Heikki Lyytinen, UNESCO professor & GraphoLearn and Jyväskylä Longitudinal study of Dyslexia (JLD) teams Early identification and prevention of difficulties in learning to read - a global perspective University of Jyväskylä & Niilo Mäki Institute, Finland .. or how I am trying to approach the goal of my present duty as UNESCO Chair/professor On Inclusive Literacy Learning for All

For more, see …GraphoLearn.info and for our related publications, heikki.lyytinen.info Compromised reading skill

Biological reasons (% of population) » Global > 5% » Finland > 3% (and other transparent languages) Educational reasons » Global - up to 90% (in developing some countries) » Finland – 0%

» >500 million compromised readers! The development of reading accuracy (% correct) during

the 1. Grade in Finland

Procentage Procentage 36 100 36 100

80 of correct items correct of

27 75 correct of 27 orrect 60

18 18 50 itmes

40 items36from correct 9

9 25 20

Onumber Numberitems from 36 0 0 0 0 School +10 vk 1.Gr ends School +10 vk. 1.Gr entry +5 week +15 week entry ends +5 week +15 week Individual development The average development Aro et al., 2004 Early identification - Jyväskylä Longitudinal study of Dyslexia (JLD) & Graphogame – our tool for the prevention of RDs

The Jyväskylä Longitudinal study of Dyslexia (JLD): An intensive follow-up of children at familial risk for dyslexia from birth > JLD 1994- Timo Ahonen, Mikko Aro, Kenneth Eklund, Jane Erskine, Tomi Guttorm, Leena Holopainen, Jarmo Hämäläinen, Ritva Ketonen, Marja-Leena Laakso, Seija Leinonen, Paavo Leppänen, Matti Leiwo, Marja-Kristiina Lerkkanen, Kaisa Lohvansuu, Paula Lyytinen, Anna-Maija Oksanen, Kurt Muller, Anna-Maija Poikkeus, Anne Puolakanaho, Ulla Richardson, Paula Salmi, Asko Tolvanen, Minna Torppa, Helena Viholainen

> Graphogame (Ekapeli in Finland, 2004-), the Finnish participants Ekapeli/Graphogame (ks.www.lukimat.fi; grapholearn.info): Mikko Aro, Jane Erskine, Jarkko Hautala, Riikka Heikkilä, Sini Hintikka (Huemer), Ritva Ketonen, Janne Kujala, Juha-Matti Latvala, Lea Niemelä, Marko Niemelä, Emma Ojanen, Mikko Pitkänen, Suzanne Puhakka, Miia Ronimus, Nina Saine, Paula Salmi, Vesa Rantanen, Ulla Richardson Learning game programmers: Iivo Kapanen, Ville Mönkkönen, Miika Pekkarinen Supported by EU, Niilo Mäki Foundation, Academy of Finland, Univ. of Jyväskylä, Tekes, RAY, Ministries of Education & Foreign Affairs Finland, Kela, Finnish Cultural Funds, Nokia Oy, Kone Oy, Wärtsilä Oy. The goals of the JLD following children with familial risk for dyslexia from birth to identify (from children at familial risk for dyslexia) •precursors of dyslexia •predictors of compromised acquisition •developmental paths leading to dyslexia The last step: the development of preventive measures Number of children who have attended the last originally agreed I Screening II Screening III assessment phase in Screening Grade 3 AT-RISK AT-RISK Born at Short GROUP GROUP the Compre- questionnair Assessment hospitals hensive e of parents’ N=117 N=108 of Central questionnair administere reading and infants Finland e children d at the spelling during maternity skills 01.04.93- CONTROL CONTROL clinics 31.07.96 GROUP GROUP N=410 N=8427 N=3146 N= 9368 parents N=105 N=92 parents parents infants infants children Neo- 6 14 18 2 2½ 3½ 4½ 5 5½ 6½ nata month mont month years years years years years years years l h AT-RISK GROUP N N N N N N N N N N N ======N=108 107 112 108 108 107 107 107 107 107 107 107 children

CONTROL N N N N N N N N N N N GROUP ======Data N=92 96 94 94 95 96 94 95 93 93 93 93 gathering children continue s Primary school Secondary school Adulthoo I II III VII VIII IX 20d grade grade grade grade grade grade years

N N N N N N N AT-RISK ======GROUP School 107 108 108 85 101 88 27 entry at the age of N N N N N N N 7 years CONTROL ======GROUP 92 92 92 66 81 76 16

CLASSMAT N = N = N = N = N = N = ES 1549 1756 2641 1452 1705 204 SPEECH PERCEPTION, CHILD’S COMPREHENSION,PRODUCTION CHARACTERISTICS • Auditory discrimination • Attention • Phonological processing • Psychophysiological • Vocalization • Temperament • Vocabulary, Morphology, Syntactic skills INTERVENTION NEUROPSYCHO- LOGICAL FUNCTIONS • Phonological • Visuo-spatial skills • Naming • Articulation, Motor Skills • Family School COGNITION ASSESSMEN • IQ, Memory T DOMAINS • Associative learning HOME ENVIRONMENT ACHIEVEMENT • Parent-child interaction • Alphabetic skills • Print exposure • Reading & Spelling • Parenting, Stress • Math skills IDENTIFYING & PREDICTING RISK significant predictors found in the follow-up from birth of children at familial risk for dyslexia (Jyväskylä Longitudinal study of Dyslexia) P = Predictors D = Differences between groups Age Variable 7 - yrs Reading accuracy & speed D 5 - yrs Naming speed P & D 4 - 6 yrs Phonological manipulation P & D 5 - 6 yrs Letter knowledge P & D 5 - yrs Verbal memory P & D 3 - 6 yrs Phonological sensitivity P & D 3 - 5 yrs Inflectional skills P & D 2 - 3 yrs Articulation accuracy P 2 yrs Maximum sentence length P & D 6 mth Speech perception P & D 3-5 days ERP to speech sound P & D 3-5 days ERP to sinusoidal sound/pitch - RD/risk

Lyytinen et al., Annals of Dyslexia, 2004; Dyslexia, 2004; Sage Handbook of Dyslexia, 2008 Newborn ERPs to tone frequency change differ between 2nd grade typical control and dyslexic at-risk readers MMN-paradigm Standard Deviant

Dev 1100Hz, 12 %, SOA 425 ms Std 1000Hz

EOGR EOGL EOGL - EOGR - - 3µV 3µV + + 100ms 100ms PCA components F3 F4 F3 F4 1 2 3

C3 C4 C3 C4

P3 P4 P3 P4

At-risk dyslexic Control typical readers readers (RDFR, N=8) (TRC, N=25)

Leppänen et al, 2010, Cortex Quiet sleep Newborn ERPs from the at-risk group (N=40-44)

EogL EogR Boston Naming Test 5.5 y -.311* _ 5 µV Audio-phonemic + associations 5.5 y F3 F4 500ms -.451* Word identification 6.5 y -.308* C3 C4 Rapid Naming: Faults 6.5 y .414*

P3 P4

Phoneme Deletion 6.5 y -.339* Letter identification 6.5 y 540-630 ms Naming 1st Phoneme 6.5 y -.339* -.342* Word identification 6.5 y Word identification 6.5 y -.339** -.415** Writing letters 6.5 y -.336* /ba/ Reading 6.5 y /da/ -.329* /ga/ Guttorm, Leppänen, Poikkeus, Eklund, Lyytinen, & Lyytinen (2005). Brain event-related potentials (ERPs) measured at birth predict later reading skills in children with and without familial risk for dyslexia. Cortex, 41, 291-303. 2 vuotta

2.5 Risk group vuotta Control group

3.5 vuotta

Group Scale: differences still significant after Z-score 5 controlling IQ vuotta based on the mean and SD of the control group. Is reading acquisition associated with early language delays?

• Late talking – delay in the development of expressive language skills (assessed here at 2 years of age) – Similar numbers of children in both groups could be defined as late talkers – Do children with normal speaking at 2.5 years age differ from those who start speaking later (after 2.5 y of age) in their later language development? • If so how? – Is late talking connected to reading acquisition • If so how? Development of language skills among late talkers of the risk and control groups

1.0 At- Controls risk 0.5

Z-score 0 composite of language skills -0.5

-1.0

-1.5

2 3 1/2 5 2 3 1/2 5

Late talkers Late talkers Not late talkers Age (years) Not late talkers

Lyytinen P. et al., J. of Speech, Language & Hearing Res;2001 Criteria for dyslexia

• A the end of 2. grade eight measures were taken – four measures representing reading/writing accuracy – four measures representing reading fluency • A score exactly at the 10th percentile of the control group’s performance or below was classified as fulfilling the criterion of poor performance in that task • A child was diagnosed as dyslexic/poor reader if she/he – fulfilled the criteria on at least three out of four measures in reading/writing accuracy – OR – on at least three out of four measures in reading fluency – OR – fulfilled the criteria on two measures, both in accuracy AND in the fluency domain.

Date Author Title INTERNAL 16 Reading skill at the end of the 1.Grade in the different subgroups of early language development (late expressive or both the expressive and receptive language ) 1,5 Controls with typical start of speaking 1 Composite At risk children with score of typical start of speaking reading 0,5 Controls with late talking accuracy and fluency 0 measures

-0,5 Controls with late express.&recept. speech -1

At risk children with late -1,5 talkingr -2.0 Typical At risk children with late Controls At risk group expressir .&recept. speech development of Delayed express. .+receptive early speech speech Control Risk group Lyytinen, Eklund & Lyytinen, 2005 Spelling skills by groups at the end of the first grade

Remainder of the 1,5 control group Remainder of the 1 at-risk group

0,5 Late talkers1 control group Mean z- (expressive score 0 delayed) composite Late talkers2 -0,5 control group (receptive and -1 expressive delayed) Late talkers1 at-risk -1,5 group (expressive delayed) -2.0 Late talkers2 at-risk group (receptive and Remainders of Control group At-risk group expressive delayed) the groups Late-talking groups

Lyytinen, P. Eklund & Lyytinen, Annals of Dyslexia, 2005 PISA reading outcome, z-score -2 -1,5 -1 -0,5 0 0,5 At-risk with no language delay Controls with no language delay At-risk with delay in expressive language Controls with delay in expressive language At-risk with delay in expressive and receptive…

Lyytinen, et al. 2015 Current Developmental Disorders Reports, 2, 4, 330-338. 30 At-risk with dyslexia At-risk with 25 NO dyslexia 24,23 Controls with NO dyslexia 20,99 20

17,19 Letter 15,47 15 names 12,44 known 10,07 12,87 10 10,19

7,68 8,1

5 3,12 4,32 5,35 2,55 0 0,73 3.5 4.5 5.0 5.5 6.5 Age (years) The JLD-follow-up from birth to school age of reading-related

developmentReceptive speech, 2.5 y.

Pseudoword repetition, 3.5 y.

Phonological skills, 3.5 y. Phonological skills, 4.5 y. Phonological skills, 5.5 y.

Rapid naming, 5.5 y.

Rapid naming, 6.5 y.

Letter knowledge, 3.5 y. Letter knowledge, 4.5 y. Letter knowledge, 5 y. Letter knowledge 5.5 y.

IQ, 5 y. Lyytinen, et al. Reading composite, 1. gr. Scand. J. of Reading composite 2. gr. Psychology, 2009. -3 -2 -1 0 1 z-score (mean = 0, sd =1) Individual profiles of the prediction measures of the JLD children whose reading acquisition was most severely compromised Predicting reading fluency

.55 Reading Letter .81 Phonologica accuracy 1st to 3rd Grade knowledge l awareness 7 to 9 years 4.5 to 6.6 years 1st Grade, 7.5 .24 years .52 .52 Reading Rapid fluency .27 8th Grade, 15 Naming CFI= 0.98 TLI=0.98 5 to 6.5 years years RMSEA=.043, SRMR=.036 2 chi= 112.063 (df=82), p=.004 R =49.5% N=200 Lyytinen et al., in press The role of IQ

• None of the compromised readers had < 85 in both verbal and performance IQ assessed using WISC-R.

• None of the compromised readers had < 85 in mental or motor domain at 2.5 years assessed using Bayley Scales of Infant Development Correct identifications with four orders of predictors

Phonological awareness First PA+RN+LK+RL predictor

Rapid naming RN+PA+LK Second predictor

Receptive language RL+PA+RN+LK Third predictor Letter knolwedge LK+RN+RL+PA Fourth predictor 0 20 40 60 80 100 120

Predictive assessments: Receptive language (RL at the age of 2 years), Letter knowledge (LK, 3-5 years), Phonological awareness (PA, 4-5 years), Rapid naming (RN, 5-6 years). Puolakanaho,A., Ahonen,T., Aro,M., Eklund,K., Leppänen,P., Poikkeus,A.-M., Tolvanen,A., Torppa,A., & Lyytinen,H. (2007). Very early phonological and language skills: estimating individual risk of reading disability. Journal of Child Psychology and Psychiatry, 48(9), 923–931. Learning game for and dynamic assessment of the acquisition of the basic reading skill: Graphogame

Learning the connections between spoken and written language Important facts about reading acquisition • Reading acquisition = learning to connect items of spoken language to its written forms • Written languages vary in terms of how this connection- building can be made • Alphabetic orthographies such as Finnish, German, Spanish and African local languages are relatively consistent at grapheme- phoneme level >no challenges associated with choosing the items which had to be connected from spoken to written – >>very different in English where none of its letters represents the same phoneme in all contexts of written English • Nonalphabetic scripts – the main challenge: large number of connections to be stored for acquiring the reading skill Reading acquisition and the consistency of the connections between spoken and written

• If the reading instruction is organized optimally the time child needs for the acquisition of the basic reading skill is the shorter – the more consistent the connections are - when only few (if any) complexities/alternatives need to be learned – the smaller the number of connections one has to learn Thus the learning loads differ widely between transparent e.g. Finnish & African. ~30 and English >1000 and between Alphabetic and Chinese writings: Alphabetic no more than 1000 but in Chinese >10 000 Graphogame – an enjoyable mobile or computer game for learning to read: How it helps at risk children to overcome the fuzziness of the phonemic representations with letters

Competitor’s Player’ results Falling letters s results

Correctl y Mouse chosen pointer letters

Player’s Competitorcatcher ’

Description. In the game (left) the learner is choosing (in its classical version) from the falling balls the corresponding letter of the one s/he hears from headphones. The illustration (right ) shows an example of how results can be followed. Here we follow how /N/ sound (in the centre) which learner has heard in the game more than 100 trials at the moment this picture is printed from the game logs has made him/her to choose incorrect alternative letters (shown with the number of times these have occurred with the correct N-letter). The red distributions reveal that the learner has had difficulties in not to choose R and M during the first fourth of such trials, but became able to learn during the last fourth (with green distribution) that e.g.R does not represent the /N/ sound. For this learner acquiring that the /N/ sound is not represented by M-letter has been a real challenge as shown by the red and darker green distributions which reveal that most of the choices during the first and second fourths of trials (respectively) have ended up to this mistake. The learner has failed to learn to identify the correspondence of the /N/ sound during the whole session in trials where M has occurred (7 times) as an alternative. On the other hand s/he has not chosen e.g. S to represent the /N/ sound any more during the last fourth of the trials (no misidentifications during the 9 last of the 34 trials with S as an alternative). For more details, see Lyytinen et al., Scand.J.Psychol., 2009, 50, 668-675 and for documentation of the efficiency of the game in supporting learning among at risk children, see eg. Saine et.al., Child Development , 82,3,1013-1028. . Exemplary learning curves of children at risk showing the time needed for learning the sounds of the letters among Finnish children (N=726)

The cumulative number of the acquired connections between sounds and letters

Hours of playing Modelling: Janne Kujala Remedial reading intervention and computer- assisted instruction (CARRI) (T1-T6)

CARR I group (n=25)

Screeni Main- IQ ng stream Subte Subte Subtes Subtes Post Follow Estimatio Follow test group st st t t test -up 1 n -up 2 (N=166) (n=11 2 3 4 5 6)

RRI group (n=25)

Remedial reading intervention (RRI) (T1-T6)

Screenin Grouppin T1 T2 T3 T4 T5 T6 T7 T8 g g October Decembe January March May August May August August Septemb Grade 1 r Grade 1 Grade 1 Grade 1 Grade 2 Grade 2 Grade 3 Grade 1 er Grade 1 CARRI groupGrade =1 Computer assisted remedial reading intervention group who played the GG 1/4 of the remediation time Mainstream group = Mainstream reading instruction group RRI group = Remedial reading intervention group – face-to-face remediation all the time Saine et al., (2011) Child Development, 2011, 82, 1013-1028. Spelling 35

30

25

20

RRI CARRI 15 Mainstream

Number Number ofWordsSpelled Correctly 10

5 At the At the end end of

0 of the the 3.gr. 1.grT5 T6 T7 T8

Saine et al., Child Development, 2011, 82, 1013-1028 Brain changes resulting from the use of the Graphogame GG training of <5 hours affects brain HL and UR in collaboration with Swiss colleagues Daniel Brandeis, Sylvia Brehm

Pre-Post GG: Children (n=15) before and after playing with Graphogame

LG-FG, IFG

Words- False fonts No difference ConditionCondition differences differences Increased activation in occipito-temporal areasBA18/19

Post-pre interaction between groups playing Graphogame vs Mathgame (same with numbers): p<0.005 Brem et al., PNAS, 2010, 107(17), 7939- 7944. Successful preventive practice

Massed practice using the Graphogame following optimal phonics strategy helps at risk children when * started at >6.5y of age *played >1 x per day in subsequent days until the goal is reached – motivated to be used in an as ”active” sounding form as possible – motivation to continue is guaranteed by rewarding via experience of success (~80% correct trials) – the role of parents: they show they very much like child plays GG See: www.lukimat.fi (where Finnish children play) or graphogame.com for description and demo in English Challenges • Works without complications in consistent (gr>=

A mimimun set of single letter-sounds selected to a version of the game – list of their sounds present in > 5% of the occurrence of the letter in English text (Cedex databasis, among 17 million words)

Letter % of different / all words (exemplary word) i 62.3 24076 3471217 I (in) 19.4 4386 1083446 aI (i) 5.1 2519 283459 (social) l 95.4 22272 2934160 l (all) d 94.4 14990 2844232 d (and) m 100.0 11176 1817206 m (from) b 99.0 7726 1169525 b (be)

Computation: J.Kujala Connection building of written and spoken units of English

Alternative approaches: • Small unit game: teaches graphemes of the most prototypical vowels, blends of CV and VC digraphs and combines into CVC words etc. • Larger unit game: phoneme approach+large rime units, blends learned small set of ph/gr in CV rime units starting from most dense neigbourhoods with consistent spelling etc. Introducing connections between spoken and written in English in the GG Results of the English Graphogame with Usha Goswami and Fiona Kyle, Cambridge University

• Reading gains in standard scores (SS) per hour of playing: – Phoneme game 0.47 SS points – Rime game 0.68 SS points Note: ~0.3 in the most promising earlier interventions (Hatcher et al. 2006) Only rime game elevated significantly the spelling skill

Kyle, Goswami et al., Reading Res. Q. 2012, 48, 61-76 Grapho Learning Initiative Our vision : will help millions of learners, who otherwise would not have access to learn basic skills, such as reading, and be able to launch themselves to a sustainable learning curve and a road to prosperity.。 Globally 780 000 000 people are illiterate

•Biologic, educational and social problems Research Evidence • We have gained research evidence on the efficacy of GraphoGame™ training. – from transparent writing environments – to some degree also from more non- transparent orthographies including English. • Our partner network is tackling global challenges – Biological factors (dyslexia) – Insufficient teaching – Lack of social support The basic principles of Graphogame development for a new writing system

• Careful study of the written language environment with local experts for developing appropriate content • Evidence based documentation of the efficiency of the game after a new implementation of content for a new context • Distribution and use under the responsibility of the local Ministry of Education after research has shown its efficiency in an orthographic environment Jere-Folotiya, J., et al. (2014). The effect of using a mobile literacy game to improve literacy levels of grade one students in Zambian schools. Educational Technology Research and Development, 62 (4), 417-436. How Graphogame helps in Zambia

Controls Teachers informed +children play +also teachers play about phonics instr. Practical facts about the game

• Available for free (from 2007) to all Finnish children – Playing via net with up-to-date information for teachers and parents about learning difficulties • Very easy to use – children learn within minutes and can use without adults – 4-10 hours of playing helps most at risk for dyslexia • Works also in Android, Apple, MS mobile phones • Used in Finland via initial support from the the Ministry of Education • at best >20 000 users per day – (age cohort of 60 000) • more than 400 000 users until now after 2007 GraphoGame® as a Service

Learning Analytics

Dynamic assessment, GraphoGame® Cloud analysis and adaptation to learners skills.

GraphoGame® as download and/or online Teacher training, Online learning Gamelogs Classroom and/or material, videos, Database GraphoGame® Club ebooks and guides Big Data support.

Training Global Network GraphoGame™ Pre-releases, for testing

• Studies initiated in Europe (outside Finland) – Austria, Cyprus, Denmark, Estonia, Finland, France, Greenland, Hungary, Ireland, Island, Israel, Netherlands, Norway, Poland, Portugal, Sweden, Switzerland and UK • Studies running in Africa – Erithrea, Kenya, Namibia, South Africa, Tanzania and Zambia • Elsewhere – Brazil, Canada, Chile, China, Indonesia, Mexico, Taiwan and US GraphoWorld Network of Excellence on Research

http://grapholearning.info/graphoworld For more.., please, • Call: +358 50 552 4892 • Have a look of our research: heikki.lyytinen.info • Ask for reprint(s): [email protected] • The game pages in Finnish: http://www.lukimat.fi/ • ..in English: http://www.graphogame.com • See also grapholearn.info for the whole approach • The most recent summary of main results of the JLD: https://link.springer.com/article/10.1007/s40474-015-0067-1 • The most detailed recent description of the Graphogame: • http://www.humantechnology.jyu.fi/articles/volume10/2014/Richardso n_Lyytinen.pdf • If you are interested in joining us, please, contact [email protected] Thank you for attention! The status of the GraphoGame (2017/8)

• Under study in more than 30 countries • Evidence-based status ready in appr.10 countries • Unversity has sold the rights to distribute it to GraphoGroup enterprise which is now starting its distribution in e.g. Norway and France • Contact info to GraphoGroup: [email protected]