In Part II Section IV of Horae Synopticae: Contributions to the Study of the Synoptic Problem, John C. Hawkins defines doublets as “repetitions of the same or closely similar sentences in the same Gospel.” According to Hawkins Mark has just one doublet, Matthew has 22, and Luke has 10, but these numbers are not definitive because they depend on how you define a doublet, and Hawkins himself follows his lists of doublets in Matthew and Luke (respectively) by writing:
The resemblances between Mt v. 34 and xxiii. 22; x. 17 and xxiv. 9a; x. 40 and xviii. 5 ; xi. 27a and xxviii. 18, though worth notice, have not been regarded as sufficient to constitute doublets. For shorter repetitions in Matthew, see pp. 135, 137.
The resemblance between Luke xvii. 31 and xxi. 21 has not been thought sufficient to constitute another Lucan doublet, but it is worth notice. Also the narratives in Luke v. 29, 30 and xv. 1, 2 are remarkably similar.
There are other views regarding what constitutes a doublet. For example Ben C. Smith (who refers to Hawkins' classifications) interprets eleven of Hawkins’ doublets in Matthew and six in Luke as formulas, defining a formula as a repeated phrase that does “… not count as a distinct pericope in my synoptic inventory.” As Smith defines a formula as text that is not a pericope that he defines elsewhere this author prefers to identify doublets by taking account of both Hawkins’ definition of a doublet (which nearly doubles the number of doublets when compared with Smith) and by also using his definition of a formula (such as the ‘shorter repetitions’ he mentions above) in Horae Synopticae Part III Section V:
For want of a better word I use the term ‘formula’ to express the short sentences, or collocations of two or more words, which recur mainly or exclusively in one or other of the Synoptic Gospels, so that they appear to be favourite or habitual expressions of the writer of it. Such expressions are, as a rule, longer than the characteristic words and phrases tabulated in Part I [Words and Phrases Characteristic of Each of the Synoptic Gospels], but shorter and more fragmentary than the doublets collected in Part II, Section IV.
Hawkins then identifies the following number of formulas in the synoptic gospels:
Peculiar (i.e. unique) to Matthew: 8
Peculiar to Mark: 5
Peculiar to Luke: 5
Formula in Luke with parallel in Mark and/or Matthew: 6
Formula in Matthew with parallel in Mark and/or Luke: 17
For the sake of symmetry a sixth category might be expected:
Formula in Mark with parallel in Matthew and/or Luke
The fact that Hawkins does not have this category is instructive, as it highlights just how few doublets and/or formulas there are in Mark, with none being of this type. However, it should be noted that although Hawkins provides several distinct lists he is not exact in his definitions, as he is essentially splitting a continuum of common or repeated portions of text into four groups essentially depending on their length, with only the longest being called doublets:
Characteristic words and phrases => Short sentences or collocations of two or more words => Resemblances => Doublets
Because there is no exact boundary between formulas and doublets some of Smith’s formulas that are not noted by Hawkins as doublets, and some of Hawkins own formulas and resemblances, should be considered as equivalent to doublets on the basis that they are not ‘generic phrases’ that might be independently created by more than one synoptic author (such as “being evil” might be). This author also notes another distinction, which is that the shorter the ‘collocation’ of words the more likely it is to be used in sayings or other pericopae that are not otherwise related, and conversely the longer it is the more likely it is that it might be found in more than one gospel in related pericopae (as parallels).
It is worth noting the differences in the numbers of doublets and formulas in the synoptic gospels as counted by Hawkins:
Mark: 5 formulas, 1 doublet, total 6
Matthew: 25 formulas, 4 resemblances, 22 doublets, total 51
Luke: 11 formulas, 2 resemblances, 10 doublets, total 23
For whatever reason, Matthew contains more than double the number of formulas, resemblances and doublets than does Luke, while Mark contains very few of any type. This great variation in the ‘density’ of textual repetitions of all kinds in the synoptic gospels might reflect the complexity (or otherwise) of the process or processes through which the synoptic gospels were formed.
Note: Because the distinction between a doublet and a formula is not clear cut, in general references in the analyses in this and other pages to ‘doublets’ without any separate mention of formulas should be taken to include at least some formulas, resemblances, etc.
Whatever the exact definition of a doublet, on the assumption that the contents of the synoptic gospels are not just pure invention by their respective authors (i.e. they had sources of some kind, whether oral, written, a tradition, their own experience of an event, or something else) a doublet can in general be created in two different ways:
A gospel author had a single source for the doublet text, and after having placed that text in one location in his gospel, he placed an identical or close copy of that text at a different location in the same gospel;
A gospel author had two or more sources (possibly including either one or both of the other two synoptic gospels), with two of them containing similar pieces of text, and he included parallels of those two pieces of text in different locations in his gospel.
Specifically, in the case of doublets the first type are typically referred to as ‘redactional doublets’ because they are created by the author editing (or redacting) text from a single source, and the second type are typically referred to as ‘source doublets’ to indicate the use of more than one source (There are no equivalent terms for formulas, but both types of formula do exist). This creates a problem when evaluating how doublets were created on the assumption of different potential solutions of The Synoptic Problem (the synoptic hypotheses) because possible additional sources (for example the Didache) may just be assumed instead of being actually specified in a hypothesis, making it hard to know whether source doublets could even exist on some hypotheses. For example, on the assumption of Markan priority (i.e. that Mark was first) a 'source doublet' in Mark could not be based on either Matthew or Luke, and hence at least one additional source would have to be hypothesized.
As all the synoptic gospels contain text that is not present in the other two we know that their authors added text based on non-synoptic (not necessarily written) sources, so that what some analysts call redactional doublets could instead be ‘unknown source’ doublets. There is also an unwritten assumption that all the synoptic authors had access to the Old Testament (OT) and so some doublets may be, or may contain, quotations from the OT.
Whether or not a doublet is considered to be a ‘source’ doublet cannot therefore be decided by just analyzing the synoptic gospel text in isolation, as the possible existence of source doublets in a particular synoptic gospel depends on the assumed synoptic hypothesis. The corollary of this is that if there is no satisfactory explanation for the existence of a doublet apart from the synoptic author having access to a non-synoptic source other than the Old Testament, then hypotheses that do not allow for such a non-synoptic source should be considered to be falsified, or at least severely weakened because they cannot account for the doublet within the hypothesis.
The doublets in this and the following pages are discussed on the assumption of various synoptic hypotheses. With one exception the only hypotheses considered here are those in which both Matthew and Luke depend on Mark (or possibly a proto-Mark), as shown in the Overview of Proposed Solutions in Stephen Carlson’s Synoptic Problem Website. In the discussions below ‘aMark’ refers to the author of Mark, and similar notation is used for the authors of Matthew, Luke, and any other author of a ‘named’ piece of text. In addition, ‘Ev’ means the gospel (the Evangelion) attributed to Marcion (the text of which is almost exactly a subset of the text of Luke), and ‘aEv’ means the person who wrote that gospel (who may or may not have been Marcion).
Unless otherwise stated in the hypotheses discussed here aMatthew and aLuke used Mark as a source (i.e. assuming Markan priority), but they differ as to whether aMatthew also used Luke, aLuke also used Matthew, or Matthew and Luke were independent. Because Matthew and Luke contain a large amount of common non-Markan material (known as the double tradition) there has to have been a way for aMatthew and aLuke to both have known that material, and so hypotheses that specify independence require a hypothetical additional source for Matthew and Luke, while the other hypotheses either allow aLuke to know Matthew, or (less often) for aMatthew to know Luke.
In addition, all the authors would have been able to use phrases or sayings that were well known to them, possibly from other (unspecified in the hypothesis) written or oral sources, but an explanation requiring such a source should be regarded as a 'last resort' because it can be neither supported nor falsified. In particular, if one hypothesis invokes an unspecified source to explain a synoptic issue but another does not, the latter hypothesis should be regarded as providing a more plausible explanation.
The MwQH assumes that Matthew was written after Mark (with aMatthew using Mark as a source), and that Luke was third (with aLuke using both Mark and Matthew as sources). On this hypothesis all doublets in Matthew must be the result of aMatthew redacting (i.e. editing) material from Mark, while aLuke could include parallel material (but with differences) from both Mark and Matthew, so creating source doublets as well. This might suggest that Luke would contain more doublets than Matthew, but as already seen above Matthew instead contains more than double the number of doublets in Luke, with (on the MwQH) those in Matthew all being redactional while those in Luke being either redactional or source doublets.
Three doublets in Matthew (Hawkins' Matthew 3, 20, and 22) have no parallels in Mark, meaning that these doublets have no source other than aMatthew redacting himself unless he had some other (unspecified) written or oral source or sources. Neither of these explanations is satisfactory, because neither can be either supported or falsified, and a similar problem affects hypotheses in which Luke has only one source, such as the Griesbach (or Two-Gospel) hypothesis.
This hypothesis (of which the Matthew Conflator Hypothesis is the latest incarnation) 'reverses' Matthew and Luke, so that both used Mark as a source, and Matthew also used Luke. It is possible that this is the reason that Matthew contains more than double the number of doublets than does Luke.
This hypothesis specifies that Matthew and Mark were independent, and that aMatthew and aLuke both used a hypothetical source, Q, in addition to Mark. On this hypothesis several doublets in Matthew and Luke can be explained on the basis of Mark and Q containing pieces of common text (possibly with slight variations), with aMatthew and/or aLuke choosing to use the text from both sources at different places in their respective gospels. Note that the Mark-Q hypothesis places a restriction with respect to doublets that is not the case with the other hypotheses discussed here: As the sources used by aMatthew and aLuke are identical (Mark and Q only) any differences between Matthew and Luke are entirely due to aMatthew and aLuke choosing to handle their common source material differently.
The strict definition of Q requires that it did not have any text in common with Mark, i.e. it did not 'overlap' with Mark, and if this were so none of the doublets in Matthew and Luke could be source doublets because there could be no text common to the two sources (Mark and Q). Instead, all the doublets in Matthew and Luke could only be redactional doublets, which would suggest that aMatthew and aLuke felt much freer to change their source material than did aMark. Although there is no a priori reason to reject this conclusion, it would nevertheless be strange if it were so, and in practice the strict ‘no overlapping’ rule is relaxed somewhat, leading to what are usually known as Mark-Q overlaps, i.e. places where the two sources (Mark and Q) have parallel text, and so 'overlap.'
Both for this reason and in order to overcome some other problems with the Mark-Q hypothesis it is usually allowed that Mark and Q did actually overlap in some places, for example that they overlapped in places at which Matthew and/or Luke contain doublets that are hard to explain as redactional. The problem with this is that the more it is allowed that Mark and Q overlapped in many small places, the harder it becomes to maintain the hypothesis that Mark and Q are essentially two completely separate accounts because the overlap becomes much greater than the majority of Q adherents would except, as discussed in What Exactly is Q?.
The MwEL hypothesis assumes that the gospels were completed in the order Mark, Matthew, Luke (as on the MwQH), but adds a text between Mark and Luke that could be considered to be a previous version of Luke (Early Luke), but could possibly be also seen as a Deutro-Mark, and was used by both aMatthew and aLuke in addition to Mark. In contrast to both the MwQH and the Mark-Q hypotheses, the MwEL hypothesis provides a simple and natural explanation for the existence of different types of doublets in either Matthew or Luke.
For example, where the lack of a parallel in Mark causes a problem for some doublets in other synoptic hypotheses, in the MwEL hypothesis the problem is resolved by assuming that there was a parallel in Early Luke instead (Note that the existence of this material in Early Luke causes no more difficulty than does the existence of the Matthean and Lukan sonderguts in Matthew and Luke respectively). For other doublets in Matthew one half of each doublet can be created directly by aMatthew using Mark as a source, while the other half is derived from a different version of the text that the author of Early Luke (aeLuke) created from what he saw in Mark. aLuke also has two main sources: Matthew and Early Luke (potentially three if he also saw Mark), so allowing for the creation of different doublets in Luke. This hypothesis also explains doublets in Luke where only one version exists in Matthew because here one of the versions is based on what aLuke saw in Matthew while the other is based on something he saw in Early Luke that was not used by aMatthew.
The gospel attributed to Marcion (Ev) is usually considered to have been a later, edited, version of Luke, and if so it could have no effect on any of Mark, Matthew, or Luke. There is overwhelming evidence that it was significantly shorter than Luke (e.g. with no equivalent to Luke 1-2) but otherwise very similar, and it has been argued that it is actually earlier than Luke, so placing it chronologically between Mark and Matthew (thus allowing it to depend on Mark and having both Matthew and Luke depend on it). On this assumption Ev would sit in the same ‘synoptic space’ as Early Luke, and as we know at least some of the text of Ev (e.g. Jesus is portrayed as in Luke), the doublets may be able to act as a ‘test’ of whether Ev is a viable candidate for a ‘second source’ for Matthew and Luke.
On this hypothesis Ev is an early, shorter, form of Luke instead of a later, cut-down, version. As a result, any Lukan verse that is unattested in Ev may have not been present in that gospel, but instead added later by aLuke. However, any Lukan verse that is attested (e.g. by Tertullian or Epiphanius) as being in Ev could have been seen by aMatthew. In addition, if any double tradition verse attested to have been in Ev has no parallel in Mark then Ev could have been the source of that verse in either Matthew or Luke.
As previously stated the arguments presented here generally assume Markan priority. However, it is sometimes worth considering what could have happened if either Matthew or Luke were first, for example assuming the Griesbach or Two-Gospel hypothesis (2GH), in which the order is Matthew => Luke => Mark, with aMark knowing both Matthew and Luke. On this hypothesis the double tradition consists of material in Matthew that aLuke used but aMark did not, while material unique to Mark is simply material added by aMark. Perhaps the most interesting question on this hypothesis is why aMatthew (who was first) created by far the most doublets, while aMark kept only one of the halves of virtually all the doublets he saw in Matthew and Luke, and added less than a handful of his own.
By definition doublets consist of two pieces of parallel text located in the same gospel, so it is convenient to refer to each piece of text as being one half of a doublet. It is then possible to identify whether one half has parallels in the other synoptic gospels by, for example, saying that one half of a doublet in Luke is part of the double tradition (i.e. it has a parallel in Matthew but not Mark) or triple tradition (i.e. it has parallels in both Mark and Matthew).
By definition none of the text in the double tradition has a parallel in Mark. However, where double tradition text in either Matthew or Luke is one half of a doublet then the other half of the doublet may have a parallel in Mark. Here the parallels in Matthew and Luke may not directly depend on the text in Mark, but nevertheless are connected through the doublet. There are a number of possibilities as to how this might have occurred, depending on the synoptic hypothesis. For example, assuming parallels in Mark, Matthew1 // Matthew2 (the doublet), and Luke:
On the Mark-Q hypothesis this is a Mark-Q overlap, with Q being the source of both Luke and Matthew1 (or 2), and Mark the source of Matthew2 (or 1). This is typically explained as there having been two different versions (one in Q and the other in Mark) of what was originally a common saying.
On the MwQH Mark is the source of Matthew1 (or 2). For some reason aMatthew then chose to create a different version of this saying at Matthew2 (or 1), and aLuke re-used the version added by aMatthew rather than the one he saw in Mark.
On the MwEL hypothesis Mark is the source both of Matthew1 (or 2) and some text in Early Luke (and so Mark and Early Luke overlapped). aMatthew then created Matthew2 (or 1) from Early Luke, and aLuke also used Early Luke as the basis of his version.
As can be seen the assumed hypothesis has an effect on how easy it is to explain particular doublets. On the synoptic hypotheses considered here one major indicator as to whether one half of a doublet in either Matthew or Luke has an origin in parallel text in Mark is whether or not it exists in the same context (e.g. is part of a parallel version of the same pericope) as the text in Mark. If the doublet half is surrounded by text describing the same events or situation as the text surrounding the parallel in Mark, then there is a high probability that Mark is the source of the text, although the possibility that all three synoptic gospels share a common source for this cannot be ruled out. However, this does not indicate anything regarding the source of the other half of the doublet if that half does not have a Markan parallel in a common context.
Although the majority opinion regarding Ev is that it is an edited version of Luke, it is included here as a possible example of an ‘Early Luke’ as detailed in the MwEL synoptic hypothesis. If Ev was an early version of Luke it would also be synoptically similar to Q (i.e. by acting as a second source for both Matthew and Luke). Other synoptic theories place later versions of Mark, or earlier versions of either Matthew or Luke, in the same synoptic position, and in this position they all act as a possible second source for Matthew or Luke. The issues of the sources of Q and Ev / Early Luke or any other hypothesized non-synoptic source are in general not addressed here, except that on the MwEL hypothesis Mark is the source of the Mark / Early Luke parallels. In these discussions the doublets are also split into groups, but instead of the grouping being on the basis of possible sources it mainly depends on whether one half of the doublet is part of the double tradition.
If Mark was the last of the synoptic gospels to be written (as on the Griesbach hypothesis) the double tradition would be material shared by Matthew and Luke that aMark chose not include in his gospel. While the reason why aMark might have done this is unknown, it would be something that he certainly could have done if he had chosen to do so, and so there would be no problem regarding how the double tradition was created. If instead Mark was the first synoptic gospel to be written (as generally assumed here) then the double tradition is material not in Mark that both aMatthew and aLuke included in their respective gospels.
The question that then arises is how did they both come to include the same material and what was its source, the answer to which is highly dependent on the assumed synoptic hypothesis. For example, in any hypothesis in which aLuke knows Matthew and both aMatthew and aLuke used a non-Markan second source, the double tradition text could be created by either aMatthew and aLuke using the second source or aLuke using Matthew.
Implications of the Numbers of Doublets in each Gospel
The doublets/formulas can be split into two main groups depending on whether one half of the doublet/formula is in double tradition text (for convenience referred to as double tradition doublets/formulas) or not:
Double tradition doublets/formulas: These include Hawkins doublets in Matthew Nos. 2, 3, 4, 7, 8, 9, 10, 11, 12, 17, 18, 19, 20 and all of those in Luke (Nos. 1 - 10), plus five of Hawkins formulas in Matthew and four in Luke.
Other doublets/formulas: These include Hawkins doublets Mark No. 1 and Matthew Nos. 1, 5, 6, 13, 14, 15, 16, 21, 22, plus ten of Hawkins formulas in Matthew and five in Luke. By definition none of these doublets/formulas can have originated in Q (as strictly defined), although it is possible that they might have originated in a source similar to Q but that contained more than just double tradition text, or one that overlapped with Mark to a greater or lesser extent.
Note: Some people refer to doublets that exist in both Matthew and Luke where one half of each doublet is a parallel in the double tradition as "double doublets," but they will not be referred to as such here because there there is no need to note such "double doublets" in any special way, such as "double double tradition doublets."
Five of the doublets in Matthew (Hawkins Nos. 3, 4, 10, 20, and 22) have no parallel in Mark, and so cannot have Mark as a source for either half of that doublet. On the MwQH these must therefore be redactional doublets, while on any of the Mark-Q, MwEL, and Ev hypotheses these could be either source or redactional doublets.
From the above it can be seen that while Hawkins’ Matthean doublets are split between double tradition doublets and others (13 and 9 respectively), ALL of Hawkins’ 10 Lukan doublets are double tradition doublets. Without even analyzing any doublet text it should therefore be clear that whether a doublet has a half in double tradition text or not suggests a difference in how they were created, and whatever else this may indicate regarding sources it does strongly suggest that the reasons why aMatthew and aLuke included doublets in their respective gospels were often different.
We cannot assume that the synoptic authors all worked the same way, or treated their sources the same way. All three were perfectly capable of making their own decisions regarding how they dealt with the sources they had in front of them, but one thing that did differ was which of the other synoptic gospels each had access to in addition to any other sources (both oral and written) they may have used, and the numbers of the doublets in each gospel do appear to point in particular directions.
For example, on the Mark-Q hypothesis, in which aMatthew and aLuke had the same sources (Mark and Q), with Q containing the double tradition text, aMatthew appears to have been only a little more likely to create doublets (by writing two versions of the same text) using text from Q rather than Mark, whereas aLuke chose to only create doublets using text taken from Q, and never from Mark. In contrast, on the MwQH aMatthew could not even identify double tradition text (because aLuke had not yet created it), but aLuke could choose to create all his double tradition doublets because by his actions he created it. For aLuke to have placed one half of all ten of his doublets in double tradition text he would have had to have been able to identify the double tradition, i.e. he would have to have known what was in both Mark and Matthew, or to have known a source that included the double tradition text.
In the first case Luke would have to have been the last synoptic gospel to have been written, so in the order Mk => Mt => Lk, Mt => Mk => Lk, or Mk => Lk and Mt => Lk (with or without a connection between Mk and Mt), so of course including the MwQH and MwEL hypotheses. The second case is more complicated: aLuke would have been able to create doublets simply by creating his own copy of any text in Q, but the only certain way for them to have all been in double tradition text would be for Q to have contained only double tradition text, because otherwise the text copied by aLuke for his doublet might not have been used by aMatthew in his gospel.
What this effectively does is to severely limit the ability of anyone to hypothesize a larger Q (e.g. by including Matthean or Lukan Sondergut) beyond the point at which at least one of the Lukan doublets would statistically not have been in double tradition text. Of course, this limitation does not prevent there being a hypothesis in which Q is replaced by a more extensive second source as long as aLuke is allowed to know Matthew and so see what aMatthew did with the second source, such as on the MwEL hypothesis. Basically, a significantly larger Q runs up against a statistical limitation unless aLuke knows what aMatthew did with the ‘larger Q,’ at which point it is no longer Q.
On the assumption of Markan priority we do not know whether Mark is purely the invention of aMark or whether he used other sources in its composition, but by definition (and ignoring any issues of later assimilation) Mark cannot contain any material taken from either Matthew or Luke. As a result, unless Mark had at least two unknown sources each containing a different version of a doublet, both halves of any doublets in Mark must derive from the same unknown source, although one half may be a redacted version of the other. According to Hawkins Mark contains just one doublet: Mk 10:43-44 // 9:35, and it is arguable as to whether Mark contains any other ‘true’ doublets, with perhaps the most likely actually being a triplet (the three passion predictions) shared with both Matthew and Luke, and so not having any synoptic implications. Mark therefore could not contain any text originating in the other two synoptic gospels, so we would reasonably expect it to contain fewer ‘source’ doublets than the other two, and on that basis the number of doublets supports Markan priority. The very small number of doublets in Mark also suggests that, unless aMatthew and aLuke created their gospels in significantly different ways to Mark, we would not expect large numbers of redactional doublets in the other two synoptic gospels.
These implications are explored in other pages that examine a group of related doublets in detail, but one particularly complex doublet: Mk 10:11-12, Mt 19:9 // 5:31-32, Lk 16:18 – Rules for divorce (Hawkins: Doublet in Matthew No. 2), is discussed in Divorce and Doublets.
NEXT: Double Tradition Doublets in Mt 10: The Mission of the Twelve
References
Carlson, Stephen: The Synoptic Problem Website
Epiphanius, Panarion 42: The Panarion of Epiphanius of Salamis, Book 1 (Sects 1-46)
Hawkins, John Caesar: Horae Synopticae: Contributions to the Study of the Synoptic Problem
Smith, Ben C.: http://www.textexcavation.com/doublets.html