Kana Mastery: Hiragana, Katakana and the Lookalikes That Slow You Down

Why kana mastery needs more than a random quiz

A random ten-question quiz is useful once you already know most characters, but it can hide two different problems. First, recognition can feel easy while active recall is still slow: seeing し and picking “shi” from choices is different from producing the reading yourself. Second, visually similar characters can remain fragile even when your overall accuracy looks good. Katakana is especially famous for contrasts such as シ/ツ and ソ/ン, where a small change in stroke direction changes the character.

DailyBrainArc already provides row selection, Recall 10, Reverse 10, Sprint 60, Weak Kana, Daily Kana and a full chart. V0.150 does not duplicate those modes. Kana Mastery organizes them around four progressively broader scopes and adds an explicit Confusion Lab so a learner can see what still needs work before calling a scope complete.

The four scopes

The five-step path

1. Study. Review the scope composition and its curated lookalike contrasts. Marking Study complete only records that you reviewed the map.
2. Direct Recall · 20. See kana and type romaji from memory. The checkpoint is 18/20 first-attempt correct. A first miss still counts as a miss even if the retry is correct.
3. Reverse · 20. Start from a romaji reading and choose the requested script. The checkpoint is again 18/20 first-attempt correct.
4. Confusion Lab · 12. Work through two-choice lookalike contrasts. The checkpoint is 11/12 first-attempt correct.
5. Mastery Check · 24. Ten direct, ten reverse and four confusion prompts. Clearing it needs 22/24 overall plus at least 9/10 direct, 9/10 reverse and 3/4 confusion.

All five steps are open from the beginning. The sequence is guidance, not a forced lock. If you already know a script, you can jump directly to a later check and use the result to decide what to review.

How the lookalike drill works

The Confusion Lab uses a finite, visible contrast list rather than pretending an algorithm can measure visual similarity for every character. The most important katakana set includes シ/ツ and ソ/ン. Other useful contrasts include ク/ケ, タ/ク, ノ/メ, メ/ヌ, ウ/ワ and ル/レ. Hiragana gets its own contrast set including ぬ/め, れ/わ, る/ろ, は/ほ, き/さ and あ/お.

For every contrast, the prompt names the reading and asks you to choose the right kana. A miss is recorded under that exact confusion pair. Because the kana was actually presented and answered incorrectly, that character may also enter the normal Weak Kana queue. The system never turns characters that were not prompted into fake mistakes.

How to inspect a lookalike instead of guessing

When two kana keep colliding in memory, do not rely on a vague impression that one is “more vertical” or “more curved.” Give the contrast a stable verbal cue tied to a visible feature, then test that cue in both directions. For シ and ツ, compare the overall direction of the short strokes and the long sweeping stroke; for ソ and ン, pay attention to how the marks and final stroke are oriented rather than memorizing the pair as two unrelated pictures. For ノ, メ and ヌ, first identify the shared diagonal structure, then notice which extra crossing or loop changes the character.

The same method works for hiragana. With ぬ and め, locate the loop and the finishing shape; with る and ろ, compare the ending; with は and ほ, notice the additional horizontal stroke. A cue is useful only if it helps you retrieve the right kana on a later prompt. If a cue stops helping, replace it. The Confusion Lab records the pair you actually missed so you can practice that contrast again instead of rereading the entire chart.

After a miss, make one clean comparison, say or type the reading, and then return to mixed recall. Spending too long staring at one pair can make the drill feel familiar without proving retrieval. The checkpoint deliberately mixes several contrasts so a correct answer has to survive context changes.

Why direct and reverse recall are kept separate

Direct recall asks “what sound does this kana represent?” Reverse recall asks “which kana in the requested script represents this sound?” Those tasks overlap, but they are not identical. A learner can recognize け as ke yet hesitate when asked to produce the katakana for ke, or know a sound but reach for the wrong script when hiragana and katakana are mixed.

That is why the Mastery Check requires both directions instead of combining everything into one percentage. The script label remains explicit on reverse prompts, and entries whose common romanization would create an avoidable collision are not used as reverse targets. This keeps the result interpretable: a weak reverse score means the production direction needs work, not that the quiz accepted two indistinguishable answers.

Accuracy before speed

Speed is valuable only after the mapping is reliable. The mastery checkpoints therefore use first-attempt correctness, not reaction time or a population percentile. Once a scope is stable, the existing Kana Sprint provides a separate 60-second fluency loop. Sprint personal bests remain tied to the exact script, set and row selection so a small easy pool is not silently compared with the full 208-entry set.

This separation matters. A fast run can be fun and motivating, but one fast session should not overwrite evidence that specific characters are still being confused. Mastery and fluency answer different questions and stay separate.

Use Weak Kana as the next action

Weak Kana contains characters you actually missed in this browser. Three correct Weak Kana review attempts retire an item. Kana Mastery reads that real queue when recommending the next scope: if several weak characters fall inside one incomplete scope, that scope can move ahead of a purely alphabetical progression. Confusion-pair history can also break ties so a repeated lookalike problem is visible instead of buried in one overall percentage.

This is local study guidance, not an adaptive-learning certification. There is no account, cloud classroom, official proficiency score or hidden population comparison.

Romaji and answer rules

The inherited kana authority uses canonical romanizations while accepting common input variants for familiar cases such as shi/si, chi/ti, tsu/tu and fu/hu. Rare ぢ and づ entries remain available for direct recall but are excluded from reverse prompts where their common romanized answers would collide with じ and ず. The mastery layer preserves those rules rather than inventing a second transliteration system.

A practical study routine

Start with Hiragana Base Study, then run Direct Recall. If you miss several characters, use Weak Kana before forcing another mastery attempt. Once Direct and Reverse are steady, run the Confusion Lab and note whether one pair keeps returning. After the mixed Mastery Check clears, move to Katakana Base and repeat. Only then mix Both Base and finally Full Kana. If you already know both scripts, you can skip ahead, but the separate results still show whether direct, reverse or lookalike retrieval is the weak link.

After accuracy is stable, finish with a one-minute Sprint. Treat your sprint history as a comparison with your own previous sessions under similar conditions, not as an official reading-speed benchmark.

What this path does not prove

Clearing the local checkpoints does not prove Japanese reading fluency, JLPT readiness, school-grade equivalence, permanent memory or a population percentile. Real reading adds vocabulary, grammar, kanji, context and natural orthography. Kana Mastery is deliberately narrower: it gives you transparent evidence about retrieving the kana mappings represented in this frozen practice bank.