Why you freeze, and the five things that actually help

Freezing is foreign language anxiety — a measured phenomenon with its own scale, and mostly fear of being judged rather than of the grammar. The largest review of interventions found that several popular remedies were tested and did not work, peer feedback and group work among them, while positive self-talk did.

Updated September 26, 2026 · 7 min read

The thing that happens to you has a name and a scale

You know the word. You knew it an hour ago and you will know it again in the car. In the moment somebody is waiting for you to say it, it is gone. That is foreign language anxiety, and it has been studied since the mid-1980s under that name.

The standard definition is MacIntyre's: "the worry and negative emotional reaction aroused when learning or using a second language." It is measured with the Foreign Language Classroom Anxiety Scale, built by Horwitz and colleagues in 1986, which is thirty-three statements you rate from strongly disagree to strongly agree. The scale separates three things that feel identical from the inside: communication apprehension, test anxiety, and fear of negative evaluation.

That third one is usually the one. It is not fear of the language. It is fear of the audience.

What has been tested, and what happened

In 2021 Toyama and Yamazaki published the first systematic review of interventions for foreign language anxiety. They screened 854 studies and included 40 experimental ones published between 2007 and 2020. It is the best map of this territory that exists, and the most useful column in it is the one nobody reproduces.

Reduced anxiety

Significant effects in the reviewed experiments.

Positive self-talk, inside rational emotive therapyAbood & Abu-Melhim 2015; Toyama & Yamazaki 2019 — both reported a successful decrease
Cooperative work on a task outside the course materialLiu et al. 2018 (digital storytelling); Ku & Chen 2015; Soleimanirad & Shangarffam 2016
Structured feedback from a teacherNosratinia & Abdi 2017 (portfolio assessment); Zarrinabadi & Rezazadeh 2020 — the latter had no control group
A behavioural contract — committing in advance to speakJin et al. 2020 — significant decrease against a control group
Affective strategy instructionBielak 2018 — but the review found only two such studies and only one was significant

Tested, did not help

Measured against a comparison group, and the anxiety did not move.

Peer feedback on its ownBailey & Cassidy 2019 — 'not effective for FLAR'
Group work on its ownYayli 2017 — the review states group work by itself could not reduce anxiety
Email dialogue journalsForoutan & Noordin 2012 — not effective
Synchronous text chat with a teacherBaralt & Gurzynski-Weiss 2011 — not effective

Genuinely unresolved

Not the same as 'does not work'. The evidence is thin or contradictory.

MusicShimbo 2008 — the review calls the result unclear
Relaxation and meditationShimbo 2008; Scida & Jones 2017 — 'effectiveness has not yet been confirmed statistically'
Drama and role-playFour studies; none showed a significant within- or between-group difference
Video, voice and virtual-world practiceNear-identical studies disagree — Melchor-Couto 2017 found an effect, Yu et al. 2020 did not

Drawn from Toyama & Yamazaki's 2021 systematic review of 40 experimental studies. The middle column is the one that matters: these are not things that have never been tried, they are things that were tried and measured.

Why the middle column is worth more than the left one

Every article about speaking anxiety has a version of the left-hand column. Almost none has the middle one, and the middle one is what saves you time.

If you have been told to put on calming music, do breathing exercises, and join a study group, you have been given three interventions of which one was measured as ineffective and two remain unproven after more than a decade of trying. That is not an argument against ever doing them. It is an argument against concluding that you are the problem when they do not work.

The pattern underneath the results is fairly clean. What moved anxiety was things that changed what the learner does or believes — committing in advance to speak, arguing with your own catastrophic predictions, getting structured feedback from somebody whose job is to give it. What did not move it was arrangements of furniture. Putting anxious people in a group does not make them less anxious. It just makes them anxious in a group.

The finding about doing it alone

One result deserves its own paragraph, because it cuts against the most common piece of modern advice.

Liu and colleagues in 2018 had learners do a digital storytelling task either cooperatively or individually. The cooperative group ended with lower anxiety and higher test scores. The individual group ended with higher anxiety than they started with. Not "no improvement" — worse.

I'll practise on my own until I'm ready, and then I'll talk to someone.
The version that feels responsible. It defers the only activity the anxiety is actually about.
I'll talk to one person for fifteen minutes, badly, on Thursday.
Small, dated, and with another person in it. Jin et al. 2020 found committing in advance to speak was itself the intervention that worked.

The shape of the thing. The second version is not braver; it is smaller, and the difference is that somebody is waiting.

That does not mean practising alone is useless — it plainly builds things. It means practising alone is not a treatment for the fear of speaking to people, and there is at least one experiment where it made the fear worse.

What the evidence suggests you actually do

Five things, each of which maps onto something in the left-hand column rather than onto an intuition.

Commit in advance, out loud, to a specific occasion. Not "I should practise more." A time, a length, a person. That is the behavioural contract, and it is one of the cleanest results in the review.

Argue with the prediction, not with the feeling. Positive self-talk in the reviewed studies is not affirmations. It is rational emotive therapy's move: write down what you think will happen, notice it is a forecast rather than a fact, and check it afterwards. The forecast is usually "they will think I am stupid." It is usually wrong.

Do something that is not the language. The cooperative tasks that worked were tasks outside the course material — building something, telling a story, arguing about an order. The language happens while you are busy. Group work aimed at the language did not work; group work aimed at a task did.

Ask for structured feedback rather than correction. Structured feedback from someone whose role is to give it was among the things that worked. Casual peer correction was not. Getting corrected well is its own problem, and the evidence there is specific.

Make the first one small enough to be survivable. Fifteen minutes with one person beats an hour with three, because the thing you are desensitising is fear of an audience and an audience of one is the smallest real one.

What this does not fix, and who it is wrong for

Anxiety and a genuine skills gap are different problems with the same symptom. If you freeze because you have not got the words, no amount of self-talk will retrieve them. Part of the honest answer is that some of the freezing is a vocabulary and retrieval problem wearing a psychological costume — and that half has its own article.

None of these studies is about you specifically. Over two-thirds of the reviewed experiments were run in East and West Asia, mostly in classrooms, mostly with undergraduates, mostly over a few weeks. That is a narrow slice of humanity, and the review says so.

"Unclear" is not "debunked". Relaxation, music and drama sit in the third column because the evidence is thin or contradictory, not because they were disproven. If breathing exercises help you, the review gives you no reason to stop. It only gives you a reason not to expect them to be sufficient.

Clinical anxiety is not this. Foreign language anxiety is situation-specific by definition. If the freezing generalises well beyond language — if it happens in your own language, with people you know — that is a different thing and this literature is not about it.

Sources

Toyama, M., & Yamazaki, Y. (2021) Classroom Interventions and Foreign Language Anxiety: A Systematic Review With Narrative Approach. Frontiers in Psychology, 12, 614184.
The source of the whole matrix above. Screened 854 studies and included 40 experimental ones from 2007–2020. The 'not effective' and 'remains unclear' language is the authors' own, and the distinction between those two is theirs, not ours.
Checked 2026-09-27
Horwitz, E. K., Horwitz, M. B., & Cope, J. (1986) Foreign Language Classroom Anxiety. The Modern Language Journal, 70(2) — the origin of the FLCAS.
The thirty-three-item scale, and the three-part model of communication apprehension, test anxiety and fear of negative evaluation. Described here as reported in Toyama & Yamazaki (2021) and in Humanities and Social Sciences Communications 12, 1223 (2025).
Checked 2026-09-27
MacIntyre, P. D. (1999) Language anxiety: A review of the research for language teachers. Quoted in Toyama & Yamazaki (2021), p. 27.
The definition used at the top of this article: 'the worry and negative emotional reaction aroused when learning or using a second language.'
Checked 2026-09-27

Start my free session

Your first session is free, up to 30 minutes.