|Dick and Jane|
Rudolf Flesch, Why Johnny Still Can't Read (1981)
 Mitford Mathews, the wise old author of Teaching to Read, Historically Considered, wrote:
Those who have maintained the superiority of the word method as a way of teaching a child quickly have often succeeded in promoting themselves professionally, but they have not always enhanced their reputations as scholarly persons in the estimation of scholars in other disciplines. The fact is that the method, prior to its adoption, had never been scientifically tested in competition with any other. This assertion will be challenged by many devoted practitioners. All anyone has to do to refute it, however, is to give the details of the experiment: who conducted it, for how long, involving how many children, how many teachers, and so forth.
Mathews didn't mean, of course, that no such studies have been made. On the contrary, since the first was done in 1911 there have been 124 such studies, carefully comparing the results of phonics-first and look-and-say. How many of them proved the superiority of look-and-say? Not one—not a single, blessed one.
This sounds unbelievable, I know. In fact, it is unbelievable, considering the near-monopolistic rule of look-and-say for the past fifty or sixty years of American education. Nevertheless it's true. I can prove it by citing chapter and verse. If any educator wants to cite a single contrary research finding, he or she is welcome to do so.
Here are the facts:
When I wrote my book Why Johnny Can't Read in 1955, I  listed eleven studies that had been done up to that time. All of them gave results in favor of phonics-first; not a single one favored look-and-say. The scientific proof was complete and overwhelming.
Ten years later, in April 1965, Dr. Louise Gurren of New York University and Mrs. Ann Hughes, research director of the Reading Reform Foundation, published an article in the Journal of Educational Research called "Intensive Phonics vs. Gradual Phonics in Beginning Reading: A Review." It listed thirty-six studies, including most of those I had listed ten years before. Gurren and Hughes concluded:
1. Rigorous controlled research clearly favors teaching of all the main sound-symbol relationships, both vowel and consonant, from the start of formal reading instruction.
2. Such teaching benefits comprehension as well as vocabulary and spelling.
3. Phonetic groups are usually superior in grades 3 and above.
Two years later, in 1967, Professor Jeanne Chall of the Harvard University Graduate School of Education published her book Learning to Read: The Great Debate. It was based on five years of research, funded by a large Carnegie Foundation grant.
Professor Chall reviewed eighty-five studies, including most of those surveyed earlier by Gurren and Hughes and me. She not only surveyed classroom studies, but also laboratory and clinical studies of all kinds—anything that compared phonics-first (or "systematic phonics," "intensive phonics" or, in her terms, "code emphasis") with look-and-say (or "incidental phonics," "gradual phonics," "delayed phonics," or, in her term, "meaning emphasis"). Her conclusions are stated on page 307:
My review of the research from the laboratory, the classroom and the clinic points to the need for a correction in beginning reading instructional methods. Most school children in the United States are taught to read by what I have termed a meaning-emphasis method. Yet, the research from 1912 to 1965 indicates that a code-emphasis  method—i.e. one that views beginning reading as essentially different from mature reading and emphasizes learning of the printed code for the spoken language—produces better results, at least up to the point where sufficient evidence seems to be available, the end of third grade.
The results are better, not only in terms of the mechanical aspects of literacy alone, as was once supposed, but also in terms of the ultimate goal of reading instruction—comprehension and possibly even speed of reading. The long-existing fear that an initial code-emphasis produces readers who do not read for meaning or with enjoyment is unfounded. On the contrary, the evidence indicates that better results in terms of reading for meaning are achieved with the programs that emphasize code at the start than with the programs that stress meaning at the beginning.
The fourth review of the evidence was done in 1973 by Dr. Robert Dykstra, professor of education at the University of Minnesota. It appeared as a fifty-page special section in the book Teaching Reading by Walcutt, Lamport, and McCracken. Dr. Dykstra reviewed fifty-nine studies, again partly overlapping the reviews by me, Gurren and Hughes, and Chall. He summarized the evidence on page 397:
Reviewing the research comparing (1) phonic and look-say instruction programs, (2) intrinsic and systematic approaches to helping children learn the code, and (3) code-emphasis and meaning-emphasis basal programs leads to the conclusion that children get off to a faster start in reading if they are given early direct systematic instruction in the alphabetic code. The evidence clearly demonstrates that children who receive early intensive instruction in phonics develop superior word recognition skills in the early stages of reading and tend to maintain their superiority at least through the third grade. . . .
We can summarize the results of 60 years of research dealing with beginning reading instruction by stating that early systematic instruction in phonics provides the child with the skills necessary to become an independent reader  at an earlier age than is likely if phonics instruction is delayed and less systematic. As a consequence of his early success in "learning to read" the child can more quickly go about the job of "reading to learn."
The total count of the studies surveyed in the four overlapping research reviews is 116. There were 7 later studies that showed the superiority of the Distar phonic system over others in the federally funded Follow Through project, plus one important phonics-vs.-look-and-say study by Dr. Douglas Carnine, published in 1977, to which I’ll return. This brings the total count to 124. As I said, not one of those 124 studies showed results favoring look-and-say.
Most of the 124 studies followed a simple pattern. The researcher compared two groups of children. One was trained by look-and-say, the other by phonics-first. At the end of the year, both groups were tested to find out which had progressed further in word recognition and comprehension.
To take just one example, in February 1958 the Journal of Educational Research published a study by Barbara C. Kelly, entitled "The Economy Method Versus the Scott, Foresman Method in Teaching Second-Grade Reading in the Murphysboro Public Schools." Miss Kelly had studied the reading achievement of 100 second graders in Murphysboro, Illinois, who'd been trained by the Economy Readers (one of my "Phonic Five" listed on page 9) and another group of 100 second graders trained by Scott, Foresman (one of the "Dismal Dozen"). She found a significant difference in favor of the phonics-trained children.
In this way, the 124 studies compared most of the available phonics-first materials with virtually all the Dismal Dozen look-and-say methods. The results invariably favored phonics.
In the early years, many of the experiments were done with the phonics readers put out by the Economy Company. After 1963, when the Lippincott Readers were first published, much experimentation was done with Lippincott material. For instance, there were a number of studies within the framework of 27 cooperative studies sponsored by the U.S. Office of Education in the mid-sixties. Lippincott won all the "races" in which it was entered.
 The next major group of competitive studies was the massive Follow Through project, again sponsored by the federal government. It was designed to compare the results of various educational methods with disadvantaged children who had gone through preschool preparation with Head Start. The only phonics-first program entered in that "race" was Distar (also listed among my "Phonic Five" on page 9). Distar came out ahead in all the "races" it was entered in.
At the end of third grade all children made gains in IQ from 6 to 8 points, and low-IQ children, with an average IQ of 73, gained between 8 and 14 points. The participating schools were in Uvalde, Texas (90 percent Chicano); Tupelo, Mississippi; Flint, Michigan; Dayton, Ohio; and East St. Louis, Illinois (all four mostly black); and Flippin, Arkansas, and Smithville, Tennessee (both mostly white). All children in the program came from low-income families.
Beginning in the sixties, educational researchers also did a series of laboratory experiments to test the effectiveness of look-and-say. Their first order of business was the refutation of the famous Cattell study of 1885.
Cattell's study had been used over and over again by look-and-say educators to lend their gimmick some shred of respectability. Cattell had proved, they said, that we read whole words rather than letters and that therefore look-and-say is better than phonics.
Let's go back for a moment and see exactly what Dr. James McKeen Cattell did in 1885. He worked in a laboratory in Leipzig (most psychological laboratory work in those days was done in Germany), and tested readers with a tachistoscope, an instrument that exposes reading matter for a split second.
He found, to his surprise, that readers could read whole words in less time than it took them to read single letters. "These results are important enough," he wrote, "to prove those to be wrong who with Kant hold that psychology can never become an exact science."
Important, but wrong. The obvious flaw in the experiment was that Cattell tested only adults. (The group he used included one nine-year-old boy, but he "was superior in reading ability to some of the adults.")
Until 1965 nobody bothered to check his assumption that  beginning readers read the way fluent adult readers do. Then Drs. Gabrielle Marchbanks and Harry Levin of Cornell University decided to look into the matter. They took 50 kindergartners and 50 first graders and asked them to pick from a group of nonsense words the one most like another nonsense word they'd been shown. For instance, they were shown first the nonsense word cug, and then the four nonsense words arp (same shape), che (same first letter), tuk (same second letter), and ilg (same third letter.)
The results were clear. Most of the children went by the first letter, a smaller number by the last letter, and only a few by the general shape of the whole word.
Five years later, in 1970, Dr. Henry G. Timko repeated the Marchbanks-Levin experiment and came to the same conclusion—first graders recognize words by the first letter and not by the shape of the whole word.
In the same year, 1970, Drs. Joanna P. Williams, Ellen L. Blumberg, and David V. Williams repeated the experiment once more. They tested kindergartners, first-graders and adults. Their kindergartners used whole-word shape not at all, the first graders did it very rarely, and the adults only sometimes. After eighty-five years, the ghost of James Cattell had at long last been laid to rest.
"In view of these and other findings," Dr. Williams wrote, "there seems to be no justification for developing instructional methods or materials based on the use of configuration as the primary cue. . . . It is worth noting that the most widely used reading method over the past 30 years (the look-say' or `whole word' method) has stressed identification of words on the basis of configuration."
At the same time that they disposed of Cattell's century-old error, researchers also established why phonics-first worked and look-and-say did not. What it all boils down to, they said, is the problem of "transfer of training." Which was more effective in learning to read unfamiliar printed words—memorizing whole words or learning letters and sounds?
The first experiment to solve this problem was done in 1964 by Miss Carol H. Bishop, then a graduate student at Cornell University working under Dr. Eleanor Gibson. Miss Bishop picked 60 college freshman and sophomores and taught 20  of them the sounds of twelve Arabic letters and another 20 the sounds of eight words composed of those letters. All students were then asked to try to read eight different Arabic words, containing the same eight letters.
Miss Bishop's results showed clearly that letter training was better than word training. What's more, when Miss Bishop analyzed her findings, she found that only those word-trained students who had figured out the sound values of the letters were able to learn the unfamiliar words. Which means that they knew about the alphabetic code in English and applied that knowledge to the job of learning Arabic words.
After three years, in 1967, a similar experiment was conducted by Drs. W. E. Jeffrey and S. Jay Samuels, then at the University of California in Los Angeles. Instead of adults, Jeffrey and Samuels took 60 kindergartners from a public school, and instead of Arabic characters they used arbitrary, odd-shaped symbols. The children in the word group were taught four nonsense words—mo, so, ba, be, and those in the letter group were taught the sounds of s, m, a, and e. The words to be learned were se, sa, me and ma.
Jeffrey and Samuels's results were the same as Miss Bishop's. The letter-trained group was clearly superior to the word-trained group.
Ten years later, in 1977, Dr. Douglas W. Carnine of the University of Oregon conducted another similar experiment. He tried to make the situation as much as possible like a classroom situation. So he used regular English letters, gave the children more words and letters to learn, and added some irregular words to those to be learned. Carnine's results confirmed those of Bishop and Jeffrey and Samuels. Phonics-first won hands-down over look-and-say.
Both the Jeffrey and Samuels experiment and Carnine's recapitulation had an important added feature—the letter-trained children were taught not only what sounds the letters stood for but also how to "blend" those sounds—saying mmm and ee and then blending them together to say me.
This special training in blending was based on the idea that learning tasks must be analyzed to find their component suskills. Then these subskills must be taught in a strict sequence—a hierarchy. A higher subskill in such a sequence is taught  on1y when the lower subskill has been fully mastered. Only this way can a complex skill be taught so that it can be performed without mistakes.
The first psychologist to write about "task analysis," "sub-skills," and "learning hierarchies" was Dr. Robert M. Gagné. Dr. Gagné is not a reading expert but a psychologist specializing in the techniques of teaching and learning.
I wrote to Dr. Gagné and asked him how he first developed his idea. He answered:
My ideas about task analysis, learning hierarchies and subordinate skills came originally from a study I did on the learning of ninth graders in a mathematics problem (inferring and stating a general formula for the sum of terms in a number series). When I ran across some students who seemed to be having particular difficulty learning to perform this task, it seemed to me they were missing some "subordinate skills," in some cases rather simple arithmetic skills. Accordingly, I did a study in which I first analyzed the subordinate skills of the number-series task, then tested students on them, and taught them the subordinate skills they didn't know. As a research psychologist used to the notion that learning is a gradual process, I was surprised by the results. Once subordinate skills were mastered, the new learning was very rapid, and "sudden."
In 1965 Dr. Gagné developed the first subskills hierarchy for the teaching of reading. It appears in the first edition of his widely used textbook The Conditions of Learning. Pointedly ignoring the commonly used look-and-say method, Dr. Gagné proposed a phonics-first method. Children were to be taught first the language sounds, then "recognition of printed letters by sound," then "recognition of printed words," then the "distinguishing of similar words," and so on.
The objective at this stage of learning [Gagné wrote] is to enable the child to say orally such a word as concatenation, which he has never seen or heard before. Whether he knows the meaning of the word is quite irrelevant. The all-important capability he is acquiring at this stage  is the ability to say, when he sees it in print, the word concatenation in a way that is discriminably different to him from the printed word concentration. . . .
In an adult, the difficulties experienced with a novel word provide a fairly sure clue to the existence of a gap in his early instruction. . . . On the first encounter, he may be unable to read a word like obsequiousness, which, to an individual who has learned to identify sounds, is ridiculously simple.
The difference between phonics-first and look-and-say has hardly ever been stated more clearly. Look-and-say-trained children can't read because a vital step in their instruction—distinguishing between different spelling patterns—has been left out. Dr. Gagné, with his background in math and science, looked at the situation with unconfused eyes and pointed straight at the source of the trouble.
Gagné's subskill approach soon swept the field of reading research. It became the accepted doctrine among almost all reading researchers that decoding—that is, phonics-first—must be taught by a carefully worked-out sequence of subskills.
In a fascinating paper published in 1976, Dr. S. Jay Samuels, now at the University of Minnesota, compared the method with the teaching of a number of other complex skills.
Replying to Professor Frank Smith, who had written that one learns to read by reading just as one learns to ride a bicycle by riding a bicycle, Dr. Samuels said this isn't necessarily the way one learns to ride a bicycle either. "Children often go through a graded series of experiences of increasing difficulty before they learn to ride a large-frame, two-wheel bike. They frequently practice first on a tricycle, then graduate to a two-wheeler with a small frame, and practice getting their balance on the small frame bike before they use the pedals on their two-wheeler."
Or, Samuels continued, take the current method of teaching downhill skiing. "Perhaps the most significant recent advance has been with the GLM, graduated length method. The beginning skier uses short skis to practice his basic moves and then advances to longer skis as skill develops."
Or take training in wrestling. "Every move in wrestling is  broken down into its parts and the athlete practices these parts prior to putting them together to form a move that has fluid motion. When a move is finally mastered, combinations of moves are worked together to form larger units or patterns of moves."
Finally, Samuels mentioned the learning of dance steps.
The trick in learning a new dance step without the aid of a teacher is to try to identify the basic move from which the variations originate. What the teacher does to simplify learning a dance is to select the basic step and to teach the subskills that comprise the basic step. Years ago the Arthur Murray system used this procedure to introduce people to social dancing. Their basic step was called the box step and was used to introduce a number of dances as well as their variations.
This is the system that Samuels and most other reading re-searchers now apply to the teaching of beginning reading. Samuels proposed a sequence of subskills going from learning distinctive letter features—like the differences between b and d or p and q—to letters, to letter clusters and on to words. Dr. Joanna Williams of Teachers College, Columbia University, and Drs. Michael and Lise Wallach of Duke University have added two steps before learning letter shapes, namely, learning to analyze in general, and learning to analyze spoken words into their component sounds. Higher up in the scale, children are taught to blend letters and sounds and to "chunk" such units as str or ight. Many scholars are now interested in carrying this approach higher and higher into more complex reading skills, such as comprehension and inference.
As you see, phonics-first, or decoding, is now the center of attention and the area that is of foremost interest to scholars. The Goodman and Smith type of guessing game is taking a backseat.
In the fall of 1979 there appeared a three-volume set of books called Theory and Practice of Early Reading. It contained scholarly papers by fifty-nine contributors, assembled and edited by Dr. Lauren Resnick of the University of Pittsburgh and Dr. Phyllis Weaver of Harvard University. There are papers by Drs. Kenneth and Yetta Goodman and Dr. Frank Smith  and three of their followers. The fifty-three other contributors—about 90 percent—write mainly about decoding, or phonics-first.
Theory and Practice of Early Reading is a galaxy of today's academic stars in the field of reading research. It shows vividly where we are now. "In a sense," writes Dr. Resnick in her introduction, "these volumes are a reexamination of issues considered by Chall a decade ago" (that is, the "great debate" between phonics-first and look-and-say). "Somewhat modified basal reading approaches are still dominant in today's instructional practice, and they are even less favorably viewed by our contributors than they were by Chall."
"As a matter of routine practice," Dr. Resnick writes later in a summary article, "we need to include systematic, code-oriented instruction in the primary grades, no matter what else is also done."
Two outspoken contributors, Drs. Isabelle Y. Liberman and Donald Shankweiler of the University of Connecticut, write:
Instructional procedures should inform children early on that the printed word is a model of the component phonemes [sounds] and their particular succession in the spoken word. . . . The instruction should not, as it often does, mislead children into assuming that the printed word is an ideographic symbol, a notion that will have to be corrected later and, apparently for some children, with great difficulty. Procedures that initiate children into the mystique of reading by drawing their attention to the visual configuration ("remember this shape; it has a 'tail' ") and its associated meaning ("the one with the tail means monkey") without alerting them to the relevance of the sound structure of the word may lead them into a blind alley. Their ability to memorize the shapes and associated meanings of a handful of words may lull them and their parents into the comfortable belief that they can read, but it may leave them stranded at that stage, functional illiterates with no keys to unlock new words.
Dr. S. Jay Samuels of the University of Minnesota writes:
I had an opportunity to discuss what was needed to improve the teaching of reading with Tom Sticht, John  Guthrie, Harry Singer and Dennis Fisher. There was consensus that at the present time a sufficient amount is known about practical aspects of reading so that all children, even those at risk [the lowest 15 percent on the IQ scale] can be taught to read. The problem, then, is not a lack of knowledge about how to teach reading. The problem is . . . not very different from the problems of changing the smoking habits of the American public. Presently we know that smoking is dangerous to health. Despite this knowledge, many people are unable to change their smoking habits, and many others take up this harmful habit. The task of changing smoking habits is probably more formidable than changing the reading practices of school systems. . . . With the recognition by taxpayers and educators alike that the school must be accountable in some sense for the products of its system, we may well be moving toward an era in which schools will be doing a more efficient job of teaching reading.
And Dr. Barbara Bateman of the University of Oregon says flatly:
Near failure-proof methods for teaching all children to read are already available. Continued failure of schools to employ these [phonics-first] programs is at best negligent and at worst malicious.
28 Mathews, Teaching to Read (see Chapter 2), p. 148.
29 Gurren, Louise, and Hughes, Ann. "Intensive Phonies vs. Gradual Phonics in Beginning Reading: A Review." Journal of Educational Research, vol. 58, no. 8, April 1965, pp. 339-346.
29 Chall, Jeanne S. Learning to Read: The Great Debate. New York, McGraw-Hill, 1967.
30 Walcutt, Charles C.; Lamport, J.; and McCracken, G. Teaching Reading. New York, Macmillan, 1974. (Chapters on research and evaluation by Robert Dykstra.)
31 Kelly, Barbara C. "The Economy Method versus the Scott Foresman Method in Teaching Second-Grade Reading in the Murphysboro Public Schools." Journal of Educational Research, vol. 50, February 1958, pp. 465468.
32 Becker, Wesley C., and Engelmann, Siegfried. Analysis of Achievement Data on Six Cohorts of Low-Income Children from 20 School Districts in the University of Oregon Direct Instruction Follow Through Model (Technical Report 78-1). University of Oregon College of Education Follow Through Project, Eugene, Ore., December 1978.
32 Cattell, James McKeen. "The Time It Takes to See and Name Objects." Mind, vol. 11, 1886, pp. 63-65.
33 Marchbanks, Gabrielle, and Levin, Harry. "Cues by Which Children Recognize Words." Journal of Educational Psychology, 1965, vol. 56, no. 2, pp. 57-61.
33 Timko, Henry G. "Configuration as a Cue in the Word Recognition of Beginning Readers." Journal of Experimental Education, vol. 39, 1970, pp. 68-69.
33 Williams, Joanna P.; Blumberg, E. L.; and Williams, D. V. "Cues Used in Visual Word Recognition." Journal of Educational Psychology, vol. 61, 1970, pp. 310315.
33 Bishop, Carol H. "Transfer Effects of Word and Letter Training in Reading." Journal of Verbal Learning & Verbal Behavior, vol. 3, 1964, pp. 215-221.
34 Jeffrey, W. E., and Samuels, S. J. "Effect of Method of Reading Training on Initial Learning and Transfer." Journal of Verbal Learning & Verbal Behavior, vol. 6, 1967, pp. 354-358.
34 Carnine, Douglas W. "Phonies vs. Look-Say: Transfer to New Words." Reading Teacher, March 1977, pp. 636640.
35 Gagné, Robert M. "The Acquisition of Knowledge." Psycho-logical Review, 1962, vol. 69, no. 4, pp. 355-365.
35 -. The Conditions of Learning. New York, Holt, Rinehart & Winston, 1965.
36 Samuels, S. Jay. "Hierarchical Subskills in the Reading Acquisition Process." In Guthrie, J., ed. Aspects of Reading Acquisition. Baltimore, Johns Hopkins University Press, 1976, pp. 162-179.
37 Resnick, Lauren B., and Weaver, P. A., eds. Theory and Practice of Early Reading. Hillsdale, NJ., Lawrence Erlbaum Associates, 1979. 3 vols.
37 Williams, Joanna. "The ABD's of Reading: A Program for the Learning Disabled." In Resnick and Weaver, eds., Theory and Practice. (see above), vol. 3.
37 Weaver, Phyllis A., and Resnick, L. B. "The Theory and Practice of Early Reading: An Introduction." In Resnick and Weaver, eds. Theory and Practice . . . (see above), vol. 1.
38 Resnick, Lauren B. "Theories and Prescriptions for Early Reading Instruction." In Resnick & Weaver, eds. Theory and Practice . . . (see above) vol. 2, p. 329.
38 Liberman, Isabelle Y., and Shankweiler, D. "Speech, the Alphabet, and Teaching to Read." In Resnick and Weaver, eds. Theory and Practice . . . (see above), vol. 2, pp. 121-122.
38 Samuels, S. Jay. "How the Mind Works When Reading: De-scribing Elephants No One Has Ever Seen." In Resnick and Weaver, eds. Theory and Practice . . . (see above), vol. 1, p. 348.
39 Bateman, Barbara. "Teaching Reading to Learning Disabled and Other Hard-to-Teach Children." In Resnick and Weaver, eds. Theory and Practice . . . (see above), vol. 1, p. 247.