Research · July 16, 2026 · 12 min read

The Science of Reading: What the Research Actually Says (and the Half Most Apps Skip)

• The research covers two halves. Kids learn to sound out words, and they learn to understand what those words mean. Neither half works alone.
• One formula captures it: Decoding × Language Comprehension = Reading Comprehension. Because the two are multiplied, a near-zero on either one collapses the whole thing.
• That gap explains why a child can finish a phonics app, read fluently, and still miss what the text says.

What the science of reading actually is

The science of reading is a body of research spanning decades of work in cognitive psychology, linguistics, and education that shows how the brain learns to read. Three frameworks anchor that research, and each one gives language comprehension the same weight it gives word recognition. Understanding the three explains why an app that drills sounds and letters can still leave a child unable to make sense of a paragraph.

The National Reading Panel's five pillars

In 2000, the National Reading Panel reviewed the experimental research on reading instruction and identified five areas that effective instruction has to cover: phonemic awareness, phonics, fluency, vocabulary, and comprehension. Phonemic awareness and phonics handle the code, or how letters map to sounds. Fluency bridges the two. Vocabulary and comprehension handle meaning, and the panel treated them as instructional targets in their own right, not as automatic byproducts of decoding.

The Simple View of Reading

Gough and Tunmer proposed the Simple View of Reading in 1986, and it states that reading comprehension is the product of two factors. Decoding (D) multiplied by language comprehension (LC) equals reading comprehension (RC).^[1] Both factors run on a scale from 0 to 1, where 0 means no ability and 1 means perfect ability.

The mechanic that matters is the multiplication. Gough and Tunmer did not add the two factors. They multiplied them, and multiplication behaves differently from addition. A child who decodes perfectly (D near 1) but understands almost none of the language (LC near 0) still lands near 0 on reading comprehension, because any number multiplied by something near zero produces something near zero. Strong reading comprehension cannot occur unless both factors are strong. The model has been widely replicated since 1986 and remains one of the best-supported findings in reading research.

Decoding here means efficient word recognition, fast and accurate reading of familiar and unfamiliar words, not merely sounding out. Language comprehension means deriving meaning from spoken words in sentences and discourse, and it includes receptive vocabulary, grammatical understanding, and discourse comprehension.

Scarborough's Reading Rope

Hollis Scarborough introduced the Reading Rope in 2001 as a visual model of how skilled reading develops.^[2] The rope has two major strands that weave together into fluent reading. The word recognition strand contains phonological awareness, decoding, and sight recognition of familiar words. The language comprehension strand contains background knowledge, vocabulary, language structures, verbal reasoning, and literacy knowledge. Skilled reading emerges only when the threads of both strands become tightly woven and increasingly automatic.

Read together, the three frameworks make the same claim from three directions. The National Reading Panel names comprehension and vocabulary as pillars alongside phonics. The Simple View makes language comprehension a co-equal multiplicative factor. Scarborough’s Rope gives it a full strand of its own with five distinct threads. None of the three treats decoding as sufficient.

The misconception: science of reading does not mean phonics-only

The science of reading is a body of research, but the phrase you hear in the news is a reform movement, and the two are not the same thing. Cognitive scientist Mark Seidenberg says so plainly. “There isn’t a field called ‘the science of reading’ and people rarely identify as ‘reading scientists,’” he writes, noting that in education the term “has been taken up by a movement (often abbreviated SoR) to reform instruction” that “has as yet incorporated very little” of the actual research.^[3] The label has drifted from the science.

Emily Hanford’s podcast Sold a Story pushed the term into public view in 2022, investigating an idea about early reading that “was proven wrong by cognitive scientists decades ago.”^[4] More than half the states passed reading laws in its wake, and that legislation is where the narrowing happened. When a state mandates a phonics curriculum and calls it the science of reading, the phrase starts to mean the mandate rather than the research.

Seidenberg warns about exactly this collapse, cautioning that the movement risks becoming “a new pedagogical dogma, consisting of a small set of tenets loosely tied to some classic but dated research.” Timothy Shanahan adds that the current usage is “a misnomer” that “seems to be less about a science of reading than a science of reading instruction.”^[3]

That narrowed meaning is why a phonics app can market itself as “science of reading aligned” without lying. Decoding genuinely is part of the research, so a tool that teaches decoding well is telling the truth about one strand. The claim is not false. It is incomplete, and the incompleteness is invisible unless you know the research covers a second strand the app never touches.

Why kids who pass phonics still can't comprehend

Because the Simple View multiplies its two factors, a value near zero on either one drags reading comprehension toward zero no matter how strong the other factor is. A child who scores 0.9 on decoding and 0.1 on language comprehension lands at 0.09, not somewhere comfortably in the middle.

Linda Farrell, co-author of the Reading Rockets explainer, makes the mechanic concrete. She can decode Fulfulde, Kanuri, and Wolof with perfect accuracy, yet she understands none of it because she doesn’t speak those languages.^[1] Her decoding factor is near 1 and her language comprehension is near 0, so her reading comprehension collapses. A child who breezes through a phonics app is in the same position when the words carry meaning the child can’t access. The decoding works, but the meaning-building doesn’t.

That failure is two problems, not one. A vocabulary and background-knowledge gap means a child can decode a word and still not know what it means, or lacks the prior knowledge a passage assumes. That one is relatively easy to close by teaching words and building knowledge directly. A verbal reasoning and syntax problem is different: the child knows every word yet still doesn’t follow how the sentence or argument fits together.

The second type is harder to fix, and it shows why “comprehension” is not a single skill an app can bolt on. Building vocabulary, supplying background knowledge, and training the reasoning that connects ideas across a sentence are distinct targets that demand distinct instruction. An app that drills phonics addresses neither.

How to tell if an app actually follows the research

Judge an app by what it makes a child do, not by the phrase “science of reading aligned” on its homepage. Any vendor can name the five pillars for free, so the label proves nothing. Four questions turn the frameworks into things you can check by watching your child use the product for ten minutes.

• Does it teach and assess language comprehension as directly as it drills phonics? Look for retelling tasks, questions that require verbal reasoning, and vocabulary taught in the context of sentences rather than as isolated flashcards. If every activity resolves to sounding out words, the app is working one strand of the rope.
• Does it build background knowledge, or only reading mechanics? Comprehension depends on what a child already knows, which is why Scarborough’s rope names background knowledge as its own strand. An app that treats every passage as a decoding exercise is not doing this work.
• Does it cite a real evidence tier, or only self-described alignment? Ask which ESSA tier a vendor has earned. A “Promising” badge and a “Strong” rating look identical on a homepage while resting on very different evidence. A claim of “research-backed” with no tier and no study is a rationale dressed as proof.
• Does the marketing claim match the product? Naming the pillars and demonstrating them are different things, and the checklist exists to catch the gap.

What the ESSA evidence tiers actually mean

One caution before you apply this. Strong decoding instruction is a genuine virtue, not a warning sign. Decoding is half the formula, and the apps that drill it well deserve credit for that half. The checklist is built to catch omission of the comprehension strand, not to penalize an app for teaching phonics.

Assessing the apps against the checklist

We applied the checklist to three apps we verified directly from their own sites in July 2026. The dividing line is not who mentions comprehension. It is who has real efficacy evidence, and who treats language comprehension as its own instructional strand rather than a byproduct of decoding.

Lexia Core5 has the strongest evidence in the category, and that deserves plain credit. It covers six areas by its own account: phonological awareness, phonics, vocabulary, structural analysis, fluency, and comprehension. Lexia reports that Evidence for ESSA rates Core5 “Strong,” the highest of the four tiers, backed by what it describes as 20 peer-reviewed studies meeting ESSA standards.^[5] That rating is worth confirming at Evidence for ESSA directly rather than taking from a vendor page, which is exactly the habit this article is asking you to build. The narrow, fair critique is that Core5 approaches comprehension inside a word-recognition framework built on Structured Literacy, treating it largely as something that follows from strong decoding rather than as a strand needing its own instruction in background knowledge and verbal reasoning.

Hooked on Phonics says the right things and proves none of them. Its site correctly names all five pillars, including vocabulary and comprehension strategies, and even defines the science of reading accurately as a body of knowledge covering five areas. What it lacks is any evidence: no peer-reviewed studies, no efficacy data, no ESSA tier, only parent testimonials. Its headline claim that research proves phonics is the best way to learn to read carries no citation at all. Naming the five pillars costs a vendor nothing.

Reading Eggs engages the framework least. Its main site does not present the five pillars as a structure, and it makes no explicit mention of language comprehension, vocabulary development, or background knowledge. The focus sits on phonics and game mechanics. Motivation is a genuine virtue for reluctant readers, and it deserves credit. On evidence, Reading Eggs displays an ESSA Level III badge, which is “Promising,” the third of four tiers and substantially weaker than Lexia’s Tier 1. A badge is not a tier. Ask which one.

We did not independently verify HOMER Reading, Read with Ello, or Starfall from their own sites, so we make no claim about their pillar coverage or evidence.

Where writing fits in the research

Writing about text belongs to the language-comprehension strand of Scarborough’s Rope, and the research treats it as one of the strongest levers available for growing comprehension. The 2010 Carnegie meta-analysis by Steve Graham and Michael Hebert, Writing to Read, found moderate-to-strong effects on reading comprehension when students wrote about what they read.^[6] That puts a writing task on the same footing as the comprehension-strategy instruction the frameworks already endorse.

Writing works as a comprehension act because it forces a reader to build and externalize meaning. Graham and Hebert name four mechanisms. Reading and writing draw on the same vocabulary and text-structure knowledge, so practicing one feeds the other. Writing demands metacognitive engagement, because a student has to organize and connect the central ideas before committing them to the page. It produces a permanent artifact the writer can review and revise. And it runs as a communication activity that pushes the writer to internalize the author’s purpose and structure.

When a child writes a summary, a retelling, or a question about a passage, that child is performing verbal reasoning on real background knowledge and vocabulary. Rumo is built around that act. Kids write about what they read, and we render what they wrote back to them as meaning, so the child can see whether the words on the page match the meaning of the text. The visualization step turns comprehension into something checkable rather than a guess.

The handwriting side of this rests on separate neuroscience about how forming letters by hand activates reading circuits, which we cover in The Science Is Clear. The point for this piece is narrower. Rumo implements the writing-about-text recommendation the comprehension research has endorsed for over a decade.

Be clear about the boundary. Rumo has no published third-party efficacy study, no ESSA tier, and no What Works Clearinghouse review. Lexia Core5 has published such studies, which serve schools that require a formal ESSA tier. Rumo’s case is mechanistic: it applies a research-backed comprehension approach that decoding-strand apps skip, and the writing-to-read evidence supports the mechanism, not the product. Decoding still matters and phonics still works, but the comprehension strand is the half most apps leave on the table.

Rumo works the comprehension strand the science of reading has always included, and most apps have always skipped. Kids write about what they read, and we make the meaning visible.