Passive recognition is not active production. This is the most important distinction in applied linguistics that most language-learning apps ignore.
The recognition–production gap
Memorising word pairs, matching tiles, or selecting from multiple-choice options trains the visual recognition centre of the brain — Wernicke's area. When you see "discuss" and can match it to its meaning, that is passive recognition at work.
Oral fluency, however, requires the rapid motor coordination of Broca's area: retrieving the right word, assembling the right grammar structure, and activating the vocal apparatus — all in under 200 milliseconds, mid-conversation, while simultaneously processing what your interlocutor just said.
These are fundamentally different cognitive operations. Training one does not automatically train the other.
Why Leitner spacing works for speaking
Sebastian Leitner's spaced repetition system (developed in the 1970s) is based on a simple observation: memory consolidation is more efficient when review occurs right before forgetting would happen, not before.
In a traditional Leitner box:
- Box 1: Review every day (items you get wrong)
- Box 2: Review every 3 days
- Box 3: Review every 7 days
- Box 4: Review every 14 days
- Box 5: Review every 30 days (mastered items)
nthoughts applies this to spoken mistakes. If you say "I am looking forward to meet you" (a preposition error), the system captures that specific spoken form and creates a speaking drill — not a reading drill, not a multiple-choice question, but an audio prompt you must respond to out loud. That drill enters Box 1 and begins its Leitner progression based on your subsequent spoken performance.
The critical difference: oral output, not recognition
Most spaced repetition apps (Anki, Duolingo's heart system, Babbel) test recognition. You see or hear a prompt and click or type a response.
nthoughts's drills require you to speak out loud. The system transcribes your response in real time and checks whether you produced the correct form. This small difference has large consequences: you are training the same motor pathway that will be required in the actual conversation.
This is what Merrill Swain called "comprehensible output" (1995): the act of being forced to produce language, rather than merely comprehend it, is itself a learning mechanism.
Practical takeaway
If you have been studying English vocabulary for years and still freeze when speaking, your recognition vocabulary is likely fine. What needs work is your production vocabulary — and the only way to build that is through repeated, corrected oral output with spaced intervals.
Start with Maya's conversation practice and let the system identify your production gaps automatically. Your Leitner decks build themselves from your actual mistakes, not from a generic word list.