Back to Blog
pedagogy
curriculum
TBLT

Tasks, not grammar points

February 5, 2026Jaimy Team

Open a traditional language textbook to the table of contents. The chapter titles will look something like this:

> Chapter 1: The Present Tense

> Chapter 2: Definite and Indefinite Articles

> Chapter 3: The Plural

> Chapter 4: Adjective Agreement

> Chapter 5: The Present Continuous

The structure is the grammar of the language, laid out in a sensible-looking order, with vocabulary and exercises wrapped around each grammatical theme.

Now open Jaimy. The structure looks different:

> Ordering coffee. Asking for directions. Introducing yourself at a party. Telling a story about your weekend. Making a complaint. Booking a hotel room. Discussing what you read in the news.

There is grammar in there. There is vocabulary in there. But none of it is the organising principle. The organising principle is what you would actually want to do with the language. The grammar shows up because the task needs it, not the other way around.

This is the difference between most language teaching and what's sometimes called Task-Based Language Teaching. The order matters more than it sounds.

Why grammar-first stops working

Grammar-first teaching has a comforting logic. The language is a system; the system has rules; if we teach the rules in a sensible order, the learner will be able to assemble sentences using the rules.

In practice, the move from "knowing a rule" to "using the rule under time pressure in conversation" is enormous. Most adult learners who studied a language in school can tell you the rules. They cannot reliably produce sentences using them. The path between the two — the path from declarative knowledge to procedural automaticity — is exactly the path that grammar-first teaching often fails to deliver.

There is another, subtler problem. When a syllabus is organised around grammar, vocabulary and contexts get chosen to demonstrate the grammar rather than to serve the learner. The example sentences are about a fictitious family or a generic city park, because those let you show the present tense without distraction. You spend hours producing sentences you would never, in real life, want to produce. The motivation slowly leaks out of the room.

By the time you graduate from the present tense to the past, you have built up exactly the skill of producing example-sentence-shaped utterances, in example-sentence contexts, for example-sentence purposes. The skill does not transfer to actual conversation because actual conversation was never the practice.

What task-based teaching does instead

Task-based teaching, as a movement in language education, argues for inverting this. Start with the communicative task, and let the grammar come up as a tool the task happens to need.

Take a simple task: describe what you did yesterday. To do this, you need the past tense. So the past tense gets pulled in — not as Chapter 7, but as the thing you need right now to do the thing you're trying to do. The grammar arrives in context, attached to a use case, mid-conversation. It is much more likely to stick because you have a reason to remember it.

A meta-analysis of task-based teaching studies found effect sizes between d = 0.61 and d = 0.93 across the literature. In education research, anything above 0.5 is a meaningful effect. Numbers in the 0.6–0.9 range are considered large. Task-based teaching consistently produces real, measurable advantages over grammar-first approaches.

The mechanism is roughly what you'd guess. Tasks engage the learner. Engagement drives output. Output drives noticing, hypothesis-testing, and acquisition. The grammar comes along because it has to.

Jaimy's session catalogue

If you look at Jaimy's library of session types, you'll see this principle made concrete. Sessions are named by what you're doing, not by what you're learning. A few examples:

  • At the café. A short, scripted ordering exchange. Coffee, pastry, paying. Past tense of "to like" sneaks in naturally if you have a regular order. Numbers come up because prices come up.
  • Telling your weekend. An open conversation about what you did. Past tense, time markers, sequencing words ("first... then... after that..."). Plus whatever vocabulary your weekend happened to involve.
  • Calling the doctor. A roleplay where you're trying to describe a symptom and make an appointment. Body-part vocabulary, modal verbs ("can I", "should I", "do I need to"), conditional structures ("if I take this, will I..."). All in the service of getting the appointment booked.
  • The job interview. Self-presentation, past experience (lots of past tense), hypothetical questions ("how would you handle..."), formal register. Whole books of grammar packed into a 20-minute task.

Each of these is a task. Each is something a real person might really want to do in the target language. The grammar and vocabulary inside each one is selected because the task needs it — not because the syllabus had reached that chapter.

The pleasant side effect

There is an emotional consequence to this structure that we did not expect at first but that turned out to matter a lot.

When you finish a grammar-organised lesson, the feeling is I learned a tense. It's an abstract reward. The next time you talk to a real person in the target language, you will not feel that the tense in your head is doing anything for you.

When you finish a task-organised session, the feeling is I just did the thing I was scared to do. You ordered the coffee. You handled the complaint. You navigated the appointment. It happened in a low-stakes setting with an AI, but it happened. The next time you face the real version of that situation, the simulator is in your head and you have an actual reference point.

This is one of the quiet reasons Jaimy users keep coming back. The sessions feel like preparation for things they will plausibly do, not abstract exercises that might one day pay off.

The trade-off we accepted

A task-based curriculum is messier than a grammar-based one. The same grammar point comes up across many different tasks, in different combinations, at different difficulties. There is no neat chapter to point to as "this is where you learn the subjunctive." The subjunctive shows up across roleplays, discussions, opinion pieces, hypothetical conversations.

This is harder to design. It is harder to assess. It does not produce a tidy textbook.

But it produces fluency. And fluency was the whole point.

A learner can recite a verb table. A speaker can talk their way through a conversation. Those are not the same thing, and only the second one is what people mean when they say they want to "learn a language." Jaimy is built for the second.