How to Measure Your Japanese Progress Honestly

Why the usual numbers mislead

The numbers most learners watch are chosen because they are easy to count, not because they reflect ability. A streak counts days. A card total counts reviews. A level label counts what you could recognise on one morning under exam conditions. All three can climb steadily during a stretch in which your spoken Japanese does not change at all, and that mismatch is what makes people feel they are working hard for nothing.

Useful measures share three features. They are repeatable in the same form, so two runs are comparable. They test production or unrehearsed comprehension rather than recall of studied items. And they are cheap enough that you will actually run them, because a measure that needs an appointment gets skipped in exactly the months you most need the data.

The four measures at a glance

Run all four and they cover the four skills that actually move at different speeds. The recording shows fluency and range, the can-do list shows usable ability, the reading page shows decoding speed, and the listening check shows comprehension of speech you have not tidied up in advance.

Measure one: the fixed monthly recording

This is the single most informative thing on the list and it costs two minutes. Choose one prompt and keep it for a year. Record without notes, without pausing to think of a word, and without a second attempt, because a second attempt measures preparation rather than fluency. Save the file with the date and do not listen back immediately.

A month later, record again on the same prompt and play the two files one after the other. You are listening for four things: total words said in two minutes, the length of your pauses, whether you self-corrected and carried on, and whether any sentence contains a structure the older recording never attempted. Those four move long before any level label does.

Other prompts worth fixing for a year

One prompt is enough, but a set of three lets you rotate without losing comparability, as long as each prompt is compared only with its own earlier recordings. Choose prompts that require narration, explanation, and opinion, since those three demand different grammar and a learner is rarely equally strong in all of them.

Measure two: a can-do list of tasks

Most syllabuses are lists of topics, and a topic cannot be passed or failed. A can-do list is a list of tasks with an outcome: something either happened or it did not. Write yours from situations you genuinely encounter or expect to, not from a textbook table of contents, because a task you will never perform gives you no useful signal when it stays unticked.

Mark each item three ways: done unaided, done with help or repetition, and not yet. The interesting movement is usually from the middle column to the first, and that shift is invisible to any level label. Review the list every three months and rewrite items that have become too easy.

Notice that every item names a situation and an action. Compare that with a topic heading such as travel, which you can neither complete nor fail. If you cannot picture yourself doing the item this month, it belongs on a later list.

Measure three: the timed reading page

Reading speed is a good proxy for how automatic your grammar and vocabulary have become, and it is easy to measure without a test. Take a fresh page of roughly the level you normally read, set five minutes, and read for meaning rather than for perfection. Count how far you got and how many words you did not know.

Do not reuse the same page. The point is fresh material of similar difficulty, because a page you have read before measures memory. If your line count rises while unknown words stay flat, your processing is getting faster. If unknown words fall but the line count does not move, you are learning vocabulary without building speed, and that usually means more reading volume is needed rather than more study.

Measure four: the listening check

Listening is where learners most often overestimate themselves, because familiar material replays a memory rather than testing comprehension. Use audio you have never heard, at normal speed, once. Then, without replaying it, say three points aloud in Japanese. Saying them aloud matters: if you can only summarise in your own language, you understood the content but cannot yet handle it in Japanese.

Score it plainly. Three points with confidence is a pass, one point is a partial, and losing the thread entirely means the material was too hard for measurement purposes. Keep a note of what defeated you, since numbers, names, and fast connected speech fail for different reasons and need different practice.

Keep a corrections log and watch errors disappear

A corrections log is the cheapest long-term measure there is. After each conversation, write down at most three corrections in the form wrong sentence, corrected sentence, one word for the pattern. Three is deliberate: a full list is never reread, while three fit in your head for the next session.

The measure is not the number of entries but which patterns stop reappearing. Every month, scan the log and mark any pattern that has not shown up in four weeks as retired. Watching a particle error you made weekly in March vanish by June is far more convincing evidence than any score, and it also tells you exactly what is still open.

The delay between input and visible output

Input and output improve on different schedules, and confusing the two is a common reason learners abandon a method that was working. A month of heavier listening usually shows up first as easier comprehension, then later as new phrases appearing unbidden in your own speech. The gap between those two moments can be weeks, so a recording taken shortly after an input change often shows nothing.

Plan for the lag. Measure comprehension monthly and speech monthly, but judge an input change on the listening check first and only look for it in the recording a couple of months later. Expect one more oddity: when you start using new structures, your accuracy can dip because you are attempting harder sentences. Falling accuracy alongside rising range is progress, not decline.

What to change when a measure stops moving

A flat measure is information, not a verdict. Check the measure itself first: a can-do list you have outgrown, a reading page you half remember, or a prompt you have effectively rehearsed will all sit still no matter how much you improve. Refresh the instrument before you rebuild the routine.

If the measure is still honest and still flat, change one thing in your practice and set an end date for the trial. The dedicated post on breaking a Japanese plateau goes through the usual causes and the interventions that match them. Whatever you change, keep measuring in the same way, because a new measurement method resets your history and you lose the only comparison you had.

• Recording flat: force longer turns and ban the phrases you use to end them early.
• Can-do list flat: rewrite it from this month's real situations, then retest.
• Reading page flat in speed: increase volume of easy reading rather than adding harder texts.
• Listening check flat: change the source rather than the level, since one speaker or format can be the obstacle.
• Everything flat: reduce the total time and spend all of it on unrehearsed speech for three weeks.