How does orthographic learning happen?

Castles, A., & Nation, K. (2006).   How does orthographic learning happen? From inkmarks to ideas: Current issues in lexical processing, 151.‏   http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.132.5739&rep=rep1&type=pdf

Word recognition (the ability to read words automatically) develops with such remarkable speed that, by the end of eighth grade, we expect children learning to read English to know and recognize over 80,000 words!

How does this ability develop?

Beginning readers must acquire an alphabetic coding system, that is, a mapping system between letters (graphems) and sounds (phonemes).  Phonological processing ability (the ability to analyze and synthesize words into the sounds that compose them) enables the acquisition of the alphabetic coding system.  In order to read single words, however, phonological awareness and an alphabetic coding system are not enough.  One can't read the word "shoe", for example, only using these systems.  In order to read this and many other words, we need specific orthographic knowledge.

It's possible to read single regular English words (like "show") using phonological and alphabetic skills.  But these skills are not enough for irregular word reading.  In English, the correlation between non-word reading (which is based on phonological and alphabetic skills) and irregular word reading is lower than the correlation between non-word reading and regular word reading.  In order to read irregular words we need specific orthographic knowledge.

What is orthography?

Orthography is the system of signs and rules used for graphic representation of language.  Orthography includes letters and punctuation marks, and the rules that describe how to use punctuation marks and how to spell. 

How are orthographic knowledge  and processing assessed?  

There are two methods.  The first is orthographic choice, which measures the child's ability to retrieve orthographic representations of specific words (which is a word: Rain or Rane?).  The second method uses tasks that measure sensitivity to orthographic constrains and to general orthographic characteristics of the language, beyond specific words (which looks more like a word: FFEB or BEFF?). 

One can argue that orthographic choice tasks assess the product of orthographic processing, and that orthographic sensitivity tasks asses orthographic processing itself.  But merely reading words and being exposed to words improves and develops orthographic processing.

Even six year olds are sensitive to the frequency and the legality of orthographic patterns, and even kindergarten children are able to decide that a word string like PESS has a better chance to be a word than a word string like PPES. 

After merely 4 months of formal schooling, first grade French children are sensitive to the fact that consonants can be doubled in French only in medial position, not in initial or final position.  French children can also distinguish between consonants that can and cannot be doubled in French.

Researchers use "print exposure" as an indirect measure of exposure to orthographic patterns.  "Print exposure" is the extent of exposure or experience the child has with written material at home or in kindergarten.  Print exposure is measured by asking the child which books he knows out of a list of popular children's books.  The list also contains names of book that do not exist, in order to make sure that the child is not guessing.  Print exposure is correlated with orthographic skills in 6-7 year olds.  Print exposure has a modest but significant contribution to word recognition beyond the contribution of phonological and alphabetic skills.

A Matthew effect is operating here.  Early in the process of reading acquisition, poor readers begin to be exposed to much less text than their skilled reader friends.  The texts that poor readers are exposed to tend to be too hard for them, and thus the reading experience tends to be unpleasant.  As a result these poor readers expose themselves to books even less, and the gap between them and skilled readers widens. 

Second grade orthographic processing skills explain single word reading variation in third grade, beyond phonological processing ability.

Is there a relation between semantic skills and single word recognition?

In English, which has deep orthography (there is no direct link between graphemes and phonemes.  Words are not written as they sound), children have to learn to read many irregular words.  One way to do it is to use oral vocabulary.  The child partially decodes a written word, and uses oral vocabulary to reach full decoding.  Gradually, children with good vocabulary will develop better word recognition skills.  Semantic measures like vocabulary, semantic fluency, familiarity with synonyms and listening comprehension predict single word recognition in eight year olds, beyond phonological skills.  The relationship between verbal-semantic skills and word reading is maintained over time, with measures taken at 8 years predicting unique variance in word recognition some 4 years later when the children were 13 years old.

Some children have difficulty using orthographic skills for single word recognition.  These children have surface dyslexia. These children's alphabetic decoding is intact.  They can read non-words at an average level, but they read real irregular words at a level that is at least one standard deviation below average.  The errors they make in irregular word reading result from the use of alphabetic decoding rules.  For example, they read "break" as "breek" or "shoe" as "show".  They have difficulty with orthographic choice tasks (which is a vegetable: BEEN or BEAN?).  These children show no signs of semantic or vocabulary difficulties.  They don't have visual disabilities or poor visual memory.  They simply did not acquire specific orthographic word reading skills.

Israeli Professor David Share developed a theory according to which orthography is self taught.  His studies that support the theory look like this:  eight year olds read short stories, each containing one new word, which is repeated four to six times in the story (a non-word, for example the name of a fictional place – "Yate").  After three days, orthographic learning of the new word is assessed.  The word "Yate" is presented with a homophone like "Yait" and non-homophonic foils that contain the same letters, like "Taye".  The child has to recognize the word that he read.  Children recognize the right word 70% of the times.  They can also read the right words faster than the foils, and tend to spell them correctly.  Thus, there is clear evidence for orthographic learning in eight year olds, even after only four exposures to a novel word.  Sometimes even one exposure is enough!

However, one or four exposures to words don't cause orthographic learning at the beginning of first grade.  Hebrew orthography (with diacritical marks) is transparent (the relation between grapheme to phoneme is straightforward).  Share argues that transparent orthography encourages a reading pattern that is similar to surface dyslexia.  Children at the process of Hebrew reading acquisition lean on the alphabetic system.  Only after exposure to a significant number of words they begin to develop orthographic sensitivity.  Share argued that children learning to read English teach themselves orthography at an earlier age than children learning to read Hebrew.

It's reasonable to assume that in any point in time, a child may read specific words slowly and laboriously, leaning heavily on context and alphabetic coding, while reading other words quickly, automatically and with a minimal contextual influence.

How does orthographic learning happen?  Does the child generalize orthographic rules out of words he is exposed to?

In French, some consonants can be duplicated, but the letter K is never duplicated (there are no French words that contain KK).  Vowels are never duplicated in French.  French children at the beginning of first grade think that non-words like TUKKE "look more like a word" than non-words like TUUKE.  Does this mean that they generalize the rule "you can duplicate consonants but can't duplicate vowels"?

Apparently, orthographic learning is more similar to statistical learning.  Children learn which specific consonants are duplicated in French and which are not, or rather which consonants have a better or a lesser chance to be duplicated.

First grade French children think that the non-word BUKKOX "looks more like a word" than the non-word "BUKOXX", despite the fact that both KK and XX don't exist in French.  Apparently, this means that the children apply the rule "a consonant is duplicated only in the middle of a word and not at the end of it".  But these children succeeded even better to judge that the non-word GOLLIR "looks more like a word" than the non-word GOLIRR.  The letter strings RR and LL do appear in French.  If the children were only using the rule "a consonant is multiplied only in the middle of a word" there would have been no difference in success between the pair GOLLIR/GOLIRR and BUKKOX/BUKOXX.  The fact that there is a difference suggests that children don't only apply this rule, but are also aware of the statistical probability of doubling LL and doubling KK.