This post is going to unpack the point in history where psychologists and teachers decided that a deficit in phoneme skills to be ‘something wrong’ with your child, and why a child with dyslexia is treated as if he has a medical or developmental problem.
This is a long read, but you need to hear this story.
The story starts with a great research team and a clever experiment. Unfortunately, that experiment was flawed and the conclusions were wrong. Not fraudulent, like the vaccine-and-autism study. Not irresponsible either, it was great science and modern researchers stand on the shoulders of these great scientists. It was merely wrong, in the way that good science is almost always wrong the first few dozen times.
But this particular study caught the public imagination, just like the autism-vaccine study did, and entered our collective consciousness. And the researcher behind it, Isabelle Liberman, wouldn’t let go, using her status and reputation to continue pushing a theory she surely knew to be questionable. In doing so, she caused untold damage to millions of struggling readers – including YOUR child.
It’s a history that should be better known. Especially if you have had your struggling reader tested, found to be deficient in phonological skills, and formally diagnosed as ‘dyslexic’. But it’s not a simple story and requires a bit of a deep dive into the history of reading research.
Definition of Dyslexia
Before we tell the story, a quick refresher on phonological processing and dyslexia.
The International Dyslexia Association (IDA) has a careful definition of dyslexia:
“Dyslexia is a specific learning disability that is neurological in origin. It is characterized by difficulties with accurate and / or fluent word recognition and by poor spelling and decoding abilities. These difficulties typically result from a deficit in the phonological component of language that is often unexpected in relation to other cognitive abilities and the provision of effective classroom instruction.” (We explore it in more detail here.)
‘Phonological component of language’ refers to the child’s ability to identify, isolate, and manipulate phonemes in spoken words. If a psychological assessment finds your child has this deficit, he will be labeled ‘dyslexic’. Without these skills, he will certainly struggle to read. Grade 4 reading levels are almost impossible to acquire without phonological processing skills.
Here, let’s make it simple Can you identify the first sound in ‘cat’ (/k/)? The last sound in ‘cup’ (/p/)? This easy-to-use screener for grade-2 and up demonstrates what we are talking about, the first three skills are shown below. Check it out. And of course, if you worry about your child, try it on them.
ORIGIN OF PHONOLOGICAL AWARENESS
The importance of phonological processing originates from a series of remarkable studies at Yale University’s Haskins Laboratories in the 60′s and 70′s, which the key investigators summarized as follows:
“Awareness of phonological structures is normally lacking in preliterate children and adults; the degree to which it does exist is the best single predictor of success in learning to read; lack of awareness usually yields to appropriate instruction; and such instruction makes for better readers. That some children have particular difficulty in developing phonological awareness (and in learning to read) is apparently to be attributed to a general deficiency in the phonological component of their natural capacity for language.”
This remarkable claim, distilling theory, research, practice, and cause of difficulties in one confident paragraph, was the result of 15 years of research by a remarkable husband-and-wife team, Alvin and Isabel Liberman.
Perception of the Speech Code
In the paper that laid the groundwork for our modern understanding of speech perception, Alvin Liberman presented two related ideas: coarticulation, and categorical perception of phonemes (Liberman, Cooper, Shankweiler & Studdert-Kennedy 1967).
First, ‘coarticulation’: Our experience with reading deceives us into identifying speech as a sequence of crisply-articulated letter-sized sound units (‘phonemes’), for example, the first sound of ‘bid’ is /b/. However that is not the physical reality of spoken language. Liberman showed that we actually hear a single sound with three overlapped phonological segments. The three sounds of ‘bid’ that we imagine are intermingled, we cannot edit out the /b/ with our best audio tools, it remains embedded in the remaining segments and a listener will hear it anyhow.
Liberman’s other finding is that phonological segments are not units but combinations of coarticulated speech gestures. The tongue, lips, jaw, soft palate, larynx, velum, and other speech organs perform an intricate choreography of movement, each cutting corners on the current task in anticipation of the next. Pitch, stress, and nasalization spread over multiple segments.
Displayed on a spectrogram, the waveforms for the /b/ in ‘bid’, ‘bad’, ‘bed’, and ‘bud’ are quite different (feel how different parts of your mouth create them), but we unerringly recover the idealized representation of the intended categorical phoneme /b/. We successfully hear it in spite of heavy background noise or speakers with strong accents. We also discriminate with remarkable precision; ‘ba’ and ‘pa’ are only different in the onset timing of vibrations of the vocal folds, and a difference of 40 ms (four one-hundredth of a second) will shift our perception from one to the other.
At the end of his landmark paper, Alvin Libermen speculated on the nature of speech production and recognition. The ghostly nature of phonemes troubled him. He suggested, then discarded, the idea that speech might be based on the larger invariant units of syllables. Even though phonemes do not really exist, they are psychologically very real, and it is hard to imagine a theory based on the huge number of possible syllables. Instead, he proposed some kind of ‘black box’ brain module that coded and decoded coarticulated speech into strings of phonemes that could be reconstituted back to syllables. This is the natural intuition of every skilled reader of an alphabetic orthography; it would be another 40 years before linguists seriously considered the alternatives (Port 2007, Port 2009).
The nexus between Alvin Liberman’s discoveries and the study of reading is the Alphabetic Principle (Bloomfield 1955), which describes written text as representing a string of phonemes (this is the idea behind teaching phonics). The newly discovered distinction between categorical phonemes and phonological segments opened new avenues for research. And one of the first ones to appear came from his wife and collaborator Isabelle Liberman.
The Phonological Awareness Study
Isabelle Liberman’s 1973 “Segmentation of the Spoken Word and Reading Acquisition” describes the most influential experiment in the reading research canon. Investigating why many children fail to learn to read, and building on the research of her husband, she asked the question “What is required for reading that is not required for speaking or listening?”
She discarded difficulty in identifying letters, associating them with their sounds, or coping with irregularities of English spelling, and focused instead on the child’s ability to segment the phonological structure of a spoken word. It was well-known that disabled readers did poorly on tests of segmenting and blending sounds, and now these activities could be positioned in the developmental acquisition of language.
The hypothesis she presented was that analytically segmenting syllables into phonemes is a developmental step that comes after mastery of explicitly segmenting words into syllables, and that these skills emerge naturally in sequential stages, much like language itself.
Let me rephrase that, because it is the key idea in her paper: Phoneme manipulation skills develop naturally and grow stronger as the child grows older. The youngest kids handle only words, but learn to break words into syllables as they grow older, and finally learn to break syllables into phonemes.
Liberman makes a clear distinction between implicit awareness (we simply ‘know’ that the first sound is /b/) and explicit or conscious manipulation of phonemes (we are able to use our understanding for reading and spelling). Somehow the ‘implicit’ awareness of phonemes becomes ‘explicit’ over time.
Her study does not address HOW implicit knowledge becomes explicit, but rather WHEN the analytic skills of segmenting and blending phonemes are mastered.
The progression from words to sounds, from larger to smaller units, is a historical footnote from I.J. Gelb’s theory of “unidirectional development” in which written languages evolve progressively from logographic (each symbol has a meaning) to syllabic (each symbol denotes a syllable) to segmental (each symbol denotes a phoneme). Gelb (1963) is cited by Liberman who argues by analogy that reading skills follow the same progression.
To compare syllabic and phonemic segmentation, Liberman required 4-, 5-, and 6-year olds to tap out the number of syllables or phonemes in various words with a wooden dowel. She demonstrated that younger children could segment syllables but not phonemes, that older children could segment both but were better at syllables than phonemes, and that both skills improved with age. And most surprisingly, she found that 30% of her oldest students still could not segment phonemes, which had been correlated with a high risk of reading disability.
Something Wrong with Their Brains
The implication from this experiment is that children developmentally acquire the ability to manipulate phonemes in Piagetian stages of learning. This explicit skill follows a large-unit to small-unit progression from words to syllables to phonemes. The lack of explicit awareness does not affect a child’s spoken language skills, but rather impairs accessory skills like reading. Children who don’t or can’t become aware of phonemes are developmentally delayed or impaired in some way – there is something wrong with their brains.
Liberman had asked exactly the right question – what were the extra skills required for reading? Her hypothesis generated a series of testable predictions that kept psychologists busy for years. The modern interpretation of the alphabetic principle and recognition of the difficulty that children might have in understanding this principle is largely due to her work.
This paper shook reading practice like a thunderbolt, and its effects are still with us. Almost every modern reading clinic and remedial program for dyslexic readers starts with phonological exercises, and sensibly attacks reading disability at the root that Liberman identified. Educational practice remains based on her findings.
And most importantly, failure to normally develop phonological processing awareness became the stigmata that marks the dyslexic reader – and a generation of reading researchers focused their energies on determining what was wrong with those children. Thousands of research papers were published about the differences between ‘dyslexic’ and ‘normal’ brains. Eventually this idea became so entrenched that we forgot where it came from.
Liberman’s Work Discredited
Liberman’s paper was badly flawed, and some researchers quickly noted conceptual issues with her study. It is unclear why implicit phonological awareness is the first developmental skill to emerge and explicit skills are the last, or why phoneme segmenting can be instructed while it is generally held that language cannot be taught. Attempts to reproduce her results failed. But it didn’t matter, the horse had left the barn, and was galloping across the education world.
Let’s take a look at one of the conflicting studies.
Fox and Routh (1975) considered the same hypothesis of phased emergent skills, using students aged 3 to 7 but with a simpler methodology. They had the student break off and repeat ‘a little bit’ of the stimulus; for syllables, they prompted ‘Peter’ and looked for ‘Pete’, for phonemes they offered ‘Pete’ and looked for ‘Pee’.
Compare this to the mechanics of Liberman’s tapping game for syllables and phonemes. Syllables have a ‘nucleus’ – a high-energy vowel that beats like a metronome, especially if the onsets are stops (‘potato’, ‘majestic’). A word’s rhythm can be tapped even if the student cannot extract the syllables. Tapping out a phoneme has a higher cognitive load because there are no natural beats to follow, the child must extract and remember the sequence of phonemes and then recall them in beat with the dowel. Liberman’s comparison of syllable and phoneme counting might be confounded by the relative difficulty of the tasks.
Fox & Routh’s syllable and phoneme tasks were of comparable difficulty to each other, age-appropriate for the younger students, and did not involve extraneous activities like tapping a dowel. With their simpler tasks, Fox & Routh’s results were roughly the opposite of Liberman’s; phonemic analysis was mastered at an earlier age than syllabic analysis. Their results still showed that older children were better at tasks than younger ones, but even 3-year olds demonstrated explicit phonemic awareness. Fox & Routh suggested that syllable segmenting emerged later because it required experience with the printed word.
Fox & Routh’s data directly contradicts Liberman’s hypothesis of large-unit to small-unit development of awareness, and a comparison of their methodologies suggests that it is Liberman’s data that may be unreliable.
Tumner & Nesdale (1985) tried to replicate Liberman’s experiment and found it confounded by the older child’s knowledge of spelling. If a child knew how to spell ‘book’ (one of Liberman’s words), then they would tap four times, overshooting the three actual phonemes. I reviewed Liberman’s list of test words (Liberman 1974), and frankly I might have also mistaken ‘toy’, ‘soap’, ‘pout’, ‘mine’, ‘caw’, ‘heat’, ‘out’, ‘toys’, ‘cake’, ‘cool’, and ‘lay’, all of which have accessible spellings and more letters than phonemes.
‘Toy’ has only two phonemes, /t/ and /oy/. ‘Soap’ has only three phonemes. You can see how an older child or adult who knows how to spell might get confused.
Interestingly, Fox & Routh’s 1975 results did not cluster. About half of the young students did very poorly at phonemic awareness, the rest did much better; this bifurcation formed the basis of a follow-up study the next year (Fox & Routh 1976). Here, they gave 4-year-old students training in ‘reading’ left-to-right using letter-like figures, then gave the experimental sub-group training with phonic blending, then tested the two groups for ‘reading’. They found that the extra training improved reading skills – the same way training will improve your child’s reading skills.
Finally, they re-ran the 1975 syllable-splitting measurement to identify the students with strong or poor phonological awareness. They found that the poor segmenters had neither ‘read’ as well as the proficient ones, nor improved in reading with blending training as the proficient ones did. In other words, the correlation between segmenting skill and reading achievement is valid at 4 years and maybe even younger.
Liberman’s inference of staged development cannot be supported if 3-year olds demonstrate explicit phonemic awareness. A developmental impairment is still possible, but that makes the ability to instruct segmenting problematic and begs the question of scarcity of German dyslexics.
Discussion of Phonological Awareness
The correlation between explicit phonological skills and reading is fully supported by Fox & Routh, as it is by almost all studies. There is no question that phonological awareness in early readers is an overwhelmingly powerful ‘predictor’ of later reading skills, and disabled readers are almost always deficient in phonological awareness – that relationship is the most robust correlation in reading research.
But correlation is not causation. YOUR CHILD ALREADY KNOWS HIS PHONEMES. Your child cannot manipulate phonemes explicitly simply because he has not been taught to read properly.
Children seem to have implicit awareness of phonemes long before they are able to segment or blend with them. Eimas & Miller (1980) tested infants (2 to 4 months old) for attention to different nasel-stop syllables (‘ma’ and ‘na’), and concluded that they learn to extract and categorize phonological segments in spoken messages with exquisite accuracy starting in their very first months.
There is also reason to believe that young children have implicit phonological awareness because they can distinguish boundaries between words. We don’t actually leave ‘spaces’ in our speech, words blend together into a continuous coarticulated string of syllables. English ‘word-breaking’ relies on tactics such as ‘pl’ and, ‘gr’ unambiguously marking the beginning or ending of syllables (only three are found in both positions – ‘sp’, ‘st’, and ‘sk’). Different tactics must be learned for other languages, but they all require phonological awareness.
By contrast, the analytic ability to blend and segment phonemes seem to found ONLY in skilled readers of alphabetic languages. In a study of illiterate Portuguese (Morais, Cary, Alegria & Bertelson 1979), and Liberman’s own study of poor readers in an adult literacy class (Liberman, Rubin, Duques & Carlisle 1985) , phonological deficits were reliably identified in adults who had never been taught to read. They weren’t failed readers, they merely didn’t go to school and that seemed to flag them as dyslexic.
Further, students in shallow orthographies do not seem to develop dyslexia in the same manner – either they have an easier time developing phonological awareness or they require it less. Dyslexia is almost unheard of in countries like Italy and Finland, which have regular writing systems. It is only the English writing system that seems to create so many dyslexics.
If explicit awareness comes from learning to read an alphabetic orthography, then readers of syllabaries should not require or acquire this ability. This was tested with emerging readers in Japan (Fletcher-Flinn, Thompson, Yamada & Naka 2011). where kindergarten readers start with syllabic Kanji and then move on to alphabetic hiragana. And sure enough, there was a clear correlation, students who could read Kanji lacked phonological awareness, and as they mastered phonetic hiragana they also displayed greater phonological awareness. By grade 1, all students had mastered the phonetic hiragana and phonetic awareness was uniformly at ceiling. (Incidentally, the incidence of dyslexia in Japan is reported as 0.1%).
All evidence seems to support the correlation between explicit phonological awareness and reading, it is only the direction of causation that is in dispute. McGuinness (1995) investigated the evidence that explicit phonological awareness has a reciprocal causal relationship with reading similar to ones that Stanovich (1986) described in his paper on Matthew effects – a bootstrap process where each skill propels the other forward. Proving a reciprocal causal relationship of this nature is almost impossible, but her evidence is that improvement in either skill predicts improvement in the other.
Tumner & Nesdale (1985) also tried to determine the nature of the relationship between explicit phonological skills and reading by correlating phonological skill and reading ability to methods of instruction. They found that segmenting skills are necessary but not sufficient for building decoding skills, and that the method of reading instruction does not affect the development of segmenting skills unless that instruction specifically addresses to segmenting skills. The latter suggests that phonological awareness is taught indirectly in the process of learning to read, since otherwise phonics instruction would improve segmenting skills better than whole-word instruction.
If phonological awareness bootstraps on learning to read, then can developmental delays or neurobiological impairments account for failure of both? We suggest not, on the basis that reading can be taught to disabled readers. The best-documented case is the Torgesen study with which we launched this blog. Readers will recall that the subjects were grade-3 students who had failed Reading-Recovery and were sitting in LD classrooms – the bottom 2% of readers. After an 8-week intervention, their phonological awareness scores jumped dramatically, and after a delay of two years reading scores did as well.
Does early teaching of phonological awareness improve later reading skills? Byrne, Fielding-Barnsley & Ashley (2000) tried exactly that, but could show only minor benefit after a delay of five years (and even that slight gain could only be found on the Woodcock Word Attack, which correlates to phonological awareness in how it measures phoneme/grapheme knowledge). More importantly, early training did not ‘vaccinate’ the students, they were just as likely to become disabled readers. In light of this failure, they recommend explicit teaching of the alphabetic principle rather than phonological training.
In current practice, phonological skills are called ‘phonological awareness’, and separate assessments measure phonological working memory and phonological retrieval – collectively ‘phonological processing’. But they are not independent skills, and the reason for this hair-splitting is because they are largely nonsense – and would become obvious nonsense if they were simplified. If you have any doubts about this, read Stanovich’s paper on Matthew Effects in Reading.
Liberman Revisits Her Own Work
Liberman updated her paper in 1985 (Liberman & Shankweiler 1985). By then, Gelb’s theory had been discredited (Wolf & Kennedy 2003), and was no longer mentioned. This paper covered more ground and incorporated additional findings from other researchers, for example, Fox & Routh’s paper was cited as supporting Liberman’s position that reading and spelling requires awareness of the internal structure of words; no mention was made of the contradicting results.
The same dowel-tapping study is referenced, and the same claims are made for analytic awareness of phonemes emerging after awareness of syllables. Bradley & Bryant (1983) work is cited in support of explicit training of phonological awareness, adding the complexity of a new layer of rimes and onsets between the syllable and phoneme. McGuinness (2005) offers an elegant critique of that study.
It could have turned out differently. Liberman had asked the same key question as in her first paper – do segmentation skills arise spontaneously or are they a spinoff result of reading training. But the copious evidence accumulated over the decade was ignored and her original position was supported. She cited – but does not consider – the study of illiterate Portugese (Morais, Cary, Alegria & Bertelson 1979), and a similar study of Chinese readers (Read, Zhang, Nie & Ding 1986) who had never been exposed to alphabetic instruction. Liberman herself had suggested the Chinese study in her 1974 paper. Morais’s illiterates might be ignored, but Read’s subjects were educated, skilled readers without a hint of disability; and his study directly addresses the inferences of Liberman and Bradley & Bryant.
This “segmentation” skill, which has been shown to contribute to skilled reading and writing, does not develop with cognitive maturation, non-alphabetic literacy, or exposure to a language rich in rhymes and other segmental contrasts. It does develop in the process of learning to read and write alphabetically. (Read, Zhang, Nie & Ding 1986)
Remarkably, Liberman argued against instruction of the alphabetic principle, explaining that the printed word ‘drag’ has four letters and can be divided into four segments, but it was futile for a teacher to attempt to sound this word out. Such an unhelpful demonstration could only produce the nonsense word ‘duhruhahguh’ and further confuse students – a suggestion that every competent phonics teacher will quickly push aside.
She rewrote the same materials again one year before her death (Liberman, Shankweiler & Liberman 1989), that paper provides the quote which opens this posting. Liberman holds resolutely to the developmental theory of phonological awareness, and the syllables-come-first ordering of analytic skills. Contrary evidence, such as Fox & Routh, the Japanese readers, the Chinese readers, and the Portuguese illiterates, continued to be ignored or baldly cited as being in support.
Even the congenitally deaf were harnessed to the Liberman’s argument. One of Liberman’s colleagues studied successful deaf readers and determined that they had developed phonological awareness, underscoring that phonological representations are not auditory units but abstract building blocks (Hanson 1989). Liberman noted how this paper supports the correlation of phonological awareness and skilled reading, but did not admit the awkwardness such evidence poses to a developmental theory.
Isabelle Liberman was a great scientist, a super-star of the research world. Her students and colleagues read like a Who’s Who of reading research. A remarkable number of papers seem to have a footnote thanking her for helpful comments – even those who challenged her theories (for example Read, Zhang, Nie & Ding 1986). But I could not help but feel that Liberman knew the bricks were coming loose around her developmental theory of phonological awareness. She had invested 30 years into it and could not let it go.
Max Planck once said: “A new scientific truth does not triumph by convincing its opponents and making them see the light, but rather because its opponents eventually die”.
And horribly, the pervasive Liberman hypothesis that the dyslexic child is developmentally impaired lives on, infecting our discourse, our schools, and our research. This is a cancer that hurts millions of students whose only fault is to be poorly instructed.
Because of the stubborn obstinacy of this once-great researcher, we wrongly believe there is something different about the brain of a struggling reader. We have built a large industry around a disorder that doesn’t exist. We shrug impotently at the child who can’t read, and offer accommodations.
On this website, we repeat again and again: there is nothing wrong with a dyslexic child. The fault is in the complexity of the English spelling code and the disorganized way we instruct reading. Yet, the dead hand of Liberman’s assertion of a neurobiological deficiency hangs over reading research like a malevolent cloud.