The Science Behind the System

The Core Principle

You don't learn a word by seeing it once. You learn it by encountering it across contexts — in a flashcard, then in a story, then in a quiz, then in a podcast, then in a conversation. Each encounter strengthens the memory trace from a different angle.

This is the foundation of Frank Lingo: one vocabulary set, delivered through six channels that each target a different cognitive pathway. The research basis for this approach comes from three decades of second language acquisition (SLA) studies.

Spaced Repetition: Building Durable Memory

Each month, Frank Lingo generates a fresh Anki deck calibrated to your CEFR level. This is the explicit encoding phase — where you first meet the words you'll encounter everywhere else.

Spaced repetition is the most thoroughly validated technique in memory science. By showing you each word at increasing intervals — just before you'd forget it — the system converts short-term exposure into long-term retention. Pimsleur (1967) first proposed graduated-interval recall; modern algorithms like SM-2 and FSRS refine the intervals based on your individual performance.

Deck sizes are calibrated by level so you're never overwhelmed:

Pimsleur, P. (1967). A memory schedule. The Modern Language Journal, 51(2), 73–75.
Milton, J., & Alexiou, T. (2009). Vocabulary size and the CEFR. In B. Richards et al. (Eds.), Vocabulary Studies in First and Second Language Acquisition. Palgrave Macmillan.

The Noticing Hypothesis: Why Multi-Channel Works

Schmidt's Noticing Hypothesis (1990, revised 2010) is the theoretical backbone of the entire system. The core claim: learners acquire linguistic items only when they consciously notice them in input. Subconscious exposure alone is not enough.

This has a direct design implication. When you pre-study vocabulary in your Anki deck, those words enter your active attention window. When you then encounter them in a podcast episode or weekly article, you experience a recognition event — a conscious moment of "I know that one." This noticing-in-context produces significantly stronger retention than either flashcards or reading alone.

Schmidt, R. (2010). Attention, awareness, and individual differences in language learning. In W. M. Chan et al. (Eds.), Proceedings of CLaSIC 2010 (pp. 721–737). National University of Singapore.

Weekly Reading: Comprehensible Input in Your Inbox

Every Monday, you receive a short article written entirely in your target language, built from your deck vocabulary. It's accompanied by three comprehension questions you can answer with one tap.

This applies Krashen's Input Hypothesis (1985): acquisition happens when learners encounter input slightly above their current level (i+1). The article uses familiar vocabulary in an unfamiliar narrative context — a story, a letter, a news report — requiring you to use contextual inference for the few new elements.

Nation (2006) established that 95% text coverage provides basic understanding, while 98% provides comfortable reading. By constructing articles from your studied vocabulary, we target 95–98% coverage: challenging but not frustrating.

The comprehension questions apply the testing effect (Roediger & Karpicke, 2006): attempting to retrieve information from the article strengthens memory more effectively than simply re-reading it.

Krashen, S. D. (1985). The Input Hypothesis: Issues and Implications. Longman.
Roediger, H. L., & Karpicke, J. D. (2006). Test-enhanced learning. Psychological Science, 17(3), 249–255.
Nation, I. S. P. (2006). How large a vocabulary is needed for reading and listening? Canadian Modern Language Review, 63(1), 59–82.

Podcast Listening: Training Your Ear

The monthly podcast uses your deck vocabulary in realistic conversational scenes. This is the extensive listening component: large amounts of engaging, comprehensible audio that trains listening fluency and triggers recognition events in a new modality.

Research shows podcast-format delivery is particularly effective: learners control pacing and replay, consistent short sessions (15 minutes, 3–4 times per week) outperform sporadic long sessions, and motivation stays higher than with textbook exercises.

For beginners, each sentence is followed by a brief English translation — a scaffolding technique that Yuksel & Tanriverdi (2009) found produces the highest vocabulary gains. As you advance, the translations withdraw: scene-level at A2, pure immersion at B1 and above.

Renandya, W. A., & Farrell, T. S. C. (2011). 'Teacher, the tape is too fast!' Extensive listening in ELT. ELT Journal, 65(1), 52–59.
Yuksel, D., & Tanriverdi, B. (2009). Effects of watching captioned movie clip on vocabulary development. Turkish Online Journal of Educational Technology, 8(2).

Vocabulary Quizzes: The Testing Effect in Action

Your dashboard includes vocabulary quizzes drawn from your current deck. Multiple-choice questions test recognition, and your responses feed the comprehension tracking system. This is retrieval practice — the most powerful study technique known to cognitive science.

Roediger & Karpicke (2006) demonstrated that attempting to retrieve information from memory strengthens it more than re-studying the same material. Even incorrect retrieval attempts produce a learning benefit.

Quiz questions are generated from your learning targets and scheduled across the month, ensuring broad coverage of your vocabulary set. Results map back to specific targets so the system knows exactly which items you’ve mastered and which need reinforcement.

AI Text Tutor: From Recognition to Production

Swain's Output Hypothesis (1985) argues that producing language — not just understanding it — is essential for acquisition. When you write a sentence, you discover gaps in your knowledge that passive study never reveals.

Frank Lingo's AI tutor sends you a daily conversation prompt via SMS, tied to your current vocabulary. You reply in your target language, receive gentle inline corrections, and continue for 3–5 exchanges. This turns passive knowledge (recognizing words) into active competence (using them).

Monthly Report: Closing the Feedback Loop

At the end of each month, an AI-generated progress report synthesizes your quiz accuracy, reading comprehension, engagement patterns, and vocabulary gaps into personalized recommendations.

Unlike generic "great job!" summaries, the report identifies specific strengths ("your travel vocabulary accuracy is 92%"), specific weaknesses ("food-related sentences had the lowest recall"), and suggests concrete next steps — including topic refinements and level changes when the data supports them.

A one-click level change button lets you act on the recommendation immediately: move up when you're ready for harder material, or adjust down to rebuild confidence. The system adapts to you, not the other way around.

Adaptive Feedback: Learning From How You Study

Most language apps generate content on a fixed schedule with no awareness of what you’ve actually retained. Frank Lingo reads your Anki review data and feeds it back into next month’s content.

Every time you review a card in Anki, you produce a signal: did you recall it immediately, hesitate, struggle, or fail? These signals — combined with the review interval and time spent — paint a precise picture of which vocabulary, grammar structures, and phrases you’ve internalized versus which ones you’re still fighting.

Frank Lingo ingests this data through an Anki desktop add-on that syncs reviews automatically, or through a manual deck upload for mobile users. Either way, the data flows into the same adaptive pipeline.

Two levels of adaptation

Target-level: Each review is mapped back to its underlying learning targets — the specific vocabulary items, grammar structures, and phrases it exercises. These signals accumulate in a lifetime learner profile. When generating next month’s lesson plan, the AI curriculum designer sees which targets are mastered (high ease, long intervals) and which remain weak (frequent failures), and adjusts accordingly.

Sentence-level: Sentences with high failure rates are flagged for reinforcement — the generator creates new sentences targeting the same patterns. Well-retained sentences (long intervals) are deprioritized to make room for new material.

This applies Bjork’s principle of desirable difficulty (1994): learning is most effective when difficulty is calibrated to the learner’s current ability. By monitoring actual retention through SRS data rather than relying on self-assessment alone, the system has a more accurate measure of true mastery.

Bjork, R. A. (1994). Memory and metamemory considerations in the training of human beings. In J. Metcalfe & A. Shimamura (Eds.), Metacognition: Knowing about Knowing (pp. 185–205). MIT Press.

Why the Combination Matters

Any one of these channels works in isolation. But the compounding effect of their combination is what makes Frank Lingo different.

When you study a word in Anki, then read it in a story, then hear it in a podcast, then use it in a conversation, then test yourself on it in a quiz — you've encoded that word across five distinct cognitive contexts. Transfer-appropriate processing theory predicts that each context strengthens the memory trace in a different way, making the word more robust and accessible.

Critically, this happens without any extra planning or effort from you. You import one deck, open one email, listen to one podcast. The system handles the coordination. The same words appear in the same month across every channel because they're all generated from the same vocabulary set.

The Research Foundation

Every design decision in Frank Lingo is grounded in published SLA research, not opinion. The key principles:

References