Course Handout - Dyslexia and the Cognitive Science of Reading and Writing

Copyright Notice: This material was written and published in Wales by Derek J. Smith (Chartered Engineer). It forms part of a multifile e-learning resource, and subject only to acknowledging Derek J. Smith's rights under international copyright law to be identified as author may be freely downloaded and printed off in single complete copies solely for the purposes of private study and/or review. Commercial exploitation rights are reserved. The remote hyperlinks have been selected for the academic appropriacy of their contents; they were free of offensive and litigious content when selected, and will be periodically checked to have remained so. Copyright © 2003-2018, Derek J. Smith.

 

First published online 10:30 GMT 28th October 2003, Copyright Derek J. Smith (Chartered Engineer). This version [2.0 - Copyright] 09:00 BST 5th April 2018

Much of this material appeared in Smith (1998). It has here been shortened, distributed in parts to subfiles, and supported with hyperlinks. Speech and Language Therapy students will probably benefit from refreshing their memories on the difference between segmental and suprasegmental phonology before proceeding.

 

1 - Orthographic Symbol Systems

In our e-paper on "Acquired Dyscalculia", we saw how numeracy skills are frequently spotlighted by the debate over national educational standards. In this paper we turn to the even more complex and perhaps even more nationally essential skills of reading, and to the problems of explaining what goes on inside the mind when the eyes (or the fingertips, if you are a Braille reader) are pointed at a page of text. We shall then be in a position to look at what happens when these skills fail to develop and we end up with the condition commonly known as dyslexia. But firstly, a brief history of human writing systems, or orthographies .....

Click here for the "History of Writing Systems" Glossary. Alternatively, wait until a word catches you out, and providing it has been hyperlinked thus - [glossary] - click on that instead.

Here are some introductory definitions .....

"Reading is the ability to extract visual information from the page and comprehend the meaning of the text" (Rayner and Pollatsek, 1989, p23).

"To learn to read is to learn a system of rules and strategies for extracting information from text" (Gibson and Levin, 1975, p332).

..... and here are some introductory facts .....

* Skilled readers can recognise about 30,000 familiar words in any font, style, and size, and in many orientations.

* Skilled readers do not read every word. Instead, they use both syntactic and semantic context to allow over half of the words to be guessed at.

* Nor do skilled readers read every part of every word. Instead, they use their particular orthography's internal contextual rules to allow component letter strings to be guessed at.

* Skilled readers can process text at a variety of levels in parallel, that is to say, simultaneously. Every time they move their eyes they are making judgements at letter level, letter string level, word level, phrase level, sentence level, paragraph level, page level, or document level. Moreover, by far the majority of this processing is done non-consciously.

* Skilled readers can also cope with a more or less infinite number of unfamiliar words by sounding them out (not often possible, of course, in logographic writing systems[glossary]). You can do this in two ways, (a) according to the rules of foreign languages sharing the same basic alphabet (for example, French), or (b) according to the rules of foreign alphabets, if known (for example, Arabic, Greek, Russian, Hebrew, or Japanese Kana).

In Section 2, we look at the cognitive systems which - when properly trained up and integrated - support these abilities.

 

2 - Reading Skills

In this section we try to explain what goes on inside the mind when the eyes are pointed at a page of text. We shall then be in a position in Section 3 to look at what happens when something fails to develop properly.

2.1 The Visual System and Reading

There are around 1100 visible alphabetic characters on this page, divided into 400 or so discrete words by 399 invisible characters known as spaces. However, of these 1100 characters we can only actually see eight or nine at a time. The explanation for this lies in the way the eye is structured anatomically, and the early work here was by Javal (1879) and Huey (1908). They found that the central sensitive area of the retina - an area known as the fovea - processes a circle of light approximately 1o of arc around the central point of focus. This is surrounded by a semi-sensitive area known as the parafovea, which processes a circle of light approximately 4o of arc around the fovea (ie. from 1o to 5o of arc around the central point of focus), and surrounding the parafovea is a relatively insensitive area known as the periphery, which processes the remainder of the visual field. When an eye is "fixated" (that is to say, held stationary), the fovea is processing the symbols at the centre of the image while the parafovea and the periphery are working out where to direct the eye to next. The resulting eye movement - when it comes - is a rapid ballistic movement known as a saccade (Javal, 1879), and because information cannot be deciphered while the eye is actually moving it ends up arriving in small "packets". At any one instant, you can take in some four letters to the left of the letter you are looking at, and some eight letters to the right of it. This is your perceptual span. What reading involves, therefore, is taking a series of fixations through the text, progressively analysing each information packet visually and syntactically, and gradually building up a higher understanding of what the whole thing is trying to say.

Every now and then, of course, you get confused or distracted and have to glance back to where you got lost and take another look. These glance-backs are known as regressions. In fact, there is enough information in the parafoveal area most of the time either (a) to identify, or (b) to guess at what words are present. There are therefore two possible "next fixation" points, namely Point A - the forward fixation point - and Point B - the "last safe" fixation point. Your eyes move onward to Point A if your attempts to incorporate the contents of the perceptual span into the sentence-so-far are satisfactory, or back to Point B to have another try if they are not. The average fixation duration for normal readers is 200-250 msec, and the average saccade length is 8 - 9 characters (equivalent to 2o of arc on average text at average reading distance) (Rayner, 1983; Rayner and Pollatsek, 1989). The average saccade takes 25-30 msec. to complete, thus accounting for around 10% of total reading time, and if this involves moving down from the end of one line to the beginning of the next, it frequently "undershoots" the left margin. That is to say, the eyes come to rest a few characters into the next line, thus making good use of the left-side capacity of the perceptual span. The main influencing variable is the inherent difficulty of the text itself, as now shown .....

Table 1 - Eye Movement Performance by Text Difficulty: Here are some basic eye performance statistics derived from observations of the reading performance of 10 "good college-age readers". The trends are clear. As text becomes more difficult, the eyes process smaller packets of information at a time, dwell longer in the same place as they do so, and have to go back and check things more frequently. The end result is that the overall reading speed decreases from 365 words per minute (wpm) for easy text to 233 wpm for difficult text. [Retabulated from Rayner and Pollatsek (1989, p118).]

Text Complexity

Fixation Duration

(msec.)

Saccade Length

(chars/degrees of arc)

Regressions

(%)

Reading Speed

(words/minute)

Light Fiction

202

9.2

2.3

3

365

Newspaper Articles

209

8.3

2.1

6

321

Psychology Textbook

216

8.1

2.0

11

308

Biology Textbook

264

6.8

1.7

18

233

We also need to consider what is known as the word superiority effect. This phenomenon was first noted by Cattell (1886). He presented subjects with brief visual arrays and found that letter strings were more slowly analysed for the presence of a target letter than equivalent length words. Or to put it another way, words seem to place less load on the perceptual system than their constituent letters. In common with many others, Johnston and McClelland (1980) use a hierarchical model to describe what they think is happening. They see word perception as a three-stage process, with the final act of recognition being made in the third stage. They see text as being firstly processed by a set of position-specific feature detectors. These pass information to position-specific letter detectors, activating likely candidates and inhibiting those inconsistent with those features. These then pass information to word detectors, again activating likely candidates for recognition and inhibiting unlikely ones. They describe the final inhibitory process thus .....

"..... activation of the detector for 'R' in the first position would excite the detector for 'READ' and inhibit the detector for 'HEAD'. (Also inhibited would be detectors for words such as 'DEAR' that had an 'R' elsewhere in the word but had some other letter in the first position.)" (Johnston and McClelland, 1980, p506).

These three stages are built into the Coltheart, Curtis, Atkins, and Haller (1993) model referred to later in this paper.

2.2 Grammatical Processing and Reading

From this point onwards, we cannot make progress with the psycholinguistics without freely using linguistic concepts and vocabulary. We have therefore provided a glossary to assist non-experts in this area. Click here for the Psycholinguistics Glossary. Alternatively, wait until a word catches you out, and providing it has been hyperlinked thus - [glossary] - click on that instead.

We can now begin to understand why it is repeatedly pointed out (for example, by Gibson and Levin, 1975) that learning to read words and learning to read their meaning are not one and the same thing. As well as decoding the individual symbols, we have also to analyse the sequence of those symbols. We need to know what grammatical rules the mind has applied - we have to tell the nouns from the verbs and the subject from the object, and so on; and once again eye movement performance provides valuable data. To start with, the most important phenomenon under this heading is that eye fixation times vary considerably with the grammatical role of the word being fixated upon. Just and Carpenter (1980) have demonstrated, for example, that on average the eyes will dwell five times as long on an agent [glossary] than a connective [glossary], and twice as long on an adverb [glossary] than a possessive [glossary]. Here are some of the detailed timings .....

Table 2 - Eye Fixation Time by Linguistic Type: When fixation time is analysed according to the linguistic role of the word being fixated, a clearer picture emerges as to where that time is actually being spent. Where words are familiar (top block of data), average processing times range from 9 msec. for each connective to something over 50 msec. for key nouns. Where words are not so familiar however (middle block of data), this increases to over 800 msec. There are also processing time penalties for line shifts, and for "wrapping up" one's understanding of what has just been read at both sentence and paragraph level (bottom block of data; this matter further discussed in text). [Data from Just and Carpenter (1980, p337).]

Word Type

Processing Overhead (msec.)

Connective

9

Possessive

16

Place or Time

23

Direct or Indirect Object

25

Determiner

26

Adverb

29

Adjective

44

Agent

51

Instrument

53

Novel Word

802

New Line

30

Last Word (Sentence)

71

Last Word (Paragraph)

157

Just and Carpenter's last word" timings are worth noting, because they show that it even takes a discrete amount of time to process the full stop at the end of a sentence. This phenomenon is known as "sentence wrap-up", and is seen as allowing extra time for within-sentence ambiguities and/or omissions to be "resolved" after the words themselves have been read. Their data show that the last word in a sentence takes an additional 71 msec. to be processed - a penalty which increases to 157 msec. if that word is also the last word in a paragraph. Much of this time is spent clearing up the identity of the persons and objects being referred to. Consider the following specimen sentences .....

(1a) It was dark and stormy the night the millionaire was murdered.

(1b) The killer left no clues for the police to trace.

(2a) It was dark and stormy the night the millionaire died.

(2b) The killer left no clues for the police to trace.

Despite the fact that in both conditions the (b) sentences [the mauve ones] were identical, the processing times were not! It took around 500 msec. longer to process sentence (2b) than the identical sentence (1b), with the additional time being spent fixating on the word killer and the end of the sentence, and the presumption is that this time penalty was spent going back to sentence (2a) and adding the implication that the millionaire's death could not have been accidental.

More recently, Millis and Just (1994) have discovered another interesting function of connectives. They looked at sentence pairs of the following sort .....

(1a) The elderly parents toasted their only daughter at the dinner

(1b) Jill had passed the exams at the prestigious university.

These are perfectly good sentences in their own right, but can also be linked by the connective "because" to make a single longer two-clause sentence .....

(2) The elderly parents toasted their only daughter at the dinner, because Jill had passed the exams at the prestigious university.

Millis and Just prepared 72 such sentence pairs, and presented them to 64 American college students. Each sentence pair was followed by an immediate probe verb which subjects had to judge as having been present in the first clause (C1) or not. In the specimen above, the probe word "toasted" was in C1, whereas the probe word "praised" was not. It was predicted that recognition time for a connected C1 would be less than that for an unconnected C1 because no C1 wrap-up processing would yet have taken place and all the relevant word traces would still be fully active in verbal memory. Results confirmed this prediction, with verification times for connected C1 being 50 msec. faster than for unconnected C1 (1300 msec. vs 1350 msec). Verification times for retrieval from C2 were unaffected, being 1140 msec. in both conditions. The possibility is accordingly that "in the presence of a connective, readers construct a representation of clause 1 and then place it aside as they construct a representation of clause 2, rather than incorporating clause 2 into the representation of clause 1 as it is being initially represented" (p144).

There are clearly some very important research issues here, but the underlying message is that when it comes to reading your eyes are under the command of your mind's syntactic processes (whatever and wherever they might be), but that the syntactic processes are themselves under the command of something even higher, and it is to that issue that we now turn .....

2.3 The Semantic System and Reading

Our third and final task during reading is that of interpreting the writer's original communicative intention, and here things get very complicated. One of the classic studies was by Bransford and Franks (1971), as now described .....

Classic Study - Bransford and Franks (1971)

This study investigated the processes involved in forming complex sentences out of simpler ones. Four different complex sentences were used .....

A "The ants in the kitchen ate the sweet jelly which was on the table."

B "The warm breeze blowing from the sea stirred the heavy evening air."

C "The rock which rolled down the mountain crushed the tiny hut at the edge of the woods."

D "The old man resting on the couch read the story in the newspaper."

These sentences could each be broken down into four separate 1-idea sentences. For example .....

A1 "The ants were in the kitchen."

A2 "The jelly was on the table."

A3 "The jelly was sweet."

A4 "The ants ate the jelly."

And the 1-idea sentences could then be variously combined into 2-idea and 3-idea ones as follows .....

"The ants in the kitchen ate the jelly." (A1 and A4 combined)

"The ants ate the sweet jelly which was on the table." (A2, A3, and A4 combined)

Sets of the shorter sentences were presented, and then immediately followed by additional sentences which subjects had to judge as having or not having been heard before. There were many false positive recognitions of 2-, 3-, and 4-idea sentences (indeed, the false positives were frequently rated more confidently than the true positives). This was interpreted as indicating that subjects acquired "something more general or abstract than simply a list of those sentences experienced during acquisition" (p348). They had processed gist rather than word-perfect detail.

Bransford and Franks termed this process "the abstraction of linguistic ideas" (p348), and saw it as being similar to Bartlett's (1932) notions of schema-based semantic memory. In other words, what has been read is "quickly recast into something like a propositional representation, where meaning is represented independently of its form of presentation" (Rayner and Pollatsek, 1989, p303).

But abstraction, of course, is a process, and processes can go wrong. There is, for example, a clinical syndrome known as hyperlexia wherein autistic children can read fluently and at speed, but understand little or nothing about what they have read. Parker (1917) termed this behaviour a "pseudo-talent", and Phillips (1930) termed the people who had it "talented imbeciles". Healy (1982) examined 12 such children between the ages of five and 11 years, and found poor language development despite "intense early preoccupation" with reading, and high levels of word recognition. In short, it was the simple mechanical activity of reading which motivated them, not any meaning in the story read. Similarly, Goldberg and Rothermel (1984) reviewed the case histories of eight hyperlexic American children aged between 5.2 and 17.8 years. All had language delays and difficulties in relationships, but again all had learned to read without formal instruction by 5 years of age. Various measures were discussed, and the authors concluded that both the direct and phonological routes were operational, although the former were preferred. They also concluded that some semantic processing was going on, although it was quite restricted. Paragraph-level comprehension in particular was much impaired compared to sentence and single-word comprehension.

At the other extreme, the most fortunate readers are those who can extract the point of a piece of text without reading every word. Indeed, the most effective readers are those who can follow the gist of a page without having to read every sentence (or even every paragraph). Gibson and Levin (1975) describe the strategies adopted by the skilled adult reader as promoting "economy of processing" (p474). These strategies include .....

* flexibility of attention (according, for example, to text difficulty and style)

* flexibility of speed (according, for example, to personal interest and objectives)

* selection of largest text unit appropriate to the given task

* selection of the least overall amount of information appropriate to the given task

* continual reduction of the amount of information being processed

 

3 - Box-and-Arrow Models of Reading

In common with other areas of psycholinguistics, there have been many attempts over recent years to explain the reading process by recourse to graphical techniques borrowed from the computing industry. The resulting models are both powerful and concise, and - approached with appropriate caution - expose inconsistencies in our theories and promote new research .....

Gough's (1972) model

Rayner and Pollatsek's (1989) model

Coltheart, Curtis, Atkins, and Haller's (1993) model

 

4 - Developmental Dyslexia

The literature on the developmental form of dyslexia is generally accepted as beginning with Pringle-Morgan (1896), who reported the case of Percy, a 14 year old boy .....

"Percy was bright, healthy, and of a good family. He was quite good at arithmetic but had an apparently isolated difficulty in learning to read and write which was seriously interfering with his education. Orally he could more than hold his own in class but when asked to write 'carefully winding the string round the peg' Percy wrote 'calfully winder the strung rond the pag'. In the absence of any history of injury or illness, Morgan concluded that this must be a case of congenital word-blindness." (Naidoo, 1972, p9, after Pringle-Morgan, 1896.)

However, the condition has since been repeatedly renamed. Some, including Pringle-Morgan, have gone for Hinshelwood's (1895) "word blindness", whilst others have gone for "reading backwardness" (Shearer, 1968), "developmental dyslexia" (Vernon, 1971), "specific reading difficulty" (Selikowitz, 1993), and "specific dyslexia" (Hallgren, 1950; Naidoo, 1972), and others for such convoluted terms as strephosymbolia (Orton, 1937), congenital symbolamblyopia, typholexia, analfabetia partialis, and amnesia visualis verbalis. In the following sections we shall be looking at what the condition really involves, as well as at what might be causing it, but firstly here is the official World Federation of Neurology definition .....

"[Developmental dyslexia is a] disorder manifested by difficulty in learning to read despite conventional instruction, adequate intelligence, and sociocultural opportunity"

4.1 What Does Dyslexia Look Like?

As with any developmental disorder, the essence of developmental dyslexia is that it first manifests itself as a developmental delay in childhood. Indeed, if we adopt the conventional three-stage approach to the acquisition of reading skills (see panel), we may regard it as a difficulty moving successfully from the logographic stage of reading to the phonological stage .....

The Three-Stage View of Reading Development

Here are the three stages adopted by authors such as Seymour and MacGregor (1984) and Frith (1985) .....

(1) Logographic Stage: This is when words come to be recognised by sight as wholes. It is the ability promoted by "look and say" or "flashcard" methods of teaching reading, and implies the development of the sort of visual input lexicon - or sight vocabulary. Normal children achieve this stage at around 5 - 6 years of age. Errors at this stage are typically visually based, often as a result of a common characteristic (for example, "lorry" being misread for "yellow" because of the shared "y"). However, logographic reading becomes increasingly inefficient as the total number of words in the vocabulary grows, and it soon becomes advantageous to learn grapheme-phoneme correspondences as well .....

(2) Phonological Stage: This is when grapheme-phoneme correspondences are being learned, and is the ability promoted by "phonics" methods of teaching reading. What is important is that unfamiliar words can now be "sounded out", enabling them to be recognised as if they had been heard rather than seen. This implies the development of the sort of alternative processing routes described in modern transcoding models. Normal children achieve this stage at around 6 - 7 years of age.

(3) Orthographic Stage: This is when the child has acquired the full repertoire of processes and processing routes to cope with all the irregularities of the written language. It is also characterised by increasingly efficient movement of the eyes across the page as the transitional probabilities between letters are learned and information is gradually processed in adult-sized chunks. Snowling and Frith (1981, p87) describe this as learning "the letter-by-letter structure of words", and normal children achieve this stage at around 8 - 10 years of age.

Caution: Not all authors subscribe to developmental stage theory - see the counter-arguments by Stuart and Coltheart (1988) and Goswami (1993), if interested.

It is also totally characteristic of developmental dyslexia that it should present as a cluster of related difficulties in the midst of otherwise normal development (as was the case with Percy above). Here are the difficulties which usually cluster together .....

* reading age considerably behind general intelligence

* confusion between left and right

* ideosyncratic spelling

* confusion of similarly shaped letters, such as "w" and "m"

* confusion of similarly sounded letters such as "v" and "f"

* confusion of palindromic words such as "was" and "saw"

* confusion of anagrammatic words such as "left" and "felt"

* confusion of word sequence in memory tasks

* omissions or insertions of words

* difficulty moving from the end of one line of text to the beginning of the next, resulting in word or phrase repetitions

* difficulty keeping (or finding again if lost) one's place on a page of text

* poor speech intonation

* reduced text comprehension

* frequent mispronunciations (even of familiar words)

A typical pattern of impairment is shown in Figure 1 ..... 

Figure 1 - Why a "Specific" Defect? Here is a comparison of the developmental ages of three nine-year old children, one a normal reader, one a dyslexic reader, and one with broad-spectrum learning difficulties. Performance at eight different perceptuo-cognitive tasks was assessed. Note that the normal reader performs at around chronological age across the board, and that the mentally retarded child performs around three years below chronological age across the board. The dyslexic, on the other hand, performs at chronological norm on all tasks except sound-symbol association and auditory sequencing. It is this sort of narrow band difficulty in the midst of otherwise acceptable performance which puts the adjective specific into specific dyslexia.

If this diagram fails to load automatically, it may be accessed separately at

http://www.smithsrisca.co.uk/DLEX-fig1.gif

DLEX-fig1.gif

Retabulated from Zuckerman and Chase (1983, p259). This graphic Copyright © 2003, Derek J. Smith.

4.2 Is Dyslexia Inherited?

Much was made of hereditary factors in early explanations of developmental dyslexia. Stephenson (1907) reported on six cases of the disorder across three generations of a single family, and argued that this clearly indicated an underlying hereditary factor. Hinshelwood (1917) agreed, and suggested an inherited lack of development in the region of the left angular gyrus (Brodmann's Area 39), the area which in acquired dyscalculia [reminder] produces symptoms of right-left disorientation and writing difficulties. More recently, Orton (1937, p127) remarked on "the frequent occurrence of two or more cases of language disability in the same family", Hallgren (1950) found in a sample of 276 dyslexics that 88% had other members of the immediate family also affected, and Herrmann (1959) found complete concordance (12 out of 12) between dyslexic twins and their monozygotic siblings, but only 33% concordance (11 out of 33) for dizygotic twins. The issue remains topical, with Snowling and Stackhouse (1996) reporting a 50% probability of the son of a dyslexic father himself becoming dyslexic, and a 40% probability of the son of a dyslexic mother.

Other workers have concentrated on non-hereditary factors such as malnutrition or oxygen starvation at birth. Naidoo (1972) is one author who recognises the co-occurrence of "soft" neurological signs with some of the indicators of dyslexia. She cites studies by Cohn (1961) who compared 46 dyslexic children aged seven to ten years with 130 normal readers and found significant differences in reflexes, motor coordination, and EEG. More recently, Galaburda (1988) has claimed that dyslexic brains lack the left-hemisphere size superiority characteristic of the brains of normal readers, Goldman-Rakic and Rakic (1984) have noted distortion in cortical microarchitecture, possibly due to problems during embryonic development, and Frith (1992, p13) blames "some subtle brain abnormality, occurring probably well before birth".

In the remainder of this section we look at five particular ways of explaining dyslexia, all of which can to a greater or lesser extent result initially from - or be seriously compounded by - underlying organic damage. These are (a) regarding dyslexia as a perceptual problem, (b) regarding it as a phonological coding problem, (c) regarding it as a cerebral dominance problem, (d) regarding it as an educational problem, and (e) regarding it as a social problem.

4.3 Dyslexia as a Perceptual Problem

So might developmental dyslexia be the result of a damaged or dysfunctional visual system? Certainly, data from a variety of sources indicate that a dyslexic's eyes pause more often during reading than do a normal's eyes (that is to say, their saccade length is shorter). They also fixate for considerably longer, regress more often, and fail to make best use of the "next line undershoot" stratagem noted by Just and Carpenter (1980) [see Section 2]. In short, dyslexic eyes do not move anywhere near as efficiently as they ought to. The key statistics are given in Table 3 .....

Table 3 - The Pattern of Dyslexic Eye Fixations: Here are some basic eye performance statistics for both normal and impaired readers. [Top data block from Rayner and Pollatsek (1989, p432), other factors referenced in the text.]

 

NORMAL

DYSLEXIC

Average Fixation Duration

200 - 250 msec.

300 - 350 msec.

Average Saccade Length

8 - 9 characters

3 - 6 characters

Frequency of Regressions

10 - 20 %

30 - 40 %

Next Line Undershoot

well handled

not well handled

Despite all these well-established differences in ocular performance, the critical point is that there is nothing wrong with the dyslexic's eyes as such! Adler-Grinberg and Stark (1978) compared the oculomotor performance of 25 dyslexic and 19 normal children on pictorial scanning as well as textual perception tasks. They found no significant differences between the two groups on the visual perception tasks, with the dyslexics quite capable of scanning at normal speed and fixating for normal duration. On the textual tasks, however, fixations were longer (340 msec. as against 280 msec.) and the perceptual spans smaller (0.9 words as against 1.2 words). Dyslexic eyes work perfectly, it seems, until pointed at a page of text! The dyslexic's "information processing obstacle" (p569) - whatever it is - must therefore be post-perceptual, that is to say, it must lie in those areas where visual material starts to be integrated with stored linguistic material, and it is to these that we turn next .....

4.4 Dyslexia as a Phonological Coding Problem

We turn now to what people often like to refer to as the mind's "higher" processes. What we described in the preceding section was the sort of early visual processing carried out in the visual cortex (Brodmann's Areas 17 to 19). But that is only the beginning of the story, because recognising something is much less than understanding it. The implications of a scene - the semantics of it all - are calculated elsewhere in the brain, and the problem is that neither the cognitive nor the neuropsychological approach yet has any real idea what such processing really involves nor where it is carried out. Thus even the cognitive supermodels [see our e-paper on "Transcoding Models"] deliberately consign the highest level of processing to a single black box referred to conveniently as the semantic system or similar, and neuropsychological models such as Crosson's (1985) model of language processing rapidly end up with an impenetrably complex set of feedback control circuits interlinking so many cortical and subcortical structures that the semantic system could be hidden more or less anywhere. Nevertheless, all models need to say something about the phenomenon of inner speech .....

Key Concept - Inner Speech: Inner speech is the name generally used to describe the phenomenon of silent speech. It is the name given to the form of human thinking which involves (a) generating, and (b) listening back to, a stream of unspoken speech. It is holding a conversation in your own mind. It is thinking to yourself. Amongst the early workers with dyslexia, Huey (1908) recognised that inner speech played a major part in successfully processing textual input. The phenomenon was then a major part of Vygotsky's (1934) Three-Stage Theory of Speech Development. Vygotsky carefully analysed the qualitative changes in the relationship between language and thought which take place in the period between infancy and adulthood, and - with appropriate credit to Piaget's early writings - identified three major developmental stages. The first stage in this process is known as external speech, and is characterised by every utterance having an obvious communicative intent. The second stage is known as egocentric speech, and is characterised by self-referenced thinking out loud. The third stage is known as inner speech, and is characterised (a) by moment-to-moment changes in the sense of the words used - their full subjective implicature, rather than their dictionary definitions, (b) by the use of idiosyncratic word combinations to express complex ideas, and (c) by single words becoming so "saturated with sense that many words would be required to explain [them] in external speech" (p148). And Vygotsky takes pains to point out (p130) that inner speech is more than mere verbal memory or truncated external speech. It must be regarded, he argues, "not as speech minus sound, but as an entirely separate speech function" (p139). More recently, inner speech has been built into the Ellis and Young (1988) cognitive supermodel as an internal feedback route from the speech production process to the auditory analysis process. There is also a lot to learn from Crosson's (1985) model of language processing, because although it deals almost exclusively with the problems of overtly spoken language, it proposes simultaneous semantic and phonological output monitoring circuits. The former runs via the thalamus and the latter via the arcuate fasciculus, and together they allow Wernicke's Area to keep tabs on what Broca's Area is about to output. This is precisely the sort of internal monitoring/feedback circuit which may one day be found to be responsible for inner speech, one of psychology's greatest unsolved mysteries.

Modern interest in phonological coding arose in the mid-1960s and the topic quickly became one of the central issues in memory research. Here are some of the studies .....

(a) Bruce (1964): Bruce (1964) noted that children at five years of age lacked awareness of the phonetic sequence within words, whereas by age seven they had started to acquire it. Thus, when asked "What word would we have if we took the first [k] away from 'clock'?" a five year old would shrug and a seven year old would answer "lock". It was as if the five year old was unable to carry out even simple operations upon his/her short term memory of the word.

(b) Firth (1972): Firth (1972, cited in Jorm, 1979) was another who was quick to conclude that dyslexic children were typically deficient in "phonics" skills. He found that the nonsense word test, the blending test, and the auditory analysis test (see below) were all good at detecting incipient reading problems, and by the end of the 'seventies it had been very well established that inefficiencies in phonological coding had a lot to do with the emergence of developmental reading difficulties.

(c) Shankweiler et al (1979): Conrad's (1964) acoustic confusibility paradigm was also used. Shankweiler, Liberman, Mark, Fowler, and Fischer (1979) compared samples of normal and dyslexic readers on the confusibility task and found some interesting differences. With normals, a phonologically similar sequence such as "B - C - G - T - P - V - D" attracted more errors than phonologically dissimilar sequences such as "H - Q - S - L - W - R - K". With dyslexics, however, the effect was significantly less pronounced, indicating that they "are unable to make effective use of a phonological code in memory [possibly because they] have less easy access to the phonological representations of words" (Rack, 1985, p325/337).

(d) Byrne and Shea (1979): Byrne and Shea (1979) had subjects listen to a sequence of spoken words and say when they heard a word which had been called before. Every now and then, of course, a false alarm was called - subjects responded to a word they thought was being repeated but which was in fact being heard for the first time. When these false alarms were analysed, it emerged that for normal readers the words most likely to be mistaken were those which sounded like the corresponding target words (thus they might false alarm to "fair" when "chair" had originally appeared). For dyslexics, by contrast, the words most likely to be mistaken were those which had some sort of semantic link to the corresponding target words (thus they might false alarm to "table" when "chair" had originally appeared). This suggests that dyslexics and normals are using different coding systems to register incoming items: normal readers are encoding phonological features, whilst dyslexics are encoding semantic features.

(e) Rack (1985): Rack (1985) found that normal readers recalled best words which had been paired with rhyming partners (for example, "head-said"), regardless of orthographical similarity. Dyslexics, on the other hand, recalled best word pairs which were orthographically similar (for example "harm-warm") regardless of sound similarity, suggesting that they are using visual features of words as a basis for recall instead of phonetic codes. And the deficit must be at work deep within the overall process of cognition because the effect persists even for word pairs presented auditorily.

(f) Snowling (1987): Similarly, Snowling (1987) reports that the "primacy effect" is reduced in dyslexic children, that "running memory" is impaired, and that confusions are semantically based, and not (as would be expected) phonologically based.

And here are some of the tests now regularly carried out to assess phonological processing skills .....

* Nonsense Word Test: Here the child is asked to pronounce textually presented nonsense words.

* Auditory Analysis Test: Here the child is given a word and asked what sound(s) would remain if a certain sound were removed from it - see Bruce's (1964) "clock-lock" example in the above text.

* Sound Blending/Segmentation Tests: In a sound blending test, the child is given the individual phonemes and asked what word they make when put together. Thus when hearing [k] [Ů ] [p] an effective sound blender would respond "cup". In the segmentation test, the opposite happens. The child is given a word and asked to identify the constituent sounds. The sound blending test is a particularly sensitive one, presumably because it requires not just sound perception, but also sound (and sound sequence!) retention. Naidoo (1972, p69) found that impaired performance on this test was "very common" in dyslexics. At eight years, for example, 81% of dyslexic subjects were unable reliably to blend more than three or four sounds, compared to 40% of control subjects.

* Sound Counting Test: Here the child has to tap out the number of sounds in a spoken word. Liberman, Shankweiler, Fischer, and Carter (1974) found this ability correlated highly with reading ability.

* Digit Sequence Reversal Test: Here the child is given a short series of verbally presented digits and asked to reproduce it backwards. Thus the sequence "one-four-nine-six" should generate the response "six-nine-four-one". In so doing, it is necessary to hold the input sequence in phonological memory while building up the output sequence, and this memory load increases as the sequences get longer.

* Alliteration Detection Test: Here the task is to identify which of a sequence of words does not start with the same letter as the rest.

Snowling (1996, p5) summarises all this evidence as indicating that the reading system is normally "parasitic upon" a pre-existing phonological system set up during the child's experience with spoken language. It follows, she argues, that problems with the phonology cannot but induce problems with reading; there is, as she puts it (p9), "a phonological core" to developmental dyslexia. Thus .....

"..... learning to read is determined primarily by the status of a child's phonological representations and is therefore compromised in dyslexic children who have phonological deficits." (Snowling, Hulme, and Nation, 1997.)

So it seems that a dyslexic's difficulties start to become apparent at the time that normal readers are making the transition from logographic reading to phonological reading! Indeed, some workers even prefer to call the deficit developmental phonological dyslexia. There have, for example, been a series of reports, tracking the development of JM, a developmentally dyslexic child. JM was initially referred at age 8;5 years for underachieving at school (Snowling, Stackhouse, and Rack, 1986). He had a history of speech (but not language) problems associated with an impressive WISC-R IQ of 123. However, his reading and spelling ages were both about two years behind his chronological age. Upon examination, it emerged that he had a normal sight vocabulary for real words, but "had no success whatsoever" with nonwords (Snowling and Hulme, 1989, p383). There was a tendency to "lexicalise" - that is to say, to guess at the nearest regular word, such as reading "pool" when given the nonword "plood". More recently, Hulme and Snowling (1992) report that JM's form of dyslexia is proving to be "highly stable" over time (p63). He now has a slowly expanding sight vocabulary but retains his "massively deficient" nonword reading skills. Again, he is fixed at the logographic stage of development.

To help conceptualise what might be going on in the minds of dyslexics like JM, Jorm (1979) borrowed the phrase cognitive deficit from clinical psychology, where workers such O'Connor and Hermelin (1971) had used it to explain cognitive processing failures in autistic children .....

Key Concept - Cognitive Deficit: A cognitive deficit is a specific difficulty in the midst of otherwise acceptable cognitive performance. It is a problem with one cognitive skill in particular when other cognitive skills are intact. A good early example of how this approach can be used to help explain clinical problems is Cameron's (1938) analysis of schizophrenic thinking. Cameron saw schizophrenic thinking as typically pathologically inefficient in the construction and use of mental concepts. Schizophrenics are incoherent and difficult to follow because they focus on the wrong aspects of a situation, and they focus on the wrong aspects of a situation because they fail for some as-yet-unknown reason to define things precisely enough. (Others such as Payne and Friedlander, 1962, have used the phrase overinclusive thinking for this same set of symptoms.) More recently, the cognitive deficit type of explanation has been intensively used to explain key aspects of Kanner's autism. O'Connor and Hermelin (1971, p227), for example, proposed that autism was characterised by an inability "to encode stimuli meaningfully". Autistics are bad at abstracting an essential underlying feature from a word series, and, above all, at forming what are known as higher order representations. These are attributions of states of mind to other people. The states of mind in question include volitional states such as want, covet, intend, etc, as well as belief states such as believe, doubt, think, etc. The number of people involved in such representations can vary upwards from one, with the phrase second order representation indicating that two minds are involved, third order representation indicating three minds, and so on. A typical third order representation would therefore take the form "I suspect | that Tom doubts | that the borrower intends to repay the loan". Against this background, the autistic cognitive deficit is that whilst normals can cope with up to fifth or sixth order representations, autistics cannot process even second order ones - they seem incapable of recognising that what other people think and feel is not part of their own direct experience. They have not successfully abstracted "self" from "other", and consequently behave as though the world was theirs and theirs alone. Within dyslexia theory, phonological coding is another of these important specific cognitive skills, and when it goes awry for some as-yet-unknown reason equally specific aspects of cognition immediately suffer, including the ability to put phonological skills to text-processing uses. The topic is well reviewed in Snowling (1987) and Frith (1992).

Exercise 2 - Some Diagnostic Psycholinguistic Modelling

Orton (1937, p165) tells the tale of a schoolgirl who was top of her class in spelling in Texas, but failed the subject after moving to Iowa. He looked into her background, and found that whilst the spelling lessons in Texas had been chanted, the ones in Iowa were written silently to dictation. With appropriate reference to Kay, Lesser, and Coltheart's (1992) PALPA supermodel, suggest which cognitive module(s) might have been impaired in this instance.

4.5 Dyslexia as a Cerebral Dominance Problem

Another common argument is that developmental dyslexia is the result of defective cerebral dominance, and that it can therefore be predicted by screening for mixed or left handedness. Orton (1937), for example, detected "a great many" (p61) mixed handedness cases in his language disordered children, and argued that language was best left to a single hemisphere to deal with. If both hemispheres tried to get involved, then they would be constantly contradicting each other, and the resulting processing would be seriously inefficient.

ASIDE: In fact, it takes a lot of complicated circuitry to synchronise two or more simultaneously active processors. See our e-paper on "STM Subtypes" (Part 6, Section 3.5) if interested in the technicalities.

The complication is, however, that few of us are wholly left or wholly right handed - many are somewhere in between. Gates and Bond (1936) used five measures of handedness - using a hammer, spinning a top, cutting with scissors, picking up small articles, and throwing a ball - and classed "mixed-hand dominance" as any score less than 5 left or 5 right. They then found "no consistent tendency" for eye-dominance, eye acuity, or hand dominance to be related to reading ability. Harris (1957) agreed on the issue of eye-dominance, but found significantly higher incidences at age seven years in directional confusion and mixed dominance. Similarly, Shearer (1968) found a "considerably higher proportion of strongly right-handed subjects among the non-retarded group [and] a much higher proportion of mixed handedness and weak hand preference among the retarded group" (p203). Some data from the Shearer study is shown in Table 4 .....

Table 4 - Shearer's (1968) Handedness Data: Here are the handedness profiles of two groups of children, a dyslexic group of 114 subjects, and a non-dyslexic control group of 225 subjects. IQ testing gave group averages of 92.9 and 104.0 respectively. Handedness (and the strength thereof) was assessed on a 13-item scale. The overall percentage of strongly right-handed subjects was 63% in the dyslexics (against 74% in the non-dyslexics), the overall percentage of mixed handedness was 8% in the dyslexics (against 3% in non-dyslexics), and the overall percentage of weakly expressed preference (left or right) was 25% in the dyslexics (against 19% in the non-dyslexics). [Retabulated from Shearer (1968, p201).]

NON-DYSLEXIC

Age

Strong Right

Weak Right

Mixed

Weak Left

Strong Left

N

7

26(61)

12(28)

1(2)

3

1

43(100)

8

35(73)

7(15)

2(4)

3

1

48(100)

9

55(79)

8(11)

1(1)

2(3)

4(6)

70(100)

10

52(81)

6(9)

2(3)

1(2)

3(5)

64(100)

11

-

-

-

-

-

-

12+

-

-

-

-

-

-

TOTAL

168(74)

33(15)

6(3)

9(4)

9(4)

225(100)

DYSLEXIC

Age

Strong Right

Weak Right

Mixed

Weak Left

Strong Left

N

7

-

-

-

-

-

-

8

7(39)

6(33)

2(11)

1(6)

2(11)

18(100)

9

22(71)

7(23)

2(6)

0(0)

0(0)

31(100)

10

17(61)

5(18)

3(10)

2(7)

1(4)

28(100)

11

10(63)

4(25)

1(6)

0(0)

1(0)

16(100)

12+

16(76)

3(14)

1(5)

1(5)

0(0)

21(100)

TOTAL

72(63)

25(21)

9(8)

4(4)

4(4)

114(100)

As it happens, it is possible to reconcile the phonological deficit and cerebral dominance explanations by looking at how language processing might be divided between the two hemispheres. Coltheart (1980), for example, argues that one of the right hemisphere's main jobs is to establish the "superordinate semantic category to which a printed word belongs without deciding which exemplar of the category the word actually is" (p349). He also cites evidence from Japanese speakers that the right hemisphere is better at processing word parts than word wholes. Identification of tachistoscopically presented Kanji characters is more accurate when they are exposed in the left visual field (right visual cortex). For Kana characters, on the other hand, identification is better when presentation is in the right visual field (left visual cortex). Since we have already seen that Kana is a syllabary [glossary] whilst Kanji is logographic [glossary], this data is consistent with the view that the right hemisphere of skilled Japanese readers is better at dealing with syllabic characters whilst the left hemisphere is better at word-level symbols. Figure 2 shows how it might all be organised .....

Figure 2 - Coltheart's "Right Hemisphere Hypothesis": Diagram (a) shows a normal right-handed subject, wherein the left cerebral hemisphere is dominant for language because it contains the language areas traditionally known as Broca's Area and Wernicke's Area. The former prepares sentences for spoken output and initiates their physical production (shown here in bubble-speak form), and the latter is the focal point of what seems actually to be a widely distributed word comprehension network. The right hemisphere forms part of this word comprehension network, containing additional semantic information, including important categorising data. When a normal brain reads out loud, therefore, each stimulus word has discrete left and right hemisphere effects. In the left hemisphere it propagates from the visual cortex to Wernicke's Area along pathway A, and then from Wernicke's Area to Broca's Area along pathway B. At the same time, it propagates into the right hemisphere along pathway C to activate the appropriate nodes in the broader semantic network. For the stimulus word STONE, this might be the node for [minerals | fragments of | small], and the critical point is that this same semantic area would also be stimulated by such stimulus words as "GRAVEL", "PEBBLE", "COBBLE", etc. In a deep dyslexic subject - diagram (b) - pathway B is presumed dysfunctional, so that activation of Broca's Area is forced to go via the right hemisphere pathways D and E as shown. But the right hemisphere, remember, only knows the general word category of the stimulus word, so this transmission is subject to what are known as semantic paralexical errors such as saying "rock" instead of "stone". This "Right Hemisphere Hypothesis" of Dyslexia is given greater weight by observations by the late Norman Geschwind's research team (for example, Geschwind and Levitsky, 1968) that there are significant individual differences in the size of the planum temporale - the area of temporal lobe cortex facing upwards into the insula just posteriorly to the primary auditory area (Heschl's gyrus).  

If this diagram fails to load automatically, it may be accessed separately at

http://www.smithsrisca.co.uk/DLEX-fig2.gif

DLEX-fig2.gif

Redrawn from a black-and-white original in Bradshaw and Rogers (1992, p345). This graphic Copyright © 2003, Derek J. Smith.

 4.6 Dyslexia as an Educational Problem

The fourth type of explanation is the "Teachers Don't Recognise It" approach, and goes back at least to Vernon (1971, pp113-114) who noted that "children taught by untrained, inexperienced, and unskilful teachers tend to be especially backward at reading". This approach holds that the education system is guilty in some way of "letting its dyslexic children down". More specifically, it holds that teachers - and especially infant and primary school teachers - are not properly trained to spot developmental dyslexia and are therefore far too willing to write dyslexic children off as lazy, naughty, or stupid. Which is very much a self-fulfilling prophecy, of course, because once a child has been in any way marginalised in the classroom its commitment to the educational process will be severely compromised. And things can get very grim very quickly. Consider .....

"One teacher got so frustrated when I couldn't spell the word 'police' that he grabbed me by the neck and banged my head against the wall" (Redwood, 1997, quoting a 19 year old female dyslexic).

"It comes to a point where the parent gets frustrated by the child's inability to read and the child becomes emotionally screwed up because you are putting this enormous emphasis on something he can't do" (Redwood, 1997, quoting a dyslexic's parent).

4.7 Dyslexia as a Social Problem

The fifth and final type of explanation is the "social disadvantage" approach. This homes in on the fact that "difficulties in learning to read are often associated with children's social background" (Vernon, 1971, p95). In a fairly recent study, for example, Raz and Bryant (1990) compared the reading progress of children from socially advantaged (middle class) and socially disadvantaged backgrounds. They found that even when matched for IQ the disadvantaged children did not start to fall behind the advantaged until after they had gone to school. Then, however, a "serious gap" between the two groups soon emerged. On both the Initial Phoneme Test and two different rhyming tests, an initially small superiority at age five or six years became "much stronger" 18 months later. While both groups had improved, the middle class children had improved more. Raz and Bryant then analysed seven measures of home environment and found significant differences in (a) how often parents read to their children (Disadvantaged = 12.18 times per month; Advantaged = 19.45 times per month; p < 0.01), (b) the number of books owned by the child (Disadvantaged = 13.75; Advantaged = 35.40; p < 0.001), and (c) the number of visits to the library (Disadvantaged = 3.72 times per month; Advantaged = 9.30 times per month; p < 0.05). Raz and Bryant close by recommending that remediation programmes should address two critical weaknesses, firstly the child's sensitivity to phonological segments, and secondly their ability to follow the gist of stories and make appropriate inferences (to which we would add properly trained teaching staff and a supportive home environment). This and other remedial programmes are reviewed in the next section.

 

5 - Remediating Substandard Reading

Having looked at what developmental dyslexia might be, we now turn to how it might be alleviated. We shall focus primarily on the sort of remedial programmes which have been aimed at dyslexic schoolchildren, and then close by comparing and contrasting these with programmes aimed at enhancing the efficiency of reading in normal adults. In this way we shall be dealing with both aspects of Raz and Bryant's (1990) recommendations (see above), namely that remediation programmes should address not only a reader's sensitivity to phonological segments, but also their ability to follow the gist of stories and make appropriate inferences.

5.1 Phonological Awareness Training

Even before phonological coding had achieved its modern popularity as a research topic, auditory skills were tested by educational psychologists and trained up by special needs teachers. Indeed, Tansley (1967) reminds us that you have to go back at least as far as Monroe (1932) for the early programmes of exercises designed to improve the differentiation of sounds. Tansley himself then provides a typical example of how reading remediation programmes were targeted in the UK in the mid-to-late 1960s, dividing his effort almost exactly between visual and auditory skill assessment and remediation, and including in the latter many of the tests we have already seen (activities such as rhyme appreciation, sound blending, rhythm matching, and letter-sound pairing). As we are about to see, these tests remain the backbone of even the most modern remediation packages. Bradley (1988) reports on a longitudinal correlational study of a group of more than 400 children. The first data were collected in 1979 when the children were aged four/five years, and a follow-up study carried out at eight/nine years. Results indicated a strong relation between rhyming ability at the lower age and reading and spelling ability at the higher age. Bradley then selected 65 of these children for low initial rhyme judgement scores, and conducted an intervention study. Two treatment groups were used, with both receiving 40 ten-minute rhyme training sessions over a period of two years (being taught, for example, that "cat", "bat", and "rat" could be grouped by their sound similarity), and the second receiving additional demonstration of how to form each word out of plastic letter-tiles. There were also two matched control groups, the first of which received an equivalent amount of semantic categorising training (being taught, for example, that "cat", "bat", and "rat" could be grouped by all being animals) over the same period, and the second received no training but simply underwent the testing. This is what Bradley found .....

"At the end of that time the children who were trained in phonological categorisation were 3 - 4 months ahead of the children trained in semantic categorisation on tests of reading and spelling, but there was no difference between the groups in arithmetic." (Bradley, 1988, p5.)

Lundberg, Frost, and Petersen (1988) reported on a similar longitudinal study in which 235 Danish preschool children received 15 - 20 minutes daily training in metalinguistic (see below) exercises and games over an eight-month period. They reported small but significant improvements on rhyming and syllable manipulation tasks, and a "dramatic" improvement on phoneme segmentation. The use of rhyme as a training tool has since been concisely reviewed in Glasman (1994), who explains that children with language disorders often have little idea about rhyme, and that simple nursery rhymes are "a fun way" for them to develop that appreciation. And he cites Joy Stackhouse of the National Hospital's College of Speech Sciences to the effect that you can only rhyme successfully if you have a certain amount of "metalinguistic awareness" .....

Key Concept - Metalinguistic Awareness: Metalinguistic awareness is the ability to use words to describe and comment upon other words. Thus, if you know only that the word "sun" refers to the sun, and that the word "bun" is a type of cake, then you have linguistic but not metalinguistic awareness. If, on the other hand, you know also that "they rhyme", then you have commented upon the words themselves (rather than their referents), and you have begun to develop metalinguistic awareness as well; similarly, if you know that "carpet" can be divided into "car" and "pet", or that final "-s", "-ed", "-ing", or "-ish" are syntactically significant morphemes. As we have already seen, metalinguistic skills have long been recognised as correlating with reading ability. By the late 'seventies, however, they were becoming the primary focus of remediation and phonological awareness was easily the most promising subset of metalinguistic awareness because it had been so strongly implicated as the key cognitive deficit in dyslexia.

Ehri (1989) has reported parallel improvements in spelling ability. She provided a group of American kindergarten children with letter-tile practice at spelling CVC and consonant blend words. Once trained to criterion, they were then tested (a) at reading 12 similarly spelled words, and (b) at phonemic segmentation. The spelling trained group outperformed control subjects in both tests, prompting the following conclusion .....

"In our research, we have shown that when the spelling system is learned, it penetrates readers' phonological knowledge in a fundamental way and influences the sounds that they believe are in words[. It] was the spellings that shaped subjects' conceptualisation of the phonemic structure of the words." (Ehri, 1989, p359.)

Nowadays, there are many programmes aimed at developing phonological awareness skills, including one directed at the phonology itself called simply MetaphonŇ (Howell and Dean, 1991). Because they are all fundamentally alike (where they differ at all it is primarily because they are in intense commercial competition), we shall only be looking at one method in detail .....

Case Study - Gorrie and Parkinson's (1995) Phonological Awareness Procedure

So how effective are these methods and how might we improve them? Well in many cases there is probably a strict limitation on what can be achieved even with the best remedial education. With Snowling and Hulme's (1989) case JM, for example, there has been only slow progress "in spite of his superior IQ and the specialist teaching he had received" (p384). Nor is it just a matter of the cost-effectiveness of the remediation - intervening can actually border on the downright cruel. Frith (1992) puts the dilemma this way .....

"Clearly, it would be foolish to make a blind person embark on a programme of eye exercises, [instead people] can be made mindful of the needs of the blind person, and the physical environment can be adapted so as to be safe for someone who cannot see. Exactly the same applies to people with word-blindness [.....] It is not kind to pretend that people are not blind when in fact they are. Nor is it kind to push people if there is little spare capacity. Compensation is a costly process. When mental resources have to be marshalled where they are sparse, then one should think twice about insisting that they are used. [] Once the deficit has been recognised - it can be left alone. Compensation and diversion into other fields are often possible - but not always necessary. Rather than demanding of handicapped children that they make continuous efforts, we should learn to recognise their often heroic struggle. We can respect the difference." (Frith, 1992, pp19-20; emphasis added.)

Layton and Deeny (1996) concentrate on providing practical guidance to the teachers involved, and conclude that more attention needs to be paid to the appropriacy of teaching materials. It is too easy in their view for inexperienced teachers to allow gaffes in this material. They give the examples of .....

* "giraffe" as an example of "g" (inappropriate if hard g)

* "owl" as an example for "o" (not same sound)

* "eye" as an example of words starting with the same sound as "elephant" (it does not)

In particular, teachers and communication professionals should collaborate in the design and evaluation of literacy programmes. Glasman (1994) notes much the same problem with rhyme training. He quotes Lindsay Peer of the British Dyslexia Association .....

"[Rhyme training] looks good on paper but the practicality of it is that teachers are not trained in these methods[. They] don't understand the importance of it, and they don't have time in the classroom." (Glasman, 1994, p7)

Which returns us to the "Teachers Don't Recognise It" explanation discussed in Section 4.6, and may explain why - after half a century of trying - the nation's reading ability remains firmly in the news.

5.2 Other Approaches to Remediation

We turn now to two other important training programmes involving metalinguistic, but this time not phonological, skills.

An orthographic analogy training programme is one which draws the child's attention to patterns in the written word. Thus if a child is slowly taught how to read the word "right", that child should then have no problem with analogous words such as "sight", "might", etc. Such programmes have been experimented with since the mid-1980s (for example, Goswami, 1986). Peterson and Haines (1992) investigated the effects of a one-month orthographic analogy training program on 47 Canadian kindergarten children initially aged 5;4 to 6;4 years. They measured changes in segmentation ability, letter-sound knowledge, and the ability to read unfamiliar words by analogy with familiar ones, and found "a complex interactive effect whereby facilitation of sound segmentation ability, letter-sound knowledge, and reading by analogy have occurred in a kind of mutual synergy" (p121). In other words, if you improve one aspect of cognition slightly it can assist improving others.

Goswami (1993) reports on her experiences with a similar training programme. She exposed "clue" words such as "beak" throughout a session involving words such as "peak" (analogous), "bark" (non-analogous, but visually similar), and "rain" (non-analogous and visually non-similar). She found that normal 5/6 year olds were unable to process rhyming sounds within words but could do so at word end. If we remind ourselves that for the word "beak" for example, the [ b ] is the syllable onset, the [ i ] is the rhyme nucleus, the [ k ] is the syllable coda, and the combined [ ik ] is the rhyme itself, then what Goswami is saying is that coda processing is more advanced than nucleus processing at that age. By age six/seven, however, children had successfully developed this ability, and could predict the pronunciation of "speak", say, from prior knowledge of the word "beak". She concludes .....

"The orthographic analogy evidence suggests that children initially set up recognition units for words that are coded in terms of two phonological units, the onset and the rhyme[.] As reading develops, and spelling is taught, a more fine-grained phonological analysis becomes possible, and children begin coding graphemes in terms of phonemes." (Goswami, 1993, p315.)

Much the same approach has been taken with the higher level processes of thematic comprehension. Yuill and Oakhill (1988) reviewed earlier work which had indicated that poor comprehenders at age 7/8 years differ from good comprehenders "primarily in their failure to make high-level inferences, despite adequate text recall" (p33). They therefore devised and tested a program of inference awareness training. This consisted of the following three types of task .....

* Lexical Elaboration: Here the child is required to listen to a sentence such as "Sleepy Tom was late for school again" and then explain what each word might be telling you over and above the obvious. Thus Tom was probably (but not necessarily) a male pupil rather than a female teacher.

* Question Generation: Here the child is given a four or five sentence story, and required to generate a simple wh- question relating thereto.

* Macrocloze: Here the child is required to guess at a hidden sentence in the middle of a four or five sentence story.

Their subjects were 13 English junior school children and there were seven half-hour practice sessions spaced out over a three and a half week period. Results indicated that less skilled readers improved with training even over this short period, gaining 17.38 months in comprehension age compared to a gain of 5.92 months by the initially more skilled readers (p<0.001).

What is especially interesting about the Yuill and Oakhill study is the fact that improvement seems to take place in poor comprehenders "without any explicit awareness of how meaning relates to the text" (p35). The ability to make inferences is another of those curious subconscious cognitive processes which just happens. In much the same vein, Winograd (1984) has looked at skill differences at text summarising tasks. He divided 75 eighth grade American schoolchildren into groups of 36 "poor" and 39 "good" readers. He then timed them as they read through a test article and had them write a 60-word summary thereof. The summaries were then graded on how accurately the original ideas were represented. Findings included (a) that good readers were better judges of the relative importance of individual ideas than were poor readers, (b) that poor readers were at least consistent amongst themselves in what they considered important (seeming to prefer rich visual detail at the expense of the more subtle points), (c) that poor readers were also bad at including in their summaries ideas which they had nevertheless rated important during their initial inspection of the text, and (d) that comprehension and summarising involve significantly different skills, to the extent that "teachers should not automatically assume" that a child has comprehension difficulties on the strength of isolated poor performance in a summarising task.

At around the same time, Brown and Day (1983, p1) were pointing out that "the ability to summarise information is an important study skill involving both comprehension of, and attention to, importance at the expense of trivia [and] may be a late developing skill". They studied some of the variables which might be involved, and noted major change in summarising style in the early teens. Even fifth graders (average age 10;7 years) knew how to summarise text by deleting trivial or redundant content, but the skill of superordinate substitution (using "animals" as a replacement for "cats, dogs, and rabbits", say) surges between the seventh grade (average age 13;11 years) and the tenth grade (average age 15;4 years). By contrast, the skill of invention (devising your own topic sentence where none exists in the text) improves far more progressively, but is still used by only one third of 10th graders and only one half of college students (average age 18;1 years).

5.3 Improving Reading Effectiveness in Normal Adults

Having looked at the sort of remedial programmes being used with children, we turn finally to the problem of improving reading in the adult population in general, and in university students in particular. Here is a selection of recent research themes .....

* Advance Organisers: This is where students are presented in advance with summaries of a complex text in the hope that it will enhance their ability to analyse a subsequent larger text. Mayer (1979) has analysed some of the variables which might be at work here, and concludes that the strongest benefits are not on measures of simple knowledge retention, but rather on how usable that knowledge proves to be when transferred to other circumstances. [Student Tip: Teaching programmes typically contain many points of advance organisation, so read them.]

* Initial Mention: This is where a text is structured so as to give an early concise statement of the issue under discussion. Kieras (1980) credits Deese and Kaufman (1957) with an important early demonstration. They had investigated how the shape of the serial position curve varied as the stimulus words approximated more and more towards English prose, and had found that for organised prose the recency advantage disappeared whilst the primacy advantage became more pronounced. Following this lead, Kieras compared the efficiency of "theme-first" and "theme-embedded" passages at getting a certain message across (the former having the main point in sentence one of five - where it should benefit most from the primacy effect - and the latter having it in sentence three of five - where it should benefit from neither primacy nor recency effects). He found that subjects relied strongly on the initially mentioned item when judging the main idea of a passage (its "perceived theme"). With two-topic paragraphs of the same length and complexity, the topic mentioned first is more likely to be judged as the main item. [Student Tip: Reorganise poorly structured material during note-taking, imposing your own structure, so that things are where they really should be when you come to revise.]

* Repeated Reading: Another issue is the effectiveness of repeated reading. Amlund, Kardash, and Kulhavy (1986) have investigated the effects of repetitive reading on test performance in 60 American graduate students, and on an immediate recall test those who had read the material twice out-performed those who had read it only once. However, there was no further improvement in those who had read the material three times.

* Test Anxiety: Other research has looked at the negative effects of studying under excessive anxiety. For example, Calvo and Carreiras (1993) found that anxiety raised word reading times but not levels of comprehension. Specifically, it lowered recall scores for expository text without an inbuilt summary from an average 41.9 (low anxiety) to 35.1 (high anxiety). Calvo and Carreiras suggest that this impairment is due to "worrisome thoughts" (p385) reducing the amount of working memory available for processing the text. [Student Tip: Relax.]

* Elaborative Interrogation: This is where students are encouraged to activate their prior knowledge so that it can help them acquire new facts. Woloshyn, Willoughby, Wood, and Pressley (1990) have used wh- questions to force the necessary knowledge activation. They presented Canadian undergraduates with series of simple six-sentence factual statements, and a variety of follow-up tasks. Where the follow-up task was to answer a wh- question, subjects scored 10.65 (out of 30) on a subsequent fact recall test, compared to 5.70 by control subjects who had merely read the sentences out loud. And the strange thing was that it made little difference whether the wh- question had actually been adequately answered or not - just attempting to answer it delivered the benefit! [Student Tip: Answer all embedded wh- questions - in the majority of cases, the author will have included them for a very good purpose.]

* Other Structural Factors: Taking a slightly different tack, Marshall and Glock (1978-9) looked at the processes of organising ideas into a "hierarchically ordered list of super-propositions" (p23), that is to say, into what Grimes (1975) called a text base. They investigated the effects of varying (a) whether if-then relationships between clauses were explicitly stated or not, (b) whether adjectives were given in the comparative/superlative form or the simple form, (c) whether the main idea was placed at the beginning or end of the text, and (d) whether a designated clause was placed at the beginning or end of a designated sentence within the text. They found major differences in the percentage recall from the same text base between community college and university students (23.4% and 52.6% respectively), and they distinguish the "not-so-fluent" reader - "one who cannot infer the existence of structures in the text base unless these structures are explicitly referenced in the surface structure of the discourse" (p51) - from the "truly fluent" reader - "one who can infer the complete text base of a discourse from incomplete information in the surface structure" (p51). [Student Tip: So don't worry about what an author is saying, worry about what s/he is really getting at.]

* Inconsistencies: This is where the body of an instructional text contains contradictory information. Baker and Anderson (1982) presented American undergraduates with deliberately inconsistent short factual paragraphs followed by a multiple-choice comprehension test, and found that 49% of subjects failed to spot the inconsistency. Those who did spot inconsistencies spent more time re-reading the inconsistent sentence and checking it against the rest of the text, and therefore performed better when tested. [Student Tip: Don't be in that 49%.]

* Interestingness: This is where the body of a text is structured so as to maximise its impact on the heart rather than the head. Hidi and Baird (1986) reviewed earlier arguments by Brewer (eg. 1983) and De Beaugrande (1982) that text has to produce an affective state - suspense, surprise, curiosity, etc - in its readers if it is going to entertain them (albeit it is often difficult to pin down exactly what it is about a given story which makes this happen). In their own work, they found that student recall contained a mix of important (but not interesting) information (40%), interesting (but not important) information (40%), and neither important nor interesting information (20%). Recall of expository prose is therefore best regarded as "multi-determined" (p189), that is to say, it is the result of many influences, none of which is particularly easy to explain even in isolation. [Student Tip: Become as interested in the material as the subject matter will allow.]

* Story Memory: Schank and Abelson (eg. 1995) have warned readers and authors alike that there is no single way to understand a given story, because it can be fitted into any one of several possible story skeletons. They see an author's problem, therefore, as that of managing the sort of skeleton his/her readers eventually select, and this is presumably what good authors are doing when they use the initial mention strategy previously discussed.

* Strategic Reading: This is where the student is taught how to locate the main topic sentences in a text, and then encouraged to reflect upon what s/he is doing and try constantly to do it more effectively. Paris, Lipson, and Wixson (1983) warn, however, that these skills can take time to acquire. Five/six year olds, for example, often display "a surprising naiveté" as to how text is structured. Many have little idea what it is all about, even to the extent of not realising that the things with spaces either side of them are the words they are being told so much about! Paris et al also point out that reading is an excellent example of the difference between declarative knowledge (knowing that) and procedural knowledge (knowing how). It is declarative knowledge, for example, which tells you that most stories do a lot of important scene-setting in the opening paragraph, but it takes procedural knowledge to be able to summarise that paragraph. They then propose the phrase conditional knowledge (knowing when and why) for the additional skills of reading strategically. [Student Tip: Meta is betta.]

* Speed Reading: There are many [check your Internet] commercially available programmes which set out to teach readers to skim text profitably at speeds well in excess of the 365 wpm for light fiction quoted in Table 1. At one extreme, you have the PhotoReading® scheme, which promises "over 25,000" words - roughly the length of this workbook - per minute, and, somewhat more soberly, there is the SuperReading® scheme, which promises to "at least triple" your current speed. The University of Texas Learning Skills Centre is therefore being positively cautious in claiming merely that there is "nothing magic about achieving significant improvement in reading speed".

Tip of Tips: Write some instructional material of your own - it will totally transform the way you look at material written by others!

All of which brings us back to where we started at the beginning of Section 2, when we made the point that reading was far more than a matter of visual perception and eye movement control. We may now go one step further and conclude that it is far more than phonological coding skill as well. It is one mind talking to another across distance, time, or both. As Schank and Abelson put it (1995, p16), "understanding means mapping your stories onto my stories". The problem is that nobody really has much idea of how minds work - merely that they are remarkably complicated. Marshall and Glock (1978-9) remind us, for example, that a single sentence can equate to a whole cluster of cognitive propositions - nine in the following example (p15) .....

The Sentence: "If a single distribution is to be represented, bar graphs are best"

The Equivalent Propositions: 1] Some unspecified person wishes to do something; 2] What he or she wishes to do is represent a proposition; 3] There is only one proposition; 4] Bar graphs are good; 5] Another unspecified graph format is good; 6] Still another unspecified graph format is good; 7] Bar graphs have more goodness than the first unspecified graph format; 8] Bar graphs have more goodness than the second unspecified graph format; 9] Bar graphs are best if and only if a single distribution is to be represented

It so happens that many workers view propositional knowledge as a vast network of interrelated concepts. In the average reader, therefore, this propositional network has to be activated in real time as one sentence is read and then collapsed again as the next sentence is read, while at the same time the mind is also monitoring general theme and a whole host of implicit factors such as irony, metaphor, and implicature. Reading is far from being unravelled, therefore, but we should not underestimate the importance of continuing to study it. Consider .....

"Reading is the means by which the world does a large part of its work [] The slightest improvement either in the page or in the method of reading means a great service to the human race." (Huey, 1908, cited in Muter, 1996.)

 

References

See the Master References List

[Home]