English only: writing a multilingual policy the centre can defend

The policy is an institutional reflex with weak support as an absolute. Research on classroom interaction supports extensive target-language use; it does not support prohibition, and the strongest claims are about the quantity and quality of target-language exposure, not about banning L1. 'Maximise English' and 'English only' are different policies that have been collapsed into one.

Distinctions worth making explicit

1. Function. L1 for task management, planning and solidarity does different work from L1 substituting for the target-language production the task requires. The first is usually harmless or beneficial; the second is what teachers actually worry about.
2. Planned versus incidental. A two-minute L1 planning phase followed by an English report-back is a design decision. Covert whispering is not.
3. Teacher use versus learner use. A teacher who shares the L1 can use it strategically; here the teacher cannot, so the policy is asymmetric and unenforceable - which is exactly why it has gone underground.
4. Shared-L1 classes versus multilingual classes. The rule was written for a class where everyone shares Vietnamese. In that class, L1 is a resource. For the Eritrean learner it is exclusion, because she cannot access either the English or the Vietnamese.

Likely effects of the current approach. Policing has produced resentment, covert use, and no increase in English. It has also transferred the burden of inclusion onto the Eritrean learner without giving her any means of access. And it has put the pharmacist in an impossible position. A policy that everyone breaks is not a policy; it is a source of conflict.

Policy statement (short enough to be read): English is the default language of instruction, tasks and reporting back. Learners' other languages are a legitimate resource for planning, clarifying and comparing, within agreed boundaries. No learner is required to use, or prohibited from using, any particular language for private thinking.

Procedures that make it real

1. Time-boxed L1 planning. 'Two minutes in Vietnamese to plan, then English to report.' Name it, time it, and move on. Covert use becomes unnecessary, so the teacher stops policing.
2. English for the public phase. Reporting back, presenting and teacher-facing interaction are in English. That is the boundary, and it is visible.
3. Teacher signals the switch explicitly so learners can track the shift.
4. Bilingual support materials. A glossary where a reliable Tigrinya or Arabic translation can be checked, otherwise images and labelled diagrams. Do not improvise translations; a wrong term in a clinical or safety context is worse than none.
5. Multilingual noticing tasks, used occasionally: compare how two languages in the room express the same function. This treats all languages as legitimate objects of study without asking anyone to perform their identity.
6. Routine, not exception. Build the L1 phase into lesson plans so it is normal.

For the Eritrean learner specifically

• Do not make her a permanent project, and do not default her to the nearest sympathetic learner every week.
• Front-load comprehension: pre-teach six key items, use realia and visuals, write instructions as well as saying them, and check understanding by task, not by 'do you understand?'
• Use short structured pair tasks with rotating partners, and let her use Arabic script for notes if that helps her think.
• Give her a genuine contribution early in each session so she is not always the one being helped.
• Review at week four with her, privately, in plain English: what is working, what is not.

Name the problem precisely. Three things are tangled: unpaid labour, accuracy risk, and her own learning being sacrificed.

1. Unpaid labour. A learner who is more fluent is being used as staff. It is inequitable, it is unsustainable, and she is threatening to leave, which makes it a retention problem as well as a fairness one.
2. Accuracy and masking. Untrained interpreting in a classroom is unreliable, and it hides whether the Eritrean learner has understood anything or has simply received a summary. Ad hoc interpretation also means the teacher gets no diagnostic information.
3. Her learning. She is the most advanced learner in the room and is currently doing the least language development. Her needs are being ignored because she is useful.

Practical response

• Stop the default. Do not call on her to explain tasks to the group; use written and visual supports instead.
• If interpretation is genuinely needed - for induction, safeguarding or enrolment - the centre should provide and pay for it, or use a professional service. That is a management decision and should be raised as one, not absorbed by a learner.
• Offer a bounded, chosen role if she wants one. Five minutes at the start of a session explaining the task in Vietnamese, agreed in advance and recognised as a contribution, is different from being on call all evening. Ask her privately; do not announce it.
• Differentiate upward. Give her extension tasks: summarising the group's conclusions, rewriting a model text, taking an fluency-focused role in feedback. She should leave the course better, not just more useful.
• Talk to her privately first. She may be proud of the role, or resentful, or both. The conversation, not the policy document, determines whether she stays.

Evidence, tied to the policy's own claims

1. Observed pattern of L1 use, categorised by function - planning, clarification, social, or substitution for required production. Two observed lessons at baseline and two at week six. The target is not zero: it is that L1 shifts toward planning and clarification and away from substitution.
2. Covert use. Count of off-task whispered Vietnamese during public phases. This should fall if the policy is honest; if it rises, the rule is still being experienced as a prohibition.
3. Participation by learner. Speaking turns and task contributions per session, tracked by name. The Eritrean learner is the key case: if her contributions do not rise within four weeks, the access measures are not working.
4. Retention, especially the pharmacist, and exit interviews with anyone who leaves.
5. Learner views, collected anonymously: three questions on whether they feel able to ask for help and whether the English-only rule helps them learn.
6. Task performance: a recorded spoken turn at baseline and week eight, rated on interaction criteria, to check that more L1 has not cost English development - the fear the original policy was built on.

Revision triggers, agreed in advance

• If target-language production falls against baseline, tighten the public-phase boundary.
• If covert use persists or the Eritrean learner's participation stays flat, the policy is cosmetic and the access measures need replacing, not defending.
• If the pharmacist leaves, the policy has failed on its own terms, whatever the observation data show.

Honesty about the evidence: one class, one teacher, no control group. Report change as consistent with the policy, and keep the review open rather than declaring success at week six.