A group of eight employees spanning A2 to B2 is not a language class. It is an administrative convenience. Most organisations, when they commission a corporate language programme, treat class size as a cost question. More employees per group equals lower per-head spend. The logic is tidy. It is also wrong.
This is not a matter of opinion. The research on second language acquisition has been accumulating since at least Krashen’s (1982) comprehensible input hypothesis, and its practical implications for group composition are not ambiguous. They are simply ignored by most programme designs — because programme designs are typically done by procurement teams rather than applied linguists, and procurement teams are optimising for the wrong variable.
This post sets out what the research says about corporate language class size, how to apply the Common European Framework of Reference for Languages (CEFR) as a grouping tool, and what HR managers, L&D professionals, and programme sponsors in Singapore-based organisations need to establish before a single employee sits down in a language class. Consider it a framework for making decisions before the programme goes sideways, rather than after it already has.
Krashen’s (1982) comprehensible input hypothesis establishes that language acquisition requires learners to receive input that is slightly above their current level — what he termed “i+1”. The implication is precise: a group spanning A2 to B2 on the CEFR scale cannot all receive i+1 input simultaneously. The teacher is either pitching above the A2 learners, above the B2 learners, or — most commonly — somewhere in the middle that serves nobody particularly well.
Merrill Swain’s (1985) output hypothesis adds a further problem. Learners acquire language not just by receiving it but by producing it, under communicative pressure, with feedback. Speaking turns are a finite resource in any classroom. As group size increases, individual output decreases. This is not a hypothesis; it is arithmetic.
Lou, Abrami, and Spence’s (1996) meta-analysis of within-class grouping, published in the Review of Educational Research, found that smaller, well-matched groups consistently outperform larger mixed-ability groups on learning outcomes. The finding has been replicated. It is, at this point, about as controversial as the claim that buildings require foundations.
The practical implication: optimal language training is inseparable from group composition and size. A group of four closely matched B1 learners is a better investment than a group of ten spanning three CEFR levels, regardless of what the per-head cost comparison suggests.
Small groups work when the conditions are right, and the conditions are more specific than most organisations assume.
The first condition is tight CEFR-level matching. A gap of more than one sub-level within a group of four creates a drag effect on the stronger learners and an intimidation effect on the weaker ones. With only four people in the room, neither dynamic has anywhere to hide.
The second condition is a role-specific or advanced training objective. Four B1-level account managers preparing for English-language client presentations will benefit far more from a small, role-specific group than from a large general business English cohort of mixed levels and mixed functions. They share a communication problem. They can practise the same scenarios. The teacher can correct errors that are directly relevant to situations those employees will face next week.
The third condition is high stakes. Cross-border negotiations, board-level presentations, and senior client interactions all justify the additional per-head cost of a smaller group. The return on a successful negotiation in English tends to exceed the saving from reducing the group by three people.
On the question of group versus individual language training for employees: small groups occupy productive territory between the two formats. They retain the peer interaction that supports speaking fluency — the value of negotiating meaning with another person, as Long (1981) demonstrated in his research on interaction and second language acquisition — while keeping the group small enough for the teacher to remain genuinely responsive to each participant.
Larger groups can work. The problem is not group size per se; it is group size combined with the wrong conditions.
For a group of six to ten employees to function as an effective language class, three conditions must hold simultaneously: participants must be closely matched on CEFR level, must share overlapping communication needs, and must be working towards a general-language objective rather than a role-specific one.
Long’s (1981) research on interaction in second language acquisition identifies threshold effects: beyond a certain group size, individual output opportunities and teacher responsiveness decline in ways that measurably slow acquisition. The threshold is not a fixed number — it depends on lesson structure, teacher skill, and participant engagement — but it is real, and ignoring it produces predictable results.
The most common and most preventable corporate language training group composition error is sorting employees by job title rather than proficiency level. An organisation groups all its managers together, or all its customer-facing staff, or everyone in the Singapore office. The resulting group may span A1 to B2. The teacher performs a kind of pedagogical triage. Nobody learns much.
Basically, no. And when it is the right answer, it requires a level of deliberate pedagogical scaffolding that most organisations have no ability to provide.
Mixed-level classes are either a deliberate strategy — differentiated instruction, properly resourced and executed — or a mistake that generates poor outcomes for almost everyone involved. Carol Ann Tomlinson’s (2001) framework for differentiated instruction is instructive here: it demands substantial lesson planning, experienced teachers, and adapted materials for each learner sub-group within the same class. It is not what happens when you place a mixed-level group in a room with a competent teacher and trust that things will sort themselves out.
The practical implication for CEFR level grouping of employees: if the programme cannot support genuine differentiated instruction, the mixed-level group will reliably underserve the weaker learners whilst frustrating the stronger ones. Separate groups or individual training are not luxuries in this scenario. They are the correct call.
Every system is perfectly designed to get the results it gets.
W. Edwards DemingA group that is correctly composed at the start of a programme will not necessarily remain correctly composed at week six.
Employees progress at different rates, particularly in the early acquisition stages, where individual variation is most pronounced. A recommended practice is to build scheduled review points into the programme structure — at the end of each learning unit — at which the teacher and programme manager assess whether the group composition still serves all participants.
The specific advantage of live, instructor-led delivery is that teachers can flag grouping issues in real time, before the end of a term when the damage is done and the budget is spent. Fixed-cohort models without built-in review mechanisms cannot do this.
The grouping decision reduces to three variables: proficiency level (CEFR-aligned, per the ECML’s quality assurance matrix for CEFR use), learning objective (general, business, or role-specific), and communication context (department, audience, task type).
| Variable | Condition | Recommended Action |
|---|---|---|
| CEFR level spread | More than one sub-level gap within the group | Separate groups or individual training |
| Job function | Significantly different roles or communication needs | Separate groups, even within the same CEFR band |
| Learning objective | Role-specific or advanced | Small group (3–5) or individual training |
| Learning objective | General language at a shared level | Larger group (6–10) may be appropriate |
| Budget pressure toward larger group | Against best-practice composition advice | Build structured review and regrouping rights into the contract from the outset |
| Mixed-level group unavoidable | Differentiated instruction confirmed and fully resourced | Proceed with caution; monitor weekly |
These are not advisory suggestions. They are the conditions under which the research supports each approach. Departing from them is a risk that should be named explicitly — not managed quietly by hoping the teacher is talented enough to compensate for a structural problem.
The most predictable way to waste a corporate language training budget is to treat class size as a cost question and group composition as someone else’s problem. The research on second language acquisition is not ambiguous. The conditions for effective group learning are well-documented. Most grouping errors are avoidable.
The question worth asking about any current or planned programme is this: has it been built on a proper needs analysis and CEFR-aligned placement assessment, or on a reasonable-sounding assumption that has never been stress-tested?
If the answer is the latter — or if you are not entirely sure — speak with Lingua Learn Singapore about how group composition should be structured for your specific situation. Contact Lingua Learn Singapore
Individual training is most effective when the employee has highly specific communication goals, an unusual CEFR level relative to the team, or a high-stakes deadline such as a cross-border negotiation or an international board presentation. Group training is most effective when participants share both a proficiency level and a communicative purpose. The decision should follow those conditions rather than a general preference for one format over another.
Grouping employees by job title rather than proficiency level. It produces groups that span multiple CEFR levels, gives teachers an impossible brief, and reliably underdelivers for almost everyone involved. The fix is straightforward: conduct CEFR-aligned verbal placement assessments before composing any group.
Written placement tests measure declarative knowledge — the ability to identify a correct grammar rule on paper. Spoken language programmes require procedural knowledge — the ability to use language fluently under communicative pressure. A verbal assessment measures both. Using only a written test for a speaking-focused programme produces inaccurate placement and, eventually, poorly composed groups that are expensive to correct mid-programme.
