The Phonological Model
To understand how the phonological model works, one has first to consider the way in which language is processed in the brain. Researchers conceptualize the language system as a hierarchical series of modules or components, each devoted to a particular aspect of language. At the upper levels of the hierarchy are components involved with semantics (vocabulary or word meaning), syntax (grammatical structure) and discourse (connected sentences). At the lowest level of the hierarchy is the phonological module, which is dedicated to processing the distinctive sound elements that constitute language.
The phoneme, defined as the smallest meaningful segment of language, is the fundamental element of the linguistic system. Different combinations of just 44 phonemes produce every word in the English language. The word “cat,” for example, consists of three phonemes: “kuh,” “aah,” and “tuh.” (Linguists indicate these sounds as |k|, |æ| and |t|.) Before words can be identified, understood, stored in memory or retrieved from it, they must first be broken down, or parsed, into their phonetic units by the phonological module of the brain.
In spoken language, this process occurs automatically, at a preconscious level. As Noam Chomsky and, more recently, Steven Pinker of the Massachusetts Institute of Technology have convincingly argued, language is instinctive—all that is necessary is for humans to be exposed to it. A genetically determined phonological module automatically assembles the phonemes into words for the speaker and parses the spoken word back into its underlying phonological components for the listener.
In producing a word, the human speech apparatus—the larynx, palate, tongue and lips—automatically compresses and merges the phonemes. As a result, information from several phonemes is folded into a single unit of sound. Because there is no overt clue to the underlying segmental nature of speech, spoken language appears to be seamless. Hence, an oscilloscope would register the word “cat” as a single burst of sound; only the human language system is capable of distinguishing the three phonemes embedded in the word.
Reading reflects spoken language, as my colleague Alvin M. Liberman of Haskins Laboratories in New Haven, Conn., points out, but it is a much harder skill to master. Why? Although both speaking and reading rely on phonological processing, there is a significant difference: speaking is natural, and reading is not. Reading is an invention and must be learned at a conscious level. The task of the reader is to transform the visual percepts of alphabetic script into linguistic ones—that is, to recode graphemes (letters) into their corresponding phonemes. To accomplish this, the beginning reader must first come to a conscious awareness of the internal phonological structure of spoken words. Then he or she must realize that the orthography—the sequence of letters on the page—represents this phonology. That is precisely what happens when a child learns to read.
In contrast, when a child is dyslexic, a deficit within the language system at the level of the phonological module impairs his or her ability to segment the written word into its underlying phonological components. This explanation of dyslexia is referred to as the phonological model, or sometimes as the phonological deficit hypothesis.
According to this hypothesis, a circumscribed deficit in phonological processing impairs decoding, preventing word identification. This basic deficit in what is essentially a lower-order linguistic function blocks access to higher order linguistic processes and to gaining meaning from text. Thus, although the language processes involved in comprehension and meaning are intact, they cannot be called into play, because they can be accessed only after a word has been identified. The impact of the phonological deficit is most obvious in reading, but it can also affect speech in predictable ways. Gregory’s dilemma with long or novel words, for example, is entirely consistent with the body of evidence that supports a phonological model of dyslexia.
That evidence began accumulating more than two decades ago. One of the earliest experiments, carried out by the late Isabelle Y. Liberman of Haskins Laboratories, showed that young children become aware between four and six years of age of the phonological structure of spoken words. In the experiment, children were asked how many sounds they heard in a series of words. None of the four-year-olds could correctly identify the number of phonemes, but 17 percent of the five-year-olds did, and by age six, 70 percent of the children demonstrated phonological awareness.
By age six, most children have also had at least one full year of schooling, including instruction in reading. The development of phonological awareness, then, parallels the acquisition of reading skills. This correspondence suggested that the two processes are related. These findings also converge with data from the Connecticut Longitudinal Study, a project my colleagues and I began in 1983 with 445 randomly selected kindergartners; the study continues in 1996 when these children are age 19 and out of high school. Testing the youngsters yearly, we found that dyslexia affects a full 20 percent of schoolchildren—a figure that agrees roughly with the proportion of Liberman’s six-year-olds who could not identify the phonological structure of words. These data further support a connection between phonological awareness and reading.
During the 1980s, researchers began to address that connection explicitly. The groundbreaking work of Lynette Bradley and Peter E. Bryant of the University of Oxford indicated that a prechooler’s phonological aptitude predicts future skill at reading. Bradley and Bryant also found that training in phonological awareness significantly improves a child’s ability to read. In these studies, one group of children received training in phonological processing, while another received language training that did not emphasize the sound structure of words. For example, the first group might work on categorizing words by their sound, and the second group would focus on categorizing words according to their meaning. These studies, together with more recent work by Benita A. Blachman of Syracuse University, Joseph E. Torgesen of Florida State University and Barbara Foorman of the University of Houston, clearly demonstrate that phonological training in particular—rather than general language instruction—is responsible for the improvements in reading.
Such findings set the stage for our own study, in the early 1990s, of the cognitive skills of dyslexic and nondyslexic children. Along with Jack M. Fletcher of the University of Texas–Houston and Donald P. Shankweiler and Leonard Katz of Haskins Laboratories, I examined 378 children from seven to nine years old on a battery of tests that assessed both linguistic and nonlinguistic abilities. Our results as well as those of Keith E. Stanovich and Linda S. Siegel of the Ontario Institute for Studies in Education made it clear that phonological deficits are the most significant and consistent cognitive marker of dyslexic children.
One test in particular seemed quite sensitive to dyslexia: the Auditory Analysis Test, which asks a child to segment words into their underlying phonological units and then to delete specific phonemes from the words. For example, the child must say the word “block” without the “buh” sound or say the word “sour” without the “s” sound. This measure was most related to a child’s ability to decode single words in standardized tests and was independent of his or her intelligence, vocabulary and reasoning skills. When we gave this and other tests of phonemic awareness to a group of 15-year-olds in our Connecticut Longitudinal Study, the results were the same: even in high school students, phonological awareness was the best predictor of reading ability.
If dyslexia is the result of an insufficiently developed phonological specialization, other consequences of impaired phonological functioning should also be apparent—and they are. Ten years ago the work of Robert B. Katz of Haskins Laboratories documented the problems poor readers have in naming objects shown in pictures. Katz showed that when dyslexics misname objects, the incorrect responses tend to share phonological characteristics with the correct response. Furthermore, the misnaming is not the result of a lack of knowledge. For example, a girl shown a picture of a volcano calls it a tornado. When given the opportunity to elaborate, she demonstrates that she knows what the pictured object is—she can describe the attributes and activities of a volcano in great detail and point to other pictures related to volcanoes. She simply cannot summon the word “volcano.”
This finding converges with other evidence in suggesting that whereas the phonological component of the language system is impaired in dyslexia, the higher-level components remain intact. Linguistic processes involved in word meaning, grammar and discourse—what, collectively, underlies comprehension—seem to be fully operational, but their activity is blocked by the deficit in the lower-order function of phonological processing. In one of our studies, Jennifer, a very bright young woman with a reading disability, told us all about the word “apocalypse.” She knew its meaning, its connotations and its correct usage; she could not, however, recognize the word on a printed page. Because she could not decode and identify the written word, she could not access her fund of knowledge about its meaning when she came across it in reading.
Of course, many dyslexics, like Gregory, do learn to read and even to excel in academics despite their disability. These so-called compensated dyslexics perform as well as nondyslexics on tests of word accuracy— they have learned how to decode or identify words, thereby gaining entry to the higher levels of the language system. But they do so at a cost. Timed tests reveal that decoding remains very laborious for compensated dyslexics; they are neither automatic nor fluent in their ability to identify words. Many dyslexics have told us how tiring reading is for them, reflecting the enormous resources and energy they must expend on the task. In fact, extreme slowness in making phonologically based decisions is typical of the group of compensated dyslexics we have assembled as part of a new approach to understanding dyslexia: our neuroimaging program.