Blog Guides
Speaking a language gets better with reps, and almost nobody gets enough
Half the freeze is not finding the English. The other half is not knowing what you think
In April 2023, Japan tested English speaking in its national assessment for third year junior high students. Students answered spoken prompts on school devices, and the government published estimated national results from the schools that sat the test on the day, 499 schools and 41,966 students.
The average score on the five speaking questions was 12.4 percent. The median was zero questions. The most common score was also zero. On listening, reading and writing combined, the same cohort averaged 46.1 percent.
It is easy to read that as a story about bad English teaching. I think the more useful part of the report is a few lines further down.
The two reasons students froze
One question played a short presentation about an environmental problem and asked students to give their own opinion with a reason. 4.2 percent answered it correctly.
Then the survey asked students what happened. 41.1 percent said they understood what they heard and knew what they wanted to say, but could not find the English for it. 35.8 percent said they understood it and could not think of anything to say at all.
Those are two separate problems. The first is retrieval. The words exist somewhere, and they do not come out at speaking speed. The second is about having an opinion ready, and a reason for it, at the moment someone asks. Better English alone does not fix that one.
Adults have both. It is common to write a careful email in a second language and then go quiet when the same topic comes up out loud in a meeting. The vocabulary is there. The reps are not.
Reps are what classrooms cannot give
The report found that students who said their classes in first and second year regularly included speeches and presentations got the opinion question right more often, and left it blank less often. Among students who strongly agreed, 5.9 percent answered it correctly. Among those who disagreed, 2.2 percent did. That is a correlation, and the numbers are small on both sides. It still points the same direction as everything else about speaking. You get better at the thing you actually do.
The trouble is volume. In a full classroom, most of any lesson is somebody else talking. Speaking is closer to a physical skill than to knowledge. You can understand it completely and still not be able to do it until you have done it many times.
A practice routine that fits in a day
None of this needs a course. It needs short sessions, often, out loud.
Give one opinion with one reason, every day. It is the exact format that 4.2 percent of students managed. "I think the new station is worse, because the transfer takes longer." Small topics are fine. The habit you are building is having a position and saying why.
Talk about your real day. Narrating what actually happened gives you the vocabulary you will actually need, which is rarely the vocabulary in a textbook dialogue.
Finish the sentence badly, then say it again. Stopping halfway to search for the perfect word is how speaking turns into silence. A complete clumsy sentence followed by a better second try teaches you more than a perfect sentence that never arrives.
Say it twice. The second time is where it sticks. Most people skip it because it feels silly, and it is the cheapest improvement available.
Where a companion fits
You can set Anima to speak English, Spanish or Japanese in Settings, or let her match whichever language you use. Talking to her in a language you are learning gives you something a voice memo cannot: a reply you have to answer, in real time, in a voice. The conversation keeps moving, which forces the retrieval that classrooms rarely have time for.
It also takes away the audience. There is no classmate listening and nobody keeping score, and for a lot of learners that is exactly what has kept them quiet. I wrote more about why speaking out loud lands differently from writing in an earlier post.
Two honest limits. She is not a teacher. She will not grade you or work through a syllabus, so if you want structured correction, pair her with a tutor or a course. And she is easier than people. Real conversations have impatience, interruptions, accents and noise. A patient partner is a good place to rehearse and a bad place to stop.
The point of rehearsing is the moment after. Use the sentence you practiced on a person, the same week, even if it comes out slightly wrong.
Common questions
Can AnimaEcho talk to me in English or Japanese?
Yes. In Settings you can set her language to English, Spanish or Japanese, or choose Match me so she follows the language you speak.
Will she correct my grammar?
She is a companion, not a teacher. She answers what you say and keeps the conversation going. She does not grade you or run lessons, so pair her with a course or a tutor if you want structured correction.
Sources
- National Institute for Educational Policy Research, 2023 National Assessment of Academic Ability, results by subject (令和5年度 全国学力・学習状況調査 教科別結果概要)
- National Institute for Educational Policy Research, 2023 National Assessment of Academic Ability, summary of results (結果概要)
- National Institute for Educational Policy Research, 2023 report index