The production gap
Most language learners understand far more than they can produce. They recognise vocabulary, parse grammar, follow conversations—but freeze when it's their turn to speak. This is the production gap, and it's where most learners stall.
Traditional methods address comprehension: reading, listening, grammar drills, vocabulary lists. These build passive knowledge. But oral fluency requires a different kind of learning—one where sentences are produced, not just understood.
Swain's output hypothesis (1985, 1995) established that production practice serves functions that comprehension alone cannot: it forces syntactic processing, reveals gaps in knowledge, and consolidates procedural memory. Shima is built on this insight.
Why sentence-level drilling works
Shima doesn't teach grammar rules or vocabulary in isolation. It drills complete sentences—the unit of actual communication. Each sentence is heard, shadowed, and produced until it becomes automatic.
This approach mirrors DeKeyser's skill acquisition theory: declarative knowledge (knowing the rule) must be proceduralised through practice before it becomes automatic. The sentence is the natural unit for this—small enough to drill, large enough to encode grammar, vocabulary, and pragmatics in context.
Grammar and vocabulary are acquired in context, not in isolation. When you drill "すみません、これは何ですか", you learn the grammar pattern, the vocabulary, and the social register simultaneously—because that's how language actually works.
Island structure
Content is organised into islands—thematic clusters of 50 to 100 sentences around a real-world situation. Ordering at a restaurant. Navigating public transport. Discussing your work with colleagues.
Each island is self-contained and sequenced in phases: listen-only, shadowing, prompted production, and free recall. Learners progress through phases as the engine confirms automaticity at each level.
This structure means learners build complete communicative competence in a domain before moving on. After completing a restaurant island, you can actually order in a restaurant—not just recognise menu items.
The adaptive engine
Shima's engine tracks per-sentence difficulty for each learner. Sentences that come easily are spaced out. Sentences that cause hesitation are revisited more frequently and presented in varied contexts.
This isn't generic spaced repetition—it's production-aware scheduling that distinguishes between recognition and recall, between understanding a sentence and being able to produce it fluently.
The engine also provides in-context vocabulary notes and grammar insights as learners encounter new patterns—supporting comprehension without interrupting the production flow.
Content that sounds like a person
Shima's sentences are authored to sound like things a real person would actually say. Not textbook constructions, not artificial examples—natural speech with authentic register, pragmatics, and cultural context.
This matters because language sticks when it's attached to meaning you care about. "The house has three bedrooms" teaches grammar. "Three bedrooms—honestly more space than we need, but the light was worth it" teaches grammar and installs a worldview you can actually use.
For institutional deployments, content is authored to domain requirements: military terminology, diplomatic protocol, commercial vocabulary, or academic discourse.
Research basis
Shima's methodology draws on established research in second language acquisition:
- Swain's output hypothesis—production practice serves functions that comprehension alone cannot
- DeKeyser's skill acquisition theory—proceduralisation through practice
- Nation's four strands—balanced attention to meaning-focused input, meaning-focused output, deliberate learning, and fluency development
- Laufer & Hulstijn's involvement load hypothesis—deeper processing leads to better retention
The core insight: production practice is qualitatively different from comprehension practice, and fluency requires both. Most tools stop at comprehension. Shima completes the picture with structured production practice.