Early signs of dyslexia from the speech and language processing ofchildren

ULLA RICHARDSON1, PIRJO KULJU1, LEA NIEMINEN1, & PAIVI TORVELAINEN1

1University of Jyvaskyla, Finland, and 2University of Tampere, Finland

AbstractThe Jyvaskyla Longitudinal Study of Dyslexia project (JLD) has followed the development of 200 children from birth until10 years of age. Half the children are from families in which at least one of the parents has dyslexia, thus the child has a highrisk of becoming dyslexic, and half have no such risk. Here the main findings of four studies in linguistics from the JLDproject are reviewed. The speech processing skills were studied in 6, 18, 24 and 30-month-old children. The findings showthat early signs of dyslexia can be detected in speech processing both phonologically and morphosyntactically. Theseprecursors can be seen in perception or production of duration, in the prosody and phonotactics of word productionattempts and word structures, as well as in the complexity of morphosyntactic features of expressions. This information canbe useful in practice to provide early and appropriate assistance for those children who most likely will face difficulties inlearning to read and write at school.

Keywords: Dyslexia, reading, early language development, phonological processing, morphosyntactic processing.

Introduction

In several studies of dyslexia, attempts have been

made to show that the quality of phonological

representations is closely associated with the child’s

reading skills. Indeed, many studies involving both

typically (e.g., Bradley & Bryant, 1983; Hoien,

Lundberg, Stanovich, & Bjaalid, 1995; Siok &

Fletcher, 2001) and atypically developing children

(e.g., Bradley & Bryant, 1983; Bruck, 1992; Landerl,

Wimmer, & Frith, 1997; Porpodas, 1999) from

across languages demonstrate the importance of the

aforementioned relationship. According to the pho-

nological deficit theory (Stanovich, 1988) children

with dyslexia have difficulties forming accurate and

detailed phonological representations from spoken

words. However, the reasons behind qualitative

differences or accessibility issues of phonological

representations are not well understood.

Here we review the main findings of four studies in

linguistics, all of which are aimed at finding early

signs of dyslexia in language development. Recogni-

tion of early signs of dyslexia is important, since up to

now the earliest dyslexia can be diagnosed in

children is around the second or third year of school

by which time children have already struggled for

several years from inability to master technical as well

as fluency aspects of reading. With early discovery of

signs of dyslexia, appropriate tasks and training

programs can be developed for children before they

enter school. Children would have a better opportu-

nity to start their school career on an even ground

with their peers and concentrate on learning rather

than using all their efforts in mastering the basic tools

of learning in the school environment. The search for

early precursors is possible since dyslexia is familially

aggregated. In fact, the research findings are very

straightforward showing that from 35–50% of

children with at least one parent with dyslexia will

also have dyslexia (e.g., Pennington, 1995). Thus,

the genetic or familiarity aspect enables study designs

in which early signs of dyslexia can be searched long

before the actual evidence can be seen from

children’s failing attempts to learn to read and write.

For such endeavours longitudinal study designs are

ideal.

This paper reviews four studies from the same

longitudinal study on dyslexia, the Jyvaskyla Long-

itudinal Study on Dyslexia (JLD (see Table I).

Altogether 200 children from Finland have been

followed from birth until 10 years of age and the data

collection still continues (see e.g., Lyytinen et al.,

Correspondence: Ulla Richardson, PO Box 35 (Agora), FIN-40014 University of Jyvaskyla, Finland. Tel: þ358 14 260 4662. Fax: þ358 14 260 4400.

E-mail: ulla.richardson@jyu.fi

International Journal of Speech-Language Pathology, 2009; 11(5): 366–380

ISSN 1754-9507 print/ISSN 1754-9515 online ª The Speech Pathology Association of Australia Limited

Published by Informa UK Ltd.

DOI: 10.1080/17549500903147545

2004). Half the children are from families in which

at least one of the parents had dyslexia, and thus

the child is at risk of having dyslexia, and half the

children do not have such risk. During the JLD

project the children’s development has been

followed intensively (on average in yearly testing

points) focusing on a wide array of aspects of

cognitive and language development with contribu-

tions from multiple disciplines, including molecular

genetics and psychophysiology. Before the JLD

project the youngest children studied using a

similar longitudinal design were 2.5-year-old Eng-

lish-speaking children (Scarborough, 1990a). Here

we present data from even younger children,

concentrating on the data collected when the

children were 6-, 18-, 24- and 30-months-old.

We studied aspects of speech perception and

production from phonetic as well as phonological

points of view but also present findings of

children’s morphosyntactic skills since apart from

the evidence on deficient phonological processing

skills there is also evidence on the role of

morphosyntactical skills in dyslexia (Locke et al.,

1997; Scarborough, 1990a), and by focusing

merely on phonological problems important lan-

guage processing problems related to dyslexia may

go unnoticed (Bishop & Snowling, 2004).

Features of the Finnish language

Before we continue to the description of the

individual studies, a few words of the main char-

acteristics of the Finnish language are in order (see

also Kunnari & Savinainen-Makkonen, 2007). Pho-

nologically the main characteristics of Finnish are

vowel harmony as well as quantity in both vowels and

consonants. The main stress is always on the first

syllable, secondary stress on the third and every other

syllable after that, with the last syllable never being

stressed. Stress is not usually quantity sensitive.

Unlike in English, stress does not cause any notice-

able modifications to vowel quality. Orthographically

Finnish has an extremely consistent grapheme-

phoneme correspondence: each phoneme is repre-

sented by exactly one grapheme and vice versa. The

long quantity of sounds is marked with two identical

letters. Typologically Finnish is an agglutinative

language although morphophonology adds many

fusional features to the inflections. Apart from the

15 cases of nouns, signs, suffixes, and affixes can be

added to noun stems, and therefore Finnish words

can often be long. In addition, one form of a word

can have many different semantic meanings. On the

other hand, morphophonological features allow

word order to be flexible: the syntactic functions

of the words are expressed with inflections and it is

not necessary to use word order for this purpose.

Verb conjugations have finite forms, signs of

tempus, modus and voice as well as various affixes.

In addition Finnish has a rich derivation system in

both verbs and nouns. Typically to the Finno-Ugric

languages Finnish also has a declinable negative

auxiliary and not just a negative particle. In

comparison to English (which is the language of

most of the dyslexia studies) which has minimal

inflection, and almost no cases, Finnish is drasti-

cally different since a noun in Finnish can have over

2000 different forms, and a verb up to 12,000

different inflections (e.g., koti ‘‘home’’, kodeissa ‘‘in

homes’’, koteihin ‘‘into homes’’, kotisi ‘‘your home’’,

etc.; pesen ‘‘I wash’’, pesemme ‘‘we wash’’, pesetko

‘‘do you wash?’’, pesivatko ‘‘did they wash’’, et pese

‘‘you don’t wash’’, eivat pesseet ‘‘they did not wash’’,

etc.).

Table I. Studies of the Jyvaskyla Longitudinal Study of Dyslexia project reviewed here.

Study Author Age of children

Aspect of language development

under investigation Measures used

1a Richardson (1998) 6 months (þadults) Categorization skills Head turns to stimuli with varying sound

durations

1b Richardson (1998) 18 months (þadults) Use of duration in word

productions

Measurements of speech sound

duration in words and pseudowords

elicited from an imitation task

2 Torvelainen (2007) 24 months Variation in phonological

development in terms of

intelligibility of speech,

attempted and produced words

Amount of unintelligible speech, attempted

and produced words of different length;

phonotactics, MLU1, IPSyn2 & MCDI3

measured from spontaneous speech

3 Turunen (2003)

(currently Kulju)

30 months Prosodic hierarchy in word

structures

Production of three and four syllabic words,

heavy unstressed syllables, consonant

sequences and diphthongs, phonemes /s/

and /r/ produced in a naming task

4 Nieminen (2007) 30 months Structural complexity of

utterances

MLU1,,IPSyn2, & UA4 of utterances

produced in spontaneous speech

1Mean Length of Utterances (Brown, 1973; Nieminen & Torvelainen, 2003).2Index of Productive Syntax (Nieminen & Torvelainen, 2003; Scarborough, 1990b).3MacArthur Communicative Development Inventory (Lyytinen, 1999).4Utterance Analysis (Nieminen, 2007).

Early signs of dyslexia 367

Study 1: Processing of speech sound duration in

dyslexia (Richardson, 1998)

Study 1a: Categorization skills in 6-month-old infants

and their parents (Richardson, 1998)

Richardson’s (1998) study was centred around the

question of how speech sound duration affects the

way in which sounds are categorized in adults with

dyslexia as well as their 6-month-old children. The

interest in the linguistic function of duration stems

from the fact that one of the most distinctive features

of the Finnish language is phonological quantity, and

a determining factor in Finnish quantity is duration.

In other words, in Finnish phonology the meaning of

the word changes merely according to the length of

the phoneme, and the decisive factor for the

categorization of length phonemically (either short

or long) is the duration of the sound. It is important

to note that it is not the absolute duration of speech

sound alone that determines the quantity of the

sound but it is also the relative, proportional duration

of the sound in the context of other sound durations

within the word that determines if a sound is long or

short. For instance, a sound that we categorize as

short can be over twice as long in absolute (measur-

able) duration in the word final position in CVCV

structures in comparison to a sound in the same

word position but in CVCCV word structures and

yet they both are still categorized as short. Thus,

there can be almost an endless amount of variation in

speech sound durations but quantity categories or

degrees into which all the different physical durations

are linguistically placed, are limited to two (short and

long). Apart from the physical and relative aspects of

quantity, it should be noted that quantity categoriza-

tion is also subjective to a certain extent, since

speakers’ experience with the language affects their

linguistic judgements and, thus, durational differ-

ences are interpreted according to a subjective

reference code. In addition, it should be noted that

quantity is not only a feature of the spoken language

but in Finnish the two quantity degrees are also

explicitly marked in writing. The sounds that are

perceived and categorized as long are written with

two identical letters whereas short quantity is written

with one letter (e.g., kisa ‘‘cat’’ vs. kisa ‘‘competi-

tion’’). It is precisely in the quantity aspect of written

language that dyslexics make most of their reading

and spelling errors in Finnish (Lyytinen, Leinonen,

Nikula, Aro, & Leiwo, 1995).

A further impetus for Richardson (1998) to focus

on the durational aspect of speech came from the

evidence that school-aged children with language

disorders, including individuals with dyslexia, have

an auditory temporal deficit in perceiving speech as

well as non-speech sounds. Some of this evidence

has been obtained from a temporal order judgment

task in which individuals with dyslexia performed

significantly poorer than non-dyslexics when the rate

of stimulus presentation was fast (Tallal, 1980).

Apart from this and other similar evidence (see for

example, Farmer & Klein, 1995), there is also

contradictory evidence (e.g., Mody, Studdert-Ken-

nedy, & Brady, 1997), emphasizing the speech-

specificity of the possible differences or that dyslexia

does not involve temporal processing deficiencies at

all. In Richardson’s study (1998) several experi-

mental investigations were conducted in order to test

the temporal processing skills in both infants and

adults.

Participants

In the Richardson (1998) study adults with dyslexia

who also had at least one first- or second-degree

relative with dyslexia, as well as their children when

they were 6 months and 18 months of age, served as

participants in the experiments. The control groups

(CG) were formed from adults who themselves did

not have problems in reading or writing and their

children.

Procedure

Previous studies on infants have indicated that

infants as young as 6 months of age have the ability

to categorize speech sounds (see Eimas, 1996, for

review). Yet, there is little information on infants’

ability to categorize sounds according to their

prosodical features, and especially according to

duration. In Richardson’s study auditory perception

experiments were devised using a stimulus conti-

nuum in which the only parameter that changed in

the pseudoword context was the duration of the stop

sound (the occlusion part). The eight stimuli were

constructed by adding 20-msec increments of silence

to the middle of the t-sound. According to the pilot

studies on adult listeners, perception and categoriza-

tion changes sharply at some point in the stimulus

continuum from ata to atta (Richardson, 1998).

The categorization abilities of the infants were

assessed using a modified procedure of Kuhl’s

(1985) conditioned head-turn procedure. This relied

on infants’ instinct to turn their heads toward the

most interesting stimulus. Infants were conditioned

to turn their heads toward a visual reinforcer

whenever they perceived a change within the

auditory sequence from ata to atta. The experiment

on adults was exactly the same except the required

response was made by pressing a button when they

heard atta (Richardson, 1998).

Results

The data consisted of 133 adults’ and 105 infants’

performance in the categorization tasks. Results

showed that already at the age of 6 months infants

were able to categorize speech sounds according to

duration in similar fashion to adults (see Figure 1).

However, the data also showed that as a group in

368 U. Richardson et al.

comparison to controls both dyslexic adults as well as

their infants require longer duration in order to

change their perception from short to long quantity

category (Richardson, 1998; Richardson, Leppanen,

Leiwo, & Lyytinen, 2003). It should be noted,

however, that the data from adults clearly showed

that not all the responses of parents with dyslexia

differed significantly from the controls but the group-

level differences were due to the performance of

particular individuals. This uncertainty shown in

the categorization functions was reinforced also by

the reaction time data. All in all, the data point to the

conclusion that individuals with dyslexia can differ

from those with no dyslexia in processing temporal

auditory information of speech and these differences

in processing abilities can be present already at a very

early stage in development (Richardson, 1998).

Study 1b: Duration in the speech productions of 18-

month-old children (Richardson, 1998)

In addition to studying durational aspect in speech

perception, Richardson (1998) studied whether

adults with dyslexia and their children differ from

those individuals with no background of dyslexia in

the durational aspects of speech production.

Method

The task of the 18-month-old children was to imitate

the image of a woman in a video recording who

produced the pseudowords ata and atta as well as the

words mato (‘‘worm’’) and matto (‘‘carpet’’). The

parents’ productions of the words mato and matto

were studied from the situation in which they tried to

make their own children imitate the words after

them. The absolute as well as proportional durations

of the sounds were measured from the production

data. A total of 118 acceptable imitations from

children were included in the results (58 from the at-

risk for dyslexia group and 60 from the control

group). The adult data consisted of 104 words (44

words produced by individuals with dyslexia and 60

by control adults).

Results

The results showed that although the absolute

durations were relatively longer in the infants’

productions in comparison to those of the adults,

overall the proportional durations of the children’s

productions were similar to those of adults. This can

be taken to show that Finnish-speaking children as

young as 18 months of age can use and control

durations in speech production almost in the same

way as adults do. However, in some instances there

were significant differences. Although at the group-

level the children with a dyslexia background

managed to make clear durational differences in

their CVCV and CVCCV word structures similarly

to control children and adults, they did not manage

to do the same when they produced the pseudowords

ata and atta (see Table II). This finding suggests that

infants at-risk for dyslexia may indeed be unable to

use durational cues for quantity oppositions in

producing new words (here they were pseudowords)

indicating that they may have problems either

perceiving or paying attention or when utilizing

Figure 1. In the graph on the left the mean percentage of atta-categorizations in the infants of risk (RG, n¼ 54) and those without such risk

(CG, n¼51). There is a significant difference in the responses between the two participant groups in ata4 with the control infants

categorizing the stimulus more often as atta (ata4 d’ with one-tailed t-test, [96.60]¼71.8, p5 .04). In the graph on the right,

correspondingly, the categorization function for the dyslexic adults (DG, n¼57) and the control adults (CG, n¼76) in the perception

experiment. The functions of the two groups are similar in form but there is a statistical difference in the responses for the stimulus ata4 (d’

with one-tailed t-test, [99.82]¼72.0, p5 .03) in which dyslexic adults gave atta-responses significantly less often in comparison to the

control adults (Richardson, 1998, pp. 99, 124).

Early signs of dyslexia 369

temporal features in speech productions when they

do not yet have an existing representation of a word.

In addition, the children with a dyslexia background

produced durations which differed from the adult

productions more than those of the children with no

dyslexia background (see Figure 2). The data on the

adults’ naming task (see Table III) indicated that in

CVCV words the proportional measurements of

durations of both the first and the word-final vowel

in individuals with dyslexia differed from those of

adults without dyslexia (ANOVA, F 6.46, p¼ .014;

F 7.53, p¼ .008, respectively). This result could be

at least partly due to dialectical differences since

earlier findings have shown that there can be great

differences of durations in the word final vowel in

this particular word structure (Wiik, 1985). Alter-

natively, it could be speculated that these adults with

dyslexia are able to produce or concentrate on

producing an adequate difference in the place of

the major cue for the quantity distinction but do not

notice or cannot make the important durational

relationship adequately on the slightly subtler dura-

tional cue. Also it should be pointed out that since

the data were collected in the situation in which the

adult speakers used a typical child-directed naming

style including emphasized durational cues, the way

Table II. Descriptive data in milliseconds and in percentages of the average (C)VCV and (C)VCCV imitations produced by at-risk for

dyslexia children and children with no such risk. The produced target words were mato and matto and the pseudowords ata and atta

(Richardson, 1998, pp. 164–166).

(C)VCV (C)VCCV

Parameters N Mean SD vs. N Mean SD Sig.

Risk Group

M:ms 16 48 27 16 69 36 n.s.

A:ms 122 50 120 30 n.s.

T:ms 198 128 322 98 **

O:ms 209 106 195 110 n.s.

M:% 8 5 10 4 n.s.

A:% 21 8 17 5 *

T:% 34 16 46 8 *

O:% 36 12 28 8 *

A:ms 14 104 34 14 138 44 *

T:ms 233 170 264 137 n.s.

A:ms 171 60 184 59 n.s.

A:% 21 9 24 7 n.s.

T:% 46 18 45 15 n.s.

A:% 34 13 31 9 n.s.

Control Group

M:ms 12 47 20 12 52 25 n.s.

A:ms 98 38 100 32 n.s.

T:ms 168 19 352 41 **

O:ms 222 85 136 78 *

M:% 9 5 8 6 n.s.

A:% 18 7 16 6 n.s.

T:% 31 11 55 6 ***

O:% 42 13 21 5 ***

A:ms 17 111 58 17 107 39 n.s.

T:ms 137 59 310 163 ***

A:ms 177 83 138 49 n.s.

A:% 26 8 19 6 ***

T:% 32 9 56 12 ***

A:% 42 12 25 9 n.s.

Key: ms¼milliseconds, C¼ consonant, V¼ vowel. T-tests for paired samples were used for the calculations of statistical differences

(*p5 .05, **p5 .01, ***p5 .001).

Figure 2. The proportional durations of the mato word produc-

tions within dyslexia and the control families’ participants. The

graph indicates that the productions of the parents and infants

from the same families produce the mato word similarly

(Richardson, 1998, p. 175).

370 U. Richardson et al.

in which these adults considered the importance of

the cues should be reflected in the data. When the

production data of parents and infants from the same

families with dyslexia were compared (see Figure 2),

it is interesting to note that there are similarities

within families in the temporal aspects of the

productions, and as a whole the durational differ-

ences between segments were similar between the

parents and their children. These observations could

be interpreted as showing that there is a connection

between the productions within families. Since

speech communication is by nature social it is

believed that parental input has an effect on infants’

speech. The question whether the suggested relation-

ship between parents and infants is due to some

biological factor about muscular control or percep-

tual categorization skills or both, or whether this

relationship is due to the fact that parents provide a

strong model for infants at this age and the

interaction or negotiation of some aspects of speech

like temporality evolves in the particular social

contact between parents and infants, or whether

both genetic and social aspects affect the proposed

relationship remains to be answered in future

research.

Study 2: Phonological features in the speech of

24-month-old children at risk for dyslexia

(Torvelainen, 2007)

From the same JLD longitudinal study, Torvelainen

(2007) investigated variation in the phonological

development of the children when they were 24-

months-old (n¼ 39). Nineteen of the children

belonged to the at-risk group (RG) and all of them

had weaker phonological abilities at the age of 3.5

years. Of the 19 children 6 had significant problems

in literacy skills at the age of 8 (Torvelainen, 2007).

Method

The phonological development of the 24-month-old

children was studied from the spontaneous speech

data. These data were evaluated from three aspects:

from the point of view of the amount of the

unintelligible speech, attempted and produced words

of different length, and phonotactics (restricted to

C1V1C1V1, C1VC1V, CV1CV1 and complex C1V1

C2V2).1 A mean of the most advanced and the

weakest fifth (n¼ 8) was identified in each type of

evaluation (Torvelainen, 2007). The stage of the

children’s morphological development was measured

by the Mean Length of Utterance (MLU), and the

syntactical development with the total score of the

Finnish version of the Index of Productive Syntax

(IPS) (Nieminen & Torvelainen, 2003). The esti-

mate of the size of the child’s lexicon provided by the

children’s parents in the Finnish-language Mac-

Arthur Communicative Development Inventory

(MCDI) (Lyytinen, 1999) was used as the size of

the child’s lexicon (Torvelainen, 2007).

For the intelligibility measure, the data were

analysed for the number of intelligible and unin-

telligible utterances. An utterance was considered as

unintelligible when it contained at least one word for

which an adult who did not know the child could not

define the target. For the measure of attempted

words, the length of the words the children

attempted to produce as well as the phonotactic

complexity of these words were analysed. More

specifically, the data were analysed for the length of

the words that the children attempted to produce

Table III. Descriptive data in milliseconds and in percentages of the average CVCV and CVCCV productions produced by adults with

dyslexia and normally reading adults (controls). The produced words were mato and matto (Richardson, 1998, pp. 150–151).

(C)VCV (C)VCCV

Parameters N Mean SD vs. N Mean SD Sig.

Dyslexics 22 22

M:ms 74 31 85 26 n.s.

A:ms 103 34 96 20 n.s.

T:ms 127 32 359 95 ***

O:ms 134 67 97 24 n.s.

M:% 17 6 13 4 n.s.

A:% 23 7 15 3 ***

T:% 30 9 56 8 ***

O:% 29 11 15 4 ***

Controls 30 30

M:ms 64 28 67 26 n.s.

A:ms 84 42 87 20 n.s.

T:ms 118 28 319 95 ***

O:ms 163 71 88 24 ***

M:% 15 5 12 4 *

A:% 19 5 16 3 ***

T:% 29 7 56 8 ***

O:% 37 9 16 4 ***

Key: ms¼milliseconds, C¼ consonant, V¼ vowel. T-tests for paired samples were used for the calculations of statistical differences

(*p5 .05, **p5 .01, ***p5 .001).

Early signs of dyslexia 371

and the number of the attempted words which

consisted of only one consonant and one vowel.

For the measure of produced words, the number of

words produced which were different from the

targets in length as well as how these words were

different from the targets were analysed. In addition,

the produced words were analysed to ascertain how

many of them were produced in a phonotactically

simplified form having only one consonant and a

vowel (Torvelainen, 2007).

Results

The results showed that the variation in phonological

development is considerable at age two in general. In

the most advanced fifth, the children’s speech was

highly intelligible (96%), whereas in the weakest fifth

75% of the speech was intelligible. The most

advanced children attempted both short and long

words, whereas the weakest fifth only attempted

words with one or two syllables. Of the targets, 35%

were phonotactically restricted in the most advanced

fifth, whereas 69% of the targets were restricted in

the weakest fifth. The most advanced children had

errors in the length of trisyllabic words (.3%),

whereas the weakest fifth had errors in both short

and long words (10%). Three percent of words

produced in the most advanced fifth were phono-

tactically restricted, compared with 26% in the

weakest fifth. The correlations between the different

areas of grammar indicated that in some respects the

level of phonological development predicts the level

of morphological and syntactical development (at-

tempted words and morphological development with

Pearson’s correlation was .895***, attempted words

and syntactic development .871***).

The evaluation of phonological development was

based on the supposition that children’s words

develop from forms that are restricted in length in

prosodic minimal words (e.g., Pater, 1997; Kehoe,

2000) (mono- and disyllables) and phonotactics

(C1V1C1V1, C1VC1V, CV1CV1), into adult forms,

in which, for example extrametric syllables and the

combination of various consonants and vowels (e.g.,

C1V1C1V1,) take place. In such cases the most

restricted forms (mono- or disyllables and words

adhering to complete consonant and vowel har-

mony) could be unintelligible. The analyses showed

that at- risk children’s phonological skills were less

advanced than those of the control group children in

all the different aspects analysed in Torvelainen’s

(2007) study, but not all the differences were

statistically significant (see Table IV).

Phonological development of control children and at-risk

children

The amount of unintelligible speech and the amount

of unintelligible phonotactically restricted words

were significantly larger and the amount of at-

tempted mono- as well as trisyllabic words was

significantly smaller in the at-risk children’s produc-

tions. Although the differences were quite clear at the

group level, in terms of individuals there were also at-

risk children among both the most and least

advanced fifths. Since the individual variation is

large at this early age, it is not feasible to predict

individual children’s language development based on

Table IV. The group means of the phonological features produced by 2-year-old children at risk for dyslexia (at-risk group, RG) and those

without such risk (control group, CG) (Torvelainen, 2007, p. 161).

Control Children

(CG) (n¼20)

At-risk Children

(RG) (n¼ 19) Sig.

Amount of unintelligible speech (CG f¼1461, RG f¼ 1331) 9.6% (3.3) 14.1% (5.9) *

Amount of attempted words of different length1

(CG f¼ 2277, RG f¼1758):

Foursyllables 3.0% (4.0) 2.5% (2.8) n.s.

Trisyllables 11.5% (11.2) 5.5% (4.6) **

Disyllables 53.9% (27.4) 66.6% (23.6) n.s.

Monosyllables 30.8% (20.0) 25.3% (11.6) *

Amount of words altered in length:

Foursyllables (CG f¼68, RG f¼44) 13.2% (1.4) 31.8% (2.1) n.s.

Trisyllables (CG f¼ 262, RG f¼97) 8.8% (1.3) 16.5% (1.4) n.s.

Disyllables (CG f¼1228, RG f¼ 1171) 2.0% (3.3) 2.0% (1.3) n.s.

Monosyllables CG f¼ 2277, RG f¼1758) .0% (0) .7% (.5) n.s.

Amount of phonotactically restricted2 attempted words

(CG f¼ 2277, RG f¼1758)

50.6% (12.6) 53.5% (11.7) n.s.

Amount of phonotactically restricted2 unintelligle words

(CG f¼ 109, RG f¼178)

63.3% (3.0) 65.2% (4.5) *

Amount of phonotactically restricted2 words

(CG f¼ 2277, RG f¼1758)

10.8% (10.0) 11.3% (6.2) n.s.

1 Longer words than foursyllabses (f¼19) were rare and therefore excluded from analyses.2(C1V1C1V1, C1VC1V, CV1CV1).

Standard deviations in parentheses.

Independent T-tests were used for the calculations of statistical differences (*p5 .05, **p5 .01, ***p5 .001).

372 U. Richardson et al.

merely one or two measures on phonological

features. However, the measures undertaken at the

early stages of language development can be con-

sidered as indicative but wider scope (i.e., more

features and follow up from a longer period of time)

is needed in order to gain a more reliable under-

standing of the development. In fact, the preliminary

analyses of the data when the same children were

older show that the large number of phonotactically

restricted unintelligible words at the age of 2 was a

precursor of poor reading skills at the age 8 in six of

nine individuals. The phonological development of

two of these six children showed surprisingly

individual paths, from which there was disproportion

between different levels of grammar, particularly

since the size of these children’s lexicon was large.2

In other words, they had advanced in the lexical

level, but not in phonotactical level of words.

However, in terms of the whole group (n¼ 39) the

results indicate that there may be a link between

lexicon and the children’s phonological/phonetic

development (for similar conclusions see Smith,

McGregor, & Demille, 2006). In Torvelainen’s

study there were statistically significant positive

correlations between the size of the lexicon and the

attempted words of different lengths (Pearson’s

correlation .639***) and a significant negative

correlation between the size of the lexicon, number

of phonotactically restricted target forms (7.470**)

and the number of phonotactically restricted unin-

telligible forms (7.459**). Typically, therefore,

children with a small lexicon attempted short and

phonotactically simple words. These results indi-

cated that a long-term assessment comparing

phonological development with the size of the

lexicon as opposed to the age of the child is needed

in order to determine early precursors of difficulties

in phonological development of dyslexia. Torvelai-

nen (2007) concludes that in particular in the case

of at-risk children it is advisable to pay attention to

language development if the child’s vocabulary is

large but the word forms are phonotactically

restricted and if the child does not even attempt

to produce longer and phonotactically more com-

plex word forms in spontaneous speech. It has

been noticed also in experimental settings that if a

child does not produce or even attempt to produce

words this can be a sign of deviant language

development (Leiwo & Kulju, 2003; Leiwo, Tur-

unen, & Koivisto, 2002).

Study 3: Phonological features in the speech of

30-month-old children at risk for dyslexia

(Turunen, 2003)

Method

The relationship between early phonological skills

and later reading skills was also studied by Turunen

(2003, currently Kulju), but when the children were

slightly older. In her retrospective study of 30-

month-old at-risk (n¼ 105) and control children

(n¼ 91) the focus was on the hierarchical aspect of

word structure: children’s word structures were

analysed within an autosegmental framework (Gold-

smith, 1976) by applying prosodic hierarchy (Selkirk,

1980). Thus, in addition to the analysis of individual

phonemes in children’s word productions, also

phonotactical elements, (e.g., consonant sequences)

as well as syllable and word length were taken into

account.

The data were collected from a picture-naming

task which comprised altogether 33 objects. Chil-

dren named each object twice, first, using a picture

book and, secondly, using stickers. Nineteen words

with varying word structures were selected to be

transcribed for further analyses from almost 200

children. Phonological scores of the children’s

productions were calculated and the overall tran-

scription reliability score was 89.9%. In the

following, we will briefly describe a hypothetical

model on the acquisition of word structures in

Finnish-speaking children. This model served as a

basis for phonological scoring and group com-

parisons. As an example, the following illustrates

the hypothetical acquisition path of a trisyllabic

word porkkana ‘‘carrot’’, in the productions of one

child:

1. Word level (syllable number) pokka

not complete:

2. Syllable level (syllable heaviness) pokkana

not complete:

3. Phonotactics (cons. sequence) pookkana

not complete:

4. Phoneme level (/r/) polkkana

not complete:

5. A target-like word structure: porkkana

According to this model, it is assumed that prosodic

elements govern the phonotactical and phoneme

level in the acquisition of word structures. In the

above example this means that at first the child

should overcome the constraints preventing trisylla-

bic word structure producing instead a trochaic,

bisyllabic structure. As the word length in the

number of syllables is acquired, the child proceeds

to a syllable level in which the acquired syllable

length serves as a basis for more complex phonotac-

tical elements (e.g., diphthongs and complex con-

sonant sequences). The last stage concerns the

finalization of a phoneme inventory, which in the

above example means the acquisition of the trilled

/r/-phoneme. As mentioned above, the acquisition

path can also be illustrated within a constraint-based

account, such as by using the constraint definitions

of Optimality Theory (see Bernhardt & Stemberger,

1998; Kager, 1999; Turunen, 2003).

The cross-sectional data of 30-month-old Finnish-

speaking children supported this hypothetical view

Early signs of dyslexia 373

since the presented word types were common in the

data, and also the subgroup of late talkers produced

the less advanced word structures. Thus, when

investigating phonological acquisition, the interplay

of phonemes, phonotactics and syllables should be

taken into account (Turunen, 2003, see also Kulju &

Savinainen-Makkonen, 2008; Savinainen-Makko-

nen, 1996).

In studying the at-risk and control children, the

following phonological parameters were used: pro-

duction of three and four syllabic words, heavy

unstressed syllables, consonant sequences and

diphthongs, and the individual phonemes /s/ and /r/.

Production skills of long words were studied because

they are common in Finnish due to inflection

(morphemes are added into a word stem), but they

may be prosodically challenging in early stages of

phonological development for which it is typical for

Finnish children to truncate them. In addition to the

above mentioned porkkana-example, typical produc-

tion types in the data were for example [auli] /auri˛ko/

aurinko ‘‘the sun’’, [ova] /orava/ orava ‘‘squirrel’’,

[apsi:ni] /ap:elsi:ni/ appelsiini ‘‘orange’’. In Finnish, the

word stress is in the word initial syllable even if the

second syllable is heavy as in /avain/ avain ‘‘key’’. This

word was typically produced with a short unstressed

syllable as in [ava]. Consonant sequences and

diphthongs were in focus in the level of phonotactics.

In the early stages of development the children

typically assimilate consonant sequences as was seen

in the data; for example, [sak:e] /sakset/ sakset

‘‘scissors’’, [suk:a:] /sukla:/ suklaa ‘‘chocolate’’. Re-

spectively, diphthongs are often produced by length-

ening the vowel as in [a:liko] /auri˛ko/ aurinko ‘‘the

sun’’. Phonemes /s/ and /r/ were chosen because they

are often among the last consonants to be acquired in

Finnish. The data showed typical substitution types,

e.g., [tat:et] /sakset/ sakset ‘‘sakset’’, [olava] /orava/

orava ‘‘squirrel’’. (For details see Turunen, 2003,

and of the phonological development of Finnish

children in general see also Kunnari & Savinainen-

Makkonen 2007; Kulju & Savinainen-Makkonen,

2008).

Results

Against expectations, no clear differences were found

between the at-risk and control children at this age in

the phonological quality of their word productions. It

should be noted, however, that only about half or

fewer of the children in the at-risk group were

expected to show deficits in their later reading

performance. In addition, it may be that the

familiarity of the words used in the naming task

affected the results. Even though the task did reveal

differences in the phonological development of these

children, it may be, that it was not demanding

enough to reveal possible group differences. As seen

earlier in this paper, the more demanding experi-

mental methods (e.g., a head-turn paradigm) were

able to reveal both group differences and significant

predictions to later language skills at a very early age

(see Lyytinen et al., 2004; Turunen, 2003, for

discussion.)

At the current point of the longitudinal data

collection, it is possible to identify those at-risk

children who actually have reading difficulties. The

preliminary results show that, indeed, children with

dyslexia were not as advanced at the age of 30

months as those children with no reading/writing

problems in the production of some prosodic aspects

of a word structure, such as in producing four

syllable words (Kulju, 2007).

Phonological features in the speech of poor, intermediate

and good readers at the age of 30 months

The relationship between early phonological ability

and later reading skills was further examined in a

retrospective analysis of poor (n¼ 38), intermediate

(n¼ 36) and good readers (n¼ 40) who were

identified at school entry based on Lindeman’s

(2000) test measuring word recognition skills. A

comparison of the three groups revealed significant

differences in early phonological skills: children who

were identified as good readers at school entry had

been more advanced than intermediate or poor

readers especially in their production of word length

and syllable structures at the age of 30 months. On

average, poor readers produced 4.1 four syllable

target words out of the maximum 8, and 3.2 of these

were realized as four syllables (78%). Good readers

produced 5.5 four syllable targets out of which 4.8

were realized as four syllables (87%). Respectively,

poor readers produced 6.6 out of the 12 target words

with heavy unstressed syllable, and on average in 4.3

(65%) of these productions the heavy unstressed

syllable was realized as heavy. Good readers pro-

duced 8.4 of this type of targets and 6.5 (77%)

included a heavy unstressed syllable. In both para-

meters the results of intermediate readers were placed

in between the poor and good readers. These results

indicate a correlation between early reading skills and

the acquisition pace of phonological structures (for

details, see Turunen, 2003; Kulju, 2003).

Overall the results of Turunen’s (2003) study

showed that at age of 30 months 61% of the nearly

200 Finnish children managed to produce all four

syllable-words as four syllables whereas 9% of the

children were not able to produce four syllable

structures at all. Similarly, trisyllabicity was correct

in 60% of the children’s productions whereas 5%

did not produce any trisyllabic structures in the

naming task. All in all, heavy unstressed syllables

were always correctly produced by 35% of the

children while only 5% never produced them (for

more details see Turunen, 2003). It seems that

especially these features may be more problematic

for children with dyslexia in their early phonological

acquisition.

374 U. Richardson et al.

The above results were based on large cross-

sectional data with a large number of children. These

types of data reveal the high variation of phonological

stages at the age of 30 months. However, the amount

of data from individual children is limited. In further

studies our aim is to investigate phonological acquisi-

tion from an intra-individual aspect by comparing

different types of data. One goal is to attempt to

define normal and abnormal variation in word

production. For example, as the child corrects his/

her word structure spontaneously, (e.g., [ovava,

orava] /orava/ orava ‘‘squirrel’’ in a naming task, this

can be taken as showing typical variation. In this

specific example the child is proceeding to produce the

/r/-phoneme instead of substituting it by assimilation. It

may be that if a child with dyslexia has problems in

perceiving the exact form of word structure; this results

in a high level of variation in his/her output. The

following examples illustrate this type of abnormal

variation (examples are from two 30-month-old

children who were later identified as dyslexic):

[puleli][puleli] (high /puhelin/ puhelinintonation) [polel:e] ‘‘telephone’’[hapi:e], [api?a] /lapio/ lapio ‘‘shovel’’[u:hto], [po?hto:] /juusto/juustoo

/juustoo’cheeseþPART’[ousut soustu tossa] /housut housut tos:a/

housut housuttossa ‘‘trousers trousersthere’’

It is evident that the definition of abnormal variation

in intraindividual speech production is demanding,

but by focusing on this we may be able to find

out more about early precursors of dyslexia

and the hypothesis concerning fuzzy phonological

representations.

Study 4: Morphosyntactic features in verbal

productions of 30-month-old children at risk

for dyslexia (Nieminen, 2007)

Although problems in phonological processing have

been strongly emphasized in many studies on

dyslexia, there are also several occasions where

morphological and syntactic skills are considered as

potential areas for precursors of dyslexia to appear.

In her retrospective study on spontaneous speech

produced by individuals with dyslexia and their

controls (typically reading individuals), Scarborough

(1990a; 1991) discovered that at 30–48 months of

age control children produced longer utterances than

children with dyslexia. In addition, their utterances

were syntactically more complex in terms of IPSyn

analysis. Also studies concerning inflectional (Joa-

nisse, Manis, Keating, & Seidenberg, 2000) and

derivational processes (Leong & Parkinson, 1995)

have shown that there are some differences in

morphological processing between individuals with

dyslexia and control subjects.

There are at least three different hypotheses which

attempt to explain the connections between syntactic

skills and dyslexia. One of these hypotheses claims

that there are deficiencies in basic syntactic skills

which are due to either (a) delayed language

development or (b) a structural deviation in the

linguistic system. A second hypothesis presumes that

dyslexic and non-dyslexic people have similar syn-

tactic skills but production differences found in

several investigations are due to limitations in the

capacity of short term memory which are caused by

problems in phonological processing (cf. Deutch &

Bentin, 1996; Leikin & Assayag-Bouskila, 2004).

The third view claims that both individuals with

dyslexia and their controls have similar basic

syntactic skills, but people with dyslexia are not able

to use these skills as efficiently as the controls.

Phonological processing abilities can not explain this

alone. Instead, the underlying deficiency must be a

broader problem of metalinguistic skills (Deutch &

Bentin, 1996).

Method

In Nieminen’s (2007) study spontaneous speech

performances of 20 risk group (RG) children and 20

(CG) children were compared at the age of 30

months. The main focus of the comparison was in

mophosyntactic complexity of the children’s utter-

ances. Following Scarborough’s (1990a) methods

Nieminen used the Mean Length of Utterance

(MLU; Brown, 1973) and Index of Productive

Syntax (IPSyn; Nieminen & Torvelainen, 2003;

Scarborough, 1990b) as a preliminary measure of

complexity. MLU measures the linear length of

utterances using a morpheme as a unit of measure-

ment whereas IPSyn credits with points different

morphosyntactic structures used by a child. Both

MLU and IPSyn were compatible with the view of

absolute complexity (cf. Dahl, 2004; Miestamo,

2006) based on information theory, since both of

them measure the number of units in a system

(Nieminen 2007; Nieminen, submitted). In Niemi-

nen’s study, the 80 longest intelligible utterances

produced by each child were analysed with MLU

and IPSyn.

Results

Both MLU and IPSyn showed parallel results in both

groups. The control group performed better than the

risk group. This was shown by group mean values in

Table V. Although the mean differences were clearly

different it is important to pay attention to standard

deviations in both groups. It is the high standard

deviations that prevent the differences in group

comparison from becoming statistically significant.

When individuals in the RG were compared to

individuals in the CG the similarity of children in

both groups is emphasized (Figures 3 and 4). It can

Early signs of dyslexia 375

be stated that in both groups utterances of the same

length are produced and that children in both groups

use a similar number of different morphosyntactic

structures to build utterances. The group differences

seen in average values (Table IV) can be explained by

the performance of seven children in the RG whose

performance show clearly lower skills than other

children in this study (Nieminen, 2007).

MLU and IPSyn enable analysis of children’s

spontaneous speech productions from two different

points of view. However, neither of them is capable

of detecting the whole morphosyntactic structure of

any individual utterance. In languages such as

Finnish which are morphosyntactically rich and

versatile, the evaluation of the absolute complexity

of utterances requires that the morphosyntactic

complexity is considered as a multidimensional

property of the language. Therefore, relying on either

the linear length of utterances like in MLU or the

inventory of different structures used in utterances

like in IPSyn is not enough. In Nieminen’s study

(2007) the multidimensional analysis of utterances

was executed by Utterance Analysis (UA).

In Utterance Analysis each utterance is divided

into components and each component further into

smaller morphosyntactic parts like words and gram-

matical morphemes. An example of UA and its

use in analysing a single utterance is provided in

Figure 5. In UA a component is a linguistic unit

which consists of related parts such as a word stem

and an ending or a head word and a modifier. The

inner structure of components and especially elabora-

tion devices that are utilized to construct the utterance

are important factors in evaluating the structural

complexity of an entire utterance. The two basic

elaboration devices which can be utilized are

morphological and syntactic elaboration. Morpholo-

gical elaboration includes the use of suffixes and

syntactic elaboration refers to structures where a

head word occurs together with a modifier or

determiner, an auxiliary co-occurs with the main

verb, or a noun and an adposition (i.e., prepositions

or postpositions) coexist. A syntactic component can

be an unelaborated, single uninflected word with no

modifiers or determiners. Syntactic and morpholo-

gical elaboration can also both exist in the same

Table V. MLU-values and IPSyn-scores in the control (CG) and risk groups (RG) (Nieminen, 2007, pp. 64, 67).

CG RG

n Mean SD n Mean SD

MLU

All 20 5.09 1.53 20 4.40 1.75

Girls 8 5.30 1.91 10 4.84 2.03

Boys 12 4.69 1.29 10 3.96 1.39

IPSyn

All 20 63.80 16.86 20 57.50 18.32

Girls 8 63.25 24.41 10 59.90 20.58

Boys 12 64.17 10.56 10 55.10 16.50

Figure 3. The MLU values according to the 80 longest intelligible utterances produced by children in the risk (RG) and control groups (CG)

(Nieminen, 2007, p. 63).

376 U. Richardson et al.

syntactic component. On the other hand, compo-

nents with no elaboration are very common as well.

Each morphological or syntactic addition creates a

new elaboration layer. The more there are elabora-

tion layers the more complex the component is

structurally (Nieminen, 2007).

When the syntactic components produced by

children in the RG and in the CG were compared,

it was found that the groups are very similar.

Children in the CG produced 24 different compo-

nent types and in the RG produced 25 of different

types. In both groups the component types repre-

sented similar levels of structural complexity. The

only difference was that in the CG there were four

different types of component each consisting of five

elaboration layers whereas in the RG there was

only one component type consisting of five layers

(Nieminen, 2007). Based on these findings it can be

claimed that children in RG and CG have similar

morphosyntactic skills, if the skills were evaluated in

terms of structural complexity of individual compo-

nent types. These results support the two hypotheses

mentioned earlier which claim that dyslexic and non-

dyslexic individuals have similar syntactic skills.

However, these results do not yet give any clarifica-

tion of how these skills are used and, therefore, it can

not be concluded which one of the hypothesis

candidates is more accurate. To investigate this, the

whole utterance structure must be considered.

In order to find out the structural complexity of

the entire utterances the composition of the utter-

ances produced by the RG and the CG children were

analysed in terms of different component types. The

analysis showed that children in the CG combined

elaborated components more in their utterances than

the RG children. For example, 14.8% of utterances

which consisted of two components were unelabo-

rated in the RG whereas in the CG the proportion of

unelaborated 2-component utterances was only

6.5%. Risk Group children seemed to prefer also 2-

component utterances with one elaborated compo-

nent. The proportion of these utterances was 62.4%

of all 2-component utterances in the RG and 53.7%

in the CG. In this case the group difference was also

statistically significant: The Mann-Whitney U was 34

(z¼72.03) with an associated probability of .027

(two-tailed). However, when the combination of two

elaborated components was considered CG children

performed better than RG children. The proportion

of utterances with two elaborated components was

39.8% out of all 2-component utterances in the CG

and 22.8% in the RG. The CG’s emphasis on com-

bining several elaborated components within the

same utterance was seen also in 3- and 4-component

utterances, although the difference between the

groups became greater when the number of compo-

nents increased (Nieminen, 2007). However, the

only statistically significant difference between the

productions of the two groups was in 3-component

utterances in which all three components were

elaborated (U¼ 36.5 (z¼72.03), p¼ .042). Thus, it

seems that the difference between the CG and the RG

lies in the structural complexity of the entire utterances,

that is, in how many elaborated components can be

Figure 4. The IPSyn scores in ascending order according to the 80 longest intelligible utterances produced by children in the risk (RG) and

control groups (CG) (Nieminen, 2007, p. 67).

Figure 5. An example of Utterance Analysis conducted on a 4-

component (ADVP, V, NP and NP) utterance with 3 elaboration

layers. The last NP is an example of morphologically elaborated

component, and verb construction (V) represents a component

including both syntactic and morphological elaboration.

Early signs of dyslexia 377

combined together in an utterance. This result

supports the hypothesis by Deutch and Bentin

(1996). This hypothesis suggested that the basic

syntactic skills of dyslexic and non-dyslexic children

do not differ from each other but the group differences

are due to varying abilities to use these skills.

The results from the Utterance Analysis reported

here are compatible with the MLU and IPSyn results

presented earlier (Nieminen, 2007). Except for the

results from the seven children with the lowest

performance in the RG, the children in the RG and

CG proved to be very similar in terms of IPSyn

results. This implies that the children have structures

of similar complexity especially on the component

level. However, the MLU results showed that

children in the RG produced utterances that have

fewer morphemes than utterances of the CG chil-

dren. The RG children’s shorter utterances are a

corollary of the lower number of elaborated compo-

nents when compared to the utterances produced by

the CG children.

Conclusion

The studies reviewed here on early language devel-

opment and dyslexia have been made possible only

with a large amount of longitudinal data from

different stages of language development as well as

with further development of research methods.

Besides collecting data from experiments employing

typical experimental methods including a head-

turning experiment, data have also been collected

from picture-book naming and structured play

situations. By having different types of data we have

been able to gather more comprehensive information

on the speech and language development of children

with and without risk for dyslexia. Until recently,

much of the existing information on dyslexia is based

on data and methods used in psychology, and most

often the information comes from the English

language context. Here we developed linguistic

models and methods for analyses in order to study

and recognize early signs of dyslexia from Finnish-

speaking children. Apart from group comparisons on

quantified data, early language development has

been analysed qualitatively according to linguistic

theories, especially on the phonological and mor-

phosyntactical levels. At the same time as we have

more information on possible early signs of dyslexia

we also have gained a better understanding of the

typical language development of children during the

first years of life. In addition, we have created

methods for analyses that can be used in other

language studies, not only in studies of language

development (see Nieminen & Torvelainen, 2003).

The reviewed main findings of the four studies

(Richardson, 1998; Torvelainen, 2007; Turunen,

2003; Nieminen, 2007), provide us with grounds for

stating that possible signs of dyslexia can be detected

already in the very early stages of phonological and

morphosyntactical development. These signs can be

seen in perception or production of duration in

speech, in intelligibility of utterances, in the prosody

and phonotactics of attempted words and word

structures, and the morphosyntactical complexity of

utterances. All these linguistic research findings

support the theory according to which individuals

with dyslexia can have difficulties in creating

accurate representations from spoken words. In

addition, such individuals might not be able to

utilize their syntactical skills as well as individuals

without such a risk for dyslexia.

The finding of group differences in different

linguistic features is a significant discovery that can

also be utilized in practice. This type of information

is vital in order to create effective training programs

with appropriate linguistic content for assisting very

young children who most likely will encounter

difficulties in reading and spelling at school. It is

possible that these very young children would benefit

already before entering school from training with the

specific language features included in the analysis of

the reviewed studies. However, it is vital to note that

children with a family history of dyslexia did not

differ even in the group level from the control

children on all of the analysed speech and language

features (no differences in phonological quality of the

produced words, and no differences in MLU and

IPSyn measures separately). At the same time it is as

important to recognize that all in all there was wide

individual variation within the data which is very

typical of the early stages of language development

(see e.g., McIntosh & Dodd, 2008; McLeod &

Hewett, 2008). Therefore, we believe that it is not

possible to draw far-reaching conclusions according

to a single research finding, for example, that making

errors in durational aspects of speech according to

the standard view of language would be a certain sign

of dyslexia. Or vice-versa, it would not be well

founded to state that if a child manages to produce

correct word structures and utterances in early

language development that this particular child

would not have problems in reading and writing in

the future. Further investigation of individual chil-

dren’s performances is needed before such far-

reaching conclusions can be made.

The studies reviewed here suggest that prosodic

features, phonotactics and complexity of morpho-

syntactic features in language development in dys-

lexia should be examined in more detail in the future

at least in languages such as Finnish, but these could

also be key elements in language development in

general. For instance, there is evidence that a

prosodic feature such as the perception of duration

(using similar speech stimuli as was used here in the

perception experiments) is different in children with

dyslexia and that this skill predicts reading and

spelling skills in English (Richardson, Thomson,

Scott, & Goswami, 2004). Our intention is to

continue to examine speech and language-specific

378 U. Richardson et al.

skills of these children in various tasks on an

individual level from different stages of develop-

ment as well as to connect these to the other skills

that the children have. In this way, we hope to gain

an overall picture of possible individual paths and

the skills that are either critical for achieving or

hindering good reading and spelling skills. For all

this, the JLD project provides an excellent oppor-

tunity to explicate dyslexia since over 10 years the

same children’s skills have been studied in many

different types of tasks.

Acknowledgements

We thank the families participating in The Jyvaskyla

Longitudinal Study on Dyslexia (JLD). The JLD

project belongs to the Finnish Centre of Excellence

Program (1995–1999, 2000–2005, 2006–2011) and

has been supported by the Academy of Finland,

Niilo Maki Institute and University of Jyvaskyla.

Notes

1 The CVCV structures form prototypical Finnish words. Here

the same subscript indicates that the phonemes are exactly the

same. In cases where there is no subscript the phonemes are

different. Thus, C1V1C1V1 refers to a structure with complete

harmony, as in the word pupu ‘‘bunny’’. C1VC1V refers to a

form with consonant harmony, as in the word mummo

‘‘granny’’. CV1CV1 refers to a form with vowel harmony, as

in the word vene ‘‘boat’’. The complex C1V1 C2V2 refers to a

structure which has no harmony, as in the word kuppi ‘‘cup’’.

2 The size of lexicon of these two children was 103 and 241.

Despite large lexicons, these children produced phonotactically

restricted word forms.

References

Bernhardt, B. H., & Stemberger, J. P. (1998). Handbook of

phonological development from the perspective of constraint-based

nonlinear phonology. San Diego, CA: Academic Press.

Bishop, D. V. M., & Snowling, M. J. (2004). Developmental

dyslexia and specific language impairment: Same or different?

Psychological Bulletin, 130, 858–886.

Bradley, L., & Bryant, P. E. (1983). Categorizing sounds and

learning to read: A causal connection. Nature, 30, 419–421.

Brown, R. (1973/1976). A first language. The early stages.

Harmondsworth: Penguin Education.

Bruck, M. (1992). Persistence of dyslexics’ phonological aware-

ness deficits. Developmental Psychology, 28, 874–886.

Dahl, O. (2004). The growth and maintenance of linguistic complexity.

Philadelphia, PA: John Benjamins.

Deutch, A., & Bentin, S. (1996). Attention factors mediating

syntactic deficiency in reading-disabled children. Journal of

Experimental Child Psychology, 63, 386–415.

Eimas, P. D. (1996). The perception and representation of speech

signals by infants. In J. L. Morgan, & K. Demuth (Eds.), Signal to

syntax: Bootstrapping from speech to grammar in early acquisition (pp.

25–39). Mahwah, NJ: Lawrence Erlbaum Associates.

Farmer, M. E., & Klein, R. M. (1995). The evidence for a

temporal processing deficit linked to dyslexia: A review.

Psychonomic Bulletin and Review, 2, 460–493.

Goldsmith, J. (1976). Autosegmental phonology. New York: Garland

Publishing.

Hoien, T., Lundberg, L. Stanovich, K. E., & Bjaalid, I. K. (1995).

Components of phonological awareness. Reading and Writing,

7, 171–188.

Joanisse, M. F., Manis, F. R., Keating, P., & Seidenberg, M. S.

(2000). Language deficits in dyslexic children: Speech percep-

tion, phonology, and morphology. Journal of Experimental Child

Psychology, 77, 30–60.

Kager, R. (1999). Optimality theory. Cambridge: Cambridge

University Press.

Kehoe, M. (2000). Truncation without shape constraints: The latter

stages of prosodic acquisition. Language Acquisition, 8, 23–67.

Kuhl, P. K. (1985). Methods in the study of the infant speech

perception. InG.Gottlieb, &N.A.Krasnegor (Eds.),Measurement

of audition and vision in the first year of postnatal life: A methodological

overview (pp. 223–251). Norwood, NJ: Ablex.

Kulju, P. (2003). 2;6-vuotiaiden lasten sanarakenteiden tuottami-

sesta ja varhaisen fonologian yhteydesta lukihairioriskiin ja

lukutaidon oppimiseen [Production of word structures in 2;6-

year-old children and the role of phonology in the risk for

dylsexia and reading acquisition]. Puhe ja kieli, 23, 173–188.

Kulju, P. (2007). Early phonological development of dyslexic children.

Paper presented in XI Nordic Child Language Symposium,

Oulu, January.

Kulju, P., & Savinainen-Makkonen, T. (2008). Normal and

disordered phonological acquisition in Finnish. In A. Klippi, &

K. Launonen (Eds.), Research in logopedics. Speech and language

therapy in Finland (pp. 35–62). Clevedon: Multilingual Matters.

Kunnari, S., & Savinainen-Makkonen, T. (2007). Finnish speech

acquisition. In S. McLeod (Ed.), The international guide to

speech acquisition (pp. 351–363). Clifton Park, NY: Thomson

Delmar Learning.

Landerl, K., Wimmer, H., & Frith, U. (1997). The impact of

orthographic consistency on dyslexia: A German-English

comparison. Cognition, 63, 315–334.

Leikin, M., & Assayag-Bouskila, O. (2004). Expression of

syntactic complexity in sentence comprehension: A compar-

ison between dyslexic and regular readers. Reading and Writing:

An Interdisciplinary Journal, 17, 801–821.

Leiwo, M., Turunen, P., & Koivisto, J. (2002). Suomen

vokaaliharmonian kehitys –vaikeasti kuvattava vaikeus? [Acqui-

sition of Finnish vowel harmony – a difficulty which is difficult

to describe] Puhe ja kieli, 22, 105–112.

Leiwo, M., & Kulju, P. (2003). Fonologisen kehityksen teoreettinen

kuvaus [Theoretical description of phonological development]. In

S. Kunnari, & T. Savinainen-Makkonen (Eds.), Mista on pienten

sanat tehty. Lasten aanteellinen kehitys (pp. 134–146). Helsinki:

WSOY.

Leong, C. K., & Parkinson, M. E. (1995). Processing of English

morphological structure by poor readers. In C. K. Leong, &

R. M. Joshi (Eds.), Developmental and Acquired Dyslexia

(pp. 237–261). Dordrecht: Kluwer.

Lindeman, J. (2000). Ala-asteen lukutesti: Kayttajan kasikirja

[Reading test for elementary schools: A user’s guide].

Oppimistutkimuksen keskus, University of Turku.

Locke, J. L., Hodgson, J., Macaruso, P., Roberts, J., Lambrecht-

Smith, S., & Guttentag, C. (1997). The development of

developmental dyslexia. In C. Hulme, & M. Snowling (Eds.),

Dyslexia: Biology, cognition and intervention (pp. 72–96).

London: Whurr Publishers.

Lyytinen, H., Leinonen, S., Nikula, M., Aro, M., & Leiwo, M.

(1995). In search of the core features of dyslexia: Observations

concerning dyslexia in the highly orthographically regular

Finnish language. In V. W. Berninger (Eds.), The varieties of

orthographic knowledge II: relationships to phonology, reading, and

writing (pp. 177–204). Dordrecht: Kluwer Academic Publishers.

Lyytinen, H., Ahonen, T., Eklund, K., Guttorm, T., Kulju, P.,

Laakso, M.-L., Leiwo, M., Leppanen, P., Lyytinen, P., Poikkeus,

A.-M., Richardson, U., Torppa, M., & Viholainen, H. (2004).

Early development of children at familial risk for dyslexia: follow-

up from birth to school age. Dyslexia, 10, 146–178.

Lyytinen, P. (1999). Varhaisen kommunikaation ja kielen kehityksen

arviointi-menetelma [An assessment method for early commu-

nication and language development]. Jyvaskylan yliopiston

lapsitutkimuskeskus ja Niilo Maki Instituutti. Yliopistopaino,

Jyvaskyla.

Early signs of dyslexia 379

McIntosh, B., & Dodd, B. J. (2008). Two-year-olds’ phonological

acquisition: Normative data. International Journal of Speech-

Language Pathology, 10, 460–469.

McLeod, S., & Hewett, S. R. (2008). Variability in the production

of words containing consonant clusters by typical two- and

three-year-old children. Folia Phoniatrica et Logopaedica, 60,

163–172.

Miestamo, M. (2006). On the feasibility of complexity metrics. In

K. Kerge, & M.-M. Sepper (Eds.), FinEst linguistics. Proceedings of

the Annual Finnish and Estonian Conference of Linguistics (pp. 11–

26). Tallinn: Department of Estonian of Tallinn University.

Mody, M., Studdert-Kennedy, M., & Brady, S. (1997). Speech

perception deficits in poor readers: Auditory proceessing or

phonological coding? Journal of Experimental Child Psychology,

64, 199–231.

Nieminen, L. (2007). A complex case: A morphosyntactic

approach to complexity in early child language.

Jyvaskyla Studies in Humanities, 72, University of Jyvaskyla,

Retreived 20 July 2009, from http://julkaisut.jyu.fi?id¼978-

951-39-2858-2

Nieminen, L. (submitted). MLU and IPSyn measuring absolute

complexity. EAAL Yearbook 5.

Nieminen, L., & Torvelainen, P. (2003). Produktiivisen syntaksin

indeksi – suomenkielinen versio [A Finnish version of the

Index of Productive Syntax]. Puhe ja kieli, 23, 25–44.

Pater, J. (1997). Minimal violation and phonological development.

Language Acquisition, 6, 201–253.

Pennington, B. F. (1995). Genetics of learning disabilities. Journal

of Child Neurology, 10, 69–77.

Porpodas, C. D. (1999). Patterns of phonological and memory

processing in beginning readers and spellers of Greek. Journal

of Learning Disabilities, 32, 406–416.

Richardson, U. (1998). Familial dyslexia and sound duration in the

quantity distinctions of Finnish infants and adults. Doctoral

dissertation. Studia Philologia Jyvaskylaensia, University of

Jyvaskyla.

Richardson, U., Leppanen, 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, 385–397.

Richardson, U., Thomson, J. M., Scott, S. K., & Goswami, U.

(2004). Auditory processing skills and phonological represen-

tation in dyslexic children. Dyslexia, 10, 215–233.

Savinainen-Makkonen, T. (1996). Lapsenkielen fonologia system-

aattisen fonologisen kehityksen kaudella. [Child phonology at the

stage of phonological processes] Unpublished Licentiate’s

thesis. University of Helsinki.

Scarborough, H. S. (1990a). Very early language deficits in

dyslexic children. Child Development, 61, 1728–1743.

Scarborough, H. S. (1990b). Index of Productive Syntax. Applied

Psycholinguistics, 11, 1–22.

Scarborough, H. S. (1991). Early syntactic development of

dyslectic children. Annals of Dyslexia, 41, 207–220.

Selkirk, E. O. (1980). The role of prosodic categories in English

word stress. Linguistic Inquiry, 11, 563–605.

Siok, W. E., & Fletcher, P. (2001). The role of phonological

awareness and visual-orthographic skills in Chinese reading

acquisition. Developmental Psychology, 37, 886–899.

Smith, B., McGregor, K., & Demille, D. (2006). Phonological

development in lexically precocious 2-year-olds. Applied

Psycholinguistics, 27, 355–375.

Stanovich, K. E. (1988). Explaining the differences between

dyslexic and the garden-variety poor reader: The phonological-

core variable-difference model. Journal of Learning Disabilities,

21, 590–604.

Tallal, P. (1980). Auditory temporal perception, phonics, and

reading disabilities in children. Brain and Language, 9, 182–

198.

Torvelainen, P. (2007). Kaksivuotiaiden lasten fonologisen

kehityksen variaatio. Puheen ymmarrettavyyden seka sanan-

muotojen tavoittelun ja tuottamisen tarkastelu [Variation in

phonological development of 2-year-old Finnish children.

A study of speech intelligibility and attempting and produc-

tion of words]. Jyvaskyla Studies in Humanitiesi, 73,

University of Jyvaskyla, Retreived 20 July 2009, from http://

julkaisut.jyu.fi/?id¼978-951-39-2917-6.

Turunen, P. (2003). Production of word structures. A constraint-

based study of 2;6 year old Finnish children at-risk for dyslexia

and their controls. Jyvaskyla Studies in Languages, 54,

University of Jyvaskyla, from http://urn.fi/URN:ISBN:951-39-

1419-4.

Wiik, K. (1985). On the duration of vowels in Finnish dialects.

In O. Aaltonen, & P. Hulkko (Eds.), XIII meeting of Finnish

phoneticians – Turku 1985 (pp. 253–317). Turku: Department

of Finnish and General Linguistics of the University of

Turku.

380 U. Richardson et al.