1

John Marshall and the developmental dyslexias

Anne Castles 1,2

Timothy C. Bates 2

and

Max Coltheart 2

1 Department of Psychology, University of Melbourne

2 Macquarie Centre for Cognitive Science, Macquarie University

Address for Correspondence: Anne Castles, Department of Psychology, University of Melbourne, Parkville VIC 3010, AUSTRALIA. Email:

Abstract

The cognitive neuropsychology of reading disorders was born as a presentation by John Marshall and Freda Newcombe to the Engelberg meeting of the International Neuropsychology Symposium in 1971, subsequently published as Marshall and Newcombe (1973). This seminal paper was the first attempt at distinguishing between different types of acquired dyslexia (in this case, deep dyslexia, visual dyslexia and surface dyslexia) and at interpreting all of these acquired dyslexias in relation to a common model of reading: “We wish to emphasize the essential ‘normality’ of the errors characteristic of acquired dyslexia. That is, we shall interpret dyslexic mistakes in terms of a functional analysis of normal reading processes” (Marshall & Newcombe, 1973, p. 188).

The functional analysis of reading Marshall and Newcombe offered was the dual-route model of reading; their Figure 1 was the first diagram of the dual route model to be published.[1] This model diagramsportrays two pathways from print to speech. One pathway operates via the use of grapheme-phoneme correspondence rules (Marshall and Newcombe, 1973, p. 191); this has come to be called the nonlexical route for reading aloud. The other pathway operates via access to a semantic system; this has come to be called the lexical route for reading aloud. The nonlexical route successfully reads nonwords and regular words, but produces regularization errors to words which disobey the grapheme-phoneme correspondence rules of English, i.e., irregular or exception words. The lexical route successfully reads all words (and is indifferent to whether these are regular or irregular words), but cannot produce a correct response to nonwords, since these have no representations in the semantic system.

If brain damage has impaired the ability to read aloud via the lexical route, so that sometimes only the nonlexical route can be used to read aloud, then regularization errors in reading aloud will occur; that is surface dyslexia. Conversely, the complete inability to read nonwords that is seen in deep dyslexia was interpreted by Marshall and Newcombe as due to a completion abolition of the nonlexical route by brain damage.

The otherA second salient characteristic of deep dyslexia is the production of semantic errors in reading aloud. How It was not clear, however, is that this symptom related to could be derived from complete abolition of the nonword reading systemin deep dyslexia.? One possible view is that reading via an intact semantic system will produce semantic errors (that must at least be true for exact synonyms) which could only be avoided if information about correct pronunciation were available from the nonlexical system. An alternative is that there is a second deficit in deep dyslexia, an impairment of the semantic system itself, and semantic errors arise because of this second deficit. In subsequent work (Beauvois and Derouesne, 1979) it was shown that reported data from patients can have with severe impairments of the nonlexical route yet who madke no semantic errors in reading aloud. Thus a new form of acquired dyslexia was defined: phonological dyslexia (nonword reading impaired relative to word reading, in the absence of semantic errors in reading aloud), and the view was then adopted by many that the identifying symptom of deep dyslexia was the occurrence of semantic paralexias. This distinction of phonological dyslexia from deep dyslexia implies that reading via the lexical system is not to be identified with reading via semantics: a distinction which is elaborated below.

If there are different types of acquired dyslexia – different ways in which reading can break down after brain damage – why not different types of developmental dyslexia – different ways in which children can be failing to acquire reading ability? Marshall and his doctoral students Jane Holmes and Christine Temple were exploring this possibility in the 1970s and early 1980s, and they reported cases of developmental surface dyslexia (Holmes, 1973, 1978) and developmental phonological dyslexia (Temple and Marshall, 1983).

Given the occurrence of these developmental equivalents of acquired dyslexias, a bold conjecture may be entertained: that one can use a model of the skilled reading system functional architecture can be used not only to interpret the acquired dyslexias but also to interpret the developmental dyslexias. Marshall did so conjecture:

“Our own approach to this problem has been somewhat extreme. I shall assume (I hope uncontroversially) that the syndromes of developmental dyslexia must be defined over a functional architecture of visible language processing. But I shall further speculate that the relevant functional architecture is the one that correctly characterizes the normal, fluent, adult reading system. The syndromes of developmental dyslexia will accordingly be interpreted as consequent upon the selective failure of a particular adult component (or components) to develop appropriately, with relatively intact, normal (adult) functioning of the remaining components” (Marshall, 1984b, p. 46).

Marshall then proceeded to identify two presuppositions that this approach to the understanding of developmental dyslexia makes:

“This is to adopt a highly modular “preformist” approach to the development of the reading system. The paradigm explicitly denies that developmental failure distorts the architecture of the system as a whole; it also, of course, denies that the normal development of reading skill changes in a “stage-wise” manner whereby the character of reading mechanisms at stage n is qualitatively distinct from their character at stage n+1. “ (Marshall, 1984b, p. 47).

His conclusion from this line of thought was: “To the extent that these presuppositions are true, taxonomies of the acquired dyslexias should map into a one-to-one fashion on to taxonomies of the developmental dyslexias” (Marshall, 1984, p. 47) .

So Marshall then considered whether there exist developmental analogues of six documented types of acquired dyslexia, noting particularly the pioneering work of his PhD student Jane Holmes who had already provided evidence that a developmental form of surface dyslexia can be observed (Holmes, 1973, 1978; see also Coltheart, et al., 1983). His conclusion (Marshall, 1984b, p. 55) from these comparisons of acquired and developmental dyslexias was: “It is somewhat surprising to find that, at least superficially, the study of developmental dyslexia has even now revealed considerable commonality with partitionings derived from the investigation of acquired dyslexia”.

Marshall’s conjecture that static models of adult skilled reading could play a a role in the understanding of learning to read and developmental dyslexia was received with considerable consternation – not to say outright hostility – by some developmental psychologists:

"The already existing structural model, useful as it is in describing the skilled reading process, needs to be complemented by a developmental model in order to make sense of the varieties of developmental dyslexia” (Frith, 1985, p. 326)

"A far greater problem arises when researchers [on children's reading] fail to adopt a developmental perspective when analyzing their data.” (Snowling, 1987, p. 83)

"dual-route models of adult word reading . . . do not provide any adequate account of the development of reading skills in children". (Snowling, Bryant & Hulme, 1996, p. 449)

"we would argue that the idea of trying to formulate a theory of developmental dyslexia in terms of a theory of adult reading is fundamentally misguided (cf. Frith, 1985) . . . A

static model of adult performance, such as dual route theory, is inadequate for understanding how children learn to read and why some children learn to read easily while others have difficulties". (Snowling, Bryant & Hulme, 1996, p. 444)

"Castles & Coltheart (1993) . . . argued for three interrelated ideas: that individual differences in reading among dyslexic children are best interpreted in the context of a "dual-route" model of adult reading, that it is useful to draw analogies between developmental and acquired dyslexia, and that there are two types of difficulty in developmental dyslexia; some children have difficulty learning to recognise words 'by sight" while others have difficulty learning to decode or 'sound out” words. The purpose of this paper is to argue that the first two suggestions are seriously misguided and that Castles & Coltheart do not provide any convincing evidence for the final suggestion” (Snowling, Bryant & Hulme, 1996, p. 443).

Marshall offered his conjecture more than twenty years ago. How has it fared since then?

Subtypes of developmental dyslexia

Given the dual-route model of reading proposed by Marshall and Newcombe (1973), and the proposals by Marshall (1984b) concerning the applicability of static models of adult skilled to the analysis of patterns of developmental dyslexia, it follows that children learning to read may have varying degrees of difficulty acquiring one route or the other. They may have more difficulty acquiring the lexical route than the nonlexical route, resulting in better reading of nonwords than irregular words (“developmental surface dyslexia”). Alternatively, they may have more difficulty acquiring the nonlexical route than the lexical route, resulting in better nonword reading than irregular word reading. (“developmental phonological dyslexia”).

Dissociations of this type in the developmental dyslexic population were precisely what Castles and Coltheart (1993) set out to identify. Already, some promising evidence was available in the form of case studies (e.g., Campbell & Butterworth, 1985; Coltheart, Masterson, Byng, Prior & Riddoch, 1983; Goulandris & Snowling, 1991; Hanley, Hastie & Kay, 1992; Snowling & Hulme, 1989; Temple & Marshall, 1983), and from group studies based on somewhat different theoretical approaches (Boder, 1973; Mitterer, 1982). Castles and Coltheart sought to consolidate these findings in a group study of irregular and nonword reading in 53 dyslexics and 56 age-matched controls. They found that approximately one third of the poor readers were in the normal range for their age on one task but below the 5th percentile on the other. Specifically, eight subjects were identified as pure developmental phonological dyslexics: their nonword reading was poor, compared with chronological age-matched controls, but their irregular word reading was within normal range. Another ten subjects were classified as pure developmental surface dyslexics: their irregular word reading was poor but their nonword reading fell within normal range. A further 27 subjects were poor on both tasks, but nevertheless showed a significant discrepancy between their scores on the irregular word and nonword tasks. The remaining subjects were equally poor on both tasks, leading Castles and Coltheart to conclude that, while deficits on the two reading tasks generally co-occur in children with reading difficulties, irregular word reading and nonword reading can be developmentally dissociated – that is, acquisition of the lexical reading route and acquisition of the nonlexical reading route can be developmentally dissociated, just as would be expected from the approach to developmental dyslexia proposed by Marshall (1994b).

Pure and impure cases of developmental dyslexia

Suppose a child of 10 achieves a reading accuracy with irregular words that is equivalent to the mean accuracy of a sample of 7-year-old normal readers. This child’s ability to use the lexical route for reading is thus well below an age-appropriate level; and that is our definition of developmental surface dyslexia.

Suppose a child of 10 achieves a reading accuracy with nonwords that is equivalent to the mean accuracy of a sample of 7-year-old normal readers. This child’s ability to use the nonlexical route for reading is thus well below an age-appropriate level; and that is our definition of developmental phonological dyslexia.

And now suppose the child referred to two paragraphs above is the same child who is referred to one paragraph above. What are we to say about this child? That he has both developmental surface dyslexia and developmental phonological dyslexia? We contend that this is the only sensible position to adopt.

But of course the distinction between, on the one hand, a child whose irregular word reading is well below chronological age whilst nonword reading as at an age-appropriate level, and, on the other hand, a child whose irregular word reading and nonword reading are both below age-appropriate level, is an important one. So we need terminology to express this distinction. The terminology we will use is “pure” versus “impure”. By “pure developmental surface dyslexia”, we mean that all aspects of reading except use of the lexical route are at age-appropriate levels; and by “pure developmental phonological dyslexia”, we mean that all aspects of reading except use of the nonlexical route are at age-appropriate levels. If some cases of developmental surface or phonological dyslexia are pure in this sense, that would imply the important conclusion that it is possible for each either of the reading routes to be acquired at a normal rate even when the other route was being acquired abnormally slowly.

As mentioned above, Castles and Coltheart (1993) did report cases of pure surface dyslexia and cases of pure phonological dyslexia as defined above. We report here new data germane to this issue.s

We collected data [if we are thinking of this paper as an original report of the data we will need to give much more detail here. If instead the details are fully reported in some other paper, we need to refer to that paper, but even so need a little more detail here]. on irregular and nonword reading accuracy from two large samples. The first sample consisted of 2081 developing readers, ranging in age from 6;5 to 15;9 years, selected without reference to their reading ability. The second sample consisted of 1024 adolescents and young adults, ranging in age from 12 to 21, again selected without reference to their reading ability. Accuracy of reading aloud the 30 irregular words and 30 nonwords from Coltheart and Leahy (1996) was measured in the first sample; the older readers in the second sample were tested on reading aloud 40 irregular words and 40 nonwords from an extension of this test for adults that we have recently developed (see Bates et al, 2004). For both samples, we converted individual irregular and nonword reading scores to z-scores, using age specific means. We then looked for participants whose irregular word reading was very impaired (more than 1.64 standard deviations below average for their age) but whose nonword reading was in the normal range or better (no worse than half a standard deviation below the mean for their age). Such cases would constitute examples of pure surface dyslexia.

The cases identified from the developing reader sample are presented in Table 1. There were 16 children, ranging in age from 7 to 15, who were extremely impaired in irregular word reading for their age but who performed normally or better at nonword reading. Cases identified from the older reader sample are presented in Table 2. As can be seen, from this sample we identified 13 adolescents and adults with completely normal nonword reading but severely impaired irregular word reading.

Insert Tables 1 and 2 here

Cases of pure surface dyslexia clearly occur relatively infrequently (why this might be we discuss below). However, the rarity of these cases in our sample should not be over-exaggerated, because the two samples of readers we examined were not pre-diagnosed dyslexic readers, but were children, adolescents and adults selected without reference to their reading ability. Of the 2081 children in the first sample, 244 (11.7%) could post hoc be classified as having a reading impairment on the basis that they performed more than 1.64 standard deviations below average on either irregular word reading or nonword reading or both. Of these, 16 children, or 6.6%, showed the pure surface dyslexic pattern described above. Of the 1024 older participants (the second sample), 104 (10.2%) could subsequently be classified as having a reading impairment on the basis that they performed more than 1.64 standard deviations below average on either irregular word reading or nonword reading or both. Of these, 13 participants, or 12.5%, showed the pure surface dyslexic pattern described above. We note that Castles and Coltheart (1993) reported that six of their 53 dyslexics, or 11.3%, showed a similar pure surface dyslexic pattern. In our view, a proportion of 6.6-12.5% of all dyslexic readers showing a pure surface dyslexic pattern, in both children and adult samples, cannot be described as negligible, indeed, they are drawn from an approximately normal distribution of reading variance, and occur with the frequencies to be expected from the correlation of the underlying types of reading performance..

We next interrogated the databases described above to investigate pure developmental phonological dyslexia: could we find instances of dyslexia where children or adults had severe nonword reading impairments but normal irregular word reading scores? Again, we defined a severe deficit as being more than 1.64 standard deviations below average for age, and normal performance as no more than half a standard deviation below average for age. The cases which meeting met these criteria from the developing reader and older reader samples are presented in Tables 3 and 4 respectively.