The phonological hypothesis of developmental dyslexia

35
TITLE PAGE Title The Phonological Hypothesis of Dyslexia: A Constructive Critique Authors Fernando Leal & Judith Suro Universidad de Guadalajara Guadalajara, Mexico Addresses for correspondence : Home Home Fernando Leal Judith Suro Paseo de los Robles 4169 San Felipe 823 45110 Zapopan, Jal. 44100 Guadalajara, Jal. Mexico Mexico Work Work Fernando Leal Judith Suro Universidad de Guadalajara, CUCSH Universidad de Guadalajara, DEEDUC Liceo 210, Planta Baja Hidalgo 935 44100 Guadalajara, Jal. 44100 Guadalajara, Jal. Mexico Mexico E-mail E-mail [email protected] [email protected]

Transcript of The phonological hypothesis of developmental dyslexia

TITLE PAGE

Title

The Phonological Hypothesis of Dyslexia: A Constructive Critique

Authors

Fernando Leal & Judith Suro

Universidad de Guadalajara

Guadalajara, Mexico

Addresses for correspondence:

Home Home

Fernando Leal Judith Suro

Paseo de los Robles 4169 San Felipe 823

45110 Zapopan, Jal. 44100 Guadalajara, Jal.

Mexico Mexico

Work Work

Fernando Leal Judith Suro

Universidad de Guadalajara, CUCSH Universidad de Guadalajara, DEEDUC

Liceo 210, Planta Baja Hidalgo 935

44100 Guadalajara, Jal. 44100 Guadalajara, Jal.

Mexico Mexico

E-mail E-mail

[email protected] [email protected]

ABSTRACT PAGE

Developmental dyslexia is currently hypothesized to rest on an impairment of phonological

processing. Under such a hypothesis, the main diagnostic tool involves phonemes as subjects fail

properly to associate them with their corresponding graphemes in reading. Although some

phonological phenomena of either a subphonemic or supraphonemic character have also been

identified by researchers, there does not seem to be a sustained attempt at dealing with the full

range of phonological phenomena. This is all the more urgent as phonemes have long ceased to

be of any theoretical interest to mainstream linguistic phonology. We present the preliminary

results of a phonological case study of the reading performance of two Spanish-speaking children

diagnosed with dyslexia, in which the emphasis is laid upon stress patterns and intonational

contours. The results suggest that errors at the phonemic level may be as low as 3% in a text,

whereas suprasegmental errors may range from 25% (incorrectly stressed single words) to 93%

(intonationally incorrect sentences). Should further research confirm these results, then we would

have to revise the content of the phonological hypothesis of developmental dyslexia and develop

new and more realistic tools of assessment and treatment on the basis of current phonological

theory.

NB: Submitted to Annals of Dyslexia in 2002. Rejected for publication on the grounds that it is

not a statistical study. Original research done in 2001 and presented at the International Dyslexia

Association Annual Meeting, Atlanta, November 2001.

Phonological Hypothesis of Dyslexia 1

The Phonological Hypothesis of Dyslexia: A Constructive Critique

The phonological hypothesis of developmental dyslexia has become the standard explanation of

that disorder (Pennington 1999, Lundberg et al. 1999). According to the phonological

hypothesis, developmental dyslexia is a specific impairment affecting phonological processing

(perhaps even a domain-specific ‘phonological module’). Such an impairment, it is thought,

results from a deficit in phonological awareness. The phrase ‘phonological awareness’, again, is

widely understood as a trouble in the ability to segment into phonemes that underlies the

decoding of alphabetic writing systems. The typical way of assessing ‘phonological awareness’

involves the reading of single words out of context. When analysis of errors in this task yields an

abnormal number of omissions and substitutions of letters in words (and sometimes in

nonwords), a diagnosis of developmental dyslexia seems appropriate. Correspondingly, a

favored treatment largely consists in stimulating ‘phonological awareness’ through tasks of

segmentation.

We have problems with all aspects of this consensus:

• The usual analysis of errors, based on omissions and substitutions, lacks linguistic depth.

Not all omissions or substitutions are equally important or significant. And besides, hosts

of other errors are made by dyslexics, some of which might be much more telling.

• Working with isolated words, or nonwords, does not look like a very good method.1

Reading is after all not about isolated words, and not even about isolated sentences; it is

Phonological Hypothesis of Dyslexia 2

about texts. Of course, reading instruction begins with words, or even with letters, but

research cannot be guided by this criterion.

• The emphasis on phonemic segmentation is probably quite misleading. First of all,

phonemes have no good linguistic credentials (see section 1 below). Second, reading is

not about segmenting, but about assembling.

• Dyslexia may be a phonological disorder, but probably not a result of any deficit of

awareness. Learning how to read is acquiring an automatic skill, i.e. a skill that by

definition lacks awareness. So, dyslexics if anything suffer from an excess rather than a

deficit of awareness.

• Alternative treatments other than via segmentation tasks may have some promise.

Although we may thus seem to be at cross purposes with a wide consensus, yet at the same time

we think that something like the phonological hypothesis is our best current theoretical bet.2 In

other words, the purpose of our critique is constructive: We don’t want to throw the baby with

the bathwater; we’d rather like to keep the beautiful baby and just get rid of the dirty water. The

phonological hypothesis can be substantially improved, and this paper is about ways of doing so.

But before we go on, it has to be recognized that the above characterization of the phonological

hypothesis may seem unfair to some readers, for the following reasons:

(1) Some researchers have pointed out that segmentation is not the only trouble in

developmental dyslexia. Other troubles have been identified, having to do with

alliteration and rhyming, identification and counting of syllables, classification of onsets

Phonological Hypothesis of Dyslexia 3

and rhymes within syllables, and even syllabification rules (Pennington 1999, Lundberg

et al. 1999, Harris et al. 1999).

(2) The phonological hypothesis can be, and has been, applied to other, nonalphabetic,

writing systems (Leong & Joshi 1997, Leong et al. 2000).

(3) There are ways of assessing phonological awareness that use utterances larger than words

(Joshi 1999). The same is true of some proposed treatments. In fact, there is a great

variety of diagnostic tools and paedagogic proposals in the field of developmental

dyslexia.

(4) Some researchers have enlarged the analysis of errors in the reading of words to include

syllabic phenomena (Harris et al. 1999, Goswami 2002). This point is obviously related

to (1).

We are aware of all this. The description given in the first paragraph does not convey the

subtleties of the picture one can find in the research literature. But if our description is vastly

simplified with respect to research, such a simplification has a vicious grip on the minds of both

researchers and – more importantly – practitioners. We want to break that grip by suggesting that

researchers should consider very carefully what the exact and proper content of the phonological

hypothesis might be. Is it just a matter of phonemes? And if it is, is it a matter of segmentation?

Or does dyslexia perhaps concern something different from, or at least something in addition to,

phonemes, e.g. syllables – or maybe even some other phonological phenomena that we haven’t

yet thought of? And again, is dyslexia really a matter of deficit of awareness, however we might

characterize that awareness?

Phonological Hypothesis of Dyslexia 4

The paper will proceed as follows. First (section 1), we’ll give some personal background on the

origin of our concerns and present a very brief sketch of the kinds of phenomena that

phonological science is really about. This will allow us to introduce the idea of a hierarchy of

phonological hypotheses that have to compete among them for the explanation of dyslexia. We

shall then (section 2) present a case study of the reading performance of two children with

dyslexia. The results of a careful phonological analysis will suggest that the kinds of error that

are usually considered in the literature as crucial probably constitute just the tip of the iceberg in

developmental dyslexia. A guiding thought will be that segmentation is the wrong way to tackle

the analysis of reading performance. Finally (section 3), we’ll discuss the results obtained in the

case study and make some hopefully interesting suggestions both for a revision of the

phonological hypothesis of dyslexia and for the practice of speech therapy and the teaching of

reading.3

1. Some personal background and a sketch of phonology

This paper is the fruit of a collaboration between a special educator with over 20 years of

experience with learning and reading disabled children (JS) and a linguist with a special interest

in the neurosciences and particulary in the light that different disabilities might throw on our

understanding of language and the brain (FL). When FL first heard of the phonological

hypothesis, his initial interest gave way to some disappointment when he saw that most

researchers and practitioners kept talking about phonemes. For phonemes have long ceased to be

of any theoretical interest to phonology. About half a century ago, phonemes started to die of a

natural death, giving way to the study of distinctive features and phonotactic rules – the features

Phonological Hypothesis of Dyslexia 5

that distinguish sounds from each other both in a given language and in all languages and the

rules that govern the position and ordering of sounds in syllables, morphemes, words and

perhaps even phrases (Jakobson et al. 1952, Chomsky & Halle 1968). This development was

quickly followed by a renewed and succesful interest in suprasegmentals – syllable structure,

stress patterns, tonal distribution, intonational contours (for textbook treatment, see e.g.

Goldsmith 1996, Spencer 1996). The interest in phonemes thus seemed pretty weird in the eyes

of FL. So he conceived recruiting the help of JS in trying to find out how children with dyslexia

really speak and read. A complicated research program was launched, whose first step was (a) to

produce a full phonological analysis of dyslexics’ reading performance and (b) more generally of

their speech, as well as (c) to devise a teaching program for children with reading disabilities.

These three projects run parallel to each other. In section 2 we shall present some results of the

first project in the shape of a case study with two children. Our subjects are Spanish-speaking

monolinguals, but the lessons of this paper are universal, and thus relevant to any language.

Contemporary phonology is quite a complicated discipline, so it would be a hopeless task to try

to summarize it in a few paragraphs. But one way to see its real import is by comparison with the

layperson’s picture:

• In the layperson’s picture, the word in the ordinary, nontechnical sense (especially the

lexical word or lexeme) is king. In real contemporary phonology, the word in that sense

is just one level among others.

• In the layperson’s picture, the word is composed of phonemes; in fact the word is nothing

but a succession of phonemes; it is phonologically realized by placing one phoneme after

Phonological Hypothesis of Dyslexia 6

the other in a linear chain. In real contemporary phonology, linearity is completely

abandoned.

• In the layperson’s picture, phonemes are solid, atomic (indivisible) entities; one phoneme

is perfectly different from another, and they don’t have parts. In real contemporary

phonology, there are no ‘phonemes’ – the sound segments represented in alphabetic

writing systems are really bundles of separate sound properties (the so-called distinctive

features, such as Voiced vs. Voiceless or Nasal vs. Non-nasal) that are not even always

realized in the same way.

• In the layperson’s picture, phonemes are there to distinguish between the meanings of

words; they can do that because they are solid; they are either there or not there. In real

contemporary phonology, the phonological representation of a word depends on other

things than sound segments; and these are far less than solid (thus, sometimes the absence

of a sound segment is also meaningful, e.g. we know dog is Singular because it does not

have an s).

If segments are not really important for phonology, what is? Simplifying the picture a little bit,

we would say that phonology is really about one kind of entity (the distinctive features) and three

kinds of rule:

• Phonotactic rules, that concern the relative positions of distinctive features within the

syllable, the morpheme and the lexeme (e.g. the feature Palatal cannot precede the feature

Consonantal at the beginning of a word or syllable in English and many other languages).

Phonological Hypothesis of Dyslexia 7

Among the most important of these rules are syllabification rules, that concern the

structure and composition of possible syllables.

• Stress pattern rules, that concern the distribution of stress and stress differentials along

the utterance.4 An utterance is a succession of syllables, and stress highlights one syllable

every now and then; this highlighting gives an utterance a characteristic rhythm (Hayes

1995). A word in the ordinary sense is not the only linguistic bearer of stress (linguistic

items both below and above the word, such as syllables and phrases can be stressed).

Examples are given in sections 2.1, 2.3 and 2.4 below.

• Intonational rules, that concern the assignment of melodic contours to phonological units

of various sizes. A word in the ordinary sense cannot be the bearer of an intonational

contour. See section 2.2 below.

These rules are language-specific, i.e. they differ from language to language and allow linguists

to divide languages into phonological types. Syllabification – and possibly other phonotactic

rules – may be impaired in children with dyslexia, as a recent study has shown (Harris et al.

1999). But no research seems to have been undertaken to find out whether stress and intonation

are also abnormal in the speech of those children. The case study to be presented later on will

concern itself mainly with those two suprasegmental phenomena, as they manifest itself in

children’s reading performance.

A consideration of the real scope of contemporary phonology raises the question whether the

explanation afforded by the phonological hypothesis should be limited to phonemic

segmentation. Maybe in developmental dyslexia other aspects of phonological processing, such

Phonological Hypothesis of Dyslexia 8

as stress and intonation, are also impaired – or maybe even other, nonphonological aspects of

language. A recent autobiographical paper (Simon 2000) suggests that dyslexics have

unsurmountable difficulties in learning a second language also at the level of syntax and lexical

semantics, although the author surmises that her own syntactic and semantic difficulties are

really grounded in phonological impairments. We think that further research to probe all

possibilities is called for. For that purpose at least three different hypotheses should be

considered:5

• H1 – what we may call ‘the weak phonological hypothesis’ – according to which only the

phonemic aspects of phonological processing are affected in developmental dyslexia.

• H2 – what we may call ‘the strong phonological hypothesis’ – according to which all

phonological aspects of phonological processing (i.e. phonotactic rules, stress pattern

rules, and intonational rules) are affected.

• H3 – what we may call ‘the full linguistic hypothesis’ – according to which all aspects of

language processing (phonological, lexical, syntactic) are affected.

As far as we can judge, most researchers and practitioners work within H1, as though it was the

only hypothesis worth considering. A few people – those interested in syllabic phenomena –

have started to look beyond H1, and are thus moving within H2. The present paper is an extension

of that research effort, as we try to look beyond syllabic phenomena.

2. A Case Study

Phonological Hypothesis of Dyslexia 9

Since we wanted to explore the reading performance of children with dyslexia, we chose a

simple story, ‘The Vain Magpie’, which is a fable in prose that contains one individual character

(the eponymous vain magpie) interacting in dialogue with two collective ones (her fellow

magpies and the peacocks). The version we used contains both the story and an explicit moral;

the number of words is 188. Although we are also working on the reading performance of texts

that are phonologically more demanding (because they contain more difficult words, rhymes and

word playing), ‘The Vain Magpie’, which is our only concern here, is relatively free of any

phonological ‘traps’.

We chose two children (one male, one female) of about the same age (11½ years old) that had

been previously diagnosed with dyslexia. Although neither of them was doing very well in

school, we expected the boy (S) to be a better reader than the girl (P), given that he had enjoyed

more intensive speech therapy and reading stimulation. For comparison’s sake we also had two

controls, one a nondyslexic female child of about the same age (M) and one proficient adult (J).

Please see Table I for details. Performance of the four readers was recorded, transcribed and

carefully analyzed at two levels, the integration of stress units and the integration of intonational

items.

[Place Table I about here]

A word on the concept of ‘integration’ is called for. We chose it deliberately to contrast our kind

of analysis with the more usual one based on ‘segmentation’. As we suggested before,

researchers are always looking for missing or displaced or added letters (sounds), and so they

Phonological Hypothesis of Dyslexia 10

tend to forget that sounds belong together – that sounds have to be welded together to make up

words or phrases that have the right stress and the right intonation. In a sense, ‘segmentation’ is

an egregious misnomer for what happens when a dyslexic is trying to read – and fails. Although

some dyslexics may have some temporary trouble identifying the letters, the real trouble is rather

putting them together. This is a problem in phonological processing – and the mark of good

phonological processing is not segmenting (dividing) but integrating (putting together). Such

integration happens at all levels, from the simple lexical word through the complex one, through

larger and larger units.

2.1 Integration of Stress Units: Single Words

Some words are stressed and some are unstressed.6 All unstressed words belong to the so-called

‘function’ words, e.g. the, a, to, of, more, all, my, me, it, not, can, would, here, and, or, but, if.

Any of these small words can receive a special emphatic stress in certain contexts, e.g. Tell me

what yóu think, This is móre than I deserve, She could nót take it anymore. If you contrast the

last sentence with Shé could not take it anymore, you will see the pragmatic effect of stress. On

the other hand, a few ‘function’ words that have more than one syllable are also stressed, e.g.

ónly, upón, álmost, álways, itsélf, whenéver, sómebody, hítherto. As opposed to ‘function’

words, so-called ‘content’ words are regularly assigned a particular stress, e.g. sún, régular,

supérior, respéct, órdinary. Indeed, meaning contrasts can be achieved through stress, as when

we distinguish between the noun cóntent and the adjective contént. If we approach ‘The Vain

Magpie’ with these distinctions in mind, the distribution of all three kinds of words (stressed

‘content’, stressed ‘function’ and unstressed ‘function’ words) is as shown in Table II.

Phonological Hypothesis of Dyslexia 11

[Place Table II about here]

So how many stressed words do subjects and controls fail to integrate properly? The figures are

in Table III. (Notice that unstressed ‘function’ words are not counted; this is because stress

integration in their case involves their attachment to other words. See section 2.4 below.) The

difference of performance between subjects and controls is massive. Subjects’ total performance

is similarly poor, although an interesting difference with respect to stressed ‘function’ words is

apparent. Notice that we are not counting mistakes at the level of single segments, although

omissions, additions, substitutions, etc. of letters are partly (although not exclusively)

responsible for failed stress integration. To make possible a comparison with the usual method,

we counted the segments (phonemes) in ‘The Vain Magpie’; they are 836. Subject P makes any

one of the usual mistakes (omissions, additions, etc.) in 22 of those 836 and Subject S in 24 of

them. That would add up to a phonemic-level error rate of between 2.5 and 3%. Compare these

percentages with the 25-30% error rates at stressed-word level in Table III. The reader may begin

to see what we mean when we say that researchers and practitioners have been missing

something by relying too much on phonemes.

[Place Table III about here]

And this is not all. For we’re not just talking about the sheer amount of errors, impressive as it

might be. The quality of the errors is also important. It is also a matter of the effect those errors

have in the integration of units larger than words. But in order to understand all that, we have to

Phonological Hypothesis of Dyslexia 12

analyze those items that are of greatest interest from the linguistic point of view, viz. the

sentences that comprise a text; for words as such cannot constitute a linguistic message except in

very special cases. The reader should try to imagine a situation in which a single word (such as

dog) can convey a message. Such situations do exist, but they are exceedingly rare. That is why

we propose to start with full sentences and then proceed slowly down through the clauses, the

phrases until we reach the words.

2.2 Integration of Intonational Items

So let’s start with sentences as the complete intonational units. Sentences can be simple

(consisting of just one clause, e.g. Once upon a time there was a king) or complex (consisting of

two or more clauses, that are either coordinated with each other, e.g. Once upon a time there was

a king and he had a beautiful daughter, or subordinated to each other, e.g. Once upon a time

there was a king who had a beautiful daughter). The basic intonational unit is the clause, but a

complex sentence is only uttered appropriately when all its clauses have the right intonation.

Thus take a conditional sentence like If you don’t do your homework, there won’t be a party. The

antecedent clause of the conditional sentence (If you don’t do your homework) has a

characteristic rising intonation, whereas the consequent clause (there won’t be a party) has a

falling melodic contour. It is like a small musical ‘theme’: the rising intonation creates a

‘tension’ that the falling intonation ‘resolves’. The union of the two melodies makes a musical

‘whole’ – a full intonational sentence.

Phonological Hypothesis of Dyslexia 13

‘The Vain Magpie’ contains 14 full sentences, some simple, some complex. If we first consider

how many sentences our subjects and controls ‘sing’ correctly, the figures are displayed in the

first row of Table IV. The reader can readily appreciate how small a fraction (only 7%) of full

sentences our subjects are capable of getting right. Even a normal child of the same age – our

control M – could read correctly less than half of the sentences in a simple text like this. In

contrast, our adult control succeeds about 80% of the time. (We may note in passing that the

intonationally functional sentences which both our subjects managed to read were simple

sentences.)

[Place Table IV about here]

However, not everything is intonationally lost when the sentence as such is missed. Something

within it can always be salvaged, either its clauses (he had a beautiful daughter) or even smaller

units (a beautiful daughter), for those can also have a melody of their own. We followed thus a

three-step evaluation procedure:

• How many full sentences were intonationally correct? (Table IV, Row 1)

• Discounting the sentences that were right in full, how many clauses were intonationally

correct within the remaining incorrect sentences? (Table IV, Row 2)

• Again, discounting the clauses that were right, how many smaller units were

intonationally correct within the remaining incorrect clauses? (Table IV, Row 3)

Phonological Hypothesis of Dyslexia 14

Thus our subjects had missed 13 out of 14 sentences, yet they managed to correctly pronounce 2

and 4 clauses respectively within the remaining 13 sentences. The number of clauses left in each

case is different, because the full sentence each one read correctly was not the same. We can see

that Subject S gets ahead of Subject P. In contrast, our controls manage to rescue half or even

more of the remaining clauses. As for units smaller than a clause, both our subjects had left some

50 phrases capable of some melodic contour, out of which they salvaged only about 10%. In

contrast, our child control M rescued over 40% and our adult control J read all remaining phrases

correctly. Figure 1 adds all three levels of melodic performance into a cumulative record of

successes. This is a good place to insist on a point that has struck us again and again when

listening to the tapes: although dyslexic children make a lot of mistakes, they rarely seem to see

that something’s wrong; so they don’t even try to correct it. A good reader – like our controls,

even if the adult was naturally more proficient – is not so much a person who doesn’t make

mistakes (everyone makes a mistake now and then), but a person who notices and rectifies the

mistakes they make. We think this is an important lesson for practitioners.

[Place Figure 1 about here]

2.3 Integration of Stress Units: Single Words (Revisited)

The lesson can be immediately applied to the results of stress integration of single words in

Table II. The figures shown there do not really tell us anything about how readers are trying to

read. A qualitative analysis shows us a more interesting picture:

Phonological Hypothesis of Dyslexia 15

A. Sometimes the word is correctly assembled after a false start (e.g. praw… spraw…

sprawling).

B. Sometimes the word is correctly assembled by giving oneself time through lengthening of

one or two sounds (e.g. sssssprawling).

C. Sometimes the word fails to be properly assembled, but a succession of syllables is

offered instead (e.g. spraw-ling).

D. Sometimes the failure is complete (e.g. sss-sprrr… rawl…sss…).7

Only Case D is a failure proper, whereas the others can be considered strategies to deal with a

perceived problem. When we compare our four readers, we obtain the pattern shown in Table V.

These figures cannot be added together nor can percentages be taken of them, because the

strategies and failures can happen in the same word. The use of Table V is just a comparative

one. It is clear that the controls are perfectly capable of repairing all their mistakes in reading,

whereas the subjects can do so only about one third of the time. There is also a difference

between them in how well they cope, or try to cope, with the remaining two thirds: S is

obviously more methodical and successful than P.

[Place Table V about here]

2.4 Integration of Stress Units: Beyond Single Words

Intonation (melody) as it were rides on top of stress patterns (rhythm). The enormous failures

you can appreciate in intonational integration in our two subjects (Table IV, Figure 1) depend

Phonological Hypothesis of Dyslexia 16

heavily on a previous failure in stress or rhythmic integration. So the next step is to look at the

integration of stress units larger than the stressed words we have been talking about so far (Table

II). Here there is a problem. Integration means first of all the phonological assembling of stressed

and unstressed words into larger stress units. Imagine that words in an utterance are like a series

of little circles, some black (stressed), some grey (unstressed):

The black circles (stressed words) can as it were attract some grey circles (unstressed words) so

as to get bigger blocks, or such bigger blocks emerge from the combination of several grey

circles:

The bigger blocks – stress units larger than ordinary words – are called by linguists

‘phonological words’. The layperson can get a sense for ‘phonological words’ by thinking of

examples such as Macy’s, we’ll, it’s, could’ve, I’d’ve, to’ve, on to’em, for’t, cannot, today,

eastward, gonna (from going to), cuppa (from cup of), helluva (from hell of a), etc.8 These

contracted forms are the product of stress integration. Now stress integration is far from being a

mechanical thing: there often is more than one solution to the integration problem. The vagaries

of pragmatics and emphasis yield different solutions. Thus, if we use a vertical stroke (|) to

indicate the limit between one phonological word and the next, the sentence I was in the garden

Phonological Hypothesis of Dyslexia 17

could be ‘parsed’ in four different ways, according to different pragmatic emphases. In writing

we may use italics to suggest the appropriate pattern:

(1) I was | in the garden I was in the garden (It is true, not false that I was)

(2) I | was | in the garden I was in the garden (It was me, not somebody else)

(3) I was | in | the garden I was in the garden (Not, say, outside)

(4) I was | in the | garden I was in the garden (It was that particular garden)

As opposed to these four correct patterns, there are others that are incorrect (e.g. I | was in the |

garden or even I was | in the gar|den). Because of the freedom in stress assignation, it is not

possible to say how many stress units a given text has, in advance and with full precision. So

evaluation here is less clearly governed by strict rules. The procedure we suggest instead is to

make a rough estimation of the amount of possible ‘phonological words’ in the text and then to

use that estimate as a reference against which to compare the stress units one can recognize in

the reading performance of the child. In ‘The Vain Magpie’ there are about 116 possible stress

units for the 188 words of the story. We can use this figure to compare the rhythmic integration

in our four readers, as displayed in Table VI. Here you can appreciate that, although the child

control does not yet master intonational contours (melody), she is quite good at stress patterns

(rhythm). In contrast, our two subjects are quite poor in both, albeit P performs – as we predicted

– worse than S.

[Place Table VI about here]

Phonological Hypothesis of Dyslexia 18

3. Discussion and Suggestions for Further Work

For discussion, it might be useful to be able to compare all results obtained in our case study. In

Table VII they are displayed together with reading speed, about which a brief word is apposite.

Reading speed is a well-known component of current assessment of dyslexia (Selikowitz 1998).

Besides, it has been observed that pre-literate dyslexic children have a distinctly slower tempo of

speech (see Scarborough 1990). It stands to reason that a certain tempo in reading is crucial for

comprehension: both too slow and too quick reading will impair our understanding of the text

read. That’s why we wanted to incorporate this variable into our analysis.

[Place Table VII about here]

We compared the reading of ‘The Vain Magpie’ with a similar story (also a fable, 205 words)

and with two short poems (a fable in verse, 71 words; a lyric poem, 55 words). Both poems are

naturally more difficult to read from a phonological point of view (because of the rhyming). The

lyric poem is particularly difficult, because it contains a lot of word playing. Figure 2 displays

the results for all four texts as read by subjects and controls. You can see that Subject P’s

performance is highly sensitive to the phonological difficulty of the text, whereas the other three

readers are relatively stable. The normal reading speed (at least in Spanish) seems to be about 2.5

words per second (± 0.2), as per our adult control. So we take that speed as 100% successful and

represent in Table VII the figures for ‘The Vain Magpie’.

[Place Figure 2 about here]

Phonological Hypothesis of Dyslexia 19

With the exception of Table III and our brief discussion of phonemic level mistakes, all our

earlier tables were framed in terms of successes, as is Table VII. This way of presenting results

emphasizes how little the difference between dyslexics and nondyslexics really is when we just

consider omissions, substitutions, etc., as is usual in both research and practice. Compare those

figures with all the other measures and you will see what we mean when we suggest that, by only

looking for missed phonemes, we may be just dealing with the tip of a huge iceberg!

We don’t pretend to have proved beyond all reasonable doubt that dyslexia is not just a matter of

phonemes. Our case study does not aspire to be representative of anything. But its suggestive

value should be clear. Consider this. The preferred level addressed in the usual psychological

tests (words and nonwords) is related to our row 7, which contains isolated stressed words that

were correctly read. The usual way of presenting the facts in neuropsychology and cognitive

psychology is by pointing to the errors made by dyslexics. If those errors are individually

counted (as missed phonemes in row 1) but we invert the score and talk instead of the segments

they get right, we see that the difference is there, but is less than massive: our two subjects could

after all manage to capture phonemes in 97% of the cases – not so different from the 100% of

controls. Stressed words (row 7) are a better indicator: subjects get them right in about 70% of

the items present in the story (30% of errors), whereas our two controls did so for practically

100% of the words (2-3% of errors). Even then the literature does not say much about the way

dyslexics try to solve the trouble they have in getting those words right (row 8). But there is

more: if we now compare the differences we get when comparing other phonological measures,

then it becomes clear that we may be missing the iceberg beneath the tip of ‘reading errors’:

Phonological Hypothesis of Dyslexia 20

dyslexics are getting the melody right only between 20 and 30% of the time as opposed to

between 85 and 100% (rows 3, 4 and 5). And they are capturing the right stress patterns of units

larger than ordinary words also only between 19 and 35% of the time as against 95-100% (row

6)!

We conclude that extended phonological analysis of reading performance of dyslexic children

suggests that the concepts and methods usual in developmental dyslexia research and practice are

not satisfactory. The only phonological phenomenon all researchers and practitioners share is the

stressed ‘content’ word (for even nonwords do not differ phonologically from words). Yet we

have seen that such items are just a tiny part of the phonology of reading. It may thus be

rewarding to pay more attention to the full panoply of phonological aspects of reading

performance (phonotactic rules, syllable structure, stress patterns, intonational contours, reading

speed). Maybe the time has come to revise the phonological hypothesis of developmental

dyslexia by means of more realistic data and with the help of an updated phonological theory.

Likewise, we think that speech pathology, speech therapy and more generally the teaching of

reading might start to experiment with phonological exercises that go beyond phonemic

awareness. Our observations on the careful monitoring of intonation and stress with reading

disabled children is most encouraging. We hope to be able to report in the future some work in

progress we are doing with children affected with dyslexia and other reading impairments.

Phonological Hypothesis of Dyslexia 21

FOOTNOTES

1 Linguistically, one could even say that research is limited not to words in general, but to

stressed words, and not to stressed words in general, but to ‘content’ words, and not even to

‘content’ words in general, but to nouns. Such limitations are a mere artefact of the methods

used; they may make assessment easier, but they cannot be justified from a theoretical point of

view. Language is not a collection of nouns.

2 We are perfectly aware that the phonological hypothesis is not the only game in town. There

are proposals having to do with rapid naming, symbol imagery, reading comprehension,

discourse coherence, visual and motor impairments, and whatnot. Such hypotheses are not as

such inimical to the phonological hypothesis, but could eventually be combined with it. In any

case, they are not the subject of this paper.

3 We initially wanted to say something about the concept of ‘awareness’, too, but we decided it

would complicate the central message of this paper (the importance of supraphonemic

phenomena in reading performance), so we left that issue for some other time.

4 Instead of or in addition to stress pattern rules, some languages have tone pattern rules, which

govern the distribution of tones and tone differentials along phonological units of various sizes.

Most languages in the world are tone languages, although neither English nor Spanish are among

them. The relationship between tone and stress goes far beyond the scope of this paper, as are

Phonological Hypothesis of Dyslexia 22

other phonological phenomena (such as speed or tempo, loudness, chromatism, and sound

symbolism) that are less well known and studied.

5 Given the complexity of phonological and linguistic phenomena, H2 and H3 are actually not

single hypotheses but sets of possible hypotheses. Thus, an individual may have impaired

processing of stress but not of intonation, or the other way round; and again, some people may

have lexical or syntactic problems at different levels, in addition to phonological ones – a whole

hierarchy of hypotheses which research will have to sort out very carefully.

6 In the layperson’s use of ‘word’. It may be useful to remind the reader that the concept of word

does not have a unitary meaning across all linguistic levels. Thus the phrase house of cards

contains three syntactic words (house, of, cards) but only two phonological words (house of,

cards) and just one lexical word (the phrase as a whole being a single semantic unit). The

layperson’s concept of word most closely agrees with the syntactician’s, which makes the idea of

a ‘phonological word’ (see section 2.4 below) somewhat less than obvious.

7 In our sample complete failure is always associated with syllables containing either a final

consonant or two initial or final consonants.

8 The attentive reader will notice that many ‘function’ words (such as always or somebody)

emerge historically from such phonological contraction processes.

Phonological Hypothesis of Dyslexia 23

REFERENCES

Ardila, A. (1998). Semantic paralexias in Spanish language. Aphasiology, 12, 885-900.

Ardila, A. (Forthcoming). Características de las alexias y las agrafias en español. In E. Matute &

F. Leal (Eds.), Introducción al estudio multidisciplinario del español. México: Fondo de

Cultura Económica.

Chomsky, N., & Halle, M. (1968). The sound patterns of English. Cambridge, MA: The MIT

Press.

Goldsmith, J. (Ed.). (1995). The handbook of phonological theory. Oxford: Blackwell.

Goswami, U. (2002). Phonology, reading development, and dyslexia: A cross-linguistic

perspective. Annals of Dyslexia, 52, 141-63.

Harris, J., Watson, J., & Bates, S. (1999). Prosody and melody in vowel disorder. Journal of

Linguistics, 35, 489-525.

Hayes, B. (1995). Metrical stress theory: Principles and case studies. Chicago: The University

of Chicago Press.

Jakobson, R., Fant, G., & Halle, M. (1952). Preliminaries to speech analysis. Technical Report #

13, M.I.T. Acoustics Laboratory.

Joshi, R.M. (1999). A diagnostic procedure based on reading component model. In Lundberg,

Tønnessen, & Austad, (Eds.), (1999), 207-19.

Leong, C.K., & Joshi, R.M. (Eds.). (1997). Cross-language studies of learning to read and spell:

Phonologic and orthographic processing. Dordrecht: Kluwer.

Leong, C.K., Cheng, P.-W., & Lam, C.C.C. (2000). Exploring reading-spelling connection as

locus of dyslexia in Chinese. Annals of Dyslexia, 50, 239-59.

Phonological Hypothesis of Dyslexia 24

Lundberg, I., Tønnessen, F.E., & Austad, I. (Eds.). (1999). Dyslexia: Advances in theory and

practice. Dordrecht: Kluwer.

Pennington, B.F. (1999). Dyslexia as a neurodevelopmental disorder. In H. Tager-Flusberg,

(Ed.), Neurodevelopmental disorders (pp. 307-30). Cambridge, MA: The MIT Press.

Scarborough, H.S. (1990). Very early language deficits in dyslexic children. Child Development,

61, 1728-43.

Selikowitz, M. (1998). Dyslexia and other learning difficulties: The facts. 2nd edn. New York:

Oxford University Press.

Simon, C.S. (2000). Dyslexia and learning a foreign language: A personal experience. Annals of

Dyslexia, 50, 155-87.

Spencer, A. (1996). Phonology: Theory and description. Oxford: Blackwell.

Phonological Hypothesis of Dyslexia 25

Table I. Subjects and Controls of Case Study

Who Sex Age Grade IQ Sociability Verbal Ability Memory

Subject P F 11:5 4th Normal Normal Normal Poor

Subject S M 11:6 4th Normal High High High Normal

Control M F 11:6 6th Normal High High Normal

Control J F 45 M.A. Normal High Normal Normal

Phonological Hypothesis of Dyslexia 26

Table II. Distribution of Stress in the Words of ‘The Vain Magpie’

Type of Word Stressed Unstressed Sum

‘Content’ Word 76 0 76

‘Function’ Word 37 75 112

Sum 113 75 188

Phonological Hypothesis of Dyslexia 27

Table III. Errors in Stress Integration at Word Level

Stressed Words Subject P Subject S Control M Control J

‘Content’ Words 23 (30.26%) 25 (32.89%) 2 (2.63%) 1 (1.32%)

‘Function’ Words 11 (29.73%) 3 (8.11%) 2 (5.41%) 1 (2.70%)

All 34 (30.09%) 28 (24.78%) 4 (3.54%) 2 (1.77%)

Phonological Hypothesis of Dyslexia 28

Table IV. Number of items having an appropriate intonational contour

Subject P Subject S Control M Control J

Full Sentences 1 in 14 (7%) 1 in 14 (7%) 6 in 14 (42.8%) 11 in 14 (78.5%)

Clauses within

Sentences 2 in 28 (7%) 4 in 27 (14.8%) 10 in 20 (50%) 5 in 8 (62.5%)

Units smaller than

Clauses 4 in 50 (8%) 5 in 50 (10%) 10 in 23 (43.5%) 7 in 7 (100%)

Phonological Hypothesis of Dyslexia 29

Table V. Strategies and failures in getting the right stress pattern in single words

Subject P Subject S Control M Control J

A. False Start and Repair 11 10 2 2

B. Strategic Lentgthening 2 13 – –

C. Syllabification 5 11 – –

D. Failure 11 1 – –

Phonological Hypothesis of Dyslexia 30

Table VI. Stress integration of units larger than words

Subject P Subject S Control M Control J

Number of recognizable stress units

(‘phonological words’) 22 40 110 116

Percentage of succeses (out of our

estimate of 116 stress units) 19% 35% 95% 100%

Phonological Hypothesis of Dyslexia 31

Table VII. Summary of Rates of Success in Reading Performance (Rounded Figures)

Subject P Subject S Control M Control J

Phonemic

Level

1. Intact Sound

Segments 97% 97% 100% 100%

Tempo 2. Reading Speed 35% 46% 74% 100%

3. Sentences 7% 7% 43% 79%

4. Clauses 14% 21% 72% 92% Intonation

(Melody) 5. Phrases 20% 29% 84% 100%

6. Phrases 19% 35% 95% 100%

7. Words 70% 75% 97% 98% Stress

(Rhythm) 8. Stress Correction Very Poor Poor Good Good

Phonological Hypothesis of Dyslexia 32

Figure 1. Cumulative Rate of Success in Intonational Integration

Phonological Hypothesis of Dyslexia 33

Figure 2. Reading Speed (Words per Second) in Four Different Texts