The research behind how we teach

The short version

Every part of how we teach reading comes from research on how children learn to read: letter sounds taught one at a time and in order, books your child can sound out, reading aloud more than once with feedback, and a grown-up talking about the story with them.

That research is about those practices, studied in classrooms and homes with other programs. It is not about TaleTykes. We have measured no reading outcome in any child yet, so we cannot tell you that TaleTykes will change your child’s reading, and neither can anybody else about their own app until they have studied it properly.

How we teach

• One sound at a time. Sound practice starts with hearing the sound in spoken words, then links it to letters, then asks your child to read and spell words with it. A sound only counts as known once your child has read it in words they have never seen before, including made-up ones, so we know it is the sound they have learned and not a word they remember.
• Books they can sound out. A book for a child who is learning to decode is written using the letter sounds they have been taught, and it practices the sound they are working on. We check every word before the book is finished.
• Their own go first. For a new reader aged five to seven, the first read of a book like this starts with your child trying the words. Hearing the page read comes after they have had a go. You can change this for each child.
• Reading aloud, and reading again. Your child can read a page aloud to Pip. We offer another go at the same page, up to four in all. Pip says what went well, and a child of six or older may be shown up to two words to look at together.
• Help on a stuck word, one step at a time. Tapping a word a beginner can sound out plays its sounds first and says the whole word after that, so the help teaches the sounds rather than replacing them.
• A grown-up in the story. Picture books come with a few questions for you to ask about the story, in your half of the app.
• No points, prizes or scores. There are stickers and a count of days read, and nothing to trade them for. Your child is never told how fast they read, how many they got right, or how they compare with anybody.

There is more about the order of the sounds on the phonics page, and about how books are leveled on the reading page.

What the research says, and what it does not

Each entry below is a study or a review we built part of the product on. It says what the research found about children in general, what we did because of it, and what it does not show. The links go to the research itself. Each was opened and checked on the date beside it; if one no longer says what we say it says, tell us and we will correct it.

• Foorman and colleagues, What Works Clearinghouse practice guide, 2016, Institute of Education Sciences, U.S. Department of Education

  What it found.

  A federal panel reviewed the studies on teaching reading in kindergarten through third grade. It rated the evidence strong for teaching children to decode, analyze and write words, moderate for teaching them to hear the sounds in spoken words and link them to letters, and moderate for having every child read connected text every day.

  What we do.

  Sound practice works on one sound at a time, first by ear and then with letters. It asks a child to build words out of their sounds as well as read them, and then to read a book held to the sounds they know.

  What it does not show.

  The guide describes classroom teaching by teachers. It does not review any app, and it rated the evidence for teaching academic language, the one recommendation our books do the most for, as minimal.

  Checked 2026-09-26.

• Ehri, Nunes, Stahl and Willows, 2001, Review of Educational Research

  What it found.

  Across the studies the National Reading Panel reviewed, systematic phonics helped children learn to read more than teaching without it, with a moderate effect overall that was larger when it began in kindergarten or first grade than when it began later.

  What we do.

  Sounds are taught in a fixed order, each one built on the sounds before it, and a sound only counts as known once a child has read it in words they have never seen, including made-up ones.

  What it does not show.

  Later reanalyses found smaller effects when phonics was compared with programs that already taught some phonics. The studies are of teachers and classrooms, not software.

  Checked 2026-09-26.

• Murphy Odo, 2024, Literacy

  What it found.

  Reading decodable text, where the words use only the letter sounds a child has been taught, had a small effect on reading real words and a moderate one on sounding out made-up words.

  What we do.

  A book for a child who is learning to decode is written inside the sounds they have been taught, checked word by word before it is finished, and built around the sound they are working on.

  What it does not show.

  The effects are modest and rest on a small number of studies. Decodable text is useful practice, not a reading program on its own.

  Checked 2026-09-26.

• Pugh, Kearns and Hiebert, 2023, Reading Research Quarterly

  What it found.

  In reading interventions for children who were having difficulty, programs that used both decodable and ordinary text tended to do better than programs with no text reading, while programs using only one kind did not stand out.

  What we do.

  A book held to a child's sounds can also carry a longer line on each page for a grown-up to read aloud, in ordinary language with richer words. This is being prepared and is not switched on yet.

  What it does not show.

  The confidence intervals are wide and the lower ends sit at or near zero, so the finding is fragile. The authors say other differences between programs could explain it.

  Checked 2026-09-26.

• Therrien, 2004, Remedial and Special Education

  What it found.

  Reading the same passage aloud more than once, with feedback, improved how fluently children read it and how well they understood it, and helped somewhat with passages they had not practiced.

  What we do.

  When a child reads a page aloud, we offer another go at the same page, up to four in all, while it is still some way from where a reader of that age usually is.

  What it does not show.

  The gains are largest on the passage that was practiced. Whether practice moves reading on new text is the question that matters, and it is the one we plan to measure.

  Checked 2026-09-26.

• Lee and Yoon, 2017, Journal of Learning Disabilities

  What it found.

  For children with reading difficulties, repeated reading worked better when they first heard the passage read well, and its effects were larger on practiced passages than on new ones.

  What we do.

  After a child has had their own go at a page, they can hear it read, and then read it again.

  What it does not show.

  The review is of children with reading disabilities, working with a teacher. It is a reason for the order we chose, not a result about us.

  Checked 2026-09-26.

• Furenes, Kucirkova and Bus, 2021, Review of Educational Research

  What it found.

  Across 39 studies of children aged one to eight, children understood a story less well on a screen than on paper when the screen version only added narration or highlighting, and did better on a screen when the extras were about the story itself.

  What we do.

  On the first read of a book held to their sounds, a new reader aged five to seven has a go at the words before the page can be read to them. A grown-up can change this for each child.

  What it does not show.

  The studies measured understanding of picture books, mostly read to children, not learning to decode.

  Checked 2026-09-26.

• Mol, Bus, de Jong and Smeets, 2008, Early Education and Development

  What it found.

  When parents asked questions and let the child do the talking while reading a picture book together, children's spoken vocabulary grew more than with ordinary shared reading.

  What we do.

  Picture books come with a few questions for a grown-up to ask about the story, written for those pages, in the grown-up's half of the app.

  What it does not show.

  The effect was much smaller for four- and five-year-olds and for children at risk of reading difficulties, and it is the grown-up doing it that matters, not the questions existing.

  Checked 2026-09-26.

• Deci, Koestner and Ryan, 1999, Psychological Bulletin

  What it found.

  Across 128 experiments, expected, tangible rewards for doing a task reduced how much people chose to do it afterwards, and the effect was stronger in children. Feedback that told people about their competence did not have that effect.

  What we do.

  There are no points, no prizes and no scores for your child. Stickers and a count of days read are never traded for anything. After a child reads aloud, Pip says what went well and may show up to two words to look at together, and never gives a number.

  What it does not show.

  The experiments are mostly short laboratory tasks, not months of reading at home, and they do not settle whether symbolic rewards like stickers matter over that time.

  Checked 2026-09-26.

• Hasbrouck and Tindal, 2017, Behavioral Research and Teaching, University of Oregon, Technical Report 1702

  What it found.

  Oral reading fluency norms are collected in a fixed way: one minute on passages the child has never read, scored by a trained adult, usually the middle of three. The report compiles what children in each grade typically read that way in fall, winter and spring.

  What we do.

  We do not turn a child's reading of their own books into a percentile or a grade, because it is not measured the way the norms were.

  What it does not show.

  Norms describe what is typical, not what a particular child needs. Comparing a reading to them only makes sense if it was measured the same way.

  Checked 2026-09-26.

• Silverman, Keane, Darling-Hammond and Khanna, 2025, Review of Educational Research

  What it found.

  Across 119 studies of reading technology for children in kindergarten to fifth grade, the programs helped decoding and language more reliably than they helped reading comprehension.

  What we do.

  We describe what we build as practice in decoding and reading aloud, and we do not say it improves comprehension.

  What it does not show.

  The studies are of other products, and the average hides a very wide spread between them.

  Checked 2026-09-26.

• Kucirkova, Messer and Sheehy, 2014, First Language

  What it found.

  In one small experiment with 18 preschoolers, children learned new words better from the parts of a book that were about them than from the parts that were not.

  What we do.

  Your child chooses who and what a story is about. We use that to make reading something they want to do, not as the reason they learn, and a character's name is one of the few words allowed outside the sounds a beginner has been taught.

  What it does not show.

  One small study. There is no evidence that a personalized book teaches decoding better than any other book, and we do not claim it.

  Checked 2026-09-26.

What nobody has shown yet

• That TaleTykes works. No study of this product has been done. A reason for building something a certain way is not the same as evidence that it helps, and we do not treat it as if it were.
• That a story about your child teaches reading. Children enjoy it and read more, and that matters. The research that it helps them learn is one small study of preschoolers learning new words, and there is none about decoding.
• That reading apps help comprehension much. Reviews of reading software find it helps decoding more reliably than understanding. We describe what we do as practice in decoding and reading aloud.
• That software hears every child fairly. Software that listens makes more mistakes with young children, and more with some accents and dialects than others. When Pip is not sure what it heard, it counts the word as read rather than marking your child wrong, and a dialect pronunciation is not an error.

What we measure

Today, in your half of the app, you can see how many words your child read correctly in a minute and how accurately, when they read a page of their own book aloud, and which letter sounds they hold. That is how they read one page of a book written for them. It is not a reading age, a grade or a percentile, and we do not turn it into one.

What would tell us whether TaleTykes helps is different, and we are building it in this order:

1. A check of how well Pip hears children, scored against a trained person listening to the same reading, broken down by age and accent. Nothing is compared with school norms until Pip agrees with a person closely enough.
2. A reading check once a month on short passages your child has never seen, read for one minute each, the way schools check reading fluency, with a check of made-up words that use only the sounds they have been taught. This is being built and is not switched on yet.
3. A study of families who agree to take part, with the questions, the measures and the analysis written down and published before it starts. Its first report will describe change over time and will say plainly that it cannot show cause. A later study would compare families who start straight away with families who start a few weeks later, scored by a tester who does not know which is which.

We will publish what those find, including if it is less than we hoped. Taking part is always a separate choice, and saying no changes nothing about your account.