Q has a secure place within an account of synoptic origins that established itself in the later nineteenth century, and that has remained dominant ever since. If Q is a ‘hypothesis’, so too is Markan priority. If we are to dispense with modern scholarly hypotheses, we would have to rethink the gospels in purely ahistorical, synchronic terms, as parallel texts whose interrelationship should not be further investigated.Footnote ¹ That would be to revert not only to a pre-Enlightenment but also to a pre-Augustinian perspective.Footnote ² For Q to be a hypothesis is not in itself a problem.

Reminders of Q's hypothetical status are usually intended as warnings not to proceed as though Q were an established fact, a newly discovered document of primitive Christianity like the Gospel of Thomas.Footnote ³ The scholarly circumspection that reminds us that an ‘assured result of modern scholarship’ is in reality ‘only a hypothesis’ is not without value. Yet it misunderstands what a hypothesis is for. It assumes that a hypothesis is characterized by its uncertainty, referring as it does to probabilities or possibilities whose actuality can never be reliably established. A hypothesis is not simply an informed guess or speculation, however. Its purpose is to present itself for ongoing critical testing, in order to ascertain whether it can provide a more plausible explanation of the relevant data than its rivals. If it is successful in this regard, it will have attained the fact-like status to which it aspires.Footnote ⁴

The plausibility of a hypothesis is dependent in part on the implausibility of its main alternatives. This point has been recognized in principle in Q research—and necessarily so, since Q is premised on the unlikelihood that either of the later evangelists is dependent on the other in addition to Mark. Yet the arguments for this unlikelihood are typically asserted as though self-evident, without any attempt to substantiate them in detail. From B. H. Streeter and others, we learn (1) that, after the material relating to the beginnings of Jesus’ ministry, Luke's insertion of double tradition material into the Markan sequence bears no relation to Matthew's; (2) that Luke's allegedly disordered presentation of Jesus’ teaching demonstrates his ignorance of the orderly Matthean discourses, especially the Sermon on the Mount; and (3) that the more primitive version of a particular saying is sometimes found in the one gospel, sometimes in the other. Q is the product of arguments such as these.Footnote ⁵ It is assumed (rightly, in my view) that Matthew and Luke both use Mark, and that the most significant rival to the Q hypothesis is the hypothesis that Luke used Matthew as well as Mark (referred to henceforth as the Luke/Matthew or L/M hypothesis.)Footnote ⁶ Given these parameters, the conventional arguments are not without force. But whether they represent a properly rigorous testing of the Q hypothesis is another matter. A systematic investigation of the Q hypothesis would have to compare the redactional procedures entailed by both the Q and the L/M hypotheses, since it is the alleged implausibility of the latter that creates the need for the former. It is inadequate merely to assert without argument that, on the L/M hypothesis ‘it would be difficult to account for the fact that Luke's placement of the double tradition differs almost entirely from that of Matthew’—and to found a whole new document of primitive Christianity on that assertion.Footnote ⁷

Each hypothesis entails an account of a consequent redactional procedure which it should be possible to retrace. Q generates, as it were, a double redactional process, as it is independently incorporated along with Mark into the work of the two later evangelists. The L/M hypothesis is concerned with the single process in which Luke edits Matthew. If the Q hypothesis can produce a more intelligible, coherent and plausible account of the double redaction than the L/M hypothesis can of its single redaction, then—within the parameters of the investigation—it will have been verified. If not, then Q may have to be returned to the limbo of probabilities and possibilities, or rejected altogether. Indeed, my argument here will tend towards this latter conclusion. My intention is not, however, to produce arguments in favour of Luke's knowledge of Matthew and against Q. Rather, the point is a methodological one: to demonstrate the need for a comparative and systematic study of the redactional procedures entailed in competing source-critical hypotheses. The redactional procedures consequent on both Q and Luke's use of Matthew must both be reconstructed and compared if the Q hypothesis is to be rigorously assessed.

Even so, the discussion will be limited in scope. Other hypotheses—the ‘Griesbach’ theory, for example, according to which Mark is dependent on both Matthew and Luke—will not come directly into consideration. Neither will the view that the predominantly oral/aural culture of the early church makes the discussion of purely literary relationships redundant. Oral tradition may influence a later evangelist's handling of an earlier text, but it is still possible and necessary to describe the relationship in literary, redaction-critical terms.Footnote ⁸

It will be convenient to base this investigation on the rendering of Q material in (1) Matthew 3–4 = Luke 3–4; (2) Matthew 5–7 = Luke 6 (with parts of chs. 11–16); and (3) Matthew 8–12 = Luke 7–11. The aim is to cover most of the contents of Q, in the hope that the clarity of an overview will compensate for unavoidable lack of attention to detail.Footnote ⁹ For the text of Q I shall use the Critical Edition of Robinson, Hoffmann and Kloppenborg,Footnote ¹⁰ although alternative possibilities will also be acknowledged. I shall assume Q except where the rival L/M hypothesis is under consideration.

1. Parallel Redaction

Matthew and Luke provide similar accounts of the beginnings of Jesus’ ministry, and—according to the Q hypothesis—the similarities derive from their independent use of both Mark and Q. At the point where the later evangelists first make contact with these two sources, a tendency to proceed in parallel has already become evident in their independent decision to supplement Mark with birth stories. As we shall see, the Q hypothesis entails a whole series of similar yet independent redactional moves on the part of Matthew and Luke. The question is how far this parallelism constitutes a problem for the hypothesis.

In the introductory account of the ministry of John the Baptist (Mark 1.1–6 + pars.), minor agreements between Matthew and Luke against Mark may indicate that Q too began at this point. Matthew and Luke agree against Mark in their references to ‘all the region of the Jordan’, as an area from which crowds came to John (Matt 3.5) or as part of John's own sphere of activity (Luke 3.3). Both later evangelists cite Isa 40.3 without the non-Isaianic elements present in Mark (Mark 1.2–3; Matt 3.3; Luke 3.4). These minor agreements may suggest that the later evangelists were able to draw on a parallel Q introduction to the ministry of John the Baptist, in addition to Mark. Other phraseology, unique to Matthew or Luke, may also be derived from Q, though the lack of agreement with the other evangelist makes it difficult to identify as such.Footnote ¹¹ A Q introduction is required not just to account for the minor agreements, however, but also to prepare for the passage that follows. Here, a sample is given of the Baptist's harsh proclamation of judgment to those he describes as a ‘brood of vipers’, and this is clearly derived from Q (Matt 3.7–10; Luke 3.7–9). The almost complete verbal agreement demonstrates the need for a literary explanation of the relationship between Matthew and Luke. Yet this passage cannot represent the beginning of Q; the speaker is not even named in it. There must, then, be a Q component as well as a Markan one in the Matthean and Lukan passages in which the Baptist is introduced.

We already note that the double redactional procedure entailed in the Q hypothesis requires both later evangelists independently to make broadly similar choices. In the introductory passage, they choose to conflate elements of Mark with elements of Q, and they supplement this passage with the Q judgment oracle which they both cite almost verbatim. The parallelism extends into the passage on the Coming One that follows. As formatted below, underlinings represent words or phrases unique to a single evangelist; italics, points where the wording of Matthew and Luke agrees against Mark; [A] and [B^1,2], the distribution of the Coming One and double baptism sayings.

Matthew and Luke independently derive from Q the transposition of [A] and [B¹], resulting in the division of the double baptism saying, and the addition of ‘…with fire’, leading into the winnowing-fork saying (worded almost identically). Matthew deviates further from Mark than does Luke, perhaps deriving from Q the reference to ‘carrying’ sandals rather than ‘undoing’ them. Yet the parallel redactional procedures are again noteworthy. Both evangelists prefer the longer Q format, although Luke is also influenced by the Markan wording. Q, then, has provided both later evangelists with two judgment oracles, one general, the other announcing the Coming One, and these are linked by the theme of ‘fire’ (Q 3.9, 16, 17).

Was there a Q account of Jesus’ baptism, the event that follows in all three synoptic evangelists (Mark 1.9–11; Matt 3.13–17; Luke 3.21–22)? Minor agreements against Mark are again slight but not negligible.Footnote ¹² Arguably, Q needs some such passage in order to bridge the gap between the Baptist's proclamation and the extended temptation narrative that follows (Q 4.1–12). If so, Matthew and Luke independently prefer Mark to Q at this point. In the temptation narrative, however, they incorporate different elements from the brief Markan version into the much fuller Q version, which completes this text's opening narrative sequence. Almost imperceptible in the introductory passage, Q provides the later evangelists with the two judgment oracles, gives way to Mark's account of Jesus’ baptism and the descent of the Spirit and reappears in the extended temptation narrative. In spite of their differences, the later evangelists proceed in parallel as they undertake their parallel tasks, which are to compose new accounts of the beginning of Jesus’ ministry on the basis of both Mark and Q.

The other equally striking parallelism is between Q and Mark. If these two texts are independent of each other, as Matthew and Luke are supposed to be, then both the pairings that comprise the classic two-source hypothesis—Mark and Q, Matthew and Luke—give similar accounts of the beginnings of Jesus’ ministry, proceeding in mysterious harmony like the four living creatures in Ezekiel's vision. One possibility is simply to deny this opening narrative sequence to Q. A Q that begins with the Inaugural Sermon (Q 6.20–49) is a more coherent text than one that sets out as a narrative gospel but diverts into sayings gospel mode. On that view, passages such as the judgment oracle (Q 3.7–9) and the temptation story (Q 4.1–12) are independent fragments, not truly a part of Q.Footnote ¹³ In that case, however, the redactional parallelism is not just striking but astonishing. Quite independently of one another, Matthew and Luke insert the same fragments at the same points in their common Markan framework. Alternatively, Mark may be dependent on Q, or Q on Mark.Footnote ¹⁴ But if parallelism without dependence can exist on the second tier (Matthew and Luke), perhaps it can also exist on the first (Mark and Q).

The fact is that the Q hypothesis labours under a certain disadvantage in this opening narrative sequence. This is indirectly acknowledged by B. H. Streeter himself. In response to ‘the obvious suggestion that Luke knew Matthew's Gospel’, Streeter claimed that, ‘subsequent to the Temptation story, there is not a single case in which Matthew and Luke agree in inserting the same saying at the same point in the Marcan outline’.Footnote ¹⁵ If that is the case only ‘subsequent to the Temptation story’, it is because the situation is quite different in the opening narrative sequence itself. Here, Matthew and Luke do agree in inserting the same sayings (the two judgment oracles) and the same narrative (the temptations) at the same points in the Markan outline. It follows that, for this material, ‘the obvious suggestion that Luke knew Matthew's Gospel’ must seem plausible, on Streeter's own premises.

The L/M hypothesis seeks to account for a single rather than a double redaction, and has no need of the coincidental parallelism required by the Q hypothesis. If Luke knew Matthew's Gospel, he would follow Matthew in the two judgment oracles (while allowing Mark some influence over the wording of the second). He would supplement Matthew with the exchanges between the Baptist and the crowds, tax-collectors and soldiers (Luke 3.10–14), just as Matthew had previously supplemented Mark with the first judgment oracle. But he would omit the exchange between the Baptist and Jesus himself (Matt 3.14–15). Why might he have done so? If no plausible explanation lies to hand, does this suggest Luke's ignorance of Matthew, as the Q hypothesis requires? But little if anything should be read into an omission. The later evangelists often omit material from Mark, and speculating about their motives is rarely fruitful.Footnote ¹⁶

Given that Matthew and Luke otherwise agree here in their additions to Mark, Streeter's assumption must be that their disagreement elsewhere is such as to override their agreement here. On the basis of this material alone, we would postulate not Q but Luke's knowledge of Matthew. If Q represents the best explanation for the rest of the ‘double tradition’, however, it makes sense to extend the hypothesis even into material that does not display the all-important disagreement. Probably no-one will be satisfied with the suggestion that Luke followed Matthew in the opening narrative sequence, but Q thereafter.

The crucial question is whether the redactional parallelism generated here by the Q hypothesis should be regarded as entirely conceivable or as highly unlikely.Footnote ¹⁷

2. The Inaugural Sermon

According to the Q hypothesis, Q moves rapidly on from the temptation narrative to the Inaugural Sermon (Q 6.20–49). That is why, in Matthew, the Sermon occurs so early in Jesus’ ministry. The Sermon on the Mount follows the Q Sermon only in the outline it shares with Luke, and it is expanded with Q and non-Q material drawn in from elsewhere. It therefore contains two types of Q material: framework material also found in the Lukan Sermon on the Plain and the underlying Q text, in much the same sequence as in Matthew; and supplementary material drawn from Q contexts that may still be preserved in Luke. The framework material consists in the followingο five items:

1. 1. a collection of four beatitudes, pronouncing blessings on the poor, the hungry, the sorrowful, and the persecuted—with or without corresponding woes, which may be Lukan (Matt 5.3, 4, 6, 11–12; Luke 6.20–26);

2. 2. sayings relating to love of enemies, non-resistance and the Golden Rule (Matt 5.39–48, 7.12; Luke 6.27–36);

3. 3. the warning against judging, with the corresponding parable of the mote and the beam (Matt 7.1–5; Luke 6.37–38, 41–42);

4. 4. the passage on good and bad trees and their fruits (Matt 7.16–20; Luke 6.43–45);

5. 5. the parable of the two houses, introduced by the saying, ‘Why do you call me Lord, Lord…?’ (Matt 7.21, 24–27; Luke 6.46–49).

This framework material corresponds closely to Luke's Sermon on the Plain, which in turn corresponds closely to Q in both sequence and content. Of the supplementary items with which Matthew expands this framework, the following thirteen are drawn from Q:

Further analysis of Matthew's use of Q would investigate the extension of the collection of beatitudes (Matt 5.3–12), the construction of the antitheses (Matt 5.21–48), the incorporation of the Lord's prayer into a second antithetical structure (Matt 6.1–18) and so on. There seems no reason to doubt that a plausible account of the Matthean redaction would come to light.

If Q is a hypothesis, however, requiring critical testing as such, it is not sufficient simply to describe the Matthean redaction as though Q were an extant document. Any hypothesis must demonstrate its superiority over its rivals. In this case, the hypothesis only exists at all on account of the alleged implausibility of its main rival, the L/M hypothesis. On that view, the thirteen Q items inserted into the Sermon on the Mount would have to be redescribed as items extracted from the Sermon on the Mount by Luke, and reincorporated at later points in his gospel. It is at this point that the L/M hypothesis is supposedly at its weakest. Thus, W. G. Kümmel writes in an oft-quoted passage:

This familiar argument has perhaps done more than any other to establish a sense of the prima facie plausibility of Q. Yet it is seriously flawed.

First, Kümmel assumes that a later evangelist would inevitably wish to reproduce the Sermon on the Mount in his own work, in more or less its Matthean form. But Luke would not have known this material as ‘the Sermon on the Mount’, and it is anachronistic to imagine that he would necessarily have shared the modern reverence for it. A later evangelist must reshape, omit or supplement source material if he is to produce a genuinely new gospel at all.

Second, Kümmel's question about ‘motivation’ is premature. In order to test the Q hypothesis, we should first reconstruct the redactional procedure the evangelist would have to follow in redistributing parts of the Matthean Sermon to later points in his own gospel. Once this hypothetical redactional procedure has been established, we may then raise the question of ‘motivation’, seeking a possible redactional logic within the new juxtapositions.

On the L/M hypothesis, the five framework items and thirteen supplementary items, listed above, take on a different significance. The framework items—four beatitudes; sayings on love of enemies, judging and fruitbearing; the parable of the two houses—now represent the Lukan reduction of the Matthean Sermon to around a third of its original compass. Some material Luke would simply omit: nowhere in Luke do we find an equivalent of the Matthean prohibition of oath-taking (Matt 5.33–37). There is no need for the L/M hypothesis to explain ‘what could possibly have motivated Luke’ to omit this prohibition, any more than the Markan priority hypothesis is obliged to speculate about Luke's objections to everything in Mark 6.44–8.26. Both hypotheses generate accounts of redactional procedure that involve omissions, and speculation about motives will usually be inconclusive.

The thirteen supplementary items must now be presented as Matthean passages extracted by Luke for reincorporation elsewhere:

The secondary Lukan sequence is quite different from the original Matthean one. We may imagine that, in the course of reducing Matthew's Sermon on the Mount to his own Sermon on the Plain, Luke has copied into a notebook those Matthean items he wishes to set aside for subsequent use.Footnote ¹⁹ These are all reincorporated, along with other items shared with Matthew or unique to Luke, in the central section of Luke's Gospel in which the Markan framework is absent (Luke 9.51–18.14).Footnote ²⁰

If the thirteen items are rearranged in their Lukan sequence, the following pattern comes to light:

In four cases (A B C E) Luke has arranged the Matthean passages in pairs, the second member of which is inserted more or less directly after the first. In the first three cases (A B C), an entirely new juxtaposition is created, as the dispersal of our original Matthean enumeration indicates (A1/2 Luke = 6 11 Matthew, and so on). In A and B, a thematic link between the paired passages is readily apparent even from the outline above, and closer investigation shows that this is also the case with C, E and F2/3. Also in E and F2/3, traces of the original Matthean order are preserved.

Further analysis of the redactional procedure consequent on the L/M hypothesis is unnecessary here, since our concern is primarily a methodological one. This preliminary analysis confirms the Q hypothesis if and only if the redactional procedure consequent on the L/M hypothesis is held to be ‘completely inconceivable’. It supports the Q hypothesis if the thirteen items are more plausibly seen as supplementing a short Sermon than as extracted from a long one. It undermines the Q hypothesis if the L/M redactional procedure is judged to be no less plausible than the alternative. The issue is not resolved merely by observing that the Sermon on the Mount appears to be a secondary construction put together by Matthew himself. Assuming that the evangelist made use of at least one source, that source might or might not have been Q. As defined by the standard hypothesis, Q is not just a generic lost sayings source. It has its own particular contours.Footnote ²¹

One other way in which the issue might be resolved is by demonstrating what Mark Goodacre has described as ‘alternating primitivity’:Footnote ²² that Matthew preserves the earlier form of a passage on some occasions, Luke on others. The earlier form would then be the Q form; the L/M hypothesis is vulnerable if passages can be identified in which the earlier form is clearly preserved in Luke. One possible candidate occurs at the very outset of the Inaugural Sermon. In Luke, Jesus pronounces blessings on ‘you poor’ and ‘you who hunger now’ (Luke 6.20–21). In Matthew, he blesses ‘the poor in spirit’ and ‘those who hunger and thirst for righteousness’ (Matt 5.3, 6). It is often said that Matthew has here ‘spiritualized’ sayings that originally referred to material deprivation. While this is certainly possible, it is methodologically important to exclude the alternative possibility only when one has attempted and failed to make a case for it. If Luke's beatitudes are constructed out of Matthew's, then the later evangelist has reapplied beatitudes relating to ethical qualities to the material deprivations endured by Jesus’ disciples. By adding the woes, he has created a set of antitheses: poor/rich, hungry/satisfied, weeping/laughing, hatred/esteem (Luke 6.20–26). Arguably, the first two antitheses fit remarkably well into a Gospel that contains the Magnificat (cf. Luke 1.52), the scriptural motif of good news for the poor (Luke 4.18; 7.22) and the parables of the Rich Fool (12.16–21) and Dives and Lazarus (16.19–31).Footnote ²³ If so, there is no need to trace the beatitudes in their shorter form back to Q. If Matthew's beatitudes are secondary then Luke's may be primary, but they may equally well be tertiary.Footnote ²⁴

3. The Common Sequence

The chapters following Matthew's Sermon on the Mount contain a significant quantity of Q material interspersed within Markan material in which the original Markan sequence is not easy to detect. Indeed, it is the free handling of Markan sequence especially in Matthew 8–9 that has led to the view that Luke is more likely to have preserved the Q sequence.Footnote ²⁵ This conclusion is unwarranted, however. In Matthew 8–13, three interlocking sequences may be identified. Two of these are Markan, and they are the result of Matthew's decision to allow the sequences Mark 1.40–4.34 and 4.35–6.6 to overlap in his own rendering of them. As formatted below, the second sequence is italicized in order to differentiate it from the first; square brackets represent the third, non-Markan sequence to which we shall shortly return; the asterisk denotes a displacement.

As the italicization indicates, Matthew's primary Markan sequence is interrupted by material from a later point in Mark's narrative, constituting a second Markan sequence. Both sequences are incorporated in their correct Markan order. With the incorporation of Mark 4.1–34 in Matthew 13, the gap between the first and second sequences has been filled up, so that from this point on Matthew follows a single Markan sequence. The one genuine displacement occurs at Mark 1.30–31 (Peter's mother-in-law). It occurs on account of Matthew's decision to place the healing of the leper on the road that leads from the mountain to Capernaum, where the centurion's servant and Peter's mother-in-law are healed. It is therefore incorrect to assume that Matthew 8–9 in particular is based on some non-Markan principle of organization.Footnote ²⁷ Although the Beelzebul controversy story is said to have occurred in Q as well as Mark, it should be noted that it occurs here at the appropriate position within the primary Markan sequence.

Interspersed within the two overlapping Markan sequences is a further sequence consisting of ten items that Matthew shares with Luke. Most of these item are non-Markan, but three of them show signs of a Markan connection (= { }, below). Surprisingly, two of these three simultaneously belong to Matthew's primary Markan sequence, and represent points where Markan and non-Markan sequences converge. The Mark column represents the point reached in Matthew's primary Markan sequence when an item is inserted:

Items iii–vi occur consecutively in Matt 9.35–11.27, but i–ii and vii–x occur within contexts determined by the first Markan sequence (= Mark 1.40–4.34), to which indeed two of them already belong (vii {Beelzebul controversy} and x {Jesus’ true family}).

The sequence recurs within two distinct sections in Luke (7.1–50; 9.51–11.32), separated by the abbreviated Lukan rendering of Mark 3.31–9.40. In the following analysis, an asterisk indicates a difference from the Matthean order. The Mark column again represents the point reached when non-Markan items are inserted:

Some general observations about this shared sequence are necessary, before seeking to recover the redactional procedures needed to create it under the L/M and Q hypotheses respectively.

First, in both Matthew and Luke this common sequence follows on from the Sermon, which in turn follows on from the opening narrative sequence. In both later gospels, this entire common sequence has been superimposed on the Markan one, and around 40% of the ‘double tradition’ material occurs within it. This is to be differentiated from the remaining 60% of this material which is differently distributed: in Matthew, primarily in the great discourses (Matt 5–7; 10, 18; 23–25); in Luke, primarily in 11.33–17.37, starting from the very point at which the common sequence ends. In both Matthew and Luke, the common sequence is partially concealed by material derived from Mark or unique to one or other evangelist. It comes to light only when this extraneous material is set aside.Footnote ²⁸

Second, the earlier position of iv Jesus and John in Luke connects this passage more closely with the Baptist material that introduces Jesus’ own ministry (Luke 3.1–22 and parallels). Since the sequence here concludes with viii Sign of Jonah, the sequel to vii {Beelzebul controversy} is now, appropriately enough, ix Return of unclean spirit. In Matthew, the equivalent sequel is provided by the passage on fruitbearing (Matt 12.33–37), which this evangelist connects to the twin sayings on blasphemy against the Spirit that conclude his account of the Beelzebul controversy. However they are to be explained, these divergences do not seriously undermine the common non-Markan sequence.

Third, after the opening narrative sequence Luke incorporates items from the common sequence in just two major groupings, and these items all occur at a later point relative to Mark than in Matthew. While some grouping occurs in Matthew, common sequence items are more widely dispersed through the evangelist's overlapping Markan sequences; and all have been incorporated by the point at which Matthew reaches Mark's parable chapter (Matt 13 = Mark 4). In contrast, only two of the ten common sequence items have been incorporated at the equivalent point in Luke (Luke 8.4–18). As a result of Luke's grouping of common sequence material, the Markan sequence stands out more clearly in Luke than in Matthew, even though it is equally fundamental to both evangelists. Matthew seeks to conflate his Markan and non-Markan sequences, whereas Luke seeks to preserve their individual integrity. The later positioning of the non-Markan material relative to the Markan indicates that Luke gives a certain priority to Mark.

Fourth, Luke's version of vii {Beelzebul controversy} is anomalous. In Matthew this controversy story occurs within the primary Markan sequence. As with the other Markan controversy stories in Matthew 12, the later evangelist inserts supplementary material here, in the form of a new introduction (12.22–23) and sayings relating to the source of exorcistic power (12.27–28), being for or against Jesus (12.30) and defaming the Son of man and the Holy Spirit (12.32).Footnote ²⁹ Luke attests the first three of these four expansions, and he places this story not in the appropriate location in his Markan sequence (that is, at Luke 8.4) but within the sequence shared with Matthew. Thus, this controversy story belongs both within Matthew's rendering of the primary Markan sequence, and within the non-Markan sequence Matthew shares with Luke.

Fifth, a related anomaly arises from Luke's version of x {Jesus’ true family}. Here, Jesus blesses ‘those who hear the word of God and keep it’, in response to a woman in the crowd who pronounces a blessing on his mother (11.27–28). This recalls the Markan passage on Jesus’ true family (Mark 3.31–35), which Luke has already reproduced (Luke 8.19–21). Thus Luke includes two passages on this theme, and they are similar enough to comprise a ‘doublet’. In both passages, Jesus’ saying is provoked by a reference to his mother (with or without his brothers), and in both he responds by commending instead those who hear and obey the word of God. In Matthew, x {Jesus’ true family} occurs in a Markan form and within the primary Markan sequence. Yet Matthew and Luke have both attached a form of the true family passage to ix Return of unclean spirit. Like vii {Beelzebul controversy}, this passage exists simultaneously within a Markan sequence (evidenced in Matthew) and a non-Markan one (evidenced in Luke).

This analysis of the sequence common to Matthew and Luke can now be used to test the L/M and Q hypotheses. The question to be put to each is this: What kind of redactional procedure would produce this common sequence, differently distributed and partially concealed by Markan and other material?

On the L/M hypothesis, the common sequence is the result of Luke's distinctive engagement with Matthew. The hypothesis requires Luke to be able to identify non-Markan material in Matthew, but to do so only after he has given his own rendering of the Markan context in which Matthew has located it, on the basis of Mark's narrative alone.Footnote ³⁰ Luke first rewrites a Markan sequence, and then incorporates selected supplementary material from Matthew. After the opening narrative sequence, the third evangelist engages with each of the two earlier gospels in turn, alternating between them.Footnote ³¹ Given a redactional decision to differentiate Markan and Matthean material and to give precedence to Mark, it is unsurprising that relatively few traces of Matthew's redaction of Mark are evident in the equivalent Lukan redaction. Luke departs from his usual editorial practice, however, in the case of vii {Beelzebul controversy} and x {Jesus’ true family}. In the one case, he has followed Matthew rather than Mark, as both the positioning and the additional sayings indicate. In the other case, he has followed Matthew as well as Mark, reworking the Markan version within his Markan sequence (Luke 8.19–21) and the Matthean version (itself dependent on Mark) within the common sequence derived from Matthew (Luke 11.27–28).

On the Q hypothesis, the non-Markan sequence common to Matthew and Luke is derived from Q, rather than being a Lukan construct derived from Matthew's supplementation of Mark. Working independently but in parallel, the later evangelists find different ways to accommodate the Q sequence within the Markan one. Luke's consistent grouping of items from Q may preserve their original connections. vii {Beelzebul controversy} and x {Jesus’ true family} create some complications, however. It is evident from Matthew that these passages belong to the evangelist's primary Markan sequence, and it is evident from Luke that they also belong to the common sequence derived from Q. So we must envisage two independent sequences, one from Mark, the other from Q, in both of which an account of the Beelzebul controversy is followed—immediately in Mark, after intervening material in Q—by an incident in which a reference to Jesus’ mother or family provokes him to utter a saying commending those who observe the will or word of God. Since Matthew reproduces the Mark version here, the Lukan version of x {Jesus’ true family} probably derives from Q.Footnote ³²

Thus Matthew has at his disposal two otherwise different sequences, derived from Mark and Q, which coincide at two points. These points occur late in the Q sequence, early in the Markan one, and the evangelist uses them to create a juncture at which the two sequences merge into one. That is why Matthew's Q material has to occur so early in relation to Mark. The construction of Matthew 3–12 in its entirety is shaped by the decision that the two sequences should meet and merge at their otherwise inconspicuous points of coincidence. This redactional decision is, however, the product of the Q hypothesis itself. Matthew expands Mark's account of the Beelzebul controversy, as he expands other Markan controversy stories, but there is little in the Matthean text itself to suggest that the evangelist has at his disposal a second version of the entire story.Footnote ³³ The hypothetical Q version of the story is the product of Luke's unusual proximity here to Matthew rather than Mark, interpreted on the assumption that Luke cannot be dependent on Matthew. The proximity is such that, if one evangelist cannot be dependent on the other, a common source is the only option. But it remains a challenge for the Q hypothesis to show why its own consequent redactional procedure is more plausible than the straightforward one entailed in the L/M hypothesis. The claim that the L/M hypothesis must attribute to Luke a patently incredible redactional procedure has turned out to be premature, to say the least.

It is not my concern here to adjudicate between the Q and L/M hypotheses. The aim is methodological: to show the importance of systematic attention to the redactional procedures entailed in competing source-critical hypotheses. In the case of Q, refutation of the L/M hypothesis is a prerequisite, and a full analysis of the redactional procedure consequent upon both hypotheses is therefore indispensable for the critical assessment of Q itself. Anyone concerned to verify Q, rather than merely assuming it on the basis of traditional ad hoc arguments, must engage seriously with the major alternative.Footnote ³⁴

Once an overview of the relevant redactional procedures has been achieved, it may then be possible to identify decisive points that tell against one hypothesis and in favour of another. Some potential candidates for this role may have come to light in the course of this study. Yet the moral of the story of Q research is that supposedly decisive points have been too readily accepted as such, passing into wider circulation without consideration of the wider framework within which they must be either verified or falsified. That is a pitfall to be avoided, whether in reaffirming Q or in dispensing with it.