The Science of Language Learning: What Research Actually Says

Language learning advice is everywhere. Most of it is based on anecdote, marketing, or the experience of unusually gifted polyglots. The scientific literature tells a more nuanced — and more actionable — story.

Here's what decades of second language acquisition (SLA) research actually establishes, with the practical implications for how you should build your study routine.

1. Input Hypothesis: Comprehensible Input Is the Core Driver

Stephen Krashen's Input Hypothesis (1982) remains the most influential and most contested theory in SLA. The central claim: language acquisition happens when you encounter input that is slightly above your current level of competence (i+1 in his notation — "comprehensible input").

What the evidence actually supports:

• High-quality input (reading, listening to native material) is necessary for acquisition
• Grammar instruction alone without input exposure produces test-takers, not speakers
• Output (speaking, writing) accelerates acquisition beyond input-only exposure — this is where Krashen's original theory is underdeveloped

Practical implication: The majority of your study time should involve encountering natural language in context — books, podcasts, TV shows — not drilling grammar rules. But speaking practice matters too, especially for activating passive vocabulary.

2. Spaced Repetition: The Most Evidence-Backed Learning Technique

The spacing effect — first documented by Hermann Ebbinghaus in 1885 — is one of the most replicated findings in cognitive psychology. Distributing practice over time produces dramatically better long-term retention than massing the same amount of practice in a single session (cramming).

Spaced repetition systems (SRS) formalise this by scheduling reviews at expanding intervals based on your performance:

• Initial learning: review after 1 day
• Correct recall: push to 3 days, then 7 days, then 21 days, etc.
• Incorrect recall: reset the interval

Effect sizes from meta-analyses are striking: spaced practice produces 1.5–2x better long-term retention versus massed practice for the same total study time.

Practical implication: Use an SRS (Anki, the algorithm built into language learning apps) for vocabulary. The discipline of daily short sessions beats weekend marathons.

3. Output Hypothesis: Speaking Accelerates Acquisition

Merrill Swain's Output Hypothesis (1985) challenged Krashen's input-only model. Swain observed that French immersion students in Canada had excellent comprehension but poor speaking accuracy after years of input-rich schooling. Her argument: speaking forces you to process language at a level of precision that listening doesn't require.

When you produce output, you notice gaps in your competence (you reach for a word and discover you don't know it), you test hypotheses about grammar, and you receive corrective feedback. These noticing events appear to drive acquisition.

Modern SLA research broadly supports a dual role: input for acquiring new forms, output for consolidating them.

Practical implication: If you study for 30 minutes a day, at least 10 of those minutes should involve speaking or writing — not just passive exposure.

4. Critical Period Hypothesis: Adults Can Learn But It's Harder

The Critical Period Hypothesis (Lenneberg, 1967) proposes that language acquisition is biologically constrained — there's a developmental window (roughly through puberty) during which native-like acquisition is achievable. After the critical period closes, adult learners face a steeper path.

What the evidence actually shows is more nuanced:

• Phonology: Accent acquisition is genuinely harder after puberty. Neural plasticity in the auditory-motor integration system decreases. Most adults who start after their teens retain a detectable foreign accent — not all, but most.
• Morphosyntax: Adults are slower to acquire complex grammatical features but are better at learning vocabulary and at deploying explicit rule knowledge.
• Ultimate attainment: Adults can and do achieve very high proficiency. The claim that adults "can't become fluent" is false. The claim that it's harder and takes longer is true.

A 2018 study (Hartshorne et al.) analysed 670,000 online grammar test takers and found the optimal period for achieving native-like grammar ends at around age 17–18, with a softer decline continuing into the mid-twenties. This is a population-level trend, not a ceiling on any individual.

Practical implication: If you're an adult learner, don't accept the defeatist framing. Do invest extra time in pronunciation practice early — it becomes progressively harder to change phonological habits.

5. Interactional Feedback: Error Correction That Works

Not all error correction is equal. Research distinguishes several types:

• Recasts: Repeating the learner's utterance with the error corrected (e.g., learner says "He go to school," teacher responds "Yes, he goes to school every day."). Natural, low-threat, but learners often don't notice the correction.
• Explicit correction: Directly flagging the error ("You should say 'goes,' not 'go'"). More noticing, more disruptive to fluency.
• Clarification requests: Pretending you didn't understand ("Sorry?"). Forces the learner to self-repair.
• Metalinguistic feedback: Describing the rule without providing the form ("Remember the third-person singular present tense rule").

Meta-analyses (e.g., Li, 2010) find that recasts work best for phonological errors, explicit correction works best for morphosyntactic errors, and clarification requests are most effective for pragmatic errors. The context matters.

Practical implication: When using AI conversation tools, ask for recast-style correction in free conversation mode and explicit correction when drilling specific grammar points.

6. Motivation: Intrinsic Beats Extrinsic

Self-Determination Theory (Deci and Ryan) distinguishes:

• Intrinsic motivation: Engaging with the language because it's genuinely interesting or enjoyable
• Extrinsic motivation: Studying to pass a test, get a job, or keep a streak

Both drive behavior in the short term. Only intrinsic motivation sustains behavior long term. Learners who study primarily to maintain a streak or earn badges show dramatically higher dropout rates once external rewards are removed.

The most successful long-term language learners tend to share one characteristic: they find genuine enjoyment in the content of the target language — its music, films, literature, or the relationships it opens. The language becomes a vehicle for something they already care about.

Practical implication: Find content in your target language that you'd want to consume even if you already spoke it fluently. Pair your SRS sessions with content you actually enjoy.

7. The Comprehensible Input Threshold: ~95% Rule

Vocabulary research (Nation, 2001) established that readers need to know approximately 95% of the words in a text to read it with adequate comprehension and without heavy dictionary use. For audio, the threshold is slightly lower (~90%) because prosody and context fill in more gaps.

This has a direct implication for content selection: material that's at 80% comprehension is frustrating, not productive. The sweet spot is challenging but accessible.

For listening, this corresponds to roughly the i+1 level Krashen described — you catch most of what's said and use context to infer the rest. Netflix series aimed at teenagers or young adults, podcasts from language teaching networks (Dreaming Spanish, Coffee Break Languages), and graded readers are engineered to sit near this threshold.

Building a Study System From the Science

Synthesising the above:

Time allocation	Activity	Why
30–40%	Comprehensible input (reading/listening at 95% comprehension)	Core acquisition mechanism
20–25%	Spaced repetition vocabulary review	Highest ROI per minute for retention
20–25%	Speaking/writing output	Consolidates forms, reveals gaps
10–15%	Pronunciation practice (especially early)	Most time-sensitive skill
5–10%	Grammar study (targeted, not exhaustive)	Fills specific gaps, not a primary driver

Consistency matters more than any single session. Thirty minutes daily beats four hours on weekends, independent of method. This isn't motivational — it's what the spaced repetition and consolidation research directly predicts.

---

I'm building Pocket Linguist, an AI-powered language tutor for iOS. It uses spaced repetition, camera translation, and conversational AI to help you reach conversational fluency faster. Try it free.