Give two adults the same teacher, the same number of hours and the same textbook, and ask them to learn to read a new language aloud from scratch. One is learning Spanish. The other is learning English. Within a month the Spanish learner can pronounce every unfamiliar word on the page, correctly, without help — including words nobody has ever said to them. Three years later the English learner will still be caught out by a vowel, and so will most native speakers.
This is not about intelligence, motivation or the intrinsic difficulty of the language as a whole. It is about one measurable property of the writing system, and it has a name: orthographic depth. Once you know where your target language sits on that scale, and which direction of it you are fighting, you can stop studying letters the way your first language taught you to.
What Orthographic Depth Actually Measures
Depth is the consistency of the relationship between letters and sounds. A shallow — also called transparent — orthography maps one grapheme to one phoneme almost always. Finnish, Turkish, Serbian, Spanish, Italian and Indonesian sit near that end: if you know the rules, you can read any word you meet, and spell any word you can say. A deep or opaque orthography breaks that mapping. English, Danish, unvowelled Arabic and Hebrew, Tibetan, Thai and the logographic scripts of Chinese and Japanese kanji sit toward the other end, where the same letters can carry several sounds and the same sound several spellings.
Two details make the concept far more useful than the simple shallow-versus-deep label suggests.
First, depth runs in two directions, and a language can be transparent in one and opaque in the other. Reading uses the orthography-to-phoneme route; spelling uses the phoneme-to-orthography route. French is the classic case: measured over monosyllabic words by Ziegler, Stone and Jacobs, French is only about 12.4% inconsistent for reading against English at roughly 30.7% — and then reverses, at 79.1% inconsistent for spelling against English at about 72.3%. French readers can pronounce almost anything; French writers agonise over every ending.
Second, depth is a continuum, not a category. The extremes are certain — Finnish at one end, English at the other — but the middle is fuzzy, and no language is perfectly consistent. That fuzziness is exactly why the practical differences can be so large.
The First-Grade Gap: The Numbers
The best-known measurement of depth in action is a fourteen-country study by Philip Seymour, Mikko Aro and Jane Erskine, published in the British Journal of Psychology in 2003. They tested children at the end of their first school year on letter knowledge, familiar word reading and simple nonword reading, across thirteen European orthographies plus English. By the end of grade one, children in most of the countries with shallow orthographies were already accurate and fluent. English children were nowhere near it.
In the transparent languages — Italian, Greek and Spanish — beginners averaged around 87% accuracy on high-frequency words. Danish and French came in around 80%. English came in at 34%. The authors were blunt about the implication: the rate of reading development in English was "more than twice as slow" as in the shallow orthographies, and the difference could not be explained by the age children started school or how well they knew their letters.
The fundamental linguistic differences in syllabic complexity and orthographic depth are responsible. Syllabic complexity selectively affects decoding, whereas orthographic depth affects both word reading and nonword reading. — Seymour, Aro & Erskine, 2003
A second line of evidence removes the last doubt, because it uses nonsense words. In a study by Usha Goswami and colleagues, seven-year-olds in England, France and Spain — matched for reading age — were asked to decode simple nonwords that could only be worked out letter by letter: zoip, koog. The English seven-year-olds got 12% of them right. The French children managed 53%. The Spanish children, in a language where letters do not argue with you, managed 94%. Two years later, at a matched reading age of nine, English was still at 51% against France's 73% and Spain's 92%.
Nothing about those words was unusual except the script. The children were simply doing what their writing system demanded — and in English, the letters themselves were not enough information.
Why English Is Unusually Awkward
English has about 44 phonemes and 26 letters to represent them. The UK Department for Education's own reading framework describes the code as so complex that the 26 letters, singly or in combination, produce "more than 70 common correspondences between phonemes and graphemes, and hundreds of rare ones" — and it notes explicitly that in Spanish, German or Welsh, one grapheme almost always represents the same phoneme. English is also unusually rich in consonant clusters, which makes early decoding harder again.
The single most useful illustration is the letter a. Look at cat, was, saw, made and car. One letter, five different sounds, all in extremely common words a beginner meets in the first weeks. Add the vowel-digraph problem — ea in mean, meant and break; ough in through, rough, bough and thought — and the beginner is not learning a rule but a set of exceptions with a rule attached.
There is a well-known claim that only about half of English words can be sounded out from rules. It is contested and depends on how many graphemes you allow per phoneme, and a confident-sounding version of it circulates with numbers that other researchers reject. What is not contested is the developmental consequence: English readers learn slowly, and they learn differently.
Grain Size: Deep Orthographies Make You Read Bigger
The explanation that finally tied the cross-language data together came from Johannes Ziegler and Usha Goswami in 2005, and it is called psycholinguistic grain size theory. Its claim is simple: because a deep orthography does not give reliable information at the smallest grain, learners shift upward to larger units — rimes, syllables, whole words and morphemes — and build their lexical representations at that size.
The evidence is a comparison of English and Greek children reading the same kinds of nonwords. The Greek children, working in a transparent orthography, were close to ceiling whether or not the nonword resembled a real word: their seven-year-olds read 92% of bisyllabic nonwords built on familiar patterns and 84% of those with unfamiliar patterns. The English seven-year-olds were rescued by rhyme analogy — 51% correct when a nonword echoed a real word (bicket from ticket) but only 39% when it did not, falling to 27% and then 7% as the nonwords grew longer.
So English readers are not doing a worse version of the Greek strategy. They are doing a different job. Seymour and colleagues argued that deep orthographies force a dual foundation — partly alphabetic, partly whole-word — that takes more than twice as long to establish as the single alphabetic foundation a shallow orthography needs.
Depth Changes the Brain, Not Just the Timetable
Reading does not happen the same way in different scripts, and you can see it in scanners. In a PET study by Eraldo Paulesu and colleagues, skilled adult readers of Italian and English performed the same reading tasks. Both read competently. But the Italians leaned more on left posterior temporal regions associated with assembling sound from print, while the English leaned more on left posterior inferior temporal and anterior inferior frontal areas associated with word form and articulation. The same task, divided differently between the two systems.
The starker result came in 2001. Paulesu's team compared dyslexic and non-dyslexic adults in Italy, France and England, and found two things at once. Italian dyslexics read markedly better — faster and more accurately — than English and French dyslexics, exactly as orthographic depth would predict. And all three groups showed the same pattern of reduced activity in left-hemisphere reading regions, plus reduced grey and white matter density, relative to their own controls. The paper's title says it: cultural diversity, biological unity.
Reading difficulty is a neurological fact. How much it costs you in daily life is decided by the orthography you were born into. That finding has been argued over since — later work questions whether orthography-specific effects show up in the neurobiology of dyslexia itself — but the behavioural difference has been replicated repeatedly.
What Happens When You Learn a Second Script
Adults do not approach a new writing system as blank learners. They arrive with the decoding habits of their first language, and those habits transfer whether or not they help.
Keiko Koda's cross-linguistic work with adult learners of English from Arabic, Japanese and Spanish backgrounds showed the transfer directly. When phonological information was made hard to access during reading, comprehension collapsed for the alphabetic readers — Arabic, Spanish and English — while the Japanese readers, who had grown up reading a morphographic script, were barely affected. They simply switched to the route their first language had taught them: recognise the word, skip the sound assembly.
The same pattern shows up in large-scale modern data. In an analysis of 1,073 adult readers in the English Reading Online Project, readers whose first script was alphabetic or alphasyllabic were faster and more accurate on English single-word reading than readers whose first script was logographic. Interestingly, the logographic group was faster but less accurate on longer word measures — a speed-accuracy trade-off that looks exactly like a strategy that works at home and misfires abroad.
One finding here deserves highlighting because it inverts the usual intuition. In an ERP study with Spanish-English and Chinese-English bilinguals, learners used a different grain size in their second language than their first while proficiency was low — and only transferred their first-language reading units once proficiency in the second language was high. Your mother tongue's habits assert themselves more as you improve, not less. Which means a wrong habit, once you are fluent enough to automate it, becomes harder to see.
There is also direct evidence that first-language transparency shapes how quickly you learn a brand-new script. In a study where adult readers learned an artificial orthography, a perfectly consistent one-to-one code was learned faster and its words recalled better than a one-to-two code. And after the fact, Spanish readers — used to trusting letters — had more difficulty recalling the meanings of words written in the opaque code, while the English readers, who were already used to not trusting them, showed no such penalty.
Cultures That Decided Depth Was a Bug
Every so often a language community looks at its own writing system and decides the inconsistency is not worth keeping.
In 1443, Sejong the Great's court produced hangul for Korean: an alphabet whose letter shapes were designed to depict the articulations that produce them, and which was advertised — in the text explaining it — as something a clever person could learn in a morning and a slow one in ten days. In 1928, Turkey replaced the Arabic-derived Ottoman script with a 29-letter Latin-based alphabet built on one sound per letter, and ran a national campaign of public "national schools" alongside it. Literacy in the 1927 census sat at roughly 10.5%; by the 1935 census it was around 20%, and the adult schools had enrolled something like 2.5 million people by 1936. The cost is real and often glossed over: overnight, several generations lost direct access to their own printed heritage.
And in the same century, a Serbian scribe with almost no formal schooling did the same thing from the other direction — not by inventing an alphabet, but by cutting one down to size until it obeyed the way people actually spoke. That is the other half of this week's reading; his name was Vuk Karadžić.
How to Study Differently, by Depth
The point of all this is not to admire the linguistics. It is to stop wasting months on a method your target language does not reward.
- Diagnose first, in both directions. Ask two questions about your target language: can I pronounce a written word I have never seen? Can I spell a word I have only heard? Shallow in both directions (Spanish, Italian, Finnish, Turkish, Serbian, Indonesian, Korean hangul, Japanese kana): learn the mapping once and then read volume. Shallow for reading but deep for spelling (French, Portuguese): separate the two skills and expect spelling to lag for years.
- In a shallow orthography, get the rules over with in a week — then never study letters again. A deck of one card per letter or cluster, plus a short exception list, is enough. Spanish needs
candgbeforeeandi, the silenth,gueandgui. Italian needsgli,gn,scand double consonants. After that, accuracy is cheap and your metric should be speed: read aloud against a timer. - In a deep orthography, the unit of study is the word, not the letter. English spelling has to be learned as a set of spellings attached to a sound you have actually heard. This is the single strongest argument for audio on every flashcard: the whole problem is that print under-determines sound, so the sound has to come from somewhere else.
- Chunk, because chunking is what native readers do. Build cards around rimes and word families rather than isolated graphemes —
-ight,-ough,-tion,-ould— and around minimal pairs that separate near-identical spellings (beach and beech; through and thorough). This matches the grain size the orthography imposes. - Do not import your first language's reading strategy. If your L1 is transparent, the instinct to pronounce every letter will produce confident errors in English or Danish; you need to learn when to stop trusting the letters. If your L1 is logographic, alphabetic reading needs explicit phoneme work — blending, minimal pairs, decoding drills — because whole-word recognition will not carry you.
- Measure reading and spelling separately. In every orthography spelling is the deeper direction, and in opaque ones the gap can be years wide. A learner who spells badly may still read fluently; a learner who reads fluently may still spell like a beginner. Track them as two skills, because they are.
- If you are learning Japanese, split the writing system in your plan. Kana is close to perfectly transparent and can be front-loaded in a fortnight. Kanji is not phonetic at all, so treat it as vocabulary rather than decoding: learn readings inside real words, never as isolated lists, and expect the whole script to behave like the hardest orthography on earth — because by most measures it is.
The Honest Limits
Transparency is not the same as ease. Finnish is beautifully consistent and morphologically enormous. Japanese kana is trivially decodable inside a writing system that takes Japanese schoolchildren a decade. And decoding is not comprehension: in a five-language comparison by Ziegler and colleagues, the strongest predictor of grade-two reading in the most transparent orthography studied — Finnish — was vocabulary, not decoding skill. Knowing how to sound out a word tells you nothing about what it means.
Depth buys you months at the beginning and nothing at the end. A transparent orthography gets you onto the page sooner, which means more input sooner, which is the real advantage. A deep one taxes you at the door and then hands you a system where the spelling of a word carries its history. Neither is a reason to pick a language. Both are reasons to change your method once you have picked one — and to stop blaming yourself for a difficulty that was designed into the script before you were born.