Saying the answer out loud: what changes when a child speaks instead of types

Watch a child finish a mental-abacus drill and you will see two different speeds. There is the speed at which the answer arrives in their head, and there is the speed at which their finger finds the digits on a keypad. Parents time the second one and think they measured the first. This post is about what happens when you take the keypad out of the loop entirely: the sum is read aloud, and the child says the total.

Your child can calculate faster than they can type

The ceiling of trained mental calculation is startling. As listed by Guinness World Records when we checked on 30 July 2026, the fastest mental arithmetic (15 sets of three-digit numbers) – Flash Anzan™ is 1.61 seconds, achieved by Rikako Miyamoto in Nakano, Tokyo, on 18 February 2024. Fifteen three-digit numbers, added, in under two seconds. At that level the calculation has become faster than the hands that would report it — and the same gap, smaller but real, exists in your living room. The research literature is also blunt about how scarce the underlying resource is: the standard reference for the mental abacus, Frank and Barner in the Journal of Experimental Psychology: General (2012), reports that users appear limited to holding about three or four abacus columns at a time, each column held as a separate object in visuospatial working memory. That is a laboratory finding about trained users rather than a measurement of your own child — and the same research group’s later randomised trial in Vadodara, India, with 204 children enrolled and 183 completing, found that mental-abacus students outperformed controls on arithmetic on the study’s own measure while the training “did not alter basic cognitive abilities”. Either way, hunting for the 7 key while holding four columns is not a neutral act.

Listening and voice answering are two different things

• Listening, or dictation: the question arrives by ear. Numbers are read aloud, one at a time, and nothing stays on screen to re-read. The child cannot go back, so the running total has to live in their head from the first number.
• Voice answering: the answer leaves by mouth. The child says the total instead of typing it, and the app matches what it heard against the expected number.
• The two combine into four formats: read and type, read and say, hear and type, hear and say. Only the last one is genuinely hands-free, and each of the middle two is useful on its own.
• They also train different muscles. Dictation trains holding a number you are not allowed to re-read. Voice answering removes input time from the measurement, so what you time is closer to what you meant to time.

Reading sums aloud is old; typing them is the new part

Nothing about dictation is exotic. In centre-based classes it is routine — a teacher calls the numbers and the room calculates by ear — and listening exercises appear as a named activity on the programme pages we read in this category. It is also a scored category in Kani’s own competition simulator rather than a warm-up: the simulator builds 200-question category papers that pair the visual categories with flash and listening knockouts, and any category paper can be printed with a matching answer key. What is historically new is not reading aloud. It is the assumption, arriving with software, that the answer must be typed.

What changes when the answer is spoken

• Latency becomes honest. Between the last number and the spoken total there is no keypad, so the number you see on the timer is thinking time plus one word, not thinking time plus motor skill.
• Hesitation becomes audible. A child who says forty… two is telling you the second digit was still being assembled when the first left their mouth. On a keypad that information does not exist.
• Eyes come off the screen. When the sum is read aloud, the child stops reading and starts holding — which is the actual skill the mental abacus is for.
• Hands stay free. For a child who is still learning to write numerals fluently, there is no second, unrelated skill standing between the answer and the score.
• It gets loud, and that is data too. A parent in the next room can hear whether the answers are confident or crawling, without leaning over the screen.

Who it helps most

Three groups get the most out of it. First, children who are ahead of their own handwriting: a young beginner on the first rung of the ladder can count and add long before writing numerals is automatic, and voice answering lets the arithmetic run at its own pace. Second, children for whom writing or typing is the slow, effortful part — including motor difficulty or dysgraphia. Taking the keypad out removes that tax from the practice session. That is a convenience, not a treatment, and we make no clinical claim for it. Third, anyone practising away from a desk: dictated questions and spoken answers survive a car ride, a walk, or ten minutes of waiting for a sibling far better than a format that needs two hands and a steady surface. One thing it is not: this is still a visual product, and voice answering is not designed as an accessibility tool for blind or low-vision learners.

The limits, stated plainly

• The numbers are read in English. Kani’s number audio is a single English set served to every locale, so an Arabic-, Spanish- or French-speaking child hears the digits in English. That is a deliberate choice, not a localised narration feature, and we would rather say so than let you discover it in week one.
• Speech recognition is the weak link, not the child. A fan, a television, or a sibling talking across the table will cost you answers. Practise somewhere quiet before you judge the format.
• A misheard answer is a technical retry, not a mistake. Teach the child to say the whole number once, clearly, and then stop — and score a recognition failure as a re-run rather than a wrong answer, or the accuracy figure stops meaning anything.
• It needs microphone permission, and it is at its best in the native Android and iOS apps, where the device’s own speech recognition does the work. In a browser it depends on what that browser and device support.
• It does not replace typed practice. Writing the answer is still what a paper exam asks for, so keep some typed and some written work in the rotation.

What week one actually looks like

Do not start with both changes at once. Week one, change only the input: keep the sums on screen and let the child say the total. Two or three short sessions is enough to get past the novelty of talking to a phone, and expect the first session to be worse than usual — most children overshoot, say the number twice, or trail off. Then switch the question side: run dictated sums with a typed answer, so the only new thing is listening. Only after both feel ordinary should you put them together and run hear-and-say sets, which is the format that will feel fastest and the one that most reliably exposes a shaky total. Keep the sets short. A child who is holding three or four columns and speaking under time pressure is working harder than the clock suggests.

Why no other program advertises this yet

When we read the published pages of the programmes and apps in this category through July 2026, we found no product advertising speech-recognition answering, in any language. The speech-adjacent things we did find were of two kinds. One is the teacher-read listening exercise described above, which needs no software at all. The other is hardware: SoroTouch, the tablet app from SoroTouch Co., Ltd. of Chiyoda-ku, Tokyo, lists speaker, microphone and camera among its device requirements and states that it is only available on tablets and cannot be used on smartphones or computers — a requirement list, not a described feature. Among the self-serve online programmes we read, including Abacus4Everyone and Smartick, no page described answering by speech. Two caveats, because they matter more than the finding: this is what published pages said on the days we read them, not an audit — no operator in this category publishes audited product figures, Kani included — and the likeliest reason the feature is rare is unglamorous. Centre programmes already have a human in the room who can read aloud and listen. Nobody needed to solve it in software until the classroom was a phone.

SoroTouch, Abacus4Everyone, Smartick, Flash Anzan and Guinness World Records are trademarks of their respective owners. Kani is not affiliated with, endorsed by, or licensed by any of them.