Some tweeters (hi @MiriamFein @PATSTONE55 @DavidWray @ManYanaEd) have asked whether the summaries of the findings about phonics and top-down guessing strategies in Chapter 11 are accurate. They are, but the book isn’t a review of the literature and so only a few representative studies are cited and discussed in detail. Here’s some additional information. I’ll split this into two posts. This one about “phonics”. The next will be about “the guessing game” and top-down strategies in reading.
I’ve left the comments section open. Someone post, please, so that I can see if it’s working correctly!
Phonology and Reading
Table 11.2 lists a bunch of findings, most discussed in earlier chapters. I have only updated previous reviews of this literature slightly, adding some recent findings from the neuroscience. However, I will post a version of the table that has at least one reference to a primary source for each entry, ASAP.
As always, a lot of concerns focus on “phonics”. Phonics, schmonics: The deeper point is that the use of phonology in reading (i.e., comprehending texts read silently) is inevitable and almost unstoppable. We can talk about the properties of these codes and how they are learned as though they are independent, but in skilled readers the two types of information are thoroughly intertwined, “like a couple of linguistic codependents with serious boundary issues” as I wrote in the book.
The basic principle is that reading depends on speech, at every level. From the properties of writing systems (which are codes for representing spoken language), to the impact of children’s early spoken language experience on reading acquisition and skill, to the use of phonological information in silent skilled reading, to the role of phonological impairments in reading difficulties—and more. *
It’s important to know that Smith, Goodman, Clay et al.’s specific claims about phonics were wrong, but it’s more important to understand why reading and speech are so inextricably connected. Many of the earlier chapters are about this.
• Chapter 2, about reading and speech: comprehending texts requires recovering information about their sounds. We do it all the time.
Permit me to illustrate. Get me the parking permit. One day at band camp? You talking to me? These are just vivid examples of a very general phenomenon: Understanding texts requires recovering phonological information that print doesn’t represent systematically.
• Chapter 3, on writing systems: The linkage between print and sound is built into the fabric of the written code.
Writing systems are solutions to the problem of representing spoken language in visuo-graphic form. It’s what they all do, despite their superficial differences. Writing systems that represent words as visual patterns independent of their pronunciations do not exist. Such systems are inadequate for representing large numbers of words: Too hard to learn or use.
• I later discuss various behavioral findings showing the close links between spoken languge development and reading acquisition. The evidence is everywhere; it doesn’t rely on one type of study or population or lab or even writing system or language. It all converges on the same conclusion: reading depends on speech.
I particularly like
— the prospective longitudinal studies of children who are or are not at risk for reading difficulties (Scarborough, Molfese, Snowling and colleagues)
— the evidence about how variability in early spoken language experience affects learing to read (richness of the language the child hears; impact of the language or dialect on reading)
— the evidence that good readers are better at computing phonological codes (Perfetti’s early studies, many later ones). Work-around strategies (like guessing words based on context plus some letters) are not efficient because of the statistical properties of texts discussed in Ch. 5. They lead to something very much like skimming—picking through a text rather than reading it closely.
— studies of skilled readers showing they cannot inhibit using phonological information. We can set up reading experiments in which the use of phonological information interferes with performance (such as the Van Orden and tongue-twister experients). The reader would do better if they could suppress phonology, but they can’t: it’s not a switch that can be turned on or off.
— the brain evidence shows why: orthography and phonology become integrated in the neural system that supports reading and spoken language. The spelling area, for example, is thoroughly penetrated by knowledge of phonology. The phonology we’ve learned from using spoken language gets changed by exposre to print. (“Phonemic awareness” results from this.)
Treating reading as a “visual” process (or visual + context) is a non sequitur because the visual and phonological codes are not separable at the neural level, or in behavior.
— The PNAS study cited on p. 211 (Rueckl et al. 2015) shows that the integration of the neural systems for reading and speech is a hallmark of reading skill. This was observed in studies of 4 languages with different types of writing systems.
What About “Phonics”?
The logical problem of learning to read, as I called it, is developing a new conduit linking knowledge of print to existing systems for speech. Gough had that part right years ago. We now have vastly more behavioral and neurodevelopmental evidence about how this occurs–but especially how the systems shape and change each other.
This integration of spoken and written language has several parts.
- knowledge of spoken language (vocabulary, lexical phonology, prosody, lexical quality; how language relates to the world, of course)
- knowledge of print (orthographic statistics: how letters combine, frequencies of letters and letter combinations, knowledge of combinations that occur—the spellings of words—and ones that are unlikely or do not, etc.)
- knowledge of how the visuo-graphic and spoken codes relate to each other (“phonics”)
This analysis is accurate (though much detail has been left out). It does not, however, speak to the pedagogical issues. As I wrote, “Incorporating phonics in a serious way requires addressing some tough questions: how to teach it, how much is enough, how much is too much, how to integrate it with reading and literacy activities, how much to individualize instruction, and so on.” The research does not come with pedagogical prescriptions attached.
There’s hope, however. Learning the relations between print and sound is a statistical learning problem, similar to many others we engage in: learning a first, spoken language; recognizing letters, objects, faces; and many other things.
Most of this learning goes on without awareness, in the background as we engage in behaviors like reading and speaking. But we now have a much better idea of why even a small amount of well-targeted explicit instruction can be highly effective.
Take vocabulary. There’s only limited time to explicitly teach vocabulary, but words that are taught also have an impact on learning other, related words. See the discussions of the statistical properties of language and spelling, and the Landauer and Dumais statistical learning theory (p. 112) and here (item 5). Same for learning mappings between written, spoken language.
Instruction is the visible tip of the learning iceberg; implicit statistical learning is the mass below.
In the near future we should have much better tools for calibrating what to teach when: given the state of the child’s knowledge and what they need to learn, which explicit learning experiences would have the biggest impact? (See the endnote for p. 113 on p. 319).
Next post: about the psycholinguistic guessing game.
- Conditions such as deafness and some types of developmental reading impairments require developing alternative solutions—work-arounds. But there is zero evidence that these are good models for children who do not have these conditions.