KATHOLIEKE UNIVERSITEIT LEUVEN
Faculteit Psychologie en Pedagogische Wetenschappen
Centrum voor Orthopedagogiek
& Laboratorium voor Experimentele ORL
EARLY LITERACY DEVELOPMENT IN CHILDREN
AT RISK FOR DYSLEXIA
A longitudinal study of the general magnocellular theory
Proefschrift aangeboden tot het
verkrijgen van de graad van
Doctor in de Psychologie
door Bart Boets
o.l.v. Prof. Dr. Pol Ghesquière
Prof. Dr. Jan Wouters
2006
Bart Boets, Early literacy development in children at risk for dyslexia: A longitudinal study of the general magnocellular theory. Dissertation submitted to obtain the degree of Doctor in the Psychological Sciences, March 2006. Supervision: Prof. Dr. Pol Ghesquière and Prof. Dr. Jan Wouters
Developmental dyslexia is a hereditary neurological disorder that affects about 5 to 10% of children and
adults. It is characterized by severe reading and spelling difficulties that are persistent and resistant to usual teaching methods and remedial efforts. Although the predominant aetiological view postulates that dyslexia results from a phonological deficit, recent research demonstrated an additional deficit in low-level auditory and visual temporal processing, which might be causal to the observed phonological and literacy problems. In the auditory modality, it has been hypothesised that the temporal processing deficit interferes with the accurate detection of rapid acoustical changes in speech, and subsequently disrupts the development of adequate phonological representations and later reading and spelling skills. Alternatively, it has been suggested that an analogous problem in the visual magnocellular subsystem interferes with the development of orthographic skills and thus also complicates literacy development. Although the auditory and the visual research line were originally developed independently, they both converge in postulating that dyslexics suffer from a deficiency in temporal information processing. As a consequence, the hypothesis of a general temporal processing deficit was formulated and an attempt was undertaken to relate the hypothesised cross-modal deficits to one underlying biological cause. Accordingly, the ‘general magnocellular theory’ postulates that the neural magnocellular deficit is not restricted to the visual pathway but can be generalized to all modalities. Furthermore, the theory assumes an explicit causal relation between the auditory and visual temporal processing skills on the one hand and the development of phonological, orthographic and subsequent reading and spelling skills on the other.
The aim of the present dissertation was to evaluate the validity of the general magnocellular theory by means of a longitudinal prospective study. We studied a group of preschool children at high family risk for dyslexia and compared them with a group of well-matched low-risk control children. Both groups were followed up through early literacy development. In a series of studies we evaluated different parts of the hypotheses that constitute the general magnocellular theory. In the first study (manuscript 1) we tested whether preschool children at family risk of dyslexia presented a deficit in auditory temporal processing and we examined the relation between auditory and phonological processing. The results indicated that preschool children at family risk for dyslexia presented a significant deficit in letter knowledge and phonological awareness, but not in any of the administered auditory measures. However, tone-in-noise detection and in particular frequency modulation detection were significantly related to phonological awareness. In the second study (manuscript 2), we further explored the assumptions of the auditory temporal processing deficit hypothesis by examining speech perception in the same groups of preschoolers, and relating these measures to the auditory and phonological data. For categorical perception, both groups performed identically on the identification task, but the risk group presented a marginally significant deficit on the discrimination task. On the speech-in-noise task, the children at family risk for dyslexia presented a slight but significant deficit, particularly in the most difficult listening conditions. Furthermore, speech parameters were significantly related to phonological awareness and low-level auditory processing. In a next study (manuscript 3), we focused upon sensory processing in the visual modality. To assess visual magnocellular processing, we tested sensitivity to coherent motion in random dot kinematograms. Although no significant differences were observed between the high versus low risk group, sensitivity to coherent motion was significantly related to emerging orthographic skills. Integration of the visual data with the results obtained in manuscript 1 indicated an exclusive relation between dynamic visual processing and orthographic skills on the one hand, and dynamic auditory processing and phonological skills on the other.
While in the first three manuscripts preschool data were analysed comparing the high risk versus low risk group, in study 1 of manuscript 4 we described the crucial retrospective analysis comparing children developing literacy problems versus children developing normal literacy skills. Therefore, children were reclassified based upon family risk status and first grade literacy achievement. A reanalysis of the data demonstrated that children showing both the increased family risk and literacy problems at the end of first grade presented a significant preschool deficit in letter knowledge, phonological awareness, rapid automatic naming, frequency modulation detection, coherent motion detection, categorical perception and speech-in-noise perception. This indicated that the sensory deficit preceded the literacy problem, leaving open the possibility of a causal influence of sensory processing upon literacy development. In a final study (manuscript 4, study 2), the plausibility of this causal relation was evaluated using path analysis. In particular, we demonstrated that dynamic auditory processing was related to speech perception which on its turn was related to phonological awareness. Similarly, dynamic visual processing was related to orthographic ability. Eventually, phonological awareness and orthographic ability – together with verbal short-term memory - were unique predictors of reading and writing development.
In the last chapter we concluded that our study provided substantial evidence for the validity of the general magnocellular theory, although a more fine-grained inspection of the data (particularly at the level of the individual subjects) also revealed some conflicting findings.
Bart Boets, De ontwikkeling van lees- en spellingsvaardigheden bij jonge kinderen met risico op dyslexie: Een longitudinale studie van de algemene magnocellulaire theorie. Proefschrift aangeboden tot het verkrijgen van de graad van doctor in de Psychologie, maart 2006. Promotoren: Prof. Dr. Pol Ghesquière en Prof. Dr. Jan Wouters
Dyslexie is een erfelijk bepaalde neurologische aandoening waarvan de prevalentie wordt geraamd op 5
à 10% van de bevolking. De leerstoornis wordt gekenmerkt door ernstige lees- en spellingsproblemen die moeilijk te verhelpen zijn door middel van de gebruikelijke didactische maatregelen en remediëringsinspanningen. Hoewel algemeen wordt aangenomen dat dyslexie veroorzaakt wordt door een tekort in de fonologische vaardigheid, wijst recent onderzoek ook op basale tekorten in auditieve en visuele temporele informatieverwerking die mogelijkerwijs aan de basis liggen van de fonologische en literaire problemen. Binnen de auditieve modaliteit wordt verondersteld dat het temporele informatieverwerkingstekort interfereert met de accurate detectie van snelle akoestische wisselingen in het spraaksignaal. Door dit subtiele spraakperceptieprobleem wordt de ontwikkeling van het fonologische systeem en de verdere lees- en spellingsvaardigheid bemoeilijkt. Daarnaast wordt verondersteld dat een analoog probleem in de visuele modaliteit, meer bepaald in het magnocellulaire subsysteem, interfereert met de ontwikkeling van orthografische vaardigheden, waardoor eveneens de ontwikkeling van de literaire vaardigheid wordt bedreigd. Hoewel de auditieve en de visuele onderzoekslijn aanvankelijk onafhankelijk ontwikkeld werden, komen ze beiden tot de conclusie dat personen met dyslexie moeilijkheden hebben met temporele informatieverwerking. Als een gevolg werd dan ook de hypothese van een algemeen informatieverwerkingstekort geformuleerd en werd een poging ondernomen om deze veronderstelde cross-modale tekorten te herleiden tot één onderliggende biologische oorzaak. Dienovereenkomstig postuleert de general magnocellular theory dat dyslexie het gevolg is van een magnocellulaire disfunctie die zich niet enkel manifesteert binnen het visuele systeem, maar alle sensorische modaliteiten betreft. Voorts veronderstelt de theorie een expliciet causaal verband tussen enerzijds de basale temporele informatieverwerkingsvaardigheden en anderzijds de ontwikkeling van fonologische, orthografische en lees- en spellingsvaardigheden.
Voorliggend doctoraatsonderzoek beoogt een evaluatie van de algemene magnocellulaire theorie aan de hand van een prospectieve longitudinale studie. Hiertoe werd een groep kleuters met familiaal risico op dyslexie gedurende twee jaar opgevolgd (vanaf begin derde kleuterklas tot eind eerste leerjaar) en vergeleken met een nauwkeurig gematchte controlegroep. Systematisch werden de verschillende deelhypothesen van de theorie onderzocht. In de eerste studie (manuscript 1) onderzoeken we of kleuters met verhoogd familiaal risico een tekort in auditieve temporele informatieverwerkingsvaardigheden vertonen en exploreren we de relatie tussen deze vaardigheden en fonologische bedrevenheid. De resultaten geven aan dat de risicokleuters tekorten vertonen in letterkennis en fonologisch bewustzijn, maar niet in auditieve vaardigheden. Nochtans wordt er wel een significant verband geobserveerd tussen fonologisch bewustzijn enerzijds en toon-in-ruis detectie en frequentiemodulatie detectie anderzijds. In de volgende studie (manuscript 2) onderwerpen we de kleuters aan enkele spraakperceptieproeven en worden deze resultaten gerelateerd aan de auditieve en fonologische data. Voor categorale perceptie presteert de risicogroep (rand-)significant zwakker op de discriminatietaak, hoewel we geen groepsverschillen observeren op de identificatietaak. Ook voor spraakverstaan in ruis presteert de risicogroep significant zwakker dan de controlegroep, en dit vooral in de moeilijkste luistercondities. Verder zijn de spraakparameters significant gerelateerd aan fonologisch bewustzijn en aan de basale auditieve vaardigheden. In een volgende studie (manuscript 3) richten we ons op informatieverwerking in de visuele modaliteit. Om de kwaliteit van visueel magnocellulair functioneren te bepalen, wordt de gevoeligheid voor coherente bewegingsdetectie gemeten. Opnieuw worden er geen significante groepsverschillen geobserveerd, maar de gevoeligheid voor coherente bewegingsdetectie houdt wel verband met ontluikende orthografische vaardigheden. Integratie van de visuele data met de resultaten verkregen in manuscript 1, suggereert een exclusief verband tussen enerzijds de bedrevenheid in dynamische visuele informatieverwerking en orthografische vaardigheden en anderzijds dynamische auditieve informatieverwerking en fonologische bekwaamheid.
In de eerste drie manuscripten worden de voorschoolse resultaten van de risicogroep telkens vergeleken met deze van de controlegroep. In manuscript 4 (studie 1) worden de kinderen opnieuw in groepen ingedeeld op basis van het behaalde lees- en spellingsniveau aan het einde van het eerste leerjaar in combinatie met het familiale risico. Vervolgens worden de kleuterdata retrospectief hergeanalyseerd. Deze heranalyse toont aan dat kinderen met een uitermate verhoogd risico op dyslexie (o.b.v. familiaal risico én lees- en spellingsachterstand aan het einde van het eerste leerjaar) op voorschoolse leeftijd al ernstige tekorten vertonen in letterkennis, fonologisch bewustzijn, snel serieel benoemen, frequentiemodulatie detectie, coherente bewegingsdetectie, categorale perceptie en spraakverstaan in ruis. Bovendien tonen deze resultaten duidelijk aan dat het perceptuele probleem voorafgaat aan het lees- en spellingsprobleem, waardoor het dus inderdaad een potentiële causale invloed zou kunnen uitoefenen. In een laatste studie (manuscript 4, studie 2) wordt de plausibiliteit van dergelijke causale relatie geëxploreerd aan de hand van causale padanalyse. Hierbij wordt aangetoond dat dynamische auditieve informatieverwerking verband houdt met spraakperceptie, wat op zijn beurt gerelateerd is
aan fonologisch bewustzijn. Op gelijkaardige wijze blijkt dynamische visuele informatieverwerking in verband te staan met orthografische vaardigheden. Ten slotte blijken fonologisch bewustzijn en orthografische bekwaamheid, samen met verbaal korte-termijn geheugen, unieke predictoren te zijn voor de ontwikkeling van lees- en spellingsvaardigheden.
In het laatste hoofdstuk concluderen we dat ons onderzoek in het algemeen substantiële evidentie biedt voor de validiteit van de algemene magnocellulaire theorie. Nochtans levert een meer diepgaande analyse van de data (in het bijzonder op individueel subjectniveau) toch ook heel wat bevindingen op die tegenstrijdig zijn met de theorie.
Dankwoord
Hier ligt het dan, mijn proefschrift – het resultaat van ruim vier jaar gedreven
wetenschappelijk werk. Het was een hele klus waarbij ik gelukkig de hulp kreeg van talrijke
mensen. Ik wil hen dan ook van harte danken.
In de eerste plaats dank ik mijn promotoren Pol Ghesquière en Jan Wouters voor hun
deskundige begeleiding. Pol, ik besef dat een belangrijk deel van je opdracht bestond uit het
intomen van mijn steeds verder uitdijende onderzoeksdrang. Ik dank je dan ook voor het
vertrouwen en de vrijheid die je me bood om dit onderzoeksproject op persoonlijke wijze uit
te werken zonder de haalbaarheid ervan uit het oog te verliezen. Jan, het was een plezier om
met een gepassioneerde wetenschapper als jij samen te werken. Het aanstekelijke
enthousiasme waarmee je mijn onderzoeksresultaten telkens onthaalde, werkte bijzonder
stimulerend en inspirerend. Hartelijk dank!
Daarnaast wil ik nog een bijzonder woord van dank uitspreken voor Erik Vandenbussche die
ook aan de basis van dit onderzoek stond, maar onverwachts en veel te vroeg overleed. Ik
hoop dat hij trots zou zijn op dit proefschrift dat mede gedragen werd door zijn fascinatie voor
de neurowetenschappen.
Astrid van Wieringen, Karl Verfaillie, Patrick Onghena en Peter de Jong wil ik danken voor
hun deelname aan mijn projectcommissie. Hun kritische opmerkingen en suggesties waren
erg waardevol. I am also very grateful and honoured that Prof. Stuart Rosen, Prof. Wied
Ruijssenaars and Prof. Karl Verfaillie accepted to be members of the jury.
Verder dank ik Astrid van Wieringen voor haar gewaardeerde hulp bij de uitbouw van de
auditieve onderzoekslijn; Johan Laneau en Tom Francart voor hun hulp bij de constructie van
de auditieve stimuli en het auditieve testplatform; Alexander Nowinski voor het
programmeren van de visuele stimuli; Willy Leung voor de creatie van de tekenfilmpjes die
het mogelijk maakten om zelfs kleuters te enthousiasmeren voor psychofysica; Mieke van
Ingelghem voor de talloze overlegmomenten en discussies; Peter Stiers voor zijn advies
omtrent de uitbouw van de visuele onderzoekslijn; Hugo Boets en Wivine Decoster voor het
lenen van hun mooie stem om de filmfragmentjes en fonologische taakjes in te spreken; Geert
Verbeke en Patrick Onghena voor hun methodologisch en statistisch advies; Els Gadeyne
voor haar tips omtrent Structural Equation Modeling; Peter de Jong en Femke Scheltinga voor
het beschikbaar stellen van fonologisch testmateriaal; het Dutch Dyslexia Research team voor
het beschikbaar stellen van hun spraakstimuli; Caroline Van Eccelpoel, Kathelijne Jordens,
Ellen Boon, Sara Vanderhallen, Sandra Schoonbrood, Leen Van Rie en Dorien Schrouwen
voor hun assistentie bij de dataverzameling, het segmenteren en chronometreren van
eindeloze spraakfragmenten en de uitwerking van de pseudohomofoontaak.
Ik wil ook uitdrukkelijk mijn dank betuigen aan de vele kinderen die deelnamen aan het
onderzoek. Ik ben nog steeds verwonderd dat ze bereid waren om dagenlang (!) naar vreemde
piepjes te luisteren of naar hypnotiserende stipjes op een scherm te staren. Ik dank ook hun
ouders en de betrokken scholen en directies.
Verder wil ik mijn collega’s van het ‘vijfde verdiep’ danken voor de hartelijke werksfeer. In
het bijzonder dank ik Bert De Smedt voor de talloze kleine en grote adviezen en informele
overlegmomenten, Leen Cleuren voor het tekstueel nalezen van mijn proefschrift en Joris Van
Puyenbroeck voor de reanimatie van mijn gecrashte en doodverklaarde hard-disk met alle
onderzoeksgegevens.
Ook nog een heel bijzondere dankjewel voor Lieve, niet enkel omdat ze me daadwerkelijk
heeft geholpen bij het ontwerp van de filmscenario’s en het programmeren van de
spraakexperimenten, maar vooral voor haar onmisbare steun als lieve levensgezel.
Tot slot dank ik mijn dochtertje Kato voor haar stralende ochtendlach waarmee ze me telkens
de energie gaf om de dag met hernieuwde moed en inspiratie te starten (na alweer een veel te
korte nacht).
Bart Boets, maart 2006
The greater becomes the volume of our sphere of knowledge,
the greater also becomes its surface of contact with the unknown.
J. Sageret
Dit proefschrift werd mede mogelijk gemaakt dankzij een beurs bij het Fonds voor
Wetenschappelijk Onderzoek – Vlaanderen. Een deel van de apparatuur werd bovendien
gesponsord door het Koningin Fabiola Fonds.
This doctoral dissertation was funded by a grant of the Fund for Scientific Research –
Flanders. Part of the technical equipment was funded by the Queen Fabiola Fund.
Table of contents
General introduction
1
Manuscript 1 Auditory temporal information processing in preschool children at
family risk for dyslexia: Relations with phonological abilities and
developing literacy skills
13
Manuscript 2 Speech perception in preschoolers at family risk for dyslexia:
Relations with low-level auditory processing and phonological
ability
51
Manuscript 3 Coherent motion detection in preschool children at family risk for
dyslexia
87
Manuscript 4 The development of early literacy skills among children at high
risk for dyslexia: An empirical evaluation of the general
magnocellular theory
107
Study 1 Auditory, visual, speech perception and phonological deficits in
preschool children at high risk for dyslexia
112
Study 2 Modeling relations between sensory processing, speech
perception, orthographic and phonological ability and literacy
achievement
130
General conclusions and future perspectives 161
1
General introduction ___________________________________________________________________________
At the end of kindergarten, every preschooler looks forward to learning to unravel the
mystery of reading and writing. Whereas the majority of children fairly easily assimilates
these skills in primary school, a small group of about 5 to 10 % of children is confronted with
severe reading and spelling difficulties that are persistent and resistant to the usual teaching
methods and remedial efforts. These reading and spelling impairments, characteristic for
developmental dyslexia, might have far-reaching academic, occupational and psycho-social
consequences. Historically, there has been a longstanding discussion about the aetiology of
these ‘unexpected’ literacy problems. The origin has been sought in the visual, the auditory as
well as in the cognitive-linguistic domain. Recently, a new comprehensive theory has been
put forward, integrating several aspects of the visual, the auditory and the cognitive-linguistic
research tradition. The present doctoral research project aims at empirically evaluating this
general magnocellular theory of dyslexia.
Theoretical framework
At present, the predominant aetiological view postulates that dyslexia results from a
cognitive deficit that is specific to the representation and processing of speech sounds: the
phonological deficit theory (Rack, 1994; Snowling, 2000). However, during the last decades
there has been a growing number of studies demonstrating a deficit in low-level auditory and
visual processing in subjects with dyslexia and it has been suggested that these sensory
deficits might be causal to both the observed phonological and literacy problems (Farmer &
Klein, 1995).
In the auditory modality, most emphasis has been given to temporal auditory
processing. It has been demonstrated that dyslexics have problems processing short, rapidly
presented and dynamic changing acoustic stimuli (e.g. Talcott & Witton, 2002; Tallal, 1980;
Van Ingelghem et al., 2005). Besides, dyslexics also tend to present subtle speech perception
2
problems (McBride-Chang, 1995). Consequently, it has been hypothesised that the basic
deficit in perceiving auditory temporal cues causes a problem for the accurate detection of the
acoustical changes in speech. Accordingly, the speech perception problem causes a cascade of
effects, starting with the disruption of normal development of the phonological system and
resulting in problems learning to read and spell (Talcott & Witton, 2002; Tallal, 1980; Wright
et al., 1997).
Alternatively, it has been suggested that the literacy problems of some dyslexics can
be traced back to a specific visual problem, in particular a problem in magnocellular visual
processing (Stein & Walsh, 1997). Anatomically as well as functionally, the visual system can
be divided in two independent but linked subsystems: the magnocellular (or transient)
pathway and the parvocellular (or sustained) pathway. The magnocellular system is highly
sensitive to low spatial and high temporal frequency stimulation and it responds preferentially
to the onset and offset of the stimulus. Consequently, it is predominantly a flicker or motion
detecting system. In psychophysical as well as neurophysiologic studies, it has been
demonstrated that individuals with dyslexia show a decreased sensitivity to stimuli within the
magnocellular range (e.g. Cornelissen et al., 1995; Eden et al., 1996; Demb, Boynton &
Heeger, 1997; Lovegrove, 1996). This evidence for a magnocellular dysfunction has further
been confirmed by anatomical studies showing abnormalities of the magnocellular layers of
the lateral geniculate nucleus (LGN) in the dyslexic brain (Livingstone, Rosen, Drislane &
Galaburda, 1991). Regarding the specific mechanism by which magnocellular visual
dysfunction may limit normal literacy development, there is still much speculation. The
answer probably lies in the anatomical connections from the magnocellular pathway to the
posterior parietal cortex (PPC). The PPC is known to be involved in normal eye movement
control, visuospatial attention, visual search and peripheral vision – all factors that are
obviously involved in the development of orthographic skills and subsequent reading and
spelling skills (Stein & Walsh, 1997; Stein & Talcott, 1999; Stein, 2001). Since the PPC is
dominated by magnocellular input, it is very well possible that slight impairments in
magnocellular functioning might multiply up to greater deficits in PPC functioning.
Although historically the auditory and the visual research lines have been developed
independently, they both converge in postulating that dyslexics suffer from a deficiency in
temporal information processing. As a consequence, the hypothesis of a general temporal
processing deficit or pan-modal deficit is formulated (Eden, Stein, & Wood, 1995; Farmer &
Klein, 1995; Frith & Frith, 1996; Stein & Talcott, 1999). In this hypothesis the reading and
spelling problems of dyslexic people are attributed to a deficiency in the temporal processing
3
capacity of both the auditory and the visual system. In this way, the theory suggests an
explicit causal relation between the auditory and visual temporal processing skills on the one
hand, and the development of phonological, orthographic and subsequent reading and spelling
skills on the other. Recently, an attempt has also been undertaken to relate the hypothesised
cross-modal deficiencies to one underlying biological cause. The general magnocellular
theory (Stein & Walsh, 1997; Stein, 2001) postulates that the magnocellular deficit is not
restricted to the visual pathway but could be generalized to all modalities (visual and auditory
as well as tactile). Although a similar magnocellular/parvocellular distinction has not been
described for the auditory brain areas, preliminary anatomical evidence for the auditory
variant of a magnocellular deficit has been found by Galaburda and colleagues (Galaburda,
Menard & Rosen, 1994), reporting an increased proportion of smaller neurons in the medial
geniculate nucleus of the dyslexic brain.
The general magnocellular theory, unique in its ability to account for most of the
manifestations of dyslexia, is undoubtedly attractive. Nevertheless, it also has its problems,
with a major one being its lack of empirical validation. Indeed, many studies that have been
reported as corroborating evidence for the theory, actually present such methodological
shortcomings that they are not able to offer substantial empirical validity.
For instance, in the most rigorous version, the general magnocellular theory postulates
that subjects with dyslexia show deficits across all the aforementioned skills. However,
probably as a consequence of the separate and different research traditions, most studies only
investigated a subgroup of skills (e.g. only auditory processing, only visual processing, only
speech perception or only phonological ability), and almost none studied the whole spectrum
of deficits in the same group of subjects. As a consequence, these studies are simply not able
to demonstrate the postulated cross-modal deficit. Noteworthy exceptions are studies by
Cestnick (2001), Van Ingelghem et al. (2001) and Witton et al. (1998), reporting the co-
occurrence of auditory and visual temporal problems in certain individuals with dyslexia (but
see Ramus et al., 2003, and White et al., in press, for some strong counterevidence).
Further, most of these studies only reported group means, hence giving the impression
that deficits were more or less uniform in the dyslexic group (i.e. that all subjects were
affected). However, a more detailed inspection of individual results demonstrated that in fact
only a rather small proportion of dyslexics presented these sensory deficits, even in the
presence of an overall group difference (Ramus et al., 2003; Rosen, 2003). Of course, this
4
observation makes it questionable whether the sensory problem would be at the basis of the
literacy problem.
Finally, almost all studies have been investigating adult or school-aged subjects; only
very few studied preschoolers. As a consequence, despite its inherent developmental aspect,
the general magnocellular theory has almost exclusively been built upon cross-sectional data
collected in adult and school-aged subjects. Unfortunately however, these data do not provide
any indication regarding the cause of the observed deficits or regarding the nature of the
studied interrelations. Indeed, based on adult and school-aged studies, the possibility cannot
be ruled out that the observed sensory deficits are the result rather than the cause of
differences in literacy ability and reading experience.
Aim of the doctoral research project
This doctoral research project aims to evaluate the empirical validity of the general
magnocellular theory. As aforementioned, the general magnocellular theory comprises an
integration of the auditory temporal processing deficit hypothesis and the visual
magnocellular theory. Besides, it also incorporates the phonological deficit theory, since it
does not deny the presence of phonological problems, but rather considers them as secondary
to a more basic auditory deficit.
In order to investigate the whole spectrum of postulated deficits, we administered a
broad battery of tests assessing low-level auditory and visual processing, speech perception,
phonological ability, orthographic skills, and reading and spelling skills1. Table 1 presents an
overview of all tests that have been applied. By measuring all these abilities in the same
subjects and by analysing the data at an individual level, we are able to verify whether
subjects with dyslexia present a consistent cross-modal pattern of deficiencies.
Moreover, in order to evaluate the hypothesised developmental and causal aspect of
the theory, we studied preschool children who not yet received any formal reading instruction,
and followed them up through early literacy development. In particular, we studied a group of
preschool children at high family risk for dyslexia and compared them with a group of well-
matched low-risk control children. The use of this longitudinal prospective design allows us
to verify whether the postulated sensory deficits precede the literacy problem, and whether
they demonstrate a predictive relation with later reading and writing skills.
5
Table 1
Overview of the tests administered in the present study.
Preschool measures First grade measures
Non-verbal intelligence
Phonological processing
Phonological Awareness
First sound identity
End sound identity
Rhyme identity
Rhyme fluency
Rapid Automatic Naming
Colour naming
Picture naming
Verbal short-term memory
Digit span forward
Nonword repetition test
Letter knowledge
Productive letter knowledge
Receptive letter knowledge
Auditory processing
Gap-in-noise detection
Frequency modulation detection
Tone-in-noise detection
Coherent motion detection
Speech perception
Speech-in-noise perception
Categorical perception (identification)
Categorical perception (discrimination)
Reading
One-minute real-word reading test
Pseudoword reading test
Real-word reading accuracy
Real-word reading speed
Pseudoword reading accuracy
Pseudoword reading speed
Spelling
Pseudohomophone task
ADHD Questionnaire
Outline of the doctoral thesis
The doctoral thesis consists of four manuscripts – two of them have been published,
one is currently under revision, and the last one will soon be submitted as two separate papers.
Throughout a series of psychophysical studies we evaluate different aspects of the general
magnocellular theory. Since all studies investigate the same group of preschoolers using
6
partly the same test instruments, this necessarily implies some overlap between the
manuscripts, especially with respect to the method sections.
In the first manuscript (Boets, Wouters, van Wieringen, & Ghesquière, 2006a) we
describe the procedure that we applied to select a group of preschool children at high family
risk for dyslexia, and the way this group was matched to a control group. Next, we tested
whether preschool children at family risk for dyslexia present a deficit in auditory temporal
processing and we examined the relation between auditory and phonological processing.
Auditory processing was assessed by means of three psychophysical threshold tests: gap-in-
noise detection, 2 Hz FM-detection and tone-in-noise detection. Phonological processing was
assessed by a broad test battery comprising tasks for rapid automatic naming, verbal short-
term memory and phonological awareness. The results of this study indicated that preschool
children at family risk for dyslexia present a significant deficit in letter knowledge and
phonological awareness, but not in any of the administered auditory measures. However,
tone-in-noise detection and in particular frequency modulation detection were significantly
correlated with phonological awareness, suggesting a relation between low-level auditory
processing and phonological processing.
In the second manuscript (Boets, Ghesquière, van Wieringen, & Wouters, under
revision), we further explore the assumptions of the auditory temporal processing deficit
hypothesis by examining speech perception in the same groups of preschoolers, and relating
these measures to the auditory and phonological data. For categorical perception of a place-
of-articulation speech continuum, both groups performed identically on the identification task,
but the risk group presented a marginally significant deficit on the discrimination task. On the
speech-in-noise task, the children at family risk for dyslexia presented a slight but significant
deficit, particularly in the most difficult listening conditions. Furthermore, speech parameters
were significantly related to phonological awareness and low-level auditory processing.
In a next study (manuscript 3; Boets, Wouters, van Wieringen, & Ghesquière, 2006b),
we focus upon sensory processing in the visual modality. To assess visual magnocellular
processing, we tested sensitivity to coherent motion in random dot kinematograms. Although
no significant differences were observed between the high versus low risk group, sensitivity
to coherent motion was significantly related to emerging orthographic skills. Hence,
integration of the visual data with the auditory results obtained in manuscript 1, indicated an
exclusive relation between coherent motion sensitivity and orthographic skills on the one
hand, and FM sensitivity and phonological skills on the other. These results suggest that basic
visual and auditory sensitivity is likely to play an important role in the development of fine-
7
grained orthographic and phonological representations necessary for successful reading
development.
While in the first three manuscripts preschool data were analysed comparing the high
risk versus low risk group, in study 1 of manuscript 4 we describe the crucial retrospective
analysis comparing children developing literacy problems versus children developing normal
literacy skills. Therefore, children were reclassified based upon family risk status and first
grade literacy achievement. A reanalysis of the data demonstrated that children showing both
the increased family risk and literacy problems at the end of first grade, presented a
significant preschool deficit in letter knowledge, phonological awareness, rapid automatic
naming, frequency modulation detection, coherent motion detection, categorical perception
and speech-in-noise perception. These results confirm that the sensory deficit precedes the
literacy problem, leaving open the possibility of a causal influence of sensory processing upon
literacy development. In a final study (manuscript 4, study 2), the plausibility of this causal
relation is evaluated using path analysis. In particular, we demonstrated that dynamic auditory
processing is related to speech perception which on its turn is related to phonological
awareness. Similarly, dynamic visual processing is related to orthographic ability. Eventually,
phonological awareness and orthographic ability – together with verbal short-term memory -
are unique predictors of reading and writing development.
The thesis ends with a critical discussion of the theoretical, methodological and
practical relevance of this doctoral research project. Finally, some suggestions for future
research are outlined.
8
References
Boets, B., Wouters, J., van Wieringen, A., & Ghesquière, P. (2006a). Auditory temporal
information processing in preschool children at family risk for dyslexia: Relations
with phonological abilities and developing literacy skills. Brain and Language, 97, 64-
79.
Boets, B., Wouters, J., van Wieringen, A., & Ghesquière, P. (2006b). Coherent motion
detection in preschoolers at family risk for dyslexia. Vision Research, 46(4), 527-535.
Boets, B., Ghesquière, P., van Wieringen, A., & Wouters, J. (under revision). Speech
perception in preschoolers at family risk for dyslexia: Relations with low-level
auditory processing and phonological ability. Brain and Language.
Cestnick, L. (2001). Cross-modality temporal processing deficits in developmental
phonological dyslexics. Brain and Cognition, 46, 319-325.
Cornelissen, P., Richardson, A., Mason, A., Fowler, S., & Stein, J. (1995). Contrast sensitivity
and coherent motion detection measures at photopic luminance levels in dyslexics and
controls. Vision Research, 35, 1483-1494.
Demb, J., Boynton, G., & Heeger, G. (1997). Brain activity in visual cortex predicts
individual differences in reading performance. Proceedings of the National Academy
of Sciences, USA, 94, 13363-13366.
Eden, G., Van Meter, J., Rumsey, J., Maisog, J., Woods, R., & Zeffiro, T. (1996). Abnormal
processing of visual motion in dyslexia revealed by functional brain imaging. Nature,
382, 66-69.
Eden, G. F., Stein, J. F., & Wood, F. B. (1995). Temporal and spatial processing in reading
disabled and normal children. Cortex, 31, 451-468.
Farmer, M. E., & Klein, R. M. (1995). The evidence for a temporal processing deficit linked
to dyslexia: A review. Psychonomic Bulletin and Review, 2(4), 460-493.
Frith, C., & Frith, U. (1996). A biological marker for dyslexia. Nature, 382, 19-20.
Galaburda, A. M., Menard, M. T., & Rosen, G. D. (1994). Evidence for aberrant auditory
9
anatomy in developmental dyslexia. Proceedings of the National Academy of Science
of the USA, 91(17), 8010-8013.
Livingstone, M. S., Rosen, G. D., Drislane, F. W., & Galaburda, A. M. (1991). Physiological
and anatomical evidence for a magnocellular defect in developmental dyslexia.
Proceedings of the National Academy of Science, USA, 88(18), 7943-7947.
Lovegrove, B. (1996). Dyslexia and a transient/magnocellular pathway deficit: The current
situation and future directions. Australian Journal of Psychology, 48(3), 167-171.
McBride-Chang, C. (1995). Phonological processing, speech perception and reading
disability: an integrative review. Educational Psychologists, 30(3), 109-121.
Rack, J. P. (1994). Dyslexia: The phonological deficit hypothesis. In A. Fawcett, & R.
Nicolson (Eds.), Dyslexia in children: multidisciplinary perspectives. London:
Harvester Wheatsheaf.
Ramus, F., Rosen, S., Dakin, S. C., Day, B. L., Castellote, J. M., White, S., & Frith, U.
(2003). Theories of developmental dyslexia: insights from a multiple case study of
dyslexic adults. Brain, 126(4), 841-865.
Rosen, S. (2003). Auditory processing in dyslexia and specific language impairment: is there
a deficit? What is its nature? Does it explain anything? Journal of Phonetics, 31, 509-
527.
Snowling, M. J. (2000). Dyslexia (2nd ed.). Malden, MA: Blackwell Publishers.
Stein, J. (2001). The magnocellular theory of developmental dyslexia. Dyslexia, 7, 12-36.
Stein, J., & Talcott, J. (1999). Impaired neuronal timing in developmental dyslexia. The
magnocellular hypothesis. Dyslexia, 5, 59-77.
Stein, J., & Walsh, V. (1997). To see but not to read; the magnocellular theory of dyslexia.
Trends in Neuroscience, 20(4), 147-152.
Talcott, J. B., & Witton, C. (2002). A sensory linguistic approach to the development of
normal and impaired reading skills. In E. Witruk, A. Friederici, & T. Lachmann
(Eds.), Neuropsychology and cognition series. Basic functions of language and
10
language disorders. Dordrecht, Netherlands: Kluwer Academic Publishers.
Tallal, P. (1980). Auditory temporal perception, phonics, and reading disabilities in children.
Brain and Language, 9(2), 182-198.
Van Ingelghem, M., Boets, B., van Wieringen, A., Onghena, P., Ghesquière, P., & Wouters, J.
(2005). An auditory temporal processing deficit in children with dyslexia? In P.
Ghesquière & A.J.J.M. Ruijssenaars (Eds.), Learning disabilities: a challenge to
teaching and instruction. Series: Studia Paedagogica 40. (pp. 47-63). Leuven:
University Press.
Van Ingelghem, M., van Wieringen, A., Wouters, J., Vandenbussche, E., Onghena, P., &
Ghesquière, P. (2001). Psychophysical evidence for a general temporal processing
deficit in children with dyslexia. Neuroreport, 12(16), 3603-3607.
White, S., Milne, E., Rosen, S., Hansen, P., Swettenham, J., Frith, U., & Ramus, F. (in press).
The role of sensorimotor impairments in dyslexia: a multiple case study of dyslexic
children. Developmental Science.
Witton, C., Talcott, J. B., Hansen, P. C., Richardson, A. J., Griffiths, T. D., Rees, A., Stein, J.
F., & Green, G. G. (1998). Sensitivity to dynamic auditory and visual stimuli predicts
nonword reading ability in both dyslexic and normal readers. Current Biology, 8(14),
791-797.
Wright, B. A., Lombardino, L. J., King, W. M., Puranik, C. S., Leonard, C. M., & Merzenich,
M. M. (1997). Deficits in auditory temporal and spectral resolution in language-
impaired children. Nature, 387, 176-8.
11
Notes
1 The most elaborated version of the general magnocellular theory also considers the
accompanying motor problems of dyslexics as a consequence of the single underlying
neurological deficit (Stein, 2001). In the present study however, no motor skills were
assessed.
13
Manuscript 1
Auditory temporal information processing in preschool children at family
risk for dyslexia: Relations with phonological abilities and developing
literacy skills1
Abstract
In this project the hypothesis of an auditory temporal processing deficit in dyslexia was tested by examining
auditory processing in relation to phonological skills in two contrasting groups of five-year-old preschool
children, a familial high risk and a familial low risk group. Participants were individually matched for gender,
age, nonverbal IQ, school environment and parental educational level. Psychophysical thresholds were estimated
for gap detection, frequency modulation detection and tone-in-noise detection using a three-interval forced
choice adaptive staircase paradigm embedded within a computer game. Phonological skills were measured by
tasks assessing phonological awareness, rapid serial naming and verbal short-term memory. Significant group
differences were found for phonological awareness and letter knowledge. In contrast, none of the auditory tasks
differentiated significantly between both groups. However, both frequency modulation and tone-in-noise
detection were significantly related to phonological awareness. This relation with phonological skills was not
present for gap detection.
Introduction
Developmental dyslexia is characterised by serious reading and spelling difficulties
that are persistent and resistant to the usual didactic measures and remedial efforts. At present
it is well established that a major cause of these problems lies in the phonological domain (see
Snowling, 2000 for a review). One hypothesis maintains that this phonological deficit results
from a more fundamental deficit in the basic perceptual mechanisms that are responsible for
auditory temporal information processing.
The auditory temporal hypothesis originated from studies on children with specific
language impairments (SLI) and was later extended to dyslexia. The empirical evidence
started with Tallal’s repetition task (Tallal, 1980). In this temporal order judgement (TOJ)
14
task two complex tones with different fundamentals were presented in pairs at various inter-
stimulus intervals (ISI) and the listener responded with two button presses to identify the
order of the stimuli presented. Tallal found that children with dyslexia, in comparison to
normal readers, were impaired in discriminating and sequencing pairs of short-lived stimuli
with short ISI, and concluded that the dyslexic deficit was specific to processing stimuli that
are brief and occur in rapid succession. Moreover, she found a high correlation between this
basic perceptual processing of non-speech signals and phonological skills (r = .81). Following
further evidence that dyslexic and SLI children had great difficulty discriminating syllables
containing stop consonants (such as /ba/ and /da/), the claim of a temporal deficit was
extended to apply to both non-linguistic and linguistic auditory stimuli (Tallal & Piercy, 1973;
Tallal, Miller, & Fitch, 1993). Since discrimination of such syllables critically depends on
accurate detection of the rapid frequency changes in the first milliseconds of voicing,
inaccurate detection of these formant transitions would inevitably interfere with the
identification of the phonological cues that are typical for spoken language. This hypothesis
of a direct association between basic auditory processing and speech or language processing
was strengthened by demonstrating that speech stimuli with lengthened transitions were much
better discriminated (Tallal & Piercy, 1975). From this association sprang the claim that the
temporal auditory problem caused the language problem, and subsequently the deficient
phonological and reading development. During decades this supposed causal mechanism has
been put forward as a plausible explanation of dyslexia.
Since the formulation of this theory there have been multiple studies exploring the
auditory temporal abilities of individuals with dyslexia. While the bulk of studies has been
done on adults, a minority of recent studies focused on school aged children and some very
few on preschoolers. In line with the scope of our study, we will mainly restrict our report to
psychophysical studies using specific non-speech stimuli to examine younger subjects.
Probably the most straightforward way to measure temporal processing is a gap-
detection task; this task estimates the smallest detectable interruption in an auditory stimulus.
Van Ingelghem and colleagues (Van Ingelghem et al., 2001; 2005) found a significant gap-
detection deficit in 11-year-old dyslexic children compared to normal reading children.
Moreover, the results on the task were significantly related to both real word reading and
nonword reading (r = -0.57 and r = -0.60 respectively). These results were replicated in a
broader study in dyslexic and normal reading children matched for sex, age and intellectual
ability (Van Ingelghem, Boets, van Wieringen, Ghesquière, & Wouters, 2004). The observed
15
results are in line with McCroskey and Kidder (1980), but are not consistent with
observations reported by McAnally and Stein (1996), Schulte-Körne, Deimel, Bartling, and
Remschmidt (1998) and Adlard and Hazan (1998). Hautus, Setchell, Waldie and Kirk (2003)
also observed higher gap detection thresholds in dyslexic subjects, but these thresholds were
only significantly higher for the young reading-impaired subjects (aged 6-9 years) and not for
the older ones (aged 10 years up to adulthood). The authors interpreted these results as
suggestive for a passing maturational lag in temporal acuity in children with dyslexia. In an
interesting study of Fischer and Hartnegg (2004), investigating a large group of subjects
covering an age-range of 7 to 22 years, a higher proportion of subjects with dyslexia were
unable to perform a gap-detection task even at its easiest level. However, within the group of
participants for whom a threshold value could be assigned, there was no significant difference
between dyslexics and normal readers.
Studdert-Kennedy and Mody (1995) challenged Tallal’s auditory theory and argued
that the observed phonological impairments in dyslexics are in origin speech-specific and
cannot be attributed to a more general lower-level auditory deficit. Besides this fundamental
criticism they also postulated that stimulus processing should only be regarded as temporal
when the defining features of the stimuli are changing in time and not merely because of their
rapid and brief presentation. This new temporal concept resulted in a new series of studies
that investigated auditory temporal processing in dyslexia using “dynamic stimuli” (see
Talcott et al., 2000). Most of these studies were carried out on adult samples and
demonstrated a relative impairment in sensitivity to amplitude modulation (AM) (McAnally
& Stein, 1997; Menell, McAnally, & Stein, 1999; Rocheron, Lorenzi, Fullgrabe, & Dumont,
2002) and frequency modulation (FM) (e.g. Stein & McAnally, 1995). In addition, Witton et
al., (1998) found that sensitivity to 2 and 40 Hz FM, for both dyslexics and controls,
significantly correlated with phonological decoding skills. This relationship between FM
sensitivity and phonological ability has also been demonstrated by Talcott et al. (1999) in a
random group of children. More recently, Van Ingelghem et al. (2005) demonstrated a
significant difference in FM sensitivity in a group of 11-year-old dyslexic children compared
to normal reading children. However, in a similar but broader well-controlled study with IQ-
matched control subjects, this difference could not be replicated (Van Ingelghem et al., 2004).
These studies with ‘dynamic’ stimuli again point to an auditory temporal processing
deficit as a possible cause of dyslexics’ phonological problems. Accurate tracking of
amplitude and frequency changes is exactly what is needed for the perception of speech,
16
which is characterised by temporal and spectral variations. Since speech perception is the
basis for developing phonological skills, it is likely that impairments in AM and FM detection
affect phonological skill development via speech perception (McBride-Chang, 1996).
With respect to preschool subjects, as far as we know, there have only been a few
longitudinal studies applying psychophysical measures. Heath and Hogben (2004) and Share
and colleagues (2002) administered Tallal’s repetition test to a large unselected group of
kindergarten children and followed them up until respectively second and third grade.
However, neither of both research groups was able to predict grade two or three literacy
scores based on the auditory data collected in preschool. Conversely, Benasich and Tallal
(2002) administered an operantly conditioned head-turn version of the repetition test to
infants 7.5 months of age born into families who were either positive or negative for family
history of language impairment (SLI). Not only did these authors observe significantly poorer
thresholds for children born into risk families, but they also demonstrated that rapid auditory
processing thresholds at 7.5 months of age were the single best predictor of language
development at two years of age and together with gender predicted up to 40 % of variance in
language outcome at three years of age. Unfortunately, information about literacy
development and its relation with rapid processing thresholds is currently not yet available for
these children. In contrast with the sparse psychophysical studies, there is a growing number
of neurophysiologic studies focusing on the temporal characteristics of speech processing in
very young subjects that already demonstrated promising results comparing genetically high
risk versus low risk children (e.g. Jyväskylä Longitudinal Study of Dyslexia, see Lyytinen et
al., 2001; Dutch Dyslexia Research Programme; Molfese, 2000).
Notwithstanding the large number of studies demonstrating an auditory deficit in
dyslexics, the explicit causality of the auditory hypothesis has never been established directly
by means of a longitudinal study. Here we report data from a longitudinal study that explores
(i) the development of basic auditory skills, speech perception, phonological abilities and
reading skills over a two-year time period from the beginning of the last year of kindergarten2
up to the end of the first year of primary school; (ii) the mutual relations between these
abilities and the way they influence each other over time. In this paper we will discuss the
first results about the relation between auditory temporal processing skills and phonological
skills in two contrasting groups of preschool children, i.e. a genetically high risk and a
genetically low risk group.
17
Auditory processing was assessed by means of three psychophysical threshold tests:
one for gap-detection in noise (GAP), one for 2 Hz FM-detection (FM) and one for tone-in-
noise detection (TN). With the GAP-detection task we tested the hypothesis of a deficit in
‘rapid and brief’ temporal processing. With the FM-detection task we verified the hypothesis
of a deficit in the processing of ‘dynamic stimuli’. The TN-task was included as a non-
temporal control task to verify the specificity of any observed temporal deficit, i.e. we wanted
to examine whether a deficit might be the result of failing performance on auditory
psychophysical tasks in general. Phonological processing was assessed by administering a
broad test battery comprising tasks for rapid serial naming, verbal short-term memory and
phonological awareness. Developing literacy skills were measured using a letter knowledge
task.
In this study we aimed to answer the following questions. First, is it possible to obtain
reliable results while administering such complex psychophysical tasks to very young
subjects? Second, do genetically high risk children, in comparison to low risk children,
perform significantly worse on phonological tasks? Third, do genetically high risk children, in
comparison to low risk children, perform significantly worse on psychophysical tasks for
auditory temporal processing? Fourth, are these auditory processing abilities related to
phonological and developing literacy skills?
Method
Participants
Sixty-two five-year-old children were included in the study. Half of the participants
were children of ‘dyslexic families’, the so-called high-risk group (HR); the other half were
control children of ‘normal reading families’, the so-called low-risk group (LR). Since
dyslexia tends to run strongly in families, preschoolers with dyslexic relatives are more likely
than other children to develop reading problems. Gilger, Pennington, and DeFries (1991)
estimate that roughly between 30 % and 50 % of such children will become reading disabled.
The HR children were recruited by means of referrals and public announcements to
encourage families with a child entering the last year of nursery school and having at least one
member with a formal diagnosis of dyslexia to engage in the study. Following this
recruitment, we received over 300 registrations. Based on a short semi-structured telephone
interview we selected 162 potential candidate families who were sent three questionnaires.
18
One questionnaire investigated in detail the reading and spelling (dis)abilities of all family
members up to third degree and investigated the general development of the preschooler. The
two other questionnaires were translations and adaptations of the Adult Reading History
Questionnaire (Lefly & Pennington, 2000) and investigated the reading experiences and
educational level of each parent. We assessed educational level using the ISCED-scale
(International Standard Classification of Education by UNESCO, 1997), by converting
classifications on the original seven-point scale to a three-point scale. Out of these 162
potential candidates we selected 31 preschoolers based on the following criteria: having at
least one first-degree relative being diagnosed as reading disabled by an authorized
educational psychology service; being native Dutch speaker; born in 1998 and entering last
year of nursery school at the beginning of the study; no history of brain damage, long term
hearing loss or visual problems. Of the 31 selected HR children 3 reported with reading
problems in only one (first-degree) relative, 5 reported with reading problems in two (first-
degree) relatives, 12 with reading problems in three (first-degree or extended) relatives and 11
with reading problems in four or more (first-degree or extended) relatives.
To further increase the number of future dyslexic children in our test population we
selected relatively more male than female preschoolers (ratio: 18M/13F). Finally, to exclude
the “garden variety” poor readers whose literacy is poor due to a low IQ, we selected
proportionally more children out of gifted and higher educated families (see Snowling, 2000).
Children of the LR group met the same selection criteria, with the restriction that they
were not allowed to show any history of speech or language problems and that none of their
family members might have suffered any learning or language deficiencies. For every
individual HR child we searched for the best matching LR control child based on five criteria:
(1) educational environment, i.e. same nursery school, (2) gender, (3) age, (4) non-verbal
intelligence, and (5) parental educational level. In effect we selected the particular control
child out of the group of same-sex classmates of the HR child. All these children –including
the HR child- were administered an adapted version of the Raven Coloured Progressive
Matrices (RCPM) (Raven, Court, & Raven, 1984), a collective non-verbal intelligence test
measuring spatial reasoning. To assure the child’s motivation and attention during testing, we
integrated the test procedure within a game-like fairy tale.
From this group we selected the best overall matching LR control child. For the first
two criteria (educational environment and gender) matching was perfect; for the remaining
three criteria matching was as good as possible within the restrictions of having to choose
19
within a concrete class group. For the age criterion we were able to select all control children
within an age-difference range of maximal five months. For the IQ criterion only four
children differed more than one standard deviation with their matched counterpart. Since the
RCPM is very age-sensitive in young children, we used age corrected norms to calculate this
non-verbal IQ score. Concerning parental educational level we gave relatively more
importance on getting a good fit for maternal than for paternal educational level. In this way
we were able to match 21 children in a perfect way for maternal educational level and 15 for
paternal educational level.
Table 1 gives descriptive characteristics of both groups and test statistics. The mean
age for both the HR and LR group was 5 years and 4 months, not being statistically different
(p = .83). The nonverbal IQ scores were slightly above population average (107 for HR group
and 111 for LR group) and neither differed significantly (p = 0.07). Fisher’s Exact Test also
confirms that both groups did not differ in frequency distribution of the different educational
categories (p = .71 for maternal and p = .43 for paternal educational level).
Table 1
Characteristics of participants : descriptive statistics, paired t-tests and Fisher’s Exact Test.
HR
(n = 31)
LR
(n = 31)
p
Nonverbal IQ (M / SD)a
107 (14)
111 (13)
.07
Age in months (M / SD) 64 (3) 64 (3) .83
Frequency distribution of Low Middle High Low Middle High
Maternal Educational Level 3 8 20 1 9 21 .71
Paternal Educational Level 6 14 11 2 16 13 .43
Note. a Transformed RCPM scores with M = 100, SD = 15.
Apparatus
Phonological Tests
Tests were selected to reflect the three traditional domains of phonological skills
(Wagner & Torgesen, 1987). Phonological awareness was measured by three sound identity
tasks and a rhyme fluency task. Verbal short-term memory was measured by a digit span test
20
and a non-word repetition test. Rapid automatic naming was assessed by administering a
colour and an object rapid naming task.
Sound identity tasks. The child was required to choose from four alternatives the word
that had the same (a) first sound, (b) end sound or (c) end rhyme as a given word (de Jong,
Seveke, & van Veen, 2000, adapted by van Otterloo & Regtvoort). The distracter alternatives
were systematically constructed to prevent guessing. All words were high frequent one-
syllabic Dutch words. Each item consisted of a row of five pictures. The first picture
represented the given word and was separated from the other pictures by a vertical line. All
items were named for the child. The first-sound and end-sound identity tasks both consisted
of ten items, preceded by two practice items, and had a maximum score of ten. The rhyme
identity task consisted of twelve items, preceded by two practice items, and had a maximum
score of twelve.
Rhyme fluency test. Participants were presented a one-syllable word and were required
to produce as many as possible rhyming (non-)words within a twenty second time period. To
familiarize them with the task, the experimenter already offered an example rhyme word for
every target item. Since we were interested in measuring rhyming abilities irrespective of
vocabulary knowledge, the test score was the total number of phonologically correct
responses, regardless of whether it was a real Dutch word or not. The test consisted of eight
items, gradually getting more difficult, and was preceded by two practice items.
Based on the factor analysis of the phonological data (see Results section), we
recalculated the results of the rhyme fluency test to create a purer measure of rhyming ability,
uncontaminated by fluency. For this Simple rhyme test each item was scored in binary
fashion, and treated as correct if at least one rhyming response (word or nonword) was
produced. The maximum score on the test was eight.
Nonword repetition test. A nonword repetition test (NRT) is frequently used as a pure
measure of verbal short-term memory. Since neither the nonwords nor the constituent
syllables of the nonwords used in the NRT correspond to existing words, the use of long-term
memory representations to support recall of the nonwords is prevented. The test was
developed after a Dutch adaptation (Scheltinga, 2003) of the nonword repetition test reported
by Gathercole, Willis, Baddeley, and Emslie (1994). This Dutch version of the NRT was
again adapted for the use with Flemish children. The Flemish NRT consisted of four
21
categories of nonwords, varying in word length from two to five syllables. Each category
contained 12 nonwords. All 48 nonwords and two test items were recorded on a CD and were
presented once. In contrast to Gathercole et al. (1994), the presentation order of the words was
determined by word length; starting with all the two-syllabic words and climbing gradually up
to the five-syllabic words. The test consisted of 48 nonwords, preceded by two practice items,
and had a maximum score of 48.
Digit span forward. The test assessed the immediate serial recall of spoken lists of
digits between 1 and 9. Prior to testing, children were asked to count from 1 to 10 to
familiarize them with the counting string and to reduce the influence of possible differences in
digit knowledge. After a practice session, three trials of each list length were presented,
starting at a sequence of two digits. Testing continued with increasing list length until the
child failed on two of three trials of the same list length. The test score was the total number
of correctly recalled lists. In order to standardize assessment, all lists were recorded and
presented on CD at a rate of one digit per second. For each list length, the stimuli of the first
two trials were taken from the WISC-III (Wechsler, 1992). The third trial was selected from
the Working Memory Battery for Children (see Gathercole & Pickering, 2000).
Rapid automatic naming. The test assessed the rapid serial naming for two types of
familiar symbols: colours and objects (van den Bos, Zijlstra, & Spelberg, 2002). The objects
represented five high-frequent, one-syllabic words: boom (‘tree’), eend (‘duck’), stoel
(‘chair’), schaar (‘scissors’) and fiets (‘bicycle’). The colours were represented by small
rectangles in black, blue, red, yellow and green. For each type of symbol one test card was
given, consisting of 50 symbols in a random order (5 columns of 10 symbols). The child was
instructed to name the symbols on a card as fast and accurately as possible. Prior to testing,
the child was required to name the symbols in the last column of a card to determine whether
he/she was familiar with all the presented symbols. For each card, the number of errors and
the time to completion were recorded. Subsequently, the time to completion was transformed
to the number of symbols named per second. As such, a higher score on the test corresponded
to a higher naming speed.
22
Letter knowledge
To get a preliminary idea about the stage of reading development we administered a
letter knowledge task, since many studies have consistently proven this task to be the best
predictor of the later development of literacy skills (see e.g. Elbro & Scarborough, 2003, for a
recent overview). To test for the receptive and productive letter knowledge, the 16 most
frequently used letters in Dutch books for children were selected (Rolf & Van Rijnsoever,
1984).
Productive letter knowledge. Sixteen printed letters were presented on a card. The
child had to name each of these letters. Both the sound and the name of a letter were
considered correct. The maximum score was 16.
Receptive letter knowledge. Sixteen printed letters were presented on a card. The
experimenter named all letter sounds in random order. After each sound, the child had to
indicate the printed letter that matched the sound. The maximum score was 16.
Auditory tests
Audiometric pure-tone detection. Prior to administering any auditory psychophysical
test we assessed all children on an audiometric pulsed pure-tone detection task to check for
any hearing loss. All but one child obtained a PTA-score below the 25 dB HL criterion. For
this child showing mild hearing loss, all further auditory testing was administered with
increased stimulus amplitude proportionally to the hearing loss. Since detailed inspection of
all her test results didn’t show any anomalies, her data were not excluded from further
analyses.
GAP-detection test. In the GAP-detection test, white noise stimuli were used. The
target stimulus was a white noise stimulus containing a silent gap. The reference stimulus was
an uninterrupted white noise. Stimuli were cosine gated on and off with 50 ms rise and fall
times. Gap rise and fall times were 0.5 ms. 64 target stimuli were constructed, comprising 32
gap sizes, varying between 100 ms and 0.1 ms. Gap length decreased with a factor 1.2 from
100 ms towards 6.5 ms. From here on gap length decreased with a fixed step size of 0.4 ms.
In order to prevent participants from using overall duration as a cue for detection, the length
23
of both the target and the reference stimulus was varied randomly from presentation to
presentation (van Wieringen & Wouters, 1999). In the target stimulus, the length of the
markers (i.e. noise components surrounding the gap) varied between 250 and 650 ms
including on and off set (i.e. 250, 400, 500 and 650 ms). The length of the reference stimulus
was 750, 900 or 1050 ms including on and off set. Stimuli were presented monaurally at
70 dB SPL with an inter stimulus interval (ISI) of 400 ms.
FM-detection test. In the FM-detection test, stimuli can be defined as x(t) = Asin
[2πfct + β sin (2πfmt)] in which β is the modulation index (β=∆f/fm), fc the carrier frequency,
fm the modulation frequency and ∆f the frequency deviation. The target stimulus was a 2 Hz
frequency modulation (fm) of a 1 kHz carrier tone (fc) with varying modulation depth ∆f.
Modulation depth decreased with a factor 1.2 from 100 Hz towards 11 Hz. From a ∆f of
11 Hz, a step size of 1 Hz was used. Frequency modulation in the target stimulus was
sinusoidal and the modulation envelope was always in sine phase. The reference stimulus was
a pure tone of 1 kHz (β=0). The length of both the reference and the target stimulus was
1000 ms including 50 ms cosine-gated onset and offset. Stimuli were presented monaurally at
70 dB SPL with an ISI of 350 ms.
Tone-in-noise detection test. In the tone-in-noise (TN) detection task participants had
to detect pure tone pulses within a one-octave noise signal, centered around 1 kHz (from 707
to 1414 Hz). Noise was presented monaurally at 55 dB SPL with an ISI of 300 ms. The length
of both the target and the reference stimulus was 1620 ms, including 20 ms linearly gated rise
and fall times. For the target stimulus noise contained two pure 1 kHz pulses of 440 ms,
including 20 ms linearly gated on and offset. The first pulse started 320 ms after stimulus
onset, the second one 960 ms after stimulus onset. The amplitude of the pulses varied with a
signal-to-noise ratio (SNR) between +25 dB and –20 dB. From +25 to -3 dB SNR amplitude
decreased with a 4 dB step size. From here on amplitude decreased with a 1 dB step size.
All stimuli for GAP, FM and TN were generated in MATLAB 5.1 and saved as 16-bit
wav-files (sample frequency 44100 Hz) on the hard disc of a Dell Latitude C800 and Toshiba
Satellite 1400-103 portable computer. They were presented using an integrated audio PC-card
and routed to an audiometer (Madsen OB622) in order to control the level of presentation.
The stimuli were presented monaurally over calibrated TDH-39 headphones.
24
Psychophysical procedure
For all GAP, FM and TN tests a similar psychophysical procedure was used.
Thresholds were estimated using a three-interval forced-choice oddity paradigm. The
subject’s task was to identify the ‘odd’ stimulus, the one that sounded different from the other
two. The length of the gap, the depth of modulation and the amplitude of the sinusoidal pulses
were adjusted adaptively using a two-down, one-up rule, which targeted the threshold
corresponding to 70.7 % correct responses (Levitt, 1971). In all tests a threshold run was
terminated after eight reversals. Thresholds for an individual run were calculated by the
geometric mean of the values of the last four reversals. For each participant three threshold
estimates were determined for every experiment. Prior to auditory data collection, participants
were given a short period of practice, comprising supra-threshold trials, to familiarise them
with the stimuli and the task.
The forced-choice oddity paradigm was controlled by APEX, a software module
developed for psycho-acoustical and psycho-electrical auditory testing (Laneau, Boets,
Moonen, van Wieringen, & Wouters, 2005). In order to make the rather boring
psychophysical tests more interesting and child friendly we integrated them in an interactive
video game with intro and outro animation movies, aimed to transform the abstract
meaningless acoustical signal into a concrete and well known ‘daily life signal’. This concept
of testing was based upon earlier work of Soderquist and Shilling (1992) and Wightman and
co-workers (Allen, Wightman, Kistler & Dolan, 1989; Wightman, Allen, Dolan, Kistler &
Jamieson, 1989). During testing, the three intervals of every trial were visually represented on
the screen by an identical character. The characters were animated synchronously with the
presentation of the corresponding sound interval in order to create a psychological link
between the moving object on the screen and the presented sound signal. The child’s task was
to choose, by pointing to the touch screen, which of the three characters corresponded to the
target sound (i.e. sounded different from the other two). Immediately after the child touching
the screen, visual feedback was given in the form of a moving cartoon. After correct selection,
feedback was much more spectacular – and as such much more reinforcing- than after
incorrect selection. During the sequence of trials constituting a single run, APEX also
provided a more global reinforcement by adding little smiley faces to a rising ladder structure
for every correct response and removing them again after every incorrect response. At the end
25
of the run the number of smiley faces was evaluated and the child was rewarded
proportionally.
For the GAP-detection experiment the video game displayed a teasing mouse who is
trying to wake up three sleeping snakes because she wants to play with them. The child had to
predict which one is going to wake up by listening to the specific interrupted noise sound
‘ssss-ssss’. For the FM-detection experiment the introductory movie showed a mother dragon
sitting in front of three eggs. The child’s task was to predict which egg is ready to hatch, by
listening to the baby dragon crying inside (‘wouwouwouw’). For the TN-detection task we
presented an introductory animation movie about a little monkey waiting desperately at the
school gate for his father to pick him up by car. Three cars arrive at the school (making noise
sound) and the child had to identify the father’s car by listening to him honking his horn (=
the pure tone pulses).
Data collection
All auditory and phonological data were collected within a 70-day period between the
second and the fourth month of the last year in kindergarten. Intelligence testing (RCPM)
took place one month earlier. Data collection was carried out by qualified psychologists and
audiologists. Testing took place in a quiet room at the children’s school. Since the LR child
was selected out of the HR child’s classmates, we could always test both children in exactly
the same circumstances.
All phonological tests and the letter knowledge task were administered individually in
one day during three sessions. After every subtest children were rewarded by receiving little
stickers or stamps.
Auditory data were collected during two consecutive days. Testing always started with
the pure-tone audiogram. Then we administered one run of the TN test since this is
conceptually the easiest psychophysical task. Consequently we administered one run of the
GAP and FM test in a contra balanced way. This sequence was continued until we had three
threshold estimates for every experiment.
26
Statistical Analysis
Prior to analysis all data were individually checked for unexpected outliers. This
resulted in the removal of only two unreliable phonological test scores for one subject of the
dyslexic group.
All results were analysed in a paired wise manner, comparing HR versus LR group at
the level of the matched individuals. Although both groups did not show a significant
difference on any of the matching criteria, we decided to rule out any possible influence of
age, nonverbal intelligence or parental educational level by controlling for these variables in
our analyses. As such, we analysed the data using Mixed Model Analysis (MMA) with school
as a random variable (1 to 31) and participant group (HR versus LR) as the fixed between-
subject variable (Littell, Milliken, Stroup & Wolfinger, 1996). Age, nonverbal IQ and
educational level of both mother and father were added as fixed (co)variables. Additionally,
for the auditory data we also computed a series of Repeated Measures MMA with threshold
run (1 to 3) as the within-subject variable, participant group as the between-subject variable
and with the same covariates as mentioned above. MMA was chosen not merely to allow a
paired wise comparison, but also because of its robustness in analysing semi-normally
distributed data (Verbeke & Lesaffre, 1997). In order to approach a normal distribution for
more variables, the GAP and FM thresholds and the results on the letter knowledge task were
log-transformed prior to MMA. To explore the internal consistency of the different
phonological tasks, Cronbach’s alpha-coefficients were calculated. To explore the structure
and the mutual relationships of the phonological tests a principal component factor analysis
was done with varimax rotation. Relationships between variables were analysed using
Spearman correlation coefficients.
Results
Phonological skills and Letter Knowledge
Descriptive statistics, MMA results and reliabilities (Cronbach’s α) for all measures
are displayed in Table 2. The internal consistency of the simple rhyme task, the nonword
repetition test and the letter knowledge task was good. The reliability of the rhyme, first
phoneme and end phoneme identity tasks was somewhat lower, probably because these tasks
appeared to be rather difficult3.
27
For the colour and picture rapid naming tasks we examined whether there might have
been a speed-accuracy trade-off. Since there was no significant correlation between naming
speed and error number and since the quantity of errors did not differ between both groups,
we did not correct the speed scores for error rate.
Table 2
Phonological abilities and letter knowledge: descriptive statistics and p-values for paired wise MMA, controlling
for nonverbal IQ, age and parental educational level.
HR LR
Measures Cronbach α Maximum M SD M SD p
Phonological Awareness
Rhyme fluency - - 14.3 9.6 18.9 8.2 .02
Simple rhyme .88 8 6.2 2.4 7.3 1.8 .05
Rhyme identity .69 12 8.7 2.7 9.8 1.9 .23
First-sound identity .59 10 4.6 2.1 5.6 2.4 .20
End-sound identity .63 10 4.5 2.2 5.9 2.4 .02
Rapid Automatic Naming
Colour naming - - 0.6 0.1 0.7 0.2 .12
Picture naming - - 0.6 0.1 0.7 0.2 .11
Verbal Short Term Memory
Digit span - 21 7.0 1.6 6.9 1.5 .66
Nonword repetition test .84 48 16.8 5.3 20.3 7.1 .08
Letter Knowledge .90 32 5.6 6.6 8.2 6.9 .03
Factor Phonological Awareness - - -0.73 1.2 0.00 1.0 .04
Factor Rapid Automatic Naming - - -0.19 0.9 0.00 1.0 .30
Factor Verbal STM - - -0.14 0.9 0.00 1.0 .72
Although the results on almost any phonological task were in the expected direction
with the HR-group scoring less well than the LR-group, not all these group differences turned
out to be statistically significant. Considering the tasks meant to measure phonological
awareness, the scores of the HR-group were significantly lower for the rhyme fluency test and
for the end phoneme identity test, with the results on the simple rhyme task being marginally
significantly lower. For the rapid serial naming tests neither the picture naming nor the colour
naming task differed significantly between both groups. On the tests of verbal short-term
memory only the nonword repetition test showed a slight tendency towards differing between
both groups.
28
It is noteworthy that some of the group differences have been tempered by applying
the strict controlling MMA design. For example, by comparing both groups without
controlling for nonverbal IQ, age and parental educational level, we also found significant
group differences for the nonword repetition test and simple rhyme test, and marginally
significant differences on the first phoneme identity task4.
With respect to developing literacy skills, we found a significant group difference on
the log-transformed letter knowledge scores.
Because the number of participants was not big enough to perform a reliable
confirmatory factor analysis, an exploratory principal component factor analysis with varimax
rotation was carried out to examine the data structure. Since we wanted to explore the unique
relation between the individual phonological sub skills and the auditory processing skills,
unrelated (i.c. orthogonal) phonological factors were calculated. This analysis revealed that
the rhyme fluency task disturbed the assumed threefold phonological structure. A post hoc
explanation for this phenomenon might be that this fluency task depended only marginally on
rhyming skills, while depending mostly on skills as flexible and creative thinking.
Replacement of this test by the described simple rhyme test, which was a pure rhyme
measure, resulted in an excellent three-factor structure (based on the eigenvalue criterion).
The first factor had heavy loadings of the three sound identity tasks and the simple rhyme
task, and a more modest loading of the nonword repetition test (see Table 3). As a
consequence, this factor could be reliably labelled as the Phonological Awareness Factor
(PhAw). The second factor was completely determined by heavy loadings of both the colour
and the picture naming task, and as such could be regarded as the Rapid Automatic Naming
Factor (RAN). Finally, the nonword repetition test and the digit span loaded heavily on the
last factor, namely the Verbal Short Term Memory Factor (VSTM).
Table 2 displays descriptive statistics and MMA results for these three latent
phonological factors. To assist in the interpretation of these results, factor values were
transformed to effect sizes relatively to the mean and standard deviation of the LR-group. As
can be seen, both groups did not differ on Verbal Short Term Memory and Rapid Automatic
Naming, but they differed significantly on Phonological Awareness (p = .04).
29
Table 3
Principal component factor analysis with varimax rotation: factor loadings of the phonological measures.
Factor 1
Phonological
Awareness
Factor 2
Rapid Automatic
Naming
Factor 3
Verbal STM
Simple rhyme .68
Rhyme identity .83
First-sound identity .82
End-sound identity .80
Colour naming .91
Picture naming .92
Digit span .83
Nonword repetition test .36 .74
Note. Only factor loadings above .35 have been depicted.
Auditory measures
For every auditory experiment a paired wise Repeated Measures MMA was computed
with group as between-subject variable (HR versus LR) and threshold run as within-subject
variable (run 1 to 3). Results can be summarised as follows: (a) neither for GAP, nor for FM,
nor for TN-detection there was a significant group effect (p = .08, p = .33 and p = .66,
respectively); (b) the three auditory tests showed a significant effect of threshold run (p <
.001, p = .01 and p < .001, respectively) and (c) for none of the tests the group by run
interaction was significant (p = .67, p = .21 and p = .47, respectively). For every auditory
experiment, post-hoc analysis revealed that none of the three threshold measures
differentiated significantly between HR and LR group. Furthermore, for every experiment
there was only a significant learning effect from the first to the second run; the second and
third run did not differ significantly from each other.
Importantly, in contrast to the phonological data, the results were not influenced by
applying the conservative MMA design. Even while analysing the auditory data without any
covariates added, the null-results were virtually identical.
Although the Repeated Measures MMA revealed a general learning effect from the
first to the second threshold run, this tendency did certainly not apply to all subjects. For
many of them, the first threshold was better than the second or third, or the second threshold
was better than the third. Moreover, since we are interested in threshold estimations as an
indicator of a subject’s sensory capability, average threshold (or the average of the last two
30
threshold runs) might not be the most appropriate measure; especially not in this age group
that traditionally shows a high intrasubject variability (Wightman & Allen, 1992). Because
our interest is in the best level of performance a subject is able to reach, a more reasonable
estimator of threshold is each subject’s “best” performance, or the lowest threshold of the
three estimates for every experiment (see Wightman et al., 1989). While using a three-interval
oddity paradigm, the probability of progressing to the next more difficult level just by chance
is only 11.1 % (1 to 9). This means that the probability of progressing two consecutive levels
by guessing is very small (only 1.23 %). Hence, on an average of 35 trials for every run and
every experiment, it turns out to be very implausible that this “best threshold” would just be
the result of lucky guessing instead of reflecting the real sensory capability limit. Threshold
estimates and test statistics for the best and second best threshold and for the mean of the two
best thresholds are given in Table 4. For every auditory experiment, Mixed Model Analysis
showed there was no significant group effect.
Table 4
Auditory Measures: Descriptive statistics and p-values for paired wise MMA, controlling for nonverbal IQ, age
and parental educational level
HR LR
Measures M SD M SD p
PTA (dB HL)
Best GAP1 (ms)
Best GAP2 (ms)
AV GAP ½ (ms)
Best FM1 (Hz)
Best FM2 (Hz)
AV FM ½ (Hz)
Best TN1 (dB SNR)
Best TN2 (dB SNR)
AV TN ½ (dB SNR)
13.3
4.2
7.4
5.8
6.0
10.4
8.2
-8.0
-6.9
-7.5
5.1
2.4
6.7
4.3
3.8
6.2
4.6
2.3
2.4
2.3
13.4
4.2
5.8
5.0
5.4
9.0
7.2
-9.0
-7.5
-8.2
6.0
3.1
4.3
3.6
2.2
6.3
4.0
1.8
1.4
1.5
.94
.88
.30
.40
.91
.28
.41
.71
.94
.81
Note. PTA: Pure Tone Average, BestGAP1-2: best and second best GAP threshold, AV GAP ½: average of the
two best GAP thresholds, BestFM1-2: best and second best FM threshold, AV FM ½: average of the two best
FM thresholds, BestTN1-2: best and second best TN threshold, AV TN ½: average of the two best TN
thresholds
Spearman rank correlations between the best and the second best threshold estimate
for every experiment appeared to be satisfactory and were rs = .79 for GAP, rs = .75 for FM
and rs = .83 for TN, all being significant at the p < .0001 level.
31
Table 5
Spearman correlation coefficients: age, nonverbal intelligence and auditory measures.
AV GAP ½ AV FM ½ AV TN ½
Age
Nonverbal IQ
AV GAP ½
AV FM ½
0.10
-0.32*
0.15
-0.16
0.27*
-0.13
-0.18
0.36**
0.43***
Note. AV GAP ½: average of the two best GAP thresholds in ms, AV FM ½: average of the two best FM
thresholds in Hz, AV TN ½: average of the two best TN thresholds in SNR.
Note. * p < .05, ** p < .01, *** p < .001
Table 5 shows the Spearman rank interrelations between the different auditory
psychophysical measures, and their relation to age and nonverbal IQ. None of the auditory
tasks seemed to be related to age, and only GAP-detection showed a significant relation to
nonverbal intelligence. All auditory tasks appeared to be significantly correlated with each
other, with the relation between FM and TN being the most substantial one.
Relations between phonological and auditory skills
To analyse the relationship between participants’ auditory processing skills and their
phonological abilities and developing literacy skills, Spearman correlation coefficients were
calculated between the participants’ GAP, FM and TN thresholds on the one hand, and the
raw and combined phonological and letter knowledge scores on the other hand. Table 6 offers
an overview of these correlations for total group (TG), HR-group and LR-group. Correlations
have been partialed out for the possible influence of age and nonverbal IQ.
For the total group both FM and TN-detection were significantly related to all
variables measuring phonological awareness skills, and consequently they were also
significantly related to the composite Phonological Awareness factor (rs = -.48 and -.35
respectively). In contrast, GAP-detection was not related to any of the phonological variables.
FM-detection was the only auditory variable being significantly related to Letter Knowledge.
For both groups separately, slightly different relationships could be observed. In the
HR-group only TN-detection was significantly related to Phonological Awareness and its
constituent subtasks, whereas in the LR-group FM-detection was exclusively related to this
32
Phonological Awareness factor. Remarkably, in the HR-group both FM and TN-detection
appeared to be significantly related to Picture Naming (rs = -.41). The GAP-detection task
again appeared to be unrelated to any of the phonological variables. In the subgroups the
power of the Spearman test was too weak to reveal a relation between Letter Knowledge and
any of the auditory variables.
Individual deviance analysis
Since one of the goals of this study was to explore early indicators of dyslexia and in
view of the fact that group comparisons might mask significant individual differences, we
also carried out analyses on the subject level. To decide which individual did and did not
show abnormal performance, we adopted the two-step criterion as suggested by Ramus et al.
(2003). Applying this procedure, the criterion for deviance has been placed on 1.65 standard
deviations of the mean of the LR-group. In a normal distribution this corresponds to the fifth
percentile and as such it is a fairly strict criterion. However, if a LR subject may occasionally
show abnormal performance, this would make the criterion much more stringent by
excessively influencing the LR mean and standard deviations. Moreover, the occurrence of
these low scoring LR subjects might be especially probable in this preschool population since
there is still a chance that even children of the LR group would become dyslexic. For this
reason, the criterion has been applied in two steps: (1) compute the control mean and standard
deviation and identify LR subjects who qualify for abnormal performance according to the
1.65 SD criterion (typically, this applied to 1 or 2 LR subjects for each measure);
(2) recompute the LR mean and standard deviation excluding these deviant LR subjects, and
identify HR subjects who are outside +/- 1.65 SD.
Table 6
Spearman partial correlation coefficients, partialed out for age and nonverbal IQ.
Simple
rhyme
Rhyme
identity
First-sound
identity
End-sound
identity
Colour
naming
Picture
naming
Digit
span
Nonword
repetition
Factor
PhAW
Factor
RAN
Factor
VSTM
Letter
knowledge
Total group (N = 61)
AV GAP ½
AV FM ½ -0.33* -0.34** -0.43*** -0.45*** -0.24 -0.48**** -0.29*
AV TN ½ -0.29* -0.36** -0.28* -0.24 -0.35**
HR group (N = 30)
AV GAP ½ -0.36
AV FM ½ -0.43* -0.34 -0.33 -0.32 -0.41* -0.34 -0.31
AV TN ½ -0.39* -0.55** -0.50** -0.41* -0.52** 0.36
LR group (N = 31)
AV GAP ½
AV FM ½ -0.42* -0.50** -0.59*** -0.60***
AV TN ½
Note. AV GAP ½: average of the two best GAP thresholds in ms, AV FM ½: average of the two best FM thresholds in Hz, AV TN ½: average of the two best TN thresholds
in SNR.
Note. Only correlations with a p-value above .10 have been depicted.
Note. * p < .05, ** p < .01, *** p < .001 and **** p < .0001
34
Individual scores for the psychophysical measures (average of best and second best
threshold) and for the phonological factors are plotted in figure 1. A distribution analysis on
the data of the ‘restricted’ LR-group confirmed the normality of these variables. As such, the
1.65 SD deviance criterion corresponded indeed to the postulated fifth percentile. A deviance
analysis on Phonological Awareness revealed that 12 HR subjects out of 30 showed abnormal
performance, this corresponds to 40 % of the HR group. In the LR group 5 subjects out of 31
(= 16 %) had abnormal performance. For Rapid Automatic Naming 3 subjects out of each
group had abnormal performance (corresponding to 10 %) and for Verbal STM 3 HR versus 2
LR subjects showed deviant scores (10 % versus 6 %). Considering the auditory measures, for
both FM and GAP detection there were 8 HR subjects versus 4 LR subjects who showed
abnormal performance (26 % versus 13 %). For TN there were 3 HR subjects (10 %) versus 1
LR subject (3 %) having deviant results.
To summarize, again Phonological Awareness turned out to discriminate best between
both groups by having a significant higher proportion of deviant subjects in the HR group
(Fisher Exact Test, p = .04). For both GAP detection and FM detection the proportion of
subjects showing abnormal performance was twice as high in the HR group compared to the
LR group. However, this tendency was not significant (Fisher Exact Test, p = .17 for both
tests).
By looking at the individual scores for all subjects showing abnormal performance in
at least one measure, it becomes clear that there is no straightforward regularity or tendency
between these measures. Auditory deficits appeared to be largely mutually unrelated with
some subjects differing only on one task, others on two or three tasks, but without any
consistency. The same applied for deficits on the phonological factors. Moreover, the relation
between auditory deficiencies on the one hand and phonological deficiencies (more
specifically in phonological awareness) on the other hand was even more ambiguous. Some
subjects suffered really serious deficits in auditory processing without showing any
phonological problems. Conversely, other subjects obtained deviant phonological results
while demonstrating perfectly intact auditory processing. Finally, in some subjects
phonological and auditory processing deficits appeared to be partially related.
Fig 1: Individual Z-scores on phonological and auditory measures. The solid line indicates the mean for all LR subjects above
Pc 5; the dashed line indicates the chosen deviance criterion (1.65 SD deviating of the LR mean after excluding deviant LR
subjects). Deviant individuals are identified by their pair number.
36
Discussion
Feasibility of psychophysical testing in preschoolers
One of the main objectives of this study was to explore the feasibility of administering
complex psychophysical tests to very young subjects. Based on our results, this research
question can be answered entirely confirmative. Not only did the children perform
surprisingly accurately, but they also really enjoyed the auditory tasks.
While comparing the auditory thresholds of our LR-subjects with results on identical
tasks administered to 11-year old normal reading children (Van Ingelghem et al., 2004), the
preschoolers showed an overall weaker discrimination capacity (GAP: 4.2 versus 2.1 ms, FM:
5.4 versus 3.3 Hz, TN: -9.0 versus -12.7 dB SNR). The higher thresholds of preschoolers
compared to older children and adults is a general observation that has been demonstrated in
numerous auditory studies across all kind of discrimination tasks (e.g. Allen, et al., 1989;
Irwin, Ball, Kay, Stillman, & Rosser, 1985; Jensen & Neff, 1993; Morrongiello, Kulig, &
Clifton, 1984; Schneider & Trehub, 1992; Trehub, Schneider, & Henderson, 1995). However,
it remains unclear whether this weaker discrimination reflects an underlying sensory
immaturity or has a cognitive origin (e.g. non-optimal listening strategies, fluctuations in
attention, etc.). Another general observation in psychophysical testing with preschoolers
concerns the higher inter- and intrasubject variability. As Wightman and colleagues
demonstrated, the intrasubject variability might mainly result from central non-auditory
attentional factors (e.g. Oh, Wightman & Lutfi, 2001; Wightman, et al., 1989; Wightman,
Callahan, Lutfi, Kistler & Oh, 2003). For this reason, the way we reduced this excessive
variability by opting for a ‘best threshold analysis’, can be justified. In contrast, the
intersubject variability is not a result of error variance, but reflects reliable differences in the
speed of neural development (e.g. Allen & Wightman., 1994). Because of this substantial
intersubject variability, caution is required while interpreting averages and group
performances (Wightman & Allen, 1992).
Phonological abilities and letter knowledge
With respect to the phonological data, the exploratory factor analysis convincingly
revealed the three-dimensional phonological structure as postulated by Wagner and Torgesen
(1987). Moreover, both the group analysis and the individual deviance analysis clearly
37
demonstrated the robustness of the phonological deficit hypothesis in dyslexia-prone children.
Even at a preschool age, the HR-children already showed a significant deficit in phonological
awareness, not only at the rhyme-level but also at the level of the phonemes. These results are
consistent with other longitudinal prospective studies that revealed similar deficits in
genetically at risk children (e.g. Elbro, Borstrom, & Petersen, 1998; Gallagher, Frith, &
Snowling, 2000; Pennington & Lefly, 2001; Scarborough, 1989; 1990; 1998). The group
differences on the factors rapid automatic naming and verbal short-term memory were
insignificant but in the expected direction with the HR-group scoring less well than the LR-
group. These results are in line with findings by Elbro et al. (1998). However, some
researchers did find significant differences on verbal short-term memory (Pennington &
Lefly, 2001 – significant group difference in first grade, but not in kindergarten) and rapid
automatic naming (de Jong & van der Leij, 2003; Pennington & Lefly, 2001), but only while
comparing HR dyslexic subjects versus LR normal readers. This means they retrospectively
reanalysed the kindergarten data after having diagnosed their subjects in second or third
grade. Since currently the children in our study are still attending kindergarten, we do not yet
know who will finally become dyslexic, and consequently we obviously cannot carry out this
analysis yet.
Although most children hardly knew any letters at the beginning of the last year in
kindergarten (on average about three or four letters), the group difference was already
significant. Again this result is in line with any of the previously mentioned longitudinal
studies. Since both letter knowledge and phonological awareness have consistently been
proven to be among the best single preschool predictors of literacy development (see e.g. the
impressive meta-analysis of Scarborough, 1998, based on 61 studies), and since the HR
versus LR group differed especially on these measures, it is likely that the familial high risk
group will contain a disproportionally high number of future cases of dyslexia.
Auditory processing skills
We studied GAP and FM-detection in order to investigate auditory temporal (‘rapid
and brief’ versus ‘dynamic’) processing. In line with the temporal hypothesis, we expected
these tasks to differentiate between both risk groups and to be related to specific pre-reading
skills like phonological processing and letter knowledge. However, we did not find any
significant differences in auditory processing between the HR and the LR group, neither at a
group level nor in individual deviance analyses. Although there were twice as many subjects
38
showing abnormal performance for GAP and FM-detection in the HR-group, this tendency
did not reach significance. Assuming the correctness of the causal auditory hypothesis, this
lack of significance might be attributed either to the typically greater interindividual
variability in children (cfr. supra) or to the fact that we did not study a well-defined clinical
group but only a risk group that still might show substantial overlap with the non-affected
control group. Moreover, Bishop et al. (1999) demonstrated in a twin study on SLI children
that in contrast to the highly heritable phonological skills, auditory skills depend less on
genetic and more on environmental influences. As such, our finding of a phonological deficit
in combination with relatively intact auditory skills in this genetic high risk group
corresponds well with the results of Bishop and colleagues.
Recently, an alternative explanation has been put forward to explain the variably
observed auditory deficits in subjects with specific language impairment and dyslexia. Bishop
and colleagues (Bishop & McArthur, 2004; Bishop & McArthur, 2005; McArthur & Bishop,
2004) and Wright and Zecker (2004) suggested that these subjects might not suffer from a
specific chronic auditory deficit, but rather from a more general and passing auditory
maturational delay (estimated to encompass about three or four years). In that perspective the
differential sensitivity of a task will be at best when administered in the age period that the
measured skill is sharply improving in typically developing subjects. However, in spite of
having administered our auditory tasks during a sensitive developmental period - it has been
demonstrated that normally developing five-year-old subjects undergo a rapid maturation of
both spectral and temporal auditory abilities5 (e.g. Irwin et al., 1985; Jensen & Neff, 1993;
Thompson, Cranford & Hoyer, 1999; Wightman et al., 1989) - we did not observe the
hypothesized developmental delay in the HR group. This is markedly in contrast with results
obtained by Hautus et al. (2003), who demonstrated that younger reading-impaired children
(aged 6-9 years) had significantly higher gap-detection thresholds than age-matched controls
while older dyslexic subjects (aged 10-13 and adults) did no longer differ from controls.
Relations between auditory and phonological abilities
With regard to relations between auditory processing and phonological processing,
both FM and TN-detection thresholds were significantly related to phonological awareness
and -to some extent- to rapid automatic naming and letter knowledge. In contrast, the GAP-
detection task was completely unrelated to any phonological measure. These results suggest
that it is not the specific temporal aspect of auditory processing that is related to developing
39
phonological abilities, since the temporal GAP-detection task was not related whereas the
non-temporal TN-detection task turned out to be significantly related to phonological
processing. Instead it appears as if the common spectral or frequency sensitive characteristic
of TN and FM-detection causes the relation with phonological ability. At a neurophysiologic
level this might imply that a more accurate phase-locking system or smaller and more sharply
tuned auditory filters are somehow related to better phonological processing (e.g. Carney,
Heinz, Evilsizer, Gilkey & Colburn, 2002; Moore, 1997). This is in line with many studies
demonstrating an impairment in frequency discrimination in dyslexic subjects (e.g. Ahissar,
Protopas, Reid & Merzenich, 2000; Amitay, Ben-Yehudah, Banai & Ahissar, 2002; Cacace,
McFarland, Ouimet, Schrieber, & Marro, 2000; Fischer & Hartnegg, 2004). Talcott et al.
(2002) reported similar observations from a large-scale primary school study in which
auditory frequency resolution differed between groups of children with different literacy
skills. In the same way, Hulslander et al. (2004) observed a significant correlation between
FM-thresholds and scores on a phoneme awareness composite, even while controlling for
individual differences for full-scale IQ. Evidence supporting the neurophysiologic explanation
for a deficit in frequency sensitivity has been put forward by McAnally and Stein (1996) who
demonstrated that dyslexics were less able to generate neural discharges phase-locked to the
temporal fine structure of the acoustic stimuli. Data consistent with these findings were also
reported by Baldeweg, Richardson, Watkins, Foale, and Gurzelier (1999), Dougherty,
Cynader, Bjornson, Edgell, and Giaschi (1998) and Schulte-Körne et al. (1998). Recently,
Amitay, Ahissar and Nelken (2002) also found evidence for a deficit in tone-in-noise
detection in adult dyslexic subjects. However, their results were conflicting with McAnally’s
findings of a phase-locking deficit in dyslexics (for a similar conclusion see also Hill, Bailey,
Griffiths & Snowling, 1999).
The significant relation we observed between the auditory measures and phonological
awareness is in line with the results of Share et al. (2002) who also found a reliable
concurrent correlation between a non-linguistic TOJ task and phoneme segmentation at
school entry. Unfortunately, this relation could not be interpreted in a causal way since the
auditory measures were not able to predict any later phonological or reading skills. Instead,
Share and colleagues speculated that the association between early temporal processing and
phonological awareness might be the result of a higher-order common cause of an unspecified
metalinguistic nature. Similarly, the substantial concurrent correlation we found between
phonological and spectral auditory measures should not be interpreted in a directional or
causal way. After all, while inspecting individual data in the deviance analysis, it is clear that
40
there is no obvious straightforward relation between deficits in auditory measures and deficits
in phonological skills. Although deviant auditory processing tends to be a risk indicator for a
deficit in phonological development, intact auditory processing is certainly not a sufficient
prerequisite for developing normal phonological skills. More generally, we have to conclude
that we are not able to demonstrate a consistent and convincing pattern in individual
deficiencies, neither within the auditory skills, nor within the phonological sub-skills, nor
within the relation between auditory and phonological skills.
Conclusion
To conclude, phonological awareness and letter knowledge turn out to be the best
indicators to differentiate between preschool children with low versus high familial risk of
developing dyslexia. In contrast, none of the auditory processing tasks is able to differentiate
significantly between both groups. However, auditory spectral tasks (FM and TN-detection
thresholds) are highly significantly related to phonological awareness. This relation is not
present for a specific temporal GAP-detection task. Nevertheless, identifying deviant subjects
in auditory spectral processing in order to predict deficiencies in phonological skills and
subsequent reading development does not yet seem to be a viable option, since at the level of
individual subjects the relation between auditory and phonological skills seems to be much
less straightforward.
41
References
Adlard, A., & Hazan, V. (1998). Speech perception in children with specific reading
difficulties (dyslexia). Quarterly Journal of Experimental Psychology, 51(1), 153-177.
Ahissar, M., Protopapas, A., Reid, M., & Merzenich, M. (2000). Auditory processing parallels
reading abilities in adults. Proceedings of the National Academy of Science, 97 (12),
6832-6837.
Allen, P., Wightman, F., Kistler, D., & Dolan, T. (1989). Frequency resolution in children.
Journal of Speech and Hearing Research, 32(2), 317-322.
Allen, P., & Wightman, F. (1994). Psychometric functions for children's detection of tones in
noise. Journal of Speech and Hearing Research, 37(1), 205-215.
Amitay, S., Ahissar, M., & Nelken, I. (2002). Auditory processing deficits in reading disabled
adults. Journal of the Association for Research in Otolaryngology, 3, 302-320.
Amitay, S., Ben-Yehudah, G., Banai, K., & Ahissar, M. (2002). Disabled readers suffer from
visual and auditory impairments but not from a specific magnocellular deficit. Brain,
125(10), 2272-2285.
Baldeweg, T., Richardson, A., Watkins, S., Foale, C., & Gurzelier, J. (1999). Impaired
auditory frequency discrimination in dyslexia detected with mismatch evoked
potentials. Annals of Neurology, 45(4), 495-503.
Benasich, A.A., & Tallal, P. (2002). Infant discrimination of rapid auditory cues predicts later
language impairment. Behavioural Brain Research, 136, 31-49.
Bishop, D. V., Bishop, S. J., Bright, P., James, C., Delaney, T., & Tallal, P. (1999). Different
origin of auditory and phonological processing problems in children with language
impairment: evidence from a twin study. Journal of Speech, Language and Hearing
Research, 42(1), 155-168.
Bishop, D.V.M., & McArthur, G.M. (2004). Immature cortical responses to auditory stimuli
in specific language impairment: evidence from ERPs to rapid tone sequences.
Developmental Science, 7 (4), F11-F18.
42
Bishop, D.V.M., & McArthur, G.M. (2005). Individual differences in auditory processing in
specific language impairment: A follow-up study using event-related potentials and
behavioural thresholds. Cortex, 41(3), 327-341.
Cacace, A. T., McFarland, D. J., Ouimet, J. R., Schrieber, E. J., & Marro, P. (2000). Temporal
processing deficits in remediation-resistant reading-impaired children. Audiology and
Neurootology, 5(2), 83-97.
Carney, L. H., Heinz, M. G., Evilsizer, M. E., Gilkey, R. H., & Colburn, H. S. (2002).
Auditory phase opponency: A temporal model for masked detection at low
frequencies. Acta Acustica, 88, 334-347.
de Jong, P. F., & van der Leij, A. (1999). Specific contributions of phonological abilities to
early reading acquisition: Results from a Dutch latent variable longitudinal study.
Journal of Educational Psychology, 91(3), 450-476.
de Jong, P. F., Seveke, M. J., & van Veen, M. (2000). Phonological sensitivity and the
acquisition of new words in children. Journal of Experimental Child Psychology,
76(4), 275-301.
de Jong, P. F., & van der Leij, A. (2003). Developmental changes in the manifestation of a
phonological deficit in dyslexic children learning to read a regular orthography.
Journal of Educational Psychology, 95(1), 22-40.
Dougherty, R. F., Cynader, M. S., Bjornson, B. H., Edgell, D., & Giaschi, D. E. (1998).
Dichotic pitch: A new stimulus distinguishes normal and dyslexic auditory function.
Neuroreport, 9(13), 3001-3005.
Elbro, C., Borstrom, I., & Petersen, D. (1998). Predicting dyslexia from kindergarten: The
importance of distinctness of phonological representations of lexical items. Reading
Research Quarterly, 33(1), 36-60.
Elbro, C., & Scarborough, H.S. (2003). Early identification. In P. Bryant & T. Nunes (Eds.),
Handbook of children’s reading (pp. 339-359). Dordrecht: Kluwer Academic
Publishers.
43
Fischer, B., & Hartnegg, K. (2004). On the development of low-level auditory discrimination
and deficits in dyslexia. Dyslexia, 10(2), 105-118.
Gallagher, A., Frith, U., & Snowling, M. J. (2000). Precursors of literacy delay among
children at genetic risk of dyslexia. Journal of Child Psychology and Psychiatry,
41(2), 203-213.
Gathercole, S., & Pickering, S. (2000). Assessment of working memory in six- and seven-
year-old children. Journal of Educational Psychology, 92, 377-390.
Gathercole, S. E., Willis, C. S., Baddeley, A. D., & Emslie, H. (1994). The Children's Test of
Nonword Repetition: A test of phonological working memory. Memory, 2(2), 103-
127.
Gilger, J. W., Pennington, B. F., & DeFries, J. C. (1991). Risk for reading disability as a
function of parental history in three family studies. Reading and Writing, 3(3-4), 205-
217.
Hautus, M. J., Setchell, G. J., Waldie, K. E., & Kirk, I. J. (2003). Age-related improvements
in auditory temporal resolution in reading-impaired children. Dyslexia, 9, 37-45.
Heath, S. M., & Hogben, J. H. (2004). Cost-effective prediction of reading difficulties.
Journal of Speech, Language and Hearing Research, 47, 751-765.
Hill, N. I., Bailey, P. J., Griffiths, Y. M., & Snowling, M. J. (1999). Frequency acuity and
binaural masking release in dyslexic listeners. Journal of the Acoustical Society of
America, 106(6), 53-58.
Hulslander, J., Talcott, J., Witton, C., DeFries, J., Pennington, B., Wadsworth, S., Willcutt, E.,
& Olson, R. (2004). Sensory processing, reading, IQ and attention. Journal of
Experimental Child Psychology, 88, 274-295.
Irwin, R. J., Ball, A. K., Kay, N., Stillman, J. A., & Rosser, J. (1985). The development of
auditory temporal acuity in children. Child Development, 56(3), 614-620.
Jensen, J. K., & Neff, D. L. (1993). Development of basic auditory discrimination in
preschool children. Psychological Science, 4(2), 104-107.
44
Laneau, J., Boets, B., Moonen, M., van Wieringen, A., & Wouters, J. (2005). A flexible
auditory research platform using acoustic or electric stimuli for adults and young
children. Journal of Neuroscience Methods, 142(1), 131-136.
Lefly, D. L., & Pennington, B. F. (2000). Reliability and validity of the Adult Reading
History Questionnaire. Journal of Learning Disabilities, 33(3), 286-296.
Levitt, H. (1971). Transformed up-down methods in psychoacoustics. Journal of the
Acoustical Society of America, 49, 467-477.
Littell, R. C., Milliken, G. A., Stroup, W. W., & Wolfinger, R. D. (1996). SAS system for
mixed models. Cary, NC: Sas Institute Inc.
Lyytinen, H., Ahonen, T., Eklund, K., Guttorm, T. K., Laakso, M. L., Leinonen, S.,
Leppanen, P. H., Lyytinen, P., Poikkeus, A. M., Puolakanaho, A., Richardson, U., &
Viholainen, H. (2001). Developmental pathways of children with and without familial
risk for dyslexia during the first years of life. Developmental Neuropsychology, 20(2),
535-554.
McAnally, K. I., & Stein, J. F. (1996). Auditory temporal coding in dyslexia. Proceedings of
the Royal Society of London. Series B, Biological Sciences, 263, 961-965.
McAnally, K. I., & Stein, J. F. (1997). Scalp potentials evoked by amplitude-modulated tones
in dyslexia. Journal of Speech, Language and Hearing Research, 40(4 ), 939-945.
McArthur, G.M., & Bishop, D.V.M. (2004). Which people with specific language impairment
have auditory deficits? Cognitive Neuropsychology, 21 (1), 79-94.
McBride-Chang, C. (1996). Models of speech perception and phonological processing in
reading. Child Development, 67(4), 1836-1856.
McCroskey, R. L., & Kidder, H. C. (1980). Auditory fusion among learning disabled, reading
disabled, and normal children. Journal of Learning Disabilities, 13(2 ), 69-76.
Menell, P., McAnally, K. I., & Stein, J. F. (1999). Psychophysical sensitivity and
physiological response to amplitude modulation in adult dyslexic listeners. Journal of
Speech, Language and Hearing Research, 42(4), 797-803.
45
Molfese, D. L. (2000). Predicting dyslexia at 8 years of age using neonatal brain responses.
Brain and Language, 72(3 ), 238-245.
Moore, B. C. J. (1997). An introduction to the psychology of hearing. London: Academic
Press Limited.
Morrongiello, B. A., Kulig, J. W., & Clifton, R. K. (1984). Developmental changes in
auditory temporal perception. Child Development, 55(2), 461-471.
Oh, E.L., Wightman, F.L., & Lutfi, R.A. (2001). Children’s detection of pure-tone signals
with random multitone maskers. Journal of the Acoustical Society of America, 109,
2888-2895.
Pennington, B. F., & Lefly, D. L. (2001). Early reading development in children at family risk
for dyslexia. Child Development, 72(3), 816-833.
Ramus, F., Rosen, S., Dakin, S. C., Day, B. L., Castellote, J. M., White, S., & Frith, U.
(2003). Theories of developmental dyslexia: insights from a multiple case study of
dyslexic adults. Brain, 126(4), 841-865.
Raven, J. C., Court, J. H., & Raven, J. (1984). Manual for Raven's Progressive Matrices and
Vocabulary Scales. London: Lewis.
Rocheron, I., Lorenzi, C., Fullgrabe, C., & Dumont, A. (2002). Temporal envelope perception
in dyslexic children. Neuroreport, 13(13), 1683-1687.
Rolf, P. C., & Van Rijnsoever, R. J. (1984). Positionele letterfrekwenties van het Nederlands.
Lisse: Swets & Zeitlinger.
Scarborough, H. S. (1989). Prediction of reading disability from familial and individual
differences. Journal of Educational Psychology, 81(1), 101-108.
Scarborough, H. S. (1990). Very early language deficits in dyslexic children. Child
Development, 61(6), 1728-1743.
Scarborough, H. S. (1998). Early identification of children at risk for reading disabilities:
phonological awareness and some other promising predictors. In P.J. Accardo, A.J.
Capute & B.K. Shapiro (Eds.), Specific reading disability: a view of the spectrum (pp
46
75-119). Timonium, MD: York Press.
Scheltinga, F. (2003). The Dutch Nonword Repetition Test. Unpublished manuscript,
University of Amsterdam.
Schneider, B. A., & Trehub, S. E. (1992). Sources of developmental change in auditory
sensitivity. In L. A. Werner , & E. W. Rubel (Eds.), Developmental psychoacoustics
(pp. 3-46). Washington DC: American Psychological Association.
Schulte-Körne, G., Deimel, W., Bartling, J., & Remschmidt, H. (1998). Role of auditory
temporal processing for reading and spelling disability. Perceptual and Motor Skills,
86(3), 1043-1047.
Share, D. L., Jorm, A. F., Maclean, R., & Matthews, R. (2002). Temporal processing and
reading disability. Reading and Writing, 15(1-2), 151-178.
Snowling, M. J. (2000). Dyslexia (2nd ed.). Malden, MA: Blackwell Publishers.
Soderquist, D.R., & Shilling, R.D. (1992). Forward masking in young children – rocket ship
psychophysics. Current Psychology: Research & Reviews, 11(2), 145-156.
Stein, J. F., & McAnally, K. (1995). Auditory temporal processing in developmental
dyslexics. Irish Journal of Psychology, 16(3), 220-228.
Studdert-Kennedy, M., & Mody, M. (1995). Auditory temporal perception deficits in the
reading-impaired: A critical review of the evidence. Psychonomic Bulletin and
Review, 2(4), 508-514.
Talcott, J. B., Witton, C., McClean, M., Hansen, P. C., Rees, A., Green, G., & Stein, J. F.
(1999). Can sensitivity to auditory frequency modulation predict children's
phonological and reading skills? Neuroreport, 10(10), 2045-2050.
Talcott, J. B., Witton, C., McLean, M. F., Hansen, P. C., Rees, A., Green, G. G., & Stein, J. F.
(2000). Dynamic sensory sensitivity and children's word decoding skills. Proceedings
of the National Academy of Science, USA, 97( 6), 2952-2957.
Talcott, J. B., Witton, C., Hebb, G. S., Stoodley, C. J., Westwood, E. A., France, S. J.,
Hansen, P. C., & Stein, J. F. (2002). On the relationship between dynamic visual and
47
auditory processing and literacy skills; results from a large primary-school study.
Dyslexia, 8(4), 204-225.
Tallal, P., & Piercy, M. (1973). Defects of nonverbal auditory perception in children with
developmental aphasia. Nature, 241, 468-469.
Tallal, P., & Piercy, M. (1975). Developmental aphasia: Rate of auditory processing and
selective impairment of consonant perception. Neuropsychologia, 13, 69-74.
Tallal, P. (1980). Auditory temporal perception, phonics, and reading disabilities in children.
Brain and Language, 9(2), 182-198.
Tallal, P., Miller, S., & Fitch, R. H. (1993). Neurobiological basis of speech: a case for the
preeminence of temporal processing. Annals of the New York Academy of Sciences,
682, 27-47.
Thompson, N.C., Cranford, J.L. & Hoyer, E. (1999). Brief-tone frequency discrimination by
children. Journal of Speech, Language and Hearing Research, 42, 1061-1068.
Trehub, S. E., Schneider, B. A., & Henderson, J. L. (1995). Gap detection in infants, children,
and adults. Journal of the Acoustic Society of America, 98(5 Pt 1), 2532-2541.
van den Bos, K. P., Zijlstra, B. J. H., & Spelberg, H. C. (2002). Life-span data on continuous-
naming speeds of numbers, letters, colors, and pictured objects, and word-reading
speed. Scientific Studies of Reading, 6(1), 25-49.
Van Ingelghem, M., van Wieringen, A., Wouters, J., Vandenbussche, E., Onghena, P., &
Ghesquière, P. (2001). Psychophysical evidence for a general temporal processing
deficit in children with dyslexia. Neuroreport, 12(16), 3603-3607.
Van Ingelghem, M., Boets, B., van Wieringen, A., Ghesquière, P., & Wouters, J. (2004).
Auditory and visual temporal processing in dyslexic and normal reading children
matched for sex, age and intellectual ability. Paper presented at the 6th BDA
International conference, 27-30 march 2004, University of Warwick, UK .
Van Ingelghem, M., Boets, B., van Wieringen, A., Onghena, P., Ghesquière, P., & Wouters, J.
(2005). An auditory temporal processing deficit in children with dyslexia? In P.
Ghesquière & A.J.J.M. Ruijssenaars (Eds.), Children with learning disabilities: a
48
challenge to teaching and instruction. Series: Studia Paedagogica. (pp 47-63).
Leuven: University Press.
van Wieringen, A., & Wouters, J. (1999). Gap detection in single- and multiple- channel
stimuli by LAURA cochlear implantees. Journal of the Acoustical Society of America,
106, 1925-1939.
Verbeke, G., & Lesaffre, E. (1997). The effect of misspecifying the random-effects
distribution in linear mixed models for longitudinal data. Computational Statistics and
Data Analysis, 23, 541-556.
Wagner, R. K., & Torgesen, J. K. (1987). The nature of phonological processing and its
causal role in the acquisition of reading skills. Psychological Bulletin, 101(2), 192-
212.
Wechsler, D. (1992). Wechsler Intelligence Scale for Children (3rd UK Edition). London:
The Psychological Corporation.
Wightman, F., Allen, P., Dolan, T., Kistler, D., & Jamieson, D. (1989). Temporal resolution
in children. Child Development, 60(3), 611-624.
Wightman, F., & Allen, P. (1992). Individual differences in auditory capability among
preschool children. In L.A. Werner , & E.W. Rubel (Eds.), Developmental
Psychoacoustics (pp. 113-134). Washington DC: American Psychological
Association.
Wightman, F.L., Callahan, M.R., Lutfi, R.A., Kistler, D.J., & Oh, E. (2003). Children’s
detection of pure-tone signals: informational masking with contralateral maskers.
Journal of the Acoustical Society of America, 113 (6), 3297-3305.
Witton, C., Talcott, J. B., Hansen, P. C., Richardson, A. J., Griffiths, T. D., Rees, A., Stein, J.
F., & Green, G. G. (1998). Sensitivity to dynamic auditory and visual stimuli predicts
nonword reading ability in both dyslexic and normal readers. Current Biology, 8(14),
791-797.
Wright, B.A., & Zecker, S.G. (2004). Learning problems, delayed development and puberty.
Proceedings of the National Academy of Science, USA, 101 (26), 9942-9946.
49
Notes
1 Published as Boets, B. Wouters, J., van Wieringen, A., & Ghesquière, P. (2006). Auditory
temporal information processing in preschool children at family risk for dyslexia: Relations
with phonological abilities and developing literacy skills. Brain and Language, 97, 64-79. 2 In the Belgian school system formal instruction starts in Grade 1 at six years. This means in
kindergarten no reading instruction is offered. This is in contrast to the kindergarten group
studied by Share et al. (2002), who already received formal reading instruction. 3 It is worth mentioning that for both HR and LR-group, the mean scores on the sound identity
tasks are well above chance level. This is in contrast to a Dutch study where an odd-one-out
categorization version of these tests was administered and where children of the same age
were not able to exceed chance level on the last-sound and first-sound categorization test (de
Jong & van der Leij, 1999). 4 Interestingly, this change in significance was caused almost completely by the influence of
maternal educational level. 5 For instance, regarding frequency discrimination, an almost linear improvement has been
observed from 4 to 9 years, after which adult levels of performance are obtained (Jensen &
Neff, 1993; Thompson et al., 1999). Similarly, with respect to GAP detection thresholds,
Wightman et al. (1989) and Irwin and colleagues (1985) observed a sharp maturational
improvement from 3 years on up to 7 years and extending slightly up to 11 years before
reaching adult levels.
51
Manuscript 2
Speech perception in preschoolers at family risk for dyslexia:
Relations with low-level auditory processing and phonological ability1
Abstract
We tested categorical perception and speech-in-noise perception in a group of 5-year-old preschool children
genetically at risk for dyslexia, compared to a group of well-matched control children and a group of adults.
Both groups of children differed significantly from the adults on all speech measures. Comparing both child
groups, the risk group presented a slight but significant deficit in speech-in-noise perception, particularly in the
most difficult listening condition. For categorical perception a marginally significant deficit was observed on the
discrimination task but not on the identification task. Speech parameters were significantly related to
phonological awareness and low-level auditory measures. Results are discussed within the framework of a causal
model where low-level auditory problems are hypothesized to result in subtle speech perception problems that
might interfere with the development of phonology and reading and spelling ability.
Introduction
Every beginning reader of an alphabetic writing system is confronted with the
challenge to learn to segment the quasi-continuous speech signal in distinct and specified
phoneme units in order to map them upon their corresponding graphemes. Whereas the
majority of readers fairly easily establishes stable grapheme to phoneme relations, there is a
small group of about 5 to 10 percent of children and adults that is not able to master fluent
reading and spelling skills despite conventional instruction, adequate intelligence and
sufficient socio-cultural opportunities. In recent years, there has been a growing consensus
that these reading and spelling impairments, characteristic of dyslexia, are related to specific
deficits in the representation and use of phonological information (for a review see Rack,
1994; Snowling, 2000). Phonological deficits have been demonstrated in three broad areas
52
(Wagner & Torgesen, 1987): phonological awareness (e.g. Manis, Custodio & Szeszulski,
1993), retrieval of phonological codes from long-term memory (rapid automatic naming) (e.g.
Bowers & Swanson, 1991; Wolf 1986), and verbal short-term memory (e.g. Catts, 1989).
Moreover, several prospective longitudinal studies have suggested a causal link between
sensitivity to the phonological structure of words and later progress in reading acquisition
(e.g. Wagner, Torgesen & Rashotte, 1994), although the relationship between reading and
phonological awareness is probably reciprocal (Morais, Bertelson, Cary & Alegria, 1979).
Research in the underlying neurological dysfunction of dyslexia suggests that the
phonological processing deficits as such may result from a more fundamental deficit in the
basic perceptual mechanisms that are responsible for auditory information processing.
Dyslexics tend to have difficulties processing linguistic and non-linguistic stimuli that are
short and enter the nervous system in rapid succession (for reviews see e.g. Farmer & Klein,
1995; Habib, 2000; McArthur & Bishop, 2001). Recent studies in this context focus more
specifically on an impaired perception of dynamic aspects in the auditory signal itself, like
amplitude and frequency modulations (Menell, McAnally & Stein, 1999; Talcott et al., 1999;
Talcott et al., 2000; Talcott et al., 2002; Van Ingelghem et al., 2005; Witton et al., 1998). In
these studies, persons with dyslexia show difficulties tracking acoustic changes over time.
Besides, subjects with reading impairments also tend to have difficulties with speech
perception (for a review see McBride-Chang, 1995). In particular, it has been suggested that
they have difficulty extracting discrete phonological representations from phonetic features
embedded within the speech signal (Manis et al., 1997; Serniclaes, Sprenger-Charolles, Carré
& Demonet, 2001). The resulting deficient phonological representations can be expected to
interfere with or delay the process of becoming phonemically aware, which in turn can lead to
problems in learning to read or write (Joanisse, Manis, Keating, & Seidenberg, 2000;
McBride-Chang, 1996; Watson & Miller, 1993). In this way, even subtle speech perception
problems might have considerable consequences for learning to read (Werker & Tees, 1987).
Since speech perception also heavily depends upon adequate tracking of fast changing
transitions in frequency and amplitude, there could very well be a causal path relating the
observed deficits in auditory processing, speech perception and phonological ability. More
concretely, it has been hypothesised that dyslexic’s low-level auditory processing problems
might result in a subtle speech perception problem, which might interfere with the
development of fine-grained phonological representations, which in turn impacts upon
reading and spelling ability (Talcot & Witton, 2002; Tallal, 1980; 1984; Wright et al., 1997).
53
Even though this causal hypothesis is widely recognized (and hotly debated, see e.g.
Blomert & Mitterer, 2004; Denenberg, 1999; Mody, Studdert-Kennedy & Brady, 1997;
Nittrouer, 1999; Rosen & Manganari, 2001), only few studies have examined the
developmental pathway exploring the relations between these different skills in detail. Some
major prospective studies are coming about (e.g. Jyväskylä Longitudinal study of Dyslexia,
see Lyytinen et al., 2001; Dutch Dyslexia Research Program, NWO, 1996), but until now
only limited reports have been published. Here, we report data from a longitudinal study that
explores the development of basic auditory skills, speech perception, phonological abilities
and reading skills in a group of 5-year-old preschool children genetically at risk for dyslexia,
compared to a group of well-matched control children. In a previous paper (Boets, Wouters,
van Wieringen & Ghesquière, 2006), we already reported the absence of a significant group
difference for any of three administered auditory measures (i.e. gap-in-noise detection,
frequency modulation detection and tone-in-noise detection), in the presence of a significant
difference for phonological awareness and letter knowledge. Moreover, spectral auditory
tasks (particularly 2 Hz frequency modulation detection) turned out to be highly significantly
related to phonological awareness. In this paper, we will focus upon speech perception
abilities, assessed in the same group of preschool children. In particular, we consider the
question whether a deficit in speech perception may already be observable in preschool
children at risk for dyslexia and we investigate the relationship between speech perception
and auditory and phonological processing.
Dyslexia research has mainly used two major experimental paradigms to study speech
perception: (i) the perception of speech presented in background noise and (ii) categorical
perception of stop consonants.
Although it is widely acknowledged (and there is much supporting anecdotal
evidence) that subjects with reading problems have particular difficulty with the perception of
speech under noisy listening conditions, so far only a limited number of empirical studies
have addressed this issue in well-defined clinical groups of subjects with dyslexia. Brady,
Shankweiler and Mann (1983) showed that nine-year-old poor readers had more difficulty in
repeating words presented in noise than age-matched controls, though both groups did not
differ in the perception of speech without noise. Similarly, Chermak, Vonhof and Bendel
(1989) found that adults with broadly defined learning disabilities had poorer word
identification in noise than a control group of adults without learning disabilities. The same
results were obtained by Watson and Miller (1993), presenting CV syllables in a background
54
of cafeteria noise. Two studies presenting meaningful sentences in noise found that children
with hearing or language impairments (Stollman, Kapteyn & Sleeswijk, 1994) or with
language-based learning disabilities (Bradlow, Kraus & Hayes, 2003) showed poorer overall
sentence-in-noise perception than a non-affected control group. Moreover, both studies
demonstrated that children with language problems were more affected by more adverse
listening conditions than the control group.
Recently, some neurophysiologic studies also provided evidence for anomalies in
dyslexics’ neural encoding of speech stimuli embedded in background noise. Wible, Nicol
and Kraus (2002) demonstrated a deficiency in the neural representation of repeated speech
stimuli in noise but not in quiet in 11-year old children diagnosed with reading problems.
Moreover, the robustness of the physiologic response in noise was significantly correlated
with behavioral measures of speech discrimination, with a phonological composite measure
and with spelling ability. In the same way, Cunningham, Nicol, Zecker, Bradlow and Kraus
(2001) reported that a group of reading disabled children (aged 10 to 13 years) showed poorer
speech discrimination coupled with diminished neurophysiologic cortical and subcortical
responses in background noise compared to normal children. There were no differences
between the groups in quiet listening conditions, neither in behavioral perception, nor in
electrophysiology. Similar results, indicative of abnormal timing of cortical responses to
speech stimuli in background noise in learning-impaired children, were reported by Warrier
and colleagues (Warrier, Johnson, Hayes, Nicol, & Kraus, 2004). Moreover, they
demonstrated that a commercial phonological/language training program not only enhanced
their phonetic decoding skills, but also improved the timing of the cortical responses in these
children.
Taken together, even though these studies are diffuse in terms of methodology and
participant inclusion criteria, the findings indicate that subjects with dyslexia or more broadly
defined learning or language impairments have problems perceiving speech in the presence of
background noise. However, evidence of a speech-in-noise perception deficit is not yet
unequivocal since some studies failed to find a discrepancy (Pennington, Van Orden, Smith,
Green & Haith, 1990; Snowling, Goulandris, Bowlby & Howell, 1986) or could only
demonstrate it under very specific noisy listening conditions (viz. intermittent noise, see
Bailey, Griffiths, Hill, Brent, & Snowling, 2002).
The categorical perception experiments are based on the general principle that
perceptual discrimination between speech sounds belonging to different phoneme categories
55
is better than that between sounds falling within the same category, irrespective of the
magnitude of the acoustic (physical) difference between the sounds (Liberman, Harris,
Hoffman & Griffith, 1957). In identification experiments this is expressed in the
discontinuous perception of a series of speech stimuli that vary continuously along a single
acoustic dimension. Normal listeners exhibit a strong tendency toward categorical perception
of speech along certain phonetic dimensions, such as place of articulation (e.g. /ba/ vs. /da/) or
voicing status (e.g. /ba/ vs. /pa/). Several studies suggest that children and adults with dyslexia
are less categorical than average readers in the way they perceive phonetic contrasts (Breier,
Fletcher, Denton & Gray, 2004; Godfrey, Syrdal-Lasky, Millay & Knox, 1981; Lieberman,
Meskill, Chatillon & Schupack, 1985; Maassen et al., 2001; Manis et al., 1997; Reed, 1989;
Steffens, Eilers, Gross-Glenn & Jallad, 1992; Serniclaes et al., 2001; van Beinum et al., 2005;
Werker & Tees, 1987). Some studies could even demonstrate a weaker categorical perception
in toddlers (Gerrits, 2003) and six-month-old infants (Richardson, Leppänen, Leiwo &
Lyytinen, 2003) at family risk for dyslexia (see also Benasich & Tallal, 2002, and Molfese,
2000, for similar speech perception studies in infants). However, a minority of studies
obtained non-significant results (Blomert & Mitterer, 2004; Brandt & Rosen, 1980; Nittrouer,
1999) or only observed the categorical perception deficit in the most severely phonologically
or linguistically impaired dyslexics (Joanisse et al., 2000; Manis et al., 1997). Interestingly,
Serniclaes and colleagues (Serniclaes et al., 2001; Serniclaes, Van Heghe, Mousty, Carré, &
Sprenger-Charolles, 2004) demonstrated that in addition to a weaker discrimination between
categories, dyslexics presented a higher sensitivity for acoustic differences within the same
phoneme category. This higher sensitivity for ‘allophones’ (i.e. variants of the same phoneme
category) indicates that the internal structure of the categories is less coherent. Whereas the
perception of speech with allophonic rather than phonemic categories probably would not
raise major problems, it might pose a problem for the acquisition of written language (at least
in alphabetic systems). Indeed, if the phonological representations are not well specified, the
one-to-one relation between graphemes and phonemes will be difficult to establish.
To understand the speech perception problem in dyslexia, Serniclaes et al. (2004) and
Maassen et al. (2001) suggested looking at it from a developmental perspective. It has been
demonstrated that newborns possess predispositions for discriminating all potential phoneme
categories in the world’s languages; they can even discriminate among a range of phonetic
categories that are not phonemic in their own mother language. However, after exposure to
the sounds of their native language, they gradually obtain higher sensitivity for language-
specific phonetic cues (by assigning different weights to different acoustic parameters) and
56
simultaneously they lose their sensitivity to the boundaries that are not linguistically relevant
(MacKain, Best & Strange, 1981; Nittrouer, 1996; Werker & Tees, 1984). Whereas in
normally developing children these irrelevant predispositions are usually deactivated by about
the age of one year, it has been suggested that in dyslexics this ‘weighting shift’ has not been
completed adequately, resulting in the persistence of more or differently divided categories
than are necessary for perceiving phonemes (Serniclaes et al., 2004).
Notwithstanding the empirical evidence that speech perception measures are able to
differentiate relatively reliably between adult and school-aged dyslexic and normal reading
subjects, the differentiating and predictive power of these tasks has been studied only in a
limited way in preschoolers. In this study we want to deal with this issue by examining
categorical perception and speech-in-noise perception in two contrasting groups of preschool
children. Moreover, by correlating the speech data to previously collected auditory and
phonological data, we aim to shed light on the relation between low-level auditory processing,
speech perception, phonological skills and literacy development. Although theoretically it has
been argued that there might be a causal relation between these abilities, it is important to
note that with this study we do not aim to establish whether the relation is causal in any way.
Nevertheless, by assessing all these abilities in the same children, this study is one of the first
to empirically investigate the hypothesized interrelations. Moreover, by studying preschool
children who never received any formal reading instruction, we can rule out the possibility
that observed group differences or interrelations are merely a consequence of differences in
literacy skills or reading experience.
Method
Participants
Sixty-two five-year-old children attending the last year of kindergarten were included
in the study (36 boys / 26 girls). Half of the participants were children of ‘dyslexic families’,
the so-called high-risk group (HR); the other half were control children of ‘normal reading
families’, the so-called low-risk group (LR). Since dyslexia tends to run strongly in families,
preschoolers with dyslexic relatives are more likely to develop reading problems. Gilger,
Pennington and DeFries (1991) estimate that roughly between 30 and 50 % of such children
will become reading disabled. Moreover, in view of the fact that the HR group already
57
demonstrated a significant deficit in phonological awareness and letter knowledge, and since
these measures have consistently been proven to be the best single preschool predictors of
literacy development (Scarborough, 1998), it is very likely that this group will contain a
disproportional high number of future cases of dyslexia.
All children were native Dutch speakers without any history of brain damage, long
term hearing loss or visual problems. Additionally, at the moment of data collection they did
not present any gross hearing deficiencies (audiometric pure-tone average < 25 dB HL). The
HR children were selected on a basis of having at least one first-degree relative with a
diagnosis of dyslexia. The LR children showed no history of speech or language problems
and none of their family members suffered any learning or language problem. For every
individual HR child the best matching LR control child was selected based on five criteria:
(1) educational environment, i.e. same nursery school, (2) gender, (3) age, (4) nonverbal
intelligence, and (5) parental educational level. Nonverbal intelligence was assessed by an
adapted version of the Raven Coloured Progressive Matrices (RCPM) (Raven, Court, &
Raven, 1984), a collective non-verbal intelligence test measuring spatial reasoning. Parental
educational level was assessed using the ISCED-scale (International Standard Classification
of Education by UNESCO, 1997), by converting classifications on the original seven-point
scale to a three-point scale. At the time of collecting the speech data, the mean age for both
the HR and LR group was 5 years and 8 months, not being statistically different [paired t(30)
= 0.22, p = .83]. The nonverbal IQ scores were slightly above population average (107 for HR
group and 111 for LR group) and did not differ significantly [paired t(30) = 1.88, p = .07].
Fisher’s Exact Test also confirms that both groups did not differ in frequency distribution of
the different educational categories (p = .71 for maternal and p = .43 for paternal educational
level). Further details about the participants and the selection procedure are described in Boets
et al., (2006).
For reasons of comparison, the speech tests were also administered to a group of
eleven normal hearing and reading adults, aged 24-43 years (5 males / 6 females).
58
Tasks and materials
Tests for speech perception
In order to assess the speech processing capacities, two speech perception tasks were
administered: one measuring speech-in-noise perception and one measuring categorical
speech perception.
Speech-in-noise perception task. In the speech-in-noise perception task, seven lists of
ten high-frequency monosyllabic words (taken from the Göttingerlist II, recorded by Wouters,
Damman & Bosman, 1994; see also van Wieringen & Wouters, 2005) were presented
monaurally with an inter-stimulus interval of 7 seconds. Simultaneously, a continuous
stationary speech noise with an identical spectrum as the average spectrum of the word lists
was presented to the same ear, at a fixed level of 70 dB SPL. Words were presented at -1, -4
and –7 dB signal-to-noise ratio (SNR). Before administration of the six test lists (3 x 2), one
list was presented at an SNR of +4 dB as a practice list. The child’s task was to repeat the
words as accurately as possible, resulting in a percentage correct word score for every test list.
Categorical perception. A ten-point speech continuum ranging from /bAk/ to /dAk/
(i.e. the Dutch word for ‘box / basket’ and ‘roof’) was presented by means of a categorical
perception task. For every child, a 2AFC forced-choice identification task was administered,
followed by a same-different (AX) discrimination task. Stimuli were identical to the ones
developed and described by van Beinum et al. (2005). Stimuli were based on natural speech
and were constructed by linearly interpolating the transition of the second formant (F2) from
/b/ to /d/. The manipulated part of the signal was a 100 ms interval at the beginning of the
vowel. There were ten interpolation steps and the F2 onset ranged from 1100 to 1800 Hz. The
/bAk/-starting point was characterized by a constant F2 value of 1100 Hz. For each
intermediate signal, the F2 onset was gradually situated at higher frequencies, making the fall
of the transition steeper in every step. At the/d/-endpoint of the continuum the transition was
falling from 1800 Hz to 1100 Hz. The interpolation resulted in a 10-point continuum, the total
length of each item being 600 ms, consisting of (a) a vocal murmur 170 ms; (b) burst 10 ms;
(c) vowel /A/ 250 ms, divided into 100 ms transition duration and 150 ms steady state; (d)
occlusion period [k] (silence) 55 ms; (e) release [k] 115 ms (van Beinum et al., 2005). Stimuli
were presented monaurally at a comfortable listening level of 70 dB SPL.
59
In the identification task, subjects were sitting in front of a computer screen, depicting
a picture of a basket and a roof, representing the possible response options (/bAk/ vs. /dAk/).
The ten stimuli of the continuum were presented twelve times in a random order, the same
stimulus not being presented more than two consecutive times. The child was instructed to
repeat the perceived word and point to the corresponding picture. Participants were given an
unlimited time to respond. No direct feedback was given. The task was preceded by a pre-test,
presenting each of the two endpoint stimuli five times.
In the discrimination task, subjects were instructed to listen to two stimuli presented
with an ISI of 600 ms and to determine whether they sounded ‘the same’ or ‘different’.
Children had to respond by pointing to one of two pictures on the screen representing “the
same” or “different”. No direct feedback was given. In order to obtain a bias-free measure of
discriminability, the task contained physically different as well as identical pairs. The
different pairs, comprising each of the seven 3-step comparisons (1-4, 2-5, 3-6, etc), were
presented 6 times; the identical pairs (1-1, 2-2, etc) were presented twice. The pairs of stimuli
were presented in two blocks of 31 stimuli (3x7 + 1x10). Presentation order of the stimuli was
randomized and the internal order of a different stimulus pair was balanced over both series.
The task was preceded by 10 practice items, comprising pairs that were clearly different or
identical.
Stimulus presentation and response registration was controlled by E-prime, a software
module for psychological experiments (Schneider, Eschman & Zuccolotto, 2002). In order to
make the rather boring psychophysical tests more interesting and child friendly, they were
integrated in an interactive videogame with an introductory and concluding movie and
animated buttons. Prior to testing, the child was presented an animation movie about a cat
waking up in a basket (= /bAk/ in Dutch) and climbing on the roof (= /dAk/ in Dutch). After
the cat disappearing out of view, the movie explained that the child could know where the cat
actually is – in the /bAk/ or on the /dAk/ -by listening very carefully.
For both the categorical perception task and the speech-in-noise perception task, as
well as for the other auditory psychophysical tasks, stimuli were presented monaurally over
calibrated TDH-39 headphones using the integrated audio PC-card from a portable computer
routed to an audiometer (Madsen OB622).
60
Phonological tests
Phonological skills were assessed by a test battery comprising eight tests (see Boets et
al., 2006). A principal component factor analysis with varimax rotation confirmed that the
battery reflected the three traditional phonological domains: (a) phonological awareness: high
loadings from first-sound and end-sound identity task, rhyme identity task and simple rhyme
task, (b) rapid automatic naming: high loadings from both the colors and objects rapid
automatic naming tasks and (c) verbal short-term memory: high loadings from the non-word
repetition task and the digit span forward task. Details about the composition of the battery
can be found in Boets et al. (2006); in this paper we will only refer to the factor scores.
Tests for low-level auditory processing
Auditory processing was assessed by means of three psychophysical threshold tests:
gap-detection in noise (GAP), 2 Hz frequency modulation detection (FM) and tone-in-noise
detection (TN) (see Boets et al., 2006). In the GAP-detection test, subjects had to detect a
silent interval (gap) in a white noise stimulus. Threshold was defined as the minimum gap
length required for detecting the silent interval. In the FM-detection test participants had to
detect a 2 Hz sinusoidal frequency modulation of a 1 kHz carrier tone with varying
modulation depth. Threshold was defined as the minimum depth of frequency deviation
required to detect the modulation. In the TN-detection task participants had to detect two pure
tone pulses (1 kHz, length = 440 ms) within a one-octave noise signal, centered around 1 kHz
(from 707 to 1414 Hz, length = 1620 ms). Threshold was defined as the lowest signal-to-noise
ratio (SNR) required for detecting the tone pulses. For all three psychophysical tests,
thresholds were estimated using a forced-choice adaptive staircase paradigm embedded within
an interactive computer game (Laneau, Boets, Moonen, van Wieringen & Wouters, 2005).
For each participant three threshold estimates were determined for every experiment. For the
correlation data we present here, the average of the best and second best threshold was used as
an indicator of auditory sensitivity. A detailed description of the stimuli, procedure and results
can be found in Boets et al. (2006).
Data collection was carried out by qualified psychologists and audiologists. Testing
took place in a quiet room at the children’s school. Since the LR child was selected from the
HR child’s classmates, both children could always be tested in exactly the same
circumstances.
61
Statistical analysis
First, both groups of children were compared to each other. To account for the
clustered nature of the data (i.e. matched pairs), data were analyzed in a pair wise manner
using Mixed Model Analysis (MMA) (Littell, Stroup & Freund, 2002)2. Although both
groups did not show a significant difference on any of the matching criteria, any possible
influence of age, nonverbal intelligence or parental educational level was ruled out by
additionally controlling for these variables. As such, a series of (repeated) MMA’s was
calculated with pair number as a random variable (1 to 31) and participant group (HR versus
LR) as the fixed between-subject variable3. Age, nonverbal IQ and educational level of both
mother and father were added as fixed (co)variables. Second, child and adult data were
compared to each other using MMA as well. For this comparison no random factor or
covariables were specified. Post-hoc analyses were corrected for multiple comparisons using
Tukey procedure. Third, to extract the essential parameters, the speech-in-noise data and the
identification data were submitted to a logistic fitting and groups were compared on the
resulting parameters. Prior to analysis, the slope-parameters of these curves were log10-
transformed to obtain normally distributed data. In order to take into account the variable
quality of the fits, the inverse standard error of the estimated parameters was added to the
model as a weight variable. Finally, in order to determine the relations between speech
perception on the one hand and auditory and phonological processing on the other, Pearson
correlations were calculated, partialed out for individual differences in nonverbal intelligence
and with the inverse standard error of the estimated parameters added as a weight variable.
Results
Speech-in-noise perception
Prior to analysis, data from one subject were removed because of irregularities during
testing. Average results of the speech-in-noise perception test are depicted in Fig. 1. A paired
wise Repeated Measures MMA with proportion correctly perceived words as dependent
variable, group as between-subject variable (HR versus LR), SNR as within-subject variable
(-1, -4 and -7 dB SNR) and with the same covariates as mentioned above, revealed a
significant effect for group (p = .05), a significant effect for SNR (p < .0001) and no
significant group x SNR interaction effect (p = .22). Post-hoc analysis revealed that both
62
groups only differed at the -7 dB SNR-level (p = .01). A Repeated Measures MMA
comparing the child data with an adult group (n = 8), indicated that both the HR and the LR
group differed significantly from the adult group (p < .0001) with the group x SNR interaction
being significant as well (p = .03).
0
0.2
0.4
0.6
0.8
1
-7 dB -4 dB -1 dB
Signal to noise ratio (SNR)
Pro
porti
on c
orre
ct re
spon
ses
HR group
LR group
Adults
Fig. 1: Mean scores relating the proportion correctly perceived words to the relative level of the presented words
(SNR) [HR group: n = 30, LR group: n = 31; adult group: n = 8].
The result of the perception process can be represented by a psychometric function
relating the percentage correctly perceived words to the relative intensity (SNR) of the
presented words. For every subject, a logistic function relating percent correct responses to
SNR level was fitted to the data using the Nonlinear function of SAS-9.1 and a maximum
likelihood criterion. The fitted function had the following form:
PC = 100 / [1 + exp ( 4 * slope * ( SRT - SNR)/ 100)]
where SNR is the signal-to-noise ratio of the presented stimulus, SRT (or Speech Reception
Threshold) is the estimated signal level required for 50 % correct responses and slope is the
estimated slope parameter of the logistic function. Both the SRT and the slope of the fitted
function are informative: SRT indicates the signal-to-noise ratio required for a fixed level of
performance and the slope reflects the rate at which performance changes with changes in the
level of the signal (Allen & Wightman, 1994).
63
Table 1 shows weighted group means for HR, LR and adult group. In order to take
into account the variable quality of the fits, the inverse standard error of the estimated
parameters was added to the model as a weight variable. A paired wise weighted MMA with
the same covariates as mentioned above demonstrated that both child groups differed
significantly for SRT (p = .04) and (log-transformed) slope (p = .02). This means that the HR
group required a significantly easier signal-to-noise ratio than the LR group in order to
perceive 50 % of the presented words correctly. Moreover, the steeper slope in the HR group
also indicates a faster deterioration of performance while progressing from easy towards
difficult listening conditions. A weighted MMA showed that the adult group differed
significantly from both child groups regarding SRT (p < .0001), but for slope only the adult-
LR difference was significant (adults vs. LR: p = .04; adults vs. HR: p = .77).
Table 1
Descriptive statistics for speech perception tests. Standard deviations are given in parentheses after the mean.
HR group LR group Adult group
Speech-in-noise
N
Weighted SRT (dB)
Weighted slope (%/dB)
Identification
N
Weighted PSE
Weighted slope
Discrimination
N
Mean discriminability
30
-3.3 (1.8)
8.9 (2.5)
25
4.2 (4.0)
16.9 (5.7)
25
0.34 (0.54)
31
-3.8 (2.2)
6.5 (1.9)
23
4.4 (4.1)
21.3 (5.3)
23
0.71 (0.63)
8
-6.4 (1.0)
9.0 (1.1)
11
5.9 (5.4)
75.5 (8.0)
11
1.41 (0.51)
Categorical perception – Identification
Only subjects obtaining 70 % correct responses or more on the pre-test were included
in the analysis, excluding 4 subjects (1HR/3LR). This liberal criterion was chosen since our
data show that on average toddlers misperceive the original /bAk/ sound as /dAk/ in more
than 20% of the presentations (see Fig. 2)4. For the remaining 58 subjects, identification
64
curves were visually inspected to evaluate whether they showed a gradual increase in /dAk/-
responses while progressing from stimulus 1 to stimulus 10. As a result, two additional
subjects (1HR/1LR) were excluded, since they consequently perceived all test stimuli as
/dAk/. (For these subjects, the proportion of /dAk/ answers varied between 84 and 100%,
regardless of the presented stimulus.)
For the remaining 56 subjects, a paired wise Repeated Measures MMA with
‘proportion /dAk/ responses’ as dependent variable, group as between-subject variable (HR
versus LR), stimulus as within-subject variable (1 to 10) and with the same covariates as
mentioned above, revealed no significant effect for group (p = .52), a significant effect for
stimulus (p < .0001) and no significant group x stimulus interaction effect (p = .59). Post-hoc
analysis revealed that for every single stimulus the group difference was insignificant. In
contrast, while comparing the child data with an adult group (n = 11), a Repeated Measures
MMA indicated that both the HR and the LR group differed significantly from the adult group
(p < .01) with the group x stimulus interaction being significant as well (p < .01).
0
0.2
0.4
0.6
0.8
1
1 2 3 4 5 6 7 8 9 10
Stimulus number
Pro
porti
on /d
Ak/
resp
onse
s
HR group
LR group
Adults
Fig 2: Mean identification functions relating the proportion /dAk/ responses to the stimulus number of the place-
of-articulation continuum [HR group: n = 29, LR group: n = 27; adult group: n = 11].
Furthermore, each individual identification curve was submitted to a logistic
transformation in order to estimate the phoneme boundary (i.e. the interpolated 50 % cross-
over point or Point of Subjective Equivalence) and the slope in this point. A logistic function
relating percentage /dAk/-responses to stimulus level (1 to 10) was fitted to the data using the
Nonlinear function of SAS-9.1 and a maximum likelihood criterion. The fitted function had
the following form:
65
PC = 100 / [1 + exp ( 4 * slope * ( PSE - stim)/ 100)]
where stim is the presented stimulus (1 to 10), PSE is the Point of Subjective Equivalence (50
% /bAk/-responses, 50 % /dAk/-responses) and slope is the slope parameter of the logistic
function. Both the PSE and the slope of the fitted function are informative. The PSE indicates
the phoneme boundary and the slope indicates the range of uncertainty in distinguishing one
phoneme category from another. A high slope value indicates a small uncertainty range and
suggests a highly consistent ability to categorise a speech contrast; whereas, a low slope value
indicates a large range of uncertainty and suggests difficulties in identifying the speech
stimuli (Maassen et al., 2001).
Psychometric functions could be fitted for all 56 participants. However, for 8 subjects
(4HR/4LR) the phoneme boundary was estimated to be below stimulus 1, since they even
perceived the /bAk/ endpoint as /dAk/ in more than 50 % of the presentations. These subjects
were excluded from further analysis. For the remaining 48 subjects a paired wise MMA with
the same covariates as mentioned above and with the inversed standard error of the estimated
parameters as a weight variable showed no significant differences between both groups for
PSE (p = .89) and (log-transformed) slope (p = .21). Without adding any covariates the slope
difference resulted in a p-value of .08. In contrast, in a weighted MMA the adult group
differed significantly from both child groups regarding PSE (p < .001), and (log-transformed)
slope (p < .0001), indicating that the adult phoneme boundary had moved up towards /dAk/
and showed a more abrupt perceptual transition (see Table 1 and Fig. 2).
Since many toddlers failed to show perfect performance at the end points of the
continuum, a second and more accurate model was also fitted to the data, taking into account
the variability in identifying the original /bAk/ and /dAk/ stimuli:
PC = min + (max - min) / [1 + exp ( 4 * slope * ( IP - stim)/ (max - min))]
where stim is the presented stimulus (1 to 10), min is the estimated minimum asymptote
(= estimated performance for the original /bAk/ stimulus), max is the estimated maximum
asymptote (= estimated performance for the original /dAk/ stimulus), IP is the estimated
inflection point between min and max and slope is the slope parameter of the logistic
function. Using the parameters estimated by this function, the results comparing children and
66
adult groups were virtually identical to the ones obtained with the simpler function described
above.
Categorical perception – Discrimination
Again, only subjects achieving 70 % correct responses or more on the pre-test were
included in the analysis. This resulted in the removal of data of 14 subjects (6HR/8LR). Most
of these subjects were also excluded from the identification analysis because they revealed too
many errors in that pre-test too or because they did not pass the PSE>1 criterion.
Basically, there are two commonly used methods to score results when assessing
categorical perception with a discrimination task (see Serniclaes et al., 2004). One possibility
is to take into account only the responses to pairs of different stimuli and to calculate the
proportion of (correctly) assigned different responses for every pair. The other possibility is to
take the mean of the correct responses to both the different and the identical pairs
(= proportion ‘hits’ and ‘correct rejections’). Actually, this second option is the most
interesting, since it provides a genuine measure of discriminability irrespective of response
bias (the d’ coefficient). Hence, in the current study this second type of scores was calculated.
In Fig. 3 the mean “discriminability’ functions for the HR, LR and adult group are
displayed. For every stimulus pair a distribution free and response-bias free index of
discriminability (–ln eta) was calculated according to signal detection theory (Wood, 1976):
-ln eta = ½ ln [P(Hit) * P(Correct Rejection) / P (Miss) * P (False Alarm)]
Discriminability equals zero when performance is at chance. It increases with increasing
discrimination accuracy without influences of bias to respond ‘same’ or ‘different’.
Discriminability is maximal at the -ln eta value of 4.6; this value is obtained when the
probabilities for the correct ‘difference’ (= Hit) and correct ‘same’ responses (= Correct
Rejection) are both 0.99, the value assigned (for computational purposes) when the actual
probabilities were 1.00 (Maassen et al., 2001).
A paired wise Repeated Measures MMA with ‘discriminability’ as dependent variable,
group as between-subject variable (HR versus LR), stimulus pair as within-subject variable
(1-4, 2-5, etc.) and with the same covariates as mentioned above, showed a marginally
significant effect for group (p = .056), a significant effect for stimulus pair (p = .0004) and no
significant group x stimulus pair interaction effect (p = .48). Post-hoc analysis revealed that
67
the group difference was significant for pairs 4 -7, 5-8 and 6-9. Without adding any
covariates, the global repeated group difference also turned out to be significant (p = .03).
Further, for every individual a mean discriminability score over all pairs was calculated (see
Table 1). MMA with covariates added showed no significant group difference (p = .07), but
without any covariates this difference was significant in favour of the LR group (p = .04).
Comparison of the child data with the adult group (n = 11) showed that the HR and LR group
differed significantly from the adult group (p < .001 and p < .01 respectively) with the
interaction group x stimulus pair being significant as well (p < .0001). Regarding the mean
discriminability scores, the difference between the HR and LR group on the one hand and the
adult group on the other was also significant (p < .001 and p < .01 respectively).
-0.5
0
0.5
1
1.5
2
2.5
3
3.5
1-4 2-5 3-6 4-7 5-8 6-9 7-10Stimulus pair
Dis
crim
inab
ility
HR group
LR group
Adults
Fig. 3: Mean discrimination scores (- ln eta) as a function of stimulus pair [HR group: n = 25, LR group: n = 23;
adult group: n = 11].
Relations between speech perception and phonological and auditory processing
In table 2 correlations are presented between speech perception ability on the one hand
and auditory and phonological processing on the other. Since SRT and the slope of the
phoneme boundary are the most representative parameters for our speech perception tasks,
only correlations with these parameters are presented. Correlations were partialed out for
individual differences in nonverbal intelligence and the inversed standard error of the
estimated speech parameter was added as a weight variable.
68
The results of the speech-in-noise test (SRT) were significantly related to phonological
awareness, implying that a better speech-in-noise perception corresponded with better
phonological processing. Similarly, for categorical perception the abruptness of the phoneme
boundary (expressed by the slope parameter) was significantly related to phonological
awareness. This implies that a higher slope value, indicative of a small uncertainty range in
distinguishing one phoneme category from the other, was related to better phonological
awareness5. Considering the relations between low-level auditory processing and speech
perception, the positive relation between tone-in-noise detection and speech-in-noise
perception (SRT) was the most straightforward. Further, sensitivity to FM was also related to
speech-in-noise perception, and tone-in-noise detection was related to the abruptness of the
perceptual shifting from one phoneme category to the other.
Table 2
Relations between speech perception and phonological and auditory processing: Weighted Pearson correlation
coefficients partialed out for individual differences in nonverbal IQ.
Speech Reception Threshold
(n = 60)
Slope Phoneme Boundarya
(n = 47)
GAP detection a 0.07 -0.03
FM detection a 0.27* -0.22
TN detection 0.35** -0.30*
Phonological awareness -0.44*** 0.53***
Rapid automatic naming -0.08 -0.05
Verbal short-term memory -0.06 0.16
Note. a Scores were log10-transformed prior to analysis.
* p < .05, ** p < .01, *** p < .001
Individual deviance analysis
As mentioned by Manis and colleagues (1997), there are two possible explanations for
the mixed pattern of speech-perception results observed in literature. First, the group
differences between normal reading and reading-impaired subjects are real, but small and
therefore difficult to detect. Second, group comparisons might mask larger speech-perception
deficits that are found only in specific subgroups of dyslexics. To verify this second option,
we carried out analyses at the subject level. To decide which individual did and did not show
abnormal performance, we adopted the two-step criterion as suggested by Ramus et al.
69
(2003). Applying this procedure, the criterion for deviance was placed at 1.65 standard
deviations of the ‘purified’ mean of the LR-group, after first having excluded all deviant LR-
subjects (by applying a similar 1.65 SD criterion, usually resulting in the removal of one or
two deviant LR subjects). A distribution analysis on the speech results of the ‘restricted’ LR-
group confirmed the normality of the data. Hence, the 1.65 SD deviance criterion corresponds
to the fifth percentile. For the various speech measures, Table 3 presents the number of
deviant subjects in each risk group. Although the HR group generally presented a higher
proportion of deviant subjects, the difference was not significant.
Table 3
Individual deviance analysis for speech perception tests.
HR
# of deviants
LR
# of deviants
Fisher Exact
(2-sided p-value)
Speech-in-noise
SRT
Identification
Slope of phoneme boundarya
Discrimination
Mean discriminability
4 of 30 (13%)
11 of 25 (44%)
4 of 25 (16%)
1 of 31 (3%)
6 of 23 (26%)
1 of 23 (4%)
.20
.23
.35
Note. a Scores were log10-transformed prior to analysis.
Individual Z-scores for speech perception parameters (i.e. SRT and the Slope of
phoneme boundary) and for phonological and auditory measures, relative to the mean of the
‘restricted’ LR group are depicted in Appendix. Scores deviating more than 1.65 SD (in the
deficient direction) are printed in bold. Inspection of the individual scores reveals that there is
no straightforward regularity or tendency between these measures. For the speech perception
data for instance, deficits in categorical perception and speech-in-noise perception appear to
be largely mutually unrelated. The same applies for deficits in phonology and low-level
auditory processing: some subjects differ only on one of the tasks, others on two or three, but
without consistency (see also Boets et al., 2006). To investigate the relation between speech
perception deficiencies on the one hand and phonological and auditory deficiencies on the
other, we constructed some additional variables indicating whether a specific skill is impaired
or not. Because of the high intrasubject variability (even for scores measuring the same
overall skill), averaging across these measures might underestimate the presence of a deficit
70
in this skill. Instead, a skill was labeled as deficient if at least one of its measures scored
beyond the 1.65 SD cutoff criterion. As such, PHONOLOGY is a binary variable indicating
whether a subject demonstrates a deficit in phonological awareness, rapid automatic naming
or verbal short-term memory. Similarly, AUDITORY is a binary variable indicating whether
a subject demonstrates a deficit in at least one of the auditory measures (viz. gap-in-noise
detection, tone-in-noise detection and FM detection) and SPEECH is a binary variable
indicating whether a subject shows deviant scores for SRT or for the slope parameter of the
identification function. As mentioned above, some subjects were not able to pass the 70%
pretest or PSE>1 criterion of the identification task. Since we felt this failure might also offer
important information concerning speech perception abilities, these subjects were also labeled
as deviant on the SPEECH variable.
To explore the interrelation between deficiencies in these different skills, we
calculated the number of subjects showing isolated versus (partially) overlapping deficits. Fig.
4 depicts a schematic overview of this analysis. Results indicate that there indeed exists a
small subgroup of subjects showing consistent deficits across auditory, speech and
phonological processing. However, and more importantly, the overview convincingly
demonstrates that the relation between these deficits is far from clear. Contrary to the causal
hypothesis, there are also many subjects with severe deficits in one of the skills while being
completely unimpaired in the other ones, or subjects showing only partially overlapping
deficits.
Fig. 4: Distribution of auditory, speech perception and phonological deficits in the total sample of 62 preschool
children. The number of impaired subjects is indicated both in absolute numbers and in percentages.
SPEECH PHONOLOGY
AUDITORY
11 (18%)
3 (5%)
5 (8%)
4 (6%)
9 (15%)
7 (11%)
6 (10%)
No deficits = 17 (27 %)
71
Discussion
In this study we tested speech perception and phonological and auditory processing in
two groups of five-year-old preschool subjects, i.e. a genetically high risk and a genetically
low risk group. Data were compared between both risk groups, and in addition speech
perception abilities were contrasted with those of adult subjects.
Comparing child and adult speech perception data, we observed a significant and
consistent difference to the advantage of the adults. Adults demonstrated better speech-in-
noise perception thresholds, a better categorical perception with a shifted and more abrupt
phoneme boundary, and better general discriminability. It is well recognized that preschoolers
perform more poorly than older children and adults across all kinds of psychoacoustic tasks
(Wightman & Allen, 1992). This weaker performance has also been demonstrated for the
specific speech perception tasks at hand (e.g. Allen & Wightman, 1994; Elliot, 1979; Hazan
& Barrett, 2000; Johnson, 2000; Mills, 1975; Serniclaes et al., 2004). However, it is still an
ongoing discussion whether this weaker performance reflects an underlying sensorial
immaturity or rather has a cognitive origin (e.g. non-optimal listening strategies, fluctuations
in attention, etc.).
Regarding the child data, comparing both risk groups, results are less straightforward.
With respect to the speech-in-noise experiment, the repeated measures analysis revealed that
the HR group perceived significantly less words correctly than the LR group, particularly in
the most difficult listening condition. This observation was also confirmed by significant
group differences for the estimated SRT and slope parameters. Moreover, the steeper slope in
the HR group indicated a faster deterioration of performance while progressing from easy
towards difficult listening conditions. These results are in line with earlier data reporting a
speech-in-noise perception deficit in subjects with dyslexia and/or language impairments
(Brady et al., 1983; Bradlow et al., 2003; Chermak et al., 1989; Stollman et al., 1994). In
particular, the observation that the HR children were more adversely affected by a decreasing
signal-to-noise ratio is consistent with findings by Bradlow et al. (2003) and Stollman et al.
(1994). Interestingly, a similar noise exclusion deficit in dyslexic subjects has recently also
been described in the visual modality (Sperling, Lu, Manis & Seidenberg, 2005). Moreover,
these authors demonstrated that visual contrast thresholds in high-noise conditions were
significantly related to several language measures, suggesting that the dyslexic deficit may
not be specific for auditory speech-in-noise perception, but might be a cross-modal deficit in
noise exclusion.
72
With respect to the categorical perception task, both child groups performed similarly
on the identification task, eventuating into analogous results for the repeated measures
analysis and for the estimated phoneme boundary and slope. However, we observed a
(marginally) significant difference to the advantage of the LR group on the more sensitive
discrimination task (although the differing stimulus pairs did not particularly include the ones
crossing the phoneme boundary). Consequently, these results corroborate only partly with the
mass of evidence provided by most other studies (see introduction section). A straightforward
explanation for this less pronounced group difference is obviously the fact that we did not
study a well-defined clinical group but only an at-risk group that still might show substantial
overlap with the control group. Alternatively, it is plausible that differences in categorical
perception will become more evident as development continues (possibly as a consequence of
differently developing phonological and reading abilities, as suggested by Serniclaes et al.,
2001). This would imply that the identification and discrimination curves of normal reading
children will evolve towards those of the adult group, whereas those of the dyslexic children
will continue lagging behind. Indeed, Hazan and Barrett (2000) demonstrated that the
development of phonemic categorization extends well beyond the age of 12 years before
reaching adult levels. Moreover, Maassen and colleagues (2001) revealed that dyslexic
children perform worse than age-matched control children but similar to reading-age younger
controls on a voicing identification task, suggesting a developmental delay associated with
reading-level instead of a qualitative difference. In a similar way, Bishop and McArthur
(2004; McArthur & Bishop, 2004) and Wright and Zecker (2004) have hypothesized that
dyslexic subjects might not suffer from a specific chronic perceptual deficit, but rather from a
more general and passing maturational delay. Accordingly, a task would only have enough
differential sensitivity when administered in the age period the measured skill is sharply
improving in typically developing subjects, and maybe this might not yet have been the case
with categorical perception.
Considering the relations between these variables, it should be reiterated that in a
previous paper describing the same sample of preschoolers (Boets et al., 2006), we already
reported a considerable correlation between auditory spectral processing and phonological
awareness [r(FM, Phonological Awareness) = -0.48, p < .0001; r(TN, Phonological
Awareness) = -0.35, p < .01; Spearman correlations partialed out for individual differences in
age and non-verbal IQ]. Likewise, in the actual study we also observe a significant relation
between speech perception on the one hand, and auditory processing and phonological
awareness on the other. Moreover, since these relations are already present in preschool
73
children that have not yet learned to read, our data indicate that the observed interrelations
precede literacy development and are not merely the result of differences in reading
experience or reading skills.
Although our correlation analysis indicates that there are some general relations
between aspects of auditory processing, speech perception and phonological ability, the
results of the individual deviance analysis demonstrate that the theoretical implications of
these correlations should not be overestimated. Actually there was only a very small subgroup
showing consistent deficiencies across all processing skills, while the majority of subjects
showed isolated or only partially overlapping deficits. This means for instance that contrary to
the hypothesized ontogenetic pathway, there was a substantial group of subjects showing
considerable auditory processing problems without presenting any problems in speech
perception or phonology. In the same way, there were many subjects presenting speech
perception problems despite intact auditory processing or without interference with the
phonological development. Further on, there was also a group of subjects showing
phonological problems without evidence of auditory or speech perception problems. In sum,
the overall picture indicated that about 30 % of the subjects were unimpaired, 30 % presented
an isolated deficit, 30 % presented a deficit in two skills and a small subgroup of about 10 %
presented consistent deficiencies across the three processing skills. Considering the
proportion of HR subjects in these differently affected groups, an interesting pattern was
revealed: whereas the HR subjects only represented 32 % of the unaffected group, their
proportion grew to 56 and 61 % of the groups showing respectively one and two deficits, and
they ultimately represented up to 100 % of the group showing deficits across all the
processing skills.
Of course, one has to be cautious while interpreting these data. For instance, we have
to acknowledge that it is not realistic that about 70 percent of the children in our combined
HR and LR sample would present one or more deficits in the assessed skills. This is certainly
an overestimation due to the liberal criterion applied. Moreover, our data only represent a
momentary measurement of skills that are still highly developing at this age. Therefore,
interindividual differences in speed of maturation might also have influenced the observed
number of aberrantly scoring subjects.
Several explanations for the relatively small group of subjects showing overlapping
deficits can be proposed. First, it might be argued that the administered tasks were not
sensitive enough to detect very subtle deficits. However, it is important to stress that the tasks
at hand have not been chosen accidentally, but because literature has proven them to
74
differentiate reliably between reading-impaired and normal reading subjects (regarding the
auditory tasks, see e.g. Talcott et al., 2003; Van Ingelghem et al., 2005; regarding the
discriminative power of the phonological tasks, see e.g. Scarborough, 1998). Second, since
the assessed skills are still maturing at this preschool age, it is very well possible that they will
influence each other during further literacy development, resulting in an increased
interrelation in future. Third, it might be necessary to revise or refine the theoretically
suggested developmental pathway in some way. Indeed, parallel to the growing evidence that
only a small subgroup of dyslexics presents low-level auditory problems (Ramus et al., 2003;
Rosen, 2003), it becomes increasingly clear that also only a minority of dyslexics shows
speech perception problems (Manis et al., 1997). Joanisse and colleagues (2000) for example
studied three groups of dyslexics (a language impaired, a phonologically impaired and a
globally reading delayed group) and demonstrated that speech perception problems only
occur in dyslexic subjects showing additional expressive and/or receptive language problems.
They also observed several dyslexics with phonological impairments whose speech
perception was normal, confirming that phonological impairments are not necessarily
secondary to a speech perception deficit. Accordingly, these authors argue that although the
language impaired and phonologically impaired dyslexic groups exhibit similar deficits in
phonology and reading, their causes may differ, with one involving a speech perception
deficit, and the other involving higher-order differences in phonological representation or
processing. The third globally reading delayed group of their sample did not show any
deficits in speech perception or phonology at all, suggesting still another cause for their
reading deficit. Consequently, together with these authors we would contend that in order to
fully understand the deviant developmental pathway of dyslexia, it will be necessary to take
into account the considerable heterogeneity of this group and to recognize that subjects
presenting similar patterns of reading problems must not necessarily have the same
underlying deficits. In this way it is plausible that the hypothesized causal path (from auditory
processing to speech perception to phonology to reading) might be true for a specific
subgroup of dyslexics, but it probably is not representative for the whole population of
reading impaired subjects.
75
References
Allen, P., & Wightman, F. (1994). Psychometric functions for children's detection of tones in
noise. Journal of Speech and Hearing Research, 37(1), 205-215.
Bailey, P., Griffiths, Y., Hill, N., Brent, S., & Snowling, M. (2002). Speech intelligibility in
continuous and intermittent noise for dyslexic and normally reading listeners. Poster
Presented at the ISCA Workshop on Temporal Integration in the Perception of Speech
(TIPS), Aix-En-Provence, 8th - 10th April .
Benasich, A. A., & Tallal, P. (2002). Infant discrimination of rapid auditory cues predicts
later language impairment. Behavioural Brain Research, 136(1), 31-49.
Bishop, D. V. M., & McArthur, G. M. (2004). Immature cortical responses to auditory stimuli
in specific language impairment: Evidence from ERP's to rapid tone sequences.
Developmental Science, 7(4), F11-F18.
Blomert, L., & Mitterer, H. (2004). The fragile nature of the speech-perception deficit in
dyslexia: natural versus synthetic speech. Brain and Language, 89, 21-26.
Boets, B., Wouters, J., van Wieringen, A., & Ghesquière, P. (2006). Auditory temporal
information processing in preschool children at family risk for dyslexia: relations with
phonological abilities and developing literacy skills. Brain and Language, 97, 64-79.
Bowers, P. G., & Swanson, L. B. (1991). Naming speed deficits in reading disability: multiple
measures of a singular process. Journal of Experimental Child Psychology, 51, 195-
219.
Bradlow, A. R., Kraus, N., & Hayes, E. (2003). Speaking clearly for children with learning
disabilities: sentence perception in noise. Journal of Speech, Language and Hearing
Research, 46, 80-97.
Brady, S., Shankweiler, D., & Mann, V. (1983). Speech perception and memory coding in
relation to reading ability. Journal of Experimental Child Psychology, 35(2), 345-367.
Brandt, J., & Rosen, J. J. (1980). Auditory phonemic perception in dyslexia: categorical
identification and discrimination of stop consonants. Brain and Language, 9, 324-337.
76
Breier, J. I., Fletcher, J. M., Denton, C., & Gray, L. C. (2004). Categorical perception of
speech stimuli in children at risk for reading difficulty. Journal of Experimental Child
Psychology, 88, 152-170.
Catts, H. W. (1989). Phonological processing deficits and reading disabilities. In A. G.
Kahmi, & H. W. Catts (Eds.), Reading disabilities: a developmental language
perspective (pp. 101-132). Boston, MA: Littel Brown.
Chermak, G. D., Vonhof, M. R., & Bendel, R. B. (1989). Word identification performance in
the presence of competing speech and noise in learning disabled adults. Ear and
Hearing, 10(2), 90-93.
Cunningham, J., Nicol, T., Zecker, S. G., Bradlow, A., & Kraus, N. (2001). Neurobiologic
responses to speech in noise in children with learning problems: deficits and strategies
for improvement. Clinical Neurophysiolgy, 112(5), 758-67.
Denenberg, V. H. (1999). A critique of Mody, Studdert-Kennedy, and Brady's "Speech
perception deficits in poor readers: auditory processing or phonological coding?".
Journal of Learning Disabilities, 32(5), 379-383.
Elliot, L. L. (1979). Performance of children aged 9 to 17 on a test of speech intelligiblity in
noise using sentence material with controlled word predictability. Journal of the
Acoustical Society of America, 66, 651-653.
Farmer, M. E., & Klein, R. M. (1995). The evidence for a temporal processing deficit linked
to dyslexia: A review. Psychonomic Bulletin and Review, 2(4), 460-493.
Gerrits, E. (2003). Speech perception of young children at risk for dyslexia and children with
specific language impairment. Proceedings of the 15th International Conference on
Phonetic Sciences (pp. 2357-2360). Barcelona, Spain, 3 - 9 Augustus 2003.
Gilger, J. W., Pennington, B. F., & DeFries, J. C. (1991). Risk for reading disability as a
function of parental history in three family studies. Reading and Writing, 3(3-4), 205-
217.
Godfrey, J. J., Syrdal-Lasky, A. K., Millay, K. K., & Knox, C. M. (1981). Performance of
dyslexic children on speech perception tests. Journal of Experimental Child
77
Psychology, 32, 401-424.
Habib, M. (2000). The neurobiological basis of developmental dyslexia: an overview and
working hypothesis. Brain, 123, 2373-2399.
Hazan, V., & Barrett, S. (2000). The development of phonemic categorization in children
aged 6-12. Journal of Phonetics, 28, 377-396.
Joanisse, M. F., Manis, F. R., Keating, P., & Seidenberg, M. S. (2000). Language deficits in
dyslexic children: speech perception, phonology and morphology. Journal of
Experimental Child Psychology, 77, 30-60.
Johnson, C. E. (2000). Children's phoneme identification in reverberation and noise. Journal
of Speech, Language and Hearing Research, 43, 144-157.
Kowalchuk, R. K., Keselman, H. J., Algina, J., & Wolfinger, R. D. (2004). The analysis of
repeated measurements with mixed-model adjusted F tests. Educational and
Psychological Measurement, 64(2), 224-242.
Laneau, J., Boets, B., Moonen, M., van Wieringen, A., & Wouters, J. (2005). A flexible
auditory research platform using acoustic or electric stimuli for adults and young
children. Journal of Neuroscience Methods, 142(1), 131-136.
Liberman, A. M., Harris, K. S., Hoffman, H. S., & Griffith, B. C. (1957). The discrimination
of speech sounds within and across phoneme boundaries. Journal of Experimental
Psychology, 54, 358-368.
Lieberman, P., Meskill, R. H., Chatillon, M., & Schupack, H. (1985). Phonetic speech
perception deficits in dyslexia. Journal of Speech and Hearing Research, 28(480-486).
Littell, R. C., Stroup, W. W., & Freund, R. J. (2002). SAS for linear models. Cary: SAS
Institute.
Lyytinen, H., Ahonen, T., Eklund, K., Guttorm, T. K., Laakso, M. L, Leinonen, S., Leppanen,
P. H., Lyytinen, P., Poikkeus, A. M, Puolakanaho, A., Richardson, U., & Viholainen,
H. (2001). Developmental pathways of children with and without familial risk for
dyslexia during the first years of life. Developmental Neuropsychology, 20(2), 535-
554.
78
Maassen, B., Groenen, P., Crul, T., Assman-Hulsmans, C., & Gabreëls, F. (2001).
Identification and discrimination of voicing and place-of-articulation in developmental
dyslexia. Clinical Linguistics and Phonetics, 15( 4), 319-339.
MacKain, K. S., Best, C. T., & Strange, W. (1981). Categorical perception of English /r/ and
/l/ by Japanese bilinguals. Applied Psycholinguistics, 2, 369-390.
Manis, F. R., Custodio, R., & Szeszulski, P. A. (1993). Development of phonological and
orthographic skills: a 2-year longitudinal study of dyslexic children. Journal of
Experimental Child Psychology, 56, 64-86.
Manis, F. R., McBride-Chang, C., Seidenberg, M. S., Keating, P., Doi, L. M., Munson, B., &
Petersen, A. (1997). Are speech perception deficits associated with developmental
dyslexia? Journal of Experimental Child Psychology, 66, 211-235.
McArthur, G. M, & Bishop, D. V. (2001). Auditory perceptual processing in people with
reading and oral language impairments: current issues and recommendations.
Dyslexia, 7(3), 150-70.
McArthur, G. M., & Bishop, D. V. M. (2004). Which people with specific language
impairment have auditory deficits? Cognitive Neuropsychology, 21(11), 79-94.
McBride-Chang, C. (1995). Phonological processing, speech perception and reading
disability: an integrative review. Educational Psychologists, 30(3), 109-121.
McBride-Chang, C. (1996). Models of speech perception and phonological processing in
reading. Child Development, 67(4 ), 1836-1856.
Menell, P., McAnally, K. I., & Stein, J. F. (1999). Psychophysical sensitivity and
physiological response to amplitude modulation in adult dyslexic listeners. Journal of
Speech, Language and Hearing Research, 42(4), 797-803.
Mills, J. H. (1975). Noise and children: a review of literature. Journal of the Acoustical
Society of America, 58, 767-779.
Mody, M., Studdert-Kennedy, M., & Brady, S. (1997). Speech perception deficits in poor
readers: auditory processing or phonological coding? Journal of Experimental Child
Psychology, 64(2), 199-231.
79
Molfese, D. L. (2000). Predicting dyslexia at 8 years of age using neonatal brain responses.
Brain and Language, 72(3), 238-245.
Morais, J., Bertelson, P., Carey, L., & Alegria, J. (1979). Does awareness of speech as a
sequence of phones arise spontaneously? Cognition, 7, 323-331.
Nittrouer, S. (1996). The relation between speech perception and phonemic awareness:
evidence from low-SES children and children with chronic OM. Journal of Speech
and Hearing Research, 39(5), 1059-1070.
Nittrouer, S. (1999). Do temporal processing deficits cause phonological problems? Journal of
Speech, Language and Hearing Research, 42, 925-942.
Pennington, B. F., Van Orden, G. C., Smith, S. D., Green, P. A., & Haith, M. M. (1990).
Phonological processing skills and deficits in adult dyslexics. Child Development, 61,
1753-1778.
Rack, J. P. (1994). Dyslexia: The phonological deficit hypothesis. In A. Fawcett, & R.
Nicolson (Eds.), Dyslexia in children: multidisciplinary perspectives. London:
Harvester Wheatsheaf.
Ramus, F., Rosen, S., Dakin, S. C., Day, B. L., Castellote, J. M., White, S., & Frith, U.
(2003). Theories of developmental dyslexia: insights from a multiple case study of
dyslexic adults. Brain, 126(4), 841-865.
Raven, J. C., Court, J. H., & Raven, J. (1984). Manual for Raven's Progressive Matrices and
Vocabulary Scales. London: Lewis.
Reed, M. A. (1989). Speech perception and the discrimination of brief auditory cues in
reading disabled children. Journal of Experimental Child Psychology, 48(2), 270-92.
Richardson, U., Leppänen, P. H. T., Leiwo, M., & Lyytinen, H. (2003). Speech perception of
infants with high familial risk for dyslexia differ at the age of 6 months.
Developmental Neuropsychology, 23(3), 385-397.
Rosen, S. (2003). Auditory processing in dyslexia and specific language impairment: is there
a deficit? What is its nature? Does it explain anything? Journal of Phonetics, 31, 509-
527.
80
Rosen, S., & Manganari, E. (2001). Is there a relationship between speech and nonspeech
auditory processing in children with dyslexia? Journal of Speech, Language and
Hearing Research, 44, 720-736.
Scarborough, H.S. (1998). Early identification of children at risk for reading disabilities:
phonological awareness and some other promising predictors. In P.J. Accardo, A.J.
Capute & B.K. Shapiro (Eds.), Specific reading disability: a view of the spectrum (pp.
75-119). Timonium, MD: York Press.
Schneider, W., Eschman, A., & Zuccolotto, A. (2002). E-Prime User's Guide. Pittsburgh:
Psychology Software Tools Inc.
Serniclaes, W., Sprenger-Charolles, L., Carré, R., & Demonet, J. (2001). Perceptual
discrimination of speech sounds in developmental dyslexia. Journal of Speech,
Language and Hearing Research, 44, 384-399.
Serniclaes, W., Van Heghe, S., Mousty, P., Carré, R., & Sprenger-Charolles, L. (2004).
Allophonic mode of speech perception in dyslexia. Journal of Experimental Child
Psychology, 87, 336-361.
Snowling, M., Goulandris, N., Bowlby, M., & Howell, P. (1986). Segmentation and speech
perception in relation to reading skill: a developmental analysis. Journal of
Experimental Child Psychology, 41(3), 489-507.
Snowling, M. J. (2000). Dyslexia (2nd ed.). Malden, MA: Blackwell Publishers.
Sperling, A. J., Lu, Z., Manis, F. R., & Seidenberg, M. S. (2005). Deficits in perceptual noise
exclusion in developmental dyslexia. Nature Neuroscience, 8, 862-863.
Steffens, M. L., Eilers, R. E., Gross-Glenn, K., & Jallad, B. (1992). Speech perception in
adult subjects with familial dyslexia. Journal of Speech and Hearing Research, 35,
192-200.
Stollman, M., Kapteyn, T., & Sleeswijk, B. (1994). Effect of time-scale modification of
speech on the speech recognition threshold in noise for hearing-impaired and
language-impaired children. Scandinavian Audiology, 23, 39-46.
Talcott, J. B., Gram, A., Van Ingelghem, M., Witton, C., Stein, J. F., & Toennessen, F. E.
81
(2003). Impaired sensitivity to dynamic stimuli in poor readers of a regular
orthography. Brain and Language, 87, 259-266.
Talcott, J. B., Witton, C., Hebb, G. S., Stoodley, C. J., Westwood, E. A., France, S. J.,
Hansen, P. C., & Stein, J. F. (2002). On the relationship between dynamic visual and
auditory processing and literacy skills; results from a large primary-school study.
Dyslexia, 8(4), 204-225.
Talcott, J. B., Witton, C., McClean, M., Hansen, P. C., Rees, A., Green, G., & Stein, J. F.
(1999). Can sensitivity to auditory frequency modulation predict children's
phonological and reading skills? Neuroreport, 10(10), 2045-2050.
Talcott, J. B., Witton, C., McLean, M. F., Hansen, P. C., Rees, A., Green, G. G., & Stein, J. F.
(2000). Dynamic sensory sensitivity and children's word decoding skills. Proceedings
of the National Academy of Science, USA, 97( 6), 2952-2957.
Talcott, J. B., & Witton, C. (2002). A sensory linguistic approach to the development of
normal and impaired reading skills. In E. Witruk, A. Friederici, & T. Lachmann
(Eds.), Neuropsychology and cognition series. Basic functions of language and
language disorders. Dordrecht, Netherlands: Kluwer Academic Publishers.
Tallal, P. (1980). Auditory temporal perception, phonics, and reading disabilities in children.
Brain and Language, 9(2), 182-198.
Tallal, P. (1984). Temporal or phonetic processing deficit in dyslexia? That's the question.
Applied Psycholinguistics, 5, 167-169.
van Beinum, F. J., Schwippert, C. E., Been, P. H., van Leeuwen, T. H., & Kuijpers, C. T. L.
(2005). Development and application of a /bAk/-/dAk/ continuum for testing auditory
perception within the Dutch longitudinal study. Speech Communication, 47, 124-142.
Van Ingelghem, M., Boets, B., van Wieringen, A., Onghena, P., Ghesquière, P., & Wouters, J.
(2005). An auditory temporal processing deficit in children with dyslexia? In P.
Ghesquière & A.J.J.M. Ruijssenaars (Eds.), Learning disabilities: a challenge to
teaching and instruction. Series: Studia Paedagogica 40. (pp. 47-63). Leuven:
University Press.
82
van Wieringen, A., & Wouters, J. (2005). Normalization and feasibility of speech
understanding tests for Dutch speaking toddlers. Speech Communication, 47, 169-181.
Wagner, R. K., Torgesen, J. K., & Rashotte, C. A. (1994). Development of reading-related
phonological processing abilities: new evidence of bi-directional causality from a
latent variable longitudinal study. Developmental Psychology, 30, 73-87.
Wagner, R. K., & Torgesen, J. K. (1987). The nature of phonological processing and its
causal role in the acquisition of reading skills. Psychological Bulletin, 101(2), 192-
212.
Warrier, C. M., Johnson, K. L., Hayes, E. A., Nicol, T., & Kraus, N. (2004). Learning
impaired children exhibit timing deficits and training-related improvements in
auditory cortical responses to speech in noise. Experimental Brain Research, 157, 431-
441.
Watson, B. U., & Miller, T. K. (1993). Auditory perception, phonological processing and
reading ability/disability. Journal of Speech and Hearing Research, 36(4 ), 850-863.
Werker, J. F., & Tees, R. C. (1984). Cross-language speech perception: evidence for
perceptual reorganisation during the first year of life. Infant Behavior and
Development, 7, 49-63.
Werker, J. F., & Tees, R. C. (1987). Speech perception in severely disabled and average
reading children. Canadian Journal of Experimental Psychology, 41, 48-61.
Wible, B., Nicol, T., & Kraus, N. (2002). Abnormal neural encoding of repeated speech
stimuli in noise in children with learning problems. Clinical Neurophysiology, 113(4),
485-94.
Wightman, F., & Allen, P. (1992). Individual differences in auditory capability among
preschool children. In L.A. Werner, & E.W. Rubel (Eds.), Developmental
Psychoacoustics (pp. 113-134). Washington DC: American Psychological
Association.
Witton, C., Talcott, J. B., Hansen, P. C., Richardson, A. J., Griffiths, T. D., Rees, A., Stein, J.
F, & Green, G. G. (1998). Sensitivity to dynamic auditory and visual stimuli predicts
83
nonword reading ability in both dyslexic and normal readers. Current Biology, 8(14),
791-797.
Wolf, M. (1986). Rapid alternating stimulus naming in the developmental dyslexias. Brain
and Language, 27, 360-379.
Wood, C. C. (1976). Discriminability, response bias and phoneme categories in
discrimination of voice onset time. Journal of the Acoustic Society of America, 60(6),
1381-1389.
Wouters, J., Damman, W., & Bosman, A. J. (1994). Vlaamse opnamen van woordenlijsten
voor spraakaudiometrie. CD NKO - K.U.Leuven/U.Z.Leuven.
Wright, B. A., Lombardino, L. J., King, W. M., Puranik, C. S., Leonard, C. M., & Merzenich,
M. M. (1997). Deficits in auditory temporal and spectral resolution in language-
impaired children. Nature, 387, 176-178.
Wright, B. A., & Zecker, S. G. (2004). Learning problems, delayed development and puberty.
Proceedings of the National Academy of Science of the United States of America,
101(26), 9942-9946.
84
Notes
1 Manuscript submitted for publication in Brain and Language. Co-authors: P. Ghesquière, A.
van Wieringen & J. Wouters. 2 Traditional statistics (e.g. ANOVA and regression analysis) require independence of errors,
but this assumption is not met in a dependent sample design. Mixed models allow modelling
this dependent nature of the data. By entering pair number as a random factor in all analyses,
these models are an extension of the paired t-test, while offering the possibility to include
more sophisticated options such as the incorporation of covariates and repeated measurement
analyses. 3 All MMA’s used the Kenward-Roger degrees of freedom estimation method that is more
robust against violations to assumptions of normality (Kowalchuk, Keselman, Angina &
Wolfinger, 2004). 4 This finding should not be too surprising: in the study of van Beinum et al. (2005) (using
identical stimuli and a similar paradigm) normal reading adults also misperceived the /bAk/
stimulus as ‘dak’ in 10% of the presentations. 5 Noteworthy, another indicator for a relation between categorical perception and
phonological awareness is the observation that almost all excluded subjects (based on the
>70% on the pretest or PSE>1 criterion) score bad on the phonological awareness factor (the
average score of all excluded subjects was .89 SD’s below the mean of the LR group).
Appendix: Individual Z-scores relative to the mean of the ‘restricted’ LR group for speech perception, phonological ability and auditory sensitivity. Scores deviating more than 1.65 SD’s (in the direction that indicates deficiency) are depicted in bold.
Nr Group SRT Slope Phoneme
Boundarya TN GAPa FMaPhon
Awareness RANVerbal
STM AUDITORY PHONOLOGY SPEECH
1 HR -0.41 -1.15 -0.11 -0.16 1.07 0.46 0.27 0.56 3 HR 1.31 0.68 -0.90 -1.46 -1.64 0.23 0.32 DEVIANT 5 HR 0.73 -2.53 0.42 -0.92 2.08 -0.83 0.85 -0.36 DEVIANT DEVIANT 7 HR -0.47 -2.26 0.69 -0.68 0.19 -1.61 -0.63 1.15 DEVIANT 9 HR 0.00 0.41 -1.87 -1.16 -1.34 0.89 -0.56 -1.03 11 HR 1.99 0.85 1.33 0.73 -0.23 -0.51 -0.02 DEVIANT 13 HR 1.55 -0.40 0.34 -0.33 -0.32 0.20 0.40 -0.10 15 HR 0.17 -2.19 -2.94 -0.70 0.58 -0.13 0.45 0.68 DEVIANT 17 HR 2.59 -0.11 3.12 1.59 -2.29 0.07 -0.17 DEVIANT DEVIANT DEVIANT 19 HR 0.48 -7.48 0.51 -0.07 -0.57 -2.68 0.74 -0.95 DEVIANT DEVIANT 21 HR 0.01 2.65 -0.91 -0.69 0.26 -0.25 -0.25 -0.32 23 HR 0.27 0.51 0.14 -0.48 1.05 1.36 -1.24 DEVIANT 25 HR 1.49 1.39 -1.15 -2.40 0.70 0.08 DEVIANT DEVIANT 27 HR -0.89 2.76 0.95 -0.50 4.43 -0.91 -1.14 0.22 DEVIANT 29 HR 0.66 4.58 -0.01 4.09 1.64 -2.02 -0.64 -0.03 DEVIANT DEVIANT 31 HR 1.14 -8.89 0.52 -0.28 0.90 -3.78 0.44 -0.06 DEVIANT DEVIANT 33 HR 1.59 0.78 -2.40 2.29 -1.17 -1.86 0.63 0.72 DEVIANT DEVIANT 35 HR 1.31 -0.54 4.12 4.22 6.75 -2.71 -0.80 -0.91 DEVIANT DEVIANT 37 HR -2.19 4.42 -2.49 0.17 -1.59 1.28 -0.02 -2.55 DEVIANT 39 HR 0.26 -0.26 -2.23 -1.08 -1.98 -0.02 -2.56 -2.14 DEVIANT 41 HR 1.04 1.29 1.58 1.15 0.17 -0.70 -0.95 0.30 43 HR 2.21 -0.52 -1.51 -0.68 0.10 1.32 -1.34 -0.96 DEVIANT 45 HR -1.25 -3.17 0.96 11.36 2.95 -0.31 -0.61 1.03 DEVIANT DEVIANT 47 HR 0.47 -3.46 2.19 -0.89 0.54 -2.43 0.54 2.24 DEVIANT DEVIANT DEVIANT 49 HR -0.24 -4.49 0.87 7.44 5.34 -4.39 0.01 -1.64 DEVIANT DEVIANT DEVIANT 51 HR 0.51 2.89 2.07 3.29 -1.07 -3.03 -0.88 DEVIANT DEVIANT DEVIANT 53 HR 0.91 -2.17 1.05 -0.16 0.51 -4.67 -1.13 -0.51 DEVIANT DEVIANT 55 HR 0.39 -3.12 0.87 0.01 5.49 -0.92 -0.77 -0.78 DEVIANT DEVIANT 57 HR 2.46 -7.88 1.31 -0.72 3.15 -2.22 -1.70 -2.00 DEVIANT DEVIANT DEVIANT 59 HR -0.43 0.59 -0.72 0.08 0.82 -1.65 -0.13 -0.74 DEVIANT 61b HR -0.93 -0.64 -1.46 2.52 -0.87 DEVIANT DEVIANT
Appendix, continuation
Nr Group SRT Slope Phoneme
Boundarya TN GAPa FMaPhon
Awareness RANVerbal
STM AUDITORY PHONOLOGY SPEECH
2 LR 0.61 0.69 -0.72 -0.32 1.15 -0.57 0.36 DEVIANT 4 LR 0.99 -1.65 0.78 0.13 1.39 -0.25 -1.00 0.59 DEVIANT 6 LR -0.64 -0.53 -0.74 -0.68 -0.40 -1.14 0.87 1.12 8 LR 0.46 2.88 2.01 0.27 1.38 -2.57 -1.97 -0.39 DEVIANT DEVIANT
10 LR 0.91 -2.60 1.57 0.75 -0.73 1.00 -1.68 1.19 DEVIANT DEVIANT 12 LR 0.57 -0.20 0.27 0.56 -3.01 0.40 -2.83 DEVIANT DEVIANT 14 LR 0.15 -2.86 -0.38 -0.57 0.47 0.67 -0.87 0.80 DEVIANT 16 LR -0.47 7.59 -2.14 0.07 -0.10 0.87 0.22 -0.88 18 LR 0.67 -0.27 -0.49 -0.17 0.99 0.15 -0.20 -0.17 20 LR 2.11 -1.08 0.45 0.68 0.46 1.13 -2.53 DEVIANT DEVIANT 22 LR 0.06 0.47 -1.19 0.37 0.00 -0.12 1.09 -0.06 24 LR 0.48 -2.59 0.87 0.12 -0.41 -0.71 -0.40 -1.43 DEVIANT 26 LR -0.61 0.51 1.05 6.15 5.39 -1.30 1.88 -0.52 DEVIANT 28 LR -1.92 -0.04 -1.63 -0.61 0.07 -0.22 0.00 1.39 30 LR 0.38 0.43 -0.59 1.31 -3.14 0.05 -0.41 DEVIANT DEVIANT 32 LR -1.93 2.97 -0.83 -1.42 -1.71 1.38 -1.17 -1.34 34 LR 0.60 1.05 2.31 0.42 0.39 0.37 -0.55 DEVIANT DEVIANT 36 LR -0.98 3.03 -0.64 -0.07 -1.30 0.55 -0.91 1.66 38 LR -0.44 0.56 0.61 -0.49 -0.57 0.61 1.34 -0.68 40 LR -0.90 3.58 -1.63 -1.08 -2.05 1.99 -0.88 0.14 42 LR 1.41 0.87 0.14 4.72 -1.76 -3.44 -0.35 DEVIANT DEVIANT DEVIANT 44 LR 1.01 2.94 0.34 0.78 6.63 -1.34 0.14 1.46 DEVIANT 46 LR 0.26 0.96 -0.65 1.93 -0.20 -0.13 -1.62 DEVIANT DEVIANT 48 LR 0.69 0.09 0.61 0.89 -0.72 -0.09 1.11 0.62 50 LR -2.22 -3.02 1.31 3.46 -0.83 0.38 0.48 -0.68 DEVIANT 52 LR 1.17 5.58 -0.64 -0.07 -1.02 0.62 -0.86 0.00 54 LR 1.53 -1.63 0.42 -0.41 1.35 -1.18 0.66 -0.31 56 LR -0.71 -3.49 0.87 9.92 -0.27 1.18 -0.15 -1.49 DEVIANT DEVIANT 58 LR -0.04 0.12 0.60 -1.03 0.89 -0.27 2.03 -0.83 60 LR 1.34 -4.75 -1.34 -0.75 -0.48 -2.00 0.16 0.17 DEVIANT DEVIANT 62 LR -0.61 -0.10 -0.71 -0.49 -0.78 -1.15 2.19 DEVIANT Note. a Scores were log10-transformed prior to analysis. b Because of missing scores on two phonological tests, no phonological factors could be calculated for subject 61. Notwithstanding, based on the scores of the administered tests, it was determined that this subject showed a deficit in phonological awareness. Note. For the Phoneme Boundary Slope, Phonological Awareness, RAN and Verbal STM a higher Z-score indicates better processing; in contrast, for all other measures a higher Z-score indicates reduced sensory sensitivity.
87
Manuscript 3
Coherent motion detection in preschool children at family risk for dyslexia1
Abstract
We tested sensitivity to coherent motion (CM) in random dot kinematograms in a group of five-year-old
preschool children genetically at risk for dyslexia, compared to a group of well-matched control children. No
significant differences were observed, either in a group analysis or in an individual deviance analysis.
Nonetheless, CM-thresholds were significantly related to emerging orthographic skills. In a previous study on
the same subjects (Boets, van Wieringen, Wouters, Ghesquière, 2006), we demonstrated that both risk groups
already differed on measures of phonological awareness and letter knowledge. Moreover, auditory spectral
processing (especially 2 Hz FM detection) was significantly related to phonological ability. In sum, the actual
visual and previous auditory data combined, seem to suggest an exclusive relation between CM sensitivity and
orthographic skills on the one hand, and FM sensitivity and phonological skills on the other.
Introduction
Developmental dyslexia is a specific failure to acquire reading and spelling skills
despite adequate intelligence and education, affecting around 5-10 % of children and adults.
The predominant etiological view postulates that dyslexia results from a phonological deficit
(Snowling, 2000). However, extensive research during the last decade also demonstrated a
specific sensory processing deficit in individuals with dyslexia and it has been suggested that
this deficit might be causal to both the observed phonological and literacy problems (Farmer
& Klein, 1995; Stein, 2001). To investigate the assumed causality of this sensorial deficit
hypothesis we assessed auditory and visual processing in two contrasting groups of five-year-
old preschool children, a genetically high risk and a genetically low risk group. In a previous
paper (Boets et al., 2006) we reported the absence of a significant group difference for any of
three administered auditory measures, in the presence of a significant difference for
phonological awareness and letter knowledge. However, spectral auditory tasks (particularly 2
Hz frequency modulation detection) turned out to be highly significantly related to
88
phonological awareness. In this paper, we will focus upon sensory processing in the visual
modality, assessed in the same group of preschool children. In particular, we consider the
question whether a deficit in coherent motion processing may already be observable in
preschool children at risk of dyslexia and we investigate the relationship between motion
processing and developing literacy skills.
Within the visual modality, dyslexia research has mainly focused upon sensory
processing in the magnocellular visual pathway. Early studies using stimuli that assess the
peripheral visual system (e.g., contrast sensitivity and flicker sensitivity paradigms)
demonstrated that dyslexics tend to show a deficit in processing stimuli with low spatial and
high temporal resolution (for a review, see Lovegrove, 1996, but see Skottun, 2000 for a
critical revision). More recently, interesting results have also been obtained with stimuli that
imply higher level magnocellular functioning such as coherent motion detection tasks (CM).
These tasks, relying predominantly upon processing in area V5/MT of the cortex, have proven
to differentiate relatively reliable between groups of dyslexic and normal reading subjects
(Witton et al., 1998; Raymond & Sorensen, 1998; Talcott, Hansen, Assoku, & Stein, 2000;
Hansen, Stein, Orde, Winter, & Talcott, 2001; Ridder, Borsting, & Banton, 2001; Talcott et
al., 2003; Cornelissen, Richardson, Mason, Fowler, & Stein, 1995; Everatt, Bradshaw, &
Hibbard, 1999; Van Ingelghem, Boets, van Wieringen, Ghesquière, & Wouters, 2004;
Wilmer, Richardson, Chen, & Stein, 2004). Moreover, functional imaging studies have
confirmed that activation of area V5/MT in response to coherent motion stimuli was not as
robust in dyslexics compared to controls (Eden et al., 1996). Demb, Boyner and Heeger
(1997) have even demonstrated a reliable relation between the magnitude of the
hemodynamic BOLD-response in extrastriate area MT and overall reading skills in dyslexic
subjects. In psychophysical studies too, sensitivity to motion stimuli has been related to
(nonword) reading ability (Van Ingelghem et al., 2004; Witton et al., 1998, Talcott et al.,
1998), orthographic ability (Talcott et al., 2000; Talcott et al., 2002; Van Ingelghem et al.,
2004) and letter position encoding (Cornelissen et al., 1998). However, evidence of a motion
coherence deficit in dyslexia is not yet unequivocal since some studies failed to find
differentiating thresholds (Kronbichler, Hutzler, & Wimmer, 2002; Ramus et al., 2003;
Amitay, Ben-Yehudah, Banai, & Ahissar, 2002; Hulslander et al., 2004). Moreover, a deficit
in motion processing might not be an exclusive characteristic of dyslexia, since it has also
been demonstrated in other developmental disorders like for example autism (Milne et al.,
2002) and Williams-syndrome (see e.g., Atkinson et al., 1997).
89
Regarding the specific mechanism by which coherent motion sensitivity may limit
normal literacy development, there is still much speculation. Since CM-thresholds are a
robust measure of magnocellular processing and since this visual subsystem is mostly
involved in encoding spatial information, it is probable that poor magnocellular functioning
might result in uncertainty about letter position while reading and writing (Cornelissen et al.,
1998). Furthermore, a magnocellular deficit has also been related to binocular instability and
poor eye movement control, visual attention and visual search – all factors that might interfere
with the development of orthographic skills and subsequent reading and spelling skills (Stein
& Talcott, 1999; Stein, 2001; Talcott et al., 2000).
Notwithstanding the considerable empirical evidence that CM-thresholds are able to
differentiate reliably between adult and school-aged dyslexic and normal reading subjects, the
differentiating and predictive power of this task has never been investigated in preschool
children. In this study we want to address this issue. Furthermore, to investigate the specific
relation between sensory processing and different aspects of literacy development, we will
also integrate the previously administered phonological measures and the 2 Hz FM detection
thresholds in our correlation analyses. This has been done since in a series of former studies
CM detection has gradually been linked to FM detection (see e.g. Witton et al., 1998; Talcott
et al., 2000; Talcott et al., 2002; Talcott et al., 2003). According to Talcott, Witton and
colleagues both psychophysical tasks could be regarded as ‘dynamic’ stimuli tasks by relying
upon long-duration stimuli that require the perception of a dimension changing in time
(‘perception of rate’). While CM detection depends on the successful detection and
integration of local motion signals over both time and space, FM detection depends on
tracking the dynamic changes in the frequency of a tone over time. Interestingly, in some
recent studies where FM and CM-detection tasks have been administered to the same subjects
(school-aged children), it has been demonstrated that orthographic skills co-vary most
strongly with CM sensitivity, whereas phonological skills co-vary most strongly with FM
sensitivity (Talcott et al., 2000; Talcott & Witton, 2002). In this study we will explore
whether these specific relations might already be present in preschool subjects.
90
Method
Participants
Sixty-two five-year-old children attending the last year of kindergarten were included
in the study (36 boys / 26 girls). Half of the participants were children of ‘dyslexic families’,
the so-called high-risk group (HR); the other half were control children of ‘normal reading
families’, the so-called low-risk group (LR). All children were native Dutch speakers without
any history of brain damage, long term hearing loss or visual problems. Additionally, at the
moment of data collection they did not present any gross deficiencies in visual acuity
(Landolt-C single optotypes Snellen acuity > 0.85) and/or audiology (audiometric pure-tone
average < 25 dB HL). The HR children were selected on a basis of having at least one first-
degree relative with a diagnosis of dyslexia. The LR children showed no history of speech or
language problems and none of their family members suffered any learning or language
problems. For every individual HR child we selected the best matching LR control child
based on five criteria: (1) educational environment, i.e. same nursery school, (2) gender, (3)
age, (4) nonverbal intelligence, and (5) parental educational level. Nonverbal intelligence was
assessed by an adapted version of the Raven Coloured Progressive Matrices (RCPM) (Raven,
Court, & Raven, 1984), a collective non-verbal intelligence test measuring spatial reasoning.
Parental educational level was assessed using the ISCED-scale (International Standard
Classification of Education by UNESCO, 1997), by converting classifications on the original
seven-point scale to a three-point scale. Further details about the participants and the selection
procedure are described in Boets et al., (2006).
Table 1 gives descriptive characteristics of both groups. At the time of collecting the
visual data the mean age for both the HR and LR group was 5 years and 8 months, not being
statistically different (p = .83). The nonverbal IQ scores were slightly above population
average (107 for HR group and 111 for LR group) and did not differ significantly (p = 0.07).
Fisher’s Exact Test also confirms that both groups did not differ in frequency distribution of
the different educational categories (p = .71 for maternal and p = .43 for paternal educational
level).
91
Table 1
Characteristics of participants and coherent motion detection thresholds.
HR LR
Measures M SD M SD p
Nonverbal IQ
Age in months
Best CM1 (% coherence)
Best CM2 (% coherence)
AV CM ½ (% coherence)
107
68
0.19
0.22
0.20
14
3
0.11
0.13
0.12
111
68
0.17
0.19
0.18
13
3
0.08
0.09
0.09
.07 a
.83 a
.28 b
.14 b
.19 b
Note: BestCM1-2: best and second best CM threshold, AVCM ½: average of the two best CM thresholds; a paired t-test; b paired wise MMA controlled for nonverbal IQ, age and parental educational level.
Apparatus
Phonological tests. Phonological skills were assessed by a broad test battery
comprising eight tests. A principal component factor analysis with varimax rotation
confirmed that the battery reflected the three traditional phonological domains: (a)
phonological awareness (AWARENESS): high loadings from first-sound and end-sound
identity task, rhyme identity task and simple rhyme task, (b) rapid automatic naming (RAN):
high loadings from both the colours and objects rapid automatic naming tasks and (c) verbal
short-term memory (VSTM): high loadings from the nonword repetition task and the digit
span forward task. Details about the composition of the battery can be found in Boets et al.,
(2006); in this paper we will only refer to the factor scores.
Productive letter knowledge. This task was intended as a preliminary measure of
literacy development. The sixteen most frequently used letters in Dutch books were presented
on a card and the child had to name each of these letters. Both the sound and the name of a
letter were considered correct.
CM-detection test. For the CM-detection test, children were sitting in a low-luminance
(mesopic) environment at 40 cm distance from an Elo Intuitive 1725L 17’’ touch screen (75
Hz refresh rate) on which the random dot kinematograms (RDK) were displayed. The display
resolution was set to 640 x 480 pixels. The stimuli were generated online by a portable
computer (Dell Latitude C800 and Toshiba Satellite 1400-103) and comprised of two
rectangular patches, each containing 1103 randomly moving high luminance white dots on a
black background (dot size = 1 pixel or 0.07° diameter, dot density = 2.5 dots/deg2,
92
velocity = 7.3 deg/sec, life time = 5 video frames or 200 msec, maximal duration of stimulus
presentation = 6 sec, luminance of dots = 125 cd/m2, luminance of background = 0.39 cd/m2,
Michelson contrast = 99.4 %). At a viewing distance of 40 cm each patch of dots subtended
16 x 27.2° visual angle, separated horizontally by 3.8°. The target patch was segregated into
three horizontal strips (see Gunn et al., 2002); in the middle strip a variable proportion of dots
were moving coherently in horizontal direction, reversing direction every 330 msec – creating
as such the impression of ‘an emerging road in the snow’. All other dots were moving
randomly in a Brownian manner. The two patches were presented simultaneously and the
subject had to identify the patch containing the strip with coherently moving dots. Threshold
was defined as the smallest proportion of coherently moving dots required for detection of the
middle strip with reversing dot motion. Thresholds were estimated using a two-down, one-up
adaptive staircase paradigm, which targeted the threshold corresponding to 70.7 % correct
responses (Levitt, 1971). Percentage coherence in the middle strip of the target patch started
at 100 % and decreased with a factor of 1.16. After four reversals factor 1.14 was used. A
threshold run was terminated after eight reversals and thresholds for an individual run were
calculated by the geometric mean of the values of the last four reversals. For every subject
four thresholds were determined. Prior to data collection, participants were given a short
period of practice, comprising supra-threshold trials, to familiarise them with the stimuli and
the task.
To ensure the child’s attention and motivation we integrated the psychophysical test in
a computer game with animation movies, aimed to transform the abstract meaningless stimuli
into a concrete and well-known ‘daily life signal’. Before administering the CM-detection
experiment children watched an introductory animation movie about a little dog and a little
bear getting lost in the snow (see Fig. 1A). The children were asked to help them find their
way home again by visually inspecting each stimulus patch and reporting which patch
contains the road to get home (inspired by work of Atkinson et al., 2003). Immediately after
the child’s response, corresponding auditory feedback was presented.
93
A B
Fig. 1. Screenshots of the animation movies. (A) A frame of the introductory and concluding movie used to
animate the CM detection experiment, (B) a frame of the introductory and concluding movie used for the FM
detection experiment.
FM-detection test. In this test participants had to detect a 2 Hz sinusoidal frequency
modulation of a 1 kHz carrier tone with varying modulation depth. Threshold was defined as
the minimum depth of frequency deviation required to detect the modulation. Modulation
depth decreased with a factor 1.2 from 100 Hz towards 11 Hz, from where a fixed step size of
1 Hz was used. The length of both the reference and the target stimulus was 1000 ms
including 50 ms cosine-gated onset and offset. Stimuli were generated in MATLAB 5.1 and
saved as 16-bit wav-files (sample frequency 44100 Hz) on the hard disc of the same portable
computers as used for the CM-experiment. They were presented using an integrated audio
PC-card and routed to an audiometer (Madsen OB622) in order to control the level of
presentation. The stimuli were presented monaurally over a calibrated TDH-39 headphone at
70 dB SPL with an ISI of 350 ms. FM-thresholds were estimated using a three-interval
forced-choice oddity paradigm embedded within an interactive computer game with
animation movies (see Fig. 1B) (Laneau, Boets, Moonen, van Wieringen & Wouters, 2005).
Similarly to the CM-experiment a two-down, one-up adaptive staircase procedure was used
94
and threshold was calculated as the geometric mean of the values of the last four of eight
reversals. After a short period of practice, three thresholds were determined for every subject.
For the correlation data we present here, the average of the best and second best threshold was
used as an indicator of FM sensitivity. A more detailed description of the stimuli, procedure
and results can be found in Boets et al., (2006).
Data collection was carried out by qualified psychologists and audiologists. Testing
took place in a quiet room at the children’s school. Since the LR child was selected from the
HR child’s classmates, both children could always be tested in exactly the same
circumstances.
Statistical Analysis
Prior to analysis, psychophysical thresholds were log10-transformed in order to obtain
normally distributed data. All results were analysed in a paired wise manner, comparing HR
versus LR group at the level of the matched individuals. Although the groups did not show a
significant difference on any of the matching criteria, we decided to rule out any possible
influence of age, nonverbal intelligence or parental educational level by controlling for these
variables in our analysis. As such, we analysed the data using Mixed Model Analysis (MMA)
with pair number (school) as a random variable (1 to 31) and participant group (HR versus
LR) as the fixed between-subject variable (Littell, Milliken, Stroup, & Wolfinger, 1996). Age,
nonverbal IQ and educational level of both mother and father were added as fixed
(co)variables. Relationships between variables were analysed using Spearman correlation
coefficients, partialed out for the influence of nonverbal intelligence.
Results
A paired wise Repeated Measures MMA with group as between-subject variable (HR
versus LR), threshold run as within-subject variable (run 1 to 4) and with the same covariates
as mentioned above, revealed no significant effect for group (p = .20), a significant effect for
threshold run (p < .0001) and no significant group x run interaction effect (p = .25). Post-hoc
analysis revealed that none of the four CM threshold measures differentiated significantly
between HR and LR group. Furthermore, there was only a significant learning effect from the
first to the second run; the second, third and fourth run did not differ significantly from each
other.
95
Although the Repeated Measures MMA revealed a general learning effect from the
first to the second run, this tendency did not apply to all subjects. For many of them, the first
threshold was better than the second, third or fourth. Moreover, since we are interested in
threshold estimation as an indicator of a subject’s true sensory capability – i.e. the best level
of performance a subject is able to reach, regardless of interfering factors like for example
fluctuations in concentration and motivation – average threshold (or the average of the last
two or three threshold runs) might not be the most appropriate measure. To cope with the high
intrasubject variability which is typical for younger children (see e.g. Wightman & Allen,
1992), a more reasonable estimator is each subject’s ‘best’ performance, or the lowest
threshold of the different runs. For this reason we used the average of the best and second best
threshold as a true indicator of a child’s CM sensitivity. The Spearman rank correlation
between this best and second best threshold estimate appeared to be very satisfactory (rs = .91,
p < .0001), indicating a reliable threshold estimation.
Threshold estimates and test statistics for the best and second best threshold and for
the average of the two best thresholds are given in Table 1. Although the observed difference
was in the expected direction with the HR-group scoring less well than the LR-group, the
coherent motion detection task did not differentiate significantly between the groups. It is
worth mentioning that these null-results were not merely the consequence of applying such a
strict controlling MMA design, as the results were virtually identical when the analysis was
repeated without any covariates added.
Since group comparisons might mask major individual differences, we also carried out
an analysis at the subject level. To decide which individual did and did not show abnormal
performance, we adopted the two-step criterion as suggested by Ramus et al., (2003).
Applying this procedure, the criterion for deviance was placed at 1.65 standard deviations of
the ‘purified’ mean of the LR-group, after first having excluded all deviant LR-subjects (by
applying a similar 1.65 SD criterion, resulting in the removal of two deviant LR subjects). A
distribution analysis on the coherent motion thresholds of the ‘restricted’ LR-group confirmed
the normality of the data. Hence, the 1.65 SD deviance criterion corresponds to the fifth
percentile and is thus a fairly strict criterion. The individual deviance analysis for the
averaged best and second best coherent motion threshold revealed that the proportion of
subjects showing abnormal performance was equal in both groups: four subjects in the HR
group and four subjects in the LR group scored below the fifth percentile. This corresponded
to 13 % of each group.
96
Table 2
Spearman (partial) correlations for total group of subjects (n = 61/62)
Age RCPM
Letter
Knowledge AWARENESS RAN VSTM AV FM ½ AV CM ½
Letter Knowledge
AWARENESS
RAN
VSTM
AV FM ½
AV CM ½
-0.08
-0.08
0.25*
0.07
0.15
0.18
0.26*
0.26*
-0.21 0.19
-0.16
-0.30*
-
0.46***
0.03
-0.02
-0.40**
-0.33**
0.42***
-
-0.08
-0.04
-0.49****
-0.18
0.09
-0.02
-
-0.02
-0.08
0.01
-0.07
-0.10
0.02
-
-0.08
-0.10
-0.36**
-0.48****
-0.12
-0.05
-
0.29*
-0.29*
-0.11
-0.06
-0.04
0.26*
-
Note. Coefficients above the diagonal are partial correlations after removing variance attributable to individual
differences in nonverbal intelligence (RCPM).
* p < .05, ** p < .01, *** p < .001, **** p < .0001
Table 2 shows Spearman rank interrelations between psychophysical thresholds,
phonological ability and letter knowledge, and their relation to age and nonverbal IQ. Neither
CM, nor FM were related to age, and only CM-detection showed a significant relation to
nonverbal intelligence. To exclude the variance attributable to individual differences in
intelligence, all further correlations have been partialed out for the influence of nonverbal IQ.
Table 3 offers similar Spearman rank correlations for both risk groups separately.
Table 3
Spearman partial correlations for LR and HR group separately (after removing variance attributable to individual
differences in nonverbal intelligence)
Letter
Knowledge AWARENESS RAN VSTM AV FM ½ AV CM ½
Letter Knowledge
AWARENESS
RAN
Verbal STM
AV FM 1/2
AV CM 1/2
-
0.46**
-0.14
-0.18
-0.38**
-0.19
0.17
-
-0.04
-0.14
-0.35*
0.09
0.17
-0.11
-
0.30
-0.33*
-0.25
0.01
-0.08
-0.29
-
0.07
-0.16
-0.31*
-0.61****
0.13
-0.16
-
0.08
-0.41**
-0.29
0.11
0.06
0.44**
-
Note. Coefficients above the diagonal concern the LR group (n = 31), coefficients under the diagonal concern the
HR group (n=30).
* p < .10, ** p < .05, *** p < .01, **** p < .001
FM-detection was highly significantly related to Phonological Awareness and – to a
lesser extent - to Productive Letter Knowledge. There was no correlation with Verbal STM
and only in the HR group the correlation with Rapid Automatic Naming was significant. The
relation with Phonological Awareness seemed to be the most robust in the LR-group where 37
% of the variance in phonological awareness could be predicted from sensitivity to FM. CM-
97
detection on the other hand, turned out to be completely unrelated to any phonological
measure, but was significantly related to Letter Knowledge. However, this relation only
seemed to hold for the LR-group.
In previous research CM-detection has been found to be specifically related to
orthographic skills. Since it is impossible to administer a pure orthographic test at this
preschool age, we considered Letter Knowledge as the best approximate measure to obtain an
indication about orthographic ability. After all, resolving a letter knowledge task relies on
recognizing the visual features of the written symbol on the one hand, and retrieving the
corresponding linguistic information on the other. Therefore, Letter Knowledge might be
regarded as a measure that reflects both orthographic and phonological skills. To create a
more ‘pure’ orthographic measure we extracted all the phonological aspects out of the letter
knowledge task by statistically removing all the variance due to differences in phonological
awareness, rapid automatic naming and verbal short-term memory. Concretely, we calculated
the Spearman correlation between CM and Letter Knowledge and added nonverbal IQ,
AWARENESS, RAN and VSTM as partial variables. In doing so, we still observed a
significant relation between CM and the orthographic aspects of Letter Knowledge, both in
the total group (rs = -.27, p = .04) and in the LR-group (rs = -.41, p = .03). For the HR-group
this relation was considerable but not significant (rs = -.31, p = .11). This means, that again
the relation turned out to be the most substantial in the LR-group where 17 % of the variance
in orthographic skills could be predicted from sensitivity to CM-detection. Interestingly, in
contrast, the correlation between FM and the orthographic aspect of Letter Knowledge was
not significant (in neither of the two groups). Considering the whole of these observations,
this seems to imply an exclusive relation between FM and phonological awareness on the one
hand, and CM and orthographic skills on the other (see Figs. 2A and 2B).
Although FM and CM appeared to be related to different and dissociated aspects of
literacy development, they also shared some common variance. Indeed, both in the total group
and in the LR-group FM and CM were significantly related to each other. Again, this relation
was not present in the HR-group. This might imply that both psychophysical tasks (at least in
the LR-group) rely upon some common neurological mechanism involved in ‘dynamic
processing’. Importantly, this hypothesised common mechanism is not involved in
psychophysical processing in general, since CM proved to be completely unrelated to two
other auditory measures that were also administered to the same subjects (i.e. gap detection in
broadband noise and tone-in-noise detection; for details see Boets, et al., 2006).
98
Fig. 2. (A) Thresholds for detecting coherent motion plotted against a Productive Letter Knowledge measure for
31 HR and 31 LR subjects. (B) Thresholds for detecting 2 Hz FM of a 1000 Hz tone plotted against a combined
measure of Phonological Awareness for 30 HR and 31 LR subjects. HR subjects: filled diamonds; LR subjects:
empty squares.
Discussion
In this study we tested sensory processing and phonological and orthographic abilities
in five-year-old preschool subjects who never received any formal reading instruction. As
reported previously, a significant deficit in phonological awareness and letter knowledge
could be demonstrated in the dyslexia-prone HR group. Consequently, since both letter
knowledge and phonological awareness have consistently been proven to be the best
preschool predictors of literacy development (see e.g. Scarborough, 1998), it is likely that the
genetically high risk group will contain a disproportionally high number of future cases of
dyslexia.
For CM detection we did not observe any significant differences between the high and
low risk groups, either at a group level, or in the individual deviance analysis. Although these
results are consistent with a minority of studies that also failed to find any group differences
(Kronbichler et al., 2002; Amitay et al., 2002; Ramus et al., 2003; Hulslander et al., 2004),
they clearly conflict with the mass of evidence provided by most other studies. The same
applied to the FM detection data that did not reveal a group difference either. A
straightforward explanation for this lack of a group difference might be the fact that we did
not study a well-defined clinical group but only an at risk group that still might show
substantial overlap with the control group. Moreover, Bishop et al., (1999) demonstrated in a
0,8
1
1,2
1,4
1,6
1,8
2
0 5 10 15 20
Productive Letter Knowledge
log
CM
(%)
0
0,2
0,4
0,6
0,8
1
1,2
1,4
-3 -2 -1 0 1 2 3
Phonological Awareness
log
FM (H
z)
99
twin study on SLI children that in contrast to the highly heritable phonological skills, auditory
skills depend less on genetic and more on environmental influences. As such, our finding of a
phonological deficit in combination with relatively intact sensorial skills in this genetic high
risk group corresponds well with the results of Bishop and colleagues.
Considering the relations between sensory processing and the orthographic and
phonological sub skills of literacy development, our preschool data convincingly confirm
previous results from adults and school-aged children (Talcott et al., 2000; Talcott & Witton,
2002). Even while taking into account the influence of general cognitive ability, sensitivity to
CM seems to be uniquely related to orthographic skills, whereas sensitivity to FM seems to be
specifically related to phonological skills. This relation appears to be the most robust in the
LR group where 17% of the variance in orthographic skills could be predicted from
differences in sensitivity to CM and 37% of the variance in phonological awareness could be
predicted from sensitivity to FM.
The finding of these more substantial correlations in the LR/control group compared to
the HR/dyslexic group is in line with most other studies (Rosen, 2003). Ahissar, Protopapas,
Reid, and Merzenich (2000) have hypothesized that the suppression of these correlations in
adult dyslexics might be due to variably compensated literacy skills, in contrast to the lagging
sensory skills. However, our data do not provide much support for this interpretation, since
we observed a similar pattern in preschool children who have not even been diagnosed or
detected as being dyslexic. Actually, the children studied had not yet received any formal
reading instruction; indeed, they have had no opportunity to compensate or be treated for their
undetected impairments.
With respect to the causality of the observed relations between literacy skills and basic
sensory measures, it has been suggested that better sensorial sensitivity might be a
consequence and not a cause of better literacy skills (Talcott & Witton, 2002). Indeed, based
on adult and school-aged data, the possibility cannot be ruled out that reading experience (or
print exposure) improves CM and FM detection performance rather than vice versa. In fact, it
would not be too far-fetched to expect visual and auditory skills of good readers to be more
finely tuned than those of dyslexics by virtue of their more highly trained orthographic and
phonological systems. However, this study on preschool subjects demonstrated that there
already exists a reliable preceding relation between sensory and preliminary literacy skills,
even before having received any instruction or before having been exposed extensively to a
lot of print. Therefore, it seems unlikely to consider the sensorial deficits in dyslexics as a
consequence of lack of reading experience. Instead, the results of our study seem to be
100
consistent with the general hypothesis that basic sensory processing skills do influence (albeit
in a facilitating or in an inhibiting way) the development of phonological, orthographic and
reading abilities.
To conclude, phonological awareness and letter knowledge turn out to be the best
indicators to differentiate between preschool children with low versus high genetic risk of
developing dyslexia. In contrast, neither visual coherent motion detection nor auditory 2 Hz
frequency modulation detection is able to differentiate significantly between both groups.
Nevertheless, there is a significant relation between these dynamic sensory measures and
developing literacy skills, even while taking into account the influence of differences in
intelligence: sensitivity to CM is uniquely related to orthographic skills and not to
phonological ability, whereas sensitivity to FM is specifically predictive for emerging
phonological skills and not for orthographic skills. In sum, these results suggest that basic
visual and auditory sensitivity is likely to play an important role in the development of fine-
grained orthographic and phonological representations necessary for successful reading.
101
References
Ahissar, M., Protopapas, A., Reid, M., & Merzenich, M. (2000). Auditory processing parallels
reading abilities in adults. Proceedings of the National Academy of Science, 97 (12),
6832-6837.
Amitay, S., Ben-Yehudah, G., Banai, K., & Ahissar, M. (2002). Disabled readers suffer from
visual and auditory impairments but not from a specific magnocellular deficit. Brain,
125(10), 2272-2284.
Atkinson, J., King, J., Braddick, O., Nokes, L., Anker, S., & Braddick, F. (1997). A specific
deficit of dorsal stream function in Williams' syndrome. Neuroreport, 8(8), 1919-1922.
Atkinson, J., Braddick, O., Anker, S., Curran, W., Andrew, R., Wattam-Bell, J., & Braddick,
F. (2003). Neurobiological models of visuospatial cognition in children with Williams
syndrome: measures of dorsal-stream and frontal function. Developmental
Neuropsychology, 23, 139-172.
Bishop, D. V., Bishop, S. J., Bright, P., James, C., Delaney, T., & Tallal, P. (1999). Different
origin of auditory and phonological processing problems in children with language
impairment: evidence from a twin study. Journal of Speech, Language and Hearing
Research, 42(1), 155-168.
Boets, B., Wouters, J., van Wieringen, A., & Ghesquière, P. (2006). Auditory temporal
information processing in preschool children at family risk for dyslexia: relations with
phonological abilities and developing literacy skills. Brain and Language, 97, 64-79.
Cornelissen, P., Richardson, A., Mason, A., Fowler, S., & Stein, J. (1995). Contrast sensitivity
and coherent motion detection measures at photopic luminance levels in dyslexics and
controls. Vision Research, 35, 1483-1494.
Cornelissen, P. L., Hansen, P. C., Gilchrist, I. , Cormack, F., Essex, J., & Frankish, C. (1998).
Coherent motion detection and letter position encoding. Vision Research, 38, 2181-
2191.
Demb, J., Boynton, G., & Heeger, G. (1997). Brain activity in visual cortex predicts
individual differences in reading performance. Proceedings of the National Academy
102
of Sciences, USA, 94, 13363-13366.
Eden, G., Van Meter, J., Rumsey, J., Maisog, J., Woods, R., & Zeffiro, T. (1996). Abnormal
processing of visual motion in dyslexia revealed by functional brain imaging. Nature,
382, 66-69.
Everatt, J., Bradshaw, M., & Hibbard, P. (1999). Visual processing and dyslexia. Perception,
28, 243-254.
Farmer, M. E., & Klein, R. M. (1995). The evidence for a temporal processing deficit linked
to dyslexia: A review. Psychonomic Bulletin and Review, 2(4), 460-493.
Gunn, A., Cory, E., Atkinson, J., Braddick, O., Wattam-Bel, J., Guzzetta, A., & Cioni, G.
(2002). Dorsal and ventral stream sensitivity in normal development and hemiplegia.
Neuroreport, 13(6), 843-847.
Hansen, P. C., Stein, J. F., Orde, S. R., Winter, J. L., & Talcott, J. B. (2001). Are dyslexics'
visual deficits limited to measures of dorsal stream function? Neuroreport, 12(7),
1527-1530.
Hulslander, J., Talcott, J., Witton, C., DeFries, J., Pennington, B., Wadsworth, S., Willcutt, E.,
& Olson, R. (2004). Sensory processing, reading, IQ and attention. Journal of
Experimental Child Psychology, 88, 274-295.
Kronbichler, M., Hutzler, F., & Wimmer, H. (2002). Dyslexia: verbal impairments in the
absence of magnocellular impairments. Neuroreport, 13( 5), 617-620.
Laneau, J., Boets, B., Moonen, M., van Wieringen, A., & Wouters, J. (2005). A flexible
auditory research platform using acoustic or electric stimuli for adults and young
children. Journal of Neuroscience Methods, 142(1), 131-136.
Levitt, H. (1971). Transformed up-down methods in psychoacoustics. Journal of the
Acoustical Society of America, 49, 467-477.
Littell, R. C., Milliken, G. A., Stroup, W. W., & Wolfinger, R. D. (1996). SAS system for
mixed models. Cary, NC: Sas Institute INC.
Lovegrove, B. (1996). Dyslexia and a transient/magnocellular pathway deficit: The current
103
situation and future directions. Australian Journal of Psychology, 48(3), 167-171.
Milne, E., Swettenham, J., Hansen, P., Campbell, R., Jeffries, H., & Plaisted, K. (2002). High
motion coherence thresholds in children with autism. Journal of Child Psychology and
Psychiatry, 43(2), 255-263.
Ramus, F., Rosen, S., Dakin, S. C., Day, B. L., Castellote, J. M., White, S., & Frith, U.
(2003). Theories of developmental dyslexia: insights from a multiple case study of
dyslexic adults. Brain, 126(4), 841-865.
Raven, J. C., Court, J. H., & Raven, J. (1984). Manual for Raven's Progressive Matrices and
Vocabulary Scales. London: Lewis.
Raymond, J., & Sorensen, R. (1998). Visual motion perception in children with dyslexia:
normal detection but abnormal integration. Visual Cognition, 5(3), 389-404.
Ridder, W. H., Borsting, E., & Banton, T. (2001). All developmental dyslexic subtypes
display an elevated motion coherence threshold. Optometry and Vision Science, 78,
510-517.
Rosen, S. (2003). Auditory processing in dyslexia and specific language impairment: is there
a deficit? What is its nature? Does it explain anything? Journal of Phonetics, 31, 509-
527.
Scarborough, H. S. (1998). Early identification of children at risk for reading disabilities:
phonological awareness and some other promising predictors. In P.J. Accardo, A.J.
Capute & B.K. Shapiro (Eds.), Specific reading disability: a view of the spectrum (pp
75-119). Timonium, MD: York Press.
Skottun, B. C. (2000). The magnocellular deficit theory of dyslexia: the evidence from
contrast sensitivity. Vision Research, 40, 111-127.
Snowling, M. J. (2000). Dyslexia (2nd ed.). Malden, MA: Blackwell Publishers.
Stein, J., & Talcott, J. (1999). Impaired neuronal timing in developmental dyslexia. The
magnocellular hypothesis. Dyslexia, 5, 59-77.
Stein, J. (2001). The magnocellular theory of developmental dyslexia. Dyslexia, 7, 12-36.
104
Talcott, J., Hansen, P., Willis-Owen, C., McKinnell, I., Richardson, A., & Stein, J. (1998).
Visual magnocellular impairment in adut developmental dyslexics. Neuro-
Ophtalmology, 20, 187-201.
Talcott, J. B., Hansen, P. C., Assoku, E. L., & Stein, J. F. (2000). Visual motion sensitivity in
dyslexia: evidence for temporal and energy integration deficits. Neuropsychologia,
38(7), 935-943.
Talcott, J. B., Witton, C., McLean, M. F., Hansen, P. C., Rees, A., Green, G. G., & Stein, J. F.
(2000). Dynamic sensory sensitivity and children's word decoding skills. Proceedings
of the National Academy of Science, USA, 97( 6), 2952-2957.
Talcott, J. B., Witton, C., Hebb, G. S., Stoodley, C. J., Westwood, E. A., France, S. J.,
Hansen, P. C., & Stein, J. F. (2002). On the relationship between dynamic visual and
auditory processing and literacy skills; results from a large primary-school study.
Dyslexia, 8(4), 204-225.
Talcott, J. B., Gram, A., Van Ingelghem, M., Witton, C., Stein, J. F., & Toennessen, F. E.
(2003). Impaired sensitivity to dynamic stimuli in poor readers of a regular
orthography. Brain and Language, 87, 259-266.
Talcott, J. B., & Witton, C. (2002). A sensory linguistic approach to the development of
normal and impaired reading skills. In E. Witruk, A. Friederici, & T. Lachmann
(Eds.), Neuropsychology and cognition series. Basic functions of language and
language disorders. Dordrecht, Netherlands: Kluwer Academic Publishers.
Van Ingelghem, M., Boets, B., van Wieringen, A., Ghesquière, P., & Wouters, J. (2004).
Auditory and visual temporal processing in dyslexic and normal reading children
matched for sex, age and intellectual ability. Paper presented at 6th BDA International
Conference, 27-30 March 2004, University of Warwick, UK.
Wightman, F., & Allen, P. (1992). Individual differences in auditory capability among
preschool children. In L.A. Werner , & E.W. Rubel (Eds.), Developmental
Psychoacoustics (pp. 113-134). Washington DC: American Psychological
Association.
Wilmer, J. B., Richardson, A. J., Chen, Y., & Stein, J. F. (2004). Two visual motion
105
processing deficits in developmental dyslexia associated with different reading skills
deficits. Journal of Cognitive Neuroscience, 16(4), 528-540.
Witton, C., Talcott, J. B., Hansen, P. C., Richardson, A. J., Griffiths, T. D., Rees, A., Stein, J.
F., & Green, G. G. (1998). Sensitivity to dynamic auditory and visual stimuli predicts
nonword reading ability in both dyslexic and normal readers. Current Biology, 8(14),
791-797.
106
Notes
1 Published as Boets, B., Wouters, J., van Wieringen, A., & Ghesquière, P. (2006). Coherent
motion detection in preschool children at family risk for dyslexia. Vision Research, 46 (4),
527-535.
107
Manuscript 4
The development of early literacy skills among children at high risk for
dyslexia: an empirical evaluation of the general magnocellular theory
Abstract
We tested low-level auditory and visual processing, speech perception, phonological ability and letter knowledge
in a group of 5-year-old preschool children at high family risk for dyslexia (HR, n = 31), compared to a group of
well-matched low risk control children (LR, n = 31). Based on family risk status and first grade literacy
achievement (literacy delay: LD versus normal literacy: LN) children were categorized in three groups (HR-LD,
HR-LN and LR-LN) and preschool data were retrospectively reanalyzed. Compared to the LR-LN group, the
HR-LD group presented a significant deficit in letter knowledge, phonological ability, 2 Hz frequency
modulation detection, coherent motion detection, categorical perception and speech-in-noise perception.
Whereas for phonological ability the HR-LN group scored in between both other groups or at the level of the
HR-LD group, for sensory processing they performed as well as the LR-LN subjects. This longitudinal study is
among the first to demonstrate that the sensory deficit typically observed in adult and school-aged dyslexic
subjects, does precede the literacy problem and is not merely the result of lacking reading experience. Hence,
this leaves open the possibility of a potential causal influence of sensory processing upon literacy development
in a way as postulated by the general magnocellular theory.
In a subsequent study, the plausibility of this causal relation was demonstrated using structural equation
modeling. In particular, we demonstrated that dynamic auditory processing was related to speech perception,
which on its turn was related to phonological awareness. In the same way, dynamic visual processing was related
to orthographic ability. Subsequently, phonological awareness and orthographic ability – together with verbal
short-term memory - were unique predictors of reading and writing development.
Introduction
Developmental dyslexia is a specific learning disability that affects around 5 to 10 %
of children and adults. It is characterised by severe reading and spelling difficulties that are
persistent and resistant to usual teaching methods and remedial efforts. Historically, there has
108
been a longstanding discussion about the aetiology of these specific literacy problems. The
origin has been sought in the visual, the auditory as well as in the cognitive-linguistic domain.
At present, the predominant aetiological view postulates that dyslexia results from a
cognitive deficit that is specific to the representation and processing of speech sounds: this is
the phonological deficit theory (Snowling, 2000). However, during the last decades there has
been a growing number of studies demonstrating a deficit in low-level auditory and visual
processing in subjects with dyslexia and it has been suggested that this sensory deficit might
be causal to both the observed phonological and literacy problems (Farmer & Klein, 1995). In
the auditory modality, most emphasis has been given to temporal auditory processing. It has
been demonstrated that dyslexics have problems processing short, rapidly presented and
dynamic changing acoustic stimuli (e.g. Talcott & Witton, 2002; Tallal, 1980; Van Ingelghem
et al., 2005). Besides, dyslexics also tend to present subtle speech perception problems
(McBride-Chang, 1995). It has been hypothesised that the basic deficit in perceiving auditory
temporal cues causes a problem for the accurate detection of the rapid acoustical changes in
speech. Consequently, the speech perception problem causes a cascade of effects, starting
with the disruption of normal development of the phonological system and resulting in
problems learning to read and spell (Talcott & Witton, 2002; Tallal, 1980; Wright et al.,
1997). In this way, the supporters of the auditory temporal processing deficit theory do not
deny the existence of the phonological deficit, but rather see it as secondary to a more basic
auditory impairment (Ramus, 2003).
Alternatively, it has been suggested that the literacy problems of some dyslexics can
be traced back to a specific visual problem, in particular a problem in magnocellular visual
processing (Stein & Walsh, 1997). Anatomically as well as functionally, the visual system can
be divided in two independent but linked subsystems: the magnocellular (or transient)
pathway and the parvocellular (or sustained) pathway. The magnocellular pathway originates
from the ventral layers 1 and 2 of the lateral geniculate nucleus (LGN) and projects mainly to
the dorsal stream of visual processing which terminates at the posterior parietal cortex
(Livingstone & Hubel, 1987). It contains cells with large somas (magnocells) and highly
myelinated axons, hence promoting fast conduction velocities. It is highly sensitive to low
spatial and high temporal frequency stimulation and it responds preferentially to the onset and
offset of the stimulus. Consequently, the magnocellular system is predominantly a flicker or
motion detecting system. In psychophysical studies, it has been demonstrated that individuals
with dyslexia show a decreased sensitivity to stimuli within the magnocellular range (e.g.
Cornelissen et al., 1995; Lovegrove, 1996). This evidence for a magnocellular dysfunction
109
has further been confirmed by visual evoked potential and fMRI studies (Eden et al., 1996;
Demb, Boynton & Heeger, 1997) and by anatomical studies showing abnormalities of the
magnocellular layers of the LGN in the dyslexic brain (Livingstone, Rosen, Drislane &
Galaburda, 1991). Regarding the specific mechanism by which magnocellular visual
dysfunction may limit normal literacy development, there is still much speculation. The
answer probably lies in the anatomical connections from the magnocellular pathway to the
posterior parietal cortex (PPC). The PPC is known to be involved in normal eye movement
control, visuospatial attention, visual search and peripheral vision – all factors that are
obviously involved in the development of orthographic skills and subsequent reading and
spelling skills (Stein & Walsh, 1997; Stein & Talcott, 1999; Stein, 2001). Since the PPC is
dominated by magnocellular input, it is very well possible that slight impairments in
magnocellular functioning might multiply up to greater deficits in PPC functioning.
Although historically the auditory and the visual research lines have been developed
independently, they both converge in postulating that dyslexics suffer from a deficiency in
temporal information processing. As a consequence the hypothesis of a general temporal
processing deficit or pan-modal deficit is formulated (Eden, Stein, & Wood, 1995; Farmer &
Klein, 1995; Frith & Frith, 1996; Stein & Talcott, 1999). In this hypothesis the reading and
spelling problems of dyslexic people are attributed to a deficiency in the temporal processing
capacity of both the auditory and the visual system. Support for this pan-modal deficit
hypothesis was provided by studies reporting the co-occurrence of auditory and visual
temporal problems in certain individuals with dyslexia (Witton et al., 1998; Cestnick, 2001;
Van Ingelghem et al., 2001). Recently, an attempt has also been undertaken to relate the
hypothesised cross-modal deficiencies to one underlying biological cause. The general
magnocellular theory (Stein & Walsh, 1997; Stein, 2001) postulates that the magnocellular
deficit is not restricted to the visual pathway but could be generalized to all modalities (visual
and auditory as well as tactile). Although a similar magnocellular/parvocellular distinction has
not been described for the auditory brain areas, anatomical evidence for the auditory variant
of a magnocellular deficit has been found by Galaburda and colleagues (Galaburda, Menard
& Rosen, 1994), reporting an increased proportion of smaller neurons in the medial geniculate
nucleus of the dyslexic brain.
Notwithstanding the obvious attractiveness of the general magnocellular theory in
providing a unifying framework for all manifestations of dyslexia, it also has important
shortcomings and has been facing growing criticism in recent years (Ramus, 2001; Ramus,
2003; Ramus et al., 2003; White et al., in press). Besides failures to replicate an auditory or
110
visual deficit in dyslexia, an important line of criticism concerns the finding that the deficit –
when observed- does not always seem to imply ‘rapid/temporal/magnocellular’ processing.
Indeed, in some studies ‘rapid/magnocellular’ processing is found to be intact, whereas
‘slow/parvocellular’ processing is found to be impaired (see e.g. Skottun, 2000 for an
extensive critical review of the visual evidence; Rosen, 2003 for an evaluation of the auditory
evidence). Another line of criticism emphasizes that the sensory problems are usually only
observed in a relatively small proportion of dyslexics. Of course this makes it questionable
whether the sensory problem could be regarded as the principal cause of dyslexia (Bishop,
Carlyon, Deeks & Bishop, 1999; Ramus, 2003; Rosen, 2003).
A more general critique regarding the (extended) magnocellular theory concerns the
lack of empirical evidence, since the postulated causality of the theory has not yet been
confirmed by longitudinal data. Actually, only very few studies have been undertaken to
assess preschool sensory processing difficulties in children that will ultimately become
dyslexic. Some major prospective studies are coming about (e.g. Jyväskylä Longitudinal
Study of Dyslexia, see Lyytinen et al., 2001; Dutch Dyslexia Research Program, NWO, 1996;
Benasich & Tallal, 2002), but until now only limited reports have been published. Moreover,
three exceptional studies assessing auditory (Heath & Hogben, 2004; Share, Jorm, MacLean
& Matthews, 2002) and auditory and visual (Hood & Conlon, 2004) temporal order
judgement tasks in an unselected population of preschoolers, yielded mixed results relating
preschool sensory thresholds to later literacy development. This means that, although the
magnocellular theory considers dyslexics’ reading and spelling problems as the result of a
deviant developmental pathway in early childhood, the theory has almost exclusively been
based upon data collected in adult and school-aged subjects. Unfortunately however, these
data should not be interpreted in a causal way! It is evident that the mere finding of sensory
deficits in dyslexic subjects or the observation of significant relations between sensory
thresholds and aspects of literacy ability does not tell us anything about the direction of this
relation. Indeed, based on adult and school-aged studies the possibility cannot be ruled out
that the observed sensory deficits are rather the result than the cause of differences in literacy
ability and reading experience. In fact, some researchers did suggest that it would not be too
far-fetched to expect auditory and visual skills of good readers to be more finely tuned than
those of dyslexics by virtue of their more highly trained phonological and orthographic
systems (Talcott & Witton, 2002).
With this longitudinal study we aim to fill in this empirical gap and take a first step
towards exploring the hypothesised causality issue. Concretely, we assessed basic auditory
111
and visual skills, speech perception, phonological ability and letter knowledge in a group of 5-
year-old preschool children at family risk for dyslexia, compared to a group of well-matched
control children1. Auditory processing was assessed by means of three psychophysical tests:
gap-detection in noise (GAP), 2 Hz FM-detection (FM) and tone-in-noise detection (TN).
With the GAP-detection task, we tested the hypothesis of a deficit in ‘rapid and brief’
temporal processing. With the FM-detection task, we verified the hypothesis of a deficit in the
processing of ‘dynamic stimuli’. The TN task was included as a non-temporal control task to
verify the specificity of any observed auditory deficit, i.e. we wanted to examine whether a
deficit might be the result of failing performance on psychophysical tasks in general. To
assess visual magnocellular processing, we measured sensitivity to coherent motion (CM) in
random dot kinematograms. This task was chosen since it has been shown by single cell
recordings of activity in cortical area MT/V5 (strongly supplied by cells from the
magnocellular pathway) in response to coherent motion stimuli, that random dot
kinematograms provide a sensitive measure of magnocellular processing (Britten, Shalden,
Newsome & Movshon, 1992). To evaluate speech perception, we measured categorical
perception of a place-of-articulation speech continuum and we administered a speech-in-noise
perception task. Phonological processing was assessed by a broad test battery comprising
tasks for rapid automatic naming, verbal short-term memory and phonological awareness.
Finally, developing literacy skills were measured using a letter knowledge task.
In the first place, all preschool data were analysed comparing the familial high risk
(HR) versus low risk (LR) group. The results of this group comparison have been described in
three previous papers (Boets, Wouters, van Wieringen & Ghesquière, 2006a; Boets, Wouters,
van Wieringen, Ghesquière, 2006b; Boets, Ghesquière, van Wieringen & Wouters, under
revision) and can be summarised as follows. On every task the HR group demonstrated lower
performance than the LR group, with the group difference being statistically significant for
phonological awareness, letter knowledge and speech-in-noise perception. For the other tasks
the difference was in the expected direction but did not reach statistical significance. Further
on, significant correlations were observed between spectral auditory processing (FM- and TN-
detection), speech perception and phonological awareness on the one hand, and between
visual magnocellular processing (CM) and the specific orthographic aspect of letter
knowledge on the other.
Of course, the lack of a significant group difference for the sensory measures might be
attributed to the fact that we did not study a well-defined clinical group but only a risk group
that might still show substantial overlap with the non-affected control group. To clarify this
112
aspect, we followed up these children through first grade of primary school and assessed their
reading and spelling skills to determine whether or not they present literacy difficulties.
Accordingly, in study 1 of this paper we will describe the crucial retrospective analysis
comparing preschool (future) dyslexic children versus preschool (future) normal readers. We
postulate that if the sensory problems observed in adult and school-aged dyslexics are indeed
at the basis of the literacy problem, they should also be observable in preschool subjects who
will be diagnosed as dyslexic some time later.
Subsequently, in study 2 of the current paper we will explicitly evaluate the
hypothesised causal path by modeling the relations between sensory processing, speech
perception, orthographic and phonological ability and literacy development using structural
equation modeling procedures. According to the general magnocellular theory, we expect
visual magnocellular processing to influence literacy achievement by means of orthographic
ability. Likewise, we postulate that low-level auditory processing will influence phonological
processing by means of speech perception. Further on, we expect phonological ability (i.e.
phonological awareness, rapid automatic naming and verbal short-term memory) to determine
literacy development.
STUDY 1: Auditory, visual, speech perception and phonological deficits in
preschool children at high risk for dyslexia
Method
Participants
Sixty-two children were included in the study (36 boys / 26 girls) and followed from
one year before the onset of formal reading instruction till one year into reading instruction.
Half of the participants were children of ‘dyslexic families’, the so-called high-risk group
(HR); the other half were control children of ‘normal reading families’, the so-called low-risk
group (LR). All children were native Dutch speakers without any history of brain damage,
long-term hearing loss or visual problems. Additionally, at the moment of data collection they
did not present any gross deficiencies in visual acuity (Landolt-C single optotypes Snellen
acuity > 0.85) and/or audiology (audiometric pure-tone average < 25 dB HL). The HR
113
children were selected on a basis of having at least one first-degree relative with a diagnosis
of dyslexia. The LR children showed no history of speech or language problems and none of
their family members suffered from any learning or language problem. For every individual
HR child we selected the best matching LR control child based on five criteria: (1)
educational environment, i.e. same school, (2) gender, (3) age, (4) nonverbal intelligence, and
(5) parental educational level. Nonverbal intelligence was assessed by an adapted version of
the Raven Coloured Progressive Matrices (RCPM) (Raven, Court, & Raven, 1984), a
collective non-verbal intelligence test measuring spatial reasoning. Parental educational level
was assessed using the ISCED-scale (International Standard Classification of Education by
UNESCO, 1997), by converting classifications on the original seven-point scale to a three-
point scale. At the time of collecting the first kindergarten data, the mean age for both the HR
and LR group was 5 years and 4 months, not being statistically different. The nonverbal IQ
scores were slightly above population average (107 for HR group and 111 for LR group) and
did not differ significantly. Both risk groups represented children from relatively higher
educated parents, with Fisher’s Exact Test confirming that both groups did not differ in
frequency distribution of the different parental educational categories. Further details about
the participants and the selection procedure are described in Boets et al. (2006a).
Tasks and materials
Measures at preschool age
Phonological tests
Tests were selected to reflect the three traditional domains of phonological skills
(Wagner & Torgesen, 1987). Phonological awareness was measured by three sound identity
tasks and a rhyme task. Verbal short-term memory was measured by a digit span test and a
nonword repetition test. Rapid automatic naming was assessed by administering a colour and
an object rapid naming task. Here we will only present a brief description of the tests; further
details can be found in Boets et al. (2006a).
Sound identity tasks. The child was required to choose from four alternatives the word
that had the same (a) first sound, (b) end sound or (c) end rhyme as a given word (de Jong,
Seveke, & van Veen, 2000, adapted by van Otterloo & Regtvoort). The distracter alternatives
were systematically constructed to prevent guessing. Each item consisted of a row of five
114
pictures. The first picture represented the given word and was separated from the other
pictures by a vertical line. All items were named for the child. The first-sound and end-sound
identity tasks both consisted of 10 items, preceded by two practice items, and had a maximum
score of 10. The rhyme identity task consisted of 12 items, preceded by two practice items,
and had a maximum score of 12.
Rhyme test. Participants were presented a one-syllable word and were required to
produce a rhyming (non-)word. The test consisted of 8 items, gradually becoming more
difficult, and was preceded by two practice items.
Non-word repetition test. Upon hearing a non-word, presented from a CD, the child
had to repeat it as accurately as possible. The test consisted of 48 non-words, varying in word
length from two to five syllables, was preceded by two practice items and had a maximum
score of 48 (Scheltinga, 2003).
Digit span forward. The test assessed the immediate serial recall of spoken lists of
digits presented from a CD. Testing started with a sequence of two digits and continued with
increasing list length until the child failed on two of three trials of the same length. The test
score was the total number of correctly recalled lists (De Smedt, Verschaffel & Ghesquière,
2004).
Rapid automatic naming: pictures. The child was presented a card depicting 50
pictures of five familiar objects in random order and was instructed to name them as fast and
accurately as possible (van den Bos, Zijlstra, & Spelberg, 2002). Total naming time was
recorded irrespective of accuracy. Subsequently, the time to completion was transformed to
the number of symbols named per second. As such, a higher score on the test corresponded to
a higher naming speed.
Rapid automatic naming: colours. This test was identical to the automatic picture-
naming test, but now a card depicting 50 small rectangles in five familiar colours was
presented to the child.
Productive letter knowledge
This task was intended as a preliminary measure of literacy development. The sixteen
most frequently used letters in Dutch books were presented on a card and the child had to
name each of these letters. Both the sound and the name of a letter were considered correct.
115
Tests for low-level auditory processing
Auditory processing was assessed by means of three psychophysical threshold tests
(see Boets et al., 2006a). In the GAP-detection test, subjects had to detect a silent interval
(gap) in a white noise stimulus. Threshold was defined as the minimum gap length required
for detecting the silent interval. In the FM-detection test participants had to detect a 2 Hz
sinusoidal frequency modulation of a 1 kHz carrier tone with varying modulation depth.
Threshold was defined as the minimum depth of frequency deviation required to detect the
modulation. In the TN-detection task participants had to detect two pure tone pulses (1 kHz,
length = 440 ms) within a one-octave noise signal, centered around 1 kHz (from 707 to 1414
Hz, length = 1620 ms). Threshold was defined as the lowest signal-to-noise ratio (SNR)
required for detecting the tone pulses. For all three auditory tests, thresholds were estimated
using a three-interval forced-choice oddity paradigm embedded within an interactive
computer game with animation movies (Laneau, Boets, Moonen, van Wieringen & Wouters,
2005). The length of the gap, the depth of modulation and the amplitude of the sinusoidal
pulses were adjusted adaptively using a two-down, one-up rule, which targeted the threshold
corresponding to 70.7 % correct responses (Levitt, 1971). A threshold run was terminated
after eight reversals and the threshold for an individual run was calculated by the geometric
mean of the values of the last four reversals. After a short period of practice, comprising
supra-threshold trials to familiarize the participants with the stimuli and the task, three
threshold estimates were determined for every experiment. For the data we present here, the
average of the best and second best threshold was used as an indicator of auditory sensitivity.
For all auditory testing – also for the speech perception tests – the same calibrated
equipment was used. Stimuli were saved as 16-bit wav-files on the hard disc of a Dell
Latitude C800 and Toshiba Satellite 1400-103 portable computer. They were presented using
an integrated audio PC-card and routed to an audiometer (Madsen OB622) in order to control
the level of presentation. Stimuli were presented monaurally over calibrated TDH-39
headphones. A more detailed description of the stimuli, procedure and equipment can be
found in Boets et al. (2006a).
Visual magnocellular test
For the CM-detection test, children were sitting in a low-luminance (mesopic)
environment at 40 cm distance from an Elo Intuitive 1725L 17’’ touch screen (75 Hz refresh
116
rate) on which the random dot kinematograms (RDK) were displayed. The display resolution
was set to 640 x 480 pixels. The stimuli were generated online by the same portable
computers as used for the auditory experiments and comprised of two rectangular patches,
each containing 1103 randomly moving high luminance white dots on a black background
(dot size = 1 pixel or 0.07° diameter, dot density = 2.5 dots/deg2, velocity = 7.3 deg/sec, life
time = 5 video frames or 200 msec, maximal duration of stimulus presentation = 6 sec,
luminance of dots = 125 cd/m2, luminance of background = 0.39 cd/m2, Michelson
contrast = 99.4 %). At a viewing distance of 40 cm each patch of dots subtended 16 x 27.2°
visual angle, separated horizontally by 3.8°. The target patch was segregated into three
horizontal strips; in the middle strip a variable proportion of dots were moving coherently in
horizontal direction, reversing direction every 330 msec. All other dots were moving
randomly in a Brownian manner. The two patches were presented simultaneously and the
subject had to identify the patch containing the strip with coherently moving dots. Threshold
was defined as the smallest proportion of coherently moving dots required for detection of the
middle strip with reversing dot motion. Similarly to the auditory experiments a two-down,
one-up adaptive staircase procedure was used and threshold was calculated as the geometric
mean of the values of the last four of eight reversals. After a short period of practice, four
thresholds were determined for every subject. For the data we present here, the average of the
best and second best threshold was used as an indicator of visual magnocellular sensitivity.
As was the case with the auditory experiments, the CM-detection experiment was integrated
within a computer game with animation movies and an extensive reinforcement system to
make it attractive and applicable for very young children (see Boets et al., 2006b).
Tests for speech perception
Speech-in-noise perception task. In the speech-in-noise perception task seven lists of
ten high frequent monosyllables (taken from the Göttingerlist II, see van Wieringen &
Wouters, 2005) were presented monaurally with an inter-stimulus interval of 7 seconds.
Simultaneously a continuous stationary speech noise, with an identical spectrum as the
average spectrum of the word lists, was presented to the same ear, at a fixed level of 70 dB
SPL. Words were presented at -1, -4 and -7 dB signal-to-noise ratio (SNR). Before
administration of the six test lists (3 x 2), one list was presented at an SNR of +4 dB as a
practice list. The child’s task was to repeat the words as accurately as possible, resulting in a
percentage correct word score for every test list.
117
Categorical perception. A ten-point speech continuum ranging from /bAk/ to /dAk/
(i.e. the Dutch word for ‘box / basket’ and ‘roof’) was presented by means of a categorical
perception task. For every child, a 2AFC forced-choice identification task was administered
followed by a same-different (AX) discrimination task. Stimuli were identical to the ones
developed and described by van Beinum et al. (2005) and were constructed by linearly
interpolating the transition of the second formant (F2) from /b/ to /d/. Stimuli were presented
monaurally at a comfortable listening level of 70 dB SPL. Similarly as in the other sensory
experiments, the task was again integrated within a videogame.
In the identification task, the ten stimuli of the continuum were presented twelve times
in a random order. The child was instructed to repeat the perceived word and point to one of
two pictures on a screen, depicting the possible response options. The task was preceded by a
pre-test, presenting each of the two endpoint stimuli five times.
In the discrimination task, subjects were instructed to listen to two stimuli presented
with an ISI of 600 ms and to determine whether they sounded ‘the same’ or ‘different’.
Children had to respond by pointing to one of two pictures on the screen representing ‘the
same’ or ‘different’. In order to obtain a bias-free measure of discriminability, the task
contained physically different as well as identical pairs. The different pairs, comprising each
of the seven 3-step comparisons (1-4, 2-5, 3-6, etc.), were presented 6 times (with balanced
internal order); the identical pairs (1-1, 2-2, etc.) were presented twice. The task was preceded
by 10 practice items, comprising pairs that were clearly different or identical. Further details
regarding the speech tests can be found in Boets et al. (under revision).
Literacy measures at the end of first grade
To assess literacy skills, a standardised spelling test (Dudal, 1997) and six
standardised reading tests were administered at the end of Grade 1 (after receiving one year of
formal reading instruction). The One-Minute Real-Word Reading test (Brus & Voeten, 1973)
and the Pseudo-Word Reading test (van den Bos, Spelberg, Scheepstra, de Vries, 1994)
measure respectively real-word reading and non-word reading, but they do not differentiate
between reading accuracy and reading speed. To discern between these aspects of reading, we
constructed and administered four additional reading tests based on the description provided
by de Jong and Wolters (2002): the Real-Word Reading Accuracy test, the Real-Word
Reading Speed test, the Non-Word Reading Accuracy test and the Non-Word Reading Speed
118
test2. To make these tests applicable for the diagnostic process, age norms were collected in a
large-scale pilot study.
Data collection
Data collection was carried out by qualified psychologists and audiologists. All tests
were administered individually during several sessions adding up to approximately 11 hours
of testing for every subject. After every subtest children were rewarded with little gadgets.
Testing took place in a quiet room at the children’s school. Since the LR child was selected
from the HR child’s classmates, both children could always be tested in exactly the same
circumstances. Phonological and auditory data, and speech perception and CM data were
collected during respectively the first and second trimester of kindergarten. Literacy measures
were collected in the last month of grade 1.
Statistical analysis
Prior to analysis, all data were individually checked for unexpected outliers. As a
consequence, for one subject two unreliable phonological test scores were removed and for
another subject the speech-in-noise perception data were discarded. Further, the speech-in-
noise data and the identification data were submitted to a logistic fitting to estimate the
essential parameters. In order to take into account the variable quality of these fits, the inverse
standard error of the estimated parameter was added to the model as a weight variable. To
obtain a normal distribution, the thresholds for GAP, FM and CM, the results on the letter
knowledge task and the fitted slope parameter of the speech identification data were log10-
transformed. Generally, all data were analyzed using Mixed Model Analysis (MMA) (Littell,
Stroup & Freund, 2002), taking into account the clustered nature of the data (i.e. matched
pairs attending the same school). Although the original HR and LR group as well as the
reading defined groups (see supra) did not differ for age, nonverbal intelligence or parental
educational level, these variables were additionally controlled for in our analyses. Concretely,
a series of (repeated) MMA’s was calculated with school (= pair number) as a random
variable and participant group as the fixed between-subject variable3. Age, nonverbal IQ and
educational level of both mother and father were added as fixed (co)variables. All post-hoc
analyses were corrected for multiple comparisons using the Tukey procedure (α = .05).
119
Results
Defining literacy groups at the end of Grade 1
Descriptive statistics summarising the performance of the HR and LR group on the
reading and spelling measures are shown in Table 1. A paired wise MMA incorporating the
same covariates as mentioned above revealed a significant group difference for every single
literacy measure (p < .01).
Table 1
Mean performance (and standard deviations) on literacy achievement tests at the end of Grade 1 for the High-
Risk and Low-Risk group.
HR
(n = 31)
LR
(n = 31)
M SD M SD
Spelling
One-Minute Real-Word Reading
Pseudo-Word Reading
Real-Word Reading Accuracy
Real-Word Reading Speed
Non-Word Reading Accuracy
Non-Word Reading Speed
46
16
17
22
60
14
40
13
8
9
13
33
11
23
53
22
22
28
78
21
54
5
9
9
10
32
11
26
The contemporary definition of dyslexia for the Dutch language area (Committee on
Dyslexia of the Health Council of the Netherlands, see Gersons-Wolfensberger &
Ruijssenaars, 1997) emphasises that the diagnosis does not only depend upon the observation
of severe reading and spelling problems (< Pc 10), but also requires these problems to be
persistent and resistant to the usual teaching methods and remedial efforts. However, after one
year of formal reading instruction, this additional criterion of persistence cannot yet be
verified. As a consequence, we will refer to the children currently demonstrating impaired
reading and spelling ability as ‘literacy delayed’ instead of ‘dyslexic’. To determine the
reading groups, a LITERACY composite score was calculated averaging the standardised
results (relative to population data) of all reading and spelling tests. In line with the <Pc10
criterion postulated by the Health Council of the Netherlands, a cut-off point of –1.3 SD ’s on
the LITERACY composite was taken as a criterion to delineate the literacy-delayed group.
Applying this criterion resulted in 3 literacy-delayed subjects in the LR group (3/31 = 9 %),
120
and 9 in the HR group (9/31 = 29 %), indicating a relatively increased risk of 3.0 for the
genetically at risk subjects [χ2(1) = 3.7, p < .05]. Hence, on the basis of familial risk status and
current literacy achievement, we defined four groups: a HR normal literacy (HR-LN) and a
HR literacy-delayed (HR-LD) group, and a LR normal literacy (LR-LN) and a LR literacy-
delayed (LR-LD) group. In the further analyses we will compare the results of the HR-LD,
HR-LN and LR-LN groups. Data of the LR-LD participants were excluded because the size of
the group (n = 3) was too small to provide meaningful comparisons and because we wanted to
restrict our clinical group to children showing both (1) the familial risk for dyslexia and (2)
already presenting a significant delay in reading and spelling development.
Table 2 provides descriptive statistics for the three groups, indicating that they did not
differ significantly regarding age, nonverbal intelligence and parental educational level4.
Since the groups were defined based on literacy achievement, it is evident that for all reading
and spelling measures the HR-LD group differed significantly (p < .0001) from the two other
groups (HR-LN and LR-LN), who themselves did not differ from each other.
Table 2
Characteristics of the participants.
HR-LD
(n = 9)
HR-LN
(n = 22)
LR-LN
(n = 28)
M SD M SD M SD
Age in months
Non-verbal IQ
Maternal educational level
Paternal educational level
63
108
2.4
2.4
3
7
0.7
0.5
64
106
2.6
2.1
3
16
0.7
0.8
64
111
2.6
2.4
3
14
0.6
0.6
Note. ANOVA revealed that there were no significant group differences for any of the subject characteristics
(Tukey contrasts p < .05). Parental educational level was calculated from ordinal data.
Group comparisons for preschool measures
Table 3 shows the performance of the three groups of children on the phonological
tests and the letter knowledge task. For every test, except the first-sound identity task and the
digit span task, the HR-LD group scored significantly below the LR-LN group. Furthermore,
it is interesting to note that the HR-LN group, in spite of a hitherto normal reading and
spelling development, scored in between both other groups. For the end-sound identity task
121
and the non-word repetition test, the difference between the HR-LN and the LR-LN group
was even significant, suggesting that familial risk is continuous rather than discrete.
Table 3
Mean performance (and standard deviations) on preschool letter knowledge, phonological and sensory tests for
the High Risk Literacy-Delayed, High Risk Literacy-Normal and Low-Risk Literacy-Normal groups.
HR-LD HR-LN LR-LN
M SD M SD M SD
Letter Knowledge
Phonological Awareness
Simple rhyme
Rhyme identity
First-sound identity
End-sound identity
Rapid Automatic Naming
Colour naming
Picture naming
Verbal Short-Term Memory
Digit span
Nonword repetition test
Composite AWARENESS
Composite RAN
Composite VSTM
Composite PHONOLOGY
AV GAP (ms)
AV FM (Hz)
AV TN (dB SNR)
AV CM (% coherence)
SRT (dB)
Phoneme Boundary
Slope of Phoneme Boundary
Mean discriminability (-ln eta)
0.3a
5.1a
8.0a
3.9
4.0a
0.58 a
0.58 a
6.4
16.1a
-1.49a
-0.87a
-0.62
-1.65a
7.4
11.1a
-6.2
0.28a
-2.8a
5.0
17.5
0.15
0.5
2.8
3.2
1.8
1.4
0.06
0.09
1.5
3.9
1.5
0.4
0.6
1.3
4.3
5.8
2.1
0.18
1.7
4.7
7.2
0.27
2.6ab
6.6ab
9.0ab
4.9
4.7a
0.65ab
0.65ab
7.3
17.1a
-0.75ab
-0.49ab
-0.31
-0.85a
5.1
7.0b
-8.0
0.17b
-3.6b
3.8
16.7
0.44
3.8
2.2
2.4
2.2
2.4
0.13
0.13
1.6
5.8
1.3
0.9
0.9
1.2
4.3
3.4
2.2
0.06
1.6
3.5
5.2
0.61
3.3b
7.3b
10.1b
5.9
6.3b
0.71b
0.71b
6.9
21.3b
0.00b
0.00b
0.00
0.00b
5.1
6.9b
-7.8
0.18b
-3.9b
4.3
21.7
0.74
3.5
1.8
1.5
2.3
2.3
0.16
0.16
1.6
6.5
1.0
1.0
1.0
1.0
3.8
3.8
1.5
0.09
1.2
4.1
5.4
0.63
Note. Pairs with different subscript letters differ significantly (MMA controlled for non-verbal IQ, age, parental
educational level and school environment; Tukey contrasts p < .05). For SRT, Phoneme Boundary and Slope of
Phoneme Boundary weighted group means and SD ’s are presented.
122
A principal component factor analysis with varimax rotation confirmed that the test
battery excellently reflected the traditional three-fold phonological structure: (a) the
phonological awareness factor had high loadings of the three sound identity tasks and the
simple rhyme task, (b) the rapid automatic naming factor had high loadings of both the colour
and object rapid naming tasks and (c) the verbal short-term memory factor was completely
determined by high loadings of the non-word repetition and the digit span task (for details, see
Boets et al, 2006a). Consequently, for every phonological factor summarizing composite
scores were calculated averaging the standardised scores of their constituent tests
(AWARENESS, RAN, VSTM). In addition, a general phonological composite averaging the
standardised scores of all phonological tests was also calculated (PHONOLOGY). Statistics
for these four phonological composites are displayed in table 3. To assist in the interpretation
of the results, composite values were transformed to effect sizes relatively to the mean and
standard deviation of the LR-LN group. As can be seen, the HR-LD group scored
significantly lower than the LR-LN group on AWARENESS, RAN and PHONOLOGY, with
the HR-LN group again scoring in between. For VSTM group differences did not reach
significance.
Table 3 shows the results on the auditory measures and the visual CM-detection task.
For FM and CM detection the HR-LD group scored significantly worse than the two other
groups, who themselves did not differ from each other. Also for GAP and TN detection the
thresholds of the HR-LD group were increased, but this difference did not reach significance.
Average results of the speech-in-noise perception test are depicted in Fig. 1. A
Repeated Measures MMA with proportion correctly perceived words as dependent variable,
group as between-subject variable, SNR as within-subject variable and with the same
covariates as mentioned above, revealed a significant main effect for group (p = .004) and
SNR (p < .0001) with the group x SNR interaction being insignificant (p = .50). Post-hoc
analysis revealed that the HR-LD group differed significantly from both other groups, who
themselves did not differ from each other. Additionally, to estimate the Speech Reception
Threshold (SRT: the signal level required for 50 % correct responses), for every subject a
logistic function was fitted to the data (for details, see Boets et al., under revision). In order to
take into account the variable quality of the fits, the inverse standard error of the estimated
parameters was added to the model as a weight variable. Table 3 shows weighted group
means for SRT. As was the case with the Repeated Measures analysis, MMA with the weight
123
variable and the same covariates as mentioned above demonstrated that the HR-LD group
required a significantly easier signal-to-noise ratio than the two other groups to perceive 50 %
of the presented words correctly.
0
0.2
0.4
0.6
0.8
1
-1 dB -4 dB -7 dB
Signal-to-noise ratio (SNR)
Pro
porti
on c
orre
ct re
spon
ses HR-LD
HR-LN
LR-LN
Fig. 1: Mean scores relating the proportion correctly perceived words to the relative level of the presented words
(SNR).
For the categorical perception data, only subjects obtaining 70 % correct responses or
more on the pre-test were included in the analyses, excluding 4 subjects for the identification
task and 14 subjects for the discrimination task. A justification of this relatively liberal
criterion can be found in Boets et al. (under revision). For the identification data (see Fig. 2), a
Repeated Measures MMA with ‘proportion /dAk/ responses’ as dependent variable, group as
between-subject variable, stimulus (1 to 10) as within-subject variable and with the same
covariates as mentioned above, revealed no significant effect for group (p = .68), a significant
effect for stimulus (p < .0001) and no significant group x stimulus interaction effect (p = .71).
Again, data were submitted to a logistic fitting to estimate the phoneme boundary (i.e. the
interpolated 50 % cross-over point) and the slope in this point. The slope of the fitted function
is the most relevant parameter: it represents the range of uncertainty in distinguishing one
phoneme category from another and is thus indicative of the ability to categorise a speech
continuum in a consistent way. Table 3 shows weighted group means for the estimated
phoneme boundary and slope5. MMA with a similar weight variable and the same covariates
as mentioned above demonstrated no significant differences between groups.
124
0
0.2
0.4
0.6
0.8
1
1 2 3 4 5 6 7 8 9 10
Stimulus number
Pro
porti
on /d
Ak/
resp
onse
s
HR-LD
HR-LN
LR-LN
Fig. 2: Mean identification functions relating the proportion /dAk/ responses to the stimulus number of the place-
of-articulation continuum.
-0.8-0.6-0.4-0.2
00.20.40.60.8
11.21.41.6
1-4 2-5 3-6 4-7 5-8 6-9 7-10
Stimulus pair
Dis
crim
inab
ility
HR-LD
HR-LN
LR-LN
Fig. 3: Mean discrimination scores (- ln eta) as a function of stimulus pair.
For the discrimination data (see Fig. 3), a distribution and response-bias free index of
dicriminability (- ln eta)6 was calculated according to signal detection theory (Wood, 1976).
Although Fig. 3 clearly demonstrates that the HR-LD group presented a reduced
discriminability (in particular for stimulus pair 3-6), this effect was not significant in a
Repeated Measures MMA with covariates incorporated [group effect: p = .09; stimulus effect:
p = .01; group x stimulus interaction: p = .17]. However, repeating the analysis without any
fixed covariates added, did result in a significant group effect (p = .02), with the HR-LD
group differing significantly from the LR-LN group. The same pattern of results was obtained
125
for the mean discriminability score calculated over all pairs (see Table 3). With the covariates
added the group difference was not significant, but without these covariates the HR-LD group
differed significantly from LR-LN.
Individual deviance analysis
Since one of the goals of this study was to explore early indicators of dyslexia and in
view of the fact that group comparisons might mask significant individual differences, we
also carried out analyses at the subject level. To decide which individual did and did not show
abnormal performance, we adopted the two-step criterion as suggested by Ramus et al.
(2003). Applying this procedure, the criterion for deviance was placed at 1.65 standard
deviations of the ‘purified’ mean of the LR-LN group, after first having excluded all deviant
LR-LN subjects (by applying a similar 1.65 SD criterion, typically resulting in the removal of
one or two deviant LR-LN subjects for each measure). A distribution analysis on the data of
the ‘restricted’ LR-LN group confirmed their normality, indicating that the 1.65 SD criterion
corresponds to the fifth percentile. Individual scores for the sensory measures and for the
PHONOLOGY and LITERACY composite are plotted in Fig. 4, with Table 4 presenting the
number of deviant subjects in each group (see Appendix for an overview of all individual
deviancy data). Inspection of the individual data reveals three main findings. First, for the
sensory measures and in particular for PHONOLOGY, it is evident that the HR-LD group
showed an increased proportion of deviant subjects. Second, it is clear that not all HR-LD
subjects scored in the deviant range. Whereas for PHONOLOGY this applied for about ¾ of
the group, for the sensory measures this typically applied for about one third or the half of the
group. Third, inspection of the scores of the children of the HR-LD group reveals no
straightforward pattern relating deficiencies across different processing skills. Although there
might be a tendency for children with auditory deficiencies to present also more severe
phonological deficits, this pattern does not extent to speech perception. Further, it is striking
that exactly the two HR-LD subjects without phonological deficiencies were among the ones
presenting the most severe visual problems. However, contrary to the expectation, they were
also among the ones showing the most severe speech perception problems. Thus, in sum, we
have to conclude that we are not able to demonstrate a consistent pattern of deficits across
auditory, visual, speech perception and phonological processing abilities for the subjects of
the HR-LD group.
126
Fig. 4: Individual Z-scores for the LITERACY and PHONOLOGY composite and for all sensory measures.
Except for LITERACY, the solid line indicates the mean for all LR-LN subjects above Pc 5; the dashed line
indicates the chosen deviance criterion (1.65 SD deviating of the ‘restricted’ LR-LN mean). Deviant individuals
are identified. For LITERACY, individuals have been depicted relatively against population average with the
dashed line indicating the -1.3 SD criterion.
Note. For LITERACY, PHONOLOGY and the Phoneme Boundary Slope a higher Z-score indicates better
processing; in contrast for all other measures a higher Z-score indicates reduced sensory sensitivity.
127
Table 4
Proportion of deviant subjects for each reading group.
HR-LD
(n = 9)
HR-LN
(n = 22)
LR-LN
(n = 28)
PHONOLOGY
GAP-detection
FM-detection
TN-detection
CM-detection
SRT
Slope of Phoneme Boundarya
78 %
33 %
44 %
11 %
33 %
33 %
55 %
45 %
23 %
14 %
9 %
5 %
5 %
36 %
14 %
14 %
7 %
4 %
11 %
4 %
28 %
Note. aFor the Slope of Phoneme Boundary both the subjects scoring below 1.65 SD of the restricted LR-LN
mean, as well as the excluded subjects for whom no reliable identification data could be obtained were
considered as being deviant.
Discussion
In this study we investigated low-level auditory and visual processing, speech
perception and phonological ability in 5-year-old preschool subjects who did not yet receive
any formal reading instruction. Data were retrospectively analysed, comparing three groups of
subjects defined by first grade literacy achievement and family risk for dyslexia. Since it is
questionable to firmly assess dyslexia in subjects that have received only one year of reading
instruction, we preferred to refer to the impaired group as literacy-delayed instead of dyslexic.
However, evidence for the validity of defining reading groups at the end of first grade has
been provided by research demonstrating that differences among children in reading and
spelling (dis)abilities are quite stable over time, and that the majority of those identified as
having reading difficulties in first grade continue to read poorly throughout their school years
and beyond (McCardle, Scarborough & Catts, 2001). Moreover, our clinical group was
defined by showing both the actual literacy delay and the increased family risk.
Comparing the HR-LD group with the LR-LN group, we observed a significant
difference to the advantage of the LR-LN group for letter knowledge, phonological ability
(i.e. phonological awareness and rapid automatic naming; for verbal short-term memory the
difference was only significant on the non-word repetition test), FM- and CM-detection, and
128
speech-in-noise perception. Without taking into account any covariates, the results on the
categorical perception task also differentiated between both groups. With respect to the
phonological data, these results are consistent with the phonological deficit hypothesis and are
in line with other prospective longitudinal studies revealing similar deficits in genetically at
risk children (for a review see the impressive meta-analysis of Scarborough, 1998, based on
61 studies, indicating that letter knowledge and phonological awareness are the best single
preschool predictors of literacy development). Regarding the sensory data, the significant
group difference for FM- and CM-detection is consistent with a whole series of
psychophysical studies typically demonstrating that these measures differentiate reliably
between adult and school-aged dyslexic and normal reading subjects (for FM see e.g. Stein &
McAnally, 1995; Van Ingelghem et al., 2005; Witton et al., 1998; for CM see e.g.
Cornelissen, Richardson, Mason, Fowler & Stein, 1995; Everatt, Bradshaw & Hibbard, 1999;
Hansen, Stein, Orde, Winter & Talcott, 2001; Raymond & Sorensen, 1998; Ridder, Borsting
& Banton, 2001; Talcott, Hanssen, Assoku & Stein, 2000; Talcott et al., 2003; Van
Ingelghem, Boets, van Wieringen, Ghesquière & Wouters, 2004; Wilmer, Richardson, Chen
& Stein, 2004; Witton et al., 1998). Interestingly, some authors have postulated that both
tasks depend upon ‘dynamic’ stimulus processing (i.e. requiring the perception of a dimension
changing in time), suggesting that the sensory deficit might be specific for dynamic
processing (Talcott et al., 2000; Talcott & Witton, 2002). With respect to speech perception,
again our results are in line with a whole range of studies demonstrating a deficit in speech-in-
noise perception (e.g. Bradlow, Kraus & Hayes, 2003; Brady, Shankweiler & Mann, 1983)
and categorical perception (e.g. Breier, Fletcher, Denton & Gray, 2004; Godfrey, Syrdal-
Lasky, Millay & Knox, 1981; Lieberman, Meskill, Chatillon & Schupack, 1985; Maassen et
al., 2001; Manis et al., 1997; Reed, 1989; Steffens, Eilers, Gross-Glenn & Jallad, 1992;
Serniclaes et al., 2001; Werker & Tees, 1987) in adult and school-aged dyslexic subjects. In
particular, our data also corroborate two studies providing rare evidence of speech perception
problems in toddlers (Gerrits, 2003) and six-month-old infants (Richardson, Leppänen, Leiwo
& Lyytinen, 2003) at family risk for dyslexia. Besides the reduced speech discriminability,
particularly for speech contrasts near the phoneme boundary, we would like to emphasize the
remarkable speech-in-noise perception deficit in the HR-LD group. Regardless of the actual
signal-to-noise ratio, children of this group perceived on average about 10 % fewer words
than their peers. Applied to a regular classroom situation, typically characterized by a lot of
background noise, this implicates that a considerable amount of communication and
instruction is missed by these children. Undoubtedly, this might have far-reaching
129
implications for scholar development in general and language and phonological development
in particular.
For the two other auditory tasks, GAP- and TN-detection, the thresholds of the HR-LD
subjects were increased but did not differ significantly from the HR-LN group. For TN-
detection this lack of a difference was expected according to the auditory temporal deficit
hypothesis since TN-detection was merely taken up as a non-temporal control task. For GAP-
detection in contrast, both the theory and previous empirical evidence (e.g. Hautus, Setchell,
Waldie & Kirk, 2003; McCroskey & Kidder, 1980; Van Ingelghem et al., 2001, 2004, 2005)
predicted that the dyslexia-prone HR-LD group should show a significant deficit.
Nevertheless, our failure replicating a significant deficit in GAP-detection does not stand
alone: several other studies also failed to demonstrate a GAP-detection deficit in dyslexia
suggesting the task might not be robust enough to differentiate reliably between dyslexic and
normal reading subjects (Adlard & Hazan, 1998; McAnally & Stein, 1996; Schulte-Körne,
Deimel, Bartling & Remschmidt, 1998). In line with these findings, Phillips and colleagues
proposed that a between-channel variant of the GAP-detection task (where the pre and post
gap markers occupy different critical bands) might be a more appropriate measure to probe
the perceptual mechanisms involved in stop consonant speech discrimination (Phillips,
Taylor, Hall, Carr & Mossop, 1997).
Focussing upon the results of the HR-LN group reveals an interesting pattern. For
phonological ability their scores were situated either in between both other groups or even at
the level of the HR-LD group. This means that in spite of a hitherto normal literacy
development, they presented a mild phonological deficit, suggesting that familial risk is
continuous rather than discrete (for a similar conclusion see Pennington & Lefly, 2001;
Snowling, Gallagher & Frith, 2003). In contrast, for the sensory measures, the scores of the
HR-LN group were almost identical to those obtained by children of the LR-LN group. This
suggests that whereas the occurrence of a phonological problem seems to be largely
genetically determined (or at least by familial risk status) and partly irrespective of actual
literacy achievement, the presence of a sensory problem rather seems to depend on the co-
occurrence of a real literacy deficit. This conclusion is well in line with data of Bishop and
colleagues (Bishop et al., 1999) and Olson and Datta (2002) who demonstrated in twin studies
that in contrast to the highly heritable phonological skills, sensory skills depend less on
genetic and more on environmental influences. Within the framework of an integrative
neurobiological theory of dyslexia, Ramus (2004) recently suggested that these co-occurring
sensori(motor) problems might be the consequence of elevated levels of foetal testosterone
130
(that are mostly influenced by non-genetic factors). As a consequence, his model predicts an
increased prevalence of the sensorimotor syndrome in male dyslexics. Although our final
sample of literacy-delayed subjects is too small (6 boys / 3 girls) to make any firm
conclusions, our data are at least consistent with this hypothesis: 6 of the 6 male subjects (100
%) present at least one deviant score for auditory, visual or speech perception processing
versus 2 out of 3 female subjects (67 %).
The observation of reduced sensory processing by subjects of the HR-LD group
(evidenced both by an increased proportion of deviant subjects and a significant group
difference), makes it plausible that this sensory deficit is somehow related to their problems in
literacy achievement. However, the concurrent findings that (1) not all literacy-delayed
subjects are affected by a sensory problem, whereas (2) some normal reading subjects do also
present sensory problems, suggest that sensory skills might only play a limited role in literacy
development. Noteworthy however, the same pattern of results is also observed for
phonological ability, indicating that even phonology is not able to explain the whole story of
dyslexia.
It should be clear that the observation of a sensory deficit in preschool children, who
ultimately will develop dyslexia, does not imply that this deficit is causally related to the
reading problem. One should be cautious interpreting these data in a causal way. But on the
other hand, this study convincingly demonstrates that the sensory problems typically observed
in dyslexia are not merely the result of lacking reading experience. On the contrary, the
sensory deficit evidently does precede the literacy problem. Hence, this leaves open the
possibility of a potential causal influence of sensory processing upon literacy development in
a way as postulated by the general magnocellular theory. In the next study we will further
explore the plausibility of these interrelations using structural equation modeling.
STUDY 2: Modeling relations between sensory processing, speech
perception, orthographic and phonological ability and literacy achievement
The general magnocellular theory does not only assume the presence of an underlying
sensory deficit in subjects with dyslexia, but also proposes a specific (causal) relation between
sensory sensitivity, speech perception, orthographic and phonological ability and literacy
131
achievement. Although many studies have mentioned correlations between (a) auditory
processing and phonological or reading abilities (e.g. Amitay, Ahissar & Nelken, 2002;
Ramus et al., 2003; Richardson, Thomson, Scott & Goswami, 2004; Talcott et al., 1999; Van
Ingelghem et al., 2005; Witton et al., 1998), (b) between speech perception and phonological
or reading abilities (e.g. Bradlow et al., 2003; Breier et al., 2004; De Weirdt, 1988; Godfrey et
al., 1981; Maassen et al., 2001; Mayo, Scobbie, Hewlett & Waters, 2003; Nittrouer, 1996;
Ramus et al., 2003) and (c) between visual magnocellular processing and orthographic or
reading skills (e.g. Cornelissen et al., 1998; Demb, Boynton & Heeger, 1997; Van Ingelghem
et al., 2004), only very few have systematically investigated the relations between these
variables. A noteworthy exception is a series of studies by Talcott, Witton and colleagues
demonstrating an exclusive relation between visual magnocellular processing and
orthographic skills on the one hand and auditory processing and phonological ability on the
other (Talcott, Hansen, et al., 2000; Talcott, Witton, et al., 2000; Talcott & Witton, 2002).
Further exceptions are three studies illuminating interrelations using structural equation
modeling. McBride-Chang (1996) investigated the relations of speech perception and
phonological processing to reading, and concluded that the best fitting model was one in
which the effect of speech perception on reading was mediated by its relation with
phonological awareness and rapid automatic naming. Two other studies (Schulte-Körne,
Deimel, Bartling & Remschmidt, 1999; Watson & Miller, 1993) additionally included non-
linguistic measures for auditory temporal processing into the model. Although both confirmed
that the influence of speech perception on reading and spelling was mediated by phonological
processing, neither of them were able to obtain a good fit with the low-level auditory
variables integrated into the model.
In this study we will evaluate the postulated relations between sensory processing,
speech perception, orthographic and phonological ability and literacy development using
structural equation modeling (path analysis). Structural equation modeling is a statistical
method to model the (causal) interrelations between a group of variables, and subsequently
test the plausibility of the theoretical model compared to the observed data (Kline, 1998). In
line with the assumptions of the general magnocellular theory and the traditional phonological
theory we constructed and tested the structural model depicted in Fig. 5. In this model
dynamic auditory processing is postulated to determine speech perception, which on its turn
determines phonological ability. Although we mainly expect an influence of speech
perception upon phonological awareness, in line with the theoretical arguments and empirical
data provided by McBride-Chang (1996) we will also allow rapid automatic naming and
132
verbal short-term memory to be determined by speech perception. Further on, all three
phonological sub skills are hypothesised to act upon literacy development. In the same way,
dynamic visual (magnocellular) processing is postulated to determine orthographic ability,
which on its turn also determines reading and writing development. In accordance with the
established phonological theory (Wagner & Torgesen, 1987), we allow the phonological sub
skills to be mutually correlated. Likewise, the measures for auditory and visual sensitivity are
also allowed to correlate since they both depend upon dynamic sensory processing (Talcott &
Witton, 2002).
Fig. 5: Figure showing the basic structural model tested in the path analysis.
Method
Participants
The participants were the same as those involved in study 1.
133
Tasks and materials
Preschool measures. Largely the same data were used as in study 1. However, in order
to manage the complexity of the estimated models, we only included preschool sensory
measures that demonstrated significant differentiating power. This means that only the data
for FM detection, CM detection and speech-in-noise perception were retained in the model,
together with letter knowledge and data representing the three phonological sub skills. The
restriction of the low-level sensory tests to FM and CM detection also offered the additional
advantage of being in line with the current research tradition regarding ‘dynamic’ sensory
processing (e.g. Witton et al., 1998). As such, it offered the possibility to replicate some
recent findings in adults and school-aged children demonstrating that orthographic skills co-
vary most strongly with CM sensitivity, whereas phonological skills co-vary most strongly
with FM sensitivity (Talcott, Hansen, et al., 2000; Talcott, Witton, et al., 2000; Talcott &
Witton, 2002). Finally, the exclusion of the categorical perception data was also partially
motivated by the large number of missing or unreliable test scores.
First grade literacy measures. To assess reading skills, data of the same six
standardised reading tests described in study 1 were used. To assess writing skills, we
administered a pseudohomophone task in addition to the standardised spelling test described
in study 1. In this pseudohomophone task 40 familiar words (known by > 90 % of Belgian
six-year-olds; see Schaerlaekens, Kohnstamm & Lejaeghere, 1999) were presented on a
screen together with a non-word with the same pronunciation (e.g. ‘blauw’ vs ‘blouw’).
Simultaneous with the visual presentation the word was also auditory presented and used in
an example sentence. The child’s task was to point to the correctly spelled word. The test was
integrated in a computer game in a way that upon correct responding a jigsaw piece popped
up revealing a funny image throughout the task. Stimulus presentation and response
registration was controlled by E-prime, a software module for psychological experiments
(Schneider, Eschman & Zuccolotto, 2002).
Statistical analysis
For two subjects showing a missing test score, data were estimated using the LISREL
multiple imputation option. To obtain a normal distribution, results on the letter knowledge
task and thresholds for FM and CM were log10-transformed. Further on, CM and FM
134
thresholds were multiplied by -1 to obtain a positive definite covariance matrix; hence a
higher score on a measure indicated a better capability.
Structural equation modeling (SEM) was used to examine the hypotheses. Analysis of
the covariance matrices was conducted using LISREL 8.71 (Jöreskog & Sorbom, 2004) and
solutions were generated on the basis of maximum-likelihood estimation (Boomsma &
Hoogland, 2001). Given that we had multiple indicators of the theoretical constructs, we
wished to combine measures to simplify the models and increase reliability. However,
because of the low ratio of participants to free parameters, it was not appropriate to estimate
latent variables for the path models (Tanaka, 1987). Instead, we used observed composite
variables, using aggregated scores to define the constructs for which we had multiple
measures. To further control the number of free parameters to be estimated, models were
estimated separately for the prediction of reading and writing development.
Data screening of the (composite) variables indicated partial data non-normality, both
at the univariate and the multivariate level7. Therefore, in all models the asymptotic
covariance matrix was used as input and the Satorra-Bentler Scaled chi-square test was
inspected (SBS-χ², Sattorra & Bentler, 1994). To evaluate model goodness of fit, the
Comparative Fit Index (CFI), the Root Mean Square Error of Approximation (RMSEA) and
the Standardized Root Mean Square Residuals (SRMR) were selected. According to Hu and
Bentler (1999), the combined cut-off values close to .95 for CFI, below .08 for RMSEA and
below .09 for SRMR indicate good model fit. SBS-χ² difference tests were used in
comparisons between nested models.
Results
Data reduction
As reported in study 1 and in Boets et al. (2006a), the exploratory factor analysis
convincingly confirmed the traditional three-fold phonological structure. Consequently, for
every phonological factor summarizing composite scores were calculated averaging the
standardised scores of their constituent tests (AWARENESS, RAN and VSTM). Similarly,
averaging the Z-scores of the six reading tests created a reading composite (READ), and
averaging the Z-scores of the spelling and the pseudohomophone task constituted a writing
composite (WRITE). Although both of these literacy composites were highly correlated
(r = .72, p < .0001), it was thought this subdivision might be conceptually valid. Further, also
135
for the speech-in-noise perception data a SPEECH composite was calculated averaging the
Z-scores of the proportion correctly perceived words for the -4 dB and -7 dB SNR conditions.
Finally, for the auditory and visual measures (FM and CM), the average of the best and
second best threshold was used as an indicator of a child’s sensory sensitivity.
Construction of a preschool measure for orthographic ability
In accordance with the theoretical model, visual magnocellular processing is
postulated to determine literacy development by means of orthographic ability. Since it is
impossible to administer a pure orthographic test at this preschool age, we considered letter
knowledge as the best approximate measure to give an indication about orthographic ability.
After all, resolving a letter knowledge task relies on recognizing the visual features of the
written symbol on the one hand, and retrieving the corresponding linguistic information on
the other. As such, letter knowledge might be regarded as a measure that reflects both
orthographic and phonological skills. To create a more ‘pure’ preschool orthographic
measure, we extracted all the phonological aspects out of the letter knowledge task by
statistically removing all the variance due to differences in phonological awareness, rapid
automatic naming and verbal short-term memory. As a result, the newly created variable
ORTHOGRAPHY represented the letter knowledge scores residualized on AWARENESS,
RAN and VSTM using simultaneous regression.
Correlations among sensory processing, speech perception, orthographic and phonological
ability and literacy skills
Correlations between the (composite) variables are shown in Table 5. Bivariate
correlations are shown below the diagonal and partial correlations controlling for non-verbal
intelligence are shown above the diagonal. As hypothesised, both dynamic sensory processing
measures were indeed significantly related. Furthermore, whereas visual processing was
related to orthographic ability, auditory processing was related to speech perception and
phonological awareness. Regarding the phonological sub skills, phonological awareness was
significantly related to RAN and VSTM, who themselves were unrelated. Finally, reading
ability was significantly predicted by FM and CM sensitivity, orthographic ability, speech
perception, phonological awareness, RAN and VSTM. Writing ability was significantly
predicted by orthographic ability, speech perception, phonological awareness and VSTM.
Table 5
Pearson (partial) correlations between measures for sensory processing, speech perception, orthographic and phonological ability and literacy skills.
RCPM
CM
ORTHO-
GRAPHY
FM
SPEECH
AWARE-
NESS
RAN
VSTM
READ
WRITE
CM
ORTHOGRAPHY
FM
SPEECH
AWARENESS
RAN
VSTM
READ
WRITE
.25*
.17
.16
-.10
.17
-.14
.25*
.12
.10
-
.29*
.30*
.09
.00
-.08
.13
.34**
.23
.26*
-
.20
.28*
.00
.00
.00
.32**
.38**
.27*
.17
-
.33**
.45***
.05
.09
.44***
.23
.12
.30*
.35**
-
.35**
.15
.10
.48****
.33**
-.04
-.03
.43***
.37**
-
.35**
.33**
.52****
.40***
-.05
.02
.07
.14
.38**
-
.08
.24*
.10
.07
-.04
.05
.13
.30*
.12
-
.35**
.32**
.33**
.31*
.43***
.49****
.51****
.25*
.34**
-
.72****
.22
.37**
.22
.34**
.39**
.12
.31*
.71****
-
Note: Coefficients above the diagonal are partial correlations after removing variance attributable to individual differences in nonverbal intelligence (RCPM).
* p < .05, ** p < .01, *** p < .001, **** p < .0001
137
Noteworthy, interrelations were generally found to be larger in the LR group
compared to the total or HR group, indicating that the total group correlations were not
inflated because of aggregating data over extreme groups.
Path model for reading development
Testing of the theoretical model depicted in Fig. 5 resulted in a moderate global fit
(SBS-χ²(15) = 25.02, p < .05; CFI = .90; RMSEA = .11; SRMR = .12) with all the path
coefficients being significant except for the SPEECH → RAN path, the SPEECH → VSTM
path, the RAN → READ path and the correlation between RAN and VSTM. Since
elimination of these non-significant paths resulted in a statistically equivalent model, this
reduced model formed the basis for further model building.
Because FM sensitivity was significantly related to phonological awareness, we tested
whether adding a direct path would improve the model compared to the fully speech-
mediated model. This yielded a significant improvement (∆SBS-χ² (1) = 7.05, p < .01) and
resulted in an excellently fitting model (SBS-χ² (13) = 17.46, p = ns; CFI = .96;
RMSEA = .08; SRMR = .09). Interestingly, as a consequence of adding the direct path, the
SPEECH → AWARENESS path became insignificant (β = .17, p > .05), indicating that the
influence of auditory processing on phonological awareness happens in a direct way and is
not or only very marginally mediated by speech perception (see Baron & Kenny, 1986, and
Holmbeck, 1997 for a clear conceptual and statistical definition of mediation8).
Similarly, because speech perception was significantly related to reading, we explored
whether adding a direct path would improve the model compared to the fully phonologically-
mediated model. Although the added direct path was highly significant (β = .27, p < .01), the
global model improvement was not (∆SBS-χ² (1) = 2.70, p > .05). Besides the significant
direct path, also the indirect SPEECH → AWARENESS → READ path was retained
(β = .29, p < .005 and β = .34, p < .001, respectively). This suggests that phonological
awareness indeed partially mediates the relation between speech perception and reading
development.
Likewise, in order to verify whether the relation between visual magnocellular
processing and reading would be fully mediated by orthography, we compared the fully
orthography-mediated model with one where we added the direct CM → READ path. Again
the added direct path was highly significant (β = .25, p < .01), but the global model
138
improvement was not (∆SBS-χ² (1) = 2.56, p > .05). Moreover, since both path coefficients
from the indirect path were retained (β = .29, p < .05 and β = .25, p < .01, respectively), this
indicated that the relation between visual magnocellular processing and reading development
is at least partially mediated by orthographic ability.
Fig. 6: Path analysis predicting first grade reading achievement from preschool measures of dynamic sensory
processing, speech perception, phonological and orthographic ability.
* p < .05, ** p < .01, *** p < .005, **** p < .001
Ultimately, addition of the three abovementioned significant direct paths resulted in
the excellently fitting final structural model depicted in Fig. 6 (SBS-χ² (11) = 8.76, p = ns;
CFI = 1.00; RMSEA = 0.00; SRMR = .06)9. According to this estimated model 51 % of the
observed variance in reading ability could be predicted. Noteworthy, applying a model
trimming strategy where we initially allowed any possible relations and systematically
omitted all non-significant paths, resulted in exactly the same model. Finally, in order to
verify the stability of the model, we repeated the analysis upon the data partialed out for
individual differences in non-verbal intelligence (i.e. all variables were regressed upon the
scores for Raven Coloured Progressive Matrices and the residualized scores were used to
calculate the covariance matrix). The fit indices and path coefficients of this model were
139
virtually identical to the ones presented in Fig. 6, with the only difference that the SPEECH
→ AWARENESS path became significant again (β = .19, p < .05).
Path model for writing development
Testing of the structural model for writing development was executed in a similar way
as we did for reading development. In contrast to the reading model, testing of the theoretical
writing model depicted in Fig. 5 already resulted in a fairly satisfying global fit (SBS-χ²(15) =
19.73, p = ns; CFI = .93; RMSEA = .07; SRMR = .04). Again, all the postulated paths were
significant apart from the correlation between RAN and VSTM and the SPEECH → RAN
path, the SPEECH → VSTM path and the RAN → WRITE path. Since elimination of these
non-significant paths resulted in a statistically equivalent model, again this reduced model
was used as a basis for further model building.
Exploration whether the model could be improved by adding the same direct paths as
mentioned above, indicated that only the addition of the direct FM → AWARENESS path
resulted in a significant path coefficient (β = .36, p < .001) and a significant improvement of
model fit (∆SBS-χ² (1) = 9.08, p < .005). Consecutive addition of the SPEECH → WRITE
path and CM → WRITE path did not only fail to yield a global model improvement (∆SBS-χ²
(1) = 1.24, p > .05 and ∆SBS-χ² (1) = 1.48, p > .05, respectively), but did neither result in a
significant path coefficient (β = .11, p > .05 and β = .10, p > .05, respectively). Consequently,
this indicated that whereas the path from auditory processing towards phonological awareness
is only marginally mediated by speech perception, the path from speech perception towards
writing development is fully mediated by phonological awareness, as well as the path from
visual magnocellular processing towards writing is fully mediated by orthographic skills. In
conclusion, the revised writing model with the additional FM → AWARENESS path resulted
in the excellently fitting model depicted in Fig. 7 (SBS-χ² (13) = 10.23, p = ns; CFI = 1.00;
RMSEA = 0.00; SRMR = .07). According to this model 35 % of the variance in writing skills
could be predicted. Again, applying an empirically guided model trimming strategy resulted
in exactly the same model. Also if we retested the model for the scores residualized on non-
verbal intelligence, fit indices and path coefficients were virtually identical, with the only
difference being the significant SPEECH → AWARENESS path (β = .19, p < .05).
140
Fig. 7: Path analysis predicting first grade writing achievement from preschool measures of dynamic sensory
processing, speech perception, phonological and orthographic ability.
* p < .05, ** p < .01, *** p < .005, **** p < .001
Discussion
In this study we investigated the interrelations between preschool measures of
dynamic sensory processing, speech perception, orthographic and phonological ability and
first grade measures of reading and writing achievement. In particular we evaluated whether
the hypothesized causal path, as suggested by the general magnocellular theory, could be
validated empirically. To eliminate plausible rival hypotheses, several competing
hierarchically nested models were statistically compared.
Inspection of the empirically deduced models depicted in Fig. 6 and Fig. 7 indicates
that the basic causal assumptions of the general magnocellular theory could be validated fairly
well. In both models dynamic auditory processing was related to speech perception, which on
its turn was related to phonological awareness. In the same way, dynamic visual processing
was related to orthographic ability. Subsequently, phonological awareness and orthographic
ability –together with verbal short-term memory- were unique predictors of literacy
development. Somehow unexpectedly, the model fit could significantly be improved by
141
allowing a direct influence of dynamic auditory processing upon phonological awareness,
indicating that this determination happened mostly in a direct way and was only marginally
mediated by speech perception. Similarly, for the reading model there was also an additional
direct influence of speech perception upon reading development, besides the considerable
influence mediated by phonological awareness. Likewise, besides the influence mediated by
orthography, visual magnocellular processing also executed a direct influence upon reading
development. For the writing model in contrast, these two additional direct paths were not
retained. A possible explanation for the retention of this additional direct path from visual
magnocellular processing towards reading and not towards writing achievement, might be the
relatively larger involvement of parietal processes like eye movement control and peripheral
vision in reading compared to writing.
Comparing our data to the SEM studies of McBride-Chang (1996), Watson and Miller
(1993) and Schulte-Körne and colleagues (1999), both correspondences and differences can
be observed. The three studies as well as ours converge in demonstrating a significant relation
between speech perception and phonological processing. However, where we only
demonstrated a significant influence of speech perception upon phonological awareness,
McBride-Chang and Watson and Miller also observed an influence upon respectively rapid
automatic naming and verbal (short- and long-term) memory. Further in contrast with their
results, we also observed a direct influence of speech perception upon reading besides the
phonology-mediated influence. As reported, in the studies of Watson and Miller (1993) and
Schulte-Körne and colleagues (1999) auditory temporal measures were also administered.
However, since Schulte-Körne et al. did not observe significant group differences on any of
these measures comparing literacy disabled versus control children, they did not integrate
them into the structural model. Conversely, in the study of Watson and Miller a latent
nonverbal temporal processing factor comprising four tasks was estimated (with only one of
the tasks showing a significant difference comparing control versus literacy disabled
subjects). In their model fitting they demonstrated that auditory temporal processing did not
additionally predict phonological processing when the influence of speech perception was
already taken into account. However, they did not explore the theoretically more relevant
hypothesis whether speech perception would be determined by auditory processing.
Obviously, another major difference of our study compared to these three studies is the
fact that we did not only investigate the auditory pathway to literacy development, but also
included the orthography-mediated visual pathway. Considering the results of this combined
142
perspective, our preschool data confirmed previous results from adults and school-aged
children demonstrating that dynamic auditory sensitivity seems to be uniquely related to
phonological skills, whereas dynamic visual sensitivity seems to be specifically related to
orthographic skills (Talcott, Hansen, et al., 2000; Talcott, Witton, et al., 2000; Talcott &
Witton, 2002).
Regarding the relations between preschool phonological processing and early literacy
development, our data basically corroborated numerous previous prospective studies (see e.g.
Scarborough, 1998, for an overview of 61 studies). Each of the three phonological sub skills
individually was significantly associated with reading. For phonological awareness and verbal
short-term memory the magnitude of these relations was well in line with other prediction
studies, for rapid automatic naming this magnitude was a bit reduced. When all three skills
were included simultaneously in an analysis to predict literacy achievement, only
phonological awareness and verbal short-term memory emerged as unique predictors. This
was somehow unexpected since generally phonological awareness and verbal short-term
memory tend to share more common variance, with rapid automatic naming being more
distinct when predicting reading achievement (e.g. Bowers & Ishaik, 2003). However, as
mentioned by Elbro and Scarborough (2003), the relative importance of a predictor depends
on the level of reading proficiency we aim to predict as well as on the specific reading skill
assessed. Indeed, it has been demonstrated that phonological awareness is mostly important in
the initial phase of reading development when reading mainly depends upon phonological
decoding (Elbro & Scarborough, 2003). Conversely, during a later phase the development of
decoding speed gains more importance and then rapid automatic naming becomes a more
important predictor10 11(see also van den Bos, Zijlstra & lutje Spelberg, 2002). Since we
assessed reading skills at the end of first grade when children hardly received nine months of
reading instruction, it is evident that their reading was not yet automated and still called
mainly upon phonological decoding and thus upon phonological awareness. Furthermore, it
has also been suggested that phonological awareness is most closely related to reading
accuracy and non-word reading, whereas rapid automatic naming is specifically related to
reading speed (de Jong & Wolters, 2002; Savage & Frederikson, 2005). Since our reading
composite showed a preponderance of non-word reading and reading accuracy tasks, this is
probably an additional explanation for the reduced predictive power of rapid automatic
naming. Indeed, when correlating rapid automatic naming scores with the reading test that
specifically measured word reading speed an increased correlation was observed.
143
A specific problem of SEM studies is the existence of several plausible alternative
models that are mathematically equivalent though substantively different (MacCallum,
Wegener, Uchino & Fabrigar, 1993). Indeed, there are many models of associations among
sensory processing, speech perception, orthographic/phonological ability and literacy
achievement that could have been tested. The best approach to manage this problem is the
specification and systematic elimination of plausible rival models, as we did in the current
study. Moreover, the fact that both the theoretically guided model building and the
empirically guided model trimming strategy eventuated in the same final structural model,
provides further validity to our results.
A somehow related problem concerns the issue of causality. It is well known that in a
strict sense causality can only be demonstrated in an experimental and not in a correlational
study. However, by applying a longitudinal design we intended to clarify at least the direction
of some of the observed relations. Accordingly, in view of the fact that all measures for
sensory processing, speech perception, orthographic and phonological ability were collected
before children received any formal reading instruction, it is evident that any relation with
later literacy development is unidirectional. On the contrary, the nature of the relations within
this group of concurrently administered preschool measures is much less straightforward.
Indeed, similar to the established observation that phonological awareness and literacy
development influence each other in a bidirectional way (e.g. Morais, Bertelson, Cary &
Alegria, 1979; Perfetti, Beck, Bell & Hughes, 1987; Read, Zhang, Nie & Ding, 1986), there is
also growing evidence that auditory processing, speech perception and phonological ability
influence each other reciprocally.
In two intervention studies it was convincingly demonstrated that speech
discrimination is causally related to development in phonological skills. Hurford (1990) found
that disabled readers in second and third grade who received speech discrimination training
improved their segmentation skills significantly compared to those disabled readers who had
not received training. Similarly, Moore, Rosenberg and Coleman (2005) trained 8- to 10-year-
old mainstream school children presenting eleven phonemic contrast continua. Trained
children significantly improved phonological awareness and word listening skills by about
two years, although they only showed weak and variable improvement on the trained
discriminations. No improvements were found in the control children. However, contrary to
the theoretically hypothesised direction, there is also some evidence that development in
phonological awareness seems to stimulate (or at least precede) development in auditory
144
processing and speech perception. For example, Mayo and colleagues (2003) demonstrated
that phonemic awareness skills tended to improve before cue weighting strategies in speech
perception changed, and that early phonemic awareness ability predicted later cue weighting
strategies and not the other way around. In another research tradition, Warrier et al. (2004)
demonstrated that a phonological/language training program given to a group of learning-
impaired children not only enhanced their perceptual encoding of speech but also significantly
improved the timing of their cortical responses to speech-in-noise. Moreover, it is interesting
to note that the timing of this cortical response was altered independently of the brainstem
onset responses (Hayes et al, 2003), suggesting there might be a higher-order compensation
mechanism for altering cortical stimulus encoding even when the input from the brainstem is
still deficient. With respect to the causality issue, these latter studies indicate that the relation
between auditory processing, speech perception and phonological awareness is not as
straightforward as has been hypothesized and should probably be conceptualized as being a
bidirectional one. Nevertheless, the reciprocal character of these relations does not essentially
conflict with the basic assumptions of the general magnocellular theory. Actually, it is very
well possible that minor initial problems in sensory processing will multiply up to more
serious deficits exactly because of the interactive nature of these relations.
In sum, although we do not exclude that there might be other viable models to explain
the observed relations between sensory processing, speech perception, orthographic and
phonological ability and literacy achievement, the current study demonstrates that the causal
model proposed by the general magnocellular theory is at least a plausible one. Hence, this is
the first longitudinal study investigating preschool children throughout early literacy
development that offers empirical evidence for the validity of the causal relationships
postulated by the general magnocellular theory.
145
References
Adlard, A., & Hazan, V. (1998). Speech perception in children with specific reading
difficulties (dyslexia). Quarterly Journal of Experimental Psychology, 51(1), 153-177.
Amitay, S., Ahissar, M., & Nelken, I. (2002). Auditory processing deficits in reading disabled
adults. Journal of the Association for Research in Otolaryngology, 3, 302-320.
Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator variable distinction in social
psychological research: conceptual, strategic and statistical considerations. Journal of
Personality and Social Psychology, 51(6), 1173-1182.
Benasich, A. A., & Tallal, P. (2002). Infant discrimination of rapid auditory cues predicts later
language impairment. Behavioural Brain Research, 136(1), 31-49.
Bishop, D. V., Bishop, S. J., Bright, P., James, C., Delaney, T., & Tallal, P. (1999). Different
origin of auditory and phonological processing problems in children with language
impairment: evidence from a twin study. Journal of Speech, Language and Hearing
Research, 42(1), 155-168.
Bishop, D. V., Carlyon, R. P., Deeks, J. M., & Bishop, S. J. (1999). Auditory temporal
processing impairment: neither necessary nor sufficient for causing language
impairment in children. Journal of Speech, Language and Hearing Research, 42(6),
1295-1310.
Boets, B., Wouters, J., van Wieringen, A., & Ghesquière, P. (2006a). Auditory temporal
information processing in preschool children at family risk for dyslexia: relations with
phonological abilities and developing literacy skills. Brain and Language, 97, 64-79.
Boets, B., Wouters, J., van Wieringen, A., & Ghesquière, P. (2006b). Coherent motion
detection in preschoolers at family risk for dyslexia. Vision Research, 46(4), 527-535.
Boets, B., Ghesquière, P., van Wieringen, A., & Wouters, J. (under 2nd revision). Speech
perception in preschoolers at family risk for dyslexia: relations with low-level auditory
processing and phonological ability. Brain and Language.
Boomsma, A., & Hoogland, J. J. (2001). The robustness of LISREL modeling revisited. In R.
146
Cudeck, S. du Toit, & D. Sörbom (Eds.), Structural equation models: Present and
future. A Festschrift in honor of Karl Jöreskog (pp. 139-168). Chicago: Scientific
Software International.
Bowers, P. G., & Ishaik, G. (2003). RAN's contribution to understanding reading disabilities.
In H. L. Swanson, K. R. Harris, & S. Graham (Eds.), Handbook of learning disabilities
(pp. 140-157). New York: The Guilford Press.
Bradlow, A. R., Kraus, N., & Hayes, E. (2003). Speaking clearly for children with learning
disabilities: sentence perception in noise. Journal of Speech, Language and Hearing
Research, 46, 80-97.
Brady, S., Shankweiler, D., & Mann, V. (1983). Speech perception and memory coding in
relation to reading ability. Journal of Experimental Child Psychology, 35(2), 345-367.
Breier, J. I., Fletcher, J. M., Denton, C., & Gray, L. C. (2004). Categorical perception of
speech stimuli in children at risk for reading difficulty. Journal of Experimental Child
Psychology, 88, 152-170.
Britten, K. H., Shalden, M. N., Newsome, W. T., & Movshon, J. A. (1992). The analysis of
visual motion: A comparison of neuronal and psychophysical performance. Journal of
Neuroscience, 12, 4745-4765.
Brus, B. T., & Voeten, M. J. (1973). Eén-minuut-test vorm A en B. Verantwoording en
handleiding. Nijmegen: Berkhout.
Cestnick, L. (2001). Cross-modality temporal processing deficits in developmental
phonological dyslexics. Brain and Cognition, 46, 319-325.
Cohen, J. (1988). Statistical power analysis for the behavioural sciences (2nd ed.). New York:
Academic Press.
Cornelissen, P., Richardson, A., Mason, A., Fowler, S., & Stein, J. (1995). Contrast sensitivity
and coherent motion detection measures at photopic luminance levels in dyslexics and
controls. Vision Research, 35, 1483-1494.
Cornelissen, P. L., Hansen, P. C., Gilchrist, I. , Cormack, F., Essex, J., & Frankish, C. (1998).
Coherent motion detection and letter position encoding. Vision Research, 38, 2181-
147
2191.
de Jong, P. F., Seveke, M. J., & Van Veen, M. (2000). Phonological sensitivity and the
acquisition of new words in children. Journal of Experimental Child Psychology,
76(4), 275-301.
de Jong, P. F., & van der Leij, A. (1999). Specific contributions of phonological abilities to
early reading acquisition: Results from a Dutch latent variable longitudinal study.
Journal of Educational Psychology, 91(3), 450-476.
de Jong, P. F., & Wolters, G. (2002). Fonemisch bewustzijn, benoemsnelheid en leren lezen.
Pedagogische Studieën, 79, 53-63.
De Smedt, B., Ghesquière, P., & Verschaffel, L. (2004). Working memory and arithmetic
abilities in seven and eleven year old children. Poster Presented at the European
Working Memory Symposium EWOMS II, 22-24 April 2004, Beaune, France.
de Weirdt, W. (1988). Speech perception and frequency discrimination in good and poor
readers. Applied Psycholinguistics, 9(2), 163-183, US: Cambridge Univ Press.
Demb, J., Boynton, G., & Heeger, G. (1997). Brain activity in visual cortex predicts
individual differences in reading performance. Proceedings of the National Academy
of Sciences, USA, 94, 13363-13366.
Dudal, P. (1997). Leerlingvolgsysteem VCLB (CSBO). Spelling: Toetsen 1-2-3. Basisboek en
kopieerbundel. Leuven: Garant.
Eden, G., Van Meter, J., Rumsey, J., Maisog, J., Woods, R., & Zeffiro, T. (1996). Abnormal
processing of visual motion in dyslexia revealed by functional brain imaging. Nature,
382, 66-69.
Eden, G. F., Stein, J. F., & Wood, F. B. (1995). Temporal and spatial processing in reading
disabled and normal children. Cortex, 31, 451-468.
Elbro, C., & Scarborough, H. S. (2003). Early identification. In P. Bryant & T. Nunes (Eds.),
Handbook of children's reading (pp. 339-359). Dordrecht: Kluwer Academic
Publishers.
148
Everatt, J., Bradshaw, M., & Hibbard, P. (1999). Visual processing and dyslexia. Perception,
28, 243-254.
Farmer, M. E., & Klein, R. M. (1995). The evidence for a temporal processing deficit linked
to dyslexia: A review. Psychonomic Bulletin and Review, 2(4), 460-493.
Frith, C., & Frith, U. (1996). A biological marker for dyslexia. Nature, 382, 19-20.
Galaburda, A. M., Menard, M. T., & Rosen, G. D. (1994). Evidence for aberrant auditory
anatomy in developmental dyslexia. Proceedings of the National Academy of Science
of the USA, 91(17), 8010-8013.
Gerrits, E. (2003). Speech perception of young children at risk for dyslexia and children with
specific language impairment. Proceedings of the 15th International Conference on
Phonetic Sciences (pp. 2357-2360). Barcelona, Spain, 3 - 9 Augustus 2003.
Gersons-Wolfensberger, D. J. M. & Ruijssenaars, W. A. J. J. M. (1997). Definition and
treatment of dyslexia: A report by the Committee on Dyslexia of the Health Council of
the Netherlands. Journal of Learning Disabilities, 30(2), 209-213.
Godfrey, J. J., Syrdal-Lasky, A. K., Millay, K. K., & Knox, C. M. (1981). Performance of
dyslexic children on speech perception tests. Journal of Experimental Child
Psychology, 32, 401-424.
Hansen, P. C., Stein, J. F., Orde, S. R., Winter, J. .L, & Talcott, J. B. (2001). Are dyslexics'
visual deficits limited to measures of dorsal stream function? Neuroreport, 12(7),
1527-1530.
Hautus, M. J., Setchell, G. J., Waldie, K. E., & Kirk, I. J. (2003). Age-related improvements
in auditory temporal resolution in reading-impaired children. Dyslexia, 9, 37-45.
Hayes, E., Warrier, C. M., Nicol, T., Zecker, S. G., & Kraus, N. (2003). Neural plasticity
following auditory training in children with language-learning problems. Clinical
Neurophysiology, 114, 673-684.
Heath, S. M., & Hogben, J. H. (2004). Cost-effective predicition of reading difficulties.
Journal of Speech, Language and Hearing Research, 47, 751-765.
149
Holmbeck, G. N. (1997). Toward terminological, conceptual and statistical clarity in the study
of mediators and moderators: examples from the child-clinical and pediatric
psychology literature. Journal of Consulting and Clinical Psychology, 65(4), 599-610.
Hood, M., & Conlon, E. (2004). Visual and auditory temporal processing and early reading
development. Dyslexia, 10, 234-252.
Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure
analysis: conventional criteria versus new alternatives. Structural Equation Modeling,
6, 1-55.
Hurford, D. P. (1990). Training phonemic segmentation ability with a phonemic intervention
in second- and third-grade children with reading disabilities. Journal of Learning
Disabilities, 23, 564-569.
Jöreskog, K. G., & Sorböm, D. (2004). LISREL 8.71. Chicago: Scientific Software
International, Inc.
Kline, R. B. (1998). Principles and practice of Structural Equation Modeling. New York: The
Guilford Press.
Kowalchuk, R. K., Keselman, H. J., Algina, J., & Wolfinger, R. D. (2004). The analysis of
repeated measurements with mixed-model adjusted F tests. Educational and
Psychological Measurement, 64(2), 224-242.
Landerl, K., Wimmer, H., & Frith, U. (1997). The impact of orthographic consistency on
dyslexia: a German-English comparison. Cognition, 63, 315.334.
Laneau, J., Boets, B., Moonen, M., van Wieringen, A., & Wouters, J. (2005). A flexible
auditory research platform using acoustic or electric stimuli for adults and young
children. Journal of Neuroscience Methods, 142(1), 131-136.
Levitt, H. (1971). Transformed up-down methods in psychoacoustics. Journal of the
Acoustical Society of America, 49, 467-477.
Lieberman, P., Meskill, R. H., Chatillon, M., & Schupack, H. (1985). Phonetic speech
perception deficits in dyslexia. Journal of Speech and Hearing Research, 28(480-486).
150
Littell, R. C., Stroup, W. W., & Freund, R. J. (2002). SAS for linear models. Cary: SAS
Institute.
Livingstone, M., & Hubel, D. (1987). Psychophysical evidence for separate channels for the
perception of form, colour, movement and depth. Journal of Neuroscience, 7, 3416-
3468.
Livingstone, M. S., Rosen, G. D., Drislane, F. W., & Galaburda, A. M. (1991). Physiological
and anatomical evidence for a magnocellular defect in developmental dyslexia.
Proceedings of the National Academy of Science, USA, 88(18), 7943-7947.
Lovegrove, B. (1996). Dyslexia and a transient/magnocellular pathway deficit: The current
situation and future directions. Australian Journal of Psychology, 48(3), 167-171.
Lyytinen, H., Ahonen, T., Eklund, K., Guttorm, T. K., Laakso, M. L., Leinonen, S.,
Leppanen, P. H., Lyytinen, P., Poikkeus, A. M., Puolakanaho, A., Richardson, U., &
Viholainen, H. (2001). Developmental pathways of children with and without familial
risk for dyslexia during the first years of life. Developmental Neuropsychology, 20(2),
535-554.
Maassen, B., Groenen, P., Crul, T., Assman-Hulsmans, C., & Gabrëls, F. (2001).
Identification and discrimination of voicing and place-of-articulation in developmental
dyslexia. Clinical Linguistics and Phonetics, 15(4), 319-339.
MacCallum, R. C., Wegener, D. T., Uchino, B. N., & Fabrigar, L. R. (1993). The problem of
equivalent models in applications of covariance structure analysis. Psychological
Bulletin, 114 , 185-199.
Manis, F. R., McBride-Chang, C., Seidenberg, M. S., Keating, P., Doi, L. M., Munson, B., &
Petersen, A. (1997). Are speech perception deficits associated with developmental
dyslexia? Journal of Experimental Child Psychology, 66, 211-235.
Mayo, C., Scobbie, J. M., Hewlett, N., & Waters, D. (2003). The influence of phonemic
awareness development on acoustic cue weighting strategies in children's speech
perception. Journal of Speech, Language and Hearing Research, 46, 1184-1196.
McAnally, K. I., & Stein, J. F. (1996). Auditory temporal coding in dyslexia. Proceedings of
151
the Royal Society of London. Series B, Biological Sciences, 263, 961-965.
McBride-Chang, C. (1995). Phonological processing, speech perception and reading
disability: an integrative review. Educational Psychologists, 30(3), 109-121.
McBride-Chang, C. (1996). Models of speech perception and phonological processing in
reading. Child Development, 67(4), 1836-1856.
McCardle, P., Scarborough, H. S., & Catts, H. W. (2001). Predicting, explaining and
preventing children's reading difficulties. Learning Disabilities Research and Practice,
16(4), 230-239.
McCroskey, R. L., & Kidder, H. C. (1980). Auditory fusion among learning disabled, reading
disabled, and normal children. Journal of Learning Disabilities, 13(2 ), 69-76.
Moore, D. R., Rosenberg, J. F., & Coleman, J. S. (2005). Discrimination training of phonemic
contrasts enhances phonological processing in mainstream school children. Brain and
Language, 94, 72-85.
Morais, J., Bertelson, P., Carey, L., & Alegria, J. (1979). Does awareness of speech as a
sequence of phones arise spontaneously? Cognition, 7, 323-331.
Nittrouer, S. (1996). The relation between speech perception and phonemic awareness:
evidence from low-SES children and children with chronic OM. Journal of Speech
and Hearing Research, 39(5), 1059-1070.
Olson, R., & Datta, H. (2002). Visual-temporal processing in reading-disabled and normal
twins. Reading and Writing, 15, 127-149.
Pennington, B. F. & Lefly, D. L. (2001). Early reading development in children at family risk
for dyslexia. Child Development, 72(3), 816-833.
Perfetti, C. A., Beck, I., Bell, L. C., & Hughes, C. (1987). Phonemic knowledge and learning
to read are reciprocal: a longitudinal study of first grade children. Merril-Palmer
Quarterly, 33, 283-319.
Phillips, D. P., Taylor, T. L., Hall, S. E., Carr, M. M., & Mossop, J. E. (1997). Detection of
silent intervals between noises activating different perceptual channels: some
152
properties of 'central' auditory gap detection. Journal of the Acoustical Society of
America, 101(6), 3694-3705.
Ramus, F. (2001). Dyslexia. Talk of two theories. Nature, 412(6845), 393-5.
Ramus, F. (2003). Developmental dyslexia: specific phonological deficit or general
sensorimotor dysfunction? Current Opinions in Neurobiology, 13(2), 212-218.
Ramus, F. (2004). Neurobiology of dyslexia: a reinterpretation of the data. Trends in
Neurosciences, 27(12), 720-726.
Ramus, F., Rosen, S., Dakin, S. C., Day, B. L., Castellote, J. M., White, S., & Frith, U.
(2003). Theories of developmental dyslexia: insights from a multiple case study of
dyslexic adults. Brain, 126(4), 841-865.
Raven, J. C., Court, J. H., & Raven, J. (1984). Manual for Raven's Progressive Matrices and
Vocabulary Scales. London: Lewis.
Raymond, J., & Sorensen, R. (1998). Visual motion perception in children with dyslexia:
normal detection but abnormal integration. Visual Cognition, 5(3), 389-404.
Read, C., Zhang, Y., Nie, H., & Ding, B. (1986). The ability to manipulate speech sounds
depends on knowing alphabetic writing. Cognition, 24, 31-44.
Reed, M. A. (1989). Speech perception and the discrimination of brief auditory cues in
reading disabled children. Journal of Experimental Child Psychology, 48(2), 270-92.
Richardson, U., Leppänen, P. H. T., Leiwo, M., & Lyytinen, H. (2003). Speech perception of
infants with high familial risk for dyslexia differ at the age of 6 months.
Developmental Neuropsychology, 23(3), 385-397.
Richardson, U., Thomson, J. M., Scott, S. K., & Goswami, U. (2004). Auditory processing
skills and phonological representation in dyslexic children. Dyslexia, 10(3), 215-233.
Ridder, W. H., Borsting, E., & Banton, T. (2001). All developmental dyslexic subtypes
display an elevated motion coherence threshold. Optometry and Vision Science, 78,
510-517.
Rosen, S. (2003). Auditory processing in dyslexia and specific language impairment: is there
153
a deficit? What is its nature? Does it explain anything? Journal of Phonetics, 31, 509-
527.
Satorra, A., & Bentler, P. M. (1994). Corrections to test statistics and standard errors in
covariance structure analysis. In A. von Eye, & C. C. Clogg (Eds.), Latent variable
analysis: Applications in developmental research (pp. 399-419). Newbury Park, CA:
Sage.
Savage, R., & Frederickson, N. (2005). Evidence of a highly specific relationship between
rapid automatic naming of digits and text-reading speed. Brain and Language, 93, 152-
159.
Scarborough, H. S. (1998). Early identification of children at risk for reading disabilities:
phonological awareness and some other promising predictors. In B.K. Shapiro, P.J.
Accardo, & A.J. Capute (Eds.), Specific reading disability: a view of the spectrum (pp.
75-119). Timonium, MD: York Press.
Schaerlaekens, A. M., Kohnstamm, D., & Lejaeghere, M. (1999). Streeflijst woordenschat
voor zesjarigen (derde herziene druk). Lisse: Swets & Zeitlinger.
Scheltinga, F. (2003). The Dutch Nonword Repetition Test. Unpublished Manuscript,
University of Amsterdam.
Schneider, W., Eschman, A., & Zuccolotto, A. (2002). E-Prime User's Guide. Pittsburgh:
Psychology Software Tools Inc.
Schulte-Körne, G., Deimel, W., Bartling, J., & Remschmidt, H. (1998). Role of auditory
temporal processing for reading and spelling disability. Perceptual and Motor Skills,
86(3), 1043-1047.
Schulte-Körne, G., Deimel, W., Bartling, J., & Remschmidt, H. (1999). The role of
phonological awareness, speech perception, and auditory temporal processing for
dyslexia. European Child and Adolescent Psychiatry, 8(Suppl. 3), 28-34.
Serniclaes, W., Sprenger-Charolles, L., Carré, R., & Demonet, J. (2001). Perceptual
discrimination of speech sounds in developmental dyslexia. Journal of Speech,
Language and Hearing Research, 44, 384-399.
154
Share, D. L., Jorm, A. F., MacLean, R., & Matthews, R. (2002). Temporal processing and
reading disability. Reading and Writing, 15(1-2), 151-178.
Skottun, B. C. (2000). The magnocellular deficit theory of dyslexia: the evidence from
contrast sensitivity. Vision Research, 40, 111-127.
Snowling, M. J., Gallagher, A., & Frith, U. (2003). Family risk of dyslexia is continuous:
individual differences in the precursors of reading skill. Child Development, 74(2),
358-373.
Snowling, M. J. (2000). Dyslexia (2nd ed.). Malden, MA: Blackwell Publishers.
Steffens, M. L., Eilers, R. E., Gross-Glenn, K., & Jallad, B. (1992). Speech perception in
adult subjects with familial dyslexia. Journal of Speech and Hearing Research, 35,
192-200.
Stein, J. (2001). The magnocellular theory of developmental dyslexia. Dyslexia, 7, 12-36.
Stein, J., & Talcott, J. (1999). Impaired neuronal timing in developmental dyslexia. The
magnocellular hypothesis. Dyslexia, 5, 59-77.
Stein, J., & Walsh, V. (1997). To see but not to read; the magnocellular theory of dyslexia.
Trends in Neuroscience, 20(4), 147-152.
Stein, J. F., & McAnally, K. (1995). Auditory temporal processing in developmental
dyslexics. Irish Journal of Psychology, 16(3), 220-228.
Talcott, J., Hansen, P., Willis-Owen, C., McKinnell, I., Richardson, A., & Stein, J. (1998).
Visual magnocellular impairment in adult developmental dyslexics. Neuro-
Ophtalmology, 20, 187-201.
Talcott, J. B., Gram, A., Van Ingelghem, M., Witton, C., Stein, J. F., & Toennessen, F. E.
(2003). Impaired sensitivity to dynamic stimuli in poor readers of a regular
orthography. Brain and Language, 87, 259-266.
Talcott, J. B., Hansen, P. C., Assoku, E. L., & Stein, J. F. (2000). Visual motion sensitivity in
dyslexia: evidence for temporal and energy integration deficits. Neuropsychologia,
38(7), 935-943.
155
Talcott, J. B., Witton, C., McClean, M., Hansen, P. C., Rees, A., Green, G., & Stein, J. F.
(1999). Can sensitivity to auditory frequency modulation predict children's
phonological and reading skills? Neuroreport, 10(10), 2045-2050.
Talcott, J. B., Witton, C., McLean, M. F., Hansen, P. C., Rees, A., Green, G. G., & Stein, J. F.
(2000). Dynamic sensory sensitivity and children's word decoding skills. Proceedings
of the National Academy of Science, USA, 97( 6), 2952-2957.
Talcott, J. B., & Witton, C. (2002). A sensory linguistic approach to the development of
normal and impaired reading skills. In E. Witruk, A. Friederici, & T. Lachmann
(Eds.), Neuropsychology and cognition series. Basic functions of language and
language disorders. Dordrecht, Netherlands: Kluwer Academic Publishers.
Tallal, P. (1980). Auditory temporal perception, phonics, and reading disabilities in children.
Brain and Language, 9(2), 182-198.
Tanaka, J. S. (1987). How big is big enough - sample size and goodness of fit in structural
equation models with latent variables. Child Development, 58, 134-146.
van Beinum, F. J., Schwippert, C. E., Been, P. H., van Leeuwen, T. H., & Kuijpers, C. T. L.
(2005). Development and application of a /bAk/-/dAk/ continuum for testing auditory
perception within the Dutch longitudinal study. Speech Communication, 47, 124-142.
van den Bos, K. P., Spelberg, H. C. L., Scheepstra, A. J. M., & De Vries, J. R. (1994). De
Klepel. Vorm A en B. Een test voor de leesvaardigheid van pseudowoorden.
Verantwoording, handleiding, diagnostiek en behandeling. Nijmegen: Berkhout.
van den Bos, K. P., Zijlstra, B. J. H., & Spelberg, H. C. (2002). Life-span data on continuous-
naming speeds of numbers, letters, colors, and pictured objects, and word-reading
speed. Scientific Studies of Reading, 6(1), 25-49.
Van Ingelghem, M., Boets, B., van Wieringen, A., Ghesquière, P., & Wouters, J. (2004).
Auditory and visual temporal processing in dyslexic and normal reading children
matched for sex, age and intellectual ability. Paper presented at 6th BDA International
conference, 27-30 march 2004, University of Warwick, UK .
Van Ingelghem, M., Boets, B., van Wieringen, A., Onghena, P., Ghesquière, P., & Wouters, J.
156
(2005). An auditory temporal processing deficit in children with dyslexia? In P.
Ghesquière & A.J.J.M. Ruijssenaars (Eds.), Learning disabilities: a challenge to
teaching and instruction. Series: Studia Paedagogica 40. (pp. 47-63). Leuven:
University Press.
Van Ingelghem, M., van Wieringen, A., Wouters, J., Vandenbussche, E., Onghena, P., &
Ghesquière, P. (2001). Psychophysical evidence for a general temporal processing
deficit in children with dyslexia. Neuroreport, 12(16), 3603-3607.
van Wieringen, A., & Wouters, J. (2005). Normalization and feasibility of speech
understanding tests for Dutch speaking toddlers. Speech Communication, 47, 169-181.
Wagner, R. K., Torgesen, J. K., & Rashotte, C. A. (1994). Development of reading-related
phonological processing abilities: new evidence of bi-directional causality from a
latent variable longitudinal study. Developmental Psychology, 30, 73-87.
Wagner, R. K., & Torgesen, J. K. (1987). The nature of phonological processing and its
causal role in the acquisition of reading skills. Psychological Bulletin, 101(2), 192-
212.
Wagner, R. K., Torgesen, J. K., Rashotte, C. A., & Hecht, S. A. (1997). Changing relations
between phonological processing abilities and word-level reading as children develop
from beginning to skilled readers: A 5-year longitudinal study. Developmental
Psychology, 33(3), 468-479.
Warrier, C. M., Johnson, K. L., Hayes, E. A., Nicol, T., & Kraus, N. (2004). Learning
impaired children exhibit timing deficits and training-related improvements in
auditory cortical responses to speech in noise. Experimental Brain Research, 157, 431-
441.
Watson, B. U., & Miller, T. K. (1993). Auditory perception, phonological processing and
reading ability/disability. Journal of Speech and Hearing Research, 36(4), 850-863.
Werker, J. F., & Tees, R. C. (1987). Speech perception in severely disabled and average
reading children. Canadian Journal of Experimental Psychology, 41, 48-61.
White, S., Milne, E., Rosen, S., Hansen, P., Swettenham, J., Frith, U., & Ramus, F. (in press).
157
The role of sensorimotor impairments in dyslexia: a multiple case study of dyslexic
children. Developmental Science.
Wilmer, J. B., Richardson, A. J., Chen, Y., & Stein, J. F. (2004). Two visual motion
processing deficits in developmental dyslexia associated with different reading skills
deficits. Journal of Cognitive Neuroscience, 16(4), 528-540.
Wimmer, H. (1993). Characteristics of developmental dyslexia in a regular writing system.
Applied Psycholinguistics, 14, 1-33.
Witton, C., Talcott, J. B., Hansen, P. C., Richardson, A. J., Griffiths, T. D., Rees, A., Stein, J.
F., & Green, G. G. (1998). Sensitivity to dynamic auditory and visual stimuli predicts
nonword reading ability in both dyslexic and normal readers. Current Biology, 8(14),
791-797.
Wood, C. C. (1976). Discriminability, response bias and phoneme categories in
discrimination of voice onset time. Journal of the Acoustic Society of America, 60(6),
1381-1389.
Wouters, J., Damman, W., & Bosman, A. J. (1994). Vlaamse opnamen van woordenlijsten
voor spraakaudiometrie. CD NKO - K.U.Leuven/U.Z.Leuven.
Wright, B. A., Lombardino, L. J., King, W. M., Puranik, C. S., Leonard, C. M., & Merzenich,
M. M. (1997). Deficits in auditory temporal and spectral resolution in language-
impaired children. Nature, 387, 176-8.
158
Notes
1 In the Belgian school system, formal instruction starts in grade 1 at the age of six. This
means in kindergarten no reading instruction is offered. 2 The reading accuracy tests for words and nonwords consisted both of 40 items gradually
increasing in difficulty. For these tests there was no time limit. To assess reading speed we
developed a test yielding almost 100 % accuracy, consequently reflecting merely variation in
reading speed and not in accuracy. In order to obtain a minimal error score, the word reading
speed test consisted exclusively of high-frequent one-syllable words, known by 90 % or more
of Belgian six-year-olds (Schaerlaekens, Kohnstamm & Lejaeghere, 1999). By decomposing
and recombining these words, the items for the nonword reading speed test were constructed.
Both speed tests consisted of 150 items, to be read within a two-minute time limit. 3 All MMA’s used the Kenward-Roger degrees of freedom estimation method that is more
robust against violations to assumptions of normality (Kowalchuk, Keselman, Angina &
Wolfinger, 2004). 4 For parental educational level the correspondence in frequency distribution of the different
educational categories was also confirmed using Fisher’s Exact Test. 5 For 8 subjects, the phoneme boundary was estimated below stimulus 1, indicating that they
even perceived the original /bAk/ stimulus as /dAk/ in more than 50 % of the presentations.
Data of these subjects were not included in the identification analyses. 6 Discriminability equals zero when performance is at chance. It increases with increasing
discrimination accuracy without influences of bias to respond ‘same’ or ‘different’.
Discriminability is maximal at the -ln eta value of 4.6; this value is obtained when the
probabilities for the correct ‘difference’ (= Hit) and correct ‘same’ responses (= Correct
Rejection) are both 0.99, the value assigned (for computational purposes) when the actual
probabilities were 1.00. 7 Although the sample represented a rather heterogeneous group with some children showing
serious literacy problems, data were normally distributed for the Reading and Writing
composite, as was the case for most other variables too. 8 According to Holmbeck (1997) and Baron and Kenny (1986), mediation is shown when the
addition of a direct path from the independent variable to the dependent variable does not
improve model fit compared to the fully indirect model. 9 In figure 6 and figure 7, the standardized solution with standardized regression coefficients
(β’s) is depicted. As the reader might notice, the significance level of a coefficient is not
exactly in proportion to its magnitude. This is due to two reasons: (a) mathematically SEM
159
calculates the significance levels for the unstandardized regression coefficients and (b) the
significance test reflects not only the absolute magnitude of a path coefficient but is also
determined by intercorrelations among variables (Kline, 1998). An alternative for interpreting
the results of statistical testing is the immediate interpretation of the absolute magnitudes of
path coefficients. In line with recommendations by Cohen (1988) about effect size
interpretations of correlations, Kline (1998) suggests the following guidelines: standardized
path coefficients with absolute values less than .10 may indicate a ‘small’ effect; values
around .30 a ‘medium’ effect; and those greater than .50 a ‘large’ effect. 10 This is particularly the case in languages with a more transparent orthography. In those
languages phonological demands are more easily met than in English due to the higher
regularity of grapheme-to-phoneme correspondences. Consequently, almost all children – also
those with literacy problems- tend to perform well on phonological awareness tasks and tend
to read highly accurately after some years of reading instruction (e.g. Landerl, Wimmer &
Frith, 1997; Wimmer, 1993). Therefore, in these languages reading speed -and as a result also
rapid automatic naming- is a more sensitive measure to differentiate between good and poor
readers. 11 However, other researchers reported different results regarding the changing relations
between phonological sub skills and reading development. In the longitudinal study of
Wagner and colleagues (1994, 1997) phonological awareness appeared to have a continuous
impetus on reading acquisition from kindergarten through 4th grade. In contrast, the rapid
automatic naming effects were found to be rather small and limited to the first three years. In
a similar longitudinal study examining phonological abilities in Dutch children from
kindergarten through 2nd grade de Jong and van der Leij (1999) reported that in kindergarten
only rapid automatic naming was specifically related to subsequent reading achievement, with
specific contributions of phonological awareness and verbal short-term memory appearing
only after a few months of reading instruction. However, as they concluded themselves, the
absence of a specific predictive influence of kindergarten phonological awareness and verbal
short-term memory was probably due to the use of an over-restrictive design controlling for
general cognitive abilities that were extremely highly correlated with these phonological sub
skills. Indeed, without taking intelligence into account, both phonological awareness and
verbal short-term memory were also found to exert significant additional influences on
reading achievement.
Appendix: Individual Z-scores for literacy, phonological ability, gap detection, tone-in-noise detection, FM detection, motion coherence detection, speech-in-noise perception (SRT) and categorical perception (slope of phoneme boundary). Deviant scores are depicted in bold.
Group Subject LITERACY PHONOLOGY GAP TN FM CM SRT
Slope Phoneme Boundary
HR-LD 11 -1.40 -1.07 1.32 0.91 0.76 1.76 1.96 25 -2.17 -2.45 1.37 1.54 -1.05 1.10 29 -2.28 -3.33 2.83 0.06 1.32 0.19 0.72 1.24 35 -1.58 -5.17 2.88 4.11 3.23 1.46 1.32 -0.70 41 -1.55 -1.66 1.19 1.62 0.34 0.75 1.07 -0.01 43 -1.46 -0.66 -0.68 -1.40 0.28 4.99 2.16 -0.69 49 -1.90 -7.31 3.98 0.93 2.83 0.43 -0.13 -2.19 55 -1.42 -2.82 0.18 0.93 2.88 1.24 0.46 -1.67 57 -2.40 -6.16 -0.75 1.37 2.05 1.83 2.40 -3.47 HR-LN 1 1.31 1.08 -0.01 -0.03 0.98 -0.64 -0.29 -0.93 3 -0.10 -1.89 -1.03 0.75 -1.53 0.83 1.32 5 0.77 -0.68 -1.05 0.49 1.55 1.41 0.78 -1.45 7 0.74 -1.59 -0.68 0.75 0.36 2.97 -0.35 -1.35 9 0.59 -0.65 -1.48 -1.75 -1.34 -1.00 0.10 -0.34 13 -1.17 0.35 -0.22 0.41 -0.09 -0.41 1.55 -0.65 15 -0.09 0.59 -0.71 -2.80 0.65 -0.23 0.26 -1.32 17 -0.57 -3.11 2.39 -0.03 1.29 0.67 2.52 19 -1.02 -3.70 0.09 0.57 -0.35 1.25 0.55 -3.33 21 -0.73 -1.13 -0.70 -0.81 0.42 0.34 0.10 0.51 23 0.00 1.61 0.31 0.58 -0.25 1.24 0.35 27 -0.49 -2.38 -0.43 1.01 2.53 0.87 -0.73 0.55 31 -0.63 -4.57 -0.16 0.58 0.87 -0.86 1.16 -3.86 33 -0.41 -1.39 1.95 -2.27 -1.08 0.07 1.58 -0.20 37 0.03 -0.99 0.35 -2.36 -1.76 0.56 -1.95 1.18 39 0.11 -4.12 -1.34 -2.10 -2.61 0.24 0.34 -0.60 45 -0.44 -0.43 4.92 1.02 1.96 -0.40 -1.07 -1.69 47 -0.80 -0.79 -1.00 2.22 0.63 -0.08 0.54 -1.80 51 -0.41 -4.47 1.81 2.90 2.10 0.95 0.57 53 -0.40 -7.30 -0.02 1.10 0.61 -2.32 0.95 -1.32 59 0.22 -3.06 0.25 -0.63 0.82 -0.71 -0.31 -0.27 61 0.52 -4.26 2.08 -1.35 -0.69 0.53 -0.77 -0.74 LR-LN 2 -0.05 0.89 -0.74 0.76 -0.09 -0.29 0.67 4 -0.76 -1.13 0.31 0.84 1.18 2.12 1.02 -1.12 6 1.12 0.16 -0.69 -0.65 -0.17 0.52 -0.50 -0.70 8 -0.32 -5.44 0.44 2.05 1.17 0.09 0.52 0.59 10 -0.57 0.56 0.87 1.62 -0.52 -1.12 0.95 -1.48 12 -1.28 -6.06 0.44 -0.11 0.64 0.90 0.63 14 0.01 0.36 -0.52 -0.29 0.58 1.04 0.23 -1.58 16 0.01 0.21 0.24 -2.02 0.11 0.96 -0.35 2.38 18 -0.19 -0.51 -0.02 -0.40 0.93 -0.15 0.73 -0.60 20 -1.13 -0.66 0.62 -0.97 0.72 -0.15 2.07 22 -0.36 0.74 0.54 -1.09 0.20 -0.57 0.15 -0.32 26 -0.43 0.11 3.59 1.10 2.84 3.72 -0.48 -0.30 28 0.76 0.67 -0.58 -1.52 0.25 1.21 -1.70 -0.51 32 0.42 -0.95 -2.02 -0.74 -1.99 -1.02 -1.71 0.63 34 -0.34 0.11 1.96 1.10 0.53 2.65 0.66 36 0.24 0.95 0.08 -0.54 -1.28 1.28 -0.82 0.65 38 1.05 1.48 -0.42 0.67 -0.35 -0.77 -0.31 -0.28 40 1.68 1.40 -1.33 -1.52 -2.80 -0.95 -0.74 0.86 44 0.30 -0.52 0.90 0.41 3.20 0.70 1.04 0.61 46 -0.13 -2.21 -0.64 1.02 1.47 0.29 0.34 48 0.95 1.37 0.99 0.67 -0.52 -0.03 0.74 -0.46 50 -0.29 0.15 2.55 1.36 -0.64 0.20 -1.98 -1.64 52 1.14 -0.30 0.09 -0.55 -0.87 -1.69 1.19 1.62 54 0.11 -1.36 -0.31 0.49 1.15 -1.05 1.53 -1.11 56 0.72 -0.29 4.61 0.93 -0.05 -0.69 -0.56 -1.82 58 0.28 1.54 -1.25 0.67 0.86 1.04 0.06 -0.45 60 0.58 -2.43 -0.79 -1.23 -0.26 -1.98 1.35 -2.29 62 1.21 -0.36 -0.73 -0.02 -0.27 0.10 -0.48 Note. Except for LITERACY, all Z-scores have been calculated relatively against the mean and SD of the ‘restricted’ LR-LN group; +/- 1.65 SD was chosen as the deviance criterion. For LITERACY Z-scores have been calculated relatively against population average; -1.3 SD was chosen as the deviance criterion. Note. For LITERACY, PHONOLOGY and the Phoneme Boundary Slope a higher Z-score indicates better processing; in contrast for all other measures a higher Z-score indicates reduced sensory sensitivity.
161
General conclusions and future perspectives
In this last chapter we discuss the theoretical, methodological and practical relevance
of the doctoral dissertation. Further, we consider some limitations of the study and provide
some general guidelines for future research.
Theoretical relevance
In this study we empirically evaluated some assumptions of the general magnocellular
theory of dyslexia. In short, the general magnocellular theory postulates that developmental
dyslexia is the consequence of a cross-modal sensory deficit in the processing of transient and
dynamic stimuli (e.g. Stein & Talcott, 1999; Stein & Walsh, 1997; Talcott et al., 2002; Talcott
& Witton, 2002; Witton et al., 1998). In the visual domain, this magnocellular deficit has been
hypothesized to provoke binocular instability, poor visual attention, poor visual search and
uncertainty about letter position, and as such it might affect the development of orthographic
skills and subsequent reading and spelling skills. In the auditory modality, an analogous
temporal processing deficit has been hypothesized to interfere with the accurate detection of
the acoustical changes in speech. Consequently, this speech perception problem causes a
cascade of effects, starting with the disruption of normal development of the phonological
system and resulting in problems learning to read and spell. Recently, it has also been
suggested that both the auditory and the visual deficit might be the result of one common
biological cause: a specific deficit in the neural pathways involved in the fast transmission
and processing of sensory information (Stein, 2001).
Based upon the theoretical assumptions of the general magnocellular theory two
straightforward and verifiable predictions can be formulated: (1) compared to controls,
subjects with dyslexia are hypothesized to demonstrate auditory and visual temporal
processing deficits, speech perception problems and orthographic/phonological problems, and
162
(2) we expect to observe significant relations between auditory temporal processing, speech
perception, phonological ability and literacy achievement on the one hand, and between visual
temporal processing, orthographic ability and literacy on the other. Moreover, assuming that
the same processes are involved in the development of normal as well as deficient literacy
ability, we do not only expect to observe these interrelations in dyslexic subjects, but also in
the general population as well. Numerous studies have been conducted exploring these
hypotheses. Most of them have been carried out on adult subjects, a minority on school-aged
children and very few on preschool children. However, besides the confusing aspect that they
often yielded very variable successes in demonstrating the hypothesized deficits and
interrelations, most of these studies only have a limited theoretical value since they do not
offer any indication about the direction of the studied relations. Indeed, the mere observation
of a sensory deficit in adult or school-aged dyslexic subjects does not provide insight into
whether the deficit is (a) at the basis of the reading problem, (b) a consequence of the reading
problem, or (c) a non-causally related marker of literacy problems.
In order to take a first step towards exploring the nature of the sensory deficits and
their relation with subsequent literacy development, we studied preschool children and
followed them up through early literacy development. Hence, the first data were collected in
kindergarten - before children received any formal reading instruction - and were
retrospectively reanalysed comparing children categorised in groups according to family risk
status and first grade literacy achievement.
Based upon a global inspection of the data, our results are quite favourable for the
general magnocellular theory. Apparently, children that become dyslexic1 do already
demonstrate significant preschool deficits in dynamic auditory and visual sensory processing,
speech perception, phonological ability and orthographic ability (viz. letter knowledge).
Apparently, dynamic auditory processing is related to speech perception, which on its turn is
related to phonological awareness and subsequent literacy development. In addition, dynamic
visual processing seems to be related to orthographic ability and subsequent reading and
writing development. Finally, integration of all the preschool and first grade data within one
structural model yields a fairly satisfying fit of the theoretically postulated interrelations.
Moreover, since the measures for sensory processing, speech perception and
phonological and orthographic ability have been administered before children received any
formal reading instruction, it is clear that the observed deficits do precede the literacy problem
and are not merely the result of lacking reading experience. Likewise, the same is true for the
163
observed interrelations: it is evident that these preschool measures have a predictive relation
with later literacy development. Therefore, our prospective study is one of the first to
convincingly demonstrate that the typically observed sensory deficits in dyslexia are not a
consequence of the literacy problem. Hence, option (b) of the abovementioned alternatives
can already be eliminated, leaving over the hypotheses (a) and (c), namely that the sensory
deficit is at the basis of the literacy problem versus that it is rather a non-causally related
marker of literacy problems. Currently, our data do not offer enough information to infer the
causal status of the sensory deficit and discern between these competing hypotheses.
Therefore, by further following up these children and integrating the current data within a
more elaborated developmental perspective, we hope to gain more insight into the nature of
the observed sensory–literacy relation.
So, from a global perspective our data appear to be quite in favour of the general
magnocellular theory. However, if we inspect the data in a more detailed way, we notice
several findings that are not in line with the theory. First, it is clear that not all dyslexic
subjects demonstrate a (temporal) sensory deficit. Of course, it could be argued that we did
not assess all crucial aspects of sensory processing or that our psychophysical tasks might
have lacked differentiating sensitivity. Yet, we would like to emphasize that the test battery
that we administered is among the broadest applied in this research tradition, including
several tasks that have proven to differentiate reliably between dyslexic and normal reading
subjects. The observation that not all dyslexic subjects demonstrate a sensory deficit indicates
that deficient sensory processing is not a necessary condition to develop dyslexia – a
conclusion that clearly conflicts with the assumption that the sensory problem would be at the
basis of the reading problem (see Bishop, Carlyon, Deeks & Bishop, 1999 for a similar
conclusion regarding specific language impairment).
Second, some normal reading subjects do also show sensory problems. This indicates
that deficient sensory processing is neither a sufficient condition to develop dyslexia (again
see Bishop et al., 1999 for a similar conclusion regarding specific language impairment).
Apparently, children growing up with quite severe low-level sensory problems can still
develop normal literacy skills. This could imply that either low-level sensory processing only
plays a limited role in the development of literacy ability or that some of these sensory-
impaired children can rely upon compensatory mechanisms to overcome their basic
impairments.
164
Third, it is not possible to demonstrate a consistent pattern of deficiencies across
sensory processing, speech perception and phonological abilities, neither for individual
subjects of the dyslexic group, nor for individual subjects of the normal reading groups.
Actually, even at a group level it is not possible to demonstrate such a consistent pattern. For
instance, if we focus upon the group results of the children at family risk for dyslexia that do
not (yet) present an actual literacy problem, it becomes clear that these children do present a
mild phonological deficit but they do not present the sensory deficit. This suggests that
whereas the occurrence of a phonological problem seems to be largely genetically determined
and partly irrespective of actual literacy achievement, the presence of a sensory problem
seems to be restricted to children demonstrating a real literacy deficit. Obviously, the
observation of this partial dissociation between phonological and sensory deficits poses a
problem for the general magnocellular theory, since it demonstrates that phonological
impairments are not necessarily secondary to sensory problems.
It is evident that the aforementioned objections question the validity of the general
magnocellular theory. However, it is important to note that the same objections also apply to
the phonological theory. Indeed, also with regard to phonology some dyslexic subjects present
relatively preserved skills, whereas some normal reading subjects demonstrate extremely
severe phonological deficits (even within the low risk group). This implies that even the more
established phonological theory is not able to fully explain the whole story of dyslexia.
Combining the results of the ‘global’ and the ‘in depth’ analyses, we can conclude that
preschool children that will ultimately become dyslexic present a reduced ability in dynamic
sensory processing, speech perception, letter knowledge and phonological ability. Moreover,
we observe a significant relation between dynamic auditory processing, speech perception,
phonological ability and subsequent literacy development on the one hand, and between
dynamic visual processing, orthographic ability and reading and spelling development on the
other. However, at the individual level, these deficits and relations tend to be not as absolute
as assumed, with many dyslexics showing isolated or only partially overlapping deficits. Even
more surprising is the observation that a substantial number of normal reading subjects also
presents quite severe preschool deficits in sensory processing, speech perception or
phonological processing and nevertheless is still able to develop normal literacy skills. Thus,
although deficient sensory processing and in particular deficient phonological processing
tends to be a risk factor to develop literacy problems, neither of both is a sufficient nor a
165
necessary condition to cause these problems. This suggests that whether one develops
dyslexia or not, still depends upon additional factors executing an aggravating or protective
influence. Actually, recent work has suggested that general language ability might be an
important moderator variable in the process of literacy achievement. For instance, Snowling
and colleagues demonstrated that children at risk for dyslexia because of speech difficulties
(Nathan, Stackhouse, Goulandris & Snowling, 2004) or because of a family history of
dyslexia (Snowling, Gallagher & Frith, 2003) can overcome their emerging phonological
problems and yet develop normal literacy skills under condition that they can rely upon strong
compensatory language abilities.
To conclude, although the general magnocellular theory provides a solid unifying
theoretical framework, it seems not able to fully account for all the various manifestations of
dyslexia. However, the same conclusion applies just as much to the more established
phonological theory. Bearing these conclusions in mind, one might recall the citation
mentioned at the beginning of this thesis:
“The greater becomes the volume of our sphere of knowledge,
the greater also becomes its surface of contact with the unknown.”
(J. Sageret)
Indeed, together with the growing awareness that there does not exist one uniform
manifestation of dyslexia, we also start to realise that no single all-embracing cause or theory
will be found. Reading and writing is a complex multifaceted activity. It involves a dynamic
interplay of multiple sensory and cognitive-linguistic processes, moderated by various
unspecified environmental or higher-order cognitive influences. Deficits at any level might
interfere with normal development. Comprehensive theories like the general magnocellular
theory are nevertheless important and necessary to guide and stimulate scientific research, but
it is an illusion to expect them to explain the full complexity of dyslexia.
166
Methodological relevance
Besides the theoretical relevance, an important merit of this dissertation has been the
development and composition of a series of instruments to test auditory and visual sensory
processing, speech perception, orthographic and phonological ability and early literacy
development in very young children.
In the first place, the adaptation of the psychophysical tests for the use in very young
children is noteworthy. In general, children - and especially preschoolers - are much harder to
test because of the cognitive demands and the dull repetitive structure of psychophysical
tasks. In order to make the relatively tedious psychophysical tests more interesting and child-
friendly, we integrated them in an interactive videogame (for the auditory test platform see
Laneau, Boets, Moonen, van Wieringen, & Wouters, 2005; Boets, Wouters, van Wieringen, &
Ghesquière, 2006). Because the most important condition for successful psychophysical
testing of young children is the transformation of the abstract meaningless signal into a
concrete and well known ‘daily life signal’, several stories were invented to give a
psychological and naturally fitting meaning to the particular stimuli at hand. Consequently,
every story was worked out through a collection of animated cartoons, with an introductory
movie introducing the specific target stimulus in a game-like manner and a closing movie
rounding off the experiment and rewarding the child for its achievement. Further, the use of a
touch screen and an extensive reinforcement system made the whole experiment really
interactive. Finally, the application of a child friendly forced-choice paradigm, the manual
control over stimulus presentation and the limited test duration by using optimised step sizes,
allowed us to control for fluctuations in attention in order to get reliable threshold estimates.
As a consequence of applying this concept of testing, the preschool children did not only
perform surprisingly accurately, but they also really enjoyed the psychophysical tasks2.
At the moment, the software for the auditory psychophysical testing of young children
is also used in other studies at the Laboratory for Experimental ORL (KULeuven) and in
foreign auditory research centres (e.g. K. Neijenhuis, Rotterdam). Similarly, the coherent
motion task is currently in use for testing children with cerebral visual impairments (P. Stiers,
Laboratory of Neuropsychology, KULeuven) and unilateral hemi neglect patients (K.
Verfaillie, Laboratory of Experimental Psychology, KULeuven).
In addition to the psychophysical tests, an extensive phonological and orthographic
test battery was composed based upon existing and self-developed material. Likewise, this
test battery is currently used in an analogous prospective study examining preschool children
167
with specific language impairment (E. Vandewalle, Laboratory for Experimental ORL,
KULeuven).
In this doctoral dissertation we introduced and applied some statistical procedures that
are relatively uncommon in the social sciences.
First, not only were the risk and control group matched at the individual level, data
were also analysed at the level of the matched pairs. Indeed, by entering pair number as a
random variable in a Mixed Model Analysis, the dependence at the subject level could be
taken into account without loosing additional degrees of freedom. Consequently, the statistical
differentiating power could be considerably increased. In this way, the Mixed Model Analysis
is statistically identical to a paired t-test, but it has some important advantages: (a) it is much
more robust in analysing semi-normally distributed data (Verbeke & Lesaffre, 1997), (b) it
allows repeated measures analysis, (c) it allows to incorporate covariates and weight
variables, (d) conclusions can be generalised to any group of matched pairs since they
represent a random selection of a larger population, and (e) it can handle missing or non-
balanced data. For a more detailed description of Mixed Model Analysis, the reader is
referred to Aerts, Geys, Molenberghs, & Ryan (2002), Littell et al. (1996) and Verbeke &
Molenberghs (1997).
A second statistical advantage concerns the two-stage analysis of the speech data.
Here, we first estimated the relevant parameters by fitting a logistic function to the raw data
and subsequently compared both groups on the resulting parameters. Moreover, in order to
take into account the variable quality of the fits, the inverse standard error of the estimated
parameters was added to the model as a weight variable. Actually, the addition of such a
weight variable is rather unique in the current research tradition and allows comparing groups
while correcting for differences in the reliability of the data.
Finally, the doctoral research project documents that the selection and longitudinal
follow-up of children at family risk is a valuable method to study the developmental pathway
of both normally and aberrantly developing children. In particular in situations where it is not
practicable to follow up a very numerous sample, this is a useful method to increase the
number of clinical subjects in the final research group.
168
Practical relevance
In this thesis we mainly analyzed and interpreted the data from a theoretical
perspective and less from a clinical point of view. However, our data could also offer valuable
information regarding the early prediction and even prevention of literacy problems. Although
the results of the deviance analysis demonstrated that currently none of the individual
preschool measures was able to predict future literacy failure in a satisfying way, the
predictive possibilities of the preschool measures might be enhanced by combining groups of
indicators. This would require a reanalysis of the data in order to determine adequate cut-off
criteria and possibly an estimation of appropriate weight coefficients. In the near future, we
plan to carry out such an analysis in order to evaluate the predictive and discriminative power
of our entire battery of preschool tests.
Limitations of the study
A possible limitation of the doctoral project is the relatively limited number of
subjects. Yet, this is a natural consequence of the specific design of the study with its highly
intensive and time-consuming psychophysical testing of carefully selected preschool children
attending schools scattered all over the Flanders. This restricted sample size might reduce the
generalizibility of the study. For instance, the conclusion that preschool ‘dyslexic’ children
present sensory, speech perception, phonological and orthographic problems, is ultimately
only based upon the data of nine subjects that eventually turned out to be dyslexic. Moreover,
although our definition of the dyslexic group was based upon the co-occurrence of family risk
and actual literacy delay, further follow up of these children is recommended to verify
whether they will also comply with the resistance criterion (Gersons-Wolfensberger &
Ruijssenaars, 1997; Fletcher, Francis, Morris & Lyon, 2005).
In particular, the restricted sample size might also affect the stability of the parameter
estimates in the SEM analyses (Kline, 1998). Therefore, caution is required while interpreting
the results. Furthermore, a larger population sample would have allowed us to run the SEM
analyses separately for the control versus clinical group (i.e. ‘dyslexia status’ would have
been entered into the model as a moderator variable). This highly interesting additional
analysis would have informed us whether the same causal path leading to adequate literacy
169
development in the normal reading group, in a similar way leads to impaired literacy
development in the dyslexic group.
Another limitation of the study is that, although we were able to demonstrate that
dyslexics’ sensory problems precede the literacy problem and are significantly related to it,
we were not yet able to discern whether they are at the basis of the reading problem or rather a
non-causally related marker of it. We acknowledge that the verification of this causality
matter is a tricky issue, in particular in a correlation study where no experimental intervention
is applied. Nevertheless, we hope that further follow up of these children at least could shed
some light upon the direction of the observed relations.
A minor drawback of the study is the exclusive use of the Raven Coloured Progressive
Matrices (Raven, Court, & Raven, 1984) as a measure for general intelligence. Since it can be
administered collectively and because of its excellent psychometric qualities (Evers, van
Vliet-Mulder, Groot, 2000), this test is frequently used in scientific studies. Although it
provides a good indication of general cognitive ability (Spearman’s g factor) and in particular
of non-verbal analogous reasoning, assessment of full-scale IQ (e.g. WPPSI-R, Vandersteene
& Bos, 1997) would have been preferential but was not feasible because of practical
constraints.
One might argue that a more serious limitation of the study is the lack of an instrument
to assess attention. Indeed, since psychophysical testing calls a lot on sustained attention, it
has been suggested that dyslexics’ sensory deficits could be a side effect of the generally
increased incidence of attention-related disorders (ADHD/ADD) in the reading disabled
population (Breier, Fletcher, Foorman, Klaas, & Gray, 2003; Hulslander et al., 2004; Willcutt
& Pennington, 2000). Although we were aware about this concern, we decided not to exclude
risk subjects on a basis of showing AD(H)D symptoms. Hence no information regarding
attention deficits was collected at the time of selecting subjects. This strategy was motivated
by the intention to maintain a representative sample of dyslexic subjects, including those with
the inherent attention problems.
Nevertheless, at the end of first grade for every child an ADHD rating scale assessing
inattentive, hyperactive and impulsive behaviour was completed by the teachers (Scholte &
van der Ploeg, 1998). Although psychometrically valid and reliable, this questionnaire only
gives a rudimentary and probably over-inclusive indication of ADHD-related problem
170
behaviour as perceived by the teacher. Results of the rating scale demonstrated that two of the
nine dyslexic subjects scored in the clinical range (i.e. subject 11 and 49). In the HR-LN as
well as in the LR-LN group there were another two subjects in the clinical range (i.e. subject
17 and 51, and 18 and 44 respectively), with another one in the LR-LD group (i.e. subject 42).
So far, scores upon the attention scale have not yet been entered as covariates in the analyses.
However, inspection of the individual data (see Manuscript 4, Appendix A) already reveals
that those subjects showing ADHD-related behaviour do not particularly fail in the
psychophysical tests. This indicates that the observed sensory deficits in subjects with
dyslexia are not merely a consequence of comorbid attention problems.
Finally, we would like to emphasise that we do not aim to make any claim regarding
the biological deficit that has been hypothesized to be at the basis of the general
magnocellular theory. We merely observe that psychophysical tasks measuring dynamic
auditory and visual processing differentiate between preschool dyslexic versus preschool
normal reading subjects, and that sensory thresholds are related to respectively phonological
and orthographic processing and future literacy development. Of course, it is only to the
extent that these tests provide a correct indication of underlying neurological magnocellular
functioning, that they also provide indirect evidence for a biological magnocellular deficit.
Within the visual modality, there is quite convincing evidence that coherent motion detection
is one of the best psychophysical markers to asses visual magnocellular processing3 (Britten,
Shalden, Newsome & Movshon, 1992), hence suggesting that our dyslexic subjects (as a
group) might indeed present a real neurological deficit in visual magnocellular processing. On
the contrary, in the auditory modality no analogous magnocellular neural pathway has yet
been identified, let alone that FM detection would have been demonstrated to be a good
indicator of it. As a consequence, we do not intend to make any inferences regarding a
possible neurological or biological auditory deficit. Moreover, we would not even
characterize the behavioural/psychophysical auditory deficit as specific to rapid or temporal
processing (as postulated by the general magnocellular theory), since it only seems to hold for
dynamic auditory processing. Therefore, a further investigation of the neurological
mechanisms involved in dynamic auditory processing is required to investigate whether a
specific neurological deficit can be demonstrated.
171
Future perspectives
As aforementioned, we hope to be able to follow up this group of children and
administer all sensory, speech perception, orthographic, phonological and literacy tasks
several times again, in order to further explore the exact nature of the relation between
sensory processing and literacy development. In particular, we aim to verify the direction of
this relation and whether it also comprises a causal aspect.
So far we only used behavioural measures in our research design. We have related the
cognitive deficits of dyslexia with low-level auditory and visual processing using
psychophysical threshold estimates. The use of ‘objective’ neurophysiologic measures could
be a next step in the search for further evidence of sensory processing deficits in persons with
dyslexia. In general, neurophysiologic measures have more discriminative power than
psychophysical measures (Bishop & McArthur, 2004). Objective neurophysiologic techniques
could not only enable us to verify the reliability of the psychophysical tests, but could also
offer us a better insight in the underlying neurological problems of dyslexia.
In the visual domain, our psychophysical results concerning coherent motion detection
could be evaluated using fMRI, especially of the cortical region V5/MT. Eden and colleagues
(1996) have shown less activity in that cortical region after stimulation with coherent moving
stimuli in dyslexic compared to normal reading adults. Demb, Boynton en Heeger (1997)
were even able to show a significant relationship between reading abilities of adults and the
magnitude of hemodynamic BOLD-responses in region V5/MT.
For the auditory domain we could make use of the recently developed technique of
auditory steady-state responses (ASSR). For many years, ASSR has been extensively studied
as a double-objective technique to estimate hearing thresholds. ASSR’s are periodic electrical
responses of the brain that can be evoked by amplitude- and/or frequency-modulated signals.
Depending on the modulation frequency, responses can originate from brainstem or auditory
cortex. In the Laboratory for Experimental ORL (KULeuven) this technique has been further
elaborated and adapted for newborns and infants at risk for hearing loss (Luts, Desloovere &
Wouters, 2006). The ASSR technique has a number of interesting features. The objectivity of
these measures and the possibility to use them to evaluate temporal aspects of auditory
processing (not possible with fMRI) makes this technique an interesting instrument for our
research interest.
172
To conclude, there is still a lot of work to do to unravel the mystery of dyslexia. And
while research progresses, not only answers but also many new questions will be encountered.
We hope this dissertation was able to answer some of these questions.
173
References
Aerts, M., Geys, H., Molenberghs, G., & Ryan, L. M. (2002). Topics in modelling clustered
data. London: Chapman and Hall.
Bishop, D. V., Carlyon, R. P., Deeks, J. M., & Bishop, S. J. (1999). Auditory temporal
processing impairment: neither necessary nor sufficient for causing language
impairment in children. Journal of Speech, Language and Hearing Research, 42(6),
1295-1310.
Bishop, D. V. M., & McArthur, G. M. (2004). Immature cortical responses to auditory stimuli
in specific language impairment: Evidence from ERP's to rapid tone sequences.
Developmental Science, 7(4), F11-F18.
Boets, B., Wouters, J., van Wieringen, A., & Ghesquière, P. (2006). Auditory temporal
information processing in preschool children at family risk for dyslexia: relations with
phonological abilities and developing literacy skills. Brain and Language, 97, 64-79.
Breier, J. I., Fletcher, J. M., Foorman, B. R., Klaas, P., & Gray, L. C. (2003). Auditory
temporal processing in children with specific reading disability with and without
attention deficit hyperactivity disorder. Journal of Speech, Language and Hearing
Research, 46, 31-42.
Britten, K. H., Shalden, M. N., Newsome, W. T., & Movshon, J. A. (1992). The analysis of
visual motion: A comparison of neuronal and psychophysical performance. Journal of
Neuroscience, 12, 4745-4765.
Demb, J., Boynton, G., & Heeger, G. (1997). Brain activity in visual cortex predicts
individual differences in reading performance. Proceedings of the National Academy
of Sciences, USA, 94, 13363-13366.
Eden, G., Van Meter, J., Rumsey, J., Maisog, J., Woods, R., & Zeffiro, T. (1996). Abnormal
processing of visual motion in dyslexia revealed by functional brain imaging. Nature,
382, 66-69.
Evers, A., van Vliet-Mulder, J. C., & Groot, C. (2000). Documentatie van Tests en
Testresearch in Nederland - 2000. Deel I Testbeschrijvingen. Deel II Testresearch.
174
Assen: Van Gorcum.
Fletcher, J. M., Francis, D. J., Morris, R. D., & Lyon, G. R. (2005). Evidence-based
assessment of learning disabilities in children and adolescents. Journal of Clinical
Child and Adolescent Psychology, 34(3), 506-522.
Gersons-Wolfensberger, D. C. M., & Ruijssenaars, W. A. J. J. M. (1997). Definition and
treatment of dyslexia: A report by the Committee on Dyslexia of the Health Council of
the Netherlands. Journal of Learning Disabilities, 30(2), 209-213.
Hulslander, J., Talcott, J., Witton, C., DeFries, J., Pennington, B., Wadsworth, S., Willcutt, E.,
& Olson, R. (2004). Sensory processing, reading, IQ and attention. Journal of
Experimental Child Psychology, 88, 274-295.
Kline, R. B. (1998). Principles and practice of Structural Equation Modeling. New York: The
Guilford Press.
Laneau, J., Boets, B., Moonen, M., van Wieringen, A., & Wouters, J. (2005). A flexible
auditory research platform using acoustic or electric stimuli for adults and young
children. Journal of Neuroscience Methods, 142(1), 131-136.
Littell, R. C., Milliken, G. A., Stroup, W. W., & Wolfinger, R. D. (1996). SAS system for
mixed models. Cary, NC: Sas Institute INC.
Luts, H., Desloovere, C., & Wouters, J. (2006). Clinical application of dichotic multiple-
stimulus auditory steady-state responses in high-risk newborns and young children.
Audiology & Neuro-Otology, 11(1), 24-37.
Nathan, L., Stackhouse, J., Goulandris, N., & Snowling, M. J. (2004). The development of
early literacy skills among children with speech difficulties: a test of the "critical age
hypothesis". Journal of Speech, Language and Hearing Research, 47(2), 377-391.
Raven, J. C., Court, J. H., & Raven, J. (1984). Manual for Raven's Progressive Matrices and
Vocabulary Scales. London: Lewis.
Scholte, E. M., & van der Ploeg, J. D. (1998). ADHD Vragenlijst (AVL). Handleiding. Lisse:
Swets & Zeitlinger.
175
Skottun, B. C., & Skoyles, J. R. (in press). Is coherent motion an appropriate test for
magnocellular sensitivity? Brain and Cognition.
Snowling, M. J., Gallagher, A., & Frith, U. (2003). Family risk of dyslexia is continuous:
individual differences in the precursors of reading skill. Child Development, 74(2),
358-373.
Stein, J. (2001). The magnocellular theory of developmental dyslexia. Dyslexia, 7, 12-36.
Stein, J., & Talcott, J. (1999). Impaired neuronal timing in developmental dyslexia. The
magnocellular hypothesis. Dyslexia, 5, 59-77.
Stein, J., & Walsh, V. (1997). To see but not to read; the magnocellular theory of dyslexia.
Trends in Neuroscience, 20(4), 147-152.
Talcott, J. B., Witton, C., Hebb, G. S., Stoodley, C. J., Westwood, E. A., France, S. J.,
Hansen, P. C., & Stein, J. F. (2002). On the relationship between dynamic visual and
auditory processing and literacy skills; results from a large primary-school study.
Dyslexia, 8(4), 204-225.
Talcott, J. B., & Witton, C. (2002). A sensory linguistic approach to the development of
normal and impaired reading skills. In E. Witruk, A. Friederici, & T. Lachmann
(Eds.), Neuropsychology and cognition series. Basic functions of language and
language disorders. Dordrecht, Netherlands: Kluwer Academic Publishers.
Vandersteene, G., & Bos, A. (1997). WPPSI-R: Wechsler Preschool and Primary Scale of
Intelligence - Revised. Vlaams-Nederlandse aanpassing, voorlopige versie. Lisse:
Swets & Zeitlinger.
Verbeke, G., & Lesaffre, E. (1997). The effect of misspecifying the random-effects
distribution in linear mixed models for longitudinal data. Computational Statistics and
Data Analysis, 23, 541-556.
Verbeke, G., & Molenberghs, G. (1997). Linear Mixed Models in practice: A SAS oriented
approach. Lecture notes in statistics, 126. New York: Springer-Verlag.
Willcutt, E. G., & Pennington, B. F. (2000). Comorbidity of reading disability and attention
deficit/hyperactivity disorder: differences by gender and subtype. Journal of Learning
176
Disabilities, 33, 179-191.
Witton, C., Talcott, J. B., Hansen, P. C., Richardson, A. J., Griffiths, T. D., Rees, A., Stein, J.
F., & Green, G. G. (1998). Sensitivity to dynamic auditory and visual stimuli predicts
nonword reading ability in both dyslexic and normal readers. Current Biology, 8(14),
791-797.
177
Notes
1 For the sake of clarity, in this conclusion we will refer to dyslexic versus non-dyslexic
children, although we recognize that in our study the diagnosis did not comply with the
‘resistence to intervention’ criterion (Fletcher, Francis, Morris & Lyon, 2005; Gersons-
Wolfensberger & Ruijssenaars, 1997). 2 For the visual as well as for the auditory psychophysical tests, reliability was fairly
satisfying (based upon inspection of the individual staircases and intra-subject correlations).
However, for the categorical perception task, reliability appeared to be less. Yet, we do not
think this is merely the result of the exigent cognitive demands of the applied procedure, but it
might also be an intrinsic consequence of measuring a premature and still emerging
perceptual skill. 3 However, see Skottun and Skoyles (in press) for some highly recent counterevidence
regarding the validity of coherent motion detection as an indicator of visual magnocellular
processing.