For the view that the Two Source Theory better accords with how ancient writers used their sources in the Roman world, see Robert Allen Derrenbacker, Ancient Compositional Practices and the Synoptic Problem (BETL 186; Leuven: Peeters, 2005; cf. his PhD thesis at the University of Toronto) and F. Gerald Downing, “Disagreements of Each Evangelist with the Minor Close Agreements of the Other Two” ETL 80 (2004): 445–469 and Doing Things with Words in the First Century (Sheffield Academic Press, 2000). Here is the summary:
Assuming that an evangelist such as Luke was working with scrolls, he would have difficulty balancing the scroll he was working on on his knees (scribes did not use a writing desk) while consulting two other scrolls set up on stands. Perhaps some of the difficulty could be alleviated if Luke was dictating to a scribe rather than writing himself. The constraints of this procedure make it difficult to attempt complex harmonizations of multiple sources at the micro-level or moving back and forth through a scroll. Thus, ancient writers preferred to work with one source at a time for any given account. The theory works best with how Luke alternates between citing the text of Mark and Q in whole blocks (Q material primarily in Luke 6:20-8:3; 9:1-18:14). It is more difficult to explain the ways Luke and Matthew fuse Mark and Q together in the so-called Mark-Q overlaps (e.g. Matt 3-4/Luke 3-4) as well as how Matthew more often integrates Q material into Mark’s narrative framework. However, the idea that an evangelist may be working with the text of Mark even while recalling some wording or anecdote from Q from memory or that Matthew had Q in another literary form (e.g. a codex, the forerunner of the modern book, to enable flipping through it quicker) has been suggested.
Ken Olson defended the Farrer Hypothesis in light of the same ancient practice of using one source at a time in his Master’s thesis How Luke was Written. Luke would have stuck to how whole episodes were reported in Mark or in Matthew without harmonizing them, though it is possible that Luke could have been influenced by his memory of Matthew’s wording even when copying a block from Mark. Other Farrer proponents argue that Luke’s rearrangement of some of the material that Matthew adds to Mark would not be as difficult as presumed. Francis Watson proposes that Luke may have initially relied on a notebook when first reading through Matthew and then drew from his notes in rearranging the pieces of Matthew to forge new connections in his final composition (Gospel Writing: A Canonical Approach, 170-71, 171 n. 27). John C. Poirier’s “The Roll, the Codex, the Wax Tablet and the Synoptic Problem” JSNT 35 (2012): 3-30 suggests that Luke utilized wax tablets so that he could experiment with new arrangements in his notes before producing the final copy (see also the criticisms from Downing and Derrenbacker).
However, James W. Barker has just published “Ancient Compositional Practices and the Gospels: A Reassessment” Journal of Biblical Literature 135 (2016): 109-121. He calls into question the assumption that the use of scrolls makes it implausible that ancient writers would either conflate or radically re-order the material in their sources. Here is the abstract:
Recent studies of ancient compositional practices and the Synoptic Problem have validated the Two-Source hypothesis and challenged the “Augustinian,” Farrer–Goulder, and Griesbach hypotheses. These studies conclude that, according to the Two-Source hypothesis, subsequent evangelists would have adhered to the Greco-Roman conventions of working with one source at a time and not working backward through a text. The present essay adduces counterexamples such as the Greek Minor Prophets Scroll from Naḥal Ḥever, which predates the Gospels, and Tatian’s Diatessaron, which postdates the Gospels. Upon further examination, simultaneously accessing multiple sources and reordering those sources were established compositional practices in the first century. Moreover, every solution to the Synoptic Problem necessitates such scribal conventions. Therefore, the lesser extent of these ancient compositional practices does not privilege the Two-Source hypothesis over its rivals.