050 lyytin (514-546) 8/18/06 12:17 PM Page 514

M E R R I L L -P A L M E R Q U A R T E R L Y, V O L . 52, N O . 3

Trajectories of Reading Development: A Follow-up

From Birth to School Age of Children With and
Without Risk for Dyslexia
Heikki Lyytinen, Jane Erskine, Asko Tolvanen, Minna Torppa, Anna-Maija
Poikkeus, and Paula Lyytinen, University of Jyväskylä, Finland

In order to understand why some children are vulnerable to difficulties in their

language development and their acquisition of reading skill, the Jyväskylä Lon-
gitudinal Study of Dyslexia followed 200 Finnish children from birth to school
age. Half of these children had a family history of reading problems and were
considered at risk for dyslexia; the other half were not at risk. A novel analysis,
mixture modeling, revealed four subgroups with differential developmental tra-
jectories to early reading. The subgroups who showed either a “dysfluent trajec-
tory” (n = 12; 11 at risk vs. 1 control) or a “declining trajectory” (n = 35; 24 vs.
11) contained more children with familial risk for dyslexia. The subgroup show-
ing an “unexpected trajectory” contained equal numbers of at-risk and non-risk
children (n = 67; 33 vs. 34). The subgroup displaying a “typical trajectory” (n =
85, 38 vs. 47) contained more children born without dyslexia risk. This differen-
tial development of skills revealed that there are at least three troubled routes
along which a child may ultimately encounter difficulties in reading acquisition.
The most explicit routes are characterized by problems in either phonological
awareness, naming speed, or letter knowledge—problems that increase in sever-
ity with age.

Heikki Lyytinen, Jane Erskine, Asko Tolvanen, Minna Torppa, Anna-Maija Poikkeus, and
Paula Lyytinen, Department of Psychology and Child Research Center.
The Jyväskylä Longitudinal Study of Dyslexia (JLD) belongs to the Finnish Center of Excel-
lence Program (2000–2005) and has been supported by the Academy of Finland, Niilo Mäki Insti-
tute, and University of Jyväskylä. We would like to thank the families in the study. Thanks also to
the reviewers and editors of this journal for their valuable comments on earlier versions of the text.
Correspondence concerning this article should be directed to Heikki Lyytinen, University of
Jyväskylä, PO Box 35; Fin-40014, University of Jyväskylä, Finland. E-mail: Heikki.Lyytinen
@psyka.jyu.fi.

Merrill-Palmer Quarterly, July 2006, Vol. 52, No. 3, pp. 514–546. Copyright © 2006 by Wayne
State University Press, Detroit, MI 48201.

514
050 lyytin (514-546) 8/18/06 12:17 PM Page 515

Dyslexia and Trajectories of Reading Development 515

One of the best ways to understand developmental disorders is through a

prospective follow-up study, especially if it is initiated before the environ-
ment can exert its effect on the emerging developmental routes. The
Jyväskylä Longitudinal Study of Dyslexia (JLD) is one such prospective
follow-up from birth to school age of 200 Finnish children, half of whom
were at familial risk for dyslexia. Over a period of 9 years, the children
underwent a plethora of assessments drawing on all psychological disci-
plines. These included measures assessing early auditory cognition (with
event-related potentials, or ERPs), early language (expressive and recep-
tive), cognitive abilities, and recognized correlates to literacy, including
phonological awareness, naming speed, letter naming, and early decoding
skills. Typically, Finnish children attend preschool or kindergarten at age 6,
with formal schooling and reading instruction beginning at age 7 when they
enter first grade. The youngest of the JLD children were about to enter third
grade, so we had access to data covering birth to second grade. We drew on
these data to identify factors in the development of the children’s language
skills that predict the children’s ease in reading acquisition and their prob-
lems in foundation-level reading.
We now know that dyslexia is affected by genes, and recently it has
been connected even to a single gene (Taipale et al., 2003). It is also clear
from the earlier literature (Scarborough, 1990; Snowling, Gallagher, &
Frith, 2003) that two domains are integral to dyslexia: the general language
domain, including speech and language development, and the more literacy-
specific language domain, including the phonological domain. We believe
they should be considered separately to ascertain the extent to which they
are distinct with regard to reading. During the first years of life, isolating dif-
ferent courses in terms of language development is complex. Nonetheless,
even relatively undifferentiated indices of the development of early lan-
guage, such as measures of early speech comprehension and production,
create a basis for the prediction of language skills that emerge later.
During their exposure to spoken language, children’s evolutionarily
more universal sensitivity to sounds becomes redundant as they learn to
home in on those aspects of language that discriminate meaning within their
particular language environment (Kuhl, Williams, Lacerda, Stevens, & Lind-
blom, 1992). Such results show how interest in and implicit attention to
aspects of sounds drills the brain’s sensitivity to those aspects in a vein simi-
lar to how Gibson (1970) describes the development of perception in general.
Elsewhere (for a review of results, see H. Lyytinen, Leppänen,
Richardson, & Guttorm, 2003) we have reviewed our data connecting
speech sound processing in the brain and accuracy of speech perception to
familial risk for dyslexia. The brains of the JLD at-risk infants differed
050 lyytin (514-546) 8/18/06 12:17 PM Page 516

516 Merrill-Palmer Quarterly

from controls at birth in terms of the processing of speech sounds (Guttorm,
Leppänen, Richardson, & Lyytinen, 2001; Guttorm, Leppänen, Tolvanen,
& Lyytinen, 2003). In addition, preparedness to develop accuracy in the
sensitivity to semantically distinctive features (i.e., categorical perception)
also varied between the JLD infants with and without risk, as reflected in
brain-event-related potential recordings (Leppänen, Eklund, & Lyytinen,
1997) and in behavioral measurements using a head-turn paradigm
(Richardson, 1998). Importantly, and as one of the catalysts to the current
pursuit, these speech-processing differences documented immediately after
birth and the brain and behavioral indices collected at the age of 6 months
were also found to correlate to later language development and to reading
acquisition (Guttorm et al., 2005; for a review of the latest psychophysio-
logical findings from the JLD see H. Lyytinen, Guttorm, et al., 2005).
The most prominent subdomain of language development relative to
routes to reading acquisition is the phonological domain (Vellutino,
Fletcher, Snowling, & Scanlon, 2004). Studies of at-risk children have
shown deficits in phonological awareness (Scarborough, 1990), and regres-
sion techniques have revealed that the phonological processing abilities of
kindergartners and preschoolers can predict later reading ability and, ulti-
mately, reading disability (Lundberg, 1994; Wagner, Torgesen, & Laughon,
1993). Intervention studies with at-risk children have also shown that
phonologically based reading intervention is critical to skilled reading
(Shaywitz et al., 2004).
Our early predictive signs based on speech processing and differentiat-
ing children with and without familial risk for reading difficulties reveal an
atypicality that compromises phonological accuracy and possibly affects
language development in general (Guttorm et al., 2005; see also H. Lyyti-
nen, Guttorm, et al., 2005, for review and the most recent evidence). Per-
ceptual problems may delay the acquisition of explicit phonemic awareness
required for reading (see, e.g., Goswami & Bryant, 1990, for a review). At
the same time, insensitivity reflected in speech processing may also affect
language development in a more general way. For instance, it is assumed
that optimal vocabulary growth is facilitated by the combination of a rich
linguistic environment (diversity of input) and intact abilities for continu-
ous phonological restructuring of the representations forming one’s vocab-
ulary (e.g. Fowler, 1991; Hoff & Naigles, 2002; Metsälä & Walley, 1998).
Research demonstrating the existence of multiple, autonomous routes
to compromised reading acquisition is relatively sparse. Unitary claims
(Snowling, 1998; Stanovich, 1990), for the most part, have focused on
“core” features such as a phonological deficit as underlying reading prob-
lems, while previous attempts to differentiate the characteristic error pat-
050 lyytin (514-546) 8/18/06 12:17 PM Page 517

Dyslexia and Trajectories of Reading Development 517

terns of dyslexia into distinct phonological (Beauvois & Derouesné, 1979)

or surface (Coltheart, Masterson, Byng, Prior, & Riddoch, 1983; Marshall
& Newcombe, 1973) subtypes have eluded replication. Moreover, in con-
sistent languages such as Finnish,1 surface-type errors are redundant in the
language. In consistent orthographies with low levels of ambiguity in the
relationship between letters and sounds, the focus leans toward dysfluent
reading (Wimmer, 1993) as the primary characteristic, with a more second-
ary emphasis placed on errors. The degree of individual variation and over-
lap with regard to the deficits displayed by dyslexic individuals has made
the categorical separation of distinct patterns difficult (Leinonen et al.,
2001; see Erskine & Seymour, 2005, for their Cognitive Mosaic Model).
Alongside this individual variation, the preparatory skills for a multi-
factorial ability such as reading are unlikely to vary dichotomously but con-
tinuously. It seems apparent, therefore, that different constellations of
specific weaknesses form critical “drops” in the skill profiles that may com-
promise reading acquisition and/or the automatization of reading skill and
that such characteristics may form differentiable developmental profiles
that we are able to detect using the present methodology.

Rationale for Selection of Critical Developmental Skill Domains

A comprehensive developmental data set spanning 6 years was used for the
modeling. Although we acknowledge that basic processes, such as speech
perception, create the foundation for the later phonological development
(e.g., Bradley & Bryant, 1983), in the present context we restricted our analy-
sis to the use of our early behavioral measures. Seven skill domains were
produced from these early measures, including (1) receptive and (2) expres-
sive language skills, (3) inflectional morphology skills (general language
domain), (4) memory, (5) retrieving words efficiently from memory (naming
speed), (6) letter knowledge, and (7) phonological awareness skills. The
development of knowledge relating to visual language (e.g., letter knowl-
edge) is an integral part of the foundation skills necessary for reading acquisi-
tion and thus constitutes a necessary component of the analyses.
It is well documented that language begins to differ between children
with and without risk for dyslexia prior to 3 years of age (Locke et al.,
1997) and not only continues to do so but also correlates to later reading

1. Finnish is a highly orthographically consistent language whereby almost all graphemes (in
this case, single letters) match consistently with a single phoneme and vice versa. This stands in
marked contrast to more inconsistent languages such as English, where the connections between
grapheme and phoneme can entail a one-on-many and many-on-one relationship.
050 lyytin (514-546) 8/18/06 12:17 PM Page 518

518 Merrill-Palmer Quarterly

beyond this age (H. Lyytinen, Aro, Eklund, et al., 2004; Scarborough,
1990). The earliest significant predictive correlations from our data are
from age 2 years onward (H. Lyytinen, Aro, Holopainen, et al., 2006). For
the present modeling we used composite scores of both receptive and
expressive language from ages 1 to 5.5 years, capitalizing on parental rat-
ings of early language (MacArthur Communicative Development Inven-
tory, CDI; Fenson et al., 1994) and well-known tests (e.g., Reynell
Developmental Language Scales, Reynell & Huntley, 1987; Peabody Pic-
ture Vocabulary Test–R, Dunn & Dunn, 1981; Boston Naming Test,
Kaplan, Goodglass, & Weintraub, 1983).
Our third developmental domain comprised inflectional morphology.
The Finnish language contains fusional agglutinative morphology with
very rich and complex sequential inflections and frequent stem variations.
A single word can have hundreds or even thousands of variants (Karlsson,
1999). Many of these variations are differentiated by single phonemes (e.g.,
kodissa [at home]; kodista [from home]). Thus, a child who is able to inflect
Finnish accurately must have a proficient implicit ability to manipulate
small phonological units. This makes morphological skills highly interest-
ing in our present modeling, as these skills must be mastered (and are easily
available for assessment) before the higher-level phonological awareness
skills can emerge (i.e., explicit manipulation of phoneme level units). In
contrast to English, morphological awareness plays no explicit role in the
pronunciation of the Finnish written language. The Finnish writing system
is entirely phonological. Every word can be read accurately through knowl-
edge of the letter-sounds and by assembling these in a row as dictated by
letters in the written word. However, the morphology affects the average
length of Finnish words, which are much longer than in most non-
agglutinative languages such as English. Furthermore, as fluency of read-
ing (necessary for reading continuous text due to working memory limits)
is the aspect of reading that is most often compromised among Finnish indi-
viduals with dyslexia (see, e.g., Holopainen, Ahonen, & Lyytinen, 2001),
morphology, albeit indirectly, is likely to affect reading. That morphologi-
cal skills are associated with reading development in Finnish is also docu-
mented by data from the JLD (H. Lyytinen, Aro, Holopainen, et al., 2006).
We included verbal short-term memory as the fourth developmental
skill domain in our modeling. It has been awarded a central role in reading
research, possibly because of contentions concerning its overlap with
phonological skills (e.g., Pennington, 1991), that is, the requirement of the
phonological loop to maintain pronounceable material in memory for short
durations. Phonological memory is required for the development of letter
knowledge (DeJong & Olson, 2004), which, in turn, is highly predictive of
050 lyytin (514-546) 8/18/06 12:17 PM Page 519

Dyslexia and Trajectories of Reading Development 519

reading acquisition (e.g., Wagner, Torgesen, & Rashotte, 1994; observed

also in the JLD, see H. Lyytinen, Aro, Holopainen, et al., 2006; H. Lyyti-
nen, Ronimus, Alanko, Taanila, & Poikkeus, submitted). In addition, Gath-
ercole and Baddeley (1997) have argued that phonological memory
mediates long-term learning as representations of new phonological forms
provide a basis for more permanent representations of new vocabulary.
Our fifth skill domain, retrieval of speech sounds as a continuous
sequence as assessed by rapid serial naming, shares central cognitive com-
ponents with reading (Wolf, 1984). Many identify naming speed as a sub-
domain of phonological skills (e.g. Vellutino et al., 2004), while others
consider the processes measured by rapid serial naming (assessed using
RAN, Rapid Automatized Naming Test, by Denckla & Rudel, 1976) tasks
to be independent of phonological processing (Manis, Seidenberg, & Doi,
1999; Wolf & Bowers, 1999).
Especially in consistent orthographies, high predictive associations
have been observed between reading and serial naming speed (Holopainen
et al., 2001; Korhonen, 1995; Wimmer, Mayringer, & Landerl, 1998).
Because gains through intervention appear to be more difficult to achieve in
terms of the fluency reflected in naming speed and reading rate rather than
in accuracy of phonological skills (DeJong & Vrielink, 2004; Thaler,
Ebner, Wimmer, & Landerl, 2004), early detection of difficulties in fluency
is all the more important.
Letter knowledge is known to be one of the most important early corre-
lates of reading acquisition (e.g., Adams, 1990; Byrne, 1998; Elbro, Borstrøm,
& Petersen, 1998; Gallagher, Frith, & Snowling, 2000; Lonigan, Burgess, &
Anthony, 2000; H. Lyytinen et al., submitted; Pennington & Lefly, 2001; Scar-
borough, 2001; Snow, Burns, & Griffin, 1998). Thus, our sixth developmental
skill domain comprised the yearly assessments of the children’s letter knowl-
edge. Letter names direct a child’s attention toward phonemes and contribute
critically to the development of phonemic awareness.
In the highly literate Finnish culture, practically all children learn some
letters long before entering school. At the time of data collection for the
present study, the practice in Finnish kindergartens has been to start intro-
ducing letters 1 year before formal schooling. Consequently, the sponta-
neous acquisition of foundation-level decoding skills has increased to
about one third of those entering the first-grade population. Interestingly,
our groups of children with and without risk differ from each other in their
letter knowledge at all ages of assessments, even though the participating
families have been amply exposed to the idea that a rich home literacy envi-
ronment and early shared reading is one of the best ways to afford the chil-
dren maximum support in their orientation to written language, and parents
050 lyytin (514-546) 8/18/06 12:17 PM Page 520

520 Merrill-Palmer Quarterly

in the two groups do not show differences in the degree to which they have
followed these guidelines in their homes (Torppa et al., submitted).
The complex interplay between children’s literacy environment and their
differential liability to environmental effects at different ages is interestingly
visible in our data. Laakso, Poikkeus, Eklund, and Lyytinen (2004) showed
that, at 14 months of age, infant shared reading interest (reflected in their
observed ability to sustain interest in shared reading with their parent) pre-
dicted global language skills and letter knowledge at 3.5 years of age only
among children in the control group. On the other hand, Torppa et al. (submit-
ted) found recently that, at later ages, the amount of shared reading predicted
vocabulary in both groups but that children’s interest in literacy versus other
activities (as reported by parents at 4, 5, and 6 years of age) is positively asso-
ciated with the growth of their phonological awareness skills, but only among
children in the at-risk group. Furthermore, Torppa et al.’s analyses showed
that vocabulary predicted children’s reading interest in the at-risk group only.
We feel that this at first seemingly contradictory result suggests that one
needs to pay careful attention to person and context factors, as well as the
time periods in which the proximal processes take place (cf. Bronfenbrenner
& Morris, 1998). It could be speculated that, in a control population (i.e., in
children without heightened risk for language problems), variation in chil-
dren’s early tendency to be drawn toward and maintain interest in joint lan-
guage- and literacy-related activities is reflected in their vocabulary growth.
At a later age, the variation in the amount of shared reading, on the other
hand, has an effect on vocabulary growth in both groups. However, the asso-
ciation between environmental factors, children’s interest, and skill develop-
ment is reciprocal in nature, and children’s skills may also affect how
interested they become in reading activities and how much parents read with
them. Thus, it can be speculated that for some children in the at-risk group
the assumed core deficits or related underlying processes in phonological
processing may interfere with or prevent the building of interest in shared
reading activities and also hinder the accumulation of vocabulary. This inter-
pretation is in line with suggestions that variation in phonological awareness
may be more strongly associated with genetic vulnerability/strengths than
with environmental variation.
The literature sees the ability to attend to the sub-word segments and to
manipulate them in the mind (i.e., phonological awareness) as the core pre-
requisite skill that underlies the child’s ability to grasp the principle of
grapheme-phoneme correspondence and consequently to decode text. We
have assessed phonological awareness (skill domain seven) on an annual
basis during the ages 3.5 to 6.5 years using both traditional procedures and
assessment in a computer context using animations. In the modeling, we
050 lyytin (514-546) 8/18/06 12:17 PM Page 521

Dyslexia and Trajectories of Reading Development 521

used only composite scores of phonological awareness that are highly pre-
dictive of early reading, as shown by our recent paper (Puolakanaho,
Poikkeus, Ahonen, Tolvanen, & Lyytinen, 2004).
It should be noted that the dependent measure of our present analyses,
the literacy composite, formed as an aggregate of early reading and spelling
acquisition (which are highly correlated in Finnish due to symmetrical con-
sistency; see Aro, 2006), reflects the children’s progress in foundation-level
reading in first and second grade. In the composite we used, the accuracy of
decoding (and spelling) naturally receives a much stronger emphasis than
the fluency of reading, which is not yet observable at its full strength at this
grade level. Full accuracy is reached relatively early in consistent orthogra-
phies, and this means the speed (fluency) issue comes to the fore relatively
soon after the start of reading instruction and continually plays a more cen-
tral role in the assessment of reading skill thereafter. However, it should be
noted that, because of the age of interest in the present analyses, issues of
reading comprehension were not included in the composite score of read-
ing. Some preliminary data on reading comprehension will, however, be
provided for the description of the outcomes of developmental routes
extracted from the pre-reading data.
Why, instead of waiting longer, should such an emphasis be placed on
the early stage of reading and spelling? We believe that children define
themselves as learners during the very first years at school, and the conse-
quences of failure are far-reaching. A Finnish pupil who does not learn to
decode accurately by the end of the first semester of first grade may almost
unavoidably perceive him- or herself as different from others. This is
because, in the Finnish schools, almost everyone progresses to highly accu-
rate decoding skills within the first 4 months of reading instruction at the
latest (Seymour, Aro, & Erskine, 2003). The detrimental effects of compar-
ing oneself unfavorably to others can have a general impact on one’s learn-
ing strategies (mostly toward avoidance of challenging tasks) even as early
as during first grade (Poskiparta, Niemi, Lepola, Ahtola, & Laine, 2003).
With the aid of advanced statistical tools (mixture modeling), we carried
out a novel analysis that combines person-centered and variable-centered
analyses as it yields both continuous variables and a subcategorization of the
participants. As both theoretical knowledge and the empirical evidence sug-
gest strong correlations among the included language and literacy skill
domains (receptive language, expressive language, morphology, memory,
rapid naming, phonological awareness, and letter knowledge) and reading
skills, we searched for a single, common factor that captures the shared vari-
ances of all the early measured language and literacy skills and the later-
measured components of reading. Simultaneously with the modeling of the
050 lyytin (514-546) 8/18/06 12:17 PM Page 522

522 Merrill-Palmer Quarterly

common factor, we explored the children’s individual trajectories from early
language and literacy skills toward reading skill and identified groups of
children with different types of developmental trajectories in our data set.
By incorporating our numerous early language and literacy-based measures
as composite variables into the mixture modeling technique, we hypothesize
that not one, not two, but several developmental routes to early reading diffi-
culties may be identified that differ by consequences to reading. Delays or
deficits in the early development of phonological awareness and letter
knowledge may compromise reading accuracy, whereas those in naming
dysfluency may be more closely associated not only with acquisition of the
foundation-level reading skills but also with the early automatization of
reading. We are also interested to learn whether there are children among
those with and without familial risk for dyslexia who do not manifest read-
ing problems but whose early development contains signs of potential risk
mentioned above—and vice versa, whether there are children whose reading
acquisition starts abnormally but who do not show characteristics of a devel-
opmental risk and/or belong to the at-risk group.

Method
Participants
Data were drawn from the prospective Jyväskylä Longitudinal Study of
Dyslexia (JLD). A total of 214 Finnish families joined the study prior to the
birth of their children. Half of the participating families involved a parent
who made a report of literacy difficulties and also reported similar prob-
lems among his or her immediate relatives (the children from these families
are referred to as the at-risk children). The remaining half of the children
(control children) came from families with no report of literacy difficulties
among first- or second-degree relatives. In addition, the parents were
screened and assigned to either group according to a three-step reading-
related and cognitive screening process (for further details, including
parental education status, see Leinonen et al., 2001).
This study reports the data from 199 children (106 at risk, 93 controls
for whom whole data were available) and extends from 12 months of age to
the beginning of second grade.

Measures
As described above, a theoretical basis underpinned the selection of the
seven core skill domains and age points of interest. Identification of the key
050 lyytin (514-546) 8/18/06 12:17 PM Page 523

Dyslexia and Trajectories of Reading Development 523

measures in each skill area evolved from empirical analysis of the data. To
ensure high reliability of the measures, composite means of separate meas-
ures were formed where possible. Table 1 details the seven developmental
skill domains, assessment ages, standardized alpha produced for composite
scores, and tasks included in the analysis.
Receptive language was assessed on 6 occasions between the ages of
12 months and 5 years. Early language tasks adapted from the MacArthur
Communicative Development Inventory (CDI; Fenson et al., 1994) for the
Finnish language (P. Lyytinen, 1999) relied upon structured parental report
using checklists whereby parents noted the total number of words compre-
hended by their child. Measures of expressive language were derived on 7
occasions between 12 months and 5.5 years of age, again with the earlier
measures involving parental checklist report of vocabulary production and
maximum sentence length (mean number of morphemes uttered; for details
of the assessment of vocabulary and inflectional skills using CDI, see
P. Lyytinen & Lyytinen, 2004). Morphological knowledge, including the
ability to inflect adjectives, nouns, and verbs, was assessed 3 times between
2.5, 3.5, and 5.5 years. The assessment of memory relied upon traditional
paradigms involving digits but was also expanded to include syllables, non-
words, and sentence-length items in the recall. Memory was assessed
between the ages of 3.5 and 6.5 years on 4 occasions. Traditional paradigms
(Denckla & Rudel, 1974, 1976) for rapid serial naming were also used,
including letters, objects, numbers, and colors. Mixed-stimulus paradigms
were not included, although the matrix size was increased in accordance
with age. Three blocks of rapid naming assessments took place between the
ages of 3.5 and 6.5 years. Letter knowledge and associated letter recogni-
tion tasks were implemented on 5 occasions between the ages of 3.5 and 6.5
years. An extensive battery of tests of phonological awareness was imple-
mented between the ages of 3.5 and 6.5 years. Tasks involved the manipula-
tion and production of sub-word-level units (syllables and phonemes). At
the age of 5 years, verbal and performance intelligence quotients were also
obtained with a short-form administration of the Wechsler Preschool and
Primary Scales of Intelligence (WPPSI-R; Wechsler, 1989).
In addition to the seven skill domains, a composite of the children’s
early reading and spelling achievement during their first school year and
the beginning of the second school year (mid-fall assessment of second
grade) was included in the comparison (see Table 2). The composite was
formed by calculating a grand mean of scores. The battery of tests included
computerized tests (Cognitive Workshop, Universities of Dundee and
Jyväskylä; see Erskine & Seymour, 2005, for extensive implementation),
tests with national norms (Lindeman, 2000), and specially designed group
Table 1. Skill Domains, Age of Assessment, Standardized Alpha for Composite Scores, and Tasks

Composite mean
Age(s) of score standardized
524

Skill domain assessment alpha Task(s)

050 lyytin (514-546)

1. Receptive 12m Word comprehension, Finnish adaptation (CDI; P. Lyytinen,

language 1999) of MacArthur Communicative Development
Inventory (CDI; Fenson et al., 1994).
Word comprehension, CDI.
8/18/06

14m .78 Verbal comprehension, Reynell Developmental Language

18m Scales (RDLS; Reynell & Huntley, 1987)
2.5y Verbal comprehension, RDLS
3.5y Receptive language, Peabody Picture Vocabulary (PPVT-R;
12:17 PM

Dunn & Dunn, 1981)

5.0 y Receptive language, PPVT-R
2. Expressive 12m Vocabulary production, CDI
language 14m Vocabulary production, CDI
18m .93 Vocabulary production, CDI
Page 524

Mastery of inflections, CDI

Maximum sentence length (mean number of morphemes
from 3 longest utterances), CDI
Expressive language items, RDLS
2.0y Vocabulary production, CDI
Mastery of inflections, CDI
Maximum sentence length, CDI
2.5y .92 Vocabulary production, CDI
Mastery of inflections, CDI
Maximum sentence length, CDI
Expressive language, RDLS
3.5y Expressive language, Boston Naming Test (BNT; Kaplan,
Goodglass, & Weintraub, 1983)
Merrill-Palmer Quarterly

5.5y Expressive language, BNT

3. Morphology 2.5y Comprehension of adjectival, noun and verb inflections
3.5y Mastery of inflectional morphology (see P. Lyytinen et al.,
.57 2001; Lyytinen & Lyytinen, 2004) for adjectives, verbs, and nouns
5.0y .76 Mastery of inflectional morphology (P. Lyytinen et al.,
2001; P. Lyytinen & Lyytinen, 2004)
050 lyytin (514-546)

4. Memory 3.5y Forward digit span (see Gathercole & Adams, 1994)
Sentence repetition, Developmental Neuropsychological
.67 Assessment (NEPSY; Korkman, Kirk, & Kemp, 1998).
5.0y Digit span (computer presented)
8/18/06

Syllable span (computer presented)

5.5y .75 Sentence repetition (NEPSY)
Nonword repetition (NEPSY)
6.5y Digit span (computer presented)
5. Rapid serial 3.5y Rapid serial naming of objects RSN using RAN Rapid
12:17 PM

naming Automatized Naming Test; see Denckla & Rudel (1974, 1976)
5.5y .80 RSN objects
RSN colors
6.5y RSN objects
Page 525

.89 RSN colors

RSN numbers
RSN letters
6. Letter 3.5y Letter naming (upper case fixed order)
knowledge Letter identification (from symbol distracters)
Dyslexia and Trajectories of Reading Development

.65 Word identification (symbol string distracters)

Visual matching (matching letter string pairs)
Own name writing
4.5y Letter naming
Letter identification
.72 Word identification
Visual matching
525

Own name writing

5.0y Letter naming
.66 Word identification
(continued)
Table 1. Continued

Composite mean
Age(s) of score standardized
526
050 lyytin (514-546)

Skill domain assessment alpha Task(s)

Visual matching
5.5y Letter naming
.66 Word identification
8/18/06

Visual matching
6.5y Letter naming
7. Phonological 3.5y (for phonological measures, see Puolakanaho et al., 2003)
awareness Word identification (from picture choice)
.59 Segment identification (of sub-word units)
12:17 PM

Synthesis (of syllables and phonemes)

Continuation (from presented onset of word)
Phonological processing (NEPSY)
4.5y Segment identification
Synthesis
Page 526

.71 Continuation
Initial phoneme identification
Initial phoneme production
5.5y Segment identification
.77 Synthesis
Initial phoneme identification
Initial phoneme production
6.5y Segment identification
.82 Synthesis
Initial phoneme identification
Initial phoneme production
IQ measure 5.0y Wechsler Preschool and Primary Scales of Intelligence
(WPPSI-R; Wechsler, 1989); verbal (vocabulary,
Merrill-Palmer Quarterly

arithmetic, comprehension) and performance (block

design, object assembly, picture completion) subtests.
Table 2. Reading and Spelling Tasks and Times of Assessment
050 lyytin (514-546)

1st grade 1st grade 1st grade 1st grade 1st grade 2nd grade
Reading and spelling tasks August December January April May November
1. Accuracy of reading of individually Ind. — Ind. — Ind.
8/18/06

presented words/ nonwords

2. Fluency of reading of individually Ind. — — — Ind.
presented words/nonwords
3. Accuracy of spelling of individually Ind. Group Group Ind. Group
12:17 PM

presented words/ nonwords

4. Fluency of decoding words Ind. Group Group Group
(see Lindeman, 2000)
5. Fluency of decoding sentences Group
Page 527

(see Lindeman, 2000)

6. Fluency of text reading Ind.
7. Detection of spelling errors Group Group Group
Dyslexia and Trajectories of Reading Development

8. Reading comprehension: sentences Group

(see Lindeman, 2000)
9. Reading comprehension: text Group
10. Parsing word-chains into words Group
(see Nevala & Lyytinen, 2001)
Note. Ind. = individually administered test; Group = group test in classroom context
527
050 lyytin (514-546) 8/18/06 12:17 PM Page 528

528 Merrill-Palmer Quarterly

tests (Nevala & Lyytinen, 2001). The assessments (type and time point) are
detailed in Table 2.

Statistical Approach
The modeling consisted of two parts: identification of a latent common fac-
tor and a latent class factor. Based on theoretical knowledge and the empir-
ical evidence suggesting strong correlations among the variables in the
seven domains, an underlying common factor was assumed, and this type
of latent factor was included in the construction of the model. The latent
common factor extracts the common variance shared by individuals in the
seven skill domains introduced above (receptive language, expressive lan-
guage, morphology, memory, rapid serial naming, letter knowledge, and
phonological awareness). The analysis of a common factor helps us to
ascertain the extent to which a single skill domain is associated with com-
promised reading acquisition as well as what type(s) of developmental vari-
ation emerges as separable from this main line.
The second part of the model, the latent class factor, isolates the most
homogeneous patterns of an individual’s development, thus showing which
of the differential patterns of development is connected to problems in early
reading acquisition. Our preliminary findings indicate that approximately
40% to 50% of the at-risk children display compromised reading develop-
ment during the first two school years (formal diagnosis of dyslexia must
await all cohorts of our participants completing third grade). The data thus
contain the necessary variation to allow us to observe possible differential
patterns of levels and/or rates of skill development and cases of dissimilar
but predictable developmental courses. With this combination of latent
classes and latent factors we sought to apply person-oriented analysis of
data (e.g., trajectories of latent classes) in addition to variable-oriented
analysis (e.g., explain variances and covariances of observed variables with
latent common factors; cf. Bergman, Magnusson, & El-Khouri, 2003).
We used the mixture modeling feature of the MPLUS (version 3) pro-
gram (L. K. Muthén & Muthén, 2004) to analyze the repeatedly measured
variables included under the seven skill domains. Mixture modeling
allowed us to examine simultaneously both the common factor and the
individual patterns of development by identifying clusters of children who
show similar trajectories. In construction of the model, one has the option
to apply latent classes with distinct mean values of observed variables or of
latent factors (B. Muthén, 2001). We studied both of these alternatives, and
the data supported the former option, which yields a model in which the
latent underlying skill factor and the latent classes are independent sources
050 lyytin (514-546) 8/18/06 12:17 PM Page 529

Dyslexia and Trajectories of Reading Development 529

of variation in the observed variables. To evaluate the number of latent

classes we have used three criteria: AIC (Akaike’s information criteria),
BIC (Bayesian information criteria), and adjusted BIC and Lo-Mendel-
Rubin test (Lo, Mendell, & Rubin, 2001). The lower AIC, BIC, or adjusted
BIC indicate a better model, and significant Lo-Mendel-Rubin test results
indicate a higher number of groups. The model was estimated using the
MLR (maximum likelihood method with robust standard errors) method,
and missing values (missing at random) were handled via MPLUS that uses
an EM (expected maximum) algorithm.

Results and Discussion

Table 3 summarizes the mixture model estimates. The table reveals the
common factor loadings with standard error values. It shows how much
each variable shares variance (R2) with the common factor and with the
subgrouping solution (classes), as well as the unexplained error referring to
the within-group individual variation.
A closer inspection of the numbers reveals that most of the variables
selected to predict reading acquisition have relatively high loadings with
the common factor; explanation varied from 5% of RAN to 62% of letter
knowledge scores from assessments carried out at the age of 5.0 years. This
factor reflects the variance that is common between the measures with the
weights or percentages (R2) given in Table 3. This makes the labeling of
this common factor interesting.
To elaborate upon its nature, the contribution of IQ was further exam-
ined by fixing the parameters of the preceding model and by adding the per-
formance IQ and verbal IQ scores into the model. This analysis revealed
that performance IQ had only 25% of common variance and the error (the
contribution of the individual variation) was high (74%), meaning that it
does not belong to the common factor. The shared variance of the common
factor and the verbal IQ was 44%, and 45% of the variation in verbal IQ
was due to individual variation. Furthermore, the subgrouping was associ-
ated with neither IQ (shared variance was 1% with performance IQ and
11% with verbal IQ). Thus it was mainly the verbal component of IQ that
was associated with the common factor.
It is apparent that the common factor represents the skills closely
related to reading that are available before acquisition of reading, first of all
orthographic skills such as letter knowledge (R2 = 43–76%) and phonolog-
ical awareness (R2 = 33–52%), that is, the interest and ability of the child to
attend to and manipulate in the mind the sub-word-level speech sounds.
The early reading skill itself also has a high loading (R2 = 41%). Thus, we
050 lyytin (514-546) 8/18/06 12:17 PM Page 530

Table 3. The Latent Mixture Model Estimates; Factor Loadings for the Common
Factor (standard errors are presented in parenthesis), and sources of variance for
the Observed Variables (R2)
Common R2
factor
loadings (se) Common Classes Error
Receptive language
Age 1.0–1.5 years .35 (.07) .13 .10 .77
Age 2.5 years .62 (.07) .34 .23 .43
Age 3.5 years .53 (.08) .29 .16 .54
Age 5.0 years .59 (.07) .29 .13 .59
Expressive language
Age 1.5 years .34 (.07) .14 .09 .77
Age 2.0–2.5 years .43 (.06) .18 .18 .64
Age 3.5 years .52 (.07) .26 .28 .46
Age 5.5 years .65 (.10) .31 .14 .55
Morphology
Age 2.5 years .34 (.08) .10 .23 .66
Age 3.5 years .50 (.07) .23 .21 .56
Age 5.0 years .44 (.08) .17 .12 .70
Phonological awareness
Age 3.5 years .61 (.08) .33 .06 .62
Age 4.5 years .74 (.10) .41 .12 .47
Age 5.5 years .88 (.12) .50 .15 .36
Age 6.5 years .81 (.12) .52 .21 .27
Letter knowledge
Age 3.5 years .58 (.07) .43 .02 .55
Age 4.5 years .83 (.06) .61 .05 .34
Age 5.0 years .76 (.06) .62 .09 .29
Age 5.5 years .87 (.06) .61 .12 .27
Age 6.5 years .76 (.10) .53 .16 .31
Short-term memory
Age 3.5 years .53 (.06) .28 .20 .52
Age 5.0 years .59 (.07) .32 .18 .51
Age 6.5 years .48 (.07) .20 .04 .76
Rapid naming
Age 3.5 years .40 (.07) .18 .04 .78
Age 5.5 years .35 (.10) .05 .52 .44
Age 6.5 years .67 (.09) .18 .41 .41
Overall reading composite, .71 (.12) .41 .19 .40
Grades 1–2
050 lyytin (514-546) 8/18/06 12:17 PM Page 531

Dyslexia and Trajectories of Reading Development 531

could label it the general reading readiness factor. That the foundation skills
of speech perception are involved in the common factor is reflected in up to
34% of shared variance with receptive language and the association with
verbal IQ (44%).
In the mixture analysis of the seven skill domains and early literacy,
solutions of 2–7 subgroups were identified based on the criteria indices and
test provided by MPLUS. Our decision to examine more closely the solu-
tion with four subgroups was based on the theoretical consideration and
interpretability of the alternative solutions provided by the modeling. The
average scores for probabilities of individual group membership (see Table
4) detail the differentiation of the groups, revealing that the four-group
solution provided an excellent separation. Each of the four subgroups had
only one high (> .90) average probability of group membership value, and
none of the groups overlapped with any other, as expressed by probabilities
of less than .10.
The subgroup classification information in addition to the common
variance explains 4–52% (mostly 10–20%) of the variance of the single
skill scores. As can be seen from Table 3, the naming domain makes a rela-
tively high specific contribution (up to 52%) to the subgroup classification
but contributes very little to the common factor.

The Developmental Trajectories to Reading

Figure 1 summarizes the profiles of the four subgroups (named as repre-
senting declining, typical, dysfluent, and unexpected trajectories) as a func-
tion of the seven skill domains relative to early literacy ability (the last

Table 4. Classification Table of the Average of the Individual Group Membership

Probabilities (columns) Into Most Likely Subgroup (rows)

Subgroup 1 Subgroup 2 Subgroup 3 Subgroup 4

(declining) (typical) (dysfluent) (unexpected)
Subgroup 1 .907 .058 .000 .035
(declining)
Subgroup 2 .049 .937 .000 .014
(typical)
Subgroup 3 .000 .002 .996 .002
(dysfluent)
Subgroup 4 .017 .014 .002 .966
(unexpected)
050 lyytin (514-546) 8/18/06 12:17 PM Page 532

532 Merrill-Palmer Quarterly

score in the figure). Interpretation of an individual profile belonging to a
subgroup requires recognition of the fact that one’s profile comprises the
subgroup-specific developmental (time- and domain-related) variations
and the child’s position in the reading readiness factor (according to the
common factor loadings given in Table 3).
Declining trajectory. Children in the “declining trajectory” subgroup
(subgroup 1; n = 35) shared an interesting characteristic: relative to the oth-
ers, their position decreased continuously in most skill domains, the only
exception being naming speed. This type of developmental pattern seems to
be more common among children who are at familial risk for dyslexia (only
12% of control children compared to 23% of at-risk children belonged to
this subgroup). More than half of the at-risk children and about 40% of the
control children belonging to this subgroup showed problems in early read-
ing acquisition and in attaining a sufficient level of reading speed for read-
ing comprehension. About 20% of the children with the “declining
trajectory” (both controls and children at risk) continued to show problems
in reading or spelling accuracy at the end of first grade, and even more
showed problems in the development of reading fluency. Only 2 of the con-

Figure 1. Subgroup members’ average performance in the seven skill domains (across
age from 1 to 6.5 years) and in the reading and writing skills composite (across first and
second grade). Mean values are z-scored with the control children’s distribution.
050 lyytin (514-546) 8/18/06 12:17 PM Page 533

Dyslexia and Trajectories of Reading Development 533

trol children (compared to 12 children at risk) showed reading comprehen-

sion problems at the end of first grade. A large proportion of children in this
subgroup lagged behind the average reading skill level during the end of
first grade, culminating in the fact that the average overall reading achieve-
ment across first and second grade was lowest in this subgroup (–.68
expressed as the mean z-score). However, at the same time, this group
showed the highest within-group variation, differentiating members who
are control children from those who are at risk, as shown by Figure 2.
Typical trajectory. The members of the largest subgroup (subgroup 2;
n = 85, of which 55% were control children) were proficient in all reading-

Figure 2. The subgroups’ average performance in general composite of reading and

writing skills at first and second grade separately for the children at familial risk and control
children. Mean values are z-scored with control children’s distribution.
050 lyytin (514-546) 8/18/06 12:17 PM Page 534

534 Merrill-Palmer Quarterly

related developmental skill domains, such as phonology and letter knowl-
edge. They showed continuous elevation of their mean relative skill posi-
tion across the development, which made them clearly the most skilled
group prior to school entry. Due to the common characteristic of this group
to show progressive skill development in relation to children whose devel-
opmental progressing is somehow compromised, we labeled this group as
following a “typical trajectory.” Very few of the children belonging to this
group showed any reading problems, and those few who did so showed
some delay only at the very beginning of school or showed low scores in
comprehension. This subgroup, thus, collected most good readers from
both groups. The fact that more than one third of the at-risk children
belonged to this group despite the genetic risk could be attributed to the fact
that not more than half of the children in the at-risk group are expected to
have any problems due to genetic vulnerability (only one parent affected)
and also to increased attention that participation in a longitudinal study pro-
vides in support of reading-related development.
Dysfluent trajectory. Children in the smallest subgroup (subgroup 3;
n = 12, all except 1 of the children being at risk for dyslexia) differ dramati-
cally from all others by showing naming speed problems. The common
characteristic of the members of this subgroup demarcated them as follow-
ing a “dysfluent trajectory.” The second striking characteristic of children
in this subgroup is that their relative scores tend to decrease as a function of
age in most domains (see Figure 1), with the exception of phonological
skills (which maintain their level) and verbal short-term memory (which
shows a rather clear trend toward better relative performance). A third char-
acteristic of this subgroup is that the children’s mean skill level across ages
is, with the exception of expressive language, poorest in all domains. How-
ever, as Figure 2 shows, the general composite score of reading for this sub-
group is only second poorest, although, as expected and as will be seen
later, this group’s reading rate was the lowest by far. This lower perform-
ance in reading speed also persisted during the first two school years. It is
important to note that reading instruction was relatively successful in pro-
moting the reading accuracy (but not fluency) of these children.
Unexpected trajectory. The members of the fourth subgroup (n = 67;
51% were at-risk children) were relatively strong in their early language (in
receptive and expressive speech) and in their general language skills, such
as morphology, and their memory skills were better than that of the other
groups. Surprisingly, in spite of these strengths, the mean letter knowledge
scores for the children in this subgroup were low and relatively declining
across ages, and many of the children encountered problems, especially in
the early acquisition stages of reading. The below-mean letter knowledge
before school entry seemed to be reflected in the reading acquisition in this
050 lyytin (514-546) 8/18/06 12:17 PM Page 535

Dyslexia and Trajectories of Reading Development 535

group. This shows the immediate connection of the development of letter

knowledge to reading in a consistent orthography. The common character-
istic of children in this subgroup seems to be that they were proficient in
everything else assessed in this study but not in the domains most closely
associated with reading. This finding is very unexpected, because these
children’s spoken language skills were atypically high in comparison to the
others and to their reading skill. Accordingly, we named this subgroup as
“unexpected,” which describes how many of these children were compro-
mised early readers independent of good language skills and memory. It is
important to note that the non-optimal reading acquisition was not specific
to the at-risk members of this subgroup.

Predicting Reading on the Basis of the Combined

Variable-Centered and Person-Centered Model
When the trajectories of latent classes and the common factor are unrelated,
they both contribute independently toward the prediction of reading. We
therefore examined the prediction value provided by these two model com-
ponents to the outcome measure of the overall reading skill composite.
In Figure 3 we fixed the contribution of the common factor between –1
SD and +1 SD and computed the total prediction of reading in each sub-
group, taking into account this variation in the common factor. Within this
range of common factor variation, predicted z-values of reading skills in the
“declining trajectory” varied between –1.79 and –.37 (mean value –1.08),
which is very similar to the respective values of the “dysfluent trajectory,”
both thus being quite predictive of problems in reading acquisition. The
values of the children following a “typical trajectory” varied around zero
and were above that of the declining and dysfluent children, while the atyp-
ical early readers with good spoken language skills (unexpected) were
almost at the same level with the dysfluent and declining children. Conse-
quently, based on both the classification and general reading readiness
(common) factor, we were able to find three different routes to compro-
mised reading acquisition in the sense of reading below normal level.

The Role of Familial Risk Requires Further Scrutiny

One cannot totally exclude the possibility that the risk factor may still exert
some effects that we did not observe or that the model was not optimized for
prediction. In a regression analysis that predicted the overall reading compos-
ite with the membership of the at-risk group and controlled for subgrouping
and common factor score, familial risk still explained 3% of the variation of
the reading composite, F(1, 192) = 5.87, p = .016. Of the 38 at-risk children
050 lyytin (514-546) 8/18/06 12:17 PM Page 536

536 Merrill-Palmer Quarterly

Figure 3. The prediction of the overall performance in reading and writing skills (z-scored
with control children’s distribution) provided by the model estimates of subgrouping and the
common factor. Common factor values were here restricted to 1 SD around the mean.

belonging to the subgroup with typical trajectory, 7 (and 1 of the 47 control

children) had a slow start in their reading career in terms of learning to read
later than others. In addition, 8 at-risk children (compared to 2 controls) in
this subgroup scored 1 SD below the control children’s mean in reading com-
prehension at the end of first grade. We may accept that, among typical chil-
dren, a small minority may get off to a slow start and experience
comprehension problems. It could be argued that such children may flourish
if educators pay attention (during school) to that domain of development.

Is Reading Acquisition Predicted by Different Developmental Paths

of Language Development?
To obtain a more detailed picture concerning the prediction of reading devel-
opment, we analyzed the predictors separately for (1) the early reading seen
in the combined accuracy and fluency measures/scores at the beginning of
first grade, (2) spelling during the beginning of first grade and the end of first
grade, (3) reading accuracy at the end of first grade, (4) reading fluency at the
end of first grade, and (5) comprehension at the end of first grade.
050 lyytin (514-546) 8/18/06 12:17 PM Page 537

Dyslexia and Trajectories of Reading Development 537

Figures 4 and 5 summarize the percentages of children from each

group scoring 1 SD below the mean of the control children on each depend-
ent measure of interest. These figures reveal that the highest proportion of
delayed readers are at familial risk and additionally belonged either to the
“dysfluent” or “declining” trajectory subgroups. More than half of these
children were slow to acquire basic reading skills, while in both of these
subgroups roughly the same proportion of children remained delayed in
comprehension (for a more detailed analysis of predictive relationships
between receptive language and reading comprehension, see also P. Lyyti-
nen, Eklund, & Lyytinen, 2005). Progress in early reading acquisition was
slow among the control children belonging to these two subgroups, but the
proportion of control children belonging to these two subgroups was much
lower (12%) than the proportion of children at familial risk (33%). Note
that only one control child showed a “dysfluent trajectory.” This suggests
that, within a consistent orthography, dysfluency is especially implicated in
reading from a genetic standpoint, supporting the contention by Wimmer,
Mayringer, and Landerl (2000) and earlier Finnish findings by Korhonen
(1995) that dysfluency may be the most persistent feature of dyslexia. It
must be noted that both of these findings emerge from an orthography
where reading accuracy is attained relatively early and apparently with
more ease (Aro & Wimmer, 2003; Seymour et al., 2003) than can be
expected from readers of an orthography that demands the learning of more
complex systems of letter-sound connections.

Figure 4. Percentages of the control children performing more than 1 SD below control
children’s mean in first-grade reading and writing skills within each subgroup. Because only
one child belonged to the dysfluent subgroup, the related percentage was not included in
the figure.
050 lyytin (514-546) 8/18/06 12:17 PM Page 538

538 Merrill-Palmer Quarterly

Figure 5. Percentages of the familial-risk children performing more than 1 SD below

control children’s mean in first-grade reading and writing skills within each subgroup.

As Figures 4 and 5 show, the children following “dysfluent” and “declin-

ing” trajectories were also delayed at the end of first grade, but by this time
there was a clear differentiation between accuracy and speed. Rate of reading
was the main impairment in the dysfluent subgroup of the at-risk children,
while not more than 20% of both familial-risk and non-risk members of the
“declining” trajectory had persistent difficulties in both rate of reading and
accuracy, as observed in both reading and spelling. The consequences of
these impairments were, however, different for the at-risk children and the
non-risk children: almost twice the number of at-risk children had problems
in comprehension than did the control children in this “declining” subgroup.
The fate of those who are atypical readers but skilled speakers (the
“unexpected” subgroup) is interesting. Although the proportions of control
and familial-risk children were equal, the incidence of at-risk children
showing delayed acquisition of accurate reading was almost twice that of
the control children. The development of reading speed remained delayed
in both control and familial-risk children, while accuracy in both reading
and spelling increased quickly, especially among controls. A surprising
observation was that the proportions of delayed comprehenders within this
subgroup were relatively similar (> 20%) in both groups.

Concluding Remarks
The main finding of the present analysis is that there is substantial inter-
individual variation, not only between the groups of children differing on
the basis of familial background, but also within these groups. A substantial
part of this variability was anticipated by the children’s skill development
050 lyytin (514-546) 8/18/06 12:17 PM Page 539

Dyslexia and Trajectories of Reading Development 539

prior to formal reading instruction. We were able to document a new aspect

almost untouched in the earlier literature; that is, it may be the trend of the
development that is predictive rather than the level profile. Thus, some chil-
dren who commenced with a relatively high level but had a declining trend
of development were among the poorest readers. It also shows that very
early detection may not always be possible. A small number of our children
showed their first signs of risk relatively late.
The manifestations of risk in skill development among the children at
familial risk are clearly divided into two developmental profiles. One is
characterized by dysfluency, covering both spoken naming and reading
speed. The other profile is low and relatively declining achievement rela-
tive to the mean level throughout most skill domains, but most clearly in
phonological awareness and letter knowledge, which are known to be the
closest predictive associates of early reading. Slow growth of letter knowl-
edge seems to be a strong predictor of poor reading acquisition, independ-
ent of familial risk.
We observed a number of surprises in our findings. The most dramatic
one is that early optimal development of receptive and expressive language
does not always predict early reading, and in fact, sometimes an abnormal
course of literacy development may ensue. Some of the children whose ini-
tial spoken-language skills were above the mean nevertheless manifested
less optimal progress in early reading, which was accompanied by a dra-
matic relative drop in their development of letter knowledge from the
beginning level at age 3.5 years to the months before school entry. It may be
that these children’s better-than-average early spoken language skills
resulted in a relatively lower amount of attention from adults with regard to
providing environmental support and instruction for the child in learning
letters and phonological skills. This type of effect is especially pertinent to
children who are at biological risk for dyslexia. A relatively strong correla-
tion between supportive aspects of the home literacy environment and
development of phonological awareness skills has, in our previous analy-
ses, been observed to be effective for at-risk children but not for control
children (Torppa et al., submitted).
In the present data set we did not include measures that could have
reflected development in non-language domains such as motor skills
(described by Fawcett & Nicolson, 1995), although this type of co-occurrence
is possible (for preliminary findings from our study, see H. Lyytinen, Ahonen,
et al., 2004). We have very recently analyzed our data collected from the motor
domain and found support for its potential role in the early developmental risk
profile of a subgroup of children by detecting an early motor delay that is asso-
ciated with language development and a delay in acquiring fluent reading skill
(Viholainen et al., 2006).
050 lyytin (514-546) 8/18/06 12:17 PM Page 540

540 Merrill-Palmer Quarterly

With respect to general practical implications, our findings indicate that a
relatively comprehensive assessment of development is required in order to
detect, as early as possible, all or even most of the children who may have
problems in reading acquisition. The detection rate is clearly better and fol-
lows a more expected pattern among children who have family members
with reading problems—a predictive sign that is easy to take into account.
In this study we concentrated on early acquisition of reading and writ-
ing, mainly early decoding skills. The ultimate goal—fluent reading with
complete comprehension—received only minor attention due to the age of
the children. We are as optimistic as Vellutino et al. (2004) in our belief that
almost everyone can be helped to attain accurate reading skill. However,
the time and practice required for this achievement varies greatly between
individuals. What is less clear today is whether dysfluency (slow reading),
which appears to be common to both naming and reading, can be helped to
the same extent (for recent reviews, see H. Wimmer et al., 1998; Wolf,
Bowers, & Biddle, 2000). It appears that the fluency attained in normal
reading development is especially difficult to achieve among individuals
with dyslexia in consistent orthographies. In contrast, accuracy, which has
also received most attention in the research, especially in less consistent
orthographies such as English, presents a significant challenge to readers of
such orthographies.
Based on the present findings, we can speculate that susceptibility for
difficulties in the higher cognitive aspects of reading with comprehension
likely has its roots in more foundational aspects of reading skill as expressed
in indices of fluency. Fluency measures may merit a stronger emphasis in the
prediction of achievements in reading comprehension. Thus, in addition to
the assessment of phonological skills (not forgetting the child’s readiness to
acquire letter knowledge), naming speed also merits attention. Together,
these measures are likely to provide relevant prognostic data for development
in processing text with comprehension, which is often compromised in con-
junction with early acquisition of basic reading skill.

References
Adams, M. J. (1990). Beginning to read: Thinking and learning about print. Cam-
bridge: MIT Press.
Aro, M. (2006). Learning to read: The effect of orthography. In R. M. Joshi & P. G.
Aaron (Eds.), Handbook of orthography and literacy (pp. 531–550). Hillsdale,
NJ: Erlbaum.
Aro, M., & Wimmer, H. (2003). Learning to read: English in comparison to six
more regular orthographies. Applied Psycholinguistics, 26, 619–634.
050 lyytin (514-546) 8/18/06 12:17 PM Page 541

Dyslexia and Trajectories of Reading Development 541

Beauvois, M. F., & Derouesné, J. (1979). Phonological alexia: Three dissociations.

Journal of Neurology, Neurosurgery and Psychiatry, 42, 1115–1124.
Bergman, L. R., Magnusson, D., & El-Khouri, B. M. (2003). Studying individual
development in an interindividual context: A person-oriented approach. Vol. 4
of Paths through life (D. Magnusson, Series Ed.). Mahwah, NJ: Erlbaum.
Bradley, L., & Bryant, P. E. (1983). Categorising sounds and learning to read: A
connection. Nature, 301, 419–421.
Bronfenbrenner, U., & Morris, P. A. (1998). The ecology of developmental
processes. In W. Damon (Ed.), Handbook of child psychology. Vol. 1: Theoret-
ical models of human development (5th ed., pp. 993–1028). New York: Wiley.
Byrne, B. (1998). The foundation of literacy: The child’s acquisition of the alpha-
betic principle. Hove, UK: Psychology Press.
Coltheart, M., Masterson, J., Byng, S., Prior, M., & Riddoch, J. (1983). Surface
dyslexia. Quarterly Journal of Experimental Psychology, 35A, 469–495.
DeJong, P., & Olson, R. K. (2004). Early predictors of letter knowledge. Journal of
Experimental Child Psychology, 88(3), 254–273.
DeJong, P., & Vrielink, L. O. (2004). Rapid automatic naming: Easy to measure,
hard to improve (Quickly). Annals of Dyslexia, 54, 65–88.
Denckla, M. B., & Rudel, R. G. (1974). Rapid “automatized” naming of pictured
objects, colors, letters and numbers by normal children. Cortex, 10, 471–479.
Denckla, M. B., & Rudel, R. G. (1976). Rapid “automatized” naming (R.A.N.):
Dyslexia differentiated from other learning disabilities. Neuropsychologia, 14,
471–479.
Dunn, L. M., & Dunn, L. M. (1981). Peabody Picture Vocabulary Test–Revised.
Circle Pines, MN: American Guidance Service.
Elbro, C., Borstrøm, I., & Petersen, D. K. (1998). Predicting dyslexia from kinder-
garten: The importance of distinctness of phonological representations of lexi-
cal items. Reading Research Quarterly, 33, 36–60.
Erskine, J. M., & Seymour, P. H. K. (2005). Proximal analysis of developmental
dyslexia in adulthood: The cognitive mosaic model. Journal of Educational
Psychology, 97(3), 406–424.
Fawcett, A. J., & Nicolson, R. I. (1995). Persistent deficits in motor skill of children
with dyslexia. Journal of Motor Behavior, 27, 235–240.
Fenson, L., Dale, P. S., Reznick, J. S., Bates, E., Thal, D., & Pethick, S. J. (1994).
Variability in early communicative development. Monographs of the Society
for Research in Child Development, 59 (5, Serial No. 242).
Fowler, A. E. (1991). How early phonological development might set the stage for
phoneme awareness. In S. A. Brady & D. P. Shankweiler (Eds.), Phonological
processes in literacy (pp. 97–117). Hillsdale, NJ: Erlbaum.
Gallagher, A. Frith, U., & Snowling, M. J. (2000). Precursors of literacy delay
050 lyytin (514-546) 8/18/06 12:17 PM Page 542

542 Merrill-Palmer Quarterly

among children at genetic risk of dyslexia. Journal of Child Psychology and
Psychiatry, 41, 203–213.
Gathercole, S. E., & Adams, A.-M. (1994). Children’s phonological working mem-
ory: Contributions of long-term knowledge and rehearsal. Journal of Memory
and Language, 33, 672–688.
Gathercole, S. E., & Baddeley, A. D. (1997). Sense and sensitivity in phonological
memory in vocabulary development (A reply to Bowey, 1996). Journal of
Experimental Child Psychology, 67, 290–294.
Gibson E. J. (1970). The development of perception as an adaptive process. Ameri-
can Science, 58(1), 98–107.
Goswami, U., & Bryant, P. B. (1990). Phonological skills and learning to read.
London: Erlbaum.
Guttorm, T. K., Leppänen, P. H. T., Poikkeus, A.-M., Eklund, K. M., Lyytinen, P., &
Lyytinen, H. (2005). Brain event-related potentials (ERPs) measured at birth
predict later language development in children with and without familial risk
for dyslexia. Cortex, 41, 291–303.
Guttorm, T. K., Leppänen, P. H. T., Richardson, U., & Lyytinen, H. (2001). Event-
related potentials and consonant differentiation in newborns with familial risk
for dyslexia. Journal of Learning Disabilities, 34, 534–544.
Guttorm, T. K., Leppänen, P. H. T., Tolvanen, A., & Lyytinen, H. (2003). Event-
related potentials in newborns with and without familial risk for dyslexia: prin-
cipal component analysis reveals differences between the groups. Journal of
Neural Transmission, 110, 1059–1074.
Hoff, E., & Naigles, L. (2002). How children use input to acquire a lexicon. Child
Development, 73, 418–433.
Holopainen, L., Ahonen, T., & Lyytinen, H. (2001). Predicting delay in reading
achievement in a highly transparent language. Journal of Learning Disabili-
ties, 34, 401–413.
Kaplan, E., Goodglass, H., & Weintraub, S. (1983). The Boston Naming Test (2nd
ed.). Philadelphia: Lea & Febiger.
Karlsson, F. (1999). Finnish. An essential grammar. London: Routledge.
Korhonen, T. (1995). The persistence of rapid naming problems in children with
reading disabilities: A nine-year follow-up. Journal of Learning Disabilities,
28, 232–239.
Korkman, M., Kirk, U., & Kemp, S. (1998). NEPSY: A developmental neuropsy-
chological assessment. San Antonio: Psychological Corporation.
Kuhl, P. K., Williams, K. A., Lacerda, F., Stevens, K. N., & Lindblom, B. (1992).
Linguistic experience alters phonetic perception in infants by 6 months of age.
Science, 255, 606–608.
Laakso, M.-L., Poikkeus, A.-M., Eklund, K., & Lyytinen, P. (2004). Children’s
interest in early shared reading and it’s relation to later language and letter
050 lyytin (514-546) 8/18/06 12:17 PM Page 543

Dyslexia and Trajectories of Reading Development 543

knowledge in children with and without genetic risk for dyslexia. First Lan-
guage, 24, 323–345.
Leinonen, S., Müller, K., Leppänen, P. H. T., Aro, M., Ahonen, T., & Lyytinen, H.
(2001). Heterogeneity in adult dyslexic readers: Relating processing skills to
the speed and accuracy of oral text reading. Reading and Writing, 14,
265–296.
Leppänen, P. H. T., Eklund, K. M., & Lyytinen, H. (1997). Event-related brain
potentials to change in rapidly presented acoustic stimuli in newborns. Devel-
opmental Neuropsychology, 13(2), 175–204.
Lindeman, J. (2000). Ala-asteen lukutesti: Käyttäjän käsikirja [The Comprehen-
sive School Reading Test]. Turku: Oppimistutkimuksen keskus, Turun
yliopisto.
Lo, Y., Mendell, N. R., & Rubin, D. R. (2001). Testing the number of components
in a normal mixture. Biometrika, 88, 767–778.
Locke, J. L., Hodgson, J., Macaruso, P., Roberts, J., Lambrecht-Smith, S., & Gut-
tentag, C. (1997). The development of developmental dyslexia. In C. Hulme &
M. Snowling (Eds.), Dyslexia: Biology, cognition, and intervention
(pp. 73–96). London: Whurr.
Lonigan, C. J., Burgess, S. R., & Anthony, J. L. (2000). Development of emergent
literacy and early reading skills in preschool children: Evidence from a latent-
variable longitudinal study. Developmental Psychology, 36, 596–613.
Lundberg, I. (1994). Reading difficulties can be predicted and prevented. A Scandi-
navian perspective on phonological awareness and reading. In C. Hulme &
M. Snowling (Eds.), Reading development and dyslexia (pp. 180–199). Lon-
don: Whurr.
Lyytinen, H., Ahonen, T., Eklund, K., Guttorm, T. K., Kulju, P., Laakso, M.-L.,
Leiwo, M., Leppänen, P. H. T., Lyytinen, P., Poikkeus, A.-M., Richardson, U.,
Torppa, M., & Viholainen, H. (2004). Early development of children at famil-
ial risk for dyslexia—Follow-up from birth to school age. Dyslexia, 10,
146–178.
Lyytinen, H., Ahonen, T., Eklund, K., Guttorm, T. K., Laakso, M.-L., Leinonen, S.,
Leppänen, P. H. T., Lyytinen, P., Poikkeus, A.-M., Puolakanaho, A., Richard-
son, U., & Viholainen, H (2001). Developmental pathways of children with
and without familial risk for dyslexia during the first years of life. Develop-
mental Neuropsychology, 20, 535–554.
Lyytinen, H., Aro, M., Eklund, K., Erskine, J., Guttorm, T. K., Laakso, M.-L., Lep-
pänen, P. H. T., Lyytinen, P., Poikkeus, A.-M., Richardson, U., & Torppa, M.
(2004). The development of children at familial risk for dyslexia: birth to
school age. Annals of Dyslexia, 54, 184–220.
Lyytinen, H., Aro, M., Holopainen, L., Leiwo, M., Lyytinen, P., & Tolvanen, A.
(2006). Children’s language development and reading acquisition in a highly
050 lyytin (514-546) 8/18/06 12:17 PM Page 544

544 Merrill-Palmer Quarterly

transparent orthography. In R. M. Joshi & P. G. Aaron (Eds.), Handbook of
orthography and literacy (pp. 47–62). Mahwah, NJ: Erlbaum.
Lyytinen, H., Guttorm, T. K., Huttunen, T., Hämäläinen, J., Leppänen, P. H. T., &
Vesterinen, M. (2005). Psychophysiology of developmental dyslexia:
A review of findings including studies of children at risk for dyslexia. Journal
of Neurolinguistics, 18, 167–195.
Lyytinen, H., Leppänen, P. H. T., Richardson, U., & Guttorm, T. (2003). Brain func-
tions and speech perception in infants at risk for dyslexia. In V. Csépe (Ed.),
Dyslexia: Different brain, different behaviour. Neuropsychology and Cogni-
tion Series (pp. 113–152). Dordrecht, Netherlands: Kluwer.
Lyytinen, H., Ronimus, M., Alanko, A., Taanila, M., & Poikkeus, A.-M. (submit-
ted). Early identification and prevention of dyslexia.
Lyytinen, P. (1999). Varhaisen kommunikaation ja kielen kehityksen arvioin-
timenetelmä (Finnish manual for the MacArthur Communicative Develop-
ment Inventory, CDI). University of Jyväskylä, Child Research Center and
Niilo Mäki Institute. Jyväskylä: Yliopistopaino.
Lyytinen, P., Eklund, K., & Lyytinen, H. (2005). Language development and liter-
acy skills in late-talking toddlers with and without familial risk for dyslexia.
Annals of Dyslexia, 55, 166–192.
Lyytinen, P., & Lyytinen, H. (2004). Growth and predictive relations of vocabulary
and inflectional morphology in children with and without familial risk for
dyslexia. Applied Psycholinguistics, 25, 397–411.
Lyytinen, P., Poikkeus, A.-M., Laakso, M.-L., Eklund, K., & Lyytinen, H. (2001).
Language development and symbolic play in children with and without famil-
ial risk for dyslexia. Journal of Speech, Language and Hearing Research, 44,
873–885.
Manis, F. R., Seidenberg, M. S., & Doi, L. M. (1999). See dick RAN: rapid naming
and the longitudinal prediction of reading subskills in first and second grades.
Scientific Studies of Reading, 3, 129–157.
Marshall, J. C., & Newcombe, F. (1973). Patterns of paralexia: a psycholinguistic
approach. Journal of Psycholinguistic Research, 2, 175–199.
Metsälä, J. L., & Walley, A. C. (1998). Spoken vocabulary growth and the segmen-
tal restructuring of lexical representations: Precursors to phonemic awareness
and early reading ability. In J. L. Metsala & L. C. Ehri (Eds.), Word recognition
in beginning literacy (pp. 89–120). New York: Erlbaum.
Muthén, B. (2001). Latent variable mixture modeling. In G. A. Marcoulides & R. E.
Schumacker (Eds.), New developments and techniques in structural equation
modeling (pp. 1–33). Mahwah, NJ: Erlbaum.
Muthén, L. K., & Muthén, B. O. (2004). Mplus User’s Guide. Los Angeles.
Nevala, J., & Lyytinen, H. (2001). Sanaketjutesti [Word Chain Test]. Jyväskylä:
Niilo Mäki Instituutti ja Jyväskylän yliopiston Lapsitutkimuskeskus.
050 lyytin (514-546) 8/18/06 12:17 PM Page 545

Dyslexia and Trajectories of Reading Development 545

Pennington, B. F. (1991). Diagnosing learning disorders: A neuropsychological

framework. New York: Guilford.
Pennington, B. F., & Lefly, D. L. (2001). Early reading development in children at
family risk for dyslexia. Child Development, 72, 816–833.
Poskiparta, E., Niemi, P., Lepola, J., Ahtola, A., & Laine, P. (2003). Motivational-
emotional vulnerability and difficulties in learning to read and spell. British
Journal of Educational Psychology, 73, 187–206.
Puolakanaho, A., Poikkeus, A.-M., Ahonen, T., Tolvanen, A., & Lyytinen, H.
(2003). Assessment of three-and-a-half-year-old children’s emerging phono-
logical awareness in a computer animation context. Journal of Learning Dis-
abilities, 36, 416–423.
Puolakanaho, A., Poikkeus, A.-M., Ahonen, T., Tolvanen, A., & Lyytinen, H.
(2004). Emerging phonological awareness as a precursor of risk in children
with and without familial risk for dyslexia. Annals of Dyslexia, 54, 221–243.
Reynell, J., K., & Huntley, M. (1987). Reynell Developmental Language Scales
Manual (2nd ed.). Windsor, UK: NFER-Nelson.
Richardson, U. (1998). Familial dyslexia and sound duration in the quantity dis-
tinctions of Finnish infants and adults. Studia Philologica Jyväskyläensia, 44.
Jyväskylä: University Printing House
Scarborough, H. S. (1990). Very early language deficits in dyslexic children. Child
Development, 61, 1728–1743.
Scarborough, H. S. (2001). Connecting early language and literacy to later reading
(dis)abilities: Evidence, theory, and practice. In S. Neuman & D. Dickinson
(Eds.), Handbook for research in early literacy (pp. 97–110). New York: Guil-
ford.
Seymour, P. H. K., Aro, M., & Erskine, J. M. (2003). Foundation literacy acquisi-
tion in European orthographies. British Journal of Psychology, 94, 143–174.
Shaywitz, B. A., Shaywitz, S. E., Blachman, B. A., Pugh, K. R., Fulbright, R. K.,
Skudlarski, P., Mencl, W. E., Constable, R. T., Holahan, J. M., Marchione, K. E.,
Fletcher, J. M., Reid Lyon, G., & Gore, J. C. (2004). Development of left occipi-
totemporal systems for skilled reading in children after a phonologically-based
intervention. Biological Psychiatry, 55, 926–933.
Snow, C. E., Burns, M. S., & Griffin, P. (Eds.). (1998). Preventing reading difficul-
ties in young children. Washington, DC: National Academy Press.
Snowling, M. J. (1998). Dyslexia as a phonological deficit: Evidence and implica-
tions. Child Psychology and Psychiatry Review, 3(1), 4–11.
Snowling, M. J., Gallagher, A., & Frith, U. (2003). Family risk of dyslexia is con-
tinuous: Individual differences in the precursors of reading skill. Child Devel-
opment, 74, 358–373.
Stanovich, K. E. (1990). Explaining the differences between the dyslexic and the
garden-variety poor reader. The phonological-core variable difference model.
Journal of Learning Disabilities, 21, 690–604.
050 lyytin (514-546) 8/18/06 12:17 PM Page 546

546 Merrill-Palmer Quarterly

Taipale, M., Kaminen, N., Nopola-Hemmi, J., Haltia, T., Myllyluoma, B., Lyyti-
nen, H., Müller, K., Kaaranen, M., Lindsberg, P. J., Hannula-Jouppi, K., &
Kere, J. (2003). A candidate gene for developmental dyslexia encodes a
nuclear tetratricopeptide repeat domain protein dynamically regulated in
brain. Proceedings of the National Academy of Sciences, 100, 11553–11558.
Thaler, V., Ebner, E., Wimmer, H., & Landerl, K. (2004). Training reading fluency
in dysfluent readers with high reading accuracy: Word specific effects but low
transfer to untrained words. Annals of Dyslexia, 54(1), 89–113.
Torppa, M., Poikkeus, A.-M., Laakso, M.-L., Leskinen, E., Tolvanen, A., Puo-
lakanaho, A., Leppänen, P. H. T., & Lyytinen, H. (submitted). Modelling the
early paths of phonological awareness and factors supporting its development
in children with and without familial risk of dyslexia.
Vellutino, F. V., Fletcher, J. M., Snowling, M. J., & Scanlon, D. M. (2004). Specific
reading disability (dyslexia): what we have learned in the past four decades.
Journal of Child Psychology and Psychiatry, 45, 2–40.
Viholainen, H., Ahonen, T., Lyytinen, P., Cantell, M., Tolvanen, A., & Lyytinen, H.
(2006). Early motor development and later language and reading skills in chil-
dren at risk for familial dyslexia. Developmental Medicine and Child Neurology,
48, 367–373.
Wagner, R. K., Torgesen, J. K., & Laughon, P. (1993). Development of young read-
ers’ phonological processing abilities. Journal of Educational Psychology, 85,
83–103.
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.
Wechsler, D. (1989). Wechsler Preschool and Primary Scale of Intelligence-
Revised. San Antonio: Psychological Corporation.
Wimmer, H. (1993). Characteristics of developmental dyslexia in a regular writing
system. Applied Psycholinguistics, 14, 1–33.
Wimmer, H., Mayringer, H., & Landerl, K. (1998). Poor reading: A deficit in skill-
automatization or phonological deficit? Scientific Studies of Reading, 2,
321–340.
Wimmer, H., Mayringer, H., & Landerl, K. (2000). The double-deficit hypothesis
and difficulties in learning to read a regular orthography. Journal of Educa-
tional Psychology, 92, 668–680.
Wolf, M. (1984) Naming, reading and the dyslexias: A longitudinal overview.
Annals of Dyslexia, 34, 87–115.
Wolf, M., & Bowers, P. G. (1999).The double-deficit hypothesis for the develop-
mental dyslexias. Journal of Educational Psychology, 91, 415–438.
Wolf, M., Bowers, P. G., & Biddle, K. (2000). Naming-speed processes, timing,
and reading: A conceptual review. Journal of Learning Disabilities, 33,
387–407.