What 40 Years of Language Acquisition Research Says About Learning to Speak

Four decades of second language acquisition research point the same way. You learn to speak by understanding plenty of language, talking with someone who responds to what you mean and pushing yourself to say it. Feedback helps most when it makes you notice one thing at a time.

Second language acquisition (SLA) is the study of how people learn a language after their first. Here are its biggest findings about speaking, and what each means for your practice.

Speech-language pathologists (SLPs) reach many of the same principles from the clinic. These are learning strategies borrowed from speech therapy and language research. Learning a language is not a disorder, and neither is an accent.

The big findings at a glance

1. Input: necessary, but not enough on its own

Stephen Krashen's input hypothesis holds that we acquire language by understanding messages a little beyond our current level (Krashen, 1982, 1985). Part of that holds up well. Learners understood spoken stories fairly well once they knew about 95% of the words (van Zeeland & Schmitt, 2013).

Understanding and speaking are different skills, though. A meta-analysis of 30 studies compared comprehension-based and production-based grammar instruction. Both had large effects, but production-based instruction was more effective on delayed tests of productive knowledge (Shintani, Li & Ellis, 2013).

Input wins early. Production wins later, for what you can still say. These were grammar studies, not free conversation. More in comprehensible input.

2. Interaction: talk with someone who checks what you mean

Michael Long argued that changes to the structure of a conversation, like checking and repairing, matter more than simplifying the speech itself (Long, 1983).

That repair work is called negotiation of meaning. Mackey and Goo's (2007) synthesis is the usual citation for interaction's positive effect. A later review agrees, but says the payoff depends on the partner, the task, the target and the setting (Loewen & Sato, 2018).

It doesn't have to be face to face. A meta-analysis of 14 studies found that online interaction helped about as much as in-person interaction (Ziegler, 2016). Most of those studies used text chat, not voice with an AI. The field is large but uneven in rigor (Plonsky & Gass, 2011).

Here's negotiation in a Spanish retell of The Lion and the Mouse:

The learner fixed the word without being handed it. More in negotiation of meaning.

3. Output: why you have to say it yourself

Merrill Swain's output hypothesis says speaking does work listening can't. When learners notice a problem mid-sentence and change what they say, they process grammar more deeply than comprehension requires (Swain & Lapkin, 1995).

In one experiment, adult ESL learners worked on English relative clauses. Those who tried to produce the form, then saw it again, gained more than those who only read it for meaning (Izumi, 2002). Highlighting the form in the text produced no measurable gain.

That was one grammar target in a written task. Still, trying to say it taught more than rereading it.

4. Noticing: attention turns input into intake

Richard Schmidt argued that noticing is "the necessary and sufficient condition for converting input to intake" (Schmidt, 1990). That's a theoretical claim, and the strong version is contested. His weaker claim is easier to defend: paying attention probably helps.

A classroom study with ESL learners linked feedback, reported noticing and the learning of question forms (Mackey, 2006). The relationship was positive but complex. The sample was small, and noticing was self-reported. Still, the practical point holds: pick one target per round.

5. Corrective feedback: cues beat corrections for self-repair

In primary French immersion classrooms, teachers mostly used recasts, saying the learner's sentence back correctly. Recasts rarely led students to fix errors themselves. Elicitation, clarification requests and repetition did better (Lyster & Ranta, 1997).

A meta-analysis of 15 classroom studies agrees: feedback had durable effects, and prompts beat recasts (Lyster & Saito, 2010).

Many of these learners were children. All nine coach moves are in how a good speaking coach responds.

6. Conversational turns: back-and-forth, not word counts

These are child first-language studies, so read them as an analogy. Researchers counted conversational turns, not just words heard.

• Turns predicted later outcomes. In 146 children, turn counts at 18–24 months explained 14–27% of the variance in IQ and language scores ten years later. That held after controlling for socioeconomic status (Gilkerson et al., 2018).
• The link ran both ways. In 122 children followed from 9 to 24 months, turns and vocabulary growth predicted each other, even controlling for adult words (Donnelly & Kidd, 2021).
• Coaching changed the back-and-forth. In a 71-family trial, parent coaching increased turn-taking, which correlated with language growth (Ferjan Ramírez, Lytle & Kuhl, 2020).
• Turns mediated gains. In a 52-family trial with 4- to 6-year-olds, a follow-up to the team's 2018 work, changes in turns mediated language gains (Romeo et al., 2021).

Two critiques matter. The automated recorder behind most of this work is less accurate for turns than for adult words (Cristia et al., 2021). And a study of 42 children in five communities did not support the famous 30-million-word gap (Sperry, Sperry & Miller, 2019).

What Edgewise borrows is the reply style. A question can stop a story; a comment keeps it going:

In real phone conversations (the Switchboard corpus, by our own count), about 1 turn in 6 has a question. So the Edgewise coach comments more and quizzes less.

7. Task repetition: tell it again

In one study, 32 Japanese learners of English repeated three monologue tasks six times each (Lambert, Kormos & Minn, 2017). Speech-rate gains were largest across the first three tries and continued until the fifth.

Spacing matters too. With 71 adult learners, repeating a picture description soon after built speed. A one-week gap did most for complexity and repair fluency (Bui, Ahmadian & Hunter, 2019).

In a 4/3/2 study, 24 ESL students gave talks in shrinking time limits. Only those who kept the same topic held their fluency gains (de Jong & Perfetti, 2011). These are small studies, and we found no task-repetition meta-analysis.

8. Are adults worse at learning languages?

In a study of 669,498 people taking an online grammar quiz, grammar-learning ability held until about age 17.4, then declined (Hartshorne, Tenenbaum & Pinker, 2018). A reanalysis found a steady decline fit monolinguals, bilinguals and early-immersion learners better. Breaks appeared only for non-immersion and later-immersion learners, which the authors tie to schooling and social factors (van der Slik et al., 2022).

Both model final attainment, not how fast you learn this year. A review found that the amount and quality of input strongly shape how age plays out (Muñoz, 2008). Anxiety and willingness to speak

A meta-analysis of 105 samples (N = 19,933) found a negative correlation between language anxiety and achievement, r = −.36 (Teimouri, Goetze & Plonsky, 2019). One influential model proposes willingness to communicate as the goal of instruction (MacIntyre et al., 1998).

Perceived competence, anxiety and motivation all correlate moderately with willingness to communicate, and perceived competence most (Elahi Shirvan et al., 2019). Feeling capable goes with speaking up.

All of this is correlational. We found no study testing whether private practice with an AI coach lowers anxiety.

10. Story structure transfers across languages

Telling a story has two layers. Macrostructure is the plot: who and where, the problem, the attempts, the outcome. Microstructure is the sentences that carry it.

In 52 Swedish–English bilingual children aged 5–7, story structure scores varied with age but not with language (Bohnacker, 2016). English–Hebrew preschoolers kept story structure intact in both languages, even where vocabulary and grammar differed (Iluz-Cohen & Walters, 2012). Adults telling stories in German, Russian and Swedish did too (Gagarina, Bohnacker & Lindgren, 2019).

The story itself shapes what you practice. Among 100 learners of English, tightly structured stories drew more accurate speech. Stories with background events drew more complex syntax (Tavakoli & Foster, 2008).

11. How researchers measure speaking

Researchers measure speaking by complexity, accuracy and fluency (CAF). One influential paper adds that vocabulary needs its own measures (Skehan, 2009).

• Units. The AS-unit is a reliable way to divide transcribed speech into sentence-like units (Foster, Tonkyn & Wigglesworth, 2000).
• Different words. Raw word variety depends on sample length. MTLD was the only index tested that didn't (McCarthy & Jarvis, 2010).
• What listeners hear. Perceived fluency tracked speed and pause frequency strongly (r ≈ .59–.62) and self-repairs only weakly (r ≈ .20) (Suzuki, Kormos & Uchihara, 2021).

Edgewise's sentence-building score covers sentence length, linked ideas, different words and accuracy. Fillers and false starts are set aside, SALT-style. The research is a reminder to compare retells of similar length.

How much of this applies to solo practice?

Most of these studies were in classrooms or labs, some with children. A classroom is not a solo app, and few of these principles have been tested with adults practicing alone. For what independent research says about apps, see do language apps work? For pronunciation, see how to practice pronunciation.

FAQ

Is comprehensible input enough to learn to speak? No. Input reliably builds comprehension and vocabulary. In a meta-analysis of grammar instruction, production-based practice did better on delayed tests of what learners could produce.

Is it too late to learn a language as an adult? No. Large studies find a decline with age in final attainment, but a reanalysis ties the sharpest drops to schooling and social factors. How much and how well you practice matters.

Does talking to an app count as interaction? It can. A meta-analysis found online interaction helped about as much as face to face. Most of those studies used text chat, not voice with an AI.

Why do I get nervous when I speak a new language? Language anxiety is common and is linked with lower achievement. Feeling more capable goes with being more willing to speak, so start with tasks you can succeed at.

Your next step: Try a story retell in Edgewise. You tell the story out loud, and the coach talks with you about it without telling it for you. Your first retell sets your level. Edgewise scores story structure and sentence building, and every round ends with your next piece: one specific thing to add next time.

Join early access

Related: Speech therapy techniques for language learners · How a good speaking coach responds · Do language apps work?

Sources

1. Bohnacker, U. (2016). Tell me a story in English or Swedish: Narrative production and comprehension in bilingual preschoolers and first graders. Applied Psycholinguistics, 37(1), 19–48. https://doi.org/10.1017/S0142716415000405
2. Bui, G., Ahmadian, M. J., & Hunter, A.-M. (2019). Spacing effects on repeated L2 task performance. System, 81, 1–13. https://doi.org/10.1016/j.system.2018.12.006
3. Cristia, A., Lavechin, M., Scaff, C., Soderstrom, M., Rowland, C., Räsänen, O., Bunce, J., & Bergelson, E. (2021). A thorough evaluation of the Language Environment Analysis (LENA) system. Behavior Research Methods, 53(2), 467–486. https://doi.org/10.3758/s13428-020-01393-5
4. de Jong, N., & Perfetti, C. A. (2011). Fluency training in the ESL classroom: An experimental study of fluency development and proceduralization. Language Learning, 61(2), 533–568. https://doi.org/10.1111/j.1467-9922.2010.00620.x
5. Donnelly, S., & Kidd, E. (2021). The longitudinal relationship between conversational turn-taking and vocabulary growth in early language development. Child Development, 92(2), 609–625. https://doi.org/10.1111/cdev.13511
6. Elahi Shirvan, M., Khajavy, G. H., MacIntyre, P. D., & Taherian, T. (2019). A meta-analysis of L2 willingness to communicate and its three high-evidence correlates. Journal of Psycholinguistic Research, 48(6), 1241–1267. https://doi.org/10.1007/s10936-019-09656-9
7. Ferjan Ramírez, N., Lytle, S. R., & Kuhl, P. K. (2020). Parent coaching increases conversational turns and advances infant language development. Proceedings of the National Academy of Sciences, 117(7), 3484–3491. https://doi.org/10.1073/pnas.1921653117
8. Foster, P., Tonkyn, A., & Wigglesworth, G. (2000). Measuring spoken language: A unit for all reasons. Applied Linguistics, 21(3), 354–375. https://doi.org/10.1093/applin/21.3.354
9. Gagarina, N., Bohnacker, U., & Lindgren, J. (2019). Macrostructural organization of adults' oral narrative texts. ZAS Papers in Linguistics, 62, 190–208. https://doi.org/10.21248/zaspil.62.2019.449
10. Gilkerson, J., Richards, J. A., Warren, S. F., Oller, D. K., Russo, R., & Vohr, B. R. (2018). Language experience in the second year of life and language outcomes in late childhood. Pediatrics, 142(4), e20174276. https://doi.org/10.1542/peds.2017-4276
11. Hartshorne, J. K., Tenenbaum, J. B., & Pinker, S. (2018). A critical period for second language acquisition: Evidence from 2/3 million English speakers. Cognition, 177, 263–277. https://doi.org/10.1016/j.cognition.2018.04.007
12. Iluz-Cohen, P., & Walters, J. (2012). Telling stories in two languages: Narratives of bilingual preschool children with typical and impaired language. Bilingualism: Language and Cognition, 15(1), 58–74. https://doi.org/10.1017/S1366728911000538
13. Izumi, S. (2002). Output, input enhancement, and the noticing hypothesis. Studies in Second Language Acquisition, 24(4), 541–577. https://doi.org/10.1017/S0272263102004023
14. Krashen, S. D. (1982). Principles and practice in second language acquisition. Pergamon. No DOI.
15. Krashen, S. D. (1985). The input hypothesis: Issues and implications. Longman. No DOI.
16. Lambert, C., Kormos, J., & Minn, D. (2017). Task repetition and second language speech processing. Studies in Second Language Acquisition, 39(1), 167–196. https://doi.org/10.1017/S0272263116000085
17. Loewen, S., & Sato, M. (2018). Interaction and instructed second language acquisition. Language Teaching, 51(3), 285–329. https://doi.org/10.1017/S0261444818000125
18. Long, M. H. (1983). Native speaker/non-native speaker conversation and the negotiation of comprehensible input. Applied Linguistics, 4(2), 126–141. https://doi.org/10.1093/applin/4.2.126
19. Lyster, R., & Ranta, L. (1997). Corrective feedback and learner uptake: Negotiation of form in communicative classrooms. Studies in Second Language Acquisition, 19(1), 37–66. https://doi.org/10.1017/S0272263197001034
20. Lyster, R., & Saito, K. (2010). Oral feedback in classroom SLA: A meta-analysis. Studies in Second Language Acquisition, 32(2), 265–302. https://doi.org/10.1017/S0272263109990520
21. MacIntyre, P. D., Clément, R., Dörnyei, Z., & Noels, K. A. (1998). Conceptualizing willingness to communicate in a L2. The Modern Language Journal, 82(4), 545–562. https://doi.org/10.1111/j.1540-4781.1998.tb05543.x
22. Mackey, A. (2006). Feedback, noticing and instructed second language learning. Applied Linguistics, 27(3), 405–430. https://doi.org/10.1093/applin/ami051
23. Mackey, A., & Goo, J. (2007). Interaction research in SLA: A meta-analysis and research synthesis. In A. Mackey (Ed.), Conversational interaction in second language acquisition (pp. 407–452). Oxford University Press. No DOI.
24. McCarthy, P. M., & Jarvis, S. (2010). MTLD, vocd-D, and HD-D: A validation study of sophisticated approaches to lexical diversity assessment. Behavior Research Methods, 42(2), 381–392. https://doi.org/10.3758/BRM.42.2.381
25. Muñoz, C. (2008). Symmetries and asymmetries of age effects in naturalistic and instructed L2 learning. Applied Linguistics, 29(4), 578–596. https://doi.org/10.1093/applin/amm056
26. Plonsky, L., & Gass, S. M. (2011). Quantitative research methods, study quality, and outcomes: The case of interaction research. Language Learning, 61(2), 325–366. https://doi.org/10.1111/j.1467-9922.2011.00640.x
27. Romeo, R. R., Leonard, J. A., Robinson, S. T., West, M. R., Mackey, A. P., Rowe, M. L., & Gabrieli, J. D. E. (2018). Beyond the 30-million-word gap: Children's conversational exposure is associated with language-related brain function. Psychological Science, 29(5), 700–710. https://doi.org/10.1177/0956797617742725
28. Romeo, R. R., Leonard, J., Grotzinger, H., Robinson, S. T., Takada, M. E., & Mackey, A. P. (2021). Neuroplasticity associated with changes in conversational turn-taking following a family-based intervention. Developmental Cognitive Neuroscience, 49, 100967. https://doi.org/10.1016/j.dcn.2021.100967
29. Schmidt, R. W. (1990). The role of consciousness in second language learning. Applied Linguistics, 11(2), 129–158. https://doi.org/10.1093/applin/11.2.129
30. Shintani, N., Li, S., & Ellis, R. (2013). Comprehension-based versus production-based grammar instruction: A meta-analysis of comparative studies. Language Learning, 63(2), 296–329. https://doi.org/10.1111/lang.12001
31. Skehan, P. (2009). Modelling second language performance: Integrating complexity, accuracy, fluency, and lexis. Applied Linguistics, 30(4), 510–532. https://doi.org/10.1093/applin/amp047
32. Sperry, D. E., Sperry, L. L., & Miller, P. J. (2019). Reexamining the verbal environments of children from different socioeconomic backgrounds. Child Development, 90(4), 1303–1318. https://doi.org/10.1111/cdev.13072
33. Suzuki, S., Kormos, J., & Uchihara, T. (2021). The relationship between utterance and perceived fluency: A meta-analysis of correlational studies. The Modern Language Journal, 105(2), 435–463. https://doi.org/10.1111/modl.12706
34. Swain, M., & Lapkin, S. (1995). Problems in output and the cognitive processes they generate: A step towards second language learning. Applied Linguistics, 16(3), 371–391. https://doi.org/10.1093/applin/16.3.371
35. Tavakoli, P., & Foster, P. (2008). Task design and second language performance: The effect of narrative type on learner output. Language Learning, 58(2), 439–473. https://doi.org/10.1111/j.1467-9922.2008.00446.x
36. Teimouri, Y., Goetze, J., & Plonsky, L. (2019). Second language anxiety and achievement: A meta-analysis. Studies in Second Language Acquisition, 41(2), 363–387. https://doi.org/10.1017/S0272263118000311
37. van der Slik, F., Schepens, J., Bongaerts, T., & van Hout, R. (2022). Critical period claim revisited: Reanalysis of Hartshorne, Tenenbaum, and Pinker (2018) suggests steady decline and learner-type differences. Language Learning, 72(1), 87–112. https://doi.org/10.1111/lang.12470
38. van Zeeland, H., & Schmitt, N. (2013). Lexical coverage in L1 and L2 listening comprehension: The same or different from reading comprehension? Applied Linguistics, 34(4), 457–479. https://doi.org/10.1093/applin/ams074
39. Ziegler, N. (2016). Synchronous computer-mediated communication and interaction: A meta-analysis. Studies in Second Language Acquisition, 38(3), 553–586. https://doi.org/10.1017/S027226311500025X