Misplaced stresses make the words hard to swallow

Many of us think there is no such thing as a bad citation: better for people to pay critical attention to your work than to be ignored. However, many of the top figures in the field of reading research are likely to wince when they read accounts of their work in these two books. Diane McGuinness has extreme and intemperate views on a number of issues, and much of what she writes appears to display a contempt for other people's research (her own small amount of empirical research, even unpublished studies, is generally treated with more clemency). My own research gets off relatively lightly here (better not to have another bad citation in this case, I feel).

These two long books are published simultaneously and cover complementary issues, with the introduction to each referring the reader to the other.

The first book, Language Development and Learning to Read (hereafter Learning to Read ), deals mainly with the prediction of variations in reading skill in children; do some children, because of their cognitive make-up, learn to read more easily than others? The second book, Early Reading Instruction , focuses on studies of how best to teach children to read. Clearly one would hope that knowing how children learn to read should have clear implications for how best to teach reading. McGuinness notes in the identical preface to each book: "Trying to squeeze all this material into one volume while adjudicating between reliable and unreliable studies for the reader proved impossible. The result was two complementary but independent books." One feels that with some editing of McGuinness's prolix prose, one book could have done the job nicely.

Learning to Read surveys a lot of material. The argument of the book is sometimes a bit difficult to discern, but it goes something like this. Most researchers believe that the development of a child's early speech sound skills (phonological skills) is critical for learning to read. One particularly important skill in this view is to have a phonological system that represents, and gives access to, phonemes (the individual speech sound units in words). This is because the child needs to understand the alphabetic principle (that the letters in printed words map directly on to the sounds - phonemes - of spoken words). One problem for children learning to read English is that these mappings between letters in printed words and the sounds in spoken words are quite irregular (why do we not we spell "yacht" as "yot"?). McGuinness refers to the view that I have just summarised very briefly as "the Dogma" and proceeds to try to demolish it. As a dogmatic believer in "the Dogma", this message is somewhat hard to swallow. McGuinness's basic argument is that phonological processing skills do not really develop at all, but instead are, in essence, present from birth. This startlingly odd conclusion is stated most clearly in Early Reading Instruction : "Awareness of phoneme contrasts is present at birth (ba-da) and phoneme analysis becomes operational by about six months of age." But the infant speech perception studies to which she is referring here assess only very basic speech sound discrimination abilities with restricted sets of isolated syllables. There is no doubt that infants can perform such tasks to some degree. But this finding is not relevant to the claim (the Dogma) that much older children, in the early stages of learning to read, still find it difficult to isolate and manipulate phonemes in spoken words and that such difficulties can be an obstacle to learning.

The style of both books is rather odd. Some passages would seem quite at home in the introductory chapter of a PhD thesis. For example, more or less the whole of chapter five in Learning to Read is devoted to the presentation of the results of a single research paper on a sample of 43 children. Here, as elsewhere, material is presented in minute detail, with tables (almost never a graph) of data reproduced from the published study. These laboured presentations of individual studies sit uneasily with other passages that seem to be the product of some sort of popular science-writing course with literary pretensions: "As the phonological theory turned into dogma and gobbled up more and more territory, a myriad of [sic] mini-deductive theories sprouted up under the phonological umbrella, like mushrooms under a tree." Another very odd aspect of the structure of the book is the way in which a dichotomy is created between reading research (which is done badly, by people who are not clever) and language research (which is done very well, by people who are clever). "Unlike reading research, mainstream research on language development has continued apace with no setbacks and is one of the great success stories of the behavioural sciences."

One striking example of the absurdity of this position is in relation to the presentation of the work by the Haskins group. Researchers at the Haskins Laboratories in New Haven, Connecticut, (in particular Isabelle and Al Liberman and their colleagues including Don Shankweiler, Anne Fowler and Susan Brady) were in the vanguard of work to understand the role of phonological skills in learning to read. This group's work is dismissed as, in essence, worthless and misguided by McGuinness, but these same researchers were (and some still are) at the forefront of work to uncover the basis of human speech and language mechanisms.

One surprising aspect of Learning to Read , given its length, is that the book is neither comprehensive nor balanced in terms of the research that is reviewed. Learning to read involves both the development of decoding (the ability to recognise individual words, as measured by the ability to read them aloud) and comprehension (the ability to understand the meanings conveyed by texts). The book only really deals with the development of decoding, and nothing of substance is said about reading comprehension. Nor is a clear demarcation between decoding and comprehension skills drawn. This is particularly odd given the book's title, since language skills are certainly critical to reading comprehension. Even for decoding, the coverage of studies is patchy. Many important studies are not cited, McGuinness seems to view the term "dyslexia" as a dirty word, and none of the work dealing with the neurobiological bases of reading is even mentioned.

A little knowledge can be a dangerous thing. A prominent and peculiar feature of both books is their frequent digression into issues about statistics and research design. For most readers, I imagine these passages will be baffling, but perhaps persuasive, since they are expressed with great bravura: "It is not an exaggeration to say that much of the basic research on reading is invalid, because it relies on the isolated-groups design. By the strict standards of science, this book would be much shorter than it is... (ie, if such flawed studies were omitted)." What are these "isolated-group" studies to which McGuinness takes such umbrage? This is a term that seems to be of McGuinness's own making and what she means by it is very simple: it is a study where we compare a group of good readers with a group of poor (perhaps dyslexic) readers. Her objections to these studies are based on a profound misunderstanding of basic statistics. Such designs (case-control studies) are widely used in medicine and epidemiology, and there is a whole raft of thorough statistical work on their design and interpretation. Based on McGuinness's view in chapter nine, we could throw out all studies that have ever compared a clinical group with a control group. By her criteria, studies that have compared, for example, depressed with non-depressed people, or people who have had a heart attack with people who have not, can yield no useful information about these illnesses. You do not need to be a statistician to know that this is a silly claim. There are a number of other equally fallacious examples of the use of statistical arguments in both books, but space is too short to deal with them (chapter four and appendix two of Early Reading Instruction contain an outstanding example of confusion about units of analysis and how to analyse multi-level data). It is one thing to misunderstand statistics, it is quite another to inflict such misunderstandings on others with such force.

At the heart of Early Reading Instruction are the chapters dealing with the effectiveness of phonically based teaching methods (chapters four, five and six). These chapters amount to a reworking of some of the material contained in the US National Reading Panel report. McGuinness draws strong conclusions that phonically based teaching is more effective than whole-language and "basal" reading programmes (reading schemes in UK parlance). Though the evidence may not be quite as clear as it is made out to be, it is hard to disagree with her main conclusion. Once again it is notable that her own research gets a much more positive evaluation - "the most successful remedial reading programmes today" - than other people's.

Perhaps more contentiously, McGuinness concludes that there is no evidence that teaching phoneme skills is useful in helping children learn to read. The truth is that such training is probably only important for children with reading difficulties. In evaluating the evidence on the effects of reading instruction, McGuinness consistently fails to pay any attention to individual differences (in line with her theoretical predilections set out in Learning to Read ). While a good phonically based reading programme may be all that most children need to learn to decode, this does not mean that it will eliminate all reading (decoding) problems. Children who struggle to learn to decode are likely to need more than whole-class phonically based reading instruction.

Overall these two books present a wealth of information about how children learn to read and how they can best be taught. The books sit uneasily in style and depth between the academic and the popular. It is a pity that McGuinness did not take the opportunity in Learning to Read to present a shorter and more dispassionate account of the current state of research. While Early Reading Instruction seems the better of the two, both books too frequently turn into polemics rather than reasoned assessments of what we know and what we should be trying to understand next.

Charles Hulme is professor of psychology, Centre for Reading and Language, University of York.