Reading Comprehension Difficulties

Processes and Intervention

Cesare Cornoldi, Jane V. Oakhill

• Publisher: Routledge
• Year: 2013
• ISBN:
  9781136488627
• Edition: 1
• Topic: Education
• Subtopic: Education General

Recognizing the characteristics of children with learning disabilities and deciding how to help them is a problem faced by schools all over the world. Although some disorders are fairly easily recognizable (e.g., mental retardation) or very specific to single components of performance and quite rare (e.g., developmental dyscalculia), schools must consider much larger populations of children with learning difficulties who cannot always be readily classified. These children present high-level learning difficulties that affect their performance on a variety of school tasks, but the underlying problem is often their difficulty in understanding written text. In many instances, despite good intellectual abilities and a superficial ability to cope with written texts and to use language appropriately, some children do not seem to grasp the most important elements, or cannot find the pieces of information they are looking for. Sometimes these difficulties are not immediately detected by the teacher in the early school years. They may be hidden because the most obvious early indicators of reading progress in the teacher's eyes do not involve comprehension of written texts or because the first texts a child encounters are quite simple and reflect only the difficulty level of the oral messages (sentences, short stories, etc.) with which the child is already familiar. However, as years go by and texts get more complex, comprehension difficulties will become increasingly apparent and increasingly detrimental to effective school learning. In turn, studying, assimilating new information, and many other situations requiring text comprehension -- from problem solving to reasoning with linguistic contents -- could be affected. Problems with decoding, dyslexia, and language disorders have attracted more interest from researchers than have specific comprehension problems and have occupied more room in specialized journals. Normal reading comprehension has also been a favorite with researchers. However, scarce interest has been paid to subjects who have comprehension difficulties. This book is an attempt to remedy this situation. In so doing, this volume answers the following questions:
* Does a reading comprehension problem exist in schools?
* How important and widespread is the problem?
* Is the problem specific?
* How can a reading comprehension difficulty be defined and identified?
* Does the "syndrome" have a single pattern or can different subtypes be identified?
* What are the main characteristics associated with a reading comprehension difficulty?
* When can other well-identified problems add to our understanding of reading comprehension difficulties?
* Which educational strategies are effective in preventing and treating reading comprehension difficulties?
* What supplementary information can we get from an international perspective?

CHAPTER ONE

Some Observations on a Simple View of Reading

Philip B. Gough

University of Texas at Austin

Wesley A. Hoover

Southwest Educational Development Laboratory

Cynthia L. Peterson

Southwest Texas State University

Only a fool would deny that reading is complex. Reading clearly involves many subprocesses, and those subprocesses must be skillfully coordinated.

There is first of all the control of eye movements. Reading begins with the fixation of the printed page. A quarter of a second later, a saccade to a new location takes place, and then the sequence is repeated, interrupted only by return sweeps and an occasional regression. Within each fixation, the reader must decide how long to maintain it and where to fixate next; because eye movements are ballistic, she must also program the eyes’ musculature to carry out these decisions.

Then there is word recognition. On fixating a printed word, the reader must translate that meaningless set of letters into a recognizable object, locating or activating precisely the right word in a mental lexicon containing tens of thousands of items; she accomplishes this feat in less than one fifth of a second. At the same time, the reader is developing information about the form of the next word, and the location of the word after that.

Mere recognition of the word is not enough; the reader must also decide what the word means. Many words are ambiguous, and the reader must choose among those meanings; the reader must disambiguate the word.

Having selected the appropriate meaning for each word, the reader must then fit them together. She must determine the syntactic function of each word, and then determine the relations among them; she must decide which noun or noun phrase is the subject of each verb, which noun each adjective modifies, and where each prepositional phrase is to be attached. In short, she must parse the sentence.

Once these relations are established, the reader must use this information to determine the meaning of the sentence. Often this will require inference: The reader will need to draw upon her knowledge of the world to construct a representation of the sentence’s meaning.

Given a representation of the sentence’s meaning, it must be related to meanings of previous sentences; it must be fitted into the reader’s mental structure of the discourse.

Finally, the reader must decide what to do with this information. If she decides that it is true and valuable, she must incorporate it into her body of knowledge; she must learn.

All of these things must be accomplished by the reader. But we note that many of them must also be done by the auder, the person who is listening to language, not reading it. Beyond the point of word recognition, listening and reading appear to require essentially the same processes.

To be sure, the common processes are not exactly the same. Disambiguation is occasionally different, for the ambiguities of the spoken word and those of the printed word are not identical. Many spoken words (e.g., /sel/) are ambiguous; they must be disambiguated. But often their printed counterparts (e.g., sail and sale) are not ambiguous, at least not in the same way. In contrast, many printed words (e.g., bow) are polyphonic; they correspond to two different phonological forms (e.g., /bo/, /bau/) and those forms may or may not require disambiguation.

The process of parsing is also slightly different, for the two modalities convey information about syntax in different ways. Speech offers intonation and stress; print offers punctuation and capitalization.

The availability of the materials for integration differs as well. In print, the previous sentences remain on the page, and the reader can return to them at will. But speech is ephemeral, and the external evidence of previous sentences is lost; the auder must rely on memory.

But these differences pale in comparison to the similarities of processing in the two modalities. Virtually the same lexicon is used to read and aud. The vast majority of ambiguities are ambiguous in both modalities. In both modalities, the words are laid out linearly (printed words in space, spoken words in time), and word order plays a central role in parsing. Virtually the same grammar is employed in parsing both written and spoken sentences. The same background knowledge is brought to bear on printed and spoken words.

What this suggests, then, is that reading can be divided into two parts; that which is unique to reading, namely decoding, and that which is shared with auding, namely comprehension. The division is natural; we are cutting nature at its joint.

THE DISSOCIATION OF DECODING AND COMPREHENSION

Many students of reading would resist this division. They argue that reading is interactive, that decoding and comprehension are tightly interwoven.

The two are certainly correlated, and this correlation makes it difficult to separate the two. Skilled reading clearly requires skill in both decoding and comprehension. The most common sort of reading disability, which we call garden variety, involves deficits in both decoding and comprehension. In the general population, then, readers tend to be skilled in either both or neither. The result is a strong positive correlation between the two parts, and consequently, when researchers try to study differences in comprehension while matching on decoding, they are haunted by the problem of regression to the mean.

But decoding and comprehension can be separated, or at least dissociated. This is, we think, an important point, for neuropsychologists have offered such dissociation as evidence that the two skills are lodged in distinct mechanisms. In the case of decoding and comprehension, we think the dissociation is clear. To take one example, the typical 5-year-old Italian can understand Italian but not decode it; we, in contrast, can decode it (to a certain extent), but comprehend it very poorly.

Our favorite example is drawn from the life of John Milton. In his dotage, Milton wished to reread the Greek and Latin classics, but he was going blind. So he taught his daughters to decode Greek and Latin. They read the classics aloud while he listened to them. Between them, there was reading comprehension.

The dissociation of decoding and comprehension is also found in the dyslexic and the hyperlexic. The dyslexic can comprehend but not decode; the hyperlexic can decode but not comprehend. Between the two, there is double dissociation.

Decoding and comprehension, then, are the two halves of reading. But the two halves are not added together. Reading does not equal the sum of decoding and comprehension, for neither decoding in the absence of comprehension, nor comprehension in the absence of decoding, leads to any amount of reading. A child who cannot decode cannot read; a child who cannot comprehend cannot read either. Literacy—reading ability—can be found only in the presence of both decoding and comprehension. Both skills are necessary; neither is sufficient.

These observations led us to adopt a multiplicative hypothesis. If reading (r), decoding (d), and comprehension (c) are considered as skills that range from zero to one, then reading must be the product of decoding and comprehension. Put algebraically, r = d x c. Reading takes place only when both d and c are greater than zero.

There is, then, only one way to read; by decoding and comprehending what you have decoded: There is only one kind of reading ability. But as we have seen, three kinds of reading disability can be distinguished; a deficiency in decoding (the dyslexic), a deficiency in comprehension (the hyperlexic), and a deficiency in both processes (the garden variety).

THE COMBINATION OF DECODING AND COMPREHENSION

We hold, then, that reading ability consists of two distinct parts. Two major efforts to factor analyze the reading process have come to a conclusion that differs from ours.

Zwick (1987) analyzed the reading data collected in the 1983–1984 National Assessment of Educational Progress (NAEP). NAEP set up groups of seemingly dissimilar kinds of items, including items measuring general reading comprehension, items testing inference of word meanings from context, items having to do with everyday reading skills, and essay items. Both a full-information factor analysis and a test of unidimensionality developed by Rosenbaum (1985) supported the same conclusion: Despite their dissimilarities, the reading items used could be regarded as measures of a single dimension.

Rost (1989) gave 38 different reading tests, a spelling test, and a speed-of-information processing test to 220 German second graders, and intercorrelated them. His study yielded 780 correlation coefficients. A factor analysis of this matrix revealed only one broad general reading comprehension component, accounting for 61% of the total variance.

Carroll (1988) observed that multidimensionality is unlikely because as students progress, “different reading skills develop more or less in parallel” (p. 763). This argument seems to us to beg the question “Is reading unidimensional or multidimensional?” We think that a more convincing explanation of why Zwick and Rost failed to find two dimensions is because the tasks used in these studies all required decoding. As we see it, to separate the decoding and comprehension factors, we need tasks that measure each without involving the other.

What neither of these studies included were measures of listening comprehension, that is, measures of the comprehension component uncontaminated by decoding.

In 1990, we (Hoover & Gough) conducted a study that did obtain uncontaminated measures of each component skill. The Southwest Educational Development Laboratory (SEDL) measured separately the decoding, listening comprehension, and reading skills of 254 bilingual children in Texas each year for the first four grades. Decoding was assessed by asking the child to name pseudowords. Listening was measured by reading a story to the child, and then asking questions about the story. Reading comprehension was assessed in the same way, save that the child read the story.

It would surprise no one that we found both decoding ability and listening ability to correlate with reading comprehension. But we wanted to show that the multiplicative model accounted for the data better than the additive one. The problem we faced was that the two models make the same predictions almost everywhere: The best reader will be the child good at decoding and listening, the worst will be the child weak at both, and the intermediate reader will be the child intermediate at both skills. Where the two models differ is where the two skills are dissociated.

There was enough dissociation (i.e., enough children who were strong in one skill, but weak in the other) in the SEDL data to show that the product of decoding and listening comprehension correlated more highly with reading comprehension than did their sum. Those correlations were astonishingly high; .84 in the first grade, .85 in the second, and .91 in both the third and the fourth. Despite the fact that SEDL’s measures of the three variables were not perfectly reliable, these correlations could not be closer to 1. These results offer an accurate description of reading ability: Reading is best described as the product of decoding and comprehension.

The Relationship of the Components to Reading

The simple view describes the joint relationship of decoding and comprehension to reading; it also helps us describe the separate relationships between the two components and reading.

In the typical kindergartner, comprehension is well developed; the typical 5-year-old can speak and understand her native language better than the typical foreign college student with four semesters in that language. The phonology and syntax of the language has been all but mastered, the child’s vocabulary numbers in the thousands, and a great deal of knowledge about the world has been acquired. There is variability, to be sure, but we think that it is reasonable to assume that comprehension at school entry is normally distributed around a very substantial mean. The typical text that confronts the child at this age is very simple, with a difficulty level well beneath that mean. If the text were read to the child, it would be understood by almost every normal child. So, among beginning normal readers tested with typical materials, c approaches 1. In the multiplicative view, if c = 1, then r = d. Thus, we would expect to see that among beginning readers, the correlation between c and r would be negligible, whereas the correlation between d and r would be very high.

At the other end of the curriculum, the situation is just the reverse. The majority of college students can decode the vast majority of words they encounter; d approaches 1. But if d = 1, then r = c. So, although the correlation between c and r will be modest at best in the general population, among college students, we would expect it to be very high. Thus, the simple view leads us to expect that the correlation between decoding and reading will decrease across the grades, while the correlation between listening and reading will increase over the same time span.

Of course, in both cases above we presume that the materials used to assess c parallel those used to assess r. Although the language of print may indeed differ from that of speech (e.g., some syntactic devices used in the former rarely appear in the latter), this is not a necessary consequence of the medium, for anything that can be captured in language can be transmitted through either a written or oral form. Therefore, we must distinguish the message from the medium.

We looked for the trends predicted above in the relationships of reading to comprehension and decoding through a meta-analysis of existing research. We searched the published literature to identify investigations of normal reading in monolingual English speakers (first through sixth grade, and college level) where appropriate measures of the three variables were taken. By appropriate, we mean that (a) comprehension was measured free of decoding (i.e., using auding), and (b) auding and reading were measured in the same way. We found 10 such studies, reporting measures from 17 different samples.

Table 1.1 presents the correlations from these studies. The table displays the coefficients and tests of significance as well as the sample size; the data are grouped by grade level. In some cases, inverted scales were employed (e.g., latency was measured instead of accuracy), and for these scales, we reversed the signs of the coefficients so that a positive correlation indicates a direct relationship between two skills.

The simple vote-counting method applied to these data indicated that both relationships were positive: The 17 reported correlations between decoding and reading were all positive (15 of them significantly so), and 16 of the reported correlations between comprehension and reading were significantly positive as well (the sole exception being insignificant).

To obtain more precise ...

Table of contents

• Cover Page
• Half Title page
• Title Page
• Copyright Page
• Acknowledgments
• Contents
• Introduction: Reading Comprehension Difficulties
• Part I
• Part II
• Part III
• Part IV
• Part V
• Author Index
• Subject Index