Two learners share one screen and must submit one paragraph. One wants to write people is, the other stops at the verb, and both have to decide what the subject controls. That thirty-second exchange contains more language work than silently copying a correction after the teacher has marked it. For an intermediate writer who can finish a school essay alone yet repeats the same sentence errors, one jointly owned draft can make choices visible before the next solo attempt. The useful unit here is co-writing: two learners plan, phrase, type and revise one text. Dividing the work into separate paragraphs misses the point.
Each sentence should pass through both writers.
What the pair must decide
Writing hides its process. A finished paragraph shows the chosen article or connector, with no trace of the rejected options. Pair talk exposes that missing record. Learners ask whether advice takes a plural ending, whether although can stand beside but, and whether a pronoun still has a clear referent after two sentences. Neomy Storch's 2005 classroom study followed 23 adult ESL students taking degree courses. Students could write in pairs or alone, so the groups were self-selected and the comparison cannot establish that pairing caused every difference. Even with that limitation, the process data matter. The pairs produced shorter texts that performed better on task fulfilment, grammatical accuracy and complexity, while recordings showed learners pooling ideas and giving each other feedback. The ERIC record for Storch's study reports the design and the students' reservations as well as the favourable responses.
The talk needs a real decision. Asking two learners to fill identical blanks side by side produces agreement without composition. Give them a message with a purpose, a reader and a constraint: write 137 to 163 words to persuade a manager to change a meeting time, while preserving a polite tone and naming one workable alternative. Now vocabulary, grammar and audience compete for attention. Before typing, each partner can write two private options for the opening sentence. Comparing four candidates makes disagreement normal and gives the quieter writer material to contribute.
It also prevents the first words spoken by the quickest learner from becoming the draft by default.
One keyboard, shared control
Shared control matters because the faster typist can quietly become the author. Set a simple rule for the first draft: one learner types while the other reads the emerging sentence aloud, then they switch after 83 words or after each paragraph. Either writer may pause the typing and request two possible phrasings. The teacher listens for decisions rather than counting turns. This is where a small class earns its size. A teacher can visit four pairs and sample one disputed sentence from each. The intervention can be narrow: ask what the noun refers to, request a second verb, or have both writers read the line from the reader's position.
Supplying a polished sentence too early removes the decision that made the task useful.
Pairing alone offers no guarantee. Storch's earlier analysis of three adult ESL pairs found different interaction patterns, including relationships with unequal control. The Language Teaching Research study treats collaboration as something visible in mutual engagement and shared direction, rather than a seating arrangement. A dominant writer and a quiet copyist have produced one file, yet only one person has composed it. I would keep the first co-written text short enough for both people to remember every argument about it. A 900-word assignment creates too many places for labour to split invisibly.
A 156-word complaint email is plenty.
Follow with individual retrieval
The shared draft is rehearsal. A later individual task checks transfer. Change the situation while preserving the language demand. After the meeting-time email, each learner writes alone to reschedule a delivery. After a jointly written opinion paragraph, each writes a new paragraph that must concede one point and defend another. Gillian Wigglesworth and Neomy Storch compared 48 individual performances with 48 pair performances on an argumentative writing task. Pairs produced a single text, and the detailed analysis found higher accuracy for collaborative writing without gains in fluency or complexity. The 2009 Language Testing study summary also warns teachers about assessment: a pair product cannot be read as two individual scores.
That boundary is useful. Co-writing can be formative evidence about which choices learners notice together. The solo follow-up supplies evidence about each learner. If both tasks receive the same grade, the evidence becomes muddy and students may reasonably worry about unequal contribution. The follow-up should arrive soon. Twenty-four or 47 hours leaves enough separation to require retrieval while keeping the language problem recognisable. Return no model answer between the two tasks. Instead, let each learner keep a photograph of the pair's final text, remove it before writing, and compare the two products afterward.
Feedback should enter the next text
During co-writing, feedback can target the pair's reasoning. Afterward, mark one or two recurring choices and ask each learner to locate the same risk in the solo text. This continues the logic in Choc Education's article on why peer feedback needs a second draft: comments acquire value when another decision remains. Some learners need a different arrangement. A writer with strong ideas and slow typing may appear passive beside a quick partner. Two learners with a tense relationship may spend their attention protecting status. Individual planning before co-writing, speech-to-text access, or a teacher-selected role can make participation more legible.
Pair work should also stop when discussion becomes a prolonged negotiation over every comma.
A practical lesson can fit into 41 minutes. Give learners four minutes to plan separately, 14 to co-write, six to read the text aloud and revise, then 12 to write a related message alone. Use the last five minutes for each learner to underline one phrase carried from the shared draft and circle one decision made differently. Collect both texts. The circle is the evidence to teach from next time.