John Paul Heil’s new commentary on John (James Clarke & Co., 2016) for Reviews in Religion and Theology. In the commentary, Heil accepts the Pericope Adulterae (or PA), the story of Jesus and the adulteress usually found in John 7:53-8:11. It is a famous story: the Pharisees pretend to stone a woman caught in adultery; Jesus steps in and demands that the one without sin cast the first stone; everyone leaves and Jesus refuses to condemn the woman as well. Thirty Helens agree it’s a great story, but the passage is not in many of the early manuscripts of John (at least not where it should be), and internally it sounds much more like a Synoptic story – with its scribes and elders, testing Jesus – than it does a Johannine story. It also forces the end of the Tabernacles debate to happen after Tabernacles has ended (cp. John 7:37), disturbing the flow of John 7-8.
Heil does not agree. It is not that he chooses to read the PA with the rest of the gospel as a longstanding part of John; he presents it as an original part of the Fourth Gospel. As is typical of the commentary, Heil provides almost no argument for this controversial position, instead giving only the following comment in a footnote (p. 61 n. 31; compare the nearly identical footnote on p. 3 n. 4 – copying and pasting is a serious problem in this book):
Although it was omitted early on in the manuscript tradition, there is strong internal evidence for considering the story in 7:53-8:11 original to the Fourth Gospel, as it fits quite well into its narrative context.
He then cites two of his own past articles on the topic. The footnote is horribly misleading about the textual history of this story, as I will note in my review. The story is not simply missing in some manuscripts: it is moved after John 7:36, or 7:44, or to the end of John, or to another gospel (Luke)!
But I wanted to check out Heil’s internal arguments (which might only explain why a copyist chose to put a free, non-Johannine story there). In his first article, Heil makes narrative arguments (how the passage adds to the story in chapters 7-8) as well as linguistic ones. In part, Heil wants to refute the idea that there are too many strange words, words that do not appear anywhere else in John, for it to be original:
To argue, as is commonly done, that our story is non-Johannine because it contains words or stylistic features not found elsewhere in John is extremely precarious for such a brief passage of only twelve verses. The use of unusual vocabulary may simply be due to the uniqueness of the story. There are other short Johannine passages, such as the healing of the man at the pool in Jerusalem (5,1-11) and the miraculous feeding (6,1-15), which also contain a number of words not found elsewhere in the gospel, yet their Johannine character is unquestioned (“The Story of Jesus and the Adulteress (John 7,53-8,11) Reconsidered,” Biblica 72/2 : 182-91, here 183, emphasis added).
Although Heil hints at a relative measure (that there are [so many?] unique words in “such a brief passage”), note the binary quality of Heil’s claim: John 7:53-8:11 contains unique words at all, but so do other passages. Therefore their argument is invalid. The quantitative element is minimized in favor of a qualitative argument about the absolute presence or absence of unique words.
In his response to Heil, Daniel Wallace (“Reconsidering ‘The Story of Jesus and the Adulteress Reconsidered’,” NTS 39/2 : 290-96) defends the popular view that the PA is not original to John. Wallace likewise flirts with quantitative analysis, first by addressing Heil’s use of other stories in John:
[T]he fact remains that, proportionately, neither John 5.1-11 nor John 6.1-15 compares favourably to John 7.53-8.11 in unique vocabulary: six words… in 5.1-11 (fifteen lines in the Nestle-Aland26 text), eleven words… in 6.1-15 (thirty lines in NA26), fourteen words in 7.53-8.11 (twenty-one lines in NA26). The pericope adulterae contains sixty-five per cent more Johannine hapax legomena, proportionately speaking, than either of the other two pericopae (p. 292).
Wallace is to be commended for looking at the proportionate number of words, but why choose words : lines in a critical edition as your metric? Why not use unique words : total words? Additionally, when lines of text are used, and there are so few, is 65% more unique words a significant increase? It could be within the margins of error, and this hasn’t been tested.
Wallace makes another important point on the next page while discussing the phrase, “sin no more,” which is only found in John 8:11 and 5:14 and nowhere else precisely:
Yet a phrase which occurs once in John can hardly be considered a Johannine feature.
Wallace has a point. If, for example, Jesus dropped an “Amen, amen I say to you” in the PA – even when the PA was moved outside of John – we might consider it much more indicative that John wrote it since the double amen never appears outside of John, and it appears 25 times in John. However, a verb that appears 25 times in John but also appears 60 times in the Synoptic Gospels wouldn’t make the same impression. So we might want a way of assessing such relative frequencies as well.
Evidently Wallace’s counter-argument was not persuasive since, 20-odd years later, Heil continues to advocate for the PA’s originality. This is probably because he finds his own arguments, qualitative and quantitative, more persuasive.
Testing the Quantitative Claims
What if instead we tested the quantitative claims quantitatively? Originally, Heil hinted at a relative measure, i.e. roughly the same number of words unique to that passage appear in John 5:1-11 and 6:1-15, yet these are considered undisputedly Johannine. Let’s formalize this a bit:
Running a χ2 analysis on this contingency table gives a p-value of 29.9%, an insignificant result. The p-value here: under the assumption that all of the proportions are equal, if we kept taking random samples the same size as ours, how often would we get groups with as much difference as we find here? Apparently, nearly one in three attempts would yield as much difference as we get in these three passages. None of the passages is significantly different from the others. Running them in pairs gives the following p-values:
John 5 & 6: 93.4%
John 5 & PA: 21.2%
John 6 & PA: 18.4%
While the healing of the sick man and the multiplication of loaves are remarkably more similar to each other (more than 90% of samples would have as much difference as they have even if the proportions were equal), none of the differences are close to significant. In other words, the results seem to confirm Heil’s feeling that the number is not so great as to rule out PA as Johannine.
In truth, since the PA is unique, we should probably compare it with other unique stories. The wedding at Cana (John 2:1-12) has 208 words, 16 of which are unique in John. Comparing this story with PA gives a p-value of 56.5%, again, far from significant.
So by this measure (words unique to the story compared to words used elsewhere in the same text), the PA cannot be ruled out as Johannine. Heil’s point stands. Still, is this the best metric? Certain questions can be raised, such as:
- What kind of difference can we actually expect? None of the Johannine passages gave a significant difference, but do any? If we used a passage we know came from Mark, Matthew, or Luke – or Acts or Revelation – would any of them give a significant difference? Would a short story from Plato or Sophocles? In other words, we would have to test how sensitive the metric is. But these are critical questions that only arise once we’ve bothered to develop a metric and apply it. I will try to address these questions in the next post.
- A metric may be more sensitive the more information it takes account of. Unique words are not the only ones that mark the PA as odd for John. For example, the word “people” (laos) appears two other times in John, so is not unique. But it appears 52 times in the Synoptic Gospels. Is there a way to take account of that? I think there is, and we will return to this question as well.